Use local version of cri packages

Signed-off-by: Derek McGowan <derek@mcg.dev>
This commit is contained in:
Derek McGowan
2020-10-07 10:59:40 -07:00
parent e7a350176a
commit 1c60ae7f87
234 changed files with 8149 additions and 18348 deletions

View File

@@ -1,201 +0,0 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "{}"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright {yyyy} {name of copyright owner}
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

View File

@@ -1,189 +0,0 @@
# cri
<p align="center">
<img src="https://kubernetes.io/images/favicon.png" width="50" height="50">
<img src="https://containerd.io/img/logos/icon/black/containerd-icon-black.png" width="50" >
</p>
*Note: The standalone `cri-containerd` binary is end-of-life. `cri-containerd` is
transitioning from a standalone binary that talks to containerd to a plugin within
containerd. This github branch is for the `cri` plugin. See
[standalone-cri-containerd branch](https://github.com/containerd/cri/tree/standalone-cri-containerd)
for information about the standalone version of `cri-containerd`.*
*Note: You need to [drain your node](https://kubernetes.io/docs/tasks/administer-cluster/safely-drain-node/) before upgrading from standalone `cri-containerd` to containerd with `cri` plugin.*
[![Build Status](https://api.travis-ci.org/containerd/cri.svg?style=flat-square)](https://travis-ci.org/containerd/cri)
[![Go Report Card](https://goreportcard.com/badge/github.com/containerd/cri)](https://goreportcard.com/report/github.com/containerd/cri)
`cri` is a [containerd](https://containerd.io/) plugin implementation of Kubernetes [container runtime interface (CRI)](https://github.com/kubernetes/cri-api/blob/master/pkg/apis/runtime/v1alpha2/api.proto).
With it, you could run Kubernetes using containerd as the container runtime.
![cri](./docs/cri.png)
## Current Status
`cri` is a native plugin of containerd 1.1 and above. It is built into containerd and enabled by default.
`cri` is in GA:
* It is feature complete.
* It (the GA version) works with Kubernetes 1.10 and above.
* It has passed all [CRI validation tests](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-node/cri-validation.md).
* It has passed all [node e2e tests](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-node/e2e-node-tests.md).
* It has passed all [e2e tests](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-testing/e2e-tests.md).
See [test dashboard](https://k8s-testgrid.appspot.com/sig-node-containerd)
## Support Metrics
| CRI-Containerd Version | Containerd Version | Kubernetes Version | CRI Version |
|:----------------------:|:------------------:|:------------------:|:-----------:|
| v1.0.0-alpha.x | | 1.7, 1.8 | v1alpha1 |
| v1.0.0-beta.x | | 1.9 | v1alpha1 |
| End-Of-Life | v1.1 (End-Of-Life) | 1.10+ | v1alpha2 |
| | v1.2 (Extended) | 1.10+ | v1alpha2 |
| | v1.3 | 1.12+ | v1alpha2 |
| | v1.4 | 1.19+ (rc) | v1alpha2 |
**Note:** The support table above specifies the Kubernetes Version that was supported at time of release of the containerd - cri integration.
The following is the current support table for containerd CRI integration taking into account that Kubernetes only supports n-3 minor release versions.
| Containerd Version | Kubernetes Version | CRI Version |
|:------------------:|:------------------:|:-----------:|
| v1.2 | 1.15+ | v1alpha2 |
| v1.3 | 1.15+ | v1alpha2 |
| v1.4 | 1.19+ (rc) | v1alpha2 |
## Production Quality Cluster on GCE
For a production quality cluster on GCE brought up with `kube-up.sh` refer [here](docs/kube-up.md).
## Installing with Ansible and Kubeadm
For a multi node cluster installer and bring up steps using ansible and kubeadm refer [here](contrib/ansible/README.md).
## Custom Installation
For non ansible users, you can download the `cri-containerd` release tarball and deploy
kubernetes cluster using kubeadm as described [here](docs/installation.md).
## Getting Started for Developers
### Binary Dependencies and Specifications
The current release of the `cri` plugin has the following dependencies:
* [containerd](https://github.com/containerd/containerd)
* [runc](https://github.com/opencontainers/runc)
* [CNI](https://github.com/containernetworking/cni)
See [versions](./vendor.conf) of these dependencies `cri` is tested with.
As containerd and runc move to their respective general availability releases,
we will do our best to rebase/retest `cri` with these releases on a
weekly/monthly basis. Similarly, given that `cri` uses the Open
Container Initiative (OCI) [image](https://github.com/opencontainers/image-spec)
and [runtime](https://github.com/opencontainers/runtime-spec) specifications, we
will also do our best to update `cri` to the latest releases of these
specifications as appropriate.
### Install Dependencies
1. Install development libraries:
* **libseccomp development library.** Required by `cri` and runc seccomp support. `libseccomp-dev` (Ubuntu, Debian) / `libseccomp-devel`
(Fedora, CentOS, RHEL). On releases of Ubuntu <=Trusty and Debian <=jessie a
backport version of `libseccomp-dev` is required. See [travis.yml](.travis.yml) for an example on trusty.
* **btrfs development library.** Required by containerd btrfs support. `btrfs-tools`(Ubuntu, Debian) / `btrfs-progs-devel`(Fedora, CentOS, RHEL)
2. Install **`pkg-config`** (required for linking with `libseccomp`).
3. Install and setup a Go 1.13.15 development environment.
4. Make a local clone of this repository.
5. Install binary dependencies by running the following command from your cloned `cri/` project directory:
```bash
# Note: install.deps installs the above mentioned runc, containerd, and CNI
# binary dependencies. install.deps is only provided for general use and ease of
# testing. To customize `runc` and `containerd` build tags and/or to configure
# `cni`, please follow instructions in their documents.
make install.deps
```
### Build and Install `cri`
To build and install a version of containerd with the `cri` plugin, enter the
following commands from your `cri` project directory:
```bash
make
sudo make install
```
*NOTE: The version of containerd built and installed from the `Makefile` is only for
testing purposes. The version tag carries the suffix "-TEST".*
#### Build Tags
`cri` supports optional build tags for compiling support of various features.
To add build tags to the make option the `BUILD_TAGS` variable must be set.
```bash
make BUILD_TAGS='seccomp apparmor selinux'
```
| Build Tag | Feature | Dependency |
|-----------|------------------------------------|---------------------------------|
| seccomp | syscall filtering | libseccomp development library |
| selinux | selinux process and mount labeling | <none> |
| apparmor | apparmor profile support | <none> |
### Validate Your `cri` Setup
A Kubernetes incubator project called [cri-tools](https://github.com/kubernetes-sigs/cri-tools)
includes programs for exercising CRI implementations such as the `cri` plugin.
More importantly, cri-tools includes the program `critest` which is used for running
[CRI Validation Testing](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-node/cri-validation.md).
Run the CRI Validation test to validate your installation of `containerd` with `cri` built in:
```bash
make test-cri
```
### Running a Kubernetes local cluster
If you already have a working development environment for supported Kubernetes
version, you can try `cri` in a local cluster:
1. Start the version of `containerd` with `cri` plugin that you built and installed
above as root in a first terminal:
```bash
sudo containerd
```
2. From the Kubernetes project directory startup a local cluster using `containerd`:
```bash
CONTAINER_RUNTIME=remote CONTAINER_RUNTIME_ENDPOINT='unix:///run/containerd/containerd.sock' ./hack/local-up-cluster.sh
```
### Test
See [here](./docs/testing.md) for information about test.
## Using crictl
See [here](./docs/crictl.md) for information about using `crictl` to debug
pods, containers, and images.
## Configurations
See [here](./docs/config.md) for information about how to configure cri plugins
and [here](https://github.com/containerd/containerd/blob/master/docs/man/containerd-config.8.md)
for information about how to configure containerd
## Documentation
See [here](./docs) for additional documentation.
## Communication
For async communication and long running discussions please use issues and pull
requests on this github repo. This will be the best place to discuss design and
implementation.
For sync communication catch us in the `#containerd` and `#containerd-dev` slack
channels on Cloud Native Computing Foundation's (CNCF) slack -
`cloud-native.slack.com`. Everyone is welcome to join and chat.
[Get Invite to CNCF slack.](https://slack.cncf.io)
## Other Communications
As this project is tightly coupled to CRI and CRI-Tools and they are Kubernetes
projects, some of our project communications take place in the Kubernetes' SIG:
`sig-node.`
For more information about `sig-node`, `CRI`, and the `CRI-Tools` projects:
* [sig-node community site](https://github.com/kubernetes/community/tree/master/sig-node)
* Slack: `#sig-node` channel in Kubernetes (kubernetes.slack.com)
* Mailing List: https://groups.google.com/forum/#!forum/kubernetes-sig-node
### Reporting Security Issues
__If you are reporting a security issue, please reach out discreetly at security@containerd.io__.
## Licenses
The containerd codebase is released under the [Apache 2.0 license](https://github.com/containerd/containerd/blob/master/LICENSE.code).
The README.md file, and files in the "docs" folder are licensed under the
Creative Commons Attribution 4.0 International License under the terms and
conditions set forth in the file "[LICENSE.docs](https://github.com/containerd/containerd/blob/master/LICENSE.docs)". You may obtain a duplicate
copy of the same license, titled CC-BY-4.0, at http://creativecommons.org/licenses/by/4.0/.
## Project details
cri is a containerd sub-project. This project was originally established in
April of 2017 in the Kubernetes Incubator program. After reaching the Beta
stage, In January of 2018, the project was merged into [containerd](https://github.com/containerd/containerd).
As a containerd sub-project, you will find the:
* [Project governance](https://github.com/containerd/project/blob/master/GOVERNANCE.md),
* [Maintainers](https://github.com/containerd/project/blob/master/MAINTAINERS),
* and [Contributing guidelines](https://github.com/containerd/project/blob/master/CONTRIBUTING.md)
information in our [`containerd/project`](https://github.com/containerd/project) repository.

View File

@@ -1,192 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package cri
import (
"flag"
"path/filepath"
"github.com/containerd/containerd"
"github.com/containerd/containerd/api/services/containers/v1"
"github.com/containerd/containerd/api/services/diff/v1"
"github.com/containerd/containerd/api/services/images/v1"
introspectionapi "github.com/containerd/containerd/api/services/introspection/v1"
"github.com/containerd/containerd/api/services/namespaces/v1"
"github.com/containerd/containerd/api/services/tasks/v1"
"github.com/containerd/containerd/content"
"github.com/containerd/containerd/leases"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/platforms"
"github.com/containerd/containerd/plugin"
"github.com/containerd/containerd/services"
"github.com/containerd/containerd/snapshots"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"k8s.io/klog/v2"
criconfig "github.com/containerd/cri/pkg/config"
"github.com/containerd/cri/pkg/constants"
criplatforms "github.com/containerd/cri/pkg/containerd/platforms"
"github.com/containerd/cri/pkg/server"
)
// TODO(random-liu): Use github.com/pkg/errors for our errors.
// Register CRI service plugin
func init() {
config := criconfig.DefaultConfig()
plugin.Register(&plugin.Registration{
Type: plugin.GRPCPlugin,
ID: "cri",
Config: &config,
Requires: []plugin.Type{
plugin.ServicePlugin,
},
InitFn: initCRIService,
})
}
func initCRIService(ic *plugin.InitContext) (interface{}, error) {
ic.Meta.Platforms = []imagespec.Platform{platforms.DefaultSpec()}
ic.Meta.Exports = map[string]string{"CRIVersion": constants.CRIVersion}
ctx := ic.Context
pluginConfig := ic.Config.(*criconfig.PluginConfig)
if err := criconfig.ValidatePluginConfig(ctx, pluginConfig); err != nil {
return nil, errors.Wrap(err, "invalid plugin config")
}
c := criconfig.Config{
PluginConfig: *pluginConfig,
ContainerdRootDir: filepath.Dir(ic.Root),
ContainerdEndpoint: ic.Address,
RootDir: ic.Root,
StateDir: ic.State,
}
log.G(ctx).Infof("Start cri plugin with config %+v", c)
if err := setGLogLevel(); err != nil {
return nil, errors.Wrap(err, "failed to set glog level")
}
servicesOpts, err := getServicesOpts(ic)
if err != nil {
return nil, errors.Wrap(err, "failed to get services")
}
log.G(ctx).Info("Connect containerd service")
client, err := containerd.New(
"",
containerd.WithDefaultNamespace(constants.K8sContainerdNamespace),
containerd.WithDefaultPlatform(criplatforms.Default()),
containerd.WithServices(servicesOpts...),
)
if err != nil {
return nil, errors.Wrap(err, "failed to create containerd client")
}
s, err := server.NewCRIService(c, client)
if err != nil {
return nil, errors.Wrap(err, "failed to create CRI service")
}
go func() {
if err := s.Run(); err != nil {
log.G(ctx).WithError(err).Fatal("Failed to run CRI service")
}
// TODO(random-liu): Whether and how we can stop containerd.
}()
return s, nil
}
// getServicesOpts get service options from plugin context.
func getServicesOpts(ic *plugin.InitContext) ([]containerd.ServicesOpt, error) {
plugins, err := ic.GetByType(plugin.ServicePlugin)
if err != nil {
return nil, errors.Wrap(err, "failed to get service plugin")
}
opts := []containerd.ServicesOpt{
containerd.WithEventService(ic.Events),
}
for s, fn := range map[string]func(interface{}) containerd.ServicesOpt{
services.ContentService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithContentStore(s.(content.Store))
},
services.ImagesService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithImageService(s.(images.ImagesClient))
},
services.SnapshotsService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithSnapshotters(s.(map[string]snapshots.Snapshotter))
},
services.ContainersService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithContainerService(s.(containers.ContainersClient))
},
services.TasksService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithTaskService(s.(tasks.TasksClient))
},
services.DiffService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithDiffService(s.(diff.DiffClient))
},
services.NamespacesService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithNamespaceService(s.(namespaces.NamespacesClient))
},
services.LeasesService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithLeasesService(s.(leases.Manager))
},
services.IntrospectionService: func(s interface{}) containerd.ServicesOpt {
return containerd.WithIntrospectionService(s.(introspectionapi.IntrospectionClient))
},
} {
p := plugins[s]
if p == nil {
return nil, errors.Errorf("service %q not found", s)
}
i, err := p.Instance()
if err != nil {
return nil, errors.Wrapf(err, "failed to get instance of service %q", s)
}
if i == nil {
return nil, errors.Errorf("instance of service %q not found", s)
}
opts = append(opts, fn(i))
}
return opts, nil
}
// Set glog level.
func setGLogLevel() error {
l := logrus.GetLevel()
fs := flag.NewFlagSet("klog", flag.PanicOnError)
klog.InitFlags(fs)
if err := fs.Set("logtostderr", "true"); err != nil {
return err
}
switch l {
case logrus.TraceLevel:
return fs.Set("v", "5")
case logrus.DebugLevel:
return fs.Set("v", "4")
case logrus.InfoLevel:
return fs.Set("v", "2")
// glog doesn't support following filters. Defaults to v=0.
case logrus.WarnLevel:
case logrus.ErrorLevel:
case logrus.FatalLevel:
case logrus.PanicLevel:
}
return nil
}

View File

@@ -1,50 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package annotations
// ContainerType values
// Following OCI annotations are used by katacontainers now.
// We'll switch to standard secure pod API after it is defined in CRI.
const (
// ContainerTypeSandbox represents a pod sandbox container
ContainerTypeSandbox = "sandbox"
// ContainerTypeContainer represents a container running within a pod
ContainerTypeContainer = "container"
// ContainerType is the container type (sandbox or container) annotation
ContainerType = "io.kubernetes.cri.container-type"
// SandboxID is the sandbox ID annotation
SandboxID = "io.kubernetes.cri.sandbox-id"
// SandboxLogDir is the pod log directory annotation.
// If the sandbox needs to generate any log, it will put it into this directory.
// Kubelet will be responsible for:
// 1) Monitoring the disk usage of the log, and including it as part of the pod
// ephemeral storage usage.
// 2) Cleaning up the logs when the pod is deleted.
// NOTE: Kubelet is not responsible for rotating the logs.
SandboxLogDir = "io.kubernetes.cri.sandbox-log-directory"
// UntrustedWorkload is the sandbox annotation for untrusted workload. Untrusted
// workload can only run on dedicated runtime for untrusted workload.
UntrustedWorkload = "io.kubernetes.cri.untrusted-workload"
// containerName is the name of the container in the pod
ContainerName = "io.kubernetes.cri.container-name"
)

View File

@@ -1,394 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
// Code generated by protoc-gen-gogo. DO NOT EDIT.
// source: api.proto
/*
Package cri_runtimeoptions_v1 is a generated protocol buffer package.
It is generated from these files:
api.proto
It has these top-level messages:
Options
*/
package cri_runtimeoptions_v1
import proto "github.com/gogo/protobuf/proto"
import fmt "fmt"
import math "math"
import _ "github.com/gogo/protobuf/gogoproto"
import strings "strings"
import reflect "reflect"
import io "io"
// Reference imports to suppress errors if they are not otherwise used.
var _ = proto.Marshal
var _ = fmt.Errorf
var _ = math.Inf
// This is a compile-time assertion to ensure that this generated file
// is compatible with the proto package it is being compiled against.
// A compilation error at this line likely means your copy of the
// proto package needs to be updated.
const _ = proto.GoGoProtoPackageIsVersion2 // please upgrade the proto package
type Options struct {
// TypeUrl specifies the type of the content inside the config file.
TypeUrl string `protobuf:"bytes,1,opt,name=type_url,json=typeUrl,proto3" json:"type_url,omitempty"`
// ConfigPath specifies the filesystem location of the config file
// used by the runtime.
ConfigPath string `protobuf:"bytes,2,opt,name=config_path,json=configPath,proto3" json:"config_path,omitempty"`
}
func (m *Options) Reset() { *m = Options{} }
func (*Options) ProtoMessage() {}
func (*Options) Descriptor() ([]byte, []int) { return fileDescriptorApi, []int{0} }
func (m *Options) GetTypeUrl() string {
if m != nil {
return m.TypeUrl
}
return ""
}
func (m *Options) GetConfigPath() string {
if m != nil {
return m.ConfigPath
}
return ""
}
func init() {
proto.RegisterType((*Options)(nil), "cri.runtimeoptions.v1.Options")
}
func (m *Options) Marshal() (dAtA []byte, err error) {
size := m.Size()
dAtA = make([]byte, size)
n, err := m.MarshalTo(dAtA)
if err != nil {
return nil, err
}
return dAtA[:n], nil
}
func (m *Options) MarshalTo(dAtA []byte) (int, error) {
var i int
_ = i
var l int
_ = l
if len(m.TypeUrl) > 0 {
dAtA[i] = 0xa
i++
i = encodeVarintApi(dAtA, i, uint64(len(m.TypeUrl)))
i += copy(dAtA[i:], m.TypeUrl)
}
if len(m.ConfigPath) > 0 {
dAtA[i] = 0x12
i++
i = encodeVarintApi(dAtA, i, uint64(len(m.ConfigPath)))
i += copy(dAtA[i:], m.ConfigPath)
}
return i, nil
}
func encodeVarintApi(dAtA []byte, offset int, v uint64) int {
for v >= 1<<7 {
dAtA[offset] = uint8(v&0x7f | 0x80)
v >>= 7
offset++
}
dAtA[offset] = uint8(v)
return offset + 1
}
func (m *Options) Size() (n int) {
var l int
_ = l
l = len(m.TypeUrl)
if l > 0 {
n += 1 + l + sovApi(uint64(l))
}
l = len(m.ConfigPath)
if l > 0 {
n += 1 + l + sovApi(uint64(l))
}
return n
}
func sovApi(x uint64) (n int) {
for {
n++
x >>= 7
if x == 0 {
break
}
}
return n
}
func sozApi(x uint64) (n int) {
return sovApi(uint64((x << 1) ^ uint64((int64(x) >> 63))))
}
func (this *Options) String() string {
if this == nil {
return "nil"
}
s := strings.Join([]string{`&Options{`,
`TypeUrl:` + fmt.Sprintf("%v", this.TypeUrl) + `,`,
`ConfigPath:` + fmt.Sprintf("%v", this.ConfigPath) + `,`,
`}`,
}, "")
return s
}
func valueToStringApi(v interface{}) string {
rv := reflect.ValueOf(v)
if rv.IsNil() {
return "nil"
}
pv := reflect.Indirect(rv).Interface()
return fmt.Sprintf("*%v", pv)
}
func (m *Options) Unmarshal(dAtA []byte) error {
l := len(dAtA)
iNdEx := 0
for iNdEx < l {
preIndex := iNdEx
var wire uint64
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return ErrIntOverflowApi
}
if iNdEx >= l {
return io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
wire |= (uint64(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
fieldNum := int32(wire >> 3)
wireType := int(wire & 0x7)
if wireType == 4 {
return fmt.Errorf("proto: Options: wiretype end group for non-group")
}
if fieldNum <= 0 {
return fmt.Errorf("proto: Options: illegal tag %d (wire type %d)", fieldNum, wire)
}
switch fieldNum {
case 1:
if wireType != 2 {
return fmt.Errorf("proto: wrong wireType = %d for field TypeUrl", wireType)
}
var stringLen uint64
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return ErrIntOverflowApi
}
if iNdEx >= l {
return io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
stringLen |= (uint64(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
intStringLen := int(stringLen)
if intStringLen < 0 {
return ErrInvalidLengthApi
}
postIndex := iNdEx + intStringLen
if postIndex > l {
return io.ErrUnexpectedEOF
}
m.TypeUrl = string(dAtA[iNdEx:postIndex])
iNdEx = postIndex
case 2:
if wireType != 2 {
return fmt.Errorf("proto: wrong wireType = %d for field ConfigPath", wireType)
}
var stringLen uint64
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return ErrIntOverflowApi
}
if iNdEx >= l {
return io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
stringLen |= (uint64(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
intStringLen := int(stringLen)
if intStringLen < 0 {
return ErrInvalidLengthApi
}
postIndex := iNdEx + intStringLen
if postIndex > l {
return io.ErrUnexpectedEOF
}
m.ConfigPath = string(dAtA[iNdEx:postIndex])
iNdEx = postIndex
default:
iNdEx = preIndex
skippy, err := skipApi(dAtA[iNdEx:])
if err != nil {
return err
}
if skippy < 0 {
return ErrInvalidLengthApi
}
if (iNdEx + skippy) > l {
return io.ErrUnexpectedEOF
}
iNdEx += skippy
}
}
if iNdEx > l {
return io.ErrUnexpectedEOF
}
return nil
}
func skipApi(dAtA []byte) (n int, err error) {
l := len(dAtA)
iNdEx := 0
for iNdEx < l {
var wire uint64
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return 0, ErrIntOverflowApi
}
if iNdEx >= l {
return 0, io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
wire |= (uint64(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
wireType := int(wire & 0x7)
switch wireType {
case 0:
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return 0, ErrIntOverflowApi
}
if iNdEx >= l {
return 0, io.ErrUnexpectedEOF
}
iNdEx++
if dAtA[iNdEx-1] < 0x80 {
break
}
}
return iNdEx, nil
case 1:
iNdEx += 8
return iNdEx, nil
case 2:
var length int
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return 0, ErrIntOverflowApi
}
if iNdEx >= l {
return 0, io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
length |= (int(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
iNdEx += length
if length < 0 {
return 0, ErrInvalidLengthApi
}
return iNdEx, nil
case 3:
for {
var innerWire uint64
var start int = iNdEx
for shift := uint(0); ; shift += 7 {
if shift >= 64 {
return 0, ErrIntOverflowApi
}
if iNdEx >= l {
return 0, io.ErrUnexpectedEOF
}
b := dAtA[iNdEx]
iNdEx++
innerWire |= (uint64(b) & 0x7F) << shift
if b < 0x80 {
break
}
}
innerWireType := int(innerWire & 0x7)
if innerWireType == 4 {
break
}
next, err := skipApi(dAtA[start:])
if err != nil {
return 0, err
}
iNdEx = start + next
}
return iNdEx, nil
case 4:
return iNdEx, nil
case 5:
iNdEx += 4
return iNdEx, nil
default:
return 0, fmt.Errorf("proto: illegal wireType %d", wireType)
}
}
panic("unreachable")
}
var (
ErrInvalidLengthApi = fmt.Errorf("proto: negative length found during unmarshaling")
ErrIntOverflowApi = fmt.Errorf("proto: integer overflow")
)
func init() { proto.RegisterFile("api.proto", fileDescriptorApi) }
var fileDescriptorApi = []byte{
// 183 bytes of a gzipped FileDescriptorProto
0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0xe2, 0xe2, 0x4c, 0x2c, 0xc8, 0xd4,
0x2b, 0x28, 0xca, 0x2f, 0xc9, 0x17, 0x12, 0x4d, 0x2e, 0xca, 0xd4, 0x2b, 0x2a, 0xcd, 0x2b, 0xc9,
0xcc, 0x4d, 0xcd, 0x2f, 0x28, 0xc9, 0xcc, 0xcf, 0x2b, 0xd6, 0x2b, 0x33, 0x94, 0xd2, 0x4d, 0xcf,
0x2c, 0xc9, 0x28, 0x4d, 0xd2, 0x4b, 0xce, 0xcf, 0xd5, 0x4f, 0xcf, 0x4f, 0xcf, 0xd7, 0x07, 0xab,
0x4e, 0x2a, 0x4d, 0x03, 0xf3, 0xc0, 0x1c, 0x30, 0x0b, 0x62, 0x8a, 0x92, 0x2b, 0x17, 0xbb, 0x3f,
0x44, 0xb3, 0x90, 0x24, 0x17, 0x47, 0x49, 0x65, 0x41, 0x6a, 0x7c, 0x69, 0x51, 0x8e, 0x04, 0xa3,
0x02, 0xa3, 0x06, 0x67, 0x10, 0x3b, 0x88, 0x1f, 0x5a, 0x94, 0x23, 0x24, 0xcf, 0xc5, 0x9d, 0x9c,
0x9f, 0x97, 0x96, 0x99, 0x1e, 0x5f, 0x90, 0x58, 0x92, 0x21, 0xc1, 0x04, 0x96, 0xe5, 0x82, 0x08,
0x05, 0x24, 0x96, 0x64, 0x38, 0xc9, 0x9c, 0x78, 0x28, 0xc7, 0x78, 0xe3, 0xa1, 0x1c, 0x43, 0xc3,
0x23, 0x39, 0xc6, 0x13, 0x8f, 0xe4, 0x18, 0x2f, 0x3c, 0x92, 0x63, 0x7c, 0xf0, 0x48, 0x8e, 0x71,
0xc2, 0x63, 0x39, 0x86, 0x24, 0x36, 0xb0, 0x5d, 0xc6, 0x80, 0x00, 0x00, 0x00, 0xff, 0xff, 0x07,
0x00, 0xf2, 0x18, 0xbe, 0x00, 0x00, 0x00,
}

View File

@@ -1,22 +0,0 @@
// To regenerate api.pb.go run `make proto`
syntax = "proto3";
package cri.runtimeoptions.v1;
import "github.com/gogo/protobuf/gogoproto/gogo.proto";
option (gogoproto.goproto_stringer_all) = false;
option (gogoproto.stringer_all) = true;
option (gogoproto.goproto_getters_all) = true;
option (gogoproto.marshaler_all) = true;
option (gogoproto.sizer_all) = true;
option (gogoproto.unmarshaler_all) = true;
option (gogoproto.goproto_unrecognized_all) = false;
message Options {
// TypeUrl specifies the type of the content inside the config file.
string type_url = 1;
// ConfigPath specifies the filesystem location of the config file
// used by the runtime.
string config_path = 2;
}

View File

@@ -1,54 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package atomic
import "sync/atomic"
// Bool is an atomic Boolean,
// Its methods are all atomic, thus safe to be called by
// multiple goroutines simultaneously.
type Bool interface {
Set()
Unset()
IsSet() bool
}
// NewBool creates an Bool with given default value
func NewBool(ok bool) Bool {
ab := new(atomicBool)
if ok {
ab.Set()
}
return ab
}
type atomicBool int32
// Set sets the Boolean to true
func (ab *atomicBool) Set() {
atomic.StoreInt32((*int32)(ab), 1)
}
// Unset sets the Boolean to false
func (ab *atomicBool) Unset() {
atomic.StoreInt32((*int32)(ab), 0)
}
// IsSet returns whether the Boolean is true
func (ab *atomicBool) IsSet() bool {
return atomic.LoadInt32((*int32)(ab)) == 1
}

View File

@@ -1,369 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package config
import (
"context"
"time"
"github.com/BurntSushi/toml"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/plugin"
"github.com/pkg/errors"
)
// Runtime struct to contain the type(ID), engine, and root variables for a default runtime
// and a runtime for untrusted worload.
type Runtime struct {
// Type is the runtime type to use in containerd e.g. io.containerd.runtime.v1.linux
Type string `toml:"runtime_type" json:"runtimeType"`
// Engine is the name of the runtime engine used by containerd.
// This only works for runtime type "io.containerd.runtime.v1.linux".
// DEPRECATED: use Options instead. Remove when shim v1 is deprecated.
Engine string `toml:"runtime_engine" json:"runtimeEngine"`
// PodAnnotations is a list of pod annotations passed to both pod sandbox as well as
// container OCI annotations.
PodAnnotations []string `toml:"pod_annotations" json:"PodAnnotations"`
// ContainerAnnotations is a list of container annotations passed through to the OCI config of the containers.
// Container annotations in CRI are usually generated by other Kubernetes node components (i.e., not users).
// Currently, only device plugins populate the annotations.
ContainerAnnotations []string `toml:"container_annotations" json:"ContainerAnnotations"`
// Root is the directory used by containerd for runtime state.
// DEPRECATED: use Options instead. Remove when shim v1 is deprecated.
// This only works for runtime type "io.containerd.runtime.v1.linux".
Root string `toml:"runtime_root" json:"runtimeRoot"`
// Options are config options for the runtime. If options is loaded
// from toml config, it will be toml.Primitive.
Options *toml.Primitive `toml:"options" json:"options"`
// PrivilegedWithoutHostDevices overloads the default behaviour for adding host devices to the
// runtime spec when the container is privileged. Defaults to false.
PrivilegedWithoutHostDevices bool `toml:"privileged_without_host_devices" json:"privileged_without_host_devices"`
// BaseRuntimeSpec is a json file with OCI spec to use as base spec that all container's will be created from.
BaseRuntimeSpec string `toml:"base_runtime_spec" json:"baseRuntimeSpec"`
}
// ContainerdConfig contains toml config related to containerd
type ContainerdConfig struct {
// Snapshotter is the snapshotter used by containerd.
Snapshotter string `toml:"snapshotter" json:"snapshotter"`
// DefaultRuntimeName is the default runtime name to use from the runtimes table.
DefaultRuntimeName string `toml:"default_runtime_name" json:"defaultRuntimeName"`
// DefaultRuntime is the default runtime to use in containerd.
// This runtime is used when no runtime handler (or the empty string) is provided.
// DEPRECATED: use DefaultRuntimeName instead. Remove in containerd 1.4.
DefaultRuntime Runtime `toml:"default_runtime" json:"defaultRuntime"`
// UntrustedWorkloadRuntime is a runtime to run untrusted workloads on it.
// DEPRECATED: use `untrusted` runtime in Runtimes instead. Remove in containerd 1.4.
UntrustedWorkloadRuntime Runtime `toml:"untrusted_workload_runtime" json:"untrustedWorkloadRuntime"`
// Runtimes is a map from CRI RuntimeHandler strings, which specify types of runtime
// configurations, to the matching configurations.
Runtimes map[string]Runtime `toml:"runtimes" json:"runtimes"`
// NoPivot disables pivot-root (linux only), required when running a container in a RamDisk with runc
// This only works for runtime type "io.containerd.runtime.v1.linux".
NoPivot bool `toml:"no_pivot" json:"noPivot"`
// DisableSnapshotAnnotations disables to pass additional annotations (image
// related information) to snapshotters. These annotations are required by
// stargz snapshotter (https://github.com/containerd/stargz-snapshotter).
DisableSnapshotAnnotations bool `toml:"disable_snapshot_annotations" json:"disableSnapshotAnnotations"`
// DiscardUnpackedLayers is a boolean flag to specify whether to allow GC to
// remove layers from the content store after successfully unpacking these
// layers to the snapshotter.
DiscardUnpackedLayers bool `toml:"discard_unpacked_layers" json:"discardUnpackedLayers"`
}
// CniConfig contains toml config related to cni
type CniConfig struct {
// NetworkPluginBinDir is the directory in which the binaries for the plugin is kept.
NetworkPluginBinDir string `toml:"bin_dir" json:"binDir"`
// NetworkPluginConfDir is the directory in which the admin places a CNI conf.
NetworkPluginConfDir string `toml:"conf_dir" json:"confDir"`
// NetworkPluginMaxConfNum is the max number of plugin config files that will
// be loaded from the cni config directory by go-cni. Set the value to 0 to
// load all config files (no arbitrary limit). The legacy default value is 1.
NetworkPluginMaxConfNum int `toml:"max_conf_num" json:"maxConfNum"`
// NetworkPluginConfTemplate is the file path of golang template used to generate
// cni config.
// When it is set, containerd will get cidr(s) from kubelet to replace {{.PodCIDR}},
// {{.PodCIDRRanges}} or {{.Routes}} in the template, and write the config into
// NetworkPluginConfDir.
// Ideally the cni config should be placed by system admin or cni daemon like calico,
// weaveworks etc. However, there are still users using kubenet
// (https://kubernetes.io/docs/concepts/cluster-administration/network-plugins/#kubenet)
// today, who don't have a cni daemonset in production. NetworkPluginConfTemplate is
// a temporary backward-compatible solution for them.
// TODO(random-liu): Deprecate this option when kubenet is deprecated.
NetworkPluginConfTemplate string `toml:"conf_template" json:"confTemplate"`
}
// Mirror contains the config related to the registry mirror
type Mirror struct {
// Endpoints are endpoints for a namespace. CRI plugin will try the endpoints
// one by one until a working one is found. The endpoint must be a valid url
// with host specified.
// The scheme, host and path from the endpoint URL will be used.
Endpoints []string `toml:"endpoint" json:"endpoint"`
}
// AuthConfig contains the config related to authentication to a specific registry
type AuthConfig struct {
// Username is the username to login the registry.
Username string `toml:"username" json:"username"`
// Password is the password to login the registry.
Password string `toml:"password" json:"password"`
// Auth is a base64 encoded string from the concatenation of the username,
// a colon, and the password.
Auth string `toml:"auth" json:"auth"`
// IdentityToken is used to authenticate the user and get
// an access token for the registry.
IdentityToken string `toml:"identitytoken" json:"identitytoken"`
}
// TLSConfig contains the CA/Cert/Key used for a registry
type TLSConfig struct {
InsecureSkipVerify bool `toml:"insecure_skip_verify" json:"insecure_skip_verify"`
CAFile string `toml:"ca_file" json:"caFile"`
CertFile string `toml:"cert_file" json:"certFile"`
KeyFile string `toml:"key_file" json:"keyFile"`
}
// Registry is registry settings configured
type Registry struct {
// Mirrors are namespace to mirror mapping for all namespaces.
Mirrors map[string]Mirror `toml:"mirrors" json:"mirrors"`
// Configs are configs for each registry.
// The key is the domain name or IP of the registry.
Configs map[string]RegistryConfig `toml:"configs" json:"configs"`
// Auths are registry endpoint to auth config mapping. The registry endpoint must
// be a valid url with host specified.
// DEPRECATED: Use Configs instead. Remove in containerd 1.4.
Auths map[string]AuthConfig `toml:"auths" json:"auths"`
// Headers adds additional HTTP headers that get sent to all registries
Headers map[string][]string `toml:"headers" json:"headers"`
}
// RegistryConfig contains configuration used to communicate with the registry.
type RegistryConfig struct {
// Auth contains information to authenticate to the registry.
Auth *AuthConfig `toml:"auth" json:"auth"`
// TLS is a pair of CA/Cert/Key which then are used when creating the transport
// that communicates with the registry.
TLS *TLSConfig `toml:"tls" json:"tls"`
}
// ImageDecryption contains configuration to handling decryption of encrypted container images.
type ImageDecryption struct {
// KeyModel specifies the trust model of where keys should reside.
//
// Details of field usage can be found in:
// https://github.com/containerd/cri/tree/master/docs/config.md
//
// Details of key models can be found in:
// https://github.com/containerd/cri/tree/master/docs/decryption.md
KeyModel string `toml:"key_model" json:"keyModel"`
}
// PluginConfig contains toml config related to CRI plugin,
// it is a subset of Config.
type PluginConfig struct {
// ContainerdConfig contains config related to containerd
ContainerdConfig `toml:"containerd" json:"containerd"`
// CniConfig contains config related to cni
CniConfig `toml:"cni" json:"cni"`
// Registry contains config related to the registry
Registry Registry `toml:"registry" json:"registry"`
// ImageDecryption contains config related to handling decryption of encrypted container images
ImageDecryption `toml:"image_decryption" json:"imageDecryption"`
// DisableTCPService disables serving CRI on the TCP server.
DisableTCPService bool `toml:"disable_tcp_service" json:"disableTCPService"`
// StreamServerAddress is the ip address streaming server is listening on.
StreamServerAddress string `toml:"stream_server_address" json:"streamServerAddress"`
// StreamServerPort is the port streaming server is listening on.
StreamServerPort string `toml:"stream_server_port" json:"streamServerPort"`
// StreamIdleTimeout is the maximum time a streaming connection
// can be idle before the connection is automatically closed.
// The string is in the golang duration format, see:
// https://golang.org/pkg/time/#ParseDuration
StreamIdleTimeout string `toml:"stream_idle_timeout" json:"streamIdleTimeout"`
// EnableSelinux indicates to enable the selinux support.
EnableSelinux bool `toml:"enable_selinux" json:"enableSelinux"`
// SelinuxCategoryRange allows the upper bound on the category range to be set.
// If not specified or set to 0, defaults to 1024 from the selinux package.
SelinuxCategoryRange int `toml:"selinux_category_range" json:"selinuxCategoryRange"`
// SandboxImage is the image used by sandbox container.
SandboxImage string `toml:"sandbox_image" json:"sandboxImage"`
// StatsCollectPeriod is the period (in seconds) of snapshots stats collection.
StatsCollectPeriod int `toml:"stats_collect_period" json:"statsCollectPeriod"`
// SystemdCgroup enables systemd cgroup support.
// This only works for runtime type "io.containerd.runtime.v1.linux".
// DEPRECATED: config runc runtime handler instead. Remove when shim v1 is deprecated.
SystemdCgroup bool `toml:"systemd_cgroup" json:"systemdCgroup"`
// EnableTLSStreaming indicates to enable the TLS streaming support.
EnableTLSStreaming bool `toml:"enable_tls_streaming" json:"enableTLSStreaming"`
// X509KeyPairStreaming is a x509 key pair used for TLS streaming
X509KeyPairStreaming `toml:"x509_key_pair_streaming" json:"x509KeyPairStreaming"`
// MaxContainerLogLineSize is the maximum log line size in bytes for a container.
// Log line longer than the limit will be split into multiple lines. Non-positive
// value means no limit.
MaxContainerLogLineSize int `toml:"max_container_log_line_size" json:"maxContainerLogSize"`
// DisableCgroup indicates to disable the cgroup support.
// This is useful when the containerd does not have permission to access cgroup.
DisableCgroup bool `toml:"disable_cgroup" json:"disableCgroup"`
// DisableApparmor indicates to disable the apparmor support.
// This is useful when the containerd does not have permission to access Apparmor.
DisableApparmor bool `toml:"disable_apparmor" json:"disableApparmor"`
// RestrictOOMScoreAdj indicates to limit the lower bound of OOMScoreAdj to the containerd's
// current OOMScoreADj.
// This is useful when the containerd does not have permission to decrease OOMScoreAdj.
RestrictOOMScoreAdj bool `toml:"restrict_oom_score_adj" json:"restrictOOMScoreAdj"`
// MaxConcurrentDownloads restricts the number of concurrent downloads for each image.
MaxConcurrentDownloads int `toml:"max_concurrent_downloads" json:"maxConcurrentDownloads"`
// DisableProcMount disables Kubernetes ProcMount support. This MUST be set to `true`
// when using containerd with Kubernetes <=1.11.
DisableProcMount bool `toml:"disable_proc_mount" json:"disableProcMount"`
// UnsetSeccompProfile is the profile containerd/cri will use If the provided seccomp profile is
// unset (`""`) for a container (default is `unconfined`)
UnsetSeccompProfile string `toml:"unset_seccomp_profile" json:"unsetSeccompProfile"`
// TolerateMissingHugetlbController if set to false will error out on create/update
// container requests with huge page limits if the cgroup controller for hugepages is not present.
// This helps with supporting Kubernetes <=1.18 out of the box. (default is `true`)
TolerateMissingHugetlbController bool `toml:"tolerate_missing_hugetlb_controller" json:"tolerateMissingHugetlbController"`
// DisableHugetlbController indicates to silently disable the hugetlb controller, even when it is
// present in /sys/fs/cgroup/cgroup.controllers.
// This helps with running rootless mode + cgroup v2 + systemd but without hugetlb delegation.
DisableHugetlbController bool `toml:"disable_hugetlb_controller" json:"disableHugetlbController"`
// IgnoreImageDefinedVolumes ignores volumes defined by the image. Useful for better resource
// isolation, security and early detection of issues in the mount configuration when using
// ReadOnlyRootFilesystem since containers won't silently mount a temporary volume.
IgnoreImageDefinedVolumes bool `toml:"ignore_image_defined_volumes" json:"ignoreImageDefinedVolumes"`
}
// X509KeyPairStreaming contains the x509 configuration for streaming
type X509KeyPairStreaming struct {
// TLSCertFile is the path to a certificate file
TLSCertFile string `toml:"tls_cert_file" json:"tlsCertFile"`
// TLSKeyFile is the path to a private key file
TLSKeyFile string `toml:"tls_key_file" json:"tlsKeyFile"`
}
// Config contains all configurations for cri server.
type Config struct {
// PluginConfig is the config for CRI plugin.
PluginConfig
// ContainerdRootDir is the root directory path for containerd.
ContainerdRootDir string `json:"containerdRootDir"`
// ContainerdEndpoint is the containerd endpoint path.
ContainerdEndpoint string `json:"containerdEndpoint"`
// RootDir is the root directory path for managing cri plugin files
// (metadata checkpoint etc.)
RootDir string `json:"rootDir"`
// StateDir is the root directory path for managing volatile pod/container data
StateDir string `json:"stateDir"`
}
const (
// RuntimeUntrusted is the implicit runtime defined for ContainerdConfig.UntrustedWorkloadRuntime
RuntimeUntrusted = "untrusted"
// RuntimeDefault is the implicit runtime defined for ContainerdConfig.DefaultRuntime
RuntimeDefault = "default"
// KeyModelNode is the key model where key for encrypted images reside
// on the worker nodes
KeyModelNode = "node"
)
// ValidatePluginConfig validates the given plugin configuration.
func ValidatePluginConfig(ctx context.Context, c *PluginConfig) error {
if c.ContainerdConfig.Runtimes == nil {
c.ContainerdConfig.Runtimes = make(map[string]Runtime)
}
// Validation for deprecated untrusted_workload_runtime.
if c.ContainerdConfig.UntrustedWorkloadRuntime.Type != "" {
log.G(ctx).Warning("`untrusted_workload_runtime` is deprecated, please use `untrusted` runtime in `runtimes` instead")
if _, ok := c.ContainerdConfig.Runtimes[RuntimeUntrusted]; ok {
return errors.Errorf("conflicting definitions: configuration includes both `untrusted_workload_runtime` and `runtimes[%q]`", RuntimeUntrusted)
}
c.ContainerdConfig.Runtimes[RuntimeUntrusted] = c.ContainerdConfig.UntrustedWorkloadRuntime
}
// Validation for deprecated default_runtime field.
if c.ContainerdConfig.DefaultRuntime.Type != "" {
log.G(ctx).Warning("`default_runtime` is deprecated, please use `default_runtime_name` to reference the default configuration you have defined in `runtimes`")
c.ContainerdConfig.DefaultRuntimeName = RuntimeDefault
c.ContainerdConfig.Runtimes[RuntimeDefault] = c.ContainerdConfig.DefaultRuntime
}
// Validation for default_runtime_name
if c.ContainerdConfig.DefaultRuntimeName == "" {
return errors.New("`default_runtime_name` is empty")
}
if _, ok := c.ContainerdConfig.Runtimes[c.ContainerdConfig.DefaultRuntimeName]; !ok {
return errors.New("no corresponding runtime configured in `runtimes` for `default_runtime_name`")
}
// Validation for deprecated runtime options.
if c.SystemdCgroup {
if c.ContainerdConfig.Runtimes[c.ContainerdConfig.DefaultRuntimeName].Type != plugin.RuntimeLinuxV1 {
return errors.Errorf("`systemd_cgroup` only works for runtime %s", plugin.RuntimeLinuxV1)
}
log.G(ctx).Warning("`systemd_cgroup` is deprecated, please use runtime `options` instead")
}
if c.NoPivot {
if c.ContainerdConfig.Runtimes[c.ContainerdConfig.DefaultRuntimeName].Type != plugin.RuntimeLinuxV1 {
return errors.Errorf("`no_pivot` only works for runtime %s", plugin.RuntimeLinuxV1)
}
// NoPivot can't be deprecated yet, because there is no alternative config option
// for `io.containerd.runtime.v1.linux`.
}
for _, r := range c.ContainerdConfig.Runtimes {
if r.Engine != "" {
if r.Type != plugin.RuntimeLinuxV1 {
return errors.Errorf("`runtime_engine` only works for runtime %s", plugin.RuntimeLinuxV1)
}
log.G(ctx).Warning("`runtime_engine` is deprecated, please use runtime `options` instead")
}
if r.Root != "" {
if r.Type != plugin.RuntimeLinuxV1 {
return errors.Errorf("`runtime_root` only works for runtime %s", plugin.RuntimeLinuxV1)
}
log.G(ctx).Warning("`runtime_root` is deprecated, please use runtime `options` instead")
}
}
// Validation for deprecated auths options and mapping it to configs.
if len(c.Registry.Auths) != 0 {
if c.Registry.Configs == nil {
c.Registry.Configs = make(map[string]RegistryConfig)
}
for endpoint, auth := range c.Registry.Auths {
config := c.Registry.Configs[endpoint]
config.Auth = &auth
c.Registry.Configs[endpoint] = config
}
log.G(ctx).Warning("`auths` is deprecated, please use registry`configs` instead")
}
// Validation for stream_idle_timeout
if c.StreamIdleTimeout != "" {
if _, err := time.ParseDuration(c.StreamIdleTimeout); err != nil {
return errors.Wrap(err, "invalid stream idle timeout")
}
}
return nil
}

View File

@@ -1,75 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package config
import (
"github.com/BurntSushi/toml"
"github.com/containerd/containerd"
"github.com/containerd/cri/pkg/streaming"
)
// DefaultConfig returns default configurations of cri plugin.
func DefaultConfig() PluginConfig {
return PluginConfig{
CniConfig: CniConfig{
NetworkPluginBinDir: "/opt/cni/bin",
NetworkPluginConfDir: "/etc/cni/net.d",
NetworkPluginMaxConfNum: 1, // only one CNI plugin config file will be loaded
NetworkPluginConfTemplate: "",
},
ContainerdConfig: ContainerdConfig{
Snapshotter: containerd.DefaultSnapshotter,
DefaultRuntimeName: "runc",
NoPivot: false,
Runtimes: map[string]Runtime{
"runc": {
Type: "io.containerd.runc.v2",
Options: new(toml.Primitive),
},
},
},
DisableTCPService: true,
StreamServerAddress: "127.0.0.1",
StreamServerPort: "0",
StreamIdleTimeout: streaming.DefaultConfig.StreamIdleTimeout.String(), // 4 hour
EnableSelinux: false,
SelinuxCategoryRange: 1024,
EnableTLSStreaming: false,
X509KeyPairStreaming: X509KeyPairStreaming{
TLSKeyFile: "",
TLSCertFile: "",
},
SandboxImage: "k8s.gcr.io/pause:3.2",
StatsCollectPeriod: 10,
SystemdCgroup: false,
MaxContainerLogLineSize: 16 * 1024,
Registry: Registry{
Mirrors: map[string]Mirror{
"docker.io": {
Endpoints: []string{"https://registry-1.docker.io"},
},
},
},
MaxConcurrentDownloads: 3,
DisableProcMount: false,
TolerateMissingHugetlbController: true,
DisableHugetlbController: true,
IgnoreImageDefinedVolumes: false,
}
}

View File

@@ -1,71 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package config
import (
"os"
"path/filepath"
"github.com/containerd/containerd"
"github.com/containerd/cri/pkg/streaming"
)
// DefaultConfig returns default configurations of cri plugin.
func DefaultConfig() PluginConfig {
return PluginConfig{
CniConfig: CniConfig{
NetworkPluginBinDir: filepath.Join(os.Getenv("ProgramFiles"), "containerd", "cni", "bin"),
NetworkPluginConfDir: filepath.Join(os.Getenv("ProgramFiles"), "containerd", "cni", "conf"),
NetworkPluginMaxConfNum: 1,
NetworkPluginConfTemplate: "",
},
ContainerdConfig: ContainerdConfig{
Snapshotter: containerd.DefaultSnapshotter,
DefaultRuntimeName: "runhcs-wcow-process",
NoPivot: false,
Runtimes: map[string]Runtime{
"runhcs-wcow-process": {
Type: "io.containerd.runhcs.v1",
},
},
},
DisableTCPService: true,
StreamServerAddress: "127.0.0.1",
StreamServerPort: "0",
StreamIdleTimeout: streaming.DefaultConfig.StreamIdleTimeout.String(), // 4 hour
EnableTLSStreaming: false,
X509KeyPairStreaming: X509KeyPairStreaming{
TLSKeyFile: "",
TLSCertFile: "",
},
SandboxImage: "mcr.microsoft.com/oss/kubernetes/pause:1.4.0",
StatsCollectPeriod: 10,
MaxContainerLogLineSize: 16 * 1024,
Registry: Registry{
Mirrors: map[string]Mirror{
"docker.io": {
Endpoints: []string{"https://registry-1.docker.io"},
},
},
},
MaxConcurrentDownloads: 3,
IgnoreImageDefinedVolumes: false,
// TODO(windows): Add platform specific config, so that most common defaults can be shared.
}
}

View File

@@ -1,26 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package constants
// TODO(random-liu): Merge annotations package into this package.
const (
// K8sContainerdNamespace is the namespace we use to connect containerd.
K8sContainerdNamespace = "k8s.io"
// CRIVersion is the CRI version supported by the CRI plugin.
CRIVersion = "v1alpha2"
)

View File

@@ -1,118 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package opts
import (
"context"
"io/ioutil"
"os"
"path/filepath"
"github.com/containerd/containerd"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/mount"
"github.com/containerd/continuity/fs"
"github.com/pkg/errors"
)
// WithNewSnapshot wraps `containerd.WithNewSnapshot` so that if creating the
// snapshot fails we make sure the image is actually unpacked and and retry.
func WithNewSnapshot(id string, i containerd.Image) containerd.NewContainerOpts {
f := containerd.WithNewSnapshot(id, i)
return func(ctx context.Context, client *containerd.Client, c *containers.Container) error {
if err := f(ctx, client, c); err != nil {
if !errdefs.IsNotFound(err) {
return err
}
if err := i.Unpack(ctx, c.Snapshotter); err != nil {
return errors.Wrap(err, "error unpacking image")
}
return f(ctx, client, c)
}
return nil
}
}
// WithVolumes copies ownership of volume in rootfs to its corresponding host path.
// It doesn't update runtime spec.
// The passed in map is a host path to container path map for all volumes.
func WithVolumes(volumeMounts map[string]string) containerd.NewContainerOpts {
return func(ctx context.Context, client *containerd.Client, c *containers.Container) (err error) {
if c.Snapshotter == "" {
return errors.New("no snapshotter set for container")
}
if c.SnapshotKey == "" {
return errors.New("rootfs not created for container")
}
snapshotter := client.SnapshotService(c.Snapshotter)
mounts, err := snapshotter.Mounts(ctx, c.SnapshotKey)
if err != nil {
return err
}
root, err := ioutil.TempDir("", "ctd-volume")
if err != nil {
return err
}
// We change RemoveAll to Remove so that we either leak a temp dir
// if it fails but not RM snapshot data.
// refer to https://github.com/containerd/containerd/pull/1868
// https://github.com/containerd/containerd/pull/1785
defer os.Remove(root) // nolint: errcheck
if err := mount.All(mounts, root); err != nil {
return errors.Wrap(err, "failed to mount")
}
defer func() {
if uerr := mount.Unmount(root, 0); uerr != nil {
log.G(ctx).WithError(uerr).Errorf("Failed to unmount snapshot %q", c.SnapshotKey)
if err == nil {
err = uerr
}
}
}()
for host, volume := range volumeMounts {
src := filepath.Join(root, volume)
if _, err := os.Stat(src); err != nil {
if os.IsNotExist(err) {
// Skip copying directory if it does not exist.
continue
}
return errors.Wrap(err, "stat volume in rootfs")
}
if err := copyExistingContents(src, host); err != nil {
return errors.Wrap(err, "taking runtime copy of volume")
}
}
return nil
}
}
// copyExistingContents copies from the source to the destination and
// ensures the ownership is appropriately set.
func copyExistingContents(source, destination string) error {
dstList, err := ioutil.ReadDir(destination)
if err != nil {
return err
}
if len(dstList) != 0 {
return errors.Errorf("volume at %q is not initially empty", destination)
}
return fs.CopyDir(destination, source)
}

View File

@@ -1,113 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package opts
import (
"context"
"os"
"path/filepath"
"strings"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/oci"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// DefaultSandboxCPUshares is default cpu shares for sandbox container.
// TODO(windows): Revisit cpu shares for windows (https://github.com/containerd/cri/issues/1297)
const DefaultSandboxCPUshares = 2
// WithRelativeRoot sets the root for the container
func WithRelativeRoot(root string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
if s.Root == nil {
s.Root = &runtimespec.Root{}
}
s.Root.Path = root
return nil
}
}
// WithoutRoot sets the root to nil for the container.
func WithoutRoot(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
s.Root = nil
return nil
}
// WithProcessArgs sets the process args on the spec based on the image and runtime config
func WithProcessArgs(config *runtime.ContainerConfig, image *imagespec.ImageConfig) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
command, args := config.GetCommand(), config.GetArgs()
// The following logic is migrated from https://github.com/moby/moby/blob/master/daemon/commit.go
// TODO(random-liu): Clearly define the commands overwrite behavior.
if len(command) == 0 {
// Copy array to avoid data race.
if len(args) == 0 {
args = append([]string{}, image.Cmd...)
}
if command == nil {
command = append([]string{}, image.Entrypoint...)
}
}
if len(command) == 0 && len(args) == 0 {
return errors.New("no command specified")
}
return oci.WithProcessArgs(append(command, args...)...)(ctx, client, c, s)
}
}
// mounts defines how to sort runtime.Mount.
// This is the same with the Docker implementation:
// https://github.com/moby/moby/blob/17.05.x/daemon/volumes.go#L26
type orderedMounts []*runtime.Mount
// Len returns the number of mounts. Used in sorting.
func (m orderedMounts) Len() int {
return len(m)
}
// Less returns true if the number of parts (a/b/c would be 3 parts) in the
// mount indexed by parameter 1 is less than that of the mount indexed by
// parameter 2. Used in sorting.
func (m orderedMounts) Less(i, j int) bool {
return m.parts(i) < m.parts(j)
}
// Swap swaps two items in an array of mounts. Used in sorting
func (m orderedMounts) Swap(i, j int) {
m[i], m[j] = m[j], m[i]
}
// parts returns the number of parts in the destination of a mount. Used in sorting.
func (m orderedMounts) parts(i int) int {
return strings.Count(filepath.Clean(m[i].ContainerPath), string(os.PathSeparator))
}
// WithAnnotation sets the provided annotation
func WithAnnotation(k, v string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Annotations == nil {
s.Annotations = make(map[string]string)
}
s.Annotations[k] = v
return nil
}
}

View File

@@ -1,721 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package opts
import (
"context"
"fmt"
"io/ioutil"
"os"
"path/filepath"
"sort"
"strconv"
"strings"
"sync"
"syscall"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/mount"
"github.com/containerd/containerd/oci"
"github.com/opencontainers/runc/libcontainer/devices"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/opencontainers/selinux/go-selinux/label"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/sys/unix"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
osinterface "github.com/containerd/cri/pkg/os"
"github.com/containerd/cri/pkg/util"
)
// WithAdditionalGIDs adds any additional groups listed for a particular user in the
// /etc/groups file of the image's root filesystem to the OCI spec's additionalGids array.
func WithAdditionalGIDs(userstr string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
gids := s.Process.User.AdditionalGids
if err := oci.WithAdditionalGIDs(userstr)(ctx, client, c, s); err != nil {
return err
}
// Merge existing gids and new gids.
s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, gids)
return nil
}
}
func mergeGids(gids1, gids2 []uint32) []uint32 {
gidsMap := make(map[uint32]struct{})
for _, gid1 := range gids1 {
gidsMap[gid1] = struct{}{}
}
for _, gid2 := range gids2 {
gidsMap[gid2] = struct{}{}
}
var gids []uint32
for gid := range gidsMap {
gids = append(gids, gid)
}
sort.Slice(gids, func(i, j int) bool { return gids[i] < gids[j] })
return gids
}
// WithoutRunMount removes the `/run` inside the spec
func WithoutRunMount(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
var (
mounts []runtimespec.Mount
current = s.Mounts
)
for _, m := range current {
if filepath.Clean(m.Destination) == "/run" {
continue
}
mounts = append(mounts, m)
}
s.Mounts = mounts
return nil
}
// WithoutDefaultSecuritySettings removes the default security settings generated on a spec
func WithoutDefaultSecuritySettings(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
// Make sure no default seccomp/apparmor is specified
s.Process.ApparmorProfile = ""
if s.Linux != nil {
s.Linux.Seccomp = nil
}
// Remove default rlimits (See issue #515)
s.Process.Rlimits = nil
return nil
}
// WithMounts sorts and adds runtime and CRI mounts to the spec
func WithMounts(osi osinterface.OS, config *runtime.ContainerConfig, extra []*runtime.Mount, mountLabel string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, _ *containers.Container, s *runtimespec.Spec) (err error) {
// mergeMounts merge CRI mounts with extra mounts. If a mount destination
// is mounted by both a CRI mount and an extra mount, the CRI mount will
// be kept.
var (
criMounts = config.GetMounts()
mounts = append([]*runtime.Mount{}, criMounts...)
)
// Copy all mounts from extra mounts, except for mounts overridden by CRI.
for _, e := range extra {
found := false
for _, c := range criMounts {
if filepath.Clean(e.ContainerPath) == filepath.Clean(c.ContainerPath) {
found = true
break
}
}
if !found {
mounts = append(mounts, e)
}
}
// Sort mounts in number of parts. This ensures that high level mounts don't
// shadow other mounts.
sort.Sort(orderedMounts(mounts))
// Mount cgroup into the container as readonly, which inherits docker's behavior.
s.Mounts = append(s.Mounts, runtimespec.Mount{
Source: "cgroup",
Destination: "/sys/fs/cgroup",
Type: "cgroup",
Options: []string{"nosuid", "noexec", "nodev", "relatime", "ro"},
})
// Copy all mounts from default mounts, except for
// - mounts overridden by supplied mount;
// - all mounts under /dev if a supplied /dev is present.
mountSet := make(map[string]struct{})
for _, m := range mounts {
mountSet[filepath.Clean(m.ContainerPath)] = struct{}{}
}
defaultMounts := s.Mounts
s.Mounts = nil
for _, m := range defaultMounts {
dst := filepath.Clean(m.Destination)
if _, ok := mountSet[dst]; ok {
// filter out mount overridden by a supplied mount
continue
}
if _, mountDev := mountSet["/dev"]; mountDev && strings.HasPrefix(dst, "/dev/") {
// filter out everything under /dev if /dev is a supplied mount
continue
}
s.Mounts = append(s.Mounts, m)
}
for _, mount := range mounts {
var (
dst = mount.GetContainerPath()
src = mount.GetHostPath()
)
// Create the host path if it doesn't exist.
// TODO(random-liu): Add CRI validation test for this case.
if _, err := osi.Stat(src); err != nil {
if !os.IsNotExist(err) {
return errors.Wrapf(err, "failed to stat %q", src)
}
if err := osi.MkdirAll(src, 0755); err != nil {
return errors.Wrapf(err, "failed to mkdir %q", src)
}
}
// TODO(random-liu): Add cri-containerd integration test or cri validation test
// for this.
src, err := osi.ResolveSymbolicLink(src)
if err != nil {
return errors.Wrapf(err, "failed to resolve symlink %q", src)
}
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
options := []string{"rbind"}
switch mount.GetPropagation() {
case runtime.MountPropagation_PROPAGATION_PRIVATE:
options = append(options, "rprivate")
// Since default root propagation in runc is rprivate ignore
// setting the root propagation
case runtime.MountPropagation_PROPAGATION_BIDIRECTIONAL:
if err := ensureShared(src, osi.(osinterface.UNIX).LookupMount); err != nil {
return err
}
options = append(options, "rshared")
s.Linux.RootfsPropagation = "rshared"
case runtime.MountPropagation_PROPAGATION_HOST_TO_CONTAINER:
if err := ensureSharedOrSlave(src, osi.(osinterface.UNIX).LookupMount); err != nil {
return err
}
options = append(options, "rslave")
if s.Linux.RootfsPropagation != "rshared" &&
s.Linux.RootfsPropagation != "rslave" {
s.Linux.RootfsPropagation = "rslave"
}
default:
log.G(ctx).Warnf("Unknown propagation mode for hostPath %q", mount.HostPath)
options = append(options, "rprivate")
}
// NOTE(random-liu): we don't change all mounts to `ro` when root filesystem
// is readonly. This is different from docker's behavior, but make more sense.
if mount.GetReadonly() {
options = append(options, "ro")
} else {
options = append(options, "rw")
}
if mount.GetSelinuxRelabel() {
if err := label.Relabel(src, mountLabel, false); err != nil && err != unix.ENOTSUP {
return errors.Wrapf(err, "relabel %q with %q failed", src, mountLabel)
}
}
s.Mounts = append(s.Mounts, runtimespec.Mount{
Source: src,
Destination: dst,
Type: "bind",
Options: options,
})
}
return nil
}
}
// Ensure mount point on which path is mounted, is shared.
func ensureShared(path string, lookupMount func(string) (mount.Info, error)) error {
mountInfo, err := lookupMount(path)
if err != nil {
return err
}
// Make sure source mount point is shared.
optsSplit := strings.Split(mountInfo.Optional, " ")
for _, opt := range optsSplit {
if strings.HasPrefix(opt, "shared:") {
return nil
}
}
return errors.Errorf("path %q is mounted on %q but it is not a shared mount", path, mountInfo.Mountpoint)
}
// ensure mount point on which path is mounted, is either shared or slave.
func ensureSharedOrSlave(path string, lookupMount func(string) (mount.Info, error)) error {
mountInfo, err := lookupMount(path)
if err != nil {
return err
}
// Make sure source mount point is shared.
optsSplit := strings.Split(mountInfo.Optional, " ")
for _, opt := range optsSplit {
if strings.HasPrefix(opt, "shared:") {
return nil
} else if strings.HasPrefix(opt, "master:") {
return nil
}
}
return errors.Errorf("path %q is mounted on %q but it is not a shared or slave mount", path, mountInfo.Mountpoint)
}
func addDevice(s *runtimespec.Spec, rd runtimespec.LinuxDevice) {
for i, dev := range s.Linux.Devices {
if dev.Path == rd.Path {
s.Linux.Devices[i] = rd
return
}
}
s.Linux.Devices = append(s.Linux.Devices, rd)
}
// WithDevices sets the provided devices onto the container spec
func WithDevices(osi osinterface.OS, config *runtime.ContainerConfig) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
if s.Linux.Resources == nil {
s.Linux.Resources = &runtimespec.LinuxResources{}
}
for _, device := range config.GetDevices() {
path, err := osi.ResolveSymbolicLink(device.HostPath)
if err != nil {
return err
}
dev, err := devices.DeviceFromPath(path, device.Permissions)
if err != nil {
return err
}
rd := runtimespec.LinuxDevice{
Path: device.ContainerPath,
Type: string(dev.Type),
Major: dev.Major,
Minor: dev.Minor,
UID: &dev.Uid,
GID: &dev.Gid,
}
addDevice(s, rd)
s.Linux.Resources.Devices = append(s.Linux.Resources.Devices, runtimespec.LinuxDeviceCgroup{
Allow: true,
Type: string(dev.Type),
Major: &dev.Major,
Minor: &dev.Minor,
Access: string(dev.Permissions),
})
}
return nil
}
}
// WithCapabilities sets the provided capabilties from the security context
func WithCapabilities(sc *runtime.LinuxContainerSecurityContext) oci.SpecOpts {
capabilities := sc.GetCapabilities()
if capabilities == nil {
return nullOpt
}
var opts []oci.SpecOpts
// Add/drop all capabilities if "all" is specified, so that
// following individual add/drop could still work. E.g.
// AddCapabilities: []string{"ALL"}, DropCapabilities: []string{"CHOWN"}
// will be all capabilities without `CAP_CHOWN`.
if util.InStringSlice(capabilities.GetAddCapabilities(), "ALL") {
opts = append(opts, oci.WithAllCapabilities)
}
if util.InStringSlice(capabilities.GetDropCapabilities(), "ALL") {
opts = append(opts, oci.WithCapabilities(nil))
}
var caps []string
for _, c := range capabilities.GetAddCapabilities() {
if strings.ToUpper(c) == "ALL" {
continue
}
// Capabilities in CRI doesn't have `CAP_` prefix, so add it.
caps = append(caps, "CAP_"+strings.ToUpper(c))
}
opts = append(opts, oci.WithAddedCapabilities(caps))
caps = []string{}
for _, c := range capabilities.GetDropCapabilities() {
if strings.ToUpper(c) == "ALL" {
continue
}
caps = append(caps, "CAP_"+strings.ToUpper(c))
}
opts = append(opts, oci.WithDroppedCapabilities(caps))
return oci.Compose(opts...)
}
// WithoutAmbientCaps removes the ambient caps from the spec
func WithoutAmbientCaps(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
if s.Process.Capabilities == nil {
s.Process.Capabilities = &runtimespec.LinuxCapabilities{}
}
s.Process.Capabilities.Ambient = nil
return nil
}
// WithDisabledCgroups clears the Cgroups Path from the spec
func WithDisabledCgroups(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
s.Linux.CgroupsPath = ""
return nil
}
// WithSelinuxLabels sets the mount and process labels
func WithSelinuxLabels(process, mount string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
s.Linux.MountLabel = mount
s.Process.SelinuxLabel = process
return nil
}
}
// WithResources sets the provided resource restrictions
func WithResources(resources *runtime.LinuxContainerResources, tolerateMissingHugetlbController, disableHugetlbController bool) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
if resources == nil {
return nil
}
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
if s.Linux.Resources == nil {
s.Linux.Resources = &runtimespec.LinuxResources{}
}
if s.Linux.Resources.CPU == nil {
s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
}
if s.Linux.Resources.Memory == nil {
s.Linux.Resources.Memory = &runtimespec.LinuxMemory{}
}
var (
p = uint64(resources.GetCpuPeriod())
q = resources.GetCpuQuota()
shares = uint64(resources.GetCpuShares())
limit = resources.GetMemoryLimitInBytes()
hugepages = resources.GetHugepageLimits()
)
if p != 0 {
s.Linux.Resources.CPU.Period = &p
}
if q != 0 {
s.Linux.Resources.CPU.Quota = &q
}
if shares != 0 {
s.Linux.Resources.CPU.Shares = &shares
}
if cpus := resources.GetCpusetCpus(); cpus != "" {
s.Linux.Resources.CPU.Cpus = cpus
}
if mems := resources.GetCpusetMems(); mems != "" {
s.Linux.Resources.CPU.Mems = resources.GetCpusetMems()
}
if limit != 0 {
s.Linux.Resources.Memory.Limit = &limit
}
if !disableHugetlbController {
if isHugetlbControllerPresent() {
for _, limit := range hugepages {
s.Linux.Resources.HugepageLimits = append(s.Linux.Resources.HugepageLimits, runtimespec.LinuxHugepageLimit{
Pagesize: limit.PageSize,
Limit: limit.Limit,
})
}
} else {
if !tolerateMissingHugetlbController {
return errors.Errorf("huge pages limits are specified but hugetlb cgroup controller is missing. " +
"Please set tolerate_missing_hugetlb_controller to `true` to ignore this error")
}
logrus.Warn("hugetlb cgroup controller is absent. skipping huge pages limits")
}
}
return nil
}
}
var (
supportsHugetlbOnce sync.Once
supportsHugetlb bool
)
func isHugetlbControllerPresent() bool {
supportsHugetlbOnce.Do(func() {
supportsHugetlb = false
if IsCgroup2UnifiedMode() {
supportsHugetlb, _ = cgroupv2HasHugetlb()
} else {
supportsHugetlb, _ = cgroupv1HasHugetlb()
}
})
return supportsHugetlb
}
var (
_cgroupv1HasHugetlbOnce sync.Once
_cgroupv1HasHugetlb bool
_cgroupv1HasHugetlbErr error
_cgroupv2HasHugetlbOnce sync.Once
_cgroupv2HasHugetlb bool
_cgroupv2HasHugetlbErr error
isUnifiedOnce sync.Once
isUnified bool
)
// cgroupv1HasHugetlb returns whether the hugetlb controller is present on
// cgroup v1.
func cgroupv1HasHugetlb() (bool, error) {
_cgroupv1HasHugetlbOnce.Do(func() {
if _, err := ioutil.ReadDir("/sys/fs/cgroup/hugetlb"); err != nil {
_cgroupv1HasHugetlbErr = errors.Wrap(err, "readdir /sys/fs/cgroup/hugetlb")
_cgroupv1HasHugetlb = false
} else {
_cgroupv1HasHugetlbErr = nil
_cgroupv1HasHugetlb = true
}
})
return _cgroupv1HasHugetlb, _cgroupv1HasHugetlbErr
}
// cgroupv2HasHugetlb returns whether the hugetlb controller is present on
// cgroup v2.
func cgroupv2HasHugetlb() (bool, error) {
_cgroupv2HasHugetlbOnce.Do(func() {
controllers, err := ioutil.ReadFile("/sys/fs/cgroup/cgroup.controllers")
if err != nil {
_cgroupv2HasHugetlbErr = errors.Wrap(err, "read /sys/fs/cgroup/cgroup.controllers")
return
}
_cgroupv2HasHugetlb = strings.Contains(string(controllers), "hugetlb")
})
return _cgroupv2HasHugetlb, _cgroupv2HasHugetlbErr
}
// IsCgroup2UnifiedMode returns whether we are running in cgroup v2 unified mode.
func IsCgroup2UnifiedMode() bool {
isUnifiedOnce.Do(func() {
var st syscall.Statfs_t
if err := syscall.Statfs("/sys/fs/cgroup", &st); err != nil {
panic("cannot statfs cgroup root")
}
isUnified = st.Type == unix.CGROUP2_SUPER_MAGIC
})
return isUnified
}
// WithOOMScoreAdj sets the oom score
func WithOOMScoreAdj(config *runtime.ContainerConfig, restrict bool) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
resources := config.GetLinux().GetResources()
if resources == nil {
return nil
}
adj := int(resources.GetOomScoreAdj())
if restrict {
var err error
adj, err = restrictOOMScoreAdj(adj)
if err != nil {
return err
}
}
s.Process.OOMScoreAdj = &adj
return nil
}
}
// WithSysctls sets the provided sysctls onto the spec
func WithSysctls(sysctls map[string]string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
if s.Linux.Sysctl == nil {
s.Linux.Sysctl = make(map[string]string)
}
for k, v := range sysctls {
s.Linux.Sysctl[k] = v
}
return nil
}
}
// WithPodOOMScoreAdj sets the oom score for the pod sandbox
func WithPodOOMScoreAdj(adj int, restrict bool) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
if restrict {
var err error
adj, err = restrictOOMScoreAdj(adj)
if err != nil {
return err
}
}
s.Process.OOMScoreAdj = &adj
return nil
}
}
// WithSupplementalGroups sets the supplemental groups for the process
func WithSupplementalGroups(groups []int64) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Process == nil {
s.Process = &runtimespec.Process{}
}
var guids []uint32
for _, g := range groups {
guids = append(guids, uint32(g))
}
s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, guids)
return nil
}
}
// WithPodNamespaces sets the pod namespaces for the container
func WithPodNamespaces(config *runtime.LinuxContainerSecurityContext, pid uint32) oci.SpecOpts {
namespaces := config.GetNamespaceOptions()
opts := []oci.SpecOpts{
oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.NetworkNamespace, Path: GetNetworkNamespace(pid)}),
oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.IPCNamespace, Path: GetIPCNamespace(pid)}),
oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.UTSNamespace, Path: GetUTSNamespace(pid)}),
}
if namespaces.GetPid() != runtime.NamespaceMode_CONTAINER {
opts = append(opts, oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.PIDNamespace, Path: GetPIDNamespace(pid)}))
}
return oci.Compose(opts...)
}
// WithDefaultSandboxShares sets the default sandbox CPU shares
func WithDefaultSandboxShares(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Linux == nil {
s.Linux = &runtimespec.Linux{}
}
if s.Linux.Resources == nil {
s.Linux.Resources = &runtimespec.LinuxResources{}
}
if s.Linux.Resources.CPU == nil {
s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
}
i := uint64(DefaultSandboxCPUshares)
s.Linux.Resources.CPU.Shares = &i
return nil
}
// WithoutNamespace removes the provided namespace
func WithoutNamespace(t runtimespec.LinuxNamespaceType) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Linux == nil {
return nil
}
var namespaces []runtimespec.LinuxNamespace
for i, ns := range s.Linux.Namespaces {
if ns.Type != t {
namespaces = append(namespaces, s.Linux.Namespaces[i])
}
}
s.Linux.Namespaces = namespaces
return nil
}
}
func nullOpt(_ context.Context, _ oci.Client, _ *containers.Container, _ *runtimespec.Spec) error {
return nil
}
func getCurrentOOMScoreAdj() (int, error) {
b, err := ioutil.ReadFile("/proc/self/oom_score_adj")
if err != nil {
return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
}
s := strings.TrimSpace(string(b))
i, err := strconv.Atoi(s)
if err != nil {
return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
}
return i, nil
}
func restrictOOMScoreAdj(preferredOOMScoreAdj int) (int, error) {
currentOOMScoreAdj, err := getCurrentOOMScoreAdj()
if err != nil {
return preferredOOMScoreAdj, err
}
if preferredOOMScoreAdj < currentOOMScoreAdj {
return currentOOMScoreAdj, nil
}
return preferredOOMScoreAdj, nil
}
const (
// netNSFormat is the format of network namespace of a process.
netNSFormat = "/proc/%v/ns/net"
// ipcNSFormat is the format of ipc namespace of a process.
ipcNSFormat = "/proc/%v/ns/ipc"
// utsNSFormat is the format of uts namespace of a process.
utsNSFormat = "/proc/%v/ns/uts"
// pidNSFormat is the format of pid namespace of a process.
pidNSFormat = "/proc/%v/ns/pid"
)
// GetNetworkNamespace returns the network namespace of a process.
func GetNetworkNamespace(pid uint32) string {
return fmt.Sprintf(netNSFormat, pid)
}
// GetIPCNamespace returns the ipc namespace of a process.
func GetIPCNamespace(pid uint32) string {
return fmt.Sprintf(ipcNSFormat, pid)
}
// GetUTSNamespace returns the uts namespace of a process.
func GetUTSNamespace(pid uint32) string {
return fmt.Sprintf(utsNSFormat, pid)
}
// GetPIDNamespace returns the pid namespace of a process.
func GetPIDNamespace(pid uint32) string {
return fmt.Sprintf(pidNSFormat, pid)
}

View File

@@ -1,224 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package opts
import (
"context"
"path/filepath"
"sort"
"strings"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/oci"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
osinterface "github.com/containerd/cri/pkg/os"
)
// WithWindowsNetworkNamespace sets windows network namespace for container.
// TODO(windows): Move this into container/containerd.
func WithWindowsNetworkNamespace(path string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Windows == nil {
s.Windows = &runtimespec.Windows{}
}
if s.Windows.Network == nil {
s.Windows.Network = &runtimespec.WindowsNetwork{}
}
s.Windows.Network.NetworkNamespace = path
return nil
}
}
// namedPipePath returns true if the given path is to a named pipe.
func namedPipePath(p string) bool {
return strings.HasPrefix(p, `\\.\pipe\`)
}
// cleanMount returns a cleaned version of the mount path. The input is returned
// as-is if it is a named pipe path.
func cleanMount(p string) string {
if namedPipePath(p) {
return p
}
return filepath.Clean(p)
}
// WithWindowsMounts sorts and adds runtime and CRI mounts to the spec for
// windows container.
func WithWindowsMounts(osi osinterface.OS, config *runtime.ContainerConfig, extra []*runtime.Mount) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, _ *containers.Container, s *runtimespec.Spec) error {
// mergeMounts merge CRI mounts with extra mounts. If a mount destination
// is mounted by both a CRI mount and an extra mount, the CRI mount will
// be kept.
var (
criMounts = config.GetMounts()
mounts = append([]*runtime.Mount{}, criMounts...)
)
// Copy all mounts from extra mounts, except for mounts overridden by CRI.
for _, e := range extra {
found := false
for _, c := range criMounts {
if cleanMount(e.ContainerPath) == cleanMount(c.ContainerPath) {
found = true
break
}
}
if !found {
mounts = append(mounts, e)
}
}
// Sort mounts in number of parts. This ensures that high level mounts don't
// shadow other mounts.
sort.Sort(orderedMounts(mounts))
// Copy all mounts from default mounts, except for
// mounts overridden by supplied mount;
mountSet := make(map[string]struct{})
for _, m := range mounts {
mountSet[cleanMount(m.ContainerPath)] = struct{}{}
}
defaultMounts := s.Mounts
s.Mounts = nil
for _, m := range defaultMounts {
dst := cleanMount(m.Destination)
if _, ok := mountSet[dst]; ok {
// filter out mount overridden by a supplied mount
continue
}
s.Mounts = append(s.Mounts, m)
}
for _, mount := range mounts {
var (
dst = mount.GetContainerPath()
src = mount.GetHostPath()
)
// In the case of a named pipe mount on Windows, don't stat the file
// or do other operations that open it, as that could interfere with
// the listening process. filepath.Clean also breaks named pipe
// paths, so don't use it.
if !namedPipePath(src) {
if _, err := osi.Stat(src); err != nil {
// If the source doesn't exist, return an error instead
// of creating the source. This aligns with Docker's
// behavior on windows.
return errors.Wrapf(err, "failed to stat %q", src)
}
var err error
src, err = osi.ResolveSymbolicLink(src)
if err != nil {
return errors.Wrapf(err, "failed to resolve symlink %q", src)
}
// hcsshim requires clean path, especially '/' -> '\'.
src = filepath.Clean(src)
dst = filepath.Clean(dst)
}
var options []string
// NOTE(random-liu): we don't change all mounts to `ro` when root filesystem
// is readonly. This is different from docker's behavior, but make more sense.
if mount.GetReadonly() {
options = append(options, "ro")
} else {
options = append(options, "rw")
}
s.Mounts = append(s.Mounts, runtimespec.Mount{
Source: src,
Destination: dst,
Options: options,
})
}
return nil
}
}
// WithWindowsResources sets the provided resource restrictions for windows.
func WithWindowsResources(resources *runtime.WindowsContainerResources) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if resources == nil {
return nil
}
if s.Windows == nil {
s.Windows = &runtimespec.Windows{}
}
if s.Windows.Resources == nil {
s.Windows.Resources = &runtimespec.WindowsResources{}
}
if s.Windows.Resources.CPU == nil {
s.Windows.Resources.CPU = &runtimespec.WindowsCPUResources{}
}
if s.Windows.Resources.Memory == nil {
s.Windows.Resources.Memory = &runtimespec.WindowsMemoryResources{}
}
var (
count = uint64(resources.GetCpuCount())
shares = uint16(resources.GetCpuShares())
max = uint16(resources.GetCpuMaximum())
limit = uint64(resources.GetMemoryLimitInBytes())
)
if count != 0 {
s.Windows.Resources.CPU.Count = &count
}
if shares != 0 {
s.Windows.Resources.CPU.Shares = &shares
}
if max != 0 {
s.Windows.Resources.CPU.Maximum = &max
}
if limit != 0 {
s.Windows.Resources.Memory.Limit = &limit
}
return nil
}
}
// WithWindowsDefaultSandboxShares sets the default sandbox CPU shares
func WithWindowsDefaultSandboxShares(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Windows == nil {
s.Windows = &runtimespec.Windows{}
}
if s.Windows.Resources == nil {
s.Windows.Resources = &runtimespec.WindowsResources{}
}
if s.Windows.Resources.CPU == nil {
s.Windows.Resources.CPU = &runtimespec.WindowsCPUResources{}
}
i := uint16(DefaultSandboxCPUshares)
s.Windows.Resources.CPU.Shares = &i
return nil
}
// WithWindowsCredentialSpec assigns `credentialSpec` to the
// `runtime.Spec.Windows.CredentialSpec` field.
func WithWindowsCredentialSpec(credentialSpec string) oci.SpecOpts {
return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
if s.Windows == nil {
s.Windows = &runtimespec.Windows{}
}
s.Windows.CredentialSpec = credentialSpec
return nil
}
}

View File

@@ -1,38 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package opts
import (
"context"
"github.com/containerd/containerd"
"github.com/containerd/containerd/runtime/linux/runctypes"
)
// WithContainerdShimCgroup returns function that sets the containerd
// shim cgroup path
func WithContainerdShimCgroup(path string) containerd.NewTaskOpts {
return func(_ context.Context, _ *containerd.Client, r *containerd.TaskInfo) error {
r.Options = &runctypes.CreateOptions{
ShimCgroup: path,
}
return nil
}
}
//TODO: Since Options is an interface different WithXXX will be needed to set different
// combinations of CreateOptions.

View File

@@ -1,28 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package platforms
import (
"github.com/containerd/containerd/platforms"
)
// Default returns the current platform's default platform specification.
func Default() platforms.MatchComparer {
return platforms.Default()
}

View File

@@ -1,77 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package platforms
import (
"fmt"
"strconv"
"strings"
"github.com/containerd/containerd/platforms"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
"golang.org/x/sys/windows"
)
type matchComparer struct {
defaults platforms.Matcher
osVersionPrefix string
}
// Match matches platform with the same windows major, minor
// and build version.
func (m matchComparer) Match(p imagespec.Platform) bool {
if m.defaults.Match(p) {
// TODO(windows): Figure out whether OSVersion is deprecated.
return strings.HasPrefix(p.OSVersion, m.osVersionPrefix)
}
return false
}
// Less sorts matched platforms in front of other platforms.
// For matched platforms, it puts platforms with larger revision
// number in front.
func (m matchComparer) Less(p1, p2 imagespec.Platform) bool {
m1, m2 := m.Match(p1), m.Match(p2)
if m1 && m2 {
r1, r2 := revision(p1.OSVersion), revision(p2.OSVersion)
return r1 > r2
}
return m1 && !m2
}
func revision(v string) int {
parts := strings.Split(v, ".")
if len(parts) < 4 {
return 0
}
r, err := strconv.Atoi(parts[3])
if err != nil {
return 0
}
return r
}
// Default returns the current platform's default platform specification.
func Default() platforms.MatchComparer {
major, minor, build := windows.RtlGetNtVersionNumbers()
return matchComparer{
defaults: platforms.Only(platforms.DefaultSpec()),
osVersionPrefix: fmt.Sprintf("%d.%d.%d", major, minor, build),
}
}

View File

@@ -1,46 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package util
import (
"time"
"github.com/containerd/containerd/namespaces"
"golang.org/x/net/context"
"github.com/containerd/cri/pkg/constants"
)
// deferCleanupTimeout is the default timeout for containerd cleanup operations
// in defer.
const deferCleanupTimeout = 1 * time.Minute
// DeferContext returns a context for containerd cleanup operations in defer.
// A default timeout is applied to avoid cleanup operation pending forever.
func DeferContext() (context.Context, context.CancelFunc) {
return context.WithTimeout(NamespacedContext(), deferCleanupTimeout)
}
// NamespacedContext returns a context with kubernetes namespace set.
func NamespacedContext() context.Context {
return WithNamespace(context.Background())
}
// WithNamespace adds kubernetes namespace to the context.
func WithNamespace(ctx context.Context) context.Context {
return namespaces.WithNamespace(ctx, constants.K8sContainerdNamespace)
}

View File

@@ -1,57 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package ioutil
import "io"
// writeCloseInformer wraps a reader with a close function.
type wrapReadCloser struct {
reader *io.PipeReader
writer *io.PipeWriter
}
// NewWrapReadCloser creates a wrapReadCloser from a reader.
// NOTE(random-liu): To avoid goroutine leakage, the reader passed in
// must be eventually closed by the caller.
func NewWrapReadCloser(r io.Reader) io.ReadCloser {
pr, pw := io.Pipe()
go func() {
_, _ = io.Copy(pw, r)
pr.Close()
pw.Close()
}()
return &wrapReadCloser{
reader: pr,
writer: pw,
}
}
// Read reads up to len(p) bytes into p.
func (w *wrapReadCloser) Read(p []byte) (int, error) {
n, err := w.reader.Read(p)
if err == io.ErrClosedPipe {
return n, io.EOF
}
return n, err
}
// Close closes read closer.
func (w *wrapReadCloser) Close() error {
w.reader.Close()
w.writer.Close()
return nil
}

View File

@@ -1,102 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package ioutil
import (
"io"
"sync"
)
// writeCloseInformer wraps passed in write closer with a close channel.
// Caller could wait on the close channel for the write closer to be
// closed.
type writeCloseInformer struct {
close chan struct{}
wc io.WriteCloser
}
// NewWriteCloseInformer creates the writeCloseInformer from a write closer.
func NewWriteCloseInformer(wc io.WriteCloser) (io.WriteCloser, <-chan struct{}) {
close := make(chan struct{})
return &writeCloseInformer{
close: close,
wc: wc,
}, close
}
// Write passes through the data into the internal write closer.
func (w *writeCloseInformer) Write(p []byte) (int, error) {
return w.wc.Write(p)
}
// Close closes the internal write closer and inform the close channel.
func (w *writeCloseInformer) Close() error {
err := w.wc.Close()
close(w.close)
return err
}
// nopWriteCloser wraps passed in writer with a nop close function.
type nopWriteCloser struct {
w io.Writer
}
// NewNopWriteCloser creates the nopWriteCloser from a writer.
func NewNopWriteCloser(w io.Writer) io.WriteCloser {
return &nopWriteCloser{w: w}
}
// Write passes through the data into the internal writer.
func (n *nopWriteCloser) Write(p []byte) (int, error) {
return n.w.Write(p)
}
// Close is a nop close function.
func (n *nopWriteCloser) Close() error {
return nil
}
// serialWriteCloser wraps a write closer and makes sure all writes
// are done in serial.
// Parallel write won't intersect with each other. Use case:
// 1) Pipe: Write content longer than PIPE_BUF.
// See http://man7.org/linux/man-pages/man7/pipe.7.html
// 2) <3.14 Linux Kernel: write is not atomic
// See http://man7.org/linux/man-pages/man2/write.2.html
type serialWriteCloser struct {
mu sync.Mutex
wc io.WriteCloser
}
// NewSerialWriteCloser creates a SerialWriteCloser from a write closer.
func NewSerialWriteCloser(wc io.WriteCloser) io.WriteCloser {
return &serialWriteCloser{wc: wc}
}
// Write writes a group of byte arrays in order atomically.
func (s *serialWriteCloser) Write(data []byte) (int, error) {
s.mu.Lock()
defer s.mu.Unlock()
return s.wc.Write(data)
}
// Close closes the write closer.
func (s *serialWriteCloser) Close() error {
s.mu.Lock()
defer s.mu.Unlock()
return s.wc.Close()
}

View File

@@ -1,105 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package ioutil
import (
"errors"
"io"
"sync"
)
// WriterGroup is a group of writers. Writer could be dynamically
// added and removed.
type WriterGroup struct {
mu sync.Mutex
writers map[string]io.WriteCloser
closed bool
}
var _ io.Writer = &WriterGroup{}
// NewWriterGroup creates an empty writer group.
func NewWriterGroup() *WriterGroup {
return &WriterGroup{
writers: make(map[string]io.WriteCloser),
}
}
// Add adds a writer into the group. The writer will be closed
// if the writer group is closed.
func (g *WriterGroup) Add(key string, w io.WriteCloser) {
g.mu.Lock()
defer g.mu.Unlock()
if g.closed {
w.Close()
return
}
g.writers[key] = w
}
// Get gets a writer from the group, returns nil if the writer
// doesn't exist.
func (g *WriterGroup) Get(key string) io.WriteCloser {
g.mu.Lock()
defer g.mu.Unlock()
return g.writers[key]
}
// Remove removes a writer from the group.
func (g *WriterGroup) Remove(key string) {
g.mu.Lock()
defer g.mu.Unlock()
w, ok := g.writers[key]
if !ok {
return
}
w.Close()
delete(g.writers, key)
}
// Write writes data into each writer. If a writer returns error,
// it will be closed and removed from the writer group. It returns
// error if writer group is empty.
func (g *WriterGroup) Write(p []byte) (int, error) {
g.mu.Lock()
defer g.mu.Unlock()
for k, w := range g.writers {
n, err := w.Write(p)
if err == nil && len(p) == n {
continue
}
// The writer is closed or in bad state, remove it.
w.Close()
delete(g.writers, k)
}
if len(g.writers) == 0 {
return 0, errors.New("writer group is empty")
}
return len(p), nil
}
// Close closes the writer group. Write will return error after
// closed.
func (g *WriterGroup) Close() {
g.mu.Lock()
defer g.mu.Unlock()
for _, w := range g.writers {
w.Close()
}
g.writers = nil
g.closed = true
}

View File

@@ -1,222 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
// Copyright 2018 CNI authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package netns
import (
"crypto/rand"
"fmt"
"os"
"path"
"runtime"
"sync"
cnins "github.com/containernetworking/plugins/pkg/ns"
"github.com/docker/docker/pkg/symlink"
"github.com/pkg/errors"
"golang.org/x/sys/unix"
osinterface "github.com/containerd/cri/pkg/os"
)
const nsRunDir = "/var/run/netns"
// Some of the following functions are migrated from
// https://github.com/containernetworking/plugins/blob/master/pkg/testutils/netns_linux.go
// newNS creates a new persistent (bind-mounted) network namespace and returns the
// path to the network namespace.
func newNS() (nsPath string, err error) {
b := make([]byte, 16)
if _, err := rand.Reader.Read(b); err != nil {
return "", errors.Wrap(err, "failed to generate random netns name")
}
// Create the directory for mounting network namespaces
// This needs to be a shared mountpoint in case it is mounted in to
// other namespaces (containers)
if err := os.MkdirAll(nsRunDir, 0755); err != nil {
return "", err
}
// create an empty file at the mount point
nsName := fmt.Sprintf("cni-%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:])
nsPath = path.Join(nsRunDir, nsName)
mountPointFd, err := os.Create(nsPath)
if err != nil {
return "", err
}
mountPointFd.Close()
defer func() {
// Ensure the mount point is cleaned up on errors
if err != nil {
os.RemoveAll(nsPath) // nolint: errcheck
}
}()
var wg sync.WaitGroup
wg.Add(1)
// do namespace work in a dedicated goroutine, so that we can safely
// Lock/Unlock OSThread without upsetting the lock/unlock state of
// the caller of this function
go (func() {
defer wg.Done()
runtime.LockOSThread()
// Don't unlock. By not unlocking, golang will kill the OS thread when the
// goroutine is done (for go1.10+)
var origNS cnins.NetNS
origNS, err = cnins.GetNS(getCurrentThreadNetNSPath())
if err != nil {
return
}
defer origNS.Close()
// create a new netns on the current thread
err = unix.Unshare(unix.CLONE_NEWNET)
if err != nil {
return
}
// Put this thread back to the orig ns, since it might get reused (pre go1.10)
defer origNS.Set() // nolint: errcheck
// bind mount the netns from the current thread (from /proc) onto the
// mount point. This causes the namespace to persist, even when there
// are no threads in the ns.
err = unix.Mount(getCurrentThreadNetNSPath(), nsPath, "none", unix.MS_BIND, "")
if err != nil {
err = errors.Wrapf(err, "failed to bind mount ns at %s", nsPath)
}
})()
wg.Wait()
if err != nil {
return "", errors.Wrap(err, "failed to create namespace")
}
return nsPath, nil
}
// unmountNS unmounts the NS held by the netns object. unmountNS is idempotent.
func unmountNS(path string) error {
if _, err := os.Stat(path); err != nil {
if os.IsNotExist(err) {
return nil
}
return errors.Wrap(err, "failed to stat netns")
}
path, err := symlink.FollowSymlinkInScope(path, "/")
if err != nil {
return errors.Wrap(err, "failed to follow symlink")
}
if err := osinterface.Unmount(path); err != nil && !os.IsNotExist(err) {
return errors.Wrap(err, "failed to umount netns")
}
if err := os.RemoveAll(path); err != nil {
return errors.Wrap(err, "failed to remove netns")
}
return nil
}
// getCurrentThreadNetNSPath copied from pkg/ns
func getCurrentThreadNetNSPath() string {
// /proc/self/ns/net returns the namespace of the main thread, not
// of whatever thread this goroutine is running on. Make sure we
// use the thread's net namespace since the thread is switching around
return fmt.Sprintf("/proc/%d/task/%d/ns/net", os.Getpid(), unix.Gettid())
}
// NetNS holds network namespace.
type NetNS struct {
path string
}
// NewNetNS creates a network namespace.
func NewNetNS() (*NetNS, error) {
path, err := newNS()
if err != nil {
return nil, errors.Wrap(err, "failed to setup netns")
}
return &NetNS{path: path}, nil
}
// LoadNetNS loads existing network namespace.
func LoadNetNS(path string) *NetNS {
return &NetNS{path: path}
}
// Remove removes network namepace. Remove is idempotent, meaning it might
// be invoked multiple times and provides consistent result.
func (n *NetNS) Remove() error {
return unmountNS(n.path)
}
// Closed checks whether the network namespace has been closed.
func (n *NetNS) Closed() (bool, error) {
ns, err := cnins.GetNS(n.path)
if err != nil {
if _, ok := err.(cnins.NSPathNotExistErr); ok {
// The network namespace has already been removed.
return true, nil
}
if _, ok := err.(cnins.NSPathNotNSErr); ok {
// The network namespace is not mounted, remove it.
if err := os.RemoveAll(n.path); err != nil {
return false, errors.Wrap(err, "remove netns")
}
return true, nil
}
return false, errors.Wrap(err, "get netns fd")
}
if err := ns.Close(); err != nil {
return false, errors.Wrap(err, "close netns fd")
}
return false, nil
}
// GetPath returns network namespace path for sandbox container
func (n *NetNS) GetPath() string {
return n.path
}
// Do runs a function in the network namespace.
func (n *NetNS) Do(f func(cnins.NetNS) error) error {
ns, err := cnins.GetNS(n.path)
if err != nil {
return errors.Wrap(err, "get netns fd")
}
defer ns.Close() // nolint: errcheck
return ns.Do(f)
}

View File

@@ -1,78 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package netns
import "github.com/Microsoft/hcsshim/hcn"
// NetNS holds network namespace for sandbox
type NetNS struct {
path string
}
// NewNetNS creates a network namespace for the sandbox
func NewNetNS() (*NetNS, error) {
temp := hcn.HostComputeNamespace{}
hcnNamespace, err := temp.Create()
if err != nil {
return nil, err
}
return &NetNS{path: hcnNamespace.Id}, nil
}
// LoadNetNS loads existing network namespace.
func LoadNetNS(path string) *NetNS {
return &NetNS{path: path}
}
// Remove removes network namepace if it exists and not closed. Remove is idempotent,
// meaning it might be invoked multiple times and provides consistent result.
func (n *NetNS) Remove() error {
hcnNamespace, err := hcn.GetNamespaceByID(n.path)
if err != nil {
if hcn.IsNotFoundError(err) {
return nil
}
return err
}
err = hcnNamespace.Delete()
if err == nil || hcn.IsNotFoundError(err) {
return nil
}
return err
}
// Closed checks whether the network namespace has been closed.
func (n *NetNS) Closed() (bool, error) {
_, err := hcn.GetNamespaceByID(n.path)
if err == nil {
return false, nil
}
if hcn.IsNotFoundError(err) {
return true, nil
}
return false, err
}
// GetPath returns network namespace path for sandbox container
func (n *NetNS) GetPath() string {
return n.path
}
// NOTE: Do function is not supported.

View File

@@ -1,102 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package os
import (
"io"
"io/ioutil"
"os"
"path/filepath"
"github.com/docker/docker/pkg/symlink"
)
// OS collects system level operations that need to be mocked out
// during tests.
type OS interface {
MkdirAll(path string, perm os.FileMode) error
RemoveAll(path string) error
Stat(name string) (os.FileInfo, error)
ResolveSymbolicLink(name string) (string, error)
FollowSymlinkInScope(path, scope string) (string, error)
CopyFile(src, dest string, perm os.FileMode) error
WriteFile(filename string, data []byte, perm os.FileMode) error
Hostname() (string, error)
}
// RealOS is used to dispatch the real system level operations.
type RealOS struct{}
// MkdirAll will call os.MkdirAll to create a directory.
func (RealOS) MkdirAll(path string, perm os.FileMode) error {
return os.MkdirAll(path, perm)
}
// RemoveAll will call os.RemoveAll to remove the path and its children.
func (RealOS) RemoveAll(path string) error {
return os.RemoveAll(path)
}
// Stat will call os.Stat to get the status of the given file.
func (RealOS) Stat(name string) (os.FileInfo, error) {
return os.Stat(name)
}
// ResolveSymbolicLink will follow any symbolic links
func (RealOS) ResolveSymbolicLink(path string) (string, error) {
info, err := os.Lstat(path)
if err != nil {
return "", err
}
if info.Mode()&os.ModeSymlink != os.ModeSymlink {
return path, nil
}
return filepath.EvalSymlinks(path)
}
// FollowSymlinkInScope will call symlink.FollowSymlinkInScope.
func (RealOS) FollowSymlinkInScope(path, scope string) (string, error) {
return symlink.FollowSymlinkInScope(path, scope)
}
// CopyFile will copy src file to dest file
func (RealOS) CopyFile(src, dest string, perm os.FileMode) error {
in, err := os.Open(src)
if err != nil {
return err
}
defer in.Close()
out, err := os.OpenFile(dest, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, perm)
if err != nil {
return err
}
defer out.Close()
_, err = io.Copy(out, in)
return err
}
// WriteFile will call ioutil.WriteFile to write data into a file.
func (RealOS) WriteFile(filename string, data []byte, perm os.FileMode) error {
return ioutil.WriteFile(filename, data, perm)
}
// Hostname will call os.Hostname to get the hostname of the host.
func (RealOS) Hostname() (string, error) {
return os.Hostname()
}

View File

@@ -1,59 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package os
import (
"github.com/containerd/containerd/mount"
"golang.org/x/sys/unix"
)
// UNIX collects unix system level operations that need to be
// mocked out during tests.
type UNIX interface {
Mount(source string, target string, fstype string, flags uintptr, data string) error
Unmount(target string) error
LookupMount(path string) (mount.Info, error)
}
// Mount will call unix.Mount to mount the file.
func (RealOS) Mount(source string, target string, fstype string, flags uintptr, data string) error {
return unix.Mount(source, target, fstype, flags, data)
}
// Unmount will call Unmount to unmount the file.
func (RealOS) Unmount(target string) error {
return Unmount(target)
}
// LookupMount gets mount info of a given path.
func (RealOS) LookupMount(path string) (mount.Info, error) {
return mount.Lookup(path)
}
// Unmount unmounts the target. It does not return an error in case the target is not mounted.
// In case the target does not exist, the appropriate error is returned.
func Unmount(target string) error {
err := unix.Unmount(target, unix.MNT_DETACH)
if err == unix.EINVAL {
// ignore "not mounted" error
err = nil
}
return err
}

View File

@@ -1,102 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package registrar
import (
"sync"
"github.com/pkg/errors"
)
// Registrar stores one-to-one name<->key mappings.
// Names and keys must be unique.
// Registrar is safe for concurrent access.
type Registrar struct {
lock sync.Mutex
nameToKey map[string]string
keyToName map[string]string
}
// NewRegistrar creates a new Registrar with the empty indexes.
func NewRegistrar() *Registrar {
return &Registrar{
nameToKey: make(map[string]string),
keyToName: make(map[string]string),
}
}
// Reserve registers a name<->key mapping, name or key must not
// be empty.
// Reserve is idempotent.
// Attempting to reserve a conflict key<->name mapping results
// in an error.
// A name<->key reservation is globally unique.
func (r *Registrar) Reserve(name, key string) error {
r.lock.Lock()
defer r.lock.Unlock()
if name == "" || key == "" {
return errors.Errorf("invalid name %q or key %q", name, key)
}
if k, exists := r.nameToKey[name]; exists {
if k != key {
return errors.Errorf("name %q is reserved for %q", name, k)
}
return nil
}
if n, exists := r.keyToName[key]; exists {
if n != name {
return errors.Errorf("key %q is reserved for %q", key, n)
}
return nil
}
r.nameToKey[name] = key
r.keyToName[key] = name
return nil
}
// ReleaseByName releases the reserved name<->key mapping by name.
// Once released, the name and the key can be reserved again.
func (r *Registrar) ReleaseByName(name string) {
r.lock.Lock()
defer r.lock.Unlock()
key, exists := r.nameToKey[name]
if !exists {
return
}
delete(r.nameToKey, name)
delete(r.keyToName, key)
}
// ReleaseByKey release the reserved name<->key mapping by key.
func (r *Registrar) ReleaseByKey(key string) {
r.lock.Lock()
defer r.lock.Unlock()
name, exists := r.keyToName[key]
if !exists {
return
}
delete(r.nameToKey, name)
delete(r.keyToName, key)
}

View File

@@ -1,88 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright The runc Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package seccomp
import (
"bufio"
"os"
"strings"
"golang.org/x/sys/unix"
)
// IsEnabled returns if the kernel has been configured to support seccomp.
// From https://github.com/opencontainers/runc/blob/v1.0.0-rc91/libcontainer/seccomp/seccomp_linux.go#L86-L102
func IsEnabled() bool {
// Try to read from /proc/self/status for kernels > 3.8
s, err := parseStatusFile("/proc/self/status")
if err != nil {
// Check if Seccomp is supported, via CONFIG_SECCOMP.
if err := unix.Prctl(unix.PR_GET_SECCOMP, 0, 0, 0, 0); err != unix.EINVAL {
// Make sure the kernel has CONFIG_SECCOMP_FILTER.
if err := unix.Prctl(unix.PR_SET_SECCOMP, unix.SECCOMP_MODE_FILTER, 0, 0, 0); err != unix.EINVAL {
return true
}
}
return false
}
_, ok := s["Seccomp"]
return ok
}
// parseStatusFile is from https://github.com/opencontainers/runc/blob/v1.0.0-rc91/libcontainer/seccomp/seccomp_linux.go#L243-L268
func parseStatusFile(path string) (map[string]string, error) {
f, err := os.Open(path)
if err != nil {
return nil, err
}
defer f.Close()
s := bufio.NewScanner(f)
status := make(map[string]string)
for s.Scan() {
text := s.Text()
parts := strings.Split(text, ":")
if len(parts) <= 1 {
continue
}
status[parts[0]] = parts[1]
}
if err := s.Err(); err != nil {
return nil, err
}
return status, nil
}

View File

@@ -1,23 +0,0 @@
// +build !linux
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package seccomp
func IsEnabled() bool {
return false
}

View File

@@ -1,34 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
// Package bandwidth provides utilities for bandwidth shaping
package bandwidth

View File

@@ -1,72 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package bandwidth
import (
"errors"
"k8s.io/apimachinery/pkg/api/resource"
)
// FakeShaper provides an implementation of the bandwith.Shaper.
// Beware this is implementation has no features besides Reset and GetCIDRs.
type FakeShaper struct {
CIDRs []string
ResetCIDRs []string
}
// Limit is not implemented
func (f *FakeShaper) Limit(cidr string, egress, ingress *resource.Quantity) error {
return errors.New("unimplemented")
}
// Reset appends a particular CIDR to the set of ResetCIDRs being managed by this shaper
func (f *FakeShaper) Reset(cidr string) error {
f.ResetCIDRs = append(f.ResetCIDRs, cidr)
return nil
}
// ReconcileInterface is not implemented
func (f *FakeShaper) ReconcileInterface() error {
return errors.New("unimplemented")
}
// ReconcileCIDR is not implemented
func (f *FakeShaper) ReconcileCIDR(cidr string, egress, ingress *resource.Quantity) error {
return errors.New("unimplemented")
}
// GetCIDRs returns the set of CIDRs that are being managed by this shaper
func (f *FakeShaper) GetCIDRs() ([]string, error) {
return f.CIDRs, nil
}

View File

@@ -1,56 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package bandwidth
import "k8s.io/apimachinery/pkg/api/resource"
// Shaper is designed so that the shaper structs created
// satisfy the Shaper interface.
type Shaper interface {
// Limit the bandwidth for a particular CIDR on a particular interface
// * ingress and egress are in bits/second
// * cidr is expected to be a valid network CIDR (e.g. '1.2.3.4/32' or '10.20.0.1/16')
// 'egress' bandwidth limit applies to all packets on the interface whose source matches 'cidr'
// 'ingress' bandwidth limit applies to all packets on the interface whose destination matches 'cidr'
// Limits are aggregate limits for the CIDR, not per IP address. CIDRs must be unique, but can be overlapping, traffic
// that matches multiple CIDRs counts against all limits.
Limit(cidr string, egress, ingress *resource.Quantity) error
// Remove a bandwidth limit for a particular CIDR on a particular network interface
Reset(cidr string) error
// Reconcile the interface managed by this shaper with the state on the ground.
ReconcileInterface() error
// Reconcile a CIDR managed by this shaper with the state on the ground
ReconcileCIDR(cidr string, egress, ingress *resource.Quantity) error
// GetCIDRs returns the set of CIDRs that are being managed by this shaper
GetCIDRs() ([]string, error)
}

View File

@@ -1,361 +0,0 @@
// +build linux
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package bandwidth
import (
"bufio"
"bytes"
"encoding/hex"
"fmt"
"net"
"regexp"
"strings"
"k8s.io/apimachinery/pkg/api/resource"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/utils/exec"
"k8s.io/klog/v2"
)
var (
classShowMatcher = regexp.MustCompile(`class htb (1:\d+)`)
classAndHandleMatcher = regexp.MustCompile(`filter parent 1:.*fh (\d+::\d+).*flowid (\d+:\d+)`)
)
// tcShaper provides an implementation of the Shaper interface on Linux using the 'tc' tool.
// In general, using this requires that the caller posses the NET_CAP_ADMIN capability, though if you
// do this within an container, it only requires the NS_CAPABLE capability for manipulations to that
// container's network namespace.
// Uses the hierarchical token bucket queuing discipline (htb), this requires Linux 2.4.20 or newer
// or a custom kernel with that queuing discipline backported.
type tcShaper struct {
e exec.Interface
iface string
}
// NewTCShaper makes a new tcShaper for the given interface
func NewTCShaper(iface string) Shaper {
shaper := &tcShaper{
e: exec.New(),
iface: iface,
}
return shaper
}
func (t *tcShaper) execAndLog(cmdStr string, args ...string) error {
klog.V(6).Infof("Running: %s %s", cmdStr, strings.Join(args, " "))
cmd := t.e.Command(cmdStr, args...)
out, err := cmd.CombinedOutput()
klog.V(6).Infof("Output from tc: %s", string(out))
return err
}
func (t *tcShaper) nextClassID() (int, error) {
data, err := t.e.Command("tc", "class", "show", "dev", t.iface).CombinedOutput()
if err != nil {
return -1, err
}
scanner := bufio.NewScanner(bytes.NewBuffer(data))
classes := sets.String{}
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
// skip empty lines
if len(line) == 0 {
continue
}
// expected tc line:
// class htb 1:1 root prio 0 rate 1000Kbit ceil 1000Kbit burst 1600b cburst 1600b
matches := classShowMatcher.FindStringSubmatch(line)
if len(matches) != 2 {
return -1, fmt.Errorf("unexpected output from tc: %s (%v)", scanner.Text(), matches)
}
classes.Insert(matches[1])
}
// Make sure it doesn't go forever
for nextClass := 1; nextClass < 10000; nextClass++ {
if !classes.Has(fmt.Sprintf("1:%d", nextClass)) {
return nextClass, nil
}
}
// This should really never happen
return -1, fmt.Errorf("exhausted class space, please try again")
}
// Convert a CIDR from text to a hex representation
// Strips any masked parts of the IP, so 1.2.3.4/16 becomes hex(1.2.0.0)/ffffffff
func hexCIDR(cidr string) (string, error) {
ip, ipnet, err := net.ParseCIDR(cidr)
if err != nil {
return "", err
}
ip = ip.Mask(ipnet.Mask)
hexIP := hex.EncodeToString([]byte(ip))
hexMask := ipnet.Mask.String()
return hexIP + "/" + hexMask, nil
}
// Convert a CIDR from hex representation to text, opposite of the above.
func asciiCIDR(cidr string) (string, error) {
parts := strings.Split(cidr, "/")
if len(parts) != 2 {
return "", fmt.Errorf("unexpected CIDR format: %s", cidr)
}
ipData, err := hex.DecodeString(parts[0])
if err != nil {
return "", err
}
ip := net.IP(ipData)
maskData, err := hex.DecodeString(parts[1])
if err != nil {
return "", err
}
mask := net.IPMask(maskData)
size, _ := mask.Size()
return fmt.Sprintf("%s/%d", ip.String(), size), nil
}
func (t *tcShaper) findCIDRClass(cidr string) (classAndHandleList [][]string, found bool, err error) {
data, err := t.e.Command("tc", "filter", "show", "dev", t.iface).CombinedOutput()
if err != nil {
return classAndHandleList, false, err
}
hex, err := hexCIDR(cidr)
if err != nil {
return classAndHandleList, false, err
}
spec := fmt.Sprintf("match %s", hex)
scanner := bufio.NewScanner(bytes.NewBuffer(data))
filter := ""
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
if len(line) == 0 {
continue
}
if strings.HasPrefix(line, "filter") {
filter = line
continue
}
if strings.Contains(line, spec) {
// expected tc line:
// `filter parent 1: protocol ip pref 1 u32 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:1` (old version) or
// `filter parent 1: protocol ip pref 1 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:1 not_in_hw` (new version)
matches := classAndHandleMatcher.FindStringSubmatch(filter)
if len(matches) != 3 {
return classAndHandleList, false, fmt.Errorf("unexpected output from tc: %s %d (%v)", filter, len(matches), matches)
}
resultTmp := []string{matches[2], matches[1]}
classAndHandleList = append(classAndHandleList, resultTmp)
}
}
if len(classAndHandleList) > 0 {
return classAndHandleList, true, nil
}
return classAndHandleList, false, nil
}
func makeKBitString(rsrc *resource.Quantity) string {
return fmt.Sprintf("%dkbit", (rsrc.Value() / 1000))
}
func (t *tcShaper) makeNewClass(rate string) (int, error) {
class, err := t.nextClassID()
if err != nil {
return -1, err
}
if err := t.execAndLog("tc", "class", "add",
"dev", t.iface,
"parent", "1:",
"classid", fmt.Sprintf("1:%d", class),
"htb", "rate", rate); err != nil {
return -1, err
}
return class, nil
}
func (t *tcShaper) Limit(cidr string, upload, download *resource.Quantity) (err error) {
var downloadClass, uploadClass int
if download != nil {
if downloadClass, err = t.makeNewClass(makeKBitString(download)); err != nil {
return err
}
if err := t.execAndLog("tc", "filter", "add",
"dev", t.iface,
"protocol", "ip",
"parent", "1:0",
"prio", "1", "u32",
"match", "ip", "dst", cidr,
"flowid", fmt.Sprintf("1:%d", downloadClass)); err != nil {
return err
}
}
if upload != nil {
if uploadClass, err = t.makeNewClass(makeKBitString(upload)); err != nil {
return err
}
if err := t.execAndLog("tc", "filter", "add",
"dev", t.iface,
"protocol", "ip",
"parent", "1:0",
"prio", "1", "u32",
"match", "ip", "src", cidr,
"flowid", fmt.Sprintf("1:%d", uploadClass)); err != nil {
return err
}
}
return nil
}
// tests to see if an interface exists, if it does, return true and the status line for the interface
// returns false, "", <err> if an error occurs.
func (t *tcShaper) interfaceExists() (bool, string, error) {
data, err := t.e.Command("tc", "qdisc", "show", "dev", t.iface).CombinedOutput()
if err != nil {
return false, "", err
}
value := strings.TrimSpace(string(data))
if len(value) == 0 {
return false, "", nil
}
// Newer versions of tc and/or the kernel return the following instead of nothing:
// qdisc noqueue 0: root refcnt 2
fields := strings.Fields(value)
if len(fields) > 1 && fields[1] == "noqueue" {
return false, "", nil
}
return true, value, nil
}
func (t *tcShaper) ReconcileCIDR(cidr string, upload, download *resource.Quantity) error {
_, found, err := t.findCIDRClass(cidr)
if err != nil {
return err
}
if !found {
return t.Limit(cidr, upload, download)
}
// TODO: actually check bandwidth limits here
return nil
}
func (t *tcShaper) ReconcileInterface() error {
exists, output, err := t.interfaceExists()
if err != nil {
return err
}
if !exists {
klog.V(4).Info("Didn't find bandwidth interface, creating")
return t.initializeInterface()
}
fields := strings.Split(output, " ")
if len(fields) < 12 || fields[1] != "htb" || fields[2] != "1:" {
if err := t.deleteInterface(fields[2]); err != nil {
return err
}
return t.initializeInterface()
}
return nil
}
func (t *tcShaper) initializeInterface() error {
return t.execAndLog("tc", "qdisc", "add", "dev", t.iface, "root", "handle", "1:", "htb", "default", "30")
}
func (t *tcShaper) Reset(cidr string) error {
classAndHandle, found, err := t.findCIDRClass(cidr)
if err != nil {
return err
}
if !found {
return fmt.Errorf("Failed to find cidr: %s on interface: %s", cidr, t.iface)
}
for i := 0; i < len(classAndHandle); i++ {
if err := t.execAndLog("tc", "filter", "del",
"dev", t.iface,
"parent", "1:",
"proto", "ip",
"prio", "1",
"handle", classAndHandle[i][1], "u32"); err != nil {
return err
}
if err := t.execAndLog("tc", "class", "del",
"dev", t.iface,
"parent", "1:",
"classid", classAndHandle[i][0]); err != nil {
return err
}
}
return nil
}
func (t *tcShaper) deleteInterface(class string) error {
return t.execAndLog("tc", "qdisc", "delete", "dev", t.iface, "root", "handle", class)
}
func (t *tcShaper) GetCIDRs() ([]string, error) {
data, err := t.e.Command("tc", "filter", "show", "dev", t.iface).CombinedOutput()
if err != nil {
return nil, err
}
result := []string{}
scanner := bufio.NewScanner(bytes.NewBuffer(data))
for scanner.Scan() {
line := strings.TrimSpace(scanner.Text())
if len(line) == 0 {
continue
}
if strings.Contains(line, "match") {
parts := strings.Split(line, " ")
// expected tc line:
// match <cidr> at <number>
if len(parts) != 4 {
return nil, fmt.Errorf("unexpected output: %v", parts)
}
cidr, err := asciiCIDR(parts[1])
if err != nil {
return nil, err
}
result = append(result, cidr)
}
}
return result, nil
}

View File

@@ -1,69 +0,0 @@
// +build !linux
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package bandwidth
import (
"errors"
"k8s.io/apimachinery/pkg/api/resource"
)
type unsupportedShaper struct {
}
// NewTCShaper makes a new unsupportedShapper for the given interface
func NewTCShaper(iface string) Shaper {
return &unsupportedShaper{}
}
func (f *unsupportedShaper) Limit(cidr string, egress, ingress *resource.Quantity) error {
return errors.New("unimplemented")
}
func (f *unsupportedShaper) Reset(cidr string) error {
return nil
}
func (f *unsupportedShaper) ReconcileInterface() error {
return errors.New("unimplemented")
}
func (f *unsupportedShaper) ReconcileCIDR(cidr string, egress, ingress *resource.Quantity) error {
return errors.New("unimplemented")
}
func (f *unsupportedShaper) GetCIDRs() ([]string, error) {
return []string{}, nil
}

View File

@@ -1,82 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
Copyright 2015 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package bandwidth
import (
"fmt"
"k8s.io/apimachinery/pkg/api/resource"
)
var minRsrc = resource.MustParse("1k")
var maxRsrc = resource.MustParse("1P")
func validateBandwidthIsReasonable(rsrc *resource.Quantity) error {
if rsrc.Value() < minRsrc.Value() {
return fmt.Errorf("resource is unreasonably small (< 1kbit)")
}
if rsrc.Value() > maxRsrc.Value() {
return fmt.Errorf("resoruce is unreasonably large (> 1Pbit)")
}
return nil
}
// ExtractPodBandwidthResources extracts the ingress and egress from the given pod annotations
func ExtractPodBandwidthResources(podAnnotations map[string]string) (ingress, egress *resource.Quantity, err error) {
if podAnnotations == nil {
return nil, nil, nil
}
str, found := podAnnotations["kubernetes.io/ingress-bandwidth"]
if found {
ingressValue, err := resource.ParseQuantity(str)
if err != nil {
return nil, nil, err
}
ingress = &ingressValue
if err := validateBandwidthIsReasonable(ingress); err != nil {
return nil, nil, err
}
}
str, found = podAnnotations["kubernetes.io/egress-bandwidth"]
if found {
egressValue, err := resource.ParseQuantity(str)
if err != nil {
return nil, nil, err
}
egress = &egressValue
if err := validateBandwidthIsReasonable(egress); err != nil {
return nil, nil, err
}
}
return ingress, egress, nil
}

View File

@@ -1,121 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"os"
"sync"
cni "github.com/containerd/go-cni"
"github.com/fsnotify/fsnotify"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
)
// cniNetConfSyncer is used to reload cni network conf triggered by fs change
// events.
type cniNetConfSyncer struct {
// only used for lastSyncStatus
sync.RWMutex
lastSyncStatus error
watcher *fsnotify.Watcher
confDir string
netPlugin cni.CNI
loadOpts []cni.CNIOpt
}
// newCNINetConfSyncer creates cni network conf syncer.
func newCNINetConfSyncer(confDir string, netPlugin cni.CNI, loadOpts []cni.CNIOpt) (*cniNetConfSyncer, error) {
watcher, err := fsnotify.NewWatcher()
if err != nil {
return nil, errors.Wrap(err, "failed to create fsnotify watcher")
}
if err := os.MkdirAll(confDir, 0700); err != nil {
return nil, errors.Wrapf(err, "failed to create cni conf dir=%s for watch", confDir)
}
if err := watcher.Add(confDir); err != nil {
return nil, errors.Wrapf(err, "failed to watch cni conf dir %s", confDir)
}
syncer := &cniNetConfSyncer{
watcher: watcher,
confDir: confDir,
netPlugin: netPlugin,
loadOpts: loadOpts,
}
if err := syncer.netPlugin.Load(syncer.loadOpts...); err != nil {
logrus.WithError(err).Error("failed to load cni during init, please check CRI plugin status before setting up network for pods")
syncer.updateLastStatus(err)
}
return syncer, nil
}
// syncLoop monitors any fs change events from cni conf dir and tries to reload
// cni configuration.
func (syncer *cniNetConfSyncer) syncLoop() error {
for {
select {
case event := <-syncer.watcher.Events:
// Only reload config when receiving write/rename/remove
// events
//
// TODO(fuweid): Might only reload target cni config
// files to prevent no-ops.
if event.Op&(fsnotify.Chmod|fsnotify.Create) > 0 {
logrus.Debugf("ignore event from cni conf dir: %s", event)
continue
}
logrus.Debugf("receiving change event from cni conf dir: %s", event)
lerr := syncer.netPlugin.Load(syncer.loadOpts...)
if lerr != nil {
logrus.WithError(lerr).
Errorf("failed to reload cni configuration after receiving fs change event(%s)", event)
}
syncer.updateLastStatus(lerr)
case err := <-syncer.watcher.Errors:
if err != nil {
logrus.WithError(err).Error("failed to continue sync cni conf change")
return err
}
}
}
}
// lastStatus retrieves last sync status.
func (syncer *cniNetConfSyncer) lastStatus() error {
syncer.RLock()
defer syncer.RUnlock()
return syncer.lastSyncStatus
}
// updateLastStatus will be called after every single cni load.
func (syncer *cniNetConfSyncer) updateLastStatus(err error) {
syncer.Lock()
defer syncer.Unlock()
syncer.lastSyncStatus = err
}
// stop stops watcher in the syncLoop.
func (syncer *cniNetConfSyncer) stop() error {
return syncer.watcher.Close()
}

View File

@@ -1,84 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"io"
"github.com/containerd/containerd"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"golang.org/x/net/context"
"k8s.io/client-go/tools/remotecommand"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
cio "github.com/containerd/cri/pkg/server/io"
)
// Attach prepares a streaming endpoint to attach to a running container, and returns the address.
func (c *criService) Attach(ctx context.Context, r *runtime.AttachRequest) (*runtime.AttachResponse, error) {
cntr, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrap(err, "failed to find container in store")
}
state := cntr.Status.Get().State()
if state != runtime.ContainerState_CONTAINER_RUNNING {
return nil, errors.Errorf("container is in %s state", criContainerStateToString(state))
}
return c.streamServer.GetAttach(r)
}
func (c *criService) attachContainer(ctx context.Context, id string, stdin io.Reader, stdout, stderr io.WriteCloser,
tty bool, resize <-chan remotecommand.TerminalSize) error {
ctx, cancel := context.WithCancel(ctx)
defer cancel()
// Get container from our container store.
cntr, err := c.containerStore.Get(id)
if err != nil {
return errors.Wrapf(err, "failed to find container %q in store", id)
}
id = cntr.ID
state := cntr.Status.Get().State()
if state != runtime.ContainerState_CONTAINER_RUNNING {
return errors.Errorf("container is in %s state", criContainerStateToString(state))
}
task, err := cntr.Container.Task(ctx, nil)
if err != nil {
return errors.Wrap(err, "failed to load task")
}
handleResizing(ctx, resize, func(size remotecommand.TerminalSize) {
if err := task.Resize(ctx, uint32(size.Width), uint32(size.Height)); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to resize task %q console", id)
}
})
opts := cio.AttachOptions{
Stdin: stdin,
Stdout: stdout,
Stderr: stderr,
Tty: tty,
StdinOnce: cntr.Config.StdinOnce,
CloseStdin: func() error {
return task.CloseIO(ctx, containerd.WithStdinCloser)
},
}
// TODO(random-liu): Figure out whether we need to support historical output.
cntr.IO.Attach(opts)
return nil
}

View File

@@ -1,343 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"path/filepath"
"time"
"github.com/containerd/containerd"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/oci"
"github.com/containerd/typeurl"
"github.com/davecgh/go-spew/spew"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
selinux "github.com/opencontainers/selinux/go-selinux"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
customopts "github.com/containerd/cri/pkg/containerd/opts"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
cio "github.com/containerd/cri/pkg/server/io"
containerstore "github.com/containerd/cri/pkg/store/container"
"github.com/containerd/cri/pkg/util"
)
func init() {
typeurl.Register(&containerstore.Metadata{},
"github.com/containerd/cri/pkg/store/container", "Metadata")
}
// CreateContainer creates a new container in the given PodSandbox.
func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateContainerRequest) (_ *runtime.CreateContainerResponse, retErr error) {
config := r.GetConfig()
log.G(ctx).Debugf("Container config %+v", config)
sandboxConfig := r.GetSandboxConfig()
sandbox, err := c.sandboxStore.Get(r.GetPodSandboxId())
if err != nil {
return nil, errors.Wrapf(err, "failed to find sandbox id %q", r.GetPodSandboxId())
}
sandboxID := sandbox.ID
s, err := sandbox.Container.Task(ctx, nil)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox container task")
}
sandboxPid := s.Pid()
// Generate unique id and name for the container and reserve the name.
// Reserve the container name to avoid concurrent `CreateContainer` request creating
// the same container.
id := util.GenerateID()
metadata := config.GetMetadata()
if metadata == nil {
return nil, errors.New("container config must include metadata")
}
containerName := metadata.Name
name := makeContainerName(metadata, sandboxConfig.GetMetadata())
log.G(ctx).Debugf("Generated id %q for container %q", id, name)
if err = c.containerNameIndex.Reserve(name, id); err != nil {
return nil, errors.Wrapf(err, "failed to reserve container name %q", name)
}
defer func() {
// Release the name if the function returns with an error.
if retErr != nil {
c.containerNameIndex.ReleaseByName(name)
}
}()
// Create initial internal container metadata.
meta := containerstore.Metadata{
ID: id,
Name: name,
SandboxID: sandboxID,
Config: config,
}
// Prepare container image snapshot. For container, the image should have
// been pulled before creating the container, so do not ensure the image.
image, err := c.localResolve(config.GetImage().GetImage())
if err != nil {
return nil, errors.Wrapf(err, "failed to resolve image %q", config.GetImage().GetImage())
}
containerdImage, err := c.toContainerdImage(ctx, image)
if err != nil {
return nil, errors.Wrapf(err, "failed to get image from containerd %q", image.ID)
}
// Run container using the same runtime with sandbox.
sandboxInfo, err := sandbox.Container.Info(ctx)
if err != nil {
return nil, errors.Wrapf(err, "failed to get sandbox %q info", sandboxID)
}
// Create container root directory.
containerRootDir := c.getContainerRootDir(id)
if err = c.os.MkdirAll(containerRootDir, 0755); err != nil {
return nil, errors.Wrapf(err, "failed to create container root directory %q",
containerRootDir)
}
defer func() {
if retErr != nil {
// Cleanup the container root directory.
if err = c.os.RemoveAll(containerRootDir); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to remove container root directory %q",
containerRootDir)
}
}
}()
volatileContainerRootDir := c.getVolatileContainerRootDir(id)
if err = c.os.MkdirAll(volatileContainerRootDir, 0755); err != nil {
return nil, errors.Wrapf(err, "failed to create volatile container root directory %q",
volatileContainerRootDir)
}
defer func() {
if retErr != nil {
// Cleanup the volatile container root directory.
if err = c.os.RemoveAll(volatileContainerRootDir); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to remove volatile container root directory %q",
volatileContainerRootDir)
}
}
}()
var volumeMounts []*runtime.Mount
if !c.config.IgnoreImageDefinedVolumes {
// Create container image volumes mounts.
volumeMounts = c.volumeMounts(containerRootDir, config.GetMounts(), &image.ImageSpec.Config)
} else if len(image.ImageSpec.Config.Volumes) != 0 {
log.G(ctx).Debugf("Ignoring volumes defined in image %v because IgnoreImageDefinedVolumes is set", image.ID)
}
// Generate container mounts.
mounts := c.containerMounts(sandboxID, config)
ociRuntime, err := c.getSandboxRuntime(sandboxConfig, sandbox.Metadata.RuntimeHandler)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox runtime")
}
log.G(ctx).Debugf("Use OCI runtime %+v for sandbox %q and container %q", ociRuntime, sandboxID, id)
spec, err := c.containerSpec(id, sandboxID, sandboxPid, sandbox.NetNSPath, containerName, config, sandboxConfig,
&image.ImageSpec.Config, append(mounts, volumeMounts...), ociRuntime)
if err != nil {
return nil, errors.Wrapf(err, "failed to generate container %q spec", id)
}
meta.ProcessLabel = spec.Process.SelinuxLabel
// handle any KVM based runtime
if err := modifyProcessLabel(ociRuntime.Type, spec); err != nil {
return nil, err
}
if config.GetLinux().GetSecurityContext().GetPrivileged() {
// If privileged don't set the SELinux label but still record it on the container so
// the unused MCS label can be release later
spec.Process.SelinuxLabel = ""
}
defer func() {
if retErr != nil {
selinux.ReleaseLabel(spec.Process.SelinuxLabel)
}
}()
log.G(ctx).Debugf("Container %q spec: %#+v", id, spew.NewFormatter(spec))
// Set snapshotter before any other options.
opts := []containerd.NewContainerOpts{
containerd.WithSnapshotter(c.config.ContainerdConfig.Snapshotter),
// Prepare container rootfs. This is always writeable even if
// the container wants a readonly rootfs since we want to give
// the runtime (runc) a chance to modify (e.g. to create mount
// points corresponding to spec.Mounts) before making the
// rootfs readonly (requested by spec.Root.Readonly).
customopts.WithNewSnapshot(id, containerdImage),
}
if len(volumeMounts) > 0 {
mountMap := make(map[string]string)
for _, v := range volumeMounts {
mountMap[filepath.Clean(v.HostPath)] = v.ContainerPath
}
opts = append(opts, customopts.WithVolumes(mountMap))
}
meta.ImageRef = image.ID
meta.StopSignal = image.ImageSpec.Config.StopSignal
// Validate log paths and compose full container log path.
if sandboxConfig.GetLogDirectory() != "" && config.GetLogPath() != "" {
meta.LogPath = filepath.Join(sandboxConfig.GetLogDirectory(), config.GetLogPath())
log.G(ctx).Debugf("Composed container full log path %q using sandbox log dir %q and container log path %q",
meta.LogPath, sandboxConfig.GetLogDirectory(), config.GetLogPath())
} else {
log.G(ctx).Infof("Logging will be disabled due to empty log paths for sandbox (%q) or container (%q)",
sandboxConfig.GetLogDirectory(), config.GetLogPath())
}
containerIO, err := cio.NewContainerIO(id,
cio.WithNewFIFOs(volatileContainerRootDir, config.GetTty(), config.GetStdin()))
if err != nil {
return nil, errors.Wrap(err, "failed to create container io")
}
defer func() {
if retErr != nil {
if err := containerIO.Close(); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to close container io %q", id)
}
}
}()
specOpts, err := c.containerSpecOpts(config, &image.ImageSpec.Config)
if err != nil {
return nil, errors.Wrap(err, "")
}
containerLabels := buildLabels(config.Labels, containerKindContainer)
runtimeOptions, err := getRuntimeOptions(sandboxInfo)
if err != nil {
return nil, errors.Wrap(err, "failed to get runtime options")
}
opts = append(opts,
containerd.WithSpec(spec, specOpts...),
containerd.WithRuntime(sandboxInfo.Runtime.Name, runtimeOptions),
containerd.WithContainerLabels(containerLabels),
containerd.WithContainerExtension(containerMetadataExtension, &meta))
var cntr containerd.Container
if cntr, err = c.client.NewContainer(ctx, id, opts...); err != nil {
return nil, errors.Wrap(err, "failed to create containerd container")
}
defer func() {
if retErr != nil {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
if err := cntr.Delete(deferCtx, containerd.WithSnapshotCleanup); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to delete containerd container %q", id)
}
}
}()
status := containerstore.Status{CreatedAt: time.Now().UnixNano()}
container, err := containerstore.NewContainer(meta,
containerstore.WithStatus(status, containerRootDir),
containerstore.WithContainer(cntr),
containerstore.WithContainerIO(containerIO),
)
if err != nil {
return nil, errors.Wrapf(err, "failed to create internal container object for %q", id)
}
defer func() {
if retErr != nil {
// Cleanup container checkpoint on error.
if err := container.Delete(); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to cleanup container checkpoint for %q", id)
}
}
}()
// Add container into container store.
if err := c.containerStore.Add(container); err != nil {
return nil, errors.Wrapf(err, "failed to add container %q into store", id)
}
return &runtime.CreateContainerResponse{ContainerId: id}, nil
}
// volumeMounts sets up image volumes for container. Rely on the removal of container
// root directory to do cleanup. Note that image volume will be skipped, if there is criMounts
// specified with the same destination.
func (c *criService) volumeMounts(containerRootDir string, criMounts []*runtime.Mount, config *imagespec.ImageConfig) []*runtime.Mount {
if len(config.Volumes) == 0 {
return nil
}
var mounts []*runtime.Mount
for dst := range config.Volumes {
if isInCRIMounts(dst, criMounts) {
// Skip the image volume, if there is CRI defined volume mapping.
// TODO(random-liu): This should be handled by Kubelet in the future.
// Kubelet should decide what to use for image volume, and also de-duplicate
// the image volume and user mounts.
continue
}
volumeID := util.GenerateID()
src := filepath.Join(containerRootDir, "volumes", volumeID)
// addOCIBindMounts will create these volumes.
mounts = append(mounts, &runtime.Mount{
ContainerPath: dst,
HostPath: src,
SelinuxRelabel: true,
})
}
return mounts
}
// runtimeSpec returns a default runtime spec used in cri-containerd.
func (c *criService) runtimeSpec(id string, baseSpecFile string, opts ...oci.SpecOpts) (*runtimespec.Spec, error) {
// GenerateSpec needs namespace.
ctx := ctrdutil.NamespacedContext()
container := &containers.Container{ID: id}
if baseSpecFile != "" {
baseSpec, ok := c.baseOCISpecs[baseSpecFile]
if !ok {
return nil, errors.Errorf("can't find base OCI spec %q", baseSpecFile)
}
spec := oci.Spec{}
if err := util.DeepCopy(&spec, &baseSpec); err != nil {
return nil, errors.Wrap(err, "failed to clone OCI spec")
}
// Fix up cgroups path
applyOpts := append([]oci.SpecOpts{oci.WithNamespacedCgroup()}, opts...)
if err := oci.ApplyOpts(ctx, nil, container, &spec, applyOpts...); err != nil {
return nil, errors.Wrap(err, "failed to apply OCI options")
}
return &spec, nil
}
spec, err := oci.GenerateSpec(ctx, nil, container, opts...)
if err != nil {
return nil, errors.Wrap(err, "failed to generate spec")
}
return spec, nil
}

View File

@@ -1,460 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"bufio"
"io"
"os"
"strconv"
"strings"
"github.com/containerd/cgroups"
"github.com/containerd/containerd/contrib/apparmor"
"github.com/containerd/containerd/contrib/seccomp"
"github.com/containerd/containerd/oci"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
selinux "github.com/opencontainers/selinux/go-selinux"
"github.com/opencontainers/selinux/go-selinux/label"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/annotations"
"github.com/containerd/cri/pkg/config"
customopts "github.com/containerd/cri/pkg/containerd/opts"
)
const (
// profileNamePrefix is the prefix for loading profiles on a localhost. Eg. AppArmor localhost/profileName.
profileNamePrefix = "localhost/" // TODO (mikebrow): get localhost/ & runtime/default from CRI kubernetes/kubernetes#51747
// runtimeDefault indicates that we should use or create a runtime default profile.
runtimeDefault = "runtime/default"
// dockerDefault indicates that we should use or create a docker default profile.
dockerDefault = "docker/default"
// appArmorDefaultProfileName is name to use when creating a default apparmor profile.
appArmorDefaultProfileName = "cri-containerd.apparmor.d"
// unconfinedProfile is a string indicating one should run a pod/containerd without a security profile
unconfinedProfile = "unconfined"
// seccompDefaultProfile is the default seccomp profile.
seccompDefaultProfile = dockerDefault
)
// containerMounts sets up necessary container system file mounts
// including /dev/shm, /etc/hosts and /etc/resolv.conf.
func (c *criService) containerMounts(sandboxID string, config *runtime.ContainerConfig) []*runtime.Mount {
var mounts []*runtime.Mount
securityContext := config.GetLinux().GetSecurityContext()
if !isInCRIMounts(etcHostname, config.GetMounts()) {
// /etc/hostname is added since 1.1.6, 1.2.4 and 1.3.
// For in-place upgrade, the old sandbox doesn't have the hostname file,
// do not mount this in that case.
// TODO(random-liu): Remove the check and always mount this when
// containerd 1.1 and 1.2 are deprecated.
hostpath := c.getSandboxHostname(sandboxID)
if _, err := c.os.Stat(hostpath); err == nil {
mounts = append(mounts, &runtime.Mount{
ContainerPath: etcHostname,
HostPath: hostpath,
Readonly: securityContext.GetReadonlyRootfs(),
})
}
}
if !isInCRIMounts(etcHosts, config.GetMounts()) {
mounts = append(mounts, &runtime.Mount{
ContainerPath: etcHosts,
HostPath: c.getSandboxHosts(sandboxID),
Readonly: securityContext.GetReadonlyRootfs(),
})
}
// Mount sandbox resolv.config.
// TODO: Need to figure out whether we should always mount it as read-only
if !isInCRIMounts(resolvConfPath, config.GetMounts()) {
mounts = append(mounts, &runtime.Mount{
ContainerPath: resolvConfPath,
HostPath: c.getResolvPath(sandboxID),
Readonly: securityContext.GetReadonlyRootfs(),
})
}
if !isInCRIMounts(devShm, config.GetMounts()) {
sandboxDevShm := c.getSandboxDevShm(sandboxID)
if securityContext.GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
sandboxDevShm = devShm
}
mounts = append(mounts, &runtime.Mount{
ContainerPath: devShm,
HostPath: sandboxDevShm,
Readonly: false,
})
}
return mounts
}
func (c *criService) containerSpec(id string, sandboxID string, sandboxPid uint32, netNSPath string, containerName string,
config *runtime.ContainerConfig, sandboxConfig *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig,
extraMounts []*runtime.Mount, ociRuntime config.Runtime) (_ *runtimespec.Spec, retErr error) {
specOpts := []oci.SpecOpts{
customopts.WithoutRunMount,
customopts.WithoutDefaultSecuritySettings,
customopts.WithRelativeRoot(relativeRootfsPath),
customopts.WithProcessArgs(config, imageConfig),
oci.WithDefaultPathEnv,
// this will be set based on the security context below
oci.WithNewPrivileges,
}
if config.GetWorkingDir() != "" {
specOpts = append(specOpts, oci.WithProcessCwd(config.GetWorkingDir()))
} else if imageConfig.WorkingDir != "" {
specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
}
if config.GetTty() {
specOpts = append(specOpts, oci.WithTTY)
}
// Add HOSTNAME env.
var (
err error
hostname = sandboxConfig.GetHostname()
)
if hostname == "" {
if hostname, err = c.os.Hostname(); err != nil {
return nil, err
}
}
specOpts = append(specOpts, oci.WithEnv([]string{hostnameEnv + "=" + hostname}))
// Apply envs from image config first, so that envs from container config
// can override them.
env := imageConfig.Env
for _, e := range config.GetEnvs() {
env = append(env, e.GetKey()+"="+e.GetValue())
}
specOpts = append(specOpts, oci.WithEnv(env))
securityContext := config.GetLinux().GetSecurityContext()
labelOptions, err := toLabel(securityContext.GetSelinuxOptions())
if err != nil {
return nil, err
}
if len(labelOptions) == 0 {
// Use pod level SELinux config
if sandbox, err := c.sandboxStore.Get(sandboxID); err == nil {
labelOptions, err = selinux.DupSecOpt(sandbox.ProcessLabel)
if err != nil {
return nil, err
}
}
}
processLabel, mountLabel, err := label.InitLabels(labelOptions)
if err != nil {
return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
}
defer func() {
if retErr != nil {
_ = label.ReleaseLabel(processLabel)
}
}()
specOpts = append(specOpts, customopts.WithMounts(c.os, config, extraMounts, mountLabel))
if !c.config.DisableProcMount {
// Apply masked paths if specified.
// If the container is privileged, this will be cleared later on.
specOpts = append(specOpts, oci.WithMaskedPaths(securityContext.GetMaskedPaths()))
// Apply readonly paths if specified.
// If the container is privileged, this will be cleared later on.
specOpts = append(specOpts, oci.WithReadonlyPaths(securityContext.GetReadonlyPaths()))
}
if securityContext.GetPrivileged() {
if !sandboxConfig.GetLinux().GetSecurityContext().GetPrivileged() {
return nil, errors.New("no privileged container allowed in sandbox")
}
specOpts = append(specOpts, oci.WithPrivileged)
if !ociRuntime.PrivilegedWithoutHostDevices {
specOpts = append(specOpts, oci.WithHostDevices, oci.WithAllDevicesAllowed)
} else {
// add requested devices by the config as host devices are not automatically added
specOpts = append(specOpts, customopts.WithDevices(c.os, config), customopts.WithCapabilities(securityContext))
}
} else { // not privileged
specOpts = append(specOpts, customopts.WithDevices(c.os, config), customopts.WithCapabilities(securityContext))
}
// Clear all ambient capabilities. The implication of non-root + caps
// is not clearly defined in Kubernetes.
// See https://github.com/kubernetes/kubernetes/issues/56374
// Keep docker's behavior for now.
specOpts = append(specOpts,
customopts.WithoutAmbientCaps,
customopts.WithSelinuxLabels(processLabel, mountLabel),
)
// TODO: Figure out whether we should set no new privilege for sandbox container by default
if securityContext.GetNoNewPrivs() {
specOpts = append(specOpts, oci.WithNoNewPrivileges)
}
// TODO(random-liu): [P1] Set selinux options (privileged or not).
if securityContext.GetReadonlyRootfs() {
specOpts = append(specOpts, oci.WithRootFSReadonly())
}
if c.config.DisableCgroup {
specOpts = append(specOpts, customopts.WithDisabledCgroups)
} else {
specOpts = append(specOpts, customopts.WithResources(config.GetLinux().GetResources(), c.config.TolerateMissingHugetlbController, c.config.DisableHugetlbController))
if sandboxConfig.GetLinux().GetCgroupParent() != "" {
cgroupsPath := getCgroupsPath(sandboxConfig.GetLinux().GetCgroupParent(), id)
specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
}
}
supplementalGroups := securityContext.GetSupplementalGroups()
for pKey, pValue := range getPassthroughAnnotations(sandboxConfig.Annotations,
ociRuntime.PodAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
ociRuntime.ContainerAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
specOpts = append(specOpts,
customopts.WithOOMScoreAdj(config, c.config.RestrictOOMScoreAdj),
customopts.WithPodNamespaces(securityContext, sandboxPid),
customopts.WithSupplementalGroups(supplementalGroups),
customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeContainer),
customopts.WithAnnotation(annotations.SandboxID, sandboxID),
customopts.WithAnnotation(annotations.ContainerName, containerName),
)
// cgroupns is used for hiding /sys/fs/cgroup from containers.
// For compatibility, cgroupns is not used when running in cgroup v1 mode or in privileged.
// https://github.com/containers/libpod/issues/4363
// https://github.com/kubernetes/enhancements/blob/0e409b47497e398b369c281074485c8de129694f/keps/sig-node/20191118-cgroups-v2.md#cgroup-namespace
if cgroups.Mode() == cgroups.Unified && !securityContext.GetPrivileged() {
specOpts = append(specOpts, oci.WithLinuxNamespace(
runtimespec.LinuxNamespace{
Type: runtimespec.CgroupNamespace,
}))
}
return c.runtimeSpec(id, ociRuntime.BaseRuntimeSpec, specOpts...)
}
func (c *criService) containerSpecOpts(config *runtime.ContainerConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
var specOpts []oci.SpecOpts
securityContext := config.GetLinux().GetSecurityContext()
// Set container username. This could only be done by containerd, because it needs
// access to the container rootfs. Pass user name to containerd, and let it overwrite
// the spec for us.
userstr, err := generateUserString(
securityContext.GetRunAsUsername(),
securityContext.GetRunAsUser(),
securityContext.GetRunAsGroup())
if err != nil {
return nil, errors.Wrap(err, "failed to generate user string")
}
if userstr == "" {
// Lastly, since no user override was passed via CRI try to set via OCI
// Image
userstr = imageConfig.User
}
if userstr != "" {
specOpts = append(specOpts, oci.WithUser(userstr))
}
if securityContext.GetRunAsUsername() != "" {
userstr = securityContext.GetRunAsUsername()
} else {
// Even if RunAsUser is not set, we still call `GetValue` to get uid 0.
// Because it is still useful to get additional gids for uid 0.
userstr = strconv.FormatInt(securityContext.GetRunAsUser().GetValue(), 10)
}
specOpts = append(specOpts, customopts.WithAdditionalGIDs(userstr))
apparmorSpecOpts, err := generateApparmorSpecOpts(
securityContext.GetApparmorProfile(),
securityContext.GetPrivileged(),
c.apparmorEnabled())
if err != nil {
return nil, errors.Wrap(err, "failed to generate apparmor spec opts")
}
if apparmorSpecOpts != nil {
specOpts = append(specOpts, apparmorSpecOpts)
}
seccompSpecOpts, err := c.generateSeccompSpecOpts(
securityContext.GetSeccompProfilePath(),
securityContext.GetPrivileged(),
c.seccompEnabled())
if err != nil {
return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
}
if seccompSpecOpts != nil {
specOpts = append(specOpts, seccompSpecOpts)
}
return specOpts, nil
}
// generateSeccompSpecOpts generates containerd SpecOpts for seccomp.
func (c *criService) generateSeccompSpecOpts(seccompProf string, privileged, seccompEnabled bool) (oci.SpecOpts, error) {
if privileged {
// Do not set seccomp profile when container is privileged
return nil, nil
}
if seccompProf == "" {
seccompProf = c.config.UnsetSeccompProfile
}
// Set seccomp profile
if seccompProf == runtimeDefault || seccompProf == dockerDefault {
// use correct default profile (Eg. if not configured otherwise, the default is docker/default)
seccompProf = seccompDefaultProfile
}
if !seccompEnabled {
if seccompProf != "" && seccompProf != unconfinedProfile {
return nil, errors.New("seccomp is not supported")
}
return nil, nil
}
switch seccompProf {
case "", unconfinedProfile:
// Do not set seccomp profile.
return nil, nil
case dockerDefault:
// Note: WithDefaultProfile specOpts must be added after capabilities
return seccomp.WithDefaultProfile(), nil
default:
// Require and Trim default profile name prefix
if !strings.HasPrefix(seccompProf, profileNamePrefix) {
return nil, errors.Errorf("invalid seccomp profile %q", seccompProf)
}
return seccomp.WithProfile(strings.TrimPrefix(seccompProf, profileNamePrefix)), nil
}
}
// generateApparmorSpecOpts generates containerd SpecOpts for apparmor.
func generateApparmorSpecOpts(apparmorProf string, privileged, apparmorEnabled bool) (oci.SpecOpts, error) {
if !apparmorEnabled {
// Should fail loudly if user try to specify apparmor profile
// but we don't support it.
if apparmorProf != "" && apparmorProf != unconfinedProfile {
return nil, errors.New("apparmor is not supported")
}
return nil, nil
}
switch apparmorProf {
// Based on kubernetes#51746, default apparmor profile should be applied
// for when apparmor is not specified.
case runtimeDefault, "":
if privileged {
// Do not set apparmor profile when container is privileged
return nil, nil
}
// TODO (mikebrow): delete created apparmor default profile
return apparmor.WithDefaultProfile(appArmorDefaultProfileName), nil
case unconfinedProfile:
return nil, nil
default:
// Require and Trim default profile name prefix
if !strings.HasPrefix(apparmorProf, profileNamePrefix) {
return nil, errors.Errorf("invalid apparmor profile %q", apparmorProf)
}
appArmorProfile := strings.TrimPrefix(apparmorProf, profileNamePrefix)
if profileExists, err := appArmorProfileExists(appArmorProfile); !profileExists {
if err != nil {
return nil, errors.Wrap(err, "failed to generate apparmor spec opts")
}
return nil, errors.Errorf("apparmor profile not found %s", appArmorProfile)
}
return apparmor.WithProfile(appArmorProfile), nil
}
}
// appArmorProfileExists scans apparmor/profiles for the requested profile
func appArmorProfileExists(profile string) (bool, error) {
if profile == "" {
return false, errors.New("nil apparmor profile is not supported")
}
profiles, err := os.Open("/sys/kernel/security/apparmor/profiles")
if err != nil {
return false, err
}
defer profiles.Close()
rbuff := bufio.NewReader(profiles)
for {
line, err := rbuff.ReadString('\n')
switch err {
case nil:
if strings.HasPrefix(line, profile+" (") {
return true, nil
}
case io.EOF:
return false, nil
default:
return false, err
}
}
}
// generateUserString generates valid user string based on OCI Image Spec
// v1.0.0.
//
// CRI defines that the following combinations are valid:
//
// (none) -> ""
// username -> username
// username, uid -> username
// username, uid, gid -> username:gid
// username, gid -> username:gid
// uid -> uid
// uid, gid -> uid:gid
// gid -> error
//
// TODO(random-liu): Add group name support in CRI.
func generateUserString(username string, uid, gid *runtime.Int64Value) (string, error) {
var userstr, groupstr string
if uid != nil {
userstr = strconv.FormatInt(uid.GetValue(), 10)
}
if username != "" {
userstr = username
}
if gid != nil {
groupstr = strconv.FormatInt(gid.GetValue(), 10)
}
if userstr == "" {
if groupstr != "" {
return "", errors.Errorf("user group %q is specified without user", groupstr)
}
return "", nil
}
if groupstr != "" {
userstr = userstr + ":" + groupstr
}
return userstr, nil
}

View File

@@ -1,117 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd/oci"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/annotations"
"github.com/containerd/cri/pkg/config"
customopts "github.com/containerd/cri/pkg/containerd/opts"
)
// No container mounts for windows.
func (c *criService) containerMounts(sandboxID string, config *runtime.ContainerConfig) []*runtime.Mount {
return nil
}
func (c *criService) containerSpec(id string, sandboxID string, sandboxPid uint32, netNSPath string, containerName string,
config *runtime.ContainerConfig, sandboxConfig *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig,
extraMounts []*runtime.Mount, ociRuntime config.Runtime) (*runtimespec.Spec, error) {
specOpts := []oci.SpecOpts{
customopts.WithProcessArgs(config, imageConfig),
}
if config.GetWorkingDir() != "" {
specOpts = append(specOpts, oci.WithProcessCwd(config.GetWorkingDir()))
} else if imageConfig.WorkingDir != "" {
specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
}
if config.GetTty() {
specOpts = append(specOpts, oci.WithTTY)
}
// Apply envs from image config first, so that envs from container config
// can override them.
env := imageConfig.Env
for _, e := range config.GetEnvs() {
env = append(env, e.GetKey()+"="+e.GetValue())
}
specOpts = append(specOpts, oci.WithEnv(env))
specOpts = append(specOpts,
// Clear the root location since hcsshim expects it.
// NOTE: readonly rootfs doesn't work on windows.
customopts.WithoutRoot,
customopts.WithWindowsNetworkNamespace(netNSPath),
oci.WithHostname(sandboxConfig.GetHostname()),
)
specOpts = append(specOpts, customopts.WithWindowsMounts(c.os, config, extraMounts))
// Start with the image config user and override below if RunAsUsername is not "".
username := imageConfig.User
windowsConfig := config.GetWindows()
if windowsConfig != nil {
specOpts = append(specOpts, customopts.WithWindowsResources(windowsConfig.GetResources()))
securityCtx := windowsConfig.GetSecurityContext()
if securityCtx != nil {
runAsUser := securityCtx.GetRunAsUsername()
if runAsUser != "" {
username = runAsUser
}
cs := securityCtx.GetCredentialSpec()
if cs != "" {
specOpts = append(specOpts, customopts.WithWindowsCredentialSpec(cs))
}
}
}
// There really isn't a good Windows way to verify that the username is available in the
// image as early as here like there is for Linux. Later on in the stack hcsshim
// will handle the behavior of erroring out if the user isn't available in the image
// when trying to run the init process.
specOpts = append(specOpts, oci.WithUser(username))
for pKey, pValue := range getPassthroughAnnotations(sandboxConfig.Annotations,
ociRuntime.PodAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
ociRuntime.ContainerAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
specOpts = append(specOpts,
customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeContainer),
customopts.WithAnnotation(annotations.SandboxID, sandboxID),
customopts.WithAnnotation(annotations.ContainerName, containerName),
)
return c.runtimeSpec(id, ociRuntime.BaseRuntimeSpec, specOpts...)
}
// No extra spec options needed for windows.
func (c *criService) containerSpecOpts(config *runtime.ContainerConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
return nil, nil
}

View File

@@ -1,36 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// Exec prepares a streaming endpoint to execute a command in the container, and returns the address.
func (c *criService) Exec(ctx context.Context, r *runtime.ExecRequest) (*runtime.ExecResponse, error) {
cntr, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrapf(err, "failed to find container %q in store", r.GetContainerId())
}
state := cntr.Status.Get().State()
if state != runtime.ContainerState_CONTAINER_RUNNING {
return nil, errors.Errorf("container is in %s state", criContainerStateToString(state))
}
return c.streamServer.GetExec(r)
}

View File

@@ -1,211 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"bytes"
"io"
"syscall"
"time"
"github.com/containerd/containerd"
containerdio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/oci"
"github.com/pkg/errors"
"golang.org/x/net/context"
"k8s.io/client-go/tools/remotecommand"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
cioutil "github.com/containerd/cri/pkg/ioutil"
cio "github.com/containerd/cri/pkg/server/io"
"github.com/containerd/cri/pkg/util"
)
// ExecSync executes a command in the container, and returns the stdout output.
// If command exits with a non-zero exit code, an error is returned.
func (c *criService) ExecSync(ctx context.Context, r *runtime.ExecSyncRequest) (*runtime.ExecSyncResponse, error) {
var stdout, stderr bytes.Buffer
exitCode, err := c.execInContainer(ctx, r.GetContainerId(), execOptions{
cmd: r.GetCmd(),
stdout: cioutil.NewNopWriteCloser(&stdout),
stderr: cioutil.NewNopWriteCloser(&stderr),
timeout: time.Duration(r.GetTimeout()) * time.Second,
})
if err != nil {
return nil, errors.Wrap(err, "failed to exec in container")
}
return &runtime.ExecSyncResponse{
Stdout: stdout.Bytes(),
Stderr: stderr.Bytes(),
ExitCode: int32(*exitCode),
}, nil
}
// execOptions specifies how to execute command in container.
type execOptions struct {
cmd []string
stdin io.Reader
stdout io.WriteCloser
stderr io.WriteCloser
tty bool
resize <-chan remotecommand.TerminalSize
timeout time.Duration
}
func (c *criService) execInternal(ctx context.Context, container containerd.Container, id string, opts execOptions) (*uint32, error) {
// Cancel the context before returning to ensure goroutines are stopped.
// This is important, because if `Start` returns error, `Wait` will hang
// forever unless we cancel the context.
ctx, cancel := context.WithCancel(ctx)
defer cancel()
spec, err := container.Spec(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get container spec")
}
task, err := container.Task(ctx, nil)
if err != nil {
return nil, errors.Wrap(err, "failed to load task")
}
pspec := spec.Process
pspec.Terminal = opts.tty
if opts.tty {
if err := oci.WithEnv([]string{"TERM=xterm"})(ctx, nil, nil, spec); err != nil {
return nil, errors.Wrap(err, "add TERM env var to spec")
}
}
pspec.Args = opts.cmd
if opts.stdout == nil {
opts.stdout = cio.NewDiscardLogger()
}
if opts.stderr == nil {
opts.stderr = cio.NewDiscardLogger()
}
execID := util.GenerateID()
log.G(ctx).Debugf("Generated exec id %q for container %q", execID, id)
volatileRootDir := c.getVolatileContainerRootDir(id)
var execIO *cio.ExecIO
process, err := task.Exec(ctx, execID, pspec,
func(id string) (containerdio.IO, error) {
var err error
execIO, err = cio.NewExecIO(id, volatileRootDir, opts.tty, opts.stdin != nil)
return execIO, err
},
)
if err != nil {
return nil, errors.Wrapf(err, "failed to create exec %q", execID)
}
defer func() {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
if _, err := process.Delete(deferCtx, containerd.WithProcessKill); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to delete exec process %q for container %q", execID, id)
}
}()
exitCh, err := process.Wait(ctx)
if err != nil {
return nil, errors.Wrapf(err, "failed to wait for process %q", execID)
}
if err := process.Start(ctx); err != nil {
return nil, errors.Wrapf(err, "failed to start exec %q", execID)
}
handleResizing(ctx, opts.resize, func(size remotecommand.TerminalSize) {
if err := process.Resize(ctx, uint32(size.Width), uint32(size.Height)); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to resize process %q console for container %q", execID, id)
}
})
attachDone := execIO.Attach(cio.AttachOptions{
Stdin: opts.stdin,
Stdout: opts.stdout,
Stderr: opts.stderr,
Tty: opts.tty,
StdinOnce: true,
CloseStdin: func() error {
return process.CloseIO(ctx, containerd.WithStdinCloser)
},
})
execCtx := ctx
if opts.timeout > 0 {
var execCtxCancel context.CancelFunc
execCtx, execCtxCancel = context.WithTimeout(ctx, opts.timeout)
defer execCtxCancel()
}
select {
case <-execCtx.Done():
// Ignore the not found error because the process may exit itself before killing.
if err := process.Kill(ctx, syscall.SIGKILL); err != nil && !errdefs.IsNotFound(err) {
return nil, errors.Wrapf(err, "failed to kill exec %q", execID)
}
// Wait for the process to be killed.
exitRes := <-exitCh
log.G(ctx).Infof("Timeout received while waiting for exec process kill %q code %d and error %v",
execID, exitRes.ExitCode(), exitRes.Error())
<-attachDone
log.G(ctx).Debugf("Stream pipe for exec process %q done", execID)
return nil, errors.Wrapf(execCtx.Err(), "timeout %v exceeded", opts.timeout)
case exitRes := <-exitCh:
code, _, err := exitRes.Result()
log.G(ctx).Infof("Exec process %q exits with exit code %d and error %v", execID, code, err)
if err != nil {
return nil, errors.Wrapf(err, "failed while waiting for exec %q", execID)
}
<-attachDone
log.G(ctx).Debugf("Stream pipe for exec process %q done", execID)
return &code, nil
}
}
// execInContainer executes a command inside the container synchronously, and
// redirects stdio stream properly.
// This function only returns when the exec process exits, this means that:
// 1) As long as the exec process is running, the goroutine in the cri plugin
// will be running and wait for the exit code;
// 2) `kubectl exec -it` will hang until the exec process exits, even after io
// is detached. This is different from dockershim, which leaves the exec process
// running in background after io is detached.
// https://github.com/kubernetes/kubernetes/blob/v1.15.0/pkg/kubelet/dockershim/exec.go#L127
// For example, if the `kubectl exec -it` process is killed, IO will be closed. In
// this case, the CRI plugin will still have a goroutine waiting for the exec process
// to exit and log the exit code, but dockershim won't.
func (c *criService) execInContainer(ctx context.Context, id string, opts execOptions) (*uint32, error) {
// Get container from our container store.
cntr, err := c.containerStore.Get(id)
if err != nil {
return nil, errors.Wrapf(err, "failed to find container %q in store", id)
}
id = cntr.ID
state := cntr.Status.Get().State()
if state != runtime.ContainerState_CONTAINER_RUNNING {
return nil, errors.Errorf("container is in %s state", criContainerStateToString(state))
}
return c.execInternal(ctx, cntr.Container, id, opts)
}

View File

@@ -1,112 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
containerstore "github.com/containerd/cri/pkg/store/container"
)
// ListContainers lists all containers matching the filter.
func (c *criService) ListContainers(ctx context.Context, r *runtime.ListContainersRequest) (*runtime.ListContainersResponse, error) {
// List all containers from store.
containersInStore := c.containerStore.List()
var containers []*runtime.Container
for _, container := range containersInStore {
containers = append(containers, toCRIContainer(container))
}
containers = c.filterCRIContainers(containers, r.GetFilter())
return &runtime.ListContainersResponse{Containers: containers}, nil
}
// toCRIContainer converts internal container object into CRI container.
func toCRIContainer(container containerstore.Container) *runtime.Container {
status := container.Status.Get()
return &runtime.Container{
Id: container.ID,
PodSandboxId: container.SandboxID,
Metadata: container.Config.GetMetadata(),
Image: container.Config.GetImage(),
ImageRef: container.ImageRef,
State: status.State(),
CreatedAt: status.CreatedAt,
Labels: container.Config.GetLabels(),
Annotations: container.Config.GetAnnotations(),
}
}
func (c *criService) normalizeContainerFilter(filter *runtime.ContainerFilter) {
if cntr, err := c.containerStore.Get(filter.GetId()); err == nil {
filter.Id = cntr.ID
}
if sb, err := c.sandboxStore.Get(filter.GetPodSandboxId()); err == nil {
filter.PodSandboxId = sb.ID
}
}
// filterCRIContainers filters CRIContainers.
func (c *criService) filterCRIContainers(containers []*runtime.Container, filter *runtime.ContainerFilter) []*runtime.Container {
if filter == nil {
return containers
}
// The containerd cri plugin supports short ids so long as there is only one
// match. So we do a lookup against the store here if a pod id has been
// included in the filter.
sb := filter.GetPodSandboxId()
if sb != "" {
sandbox, err := c.sandboxStore.Get(sb)
if err == nil {
sb = sandbox.ID
}
}
c.normalizeContainerFilter(filter)
filtered := []*runtime.Container{}
for _, cntr := range containers {
if filter.GetId() != "" && filter.GetId() != cntr.Id {
continue
}
if sb != "" && sb != cntr.PodSandboxId {
continue
}
if filter.GetState() != nil && filter.GetState().GetState() != cntr.State {
continue
}
if filter.GetLabelSelector() != nil {
match := true
for k, v := range filter.GetLabelSelector() {
got, ok := cntr.Labels[k]
if !ok || got != v {
match = false
break
}
}
if !match {
continue
}
}
filtered = append(filtered, cntr)
}
return filtered
}

View File

@@ -1,51 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// ReopenContainerLog asks the cri plugin to reopen the stdout/stderr log file for the container.
// This is often called after the log file has been rotated.
func (c *criService) ReopenContainerLog(ctx context.Context, r *runtime.ReopenContainerLogRequest) (*runtime.ReopenContainerLogResponse, error) {
container, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
}
if container.Status.Get().State() != runtime.ContainerState_CONTAINER_RUNNING {
return nil, errors.New("container is not running")
}
// Create new container logger and replace the existing ones.
stdoutWC, stderrWC, err := c.createContainerLoggers(container.LogPath, container.Config.GetTty())
if err != nil {
return nil, err
}
oldStdoutWC, oldStderrWC := container.IO.AddOutput("log", stdoutWC, stderrWC)
if oldStdoutWC != nil {
oldStdoutWC.Close()
}
if oldStderrWC != nil {
oldStderrWC.Close()
}
return &runtime.ReopenContainerLogResponse{}, nil
}

View File

@@ -1,135 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store"
containerstore "github.com/containerd/cri/pkg/store/container"
)
// RemoveContainer removes the container.
func (c *criService) RemoveContainer(ctx context.Context, r *runtime.RemoveContainerRequest) (_ *runtime.RemoveContainerResponse, retErr error) {
container, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
if err != store.ErrNotExist {
return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
}
// Do not return error if container metadata doesn't exist.
log.G(ctx).Tracef("RemoveContainer called for container %q that does not exist", r.GetContainerId())
return &runtime.RemoveContainerResponse{}, nil
}
id := container.ID
// Forcibly stop the containers if they are in running or unknown state
state := container.Status.Get().State()
if state == runtime.ContainerState_CONTAINER_RUNNING ||
state == runtime.ContainerState_CONTAINER_UNKNOWN {
logrus.Infof("Forcibly stopping container %q", id)
if err := c.stopContainer(ctx, container, 0); err != nil {
return nil, errors.Wrapf(err, "failed to forcibly stop container %q", id)
}
}
// Set removing state to prevent other start/remove operations against this container
// while it's being removed.
if err := setContainerRemoving(container); err != nil {
return nil, errors.Wrapf(err, "failed to set removing state for container %q", id)
}
defer func() {
if retErr != nil {
// Reset removing if remove failed.
if err := resetContainerRemoving(container); err != nil {
log.G(ctx).WithError(err).Errorf("failed to reset removing state for container %q", id)
}
}
}()
// NOTE(random-liu): Docker set container to "Dead" state when start removing the
// container so as to avoid start/restart the container again. However, for current
// kubelet implementation, we'll never start a container once we decide to remove it,
// so we don't need the "Dead" state for now.
// Delete containerd container.
if err := container.Container.Delete(ctx, containerd.WithSnapshotCleanup); err != nil {
if !errdefs.IsNotFound(err) {
return nil, errors.Wrapf(err, "failed to delete containerd container %q", id)
}
log.G(ctx).Tracef("Remove called for containerd container %q that does not exist", id)
}
// Delete container checkpoint.
if err := container.Delete(); err != nil {
return nil, errors.Wrapf(err, "failed to delete container checkpoint for %q", id)
}
containerRootDir := c.getContainerRootDir(id)
if err := ensureRemoveAll(ctx, containerRootDir); err != nil {
return nil, errors.Wrapf(err, "failed to remove container root directory %q",
containerRootDir)
}
volatileContainerRootDir := c.getVolatileContainerRootDir(id)
if err := ensureRemoveAll(ctx, volatileContainerRootDir); err != nil {
return nil, errors.Wrapf(err, "failed to remove volatile container root directory %q",
volatileContainerRootDir)
}
c.containerStore.Delete(id)
c.containerNameIndex.ReleaseByKey(id)
return &runtime.RemoveContainerResponse{}, nil
}
// setContainerRemoving sets the container into removing state. In removing state, the
// container will not be started or removed again.
func setContainerRemoving(container containerstore.Container) error {
return container.Status.Update(func(status containerstore.Status) (containerstore.Status, error) {
// Do not remove container if it's still running or unknown.
if status.State() == runtime.ContainerState_CONTAINER_RUNNING {
return status, errors.New("container is still running, to stop first")
}
if status.State() == runtime.ContainerState_CONTAINER_UNKNOWN {
return status, errors.New("container state is unknown, to stop first")
}
if status.Starting {
return status, errors.New("container is in starting state, can't be removed")
}
if status.Removing {
return status, errors.New("container is already in removing state")
}
status.Removing = true
return status, nil
})
}
// resetContainerRemoving resets the container removing state on remove failure. So
// that we could remove the container again.
func resetContainerRemoving(container containerstore.Container) error {
return container.Status.Update(func(status containerstore.Status) (containerstore.Status, error) {
status.Removing = false
return status, nil
})
}

View File

@@ -1,222 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"io"
"time"
"github.com/containerd/containerd"
containerdio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/containerd/nri"
v1 "github.com/containerd/nri/types/v1"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
cioutil "github.com/containerd/cri/pkg/ioutil"
cio "github.com/containerd/cri/pkg/server/io"
containerstore "github.com/containerd/cri/pkg/store/container"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// StartContainer starts the container.
func (c *criService) StartContainer(ctx context.Context, r *runtime.StartContainerRequest) (retRes *runtime.StartContainerResponse, retErr error) {
cntr, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
}
id := cntr.ID
meta := cntr.Metadata
container := cntr.Container
config := meta.Config
// Set starting state to prevent other start/remove operations against this container
// while it's being started.
if err := setContainerStarting(cntr); err != nil {
return nil, errors.Wrapf(err, "failed to set starting state for container %q", id)
}
defer func() {
if retErr != nil {
// Set container to exited if fail to start.
if err := cntr.Status.UpdateSync(func(status containerstore.Status) (containerstore.Status, error) {
status.Pid = 0
status.FinishedAt = time.Now().UnixNano()
status.ExitCode = errorStartExitCode
status.Reason = errorStartReason
status.Message = retErr.Error()
return status, nil
}); err != nil {
log.G(ctx).WithError(err).Errorf("failed to set start failure state for container %q", id)
}
}
if err := resetContainerStarting(cntr); err != nil {
log.G(ctx).WithError(err).Errorf("failed to reset starting state for container %q", id)
}
}()
// Get sandbox config from sandbox store.
sandbox, err := c.sandboxStore.Get(meta.SandboxID)
if err != nil {
return nil, errors.Wrapf(err, "sandbox %q not found", meta.SandboxID)
}
sandboxID := meta.SandboxID
if sandbox.Status.Get().State != sandboxstore.StateReady {
return nil, errors.Errorf("sandbox container %q is not running", sandboxID)
}
ioCreation := func(id string) (_ containerdio.IO, err error) {
stdoutWC, stderrWC, err := c.createContainerLoggers(meta.LogPath, config.GetTty())
if err != nil {
return nil, errors.Wrap(err, "failed to create container loggers")
}
cntr.IO.AddOutput("log", stdoutWC, stderrWC)
cntr.IO.Pipe()
return cntr.IO, nil
}
ctrInfo, err := container.Info(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get container info")
}
taskOpts := c.taskOpts(ctrInfo.Runtime.Name)
task, err := container.NewTask(ctx, ioCreation, taskOpts...)
if err != nil {
return nil, errors.Wrap(err, "failed to create containerd task")
}
defer func() {
if retErr != nil {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
// It's possible that task is deleted by event monitor.
if _, err := task.Delete(deferCtx, WithNRISandboxDelete(sandboxID), containerd.WithProcessKill); err != nil && !errdefs.IsNotFound(err) {
log.G(ctx).WithError(err).Errorf("Failed to delete containerd task %q", id)
}
}
}()
// wait is a long running background request, no timeout needed.
exitCh, err := task.Wait(ctrdutil.NamespacedContext())
if err != nil {
return nil, errors.Wrap(err, "failed to wait for containerd task")
}
nric, err := nri.New()
if err != nil {
log.G(ctx).WithError(err).Error("unable to create nri client")
}
if nric != nil {
nriSB := &nri.Sandbox{
ID: sandboxID,
}
if _, err := nric.InvokeWithSandbox(ctx, task, v1.Create, nriSB); err != nil {
return nil, errors.Wrap(err, "nri invoke")
}
}
// Start containerd task.
if err := task.Start(ctx); err != nil {
return nil, errors.Wrapf(err, "failed to start containerd task %q", id)
}
// Update container start timestamp.
if err := cntr.Status.UpdateSync(func(status containerstore.Status) (containerstore.Status, error) {
status.Pid = task.Pid()
status.StartedAt = time.Now().UnixNano()
return status, nil
}); err != nil {
return nil, errors.Wrapf(err, "failed to update container %q state", id)
}
// start the monitor after updating container state, this ensures that
// event monitor receives the TaskExit event and update container state
// after this.
c.eventMonitor.startExitMonitor(context.Background(), id, task.Pid(), exitCh)
return &runtime.StartContainerResponse{}, nil
}
// setContainerStarting sets the container into starting state. In starting state, the
// container will not be removed or started again.
func setContainerStarting(container containerstore.Container) error {
return container.Status.Update(func(status containerstore.Status) (containerstore.Status, error) {
// Return error if container is not in created state.
if status.State() != runtime.ContainerState_CONTAINER_CREATED {
return status, errors.Errorf("container is in %s state", criContainerStateToString(status.State()))
}
// Do not start the container when there is a removal in progress.
if status.Removing {
return status, errors.New("container is in removing state, can't be started")
}
if status.Starting {
return status, errors.New("container is already in starting state")
}
status.Starting = true
return status, nil
})
}
// resetContainerStarting resets the container starting state on start failure. So
// that we could remove the container later.
func resetContainerStarting(container containerstore.Container) error {
return container.Status.Update(func(status containerstore.Status) (containerstore.Status, error) {
status.Starting = false
return status, nil
})
}
// createContainerLoggers creates container loggers and return write closer for stdout and stderr.
func (c *criService) createContainerLoggers(logPath string, tty bool) (stdout io.WriteCloser, stderr io.WriteCloser, err error) {
if logPath != "" {
// Only generate container log when log path is specified.
f, err := openLogFile(logPath)
if err != nil {
return nil, nil, errors.Wrap(err, "failed to create and open log file")
}
defer func() {
if err != nil {
f.Close()
}
}()
var stdoutCh, stderrCh <-chan struct{}
wc := cioutil.NewSerialWriteCloser(f)
stdout, stdoutCh = cio.NewCRILogger(logPath, wc, cio.Stdout, c.config.MaxContainerLogLineSize)
// Only redirect stderr when there is no tty.
if !tty {
stderr, stderrCh = cio.NewCRILogger(logPath, wc, cio.Stderr, c.config.MaxContainerLogLineSize)
}
go func() {
if stdoutCh != nil {
<-stdoutCh
}
if stderrCh != nil {
<-stderrCh
}
logrus.Debugf("Finish redirecting log file %q, closing it", logPath)
f.Close()
}()
} else {
stdout = cio.NewDiscardLogger()
stderr = cio.NewDiscardLogger()
}
return
}

View File

@@ -1,47 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
tasks "github.com/containerd/containerd/api/services/tasks/v1"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// ContainerStats returns stats of the container. If the container does not
// exist, the call returns an error.
func (c *criService) ContainerStats(ctx context.Context, in *runtime.ContainerStatsRequest) (*runtime.ContainerStatsResponse, error) {
cntr, err := c.containerStore.Get(in.GetContainerId())
if err != nil {
return nil, errors.Wrap(err, "failed to find container")
}
request := &tasks.MetricsRequest{Filters: []string{"id==" + cntr.ID}}
resp, err := c.client.TaskService().Metrics(ctx, request)
if err != nil {
return nil, errors.Wrap(err, "failed to fetch metrics for task")
}
if len(resp.Metrics) != 1 {
return nil, errors.Errorf("unexpected metrics response: %+v", resp.Metrics)
}
cs, err := c.containerMetrics(cntr.Metadata, resp.Metrics[0])
if err != nil {
return nil, errors.Wrap(err, "failed to decode container metrics")
}
return &runtime.ContainerStatsResponse{Stats: cs}, nil
}

View File

@@ -1,116 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
tasks "github.com/containerd/containerd/api/services/tasks/v1"
"github.com/containerd/containerd/api/types"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
containerstore "github.com/containerd/cri/pkg/store/container"
)
// ListContainerStats returns stats of all running containers.
func (c *criService) ListContainerStats(
ctx context.Context,
in *runtime.ListContainerStatsRequest,
) (*runtime.ListContainerStatsResponse, error) {
request, containers, err := c.buildTaskMetricsRequest(in)
if err != nil {
return nil, errors.Wrap(err, "failed to build metrics request")
}
resp, err := c.client.TaskService().Metrics(ctx, &request)
if err != nil {
return nil, errors.Wrap(err, "failed to fetch metrics for tasks")
}
criStats, err := c.toCRIContainerStats(resp.Metrics, containers)
if err != nil {
return nil, errors.Wrap(err, "failed to convert to cri containerd stats format")
}
return criStats, nil
}
func (c *criService) toCRIContainerStats(
stats []*types.Metric,
containers []containerstore.Container,
) (*runtime.ListContainerStatsResponse, error) {
statsMap := make(map[string]*types.Metric)
for _, stat := range stats {
statsMap[stat.ID] = stat
}
containerStats := new(runtime.ListContainerStatsResponse)
for _, cntr := range containers {
cs, err := c.containerMetrics(cntr.Metadata, statsMap[cntr.ID])
if err != nil {
return nil, errors.Wrapf(err, "failed to decode container metrics for %q", cntr.ID)
}
containerStats.Stats = append(containerStats.Stats, cs)
}
return containerStats, nil
}
func (c *criService) normalizeContainerStatsFilter(filter *runtime.ContainerStatsFilter) {
if cntr, err := c.containerStore.Get(filter.GetId()); err == nil {
filter.Id = cntr.ID
}
if sb, err := c.sandboxStore.Get(filter.GetPodSandboxId()); err == nil {
filter.PodSandboxId = sb.ID
}
}
// buildTaskMetricsRequest constructs a tasks.MetricsRequest based on
// the information in the stats request and the containerStore
func (c *criService) buildTaskMetricsRequest(
r *runtime.ListContainerStatsRequest,
) (tasks.MetricsRequest, []containerstore.Container, error) {
var req tasks.MetricsRequest
if r.GetFilter() == nil {
return req, nil, nil
}
c.normalizeContainerStatsFilter(r.GetFilter())
var containers []containerstore.Container
for _, cntr := range c.containerStore.List() {
if r.GetFilter().GetId() != "" && cntr.ID != r.GetFilter().GetId() {
continue
}
if r.GetFilter().GetPodSandboxId() != "" && cntr.SandboxID != r.GetFilter().GetPodSandboxId() {
continue
}
if r.GetFilter().GetLabelSelector() != nil &&
!matchLabelSelector(r.GetFilter().GetLabelSelector(), cntr.Config.GetLabels()) {
continue
}
containers = append(containers, cntr)
req.Filters = append(req.Filters, "id=="+cntr.ID)
}
return req, containers, nil
}
func matchLabelSelector(selector, labels map[string]string) bool {
for k, v := range selector {
if val, ok := labels[k]; ok {
if v != val {
return false
}
} else {
return false
}
}
return true
}

View File

@@ -1,129 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"fmt"
"github.com/containerd/containerd/api/types"
v1 "github.com/containerd/containerd/metrics/types/v1"
v2 "github.com/containerd/containerd/metrics/types/v2"
"github.com/containerd/typeurl"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
containerstore "github.com/containerd/cri/pkg/store/container"
)
func (c *criService) containerMetrics(
meta containerstore.Metadata,
stats *types.Metric,
) (*runtime.ContainerStats, error) {
var cs runtime.ContainerStats
var usedBytes, inodesUsed uint64
sn, err := c.snapshotStore.Get(meta.ID)
// If snapshotstore doesn't have cached snapshot information
// set WritableLayer usage to zero
if err == nil {
usedBytes = sn.Size
inodesUsed = sn.Inodes
}
cs.WritableLayer = &runtime.FilesystemUsage{
Timestamp: sn.Timestamp,
FsId: &runtime.FilesystemIdentifier{
Mountpoint: c.imageFSPath,
},
UsedBytes: &runtime.UInt64Value{Value: usedBytes},
InodesUsed: &runtime.UInt64Value{Value: inodesUsed},
}
cs.Attributes = &runtime.ContainerAttributes{
Id: meta.ID,
Metadata: meta.Config.GetMetadata(),
Labels: meta.Config.GetLabels(),
Annotations: meta.Config.GetAnnotations(),
}
if stats != nil {
s, err := typeurl.UnmarshalAny(stats.Data)
if err != nil {
return nil, errors.Wrap(err, "failed to extract container metrics")
}
switch metrics := s.(type) {
case *v1.Metrics:
if metrics.CPU != nil && metrics.CPU.Usage != nil {
cs.Cpu = &runtime.CpuUsage{
Timestamp: stats.Timestamp.UnixNano(),
UsageCoreNanoSeconds: &runtime.UInt64Value{Value: metrics.CPU.Usage.Total},
}
}
if metrics.Memory != nil && metrics.Memory.Usage != nil {
cs.Memory = &runtime.MemoryUsage{
Timestamp: stats.Timestamp.UnixNano(),
WorkingSetBytes: &runtime.UInt64Value{
Value: getWorkingSet(metrics.Memory),
},
}
}
case *v2.Metrics:
if metrics.CPU != nil {
cs.Cpu = &runtime.CpuUsage{
Timestamp: stats.Timestamp.UnixNano(),
UsageCoreNanoSeconds: &runtime.UInt64Value{Value: metrics.CPU.UsageUsec * 1000},
}
}
if metrics.Memory != nil {
cs.Memory = &runtime.MemoryUsage{
Timestamp: stats.Timestamp.UnixNano(),
WorkingSetBytes: &runtime.UInt64Value{
Value: getWorkingSetV2(metrics.Memory),
},
}
}
default:
return &cs, errors.New(fmt.Sprintf("unxpected metrics type: %v", metrics))
}
}
return &cs, nil
}
// getWorkingSet calculates workingset memory from cgroup memory stats.
// The caller should make sure memory is not nil.
// workingset = usage - total_inactive_file
func getWorkingSet(memory *v1.MemoryStat) uint64 {
if memory.Usage == nil {
return 0
}
var workingSet uint64
if memory.TotalInactiveFile < memory.Usage.Usage {
workingSet = memory.Usage.Usage - memory.TotalInactiveFile
}
return workingSet
}
// getWorkingSetV2 calculates workingset memory from cgroupv2 memory stats.
// The caller should make sure memory is not nil.
// workingset = usage - inactive_file
func getWorkingSetV2(memory *v2.MemoryStat) uint64 {
var workingSet uint64
if memory.InactiveFile < memory.Usage {
workingSet = memory.Usage - memory.InactiveFile
}
return workingSet
}

View File

@@ -1,84 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
wstats "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/stats"
"github.com/containerd/containerd/api/types"
"github.com/containerd/typeurl"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
containerstore "github.com/containerd/cri/pkg/store/container"
)
func (c *criService) containerMetrics(
meta containerstore.Metadata,
stats *types.Metric,
) (*runtime.ContainerStats, error) {
var cs runtime.ContainerStats
var usedBytes, inodesUsed uint64
sn, err := c.snapshotStore.Get(meta.ID)
// If snapshotstore doesn't have cached snapshot information
// set WritableLayer usage to zero
if err == nil {
usedBytes = sn.Size
inodesUsed = sn.Inodes
}
cs.WritableLayer = &runtime.FilesystemUsage{
Timestamp: sn.Timestamp,
FsId: &runtime.FilesystemIdentifier{
Mountpoint: c.imageFSPath,
},
UsedBytes: &runtime.UInt64Value{Value: usedBytes},
InodesUsed: &runtime.UInt64Value{Value: inodesUsed},
}
cs.Attributes = &runtime.ContainerAttributes{
Id: meta.ID,
Metadata: meta.Config.GetMetadata(),
Labels: meta.Config.GetLabels(),
Annotations: meta.Config.GetAnnotations(),
}
if stats != nil {
s, err := typeurl.UnmarshalAny(stats.Data)
if err != nil {
return nil, errors.Wrap(err, "failed to extract container metrics")
}
wstats := s.(*wstats.Statistics).GetWindows()
if wstats == nil {
return nil, errors.New("windows stats is empty")
}
if wstats.Processor != nil {
cs.Cpu = &runtime.CpuUsage{
Timestamp: wstats.Timestamp.UnixNano(),
UsageCoreNanoSeconds: &runtime.UInt64Value{Value: wstats.Processor.TotalRuntimeNS},
}
}
if wstats.Memory != nil {
cs.Memory = &runtime.MemoryUsage{
Timestamp: wstats.Timestamp.UnixNano(),
WorkingSetBytes: &runtime.UInt64Value{
Value: wstats.Memory.MemoryUsagePrivateWorkingSetBytes,
},
}
}
}
return &cs, nil
}

View File

@@ -1,173 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store"
containerstore "github.com/containerd/cri/pkg/store/container"
)
// ContainerStatus inspects the container and returns the status.
func (c *criService) ContainerStatus(ctx context.Context, r *runtime.ContainerStatusRequest) (*runtime.ContainerStatusResponse, error) {
container, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
}
// TODO(random-liu): Clean up the following logic in CRI.
// Current assumption:
// * ImageSpec in container config is image ID.
// * ImageSpec in container status is image tag.
// * ImageRef in container status is repo digest.
spec := container.Config.GetImage()
imageRef := container.ImageRef
image, err := c.imageStore.Get(imageRef)
if err != nil {
if err != store.ErrNotExist {
return nil, errors.Wrapf(err, "failed to get image %q", imageRef)
}
} else {
repoTags, repoDigests := parseImageReferences(image.References)
if len(repoTags) > 0 {
// Based on current behavior of dockershim, this field should be
// image tag.
spec = &runtime.ImageSpec{Image: repoTags[0]}
}
if len(repoDigests) > 0 {
// Based on the CRI definition, this field will be consumed by user.
imageRef = repoDigests[0]
}
}
status := toCRIContainerStatus(container, spec, imageRef)
if status.GetCreatedAt() == 0 {
// CRI doesn't allow CreatedAt == 0.
info, err := container.Container.Info(ctx)
if err != nil {
return nil, errors.Wrapf(err, "failed to get CreatedAt in %q state", status.State)
}
status.CreatedAt = info.CreatedAt.UnixNano()
}
info, err := toCRIContainerInfo(ctx, container, r.GetVerbose())
if err != nil {
return nil, errors.Wrap(err, "failed to get verbose container info")
}
return &runtime.ContainerStatusResponse{
Status: status,
Info: info,
}, nil
}
// toCRIContainerStatus converts internal container object to CRI container status.
func toCRIContainerStatus(container containerstore.Container, spec *runtime.ImageSpec, imageRef string) *runtime.ContainerStatus {
meta := container.Metadata
status := container.Status.Get()
reason := status.Reason
if status.State() == runtime.ContainerState_CONTAINER_EXITED && reason == "" {
if status.ExitCode == 0 {
reason = completeExitReason
} else {
reason = errorExitReason
}
}
return &runtime.ContainerStatus{
Id: meta.ID,
Metadata: meta.Config.GetMetadata(),
State: status.State(),
CreatedAt: status.CreatedAt,
StartedAt: status.StartedAt,
FinishedAt: status.FinishedAt,
ExitCode: status.ExitCode,
Image: spec,
ImageRef: imageRef,
Reason: reason,
Message: status.Message,
Labels: meta.Config.GetLabels(),
Annotations: meta.Config.GetAnnotations(),
Mounts: meta.Config.GetMounts(),
LogPath: meta.LogPath,
}
}
// ContainerInfo is extra information for a container.
type ContainerInfo struct {
// TODO(random-liu): Add sandboxID in CRI container status.
SandboxID string `json:"sandboxID"`
Pid uint32 `json:"pid"`
Removing bool `json:"removing"`
SnapshotKey string `json:"snapshotKey"`
Snapshotter string `json:"snapshotter"`
RuntimeType string `json:"runtimeType"`
RuntimeOptions interface{} `json:"runtimeOptions"`
Config *runtime.ContainerConfig `json:"config"`
RuntimeSpec *runtimespec.Spec `json:"runtimeSpec"`
}
// toCRIContainerInfo converts internal container object information to CRI container status response info map.
func toCRIContainerInfo(ctx context.Context, container containerstore.Container, verbose bool) (map[string]string, error) {
if !verbose {
return nil, nil
}
meta := container.Metadata
status := container.Status.Get()
// TODO(random-liu): Change CRI status info to use array instead of map.
ci := &ContainerInfo{
SandboxID: container.SandboxID,
Pid: status.Pid,
Removing: status.Removing,
Config: meta.Config,
}
var err error
ci.RuntimeSpec, err = container.Container.Spec(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get container runtime spec")
}
ctrInfo, err := container.Container.Info(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get container info")
}
ci.SnapshotKey = ctrInfo.SnapshotKey
ci.Snapshotter = ctrInfo.Snapshotter
runtimeOptions, err := getRuntimeOptions(ctrInfo)
if err != nil {
return nil, errors.Wrap(err, "failed to get runtime options")
}
ci.RuntimeType = ctrInfo.Runtime.Name
ci.RuntimeOptions = runtimeOptions
infoBytes, err := json.Marshal(ci)
if err != nil {
return nil, errors.Wrapf(err, "failed to marshal info %v", ci)
}
return map[string]string{
"info": string(infoBytes),
}, nil
}

View File

@@ -1,186 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"syscall"
"time"
"github.com/containerd/containerd"
eventtypes "github.com/containerd/containerd/api/events"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
"github.com/containerd/cri/pkg/store"
containerstore "github.com/containerd/cri/pkg/store/container"
)
// StopContainer stops a running container with a grace period (i.e., timeout).
func (c *criService) StopContainer(ctx context.Context, r *runtime.StopContainerRequest) (*runtime.StopContainerResponse, error) {
// Get container config from container store.
container, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrapf(err, "an error occurred when try to find container %q", r.GetContainerId())
}
if err := c.stopContainer(ctx, container, time.Duration(r.GetTimeout())*time.Second); err != nil {
return nil, err
}
return &runtime.StopContainerResponse{}, nil
}
// stopContainer stops a container based on the container metadata.
func (c *criService) stopContainer(ctx context.Context, container containerstore.Container, timeout time.Duration) error {
id := container.ID
// Return without error if container is not running. This makes sure that
// stop only takes real action after the container is started.
state := container.Status.Get().State()
if state != runtime.ContainerState_CONTAINER_RUNNING &&
state != runtime.ContainerState_CONTAINER_UNKNOWN {
log.G(ctx).Infof("Container to stop %q must be in running or unknown state, current state %q",
id, criContainerStateToString(state))
return nil
}
task, err := container.Container.Task(ctx, nil)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrapf(err, "failed to get task for container %q", id)
}
// Don't return for unknown state, some cleanup needs to be done.
if state == runtime.ContainerState_CONTAINER_UNKNOWN {
return cleanupUnknownContainer(ctx, id, container)
}
return nil
}
// Handle unknown state.
if state == runtime.ContainerState_CONTAINER_UNKNOWN {
// Start an exit handler for containers in unknown state.
waitCtx, waitCancel := context.WithCancel(ctrdutil.NamespacedContext())
defer waitCancel()
exitCh, err := task.Wait(waitCtx)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrapf(err, "failed to wait for task for %q", id)
}
return cleanupUnknownContainer(ctx, id, container)
}
exitCtx, exitCancel := context.WithCancel(context.Background())
stopCh := c.eventMonitor.startExitMonitor(exitCtx, id, task.Pid(), exitCh)
defer func() {
exitCancel()
// This ensures that exit monitor is stopped before
// `Wait` is cancelled, so no exit event is generated
// because of the `Wait` cancellation.
<-stopCh
}()
}
// We only need to kill the task. The event handler will Delete the
// task from containerd after it handles the Exited event.
if timeout > 0 {
stopSignal := "SIGTERM"
if container.StopSignal != "" {
stopSignal = container.StopSignal
} else {
// The image may have been deleted, and the `StopSignal` field is
// just introduced to handle that.
// However, for containers created before the `StopSignal` field is
// introduced, still try to get the stop signal from the image config.
// If the image has been deleted, logging an error and using the
// default SIGTERM is still better than returning error and leaving
// the container unstoppable. (See issue #990)
// TODO(random-liu): Remove this logic when containerd 1.2 is deprecated.
image, err := c.imageStore.Get(container.ImageRef)
if err != nil {
if err != store.ErrNotExist {
return errors.Wrapf(err, "failed to get image %q", container.ImageRef)
}
log.G(ctx).Warningf("Image %q not found, stop container with signal %q", container.ImageRef, stopSignal)
} else {
if image.ImageSpec.Config.StopSignal != "" {
stopSignal = image.ImageSpec.Config.StopSignal
}
}
}
sig, err := containerd.ParseSignal(stopSignal)
if err != nil {
return errors.Wrapf(err, "failed to parse stop signal %q", stopSignal)
}
log.G(ctx).Infof("Stop container %q with signal %v", id, sig)
if err = task.Kill(ctx, sig); err != nil && !errdefs.IsNotFound(err) {
return errors.Wrapf(err, "failed to stop container %q", id)
}
sigTermCtx, sigTermCtxCancel := context.WithTimeout(ctx, timeout)
defer sigTermCtxCancel()
err = c.waitContainerStop(sigTermCtx, container)
if err == nil {
// Container stopped on first signal no need for SIGKILL
return nil
}
// If the parent context was cancelled or exceeded return immediately
if ctx.Err() != nil {
return ctx.Err()
}
// sigTermCtx was exceeded. Send SIGKILL
log.G(ctx).Debugf("Stop container %q with signal %v timed out", id, sig)
}
log.G(ctx).Infof("Kill container %q", id)
if err = task.Kill(ctx, syscall.SIGKILL); err != nil && !errdefs.IsNotFound(err) {
return errors.Wrapf(err, "failed to kill container %q", id)
}
// Wait for a fixed timeout until container stop is observed by event monitor.
err = c.waitContainerStop(ctx, container)
if err != nil {
return errors.Wrapf(err, "an error occurs during waiting for container %q to be killed", id)
}
return nil
}
// waitContainerStop waits for container to be stopped until context is
// cancelled or the context deadline is exceeded.
func (c *criService) waitContainerStop(ctx context.Context, container containerstore.Container) error {
select {
case <-ctx.Done():
return errors.Wrapf(ctx.Err(), "wait container %q", container.ID)
case <-container.Stopped():
return nil
}
}
// cleanupUnknownContainer cleanup stopped container in unknown state.
func cleanupUnknownContainer(ctx context.Context, id string, cntr containerstore.Container) error {
// Reuse handleContainerExit to do the cleanup.
return handleContainerExit(ctx, &eventtypes.TaskExit{
ContainerID: id,
ID: id,
Pid: 0,
ExitStatus: unknownExitCode,
ExitedAt: time.Now(),
}, cntr)
}

View File

@@ -1,150 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
gocontext "context"
"github.com/containerd/containerd"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/containerd/typeurl"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/containerd/opts"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
containerstore "github.com/containerd/cri/pkg/store/container"
"github.com/containerd/cri/pkg/util"
)
// UpdateContainerResources updates ContainerConfig of the container.
func (c *criService) UpdateContainerResources(ctx context.Context, r *runtime.UpdateContainerResourcesRequest) (retRes *runtime.UpdateContainerResourcesResponse, retErr error) {
container, err := c.containerStore.Get(r.GetContainerId())
if err != nil {
return nil, errors.Wrap(err, "failed to find container")
}
// Update resources in status update transaction, so that:
// 1) There won't be race condition with container start.
// 2) There won't be concurrent resource update to the same container.
if err := container.Status.Update(func(status containerstore.Status) (containerstore.Status, error) {
return status, c.updateContainerResources(ctx, container, r.GetLinux(), status)
}); err != nil {
return nil, errors.Wrap(err, "failed to update resources")
}
return &runtime.UpdateContainerResourcesResponse{}, nil
}
func (c *criService) updateContainerResources(ctx context.Context,
cntr containerstore.Container,
resources *runtime.LinuxContainerResources,
status containerstore.Status) (retErr error) {
id := cntr.ID
// Do not update the container when there is a removal in progress.
if status.Removing {
return errors.Errorf("container %q is in removing state", id)
}
// Update container spec. If the container is not started yet, updating
// spec makes sure that the resource limits are correct when start;
// if the container is already started, updating spec is still required,
// the spec will become our source of truth for resource limits.
oldSpec, err := cntr.Container.Spec(ctx)
if err != nil {
return errors.Wrap(err, "failed to get container spec")
}
newSpec, err := updateOCILinuxResource(ctx, oldSpec, resources,
c.config.TolerateMissingHugetlbController, c.config.DisableHugetlbController)
if err != nil {
return errors.Wrap(err, "failed to update resource in spec")
}
if err := updateContainerSpec(ctx, cntr.Container, newSpec); err != nil {
return err
}
defer func() {
if retErr != nil {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
// Reset spec on error.
if err := updateContainerSpec(deferCtx, cntr.Container, oldSpec); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to update spec %+v for container %q", oldSpec, id)
}
}
}()
// If container is not running, only update spec is enough, new resource
// limit will be applied when container start.
if status.State() != runtime.ContainerState_CONTAINER_RUNNING {
return nil
}
task, err := cntr.Container.Task(ctx, nil)
if err != nil {
if errdefs.IsNotFound(err) {
// Task exited already.
return nil
}
return errors.Wrap(err, "failed to get task")
}
// newSpec.Linux won't be nil
if err := task.Update(ctx, containerd.WithResources(newSpec.Linux.Resources)); err != nil {
if errdefs.IsNotFound(err) {
// Task exited already.
return nil
}
return errors.Wrap(err, "failed to update resources")
}
return nil
}
// updateContainerSpec updates container spec.
func updateContainerSpec(ctx context.Context, cntr containerd.Container, spec *runtimespec.Spec) error {
any, err := typeurl.MarshalAny(spec)
if err != nil {
return errors.Wrapf(err, "failed to marshal spec %+v", spec)
}
if err := cntr.Update(ctx, func(ctx gocontext.Context, client *containerd.Client, c *containers.Container) error {
c.Spec = any
return nil
}); err != nil {
return errors.Wrap(err, "failed to update container spec")
}
return nil
}
// updateOCILinuxResource updates container resource limit.
func updateOCILinuxResource(ctx context.Context, spec *runtimespec.Spec, new *runtime.LinuxContainerResources,
tolerateMissingHugetlbController, disableHugetlbController bool) (*runtimespec.Spec, error) {
// Copy to make sure old spec is not changed.
var cloned runtimespec.Spec
if err := util.DeepCopy(&cloned, spec); err != nil {
return nil, errors.Wrap(err, "failed to deep copy")
}
if cloned.Linux == nil {
cloned.Linux = &runtimespec.Linux{}
}
if err := opts.WithResources(new, tolerateMissingHugetlbController, disableHugetlbController)(ctx, nil, nil, &cloned); err != nil {
return nil, errors.Wrap(err, "unable to set linux container resources")
}
return &cloned, nil
}

View File

@@ -1,31 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd/errdefs"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// UpdateContainerResources updates ContainerConfig of the container.
// TODO(windows): Figure out whether windows support this.
func (c *criService) UpdateContainerResources(ctx context.Context, r *runtime.UpdateContainerResourcesRequest) (*runtime.UpdateContainerResourcesResponse, error) {
return nil, errdefs.ErrNotImplemented
}

View File

@@ -1,461 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"sync"
"time"
"github.com/containerd/containerd"
eventtypes "github.com/containerd/containerd/api/events"
containerdio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/events"
"github.com/containerd/typeurl"
gogotypes "github.com/gogo/protobuf/types"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/net/context"
"k8s.io/apimachinery/pkg/util/clock"
"github.com/containerd/cri/pkg/constants"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
"github.com/containerd/cri/pkg/store"
containerstore "github.com/containerd/cri/pkg/store/container"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
const (
backOffInitDuration = 1 * time.Second
backOffMaxDuration = 5 * time.Minute
backOffExpireCheckDuration = 1 * time.Second
// handleEventTimeout is the timeout for handling 1 event. Event monitor
// handles events in serial, if one event blocks the event monitor, no
// other events can be handled.
// Add a timeout for each event handling, events that timeout will be requeued and
// handled again in the future.
handleEventTimeout = 10 * time.Second
exitChannelSize = 1024
)
// eventMonitor monitors containerd event and updates internal state correspondingly.
// TODO(random-liu): Handle event for each container in a separate goroutine.
type eventMonitor struct {
c *criService
ch <-chan *events.Envelope
// exitCh receives container/sandbox exit events from exit monitors.
exitCh chan *eventtypes.TaskExit
errCh <-chan error
ctx context.Context
cancel context.CancelFunc
backOff *backOff
}
type backOff struct {
queuePool map[string]*backOffQueue
// tickerMu is mutex used to protect the ticker.
tickerMu sync.Mutex
ticker *time.Ticker
minDuration time.Duration
maxDuration time.Duration
checkDuration time.Duration
clock clock.Clock
}
type backOffQueue struct {
events []interface{}
expireTime time.Time
duration time.Duration
clock clock.Clock
}
// Create new event monitor. New event monitor will start subscribing containerd event. All events
// happen after it should be monitored.
func newEventMonitor(c *criService) *eventMonitor {
ctx, cancel := context.WithCancel(context.Background())
return &eventMonitor{
c: c,
ctx: ctx,
cancel: cancel,
exitCh: make(chan *eventtypes.TaskExit, exitChannelSize),
backOff: newBackOff(),
}
}
// subscribe starts to subscribe containerd events.
func (em *eventMonitor) subscribe(subscriber events.Subscriber) {
// note: filters are any match, if you want any match but not in namespace foo
// then you have to manually filter namespace foo
filters := []string{
`topic=="/tasks/oom"`,
`topic~="/images/"`,
}
em.ch, em.errCh = subscriber.Subscribe(em.ctx, filters...)
}
// startExitMonitor starts an exit monitor for a given container/sandbox.
func (em *eventMonitor) startExitMonitor(ctx context.Context, id string, pid uint32, exitCh <-chan containerd.ExitStatus) <-chan struct{} {
stopCh := make(chan struct{})
go func() {
defer close(stopCh)
select {
case exitRes := <-exitCh:
exitStatus, exitedAt, err := exitRes.Result()
if err != nil {
logrus.WithError(err).Errorf("Failed to get task exit status for %q", id)
exitStatus = unknownExitCode
exitedAt = time.Now()
}
em.exitCh <- &eventtypes.TaskExit{
ContainerID: id,
ID: id,
Pid: pid,
ExitStatus: exitStatus,
ExitedAt: exitedAt,
}
case <-ctx.Done():
}
}()
return stopCh
}
func convertEvent(e *gogotypes.Any) (string, interface{}, error) {
id := ""
evt, err := typeurl.UnmarshalAny(e)
if err != nil {
return "", nil, errors.Wrap(err, "failed to unmarshalany")
}
switch e := evt.(type) {
case *eventtypes.TaskOOM:
id = e.ContainerID
case *eventtypes.ImageCreate:
id = e.Name
case *eventtypes.ImageUpdate:
id = e.Name
case *eventtypes.ImageDelete:
id = e.Name
default:
return "", nil, errors.New("unsupported event")
}
return id, evt, nil
}
// start starts the event monitor which monitors and handles all subscribed events. It returns
// an error channel for the caller to wait for stop errors from the event monitor.
// start must be called after subscribe.
func (em *eventMonitor) start() <-chan error {
errCh := make(chan error)
if em.ch == nil || em.errCh == nil {
panic("event channel is nil")
}
backOffCheckCh := em.backOff.start()
go func() {
defer close(errCh)
for {
select {
case e := <-em.exitCh:
logrus.Debugf("Received exit event %+v", e)
id := e.ID
if em.backOff.isInBackOff(id) {
logrus.Infof("Events for %q is in backoff, enqueue event %+v", id, e)
em.backOff.enBackOff(id, e)
break
}
if err := em.handleEvent(e); err != nil {
logrus.WithError(err).Errorf("Failed to handle exit event %+v for %s", e, id)
em.backOff.enBackOff(id, e)
}
case e := <-em.ch:
logrus.Debugf("Received containerd event timestamp - %v, namespace - %q, topic - %q", e.Timestamp, e.Namespace, e.Topic)
if e.Namespace != constants.K8sContainerdNamespace {
logrus.Debugf("Ignoring events in namespace - %q", e.Namespace)
break
}
id, evt, err := convertEvent(e.Event)
if err != nil {
logrus.WithError(err).Errorf("Failed to convert event %+v", e)
break
}
if em.backOff.isInBackOff(id) {
logrus.Infof("Events for %q is in backoff, enqueue event %+v", id, evt)
em.backOff.enBackOff(id, evt)
break
}
if err := em.handleEvent(evt); err != nil {
logrus.WithError(err).Errorf("Failed to handle event %+v for %s", evt, id)
em.backOff.enBackOff(id, evt)
}
case err := <-em.errCh:
// Close errCh in defer directly if there is no error.
if err != nil {
logrus.WithError(err).Errorf("Failed to handle event stream")
errCh <- err
}
return
case <-backOffCheckCh:
ids := em.backOff.getExpiredIDs()
for _, id := range ids {
queue := em.backOff.deBackOff(id)
for i, any := range queue.events {
if err := em.handleEvent(any); err != nil {
logrus.WithError(err).Errorf("Failed to handle backOff event %+v for %s", any, id)
em.backOff.reBackOff(id, queue.events[i:], queue.duration)
break
}
}
}
}
}
}()
return errCh
}
// stop stops the event monitor. It will close the event channel.
// Once event monitor is stopped, it can't be started.
func (em *eventMonitor) stop() {
em.backOff.stop()
em.cancel()
}
// handleEvent handles a containerd event.
func (em *eventMonitor) handleEvent(any interface{}) error {
ctx := ctrdutil.NamespacedContext()
ctx, cancel := context.WithTimeout(ctx, handleEventTimeout)
defer cancel()
switch e := any.(type) {
case *eventtypes.TaskExit:
logrus.Infof("TaskExit event %+v", e)
// Use ID instead of ContainerID to rule out TaskExit event for exec.
cntr, err := em.c.containerStore.Get(e.ID)
if err == nil {
if err := handleContainerExit(ctx, e, cntr); err != nil {
return errors.Wrap(err, "failed to handle container TaskExit event")
}
return nil
} else if err != store.ErrNotExist {
return errors.Wrap(err, "can't find container for TaskExit event")
}
sb, err := em.c.sandboxStore.Get(e.ID)
if err == nil {
if err := handleSandboxExit(ctx, e, sb); err != nil {
return errors.Wrap(err, "failed to handle sandbox TaskExit event")
}
return nil
} else if err != store.ErrNotExist {
return errors.Wrap(err, "can't find sandbox for TaskExit event")
}
return nil
case *eventtypes.TaskOOM:
logrus.Infof("TaskOOM event %+v", e)
// For TaskOOM, we only care which container it belongs to.
cntr, err := em.c.containerStore.Get(e.ContainerID)
if err != nil {
if err != store.ErrNotExist {
return errors.Wrap(err, "can't find container for TaskOOM event")
}
return nil
}
err = cntr.Status.UpdateSync(func(status containerstore.Status) (containerstore.Status, error) {
status.Reason = oomExitReason
return status, nil
})
if err != nil {
return errors.Wrap(err, "failed to update container status for TaskOOM event")
}
case *eventtypes.ImageCreate:
logrus.Infof("ImageCreate event %+v", e)
return em.c.updateImage(ctx, e.Name)
case *eventtypes.ImageUpdate:
logrus.Infof("ImageUpdate event %+v", e)
return em.c.updateImage(ctx, e.Name)
case *eventtypes.ImageDelete:
logrus.Infof("ImageDelete event %+v", e)
return em.c.updateImage(ctx, e.Name)
}
return nil
}
// handleContainerExit handles TaskExit event for container.
func handleContainerExit(ctx context.Context, e *eventtypes.TaskExit, cntr containerstore.Container) error {
// Attach container IO so that `Delete` could cleanup the stream properly.
task, err := cntr.Container.Task(ctx,
func(*containerdio.FIFOSet) (containerdio.IO, error) {
// We can't directly return cntr.IO here, because
// even if cntr.IO is nil, the cio.IO interface
// is not.
// See https://tour.golang.org/methods/12:
// Note that an interface value that holds a nil
// concrete value is itself non-nil.
if cntr.IO != nil {
return cntr.IO, nil
}
return nil, nil
},
)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrapf(err, "failed to load task for container")
}
} else {
// TODO(random-liu): [P1] This may block the loop, we may want to spawn a worker
if _, err = task.Delete(ctx, WithNRISandboxDelete(cntr.SandboxID), containerd.WithProcessKill); err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to stop container")
}
// Move on to make sure container status is updated.
}
}
err = cntr.Status.UpdateSync(func(status containerstore.Status) (containerstore.Status, error) {
// If FinishedAt has been set (e.g. with start failure), keep as
// it is.
if status.FinishedAt != 0 {
return status, nil
}
status.Pid = 0
status.FinishedAt = e.ExitedAt.UnixNano()
status.ExitCode = int32(e.ExitStatus)
// Unknown state can only transit to EXITED state, so we need
// to handle unknown state here.
if status.Unknown {
logrus.Debugf("Container %q transited from UNKNOWN to EXITED", cntr.ID)
status.Unknown = false
}
return status, nil
})
if err != nil {
return errors.Wrap(err, "failed to update container state")
}
// Using channel to propagate the information of container stop
cntr.Stop()
return nil
}
// handleSandboxExit handles TaskExit event for sandbox.
func handleSandboxExit(ctx context.Context, e *eventtypes.TaskExit, sb sandboxstore.Sandbox) error {
// No stream attached to sandbox container.
task, err := sb.Container.Task(ctx, nil)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to load task for sandbox")
}
} else {
// TODO(random-liu): [P1] This may block the loop, we may want to spawn a worker
if _, err = task.Delete(ctx, WithNRISandboxDelete(sb.ID), containerd.WithProcessKill); err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to stop sandbox")
}
// Move on to make sure container status is updated.
}
}
err = sb.Status.Update(func(status sandboxstore.Status) (sandboxstore.Status, error) {
status.State = sandboxstore.StateNotReady
status.Pid = 0
return status, nil
})
if err != nil {
return errors.Wrap(err, "failed to update sandbox state")
}
// Using channel to propagate the information of sandbox stop
sb.Stop()
return nil
}
func newBackOff() *backOff {
return &backOff{
queuePool: map[string]*backOffQueue{},
minDuration: backOffInitDuration,
maxDuration: backOffMaxDuration,
checkDuration: backOffExpireCheckDuration,
clock: clock.RealClock{},
}
}
func (b *backOff) getExpiredIDs() []string {
var ids []string
for id, q := range b.queuePool {
if q.isExpire() {
ids = append(ids, id)
}
}
return ids
}
func (b *backOff) isInBackOff(key string) bool {
if _, ok := b.queuePool[key]; ok {
return true
}
return false
}
// enBackOff start to backOff and put event to the tail of queue
func (b *backOff) enBackOff(key string, evt interface{}) {
if queue, ok := b.queuePool[key]; ok {
queue.events = append(queue.events, evt)
return
}
b.queuePool[key] = newBackOffQueue([]interface{}{evt}, b.minDuration, b.clock)
}
// enBackOff get out the whole queue
func (b *backOff) deBackOff(key string) *backOffQueue {
queue := b.queuePool[key]
delete(b.queuePool, key)
return queue
}
// enBackOff start to backOff again and put events to the queue
func (b *backOff) reBackOff(key string, events []interface{}, oldDuration time.Duration) {
duration := 2 * oldDuration
if duration > b.maxDuration {
duration = b.maxDuration
}
b.queuePool[key] = newBackOffQueue(events, duration, b.clock)
}
func (b *backOff) start() <-chan time.Time {
b.tickerMu.Lock()
defer b.tickerMu.Unlock()
b.ticker = time.NewTicker(b.checkDuration)
return b.ticker.C
}
func (b *backOff) stop() {
b.tickerMu.Lock()
defer b.tickerMu.Unlock()
if b.ticker != nil {
b.ticker.Stop()
}
}
func newBackOffQueue(events []interface{}, init time.Duration, c clock.Clock) *backOffQueue {
return &backOffQueue{
events: events,
duration: init,
expireTime: c.Now().Add(init),
clock: c,
}
}
func (q *backOffQueue) isExpire() bool {
// return time.Now >= expireTime
return !q.clock.Now().Before(q.expireTime)
}

View File

@@ -1,390 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"fmt"
"path"
"path/filepath"
"strconv"
"strings"
"github.com/BurntSushi/toml"
runhcsoptions "github.com/Microsoft/hcsshim/cmd/containerd-shim-runhcs-v1/options"
"github.com/containerd/containerd"
"github.com/containerd/containerd/containers"
"github.com/containerd/containerd/plugin"
"github.com/containerd/containerd/reference/docker"
"github.com/containerd/containerd/runtime/linux/runctypes"
runcoptions "github.com/containerd/containerd/runtime/v2/runc/options"
"github.com/containerd/typeurl"
imagedigest "github.com/opencontainers/go-digest"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
runtimeoptions "github.com/containerd/cri/pkg/api/runtimeoptions/v1"
criconfig "github.com/containerd/cri/pkg/config"
"github.com/containerd/cri/pkg/store"
containerstore "github.com/containerd/cri/pkg/store/container"
imagestore "github.com/containerd/cri/pkg/store/image"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
const (
// errorStartReason is the exit reason when fails to start container.
errorStartReason = "StartError"
// errorStartExitCode is the exit code when fails to start container.
// 128 is the same with Docker's behavior.
// TODO(windows): Figure out what should be used for windows.
errorStartExitCode = 128
// completeExitReason is the exit reason when container exits with code 0.
completeExitReason = "Completed"
// errorExitReason is the exit reason when container exits with code non-zero.
errorExitReason = "Error"
// oomExitReason is the exit reason when process in container is oom killed.
oomExitReason = "OOMKilled"
// sandboxesDir contains all sandbox root. A sandbox root is the running
// directory of the sandbox, all files created for the sandbox will be
// placed under this directory.
sandboxesDir = "sandboxes"
// containersDir contains all container root.
containersDir = "containers"
// Delimiter used to construct container/sandbox names.
nameDelimiter = "_"
// criContainerdPrefix is common prefix for cri-containerd
criContainerdPrefix = "io.cri-containerd"
// containerKindLabel is a label key indicating container is sandbox container or application container
containerKindLabel = criContainerdPrefix + ".kind"
// containerKindSandbox is a label value indicating container is sandbox container
containerKindSandbox = "sandbox"
// containerKindContainer is a label value indicating container is application container
containerKindContainer = "container"
// imageLabelKey is the label key indicating the image is managed by cri plugin.
imageLabelKey = criContainerdPrefix + ".image"
// imageLabelValue is the label value indicating the image is managed by cri plugin.
imageLabelValue = "managed"
// sandboxMetadataExtension is an extension name that identify metadata of sandbox in CreateContainerRequest
sandboxMetadataExtension = criContainerdPrefix + ".sandbox.metadata"
// containerMetadataExtension is an extension name that identify metadata of container in CreateContainerRequest
containerMetadataExtension = criContainerdPrefix + ".container.metadata"
// defaultIfName is the default network interface for the pods
defaultIfName = "eth0"
// runtimeRunhcsV1 is the runtime type for runhcs.
runtimeRunhcsV1 = "io.containerd.runhcs.v1"
)
// makeSandboxName generates sandbox name from sandbox metadata. The name
// generated is unique as long as sandbox metadata is unique.
func makeSandboxName(s *runtime.PodSandboxMetadata) string {
return strings.Join([]string{
s.Name, // 0
s.Namespace, // 1
s.Uid, // 2
fmt.Sprintf("%d", s.Attempt), // 3
}, nameDelimiter)
}
// makeContainerName generates container name from sandbox and container metadata.
// The name generated is unique as long as the sandbox container combination is
// unique.
func makeContainerName(c *runtime.ContainerMetadata, s *runtime.PodSandboxMetadata) string {
return strings.Join([]string{
c.Name, // 0
s.Name, // 1: pod name
s.Namespace, // 2: pod namespace
s.Uid, // 3: pod uid
fmt.Sprintf("%d", c.Attempt), // 4
}, nameDelimiter)
}
// getSandboxRootDir returns the root directory for managing sandbox files,
// e.g. hosts files.
func (c *criService) getSandboxRootDir(id string) string {
return filepath.Join(c.config.RootDir, sandboxesDir, id)
}
// getVolatileSandboxRootDir returns the root directory for managing volatile sandbox files,
// e.g. named pipes.
func (c *criService) getVolatileSandboxRootDir(id string) string {
return filepath.Join(c.config.StateDir, sandboxesDir, id)
}
// getContainerRootDir returns the root directory for managing container files,
// e.g. state checkpoint.
func (c *criService) getContainerRootDir(id string) string {
return filepath.Join(c.config.RootDir, containersDir, id)
}
// getVolatileContainerRootDir returns the root directory for managing volatile container files,
// e.g. named pipes.
func (c *criService) getVolatileContainerRootDir(id string) string {
return filepath.Join(c.config.StateDir, containersDir, id)
}
// criContainerStateToString formats CRI container state to string.
func criContainerStateToString(state runtime.ContainerState) string {
return runtime.ContainerState_name[int32(state)]
}
// getRepoDigestAngTag returns image repoDigest and repoTag of the named image reference.
func getRepoDigestAndTag(namedRef docker.Named, digest imagedigest.Digest, schema1 bool) (string, string) {
var repoTag, repoDigest string
if _, ok := namedRef.(docker.NamedTagged); ok {
repoTag = namedRef.String()
}
if _, ok := namedRef.(docker.Canonical); ok {
repoDigest = namedRef.String()
} else if !schema1 {
// digest is not actual repo digest for schema1 image.
repoDigest = namedRef.Name() + "@" + digest.String()
}
return repoDigest, repoTag
}
// localResolve resolves image reference locally and returns corresponding image metadata. It
// returns store.ErrNotExist if the reference doesn't exist.
func (c *criService) localResolve(refOrID string) (imagestore.Image, error) {
getImageID := func(refOrId string) string {
if _, err := imagedigest.Parse(refOrID); err == nil {
return refOrID
}
return func(ref string) string {
// ref is not image id, try to resolve it locally.
// TODO(random-liu): Handle this error better for debugging.
normalized, err := docker.ParseDockerRef(ref)
if err != nil {
return ""
}
id, err := c.imageStore.Resolve(normalized.String())
if err != nil {
return ""
}
return id
}(refOrID)
}
imageID := getImageID(refOrID)
if imageID == "" {
// Try to treat ref as imageID
imageID = refOrID
}
return c.imageStore.Get(imageID)
}
// toContainerdImage converts an image object in image store to containerd image handler.
func (c *criService) toContainerdImage(ctx context.Context, image imagestore.Image) (containerd.Image, error) {
// image should always have at least one reference.
if len(image.References) == 0 {
return nil, errors.Errorf("invalid image with no reference %q", image.ID)
}
return c.client.GetImage(ctx, image.References[0])
}
// getUserFromImage gets uid or user name of the image user.
// If user is numeric, it will be treated as uid; or else, it is treated as user name.
func getUserFromImage(user string) (*int64, string) {
// return both empty if user is not specified in the image.
if user == "" {
return nil, ""
}
// split instances where the id may contain user:group
user = strings.Split(user, ":")[0]
// user could be either uid or user name. Try to interpret as numeric uid.
uid, err := strconv.ParseInt(user, 10, 64)
if err != nil {
// If user is non numeric, assume it's user name.
return nil, user
}
// If user is a numeric uid.
return &uid, ""
}
// ensureImageExists returns corresponding metadata of the image reference, if image is not
// pulled yet, the function will pull the image.
func (c *criService) ensureImageExists(ctx context.Context, ref string, config *runtime.PodSandboxConfig) (*imagestore.Image, error) {
image, err := c.localResolve(ref)
if err != nil && err != store.ErrNotExist {
return nil, errors.Wrapf(err, "failed to get image %q", ref)
}
if err == nil {
return &image, nil
}
// Pull image to ensure the image exists
resp, err := c.PullImage(ctx, &runtime.PullImageRequest{Image: &runtime.ImageSpec{Image: ref}, SandboxConfig: config})
if err != nil {
return nil, errors.Wrapf(err, "failed to pull image %q", ref)
}
imageID := resp.GetImageRef()
newImage, err := c.imageStore.Get(imageID)
if err != nil {
// It's still possible that someone removed the image right after it is pulled.
return nil, errors.Wrapf(err, "failed to get image %q after pulling", imageID)
}
return &newImage, nil
}
// isInCRIMounts checks whether a destination is in CRI mount list.
func isInCRIMounts(dst string, mounts []*runtime.Mount) bool {
for _, m := range mounts {
if filepath.Clean(m.ContainerPath) == filepath.Clean(dst) {
return true
}
}
return false
}
// filterLabel returns a label filter. Use `%q` here because containerd
// filter needs extra quote to work properly.
func filterLabel(k, v string) string {
return fmt.Sprintf("labels.%q==%q", k, v)
}
// buildLabel builds the labels from config to be passed to containerd
func buildLabels(configLabels map[string]string, containerType string) map[string]string {
labels := make(map[string]string)
for k, v := range configLabels {
labels[k] = v
}
labels[containerKindLabel] = containerType
return labels
}
// toRuntimeAuthConfig converts cri plugin auth config to runtime auth config.
func toRuntimeAuthConfig(a criconfig.AuthConfig) *runtime.AuthConfig {
return &runtime.AuthConfig{
Username: a.Username,
Password: a.Password,
Auth: a.Auth,
IdentityToken: a.IdentityToken,
}
}
// parseImageReferences parses a list of arbitrary image references and returns
// the repotags and repodigests
func parseImageReferences(refs []string) ([]string, []string) {
var tags, digests []string
for _, ref := range refs {
parsed, err := docker.ParseAnyReference(ref)
if err != nil {
continue
}
if _, ok := parsed.(docker.Canonical); ok {
digests = append(digests, parsed.String())
} else if _, ok := parsed.(docker.Tagged); ok {
tags = append(tags, parsed.String())
}
}
return tags, digests
}
// generateRuntimeOptions generates runtime options from cri plugin config.
func generateRuntimeOptions(r criconfig.Runtime, c criconfig.Config) (interface{}, error) {
if r.Options == nil {
if r.Type != plugin.RuntimeLinuxV1 {
return nil, nil
}
// This is a legacy config, generate runctypes.RuncOptions.
return &runctypes.RuncOptions{
Runtime: r.Engine,
RuntimeRoot: r.Root,
SystemdCgroup: c.SystemdCgroup,
}, nil
}
options := getRuntimeOptionsType(r.Type)
if err := toml.PrimitiveDecode(*r.Options, options); err != nil {
return nil, err
}
return options, nil
}
// getRuntimeOptionsType gets empty runtime options by the runtime type name.
func getRuntimeOptionsType(t string) interface{} {
switch t {
case plugin.RuntimeRuncV1:
fallthrough
case plugin.RuntimeRuncV2:
return &runcoptions.Options{}
case plugin.RuntimeLinuxV1:
return &runctypes.RuncOptions{}
case runtimeRunhcsV1:
return &runhcsoptions.Options{}
default:
return &runtimeoptions.Options{}
}
}
// getRuntimeOptions get runtime options from container metadata.
func getRuntimeOptions(c containers.Container) (interface{}, error) {
if c.Runtime.Options == nil {
return nil, nil
}
opts, err := typeurl.UnmarshalAny(c.Runtime.Options)
if err != nil {
return nil, err
}
return opts, nil
}
const (
// unknownExitCode is the exit code when exit reason is unknown.
unknownExitCode = 255
// unknownExitReason is the exit reason when exit reason is unknown.
unknownExitReason = "Unknown"
)
// unknownContainerStatus returns the default container status when its status is unknown.
func unknownContainerStatus() containerstore.Status {
return containerstore.Status{
CreatedAt: 0,
StartedAt: 0,
FinishedAt: 0,
ExitCode: unknownExitCode,
Reason: unknownExitReason,
Unknown: true,
}
}
// unknownSandboxStatus returns the default sandbox status when its status is unknown.
func unknownSandboxStatus() sandboxstore.Status {
return sandboxstore.Status{
State: sandboxstore.StateUnknown,
}
}
// getPassthroughAnnotations filters requested pod annotations by comparing
// against permitted annotations for the given runtime.
func getPassthroughAnnotations(podAnnotations map[string]string,
runtimePodAnnotations []string) (passthroughAnnotations map[string]string) {
passthroughAnnotations = make(map[string]string)
for podAnnotationKey, podAnnotationValue := range podAnnotations {
for _, pattern := range runtimePodAnnotations {
// Use path.Match instead of filepath.Match here.
// filepath.Match treated `\\` as path separator
// on windows, which is not what we want.
if ok, _ := path.Match(pattern, podAnnotationKey); ok {
passthroughAnnotations[podAnnotationKey] = podAnnotationValue
}
}
}
return passthroughAnnotations
}

View File

@@ -1,292 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"context"
"fmt"
"os"
"path"
"path/filepath"
"regexp"
"sort"
"strings"
"syscall"
"time"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/mount"
"github.com/containerd/cri/pkg/seccomp"
"github.com/containerd/cri/pkg/seutil"
runcapparmor "github.com/opencontainers/runc/libcontainer/apparmor"
"github.com/opencontainers/runtime-spec/specs-go"
"github.com/opencontainers/selinux/go-selinux/label"
"github.com/pkg/errors"
"golang.org/x/sys/unix"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
const (
// defaultSandboxOOMAdj is default omm adj for sandbox container. (kubernetes#47938).
defaultSandboxOOMAdj = -998
// defaultShmSize is the default size of the sandbox shm.
defaultShmSize = int64(1024 * 1024 * 64)
// relativeRootfsPath is the rootfs path relative to bundle path.
relativeRootfsPath = "rootfs"
// According to http://man7.org/linux/man-pages/man5/resolv.conf.5.html:
// "The search list is currently limited to six domains with a total of 256 characters."
maxDNSSearches = 6
// devShm is the default path of /dev/shm.
devShm = "/dev/shm"
// etcHosts is the default path of /etc/hosts file.
etcHosts = "/etc/hosts"
// etcHostname is the default path of /etc/hostname file.
etcHostname = "/etc/hostname"
// resolvConfPath is the abs path of resolv.conf on host or container.
resolvConfPath = "/etc/resolv.conf"
// hostnameEnv is the key for HOSTNAME env.
hostnameEnv = "HOSTNAME"
)
// getCgroupsPath generates container cgroups path.
func getCgroupsPath(cgroupsParent, id string) string {
base := path.Base(cgroupsParent)
if strings.HasSuffix(base, ".slice") {
// For a.slice/b.slice/c.slice, base is c.slice.
// runc systemd cgroup path format is "slice:prefix:name".
return strings.Join([]string{base, "cri-containerd", id}, ":")
}
return filepath.Join(cgroupsParent, id)
}
// getSandboxHostname returns the hostname file path inside the sandbox root directory.
func (c *criService) getSandboxHostname(id string) string {
return filepath.Join(c.getSandboxRootDir(id), "hostname")
}
// getSandboxHosts returns the hosts file path inside the sandbox root directory.
func (c *criService) getSandboxHosts(id string) string {
return filepath.Join(c.getSandboxRootDir(id), "hosts")
}
// getResolvPath returns resolv.conf filepath for specified sandbox.
func (c *criService) getResolvPath(id string) string {
return filepath.Join(c.getSandboxRootDir(id), "resolv.conf")
}
// getSandboxDevShm returns the shm file path inside the sandbox root directory.
func (c *criService) getSandboxDevShm(id string) string {
return filepath.Join(c.getVolatileSandboxRootDir(id), "shm")
}
func toLabel(selinuxOptions *runtime.SELinuxOption) ([]string, error) {
var labels []string
if selinuxOptions == nil {
return nil, nil
}
if err := checkSelinuxLevel(selinuxOptions.Level); err != nil {
return nil, err
}
if selinuxOptions.User != "" {
labels = append(labels, "user:"+selinuxOptions.User)
}
if selinuxOptions.Role != "" {
labels = append(labels, "role:"+selinuxOptions.Role)
}
if selinuxOptions.Type != "" {
labels = append(labels, "type:"+selinuxOptions.Type)
}
if selinuxOptions.Level != "" {
labels = append(labels, "level:"+selinuxOptions.Level)
}
return labels, nil
}
func initLabelsFromOpt(selinuxOpts *runtime.SELinuxOption) (string, string, error) {
labels, err := toLabel(selinuxOpts)
if err != nil {
return "", "", err
}
return label.InitLabels(labels)
}
func checkSelinuxLevel(level string) error {
if len(level) == 0 {
return nil
}
matched, err := regexp.MatchString(`^s\d(-s\d)??(:c\d{1,4}(\.c\d{1,4})?(,c\d{1,4}(\.c\d{1,4})?)*)?$`, level)
if err != nil {
return errors.Wrapf(err, "the format of 'level' %q is not correct", level)
}
if !matched {
return fmt.Errorf("the format of 'level' %q is not correct", level)
}
return nil
}
func (c *criService) apparmorEnabled() bool {
return runcapparmor.IsEnabled() && !c.config.DisableApparmor
}
func (c *criService) seccompEnabled() bool {
return seccomp.IsEnabled()
}
// openLogFile opens/creates a container log file.
func openLogFile(path string) (*os.File, error) {
return os.OpenFile(path, os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0640)
}
// unmountRecursive unmounts the target and all mounts underneath, starting with
// the deepest mount first.
func unmountRecursive(ctx context.Context, target string) error {
mounts, err := mount.Self()
if err != nil {
return err
}
var toUnmount []string
for _, m := range mounts {
p, err := filepath.Rel(target, m.Mountpoint)
if err != nil {
return err
}
if !strings.HasPrefix(p, "..") {
toUnmount = append(toUnmount, m.Mountpoint)
}
}
// Make the deepest mount be first
sort.Slice(toUnmount, func(i, j int) bool {
return len(toUnmount[i]) > len(toUnmount[j])
})
for i, mountPath := range toUnmount {
if err := mount.UnmountAll(mountPath, unix.MNT_DETACH); err != nil {
if i == len(toUnmount)-1 { // last mount
return err
}
// This is some submount, we can ignore this error for now, the final unmount will fail if this is a real problem
log.G(ctx).WithError(err).Debugf("failed to unmount submount %s", mountPath)
}
}
return nil
}
// ensureRemoveAll wraps `os.RemoveAll` to check for specific errors that can
// often be remedied.
// Only use `ensureRemoveAll` if you really want to make every effort to remove
// a directory.
//
// Because of the way `os.Remove` (and by extension `os.RemoveAll`) works, there
// can be a race between reading directory entries and then actually attempting
// to remove everything in the directory.
// These types of errors do not need to be returned since it's ok for the dir to
// be gone we can just retry the remove operation.
//
// This should not return a `os.ErrNotExist` kind of error under any circumstances
func ensureRemoveAll(ctx context.Context, dir string) error {
notExistErr := make(map[string]bool)
// track retries
exitOnErr := make(map[string]int)
maxRetry := 50
// Attempt to unmount anything beneath this dir first.
if err := unmountRecursive(ctx, dir); err != nil {
log.G(ctx).WithError(err).Debugf("failed to do initial unmount of %s", dir)
}
for {
err := os.RemoveAll(dir)
if err == nil {
return nil
}
pe, ok := err.(*os.PathError)
if !ok {
return err
}
if os.IsNotExist(err) {
if notExistErr[pe.Path] {
return err
}
notExistErr[pe.Path] = true
// There is a race where some subdir can be removed but after the
// parent dir entries have been read.
// So the path could be from `os.Remove(subdir)`
// If the reported non-existent path is not the passed in `dir` we
// should just retry, but otherwise return with no error.
if pe.Path == dir {
return nil
}
continue
}
if pe.Err != syscall.EBUSY {
return err
}
if e := mount.Unmount(pe.Path, unix.MNT_DETACH); e != nil {
return errors.Wrapf(e, "error while removing %s", dir)
}
if exitOnErr[pe.Path] == maxRetry {
return err
}
exitOnErr[pe.Path]++
time.Sleep(100 * time.Millisecond)
}
}
var vmbasedRuntimes = []string{
"io.containerd.kata",
}
func isVMBasedRuntime(runtimeType string) bool {
for _, rt := range vmbasedRuntimes {
if strings.Contains(runtimeType, rt) {
return true
}
}
return false
}
func modifyProcessLabel(runtimeType string, spec *specs.Spec) error {
if !isVMBasedRuntime(runtimeType) {
return nil
}
l, err := getKVMLabel(spec.Process.SelinuxLabel)
if err != nil {
return errors.Wrap(err, "failed to get selinux kvm label")
}
spec.Process.SelinuxLabel = l
return nil
}
func getKVMLabel(l string) (string, error) {
if !seutil.HasType("container_kvm_t") {
return "", nil
}
return seutil.ChangeToKVM(l)
}

View File

@@ -1,225 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"context"
"os"
"path/filepath"
"syscall"
"time"
"github.com/opencontainers/runtime-spec/specs-go"
)
// openLogFile opens/creates a container log file.
// It specifies `FILE_SHARE_DELETE` option to make sure
// log files can be rotated by kubelet.
// TODO(windows): Use golang support after 1.14. (https://github.com/golang/go/issues/32088)
func openLogFile(path string) (*os.File, error) {
path = fixLongPath(path)
if len(path) == 0 {
return nil, syscall.ERROR_FILE_NOT_FOUND
}
pathp, err := syscall.UTF16PtrFromString(path)
if err != nil {
return nil, err
}
createmode := uint32(syscall.OPEN_ALWAYS)
access := uint32(syscall.FILE_APPEND_DATA)
sharemode := uint32(syscall.FILE_SHARE_READ | syscall.FILE_SHARE_WRITE | syscall.FILE_SHARE_DELETE)
h, err := syscall.CreateFile(pathp, access, sharemode, nil, createmode, syscall.FILE_ATTRIBUTE_NORMAL, 0)
if err != nil {
return nil, err
}
return os.NewFile(uintptr(h), path), nil
}
// Copyright (c) 2009 The Go Authors. All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
// fixLongPath returns the extended-length (\\?\-prefixed) form of
// path when needed, in order to avoid the default 260 character file
// path limit imposed by Windows. If path is not easily converted to
// the extended-length form (for example, if path is a relative path
// or contains .. elements), or is short enough, fixLongPath returns
// path unmodified.
//
// See https://msdn.microsoft.com/en-us/library/windows/desktop/aa365247(v=vs.85).aspx#maxpath
//
// This is copied from https://golang.org/src/path/filepath/path_windows.go.
func fixLongPath(path string) string {
// Do nothing (and don't allocate) if the path is "short".
// Empirically (at least on the Windows Server 2013 builder),
// the kernel is arbitrarily okay with < 248 bytes. That
// matches what the docs above say:
// "When using an API to create a directory, the specified
// path cannot be so long that you cannot append an 8.3 file
// name (that is, the directory name cannot exceed MAX_PATH
// minus 12)." Since MAX_PATH is 260, 260 - 12 = 248.
//
// The MSDN docs appear to say that a normal path that is 248 bytes long
// will work; empirically the path must be less then 248 bytes long.
if len(path) < 248 {
// Don't fix. (This is how Go 1.7 and earlier worked,
// not automatically generating the \\?\ form)
return path
}
// The extended form begins with \\?\, as in
// \\?\c:\windows\foo.txt or \\?\UNC\server\share\foo.txt.
// The extended form disables evaluation of . and .. path
// elements and disables the interpretation of / as equivalent
// to \. The conversion here rewrites / to \ and elides
// . elements as well as trailing or duplicate separators. For
// simplicity it avoids the conversion entirely for relative
// paths or paths containing .. elements. For now,
// \\server\share paths are not converted to
// \\?\UNC\server\share paths because the rules for doing so
// are less well-specified.
if len(path) >= 2 && path[:2] == `\\` {
// Don't canonicalize UNC paths.
return path
}
if !filepath.IsAbs(path) {
// Relative path
return path
}
const prefix = `\\?`
pathbuf := make([]byte, len(prefix)+len(path)+len(`\`))
copy(pathbuf, prefix)
n := len(path)
r, w := 0, len(prefix)
for r < n {
switch {
case os.IsPathSeparator(path[r]):
// empty block
r++
case path[r] == '.' && (r+1 == n || os.IsPathSeparator(path[r+1])):
// /./
r++
case r+1 < n && path[r] == '.' && path[r+1] == '.' && (r+2 == n || os.IsPathSeparator(path[r+2])):
// /../ is currently unhandled
return path
default:
pathbuf[w] = '\\'
w++
for ; r < n && !os.IsPathSeparator(path[r]); r++ {
pathbuf[w] = path[r]
w++
}
}
}
// A drive's root directory needs a trailing \
if w == len(`\\?\c:`) {
pathbuf[w] = '\\'
w++
}
return string(pathbuf[:w])
}
// ensureRemoveAll wraps `os.RemoveAll` to check for specific errors that can
// often be remedied.
// Only use `ensureRemoveAll` if you really want to make every effort to remove
// a directory.
//
// Because of the way `os.Remove` (and by extension `os.RemoveAll`) works, there
// can be a race between reading directory entries and then actually attempting
// to remove everything in the directory.
// These types of errors do not need to be returned since it's ok for the dir to
// be gone we can just retry the remove operation.
//
// This should not return a `os.ErrNotExist` kind of error under any circumstances
func ensureRemoveAll(_ context.Context, dir string) error {
notExistErr := make(map[string]bool)
// track retries
exitOnErr := make(map[string]int)
maxRetry := 50
for {
err := os.RemoveAll(dir)
if err == nil {
return nil
}
pe, ok := err.(*os.PathError)
if !ok {
return err
}
if os.IsNotExist(err) {
if notExistErr[pe.Path] {
return err
}
notExistErr[pe.Path] = true
// There is a race where some subdir can be removed but after the
// parent dir entries have been read.
// So the path could be from `os.Remove(subdir)`
// If the reported non-existent path is not the passed in `dir` we
// should just retry, but otherwise return with no error.
if pe.Path == dir {
return nil
}
continue
}
if pe.Err != syscall.EBUSY {
return err
}
if exitOnErr[pe.Path] == maxRetry {
return err
}
exitOnErr[pe.Path]++
time.Sleep(100 * time.Millisecond)
}
}
func modifyProcessLabel(runtimeType string, spec *specs.Spec) error {
return nil
}

View File

@@ -1,38 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// ListImages lists existing images.
// TODO(random-liu): Add image list filters after CRI defines this more clear, and kubelet
// actually needs it.
func (c *criService) ListImages(ctx context.Context, r *runtime.ListImagesRequest) (*runtime.ListImagesResponse, error) {
imagesInStore := c.imageStore.List()
var images []*runtime.Image
for _, image := range imagesInStore {
// TODO(random-liu): [P0] Make sure corresponding snapshot exists. What if snapshot
// doesn't exist?
images = append(images, toCRIImage(image))
}
return &runtime.ListImagesResponse{Images: images}, nil
}

View File

@@ -1,519 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"crypto/tls"
"crypto/x509"
"encoding/base64"
"fmt"
"io/ioutil"
"net"
"net/http"
"net/url"
"strings"
"time"
"github.com/containerd/containerd"
"github.com/containerd/containerd/errdefs"
containerdimages "github.com/containerd/containerd/images"
"github.com/containerd/containerd/labels"
"github.com/containerd/containerd/log"
distribution "github.com/containerd/containerd/reference/docker"
"github.com/containerd/containerd/remotes/docker"
"github.com/containerd/imgcrypt"
"github.com/containerd/imgcrypt/images/encryption"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
criconfig "github.com/containerd/cri/pkg/config"
)
// For image management:
// 1) We have an in-memory metadata index to:
// a. Maintain ImageID -> RepoTags, ImageID -> RepoDigset relationships; ImageID
// is the digest of image config, which conforms to oci image spec.
// b. Cache constant and useful information such as image chainID, config etc.
// c. An image will be added into the in-memory metadata only when it's successfully
// pulled and unpacked.
//
// 2) We use containerd image metadata store and content store:
// a. To resolve image reference (digest/tag) locally. During pulling image, we
// normalize the image reference provided by user, and put it into image metadata
// store with resolved descriptor. For the other operations, if image id is provided,
// we'll access the in-memory metadata index directly; if image reference is
// provided, we'll normalize it, resolve it in containerd image metadata store
// to get the image id.
// b. As the backup of in-memory metadata in 1). During startup, the in-memory
// metadata could be re-constructed from image metadata store + content store.
//
// Several problems with current approach:
// 1) An entry in containerd image metadata store doesn't mean a "READY" (successfully
// pulled and unpacked) image. E.g. during pulling, the client gets killed. In that case,
// if we saw an image without snapshots or with in-complete contents during startup,
// should we re-pull the image? Or should we remove the entry?
//
// yanxuean: We can't delete image directly, because we don't know if the image
// is pulled by us. There are resource leakage.
//
// 2) Containerd suggests user to add entry before pulling the image. However if
// an error occurs during the pulling, should we remove the entry from metadata
// store? Or should we leave it there until next startup (resource leakage)?
//
// 3) The cri plugin only exposes "READY" (successfully pulled and unpacked) images
// to the user, which are maintained in the in-memory metadata index. However, it's
// still possible that someone else removes the content or snapshot by-pass the cri plugin,
// how do we detect that and update the in-memory metadata correspondingly? Always
// check whether corresponding snapshot is ready when reporting image status?
//
// 4) Is the content important if we cached necessary information in-memory
// after we pull the image? How to manage the disk usage of contents? If some
// contents are missing but snapshots are ready, is the image still "READY"?
// PullImage pulls an image with authentication config.
func (c *criService) PullImage(ctx context.Context, r *runtime.PullImageRequest) (*runtime.PullImageResponse, error) {
imageRef := r.GetImage().GetImage()
namedRef, err := distribution.ParseDockerRef(imageRef)
if err != nil {
return nil, errors.Wrapf(err, "failed to parse image reference %q", imageRef)
}
ref := namedRef.String()
if ref != imageRef {
log.G(ctx).Debugf("PullImage using normalized image ref: %q", ref)
}
var (
resolver = docker.NewResolver(docker.ResolverOptions{
Headers: c.config.Registry.Headers,
Hosts: c.registryHosts(r.GetAuth()),
})
isSchema1 bool
imageHandler containerdimages.HandlerFunc = func(_ context.Context,
desc imagespec.Descriptor) ([]imagespec.Descriptor, error) {
if desc.MediaType == containerdimages.MediaTypeDockerSchema1Manifest {
isSchema1 = true
}
return nil, nil
}
)
pullOpts := []containerd.RemoteOpt{
containerd.WithSchema1Conversion,
containerd.WithResolver(resolver),
containerd.WithPullSnapshotter(c.config.ContainerdConfig.Snapshotter),
containerd.WithPullUnpack,
containerd.WithPullLabel(imageLabelKey, imageLabelValue),
containerd.WithMaxConcurrentDownloads(c.config.MaxConcurrentDownloads),
containerd.WithImageHandler(imageHandler),
}
pullOpts = append(pullOpts, c.encryptedImagesPullOpts()...)
if !c.config.ContainerdConfig.DisableSnapshotAnnotations {
pullOpts = append(pullOpts,
containerd.WithImageHandlerWrapper(appendInfoHandlerWrapper(ref)))
}
if c.config.ContainerdConfig.DiscardUnpackedLayers {
// Allows GC to clean layers up from the content store after unpacking
pullOpts = append(pullOpts,
containerd.WithChildLabelMap(containerdimages.ChildGCLabelsFilterLayers))
}
image, err := c.client.Pull(ctx, ref, pullOpts...)
if err != nil {
return nil, errors.Wrapf(err, "failed to pull and unpack image %q", ref)
}
configDesc, err := image.Config(ctx)
if err != nil {
return nil, errors.Wrap(err, "get image config descriptor")
}
imageID := configDesc.Digest.String()
repoDigest, repoTag := getRepoDigestAndTag(namedRef, image.Target().Digest, isSchema1)
for _, r := range []string{imageID, repoTag, repoDigest} {
if r == "" {
continue
}
if err := c.createImageReference(ctx, r, image.Target()); err != nil {
return nil, errors.Wrapf(err, "failed to create image reference %q", r)
}
// Update image store to reflect the newest state in containerd.
// No need to use `updateImage`, because the image reference must
// have been managed by the cri plugin.
if err := c.imageStore.Update(ctx, r); err != nil {
return nil, errors.Wrapf(err, "failed to update image store %q", r)
}
}
log.G(ctx).Debugf("Pulled image %q with image id %q, repo tag %q, repo digest %q", imageRef, imageID,
repoTag, repoDigest)
// NOTE(random-liu): the actual state in containerd is the source of truth, even we maintain
// in-memory image store, it's only for in-memory indexing. The image could be removed
// by someone else anytime, before/during/after we create the metadata. We should always
// check the actual state in containerd before using the image or returning status of the
// image.
return &runtime.PullImageResponse{ImageRef: imageID}, nil
}
// ParseAuth parses AuthConfig and returns username and password/secret required by containerd.
func ParseAuth(auth *runtime.AuthConfig, host string) (string, string, error) {
if auth == nil {
return "", "", nil
}
if auth.ServerAddress != "" {
// Do not return the auth info when server address doesn't match.
u, err := url.Parse(auth.ServerAddress)
if err != nil {
return "", "", errors.Wrap(err, "parse server address")
}
if host != u.Host {
return "", "", nil
}
}
if auth.Username != "" {
return auth.Username, auth.Password, nil
}
if auth.IdentityToken != "" {
return "", auth.IdentityToken, nil
}
if auth.Auth != "" {
decLen := base64.StdEncoding.DecodedLen(len(auth.Auth))
decoded := make([]byte, decLen)
_, err := base64.StdEncoding.Decode(decoded, []byte(auth.Auth))
if err != nil {
return "", "", err
}
fields := strings.SplitN(string(decoded), ":", 2)
if len(fields) != 2 {
return "", "", errors.Errorf("invalid decoded auth: %q", decoded)
}
user, passwd := fields[0], fields[1]
return user, strings.Trim(passwd, "\x00"), nil
}
// TODO(random-liu): Support RegistryToken.
// An empty auth config is valid for anonymous registry
return "", "", nil
}
// createImageReference creates image reference inside containerd image store.
// Note that because create and update are not finished in one transaction, there could be race. E.g.
// the image reference is deleted by someone else after create returns already exists, but before update
// happens.
func (c *criService) createImageReference(ctx context.Context, name string, desc imagespec.Descriptor) error {
img := containerdimages.Image{
Name: name,
Target: desc,
// Add a label to indicate that the image is managed by the cri plugin.
Labels: map[string]string{imageLabelKey: imageLabelValue},
}
// TODO(random-liu): Figure out which is the more performant sequence create then update or
// update then create.
oldImg, err := c.client.ImageService().Create(ctx, img)
if err == nil || !errdefs.IsAlreadyExists(err) {
return err
}
if oldImg.Target.Digest == img.Target.Digest && oldImg.Labels[imageLabelKey] == imageLabelValue {
return nil
}
_, err = c.client.ImageService().Update(ctx, img, "target", "labels")
return err
}
// updateImage updates image store to reflect the newest state of an image reference
// in containerd. If the reference is not managed by the cri plugin, the function also
// generates necessary metadata for the image and make it managed.
func (c *criService) updateImage(ctx context.Context, r string) error {
img, err := c.client.GetImage(ctx, r)
if err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "get image by reference")
}
if err == nil && img.Labels()[imageLabelKey] != imageLabelValue {
// Make sure the image has the image id as its unique
// identifier that references the image in its lifetime.
configDesc, err := img.Config(ctx)
if err != nil {
return errors.Wrap(err, "get image id")
}
id := configDesc.Digest.String()
if err := c.createImageReference(ctx, id, img.Target()); err != nil {
return errors.Wrapf(err, "create image id reference %q", id)
}
if err := c.imageStore.Update(ctx, id); err != nil {
return errors.Wrapf(err, "update image store for %q", id)
}
// The image id is ready, add the label to mark the image as managed.
if err := c.createImageReference(ctx, r, img.Target()); err != nil {
return errors.Wrap(err, "create managed label")
}
}
// If the image is not found, we should continue updating the cache,
// so that the image can be removed from the cache.
if err := c.imageStore.Update(ctx, r); err != nil {
return errors.Wrapf(err, "update image store for %q", r)
}
return nil
}
// getTLSConfig returns a TLSConfig configured with a CA/Cert/Key specified by registryTLSConfig
func (c *criService) getTLSConfig(registryTLSConfig criconfig.TLSConfig) (*tls.Config, error) {
var (
tlsConfig = &tls.Config{}
cert tls.Certificate
err error
)
if registryTLSConfig.CertFile != "" && registryTLSConfig.KeyFile == "" {
return nil, errors.Errorf("cert file %q was specified, but no corresponding key file was specified", registryTLSConfig.CertFile)
}
if registryTLSConfig.CertFile == "" && registryTLSConfig.KeyFile != "" {
return nil, errors.Errorf("key file %q was specified, but no corresponding cert file was specified", registryTLSConfig.KeyFile)
}
if registryTLSConfig.CertFile != "" && registryTLSConfig.KeyFile != "" {
cert, err = tls.LoadX509KeyPair(registryTLSConfig.CertFile, registryTLSConfig.KeyFile)
if err != nil {
return nil, errors.Wrap(err, "failed to load cert file")
}
if len(cert.Certificate) != 0 {
tlsConfig.Certificates = []tls.Certificate{cert}
}
tlsConfig.BuildNameToCertificate() // nolint:staticcheck
}
if registryTLSConfig.CAFile != "" {
caCertPool, err := x509.SystemCertPool()
if err != nil {
return nil, errors.Wrap(err, "failed to get system cert pool")
}
caCert, err := ioutil.ReadFile(registryTLSConfig.CAFile)
if err != nil {
return nil, errors.Wrap(err, "failed to load CA file")
}
caCertPool.AppendCertsFromPEM(caCert)
tlsConfig.RootCAs = caCertPool
}
tlsConfig.InsecureSkipVerify = registryTLSConfig.InsecureSkipVerify
return tlsConfig, nil
}
// registryHosts is the registry hosts to be used by the resolver.
func (c *criService) registryHosts(auth *runtime.AuthConfig) docker.RegistryHosts {
return func(host string) ([]docker.RegistryHost, error) {
var registries []docker.RegistryHost
endpoints, err := c.registryEndpoints(host)
if err != nil {
return nil, errors.Wrap(err, "get registry endpoints")
}
for _, e := range endpoints {
u, err := url.Parse(e)
if err != nil {
return nil, errors.Wrapf(err, "parse registry endpoint %q from mirrors", e)
}
var (
transport = newTransport()
client = &http.Client{Transport: transport}
config = c.config.Registry.Configs[u.Host]
)
if config.TLS != nil {
transport.TLSClientConfig, err = c.getTLSConfig(*config.TLS)
if err != nil {
return nil, errors.Wrapf(err, "get TLSConfig for registry %q", e)
}
}
if auth == nil && config.Auth != nil {
auth = toRuntimeAuthConfig(*config.Auth)
}
if u.Path == "" {
u.Path = "/v2"
}
registries = append(registries, docker.RegistryHost{
Client: client,
Authorizer: docker.NewDockerAuthorizer(
docker.WithAuthClient(client),
docker.WithAuthCreds(func(host string) (string, string, error) {
return ParseAuth(auth, host)
})),
Host: u.Host,
Scheme: u.Scheme,
Path: u.Path,
Capabilities: docker.HostCapabilityResolve | docker.HostCapabilityPull,
})
}
return registries, nil
}
}
// defaultScheme returns the default scheme for a registry host.
func defaultScheme(host string) string {
if h, _, err := net.SplitHostPort(host); err == nil {
host = h
}
if host == "localhost" || host == "127.0.0.1" || host == "::1" {
return "http"
}
return "https"
}
// addDefaultScheme returns the endpoint with default scheme
func addDefaultScheme(endpoint string) (string, error) {
if strings.Contains(endpoint, "://") {
return endpoint, nil
}
ue := "dummy://" + endpoint
u, err := url.Parse(ue)
if err != nil {
return "", err
}
return fmt.Sprintf("%s://%s", defaultScheme(u.Host), endpoint), nil
}
// registryEndpoints returns endpoints for a given host.
// It adds default registry endpoint if it does not exist in the passed-in endpoint list.
// It also supports wildcard host matching with `*`.
func (c *criService) registryEndpoints(host string) ([]string, error) {
var endpoints []string
_, ok := c.config.Registry.Mirrors[host]
if ok {
endpoints = c.config.Registry.Mirrors[host].Endpoints
} else {
endpoints = c.config.Registry.Mirrors["*"].Endpoints
}
defaultHost, err := docker.DefaultHost(host)
if err != nil {
return nil, errors.Wrap(err, "get default host")
}
for i := range endpoints {
en, err := addDefaultScheme(endpoints[i])
if err != nil {
return nil, errors.Wrap(err, "parse endpoint url")
}
endpoints[i] = en
}
for _, e := range endpoints {
u, err := url.Parse(e)
if err != nil {
return nil, errors.Wrap(err, "parse endpoint url")
}
if u.Host == host {
// Do not add default if the endpoint already exists.
return endpoints, nil
}
}
return append(endpoints, defaultScheme(defaultHost)+"://"+defaultHost), nil
}
// newTransport returns a new HTTP transport used to pull image.
// TODO(random-liu): Create a library and share this code with `ctr`.
func newTransport() *http.Transport {
return &http.Transport{
Proxy: http.ProxyFromEnvironment,
DialContext: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
FallbackDelay: 300 * time.Millisecond,
}).DialContext,
MaxIdleConns: 10,
IdleConnTimeout: 30 * time.Second,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 5 * time.Second,
}
}
// encryptedImagesPullOpts returns the necessary list of pull options required
// for decryption of encrypted images based on the cri decryption configuration.
func (c *criService) encryptedImagesPullOpts() []containerd.RemoteOpt {
if c.config.ImageDecryption.KeyModel == criconfig.KeyModelNode {
ltdd := imgcrypt.Payload{}
decUnpackOpt := encryption.WithUnpackConfigApplyOpts(encryption.WithDecryptedUnpack(&ltdd))
opt := containerd.WithUnpackOpts([]containerd.UnpackOpt{decUnpackOpt})
return []containerd.RemoteOpt{opt}
}
return nil
}
const (
// targetRefLabel is a label which contains image reference and will be passed
// to snapshotters.
targetRefLabel = "containerd.io/snapshot/cri.image-ref"
// targetDigestLabel is a label which contains layer digest and will be passed
// to snapshotters.
targetDigestLabel = "containerd.io/snapshot/cri.layer-digest"
// targetImageLayersLabel is a label which contains layer digests contained in
// the target image and will be passed to snapshotters for preparing layers in
// parallel. Skipping some layers is allowed and only affects performance.
targetImageLayersLabel = "containerd.io/snapshot/cri.image-layers"
)
// appendInfoHandlerWrapper makes a handler which appends some basic information
// of images to each layer descriptor as annotations during unpack. These
// annotations will be passed to snapshotters as labels. These labels will be
// used mainly by stargz-based snapshotters for querying image contents from the
// registry.
func appendInfoHandlerWrapper(ref string) func(f containerdimages.Handler) containerdimages.Handler {
return func(f containerdimages.Handler) containerdimages.Handler {
return containerdimages.HandlerFunc(func(ctx context.Context, desc imagespec.Descriptor) ([]imagespec.Descriptor, error) {
children, err := f.Handle(ctx, desc)
if err != nil {
return nil, err
}
switch desc.MediaType {
case imagespec.MediaTypeImageManifest, containerdimages.MediaTypeDockerSchema2Manifest:
for i := range children {
c := &children[i]
if containerdimages.IsLayerType(c.MediaType) {
if c.Annotations == nil {
c.Annotations = make(map[string]string)
}
c.Annotations[targetRefLabel] = ref
c.Annotations[targetDigestLabel] = c.Digest.String()
c.Annotations[targetImageLayersLabel] = getLayers(ctx, targetImageLayersLabel, children[i:], labels.Validate)
}
}
}
return children, nil
})
}
}
// getLayers returns comma-separated digests based on the passed list of
// descriptors. The returned list contains as many digests as possible as well
// as meets the label validation.
func getLayers(ctx context.Context, key string, descs []imagespec.Descriptor, validate func(k, v string) error) (layers string) {
var item string
for _, l := range descs {
if containerdimages.IsLayerType(l.MediaType) {
item = l.Digest.String()
if layers != "" {
item = "," + item
}
// This avoids the label hits the size limitation.
if err := validate(key, layers+item); err != nil {
log.G(ctx).WithError(err).WithField("label", key).Debugf("%q is omitted in the layers list", l.Digest.String())
break
}
layers += item
}
}
return
}

View File

@@ -1,65 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/images"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store"
)
// RemoveImage removes the image.
// TODO(random-liu): Update CRI to pass image reference instead of ImageSpec. (See
// kubernetes/kubernetes#46255)
// TODO(random-liu): We should change CRI to distinguish image id and image spec.
// Remove the whole image no matter the it's image id or reference. This is the
// semantic defined in CRI now.
func (c *criService) RemoveImage(ctx context.Context, r *runtime.RemoveImageRequest) (*runtime.RemoveImageResponse, error) {
image, err := c.localResolve(r.GetImage().GetImage())
if err != nil {
if err == store.ErrNotExist {
// return empty without error when image not found.
return &runtime.RemoveImageResponse{}, nil
}
return nil, errors.Wrapf(err, "can not resolve %q locally", r.GetImage().GetImage())
}
// Remove all image references.
for i, ref := range image.References {
var opts []images.DeleteOpt
if i == len(image.References)-1 {
// Delete the last image reference synchronously to trigger garbage collection.
// This is best effort. It is possible that the image reference is deleted by
// someone else before this point.
opts = []images.DeleteOpt{images.SynchronousDelete()}
}
err = c.client.ImageService().Delete(ctx, ref, opts...)
if err == nil || errdefs.IsNotFound(err) {
// Update image store to reflect the newest state in containerd.
if err := c.imageStore.Update(ctx, ref); err != nil {
return nil, errors.Wrapf(err, "failed to update image reference %q for %q", ref, image.ID)
}
continue
}
return nil, errors.Wrapf(err, "failed to delete image reference %q for %q", ref, image.ID)
}
return &runtime.RemoveImageResponse{}, nil
}

View File

@@ -1,105 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
"github.com/containerd/containerd/log"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store"
imagestore "github.com/containerd/cri/pkg/store/image"
)
// ImageStatus returns the status of the image, returns nil if the image isn't present.
// TODO(random-liu): We should change CRI to distinguish image id and image spec. (See
// kubernetes/kubernetes#46255)
func (c *criService) ImageStatus(ctx context.Context, r *runtime.ImageStatusRequest) (*runtime.ImageStatusResponse, error) {
image, err := c.localResolve(r.GetImage().GetImage())
if err != nil {
if err == store.ErrNotExist {
// return empty without error when image not found.
return &runtime.ImageStatusResponse{}, nil
}
return nil, errors.Wrapf(err, "can not resolve %q locally", r.GetImage().GetImage())
}
// TODO(random-liu): [P0] Make sure corresponding snapshot exists. What if snapshot
// doesn't exist?
runtimeImage := toCRIImage(image)
info, err := c.toCRIImageInfo(ctx, &image, r.GetVerbose())
if err != nil {
return nil, errors.Wrap(err, "failed to generate image info")
}
return &runtime.ImageStatusResponse{
Image: runtimeImage,
Info: info,
}, nil
}
// toCRIImage converts internal image object to CRI runtime.Image.
func toCRIImage(image imagestore.Image) *runtime.Image {
repoTags, repoDigests := parseImageReferences(image.References)
runtimeImage := &runtime.Image{
Id: image.ID,
RepoTags: repoTags,
RepoDigests: repoDigests,
Size_: uint64(image.Size),
}
uid, username := getUserFromImage(image.ImageSpec.Config.User)
if uid != nil {
runtimeImage.Uid = &runtime.Int64Value{Value: *uid}
}
runtimeImage.Username = username
return runtimeImage
}
// TODO (mikebrow): discuss moving this struct and / or constants for info map for some or all of these fields to CRI
type verboseImageInfo struct {
ChainID string `json:"chainID"`
ImageSpec imagespec.Image `json:"imageSpec"`
}
// toCRIImageInfo converts internal image object information to CRI image status response info map.
func (c *criService) toCRIImageInfo(ctx context.Context, image *imagestore.Image, verbose bool) (map[string]string, error) {
if !verbose {
return nil, nil
}
info := make(map[string]string)
imi := &verboseImageInfo{
ChainID: image.ChainID,
ImageSpec: image.ImageSpec,
}
m, err := json.Marshal(imi)
if err == nil {
info["info"] = string(m)
} else {
log.G(ctx).WithError(err).Errorf("failed to marshal info %v", imi)
info["info"] = err.Error()
}
return info, nil
}

View File

@@ -1,52 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"time"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// ImageFsInfo returns information of the filesystem that is used to store images.
// TODO(windows): Usage for windows is always 0 right now. Support this for windows.
func (c *criService) ImageFsInfo(ctx context.Context, r *runtime.ImageFsInfoRequest) (*runtime.ImageFsInfoResponse, error) {
snapshots := c.snapshotStore.List()
timestamp := time.Now().UnixNano()
var usedBytes, inodesUsed uint64
for _, sn := range snapshots {
// Use the oldest timestamp as the timestamp of imagefs info.
if sn.Timestamp < timestamp {
timestamp = sn.Timestamp
}
usedBytes += sn.Size
inodesUsed += sn.Inodes
}
// TODO(random-liu): Handle content store
return &runtime.ImageFsInfoResponse{
ImageFilesystems: []*runtime.FilesystemUsage{
{
Timestamp: timestamp,
FsId: &runtime.FilesystemIdentifier{Mountpoint: c.imageFSPath},
UsedBytes: &runtime.UInt64Value{Value: usedBytes},
InodesUsed: &runtime.UInt64Value{Value: inodesUsed},
},
},
}, nil
}

View File

@@ -1,490 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"errors"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
)
// instrumentedService wraps service with containerd namespace and logs.
type instrumentedService struct {
c *criService
}
func newInstrumentedService(c *criService) grpcServices {
return &instrumentedService{c: c}
}
// checkInitialized returns error if the server is not fully initialized.
// GRPC service request handlers should return error before server is fully
// initialized.
// NOTE(random-liu): All following functions MUST check initialized at the beginning.
func (in *instrumentedService) checkInitialized() error {
if in.c.initialized.IsSet() {
return nil
}
return errors.New("server is not initialized yet")
}
func (in *instrumentedService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandboxRequest) (res *runtime.RunPodSandboxResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("RunPodsandbox for %+v", r.GetConfig().GetMetadata())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("RunPodSandbox for %+v failed, error", r.GetConfig().GetMetadata())
} else {
log.G(ctx).Infof("RunPodSandbox for %+v returns sandbox id %q", r.GetConfig().GetMetadata(), res.GetPodSandboxId())
}
}()
res, err = in.c.RunPodSandbox(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ListPodSandbox(ctx context.Context, r *runtime.ListPodSandboxRequest) (res *runtime.ListPodSandboxResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ListPodSandbox with filter %+v", r.GetFilter())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("ListPodSandbox failed")
} else {
log.G(ctx).Tracef("ListPodSandbox returns pod sandboxes %+v", res.GetItems())
}
}()
res, err = in.c.ListPodSandbox(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) PodSandboxStatus(ctx context.Context, r *runtime.PodSandboxStatusRequest) (res *runtime.PodSandboxStatusResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("PodSandboxStatus for %q", r.GetPodSandboxId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("PodSandboxStatus for %q failed", r.GetPodSandboxId())
} else {
log.G(ctx).Tracef("PodSandboxStatus for %q returns status %+v", r.GetPodSandboxId(), res.GetStatus())
}
}()
res, err = in.c.PodSandboxStatus(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) StopPodSandbox(ctx context.Context, r *runtime.StopPodSandboxRequest) (_ *runtime.StopPodSandboxResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("StopPodSandbox for %q", r.GetPodSandboxId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("StopPodSandbox for %q failed", r.GetPodSandboxId())
} else {
log.G(ctx).Infof("StopPodSandbox for %q returns successfully", r.GetPodSandboxId())
}
}()
res, err := in.c.StopPodSandbox(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) RemovePodSandbox(ctx context.Context, r *runtime.RemovePodSandboxRequest) (_ *runtime.RemovePodSandboxResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("RemovePodSandbox for %q", r.GetPodSandboxId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("RemovePodSandbox for %q failed", r.GetPodSandboxId())
} else {
log.G(ctx).Infof("RemovePodSandbox %q returns successfully", r.GetPodSandboxId())
}
}()
res, err := in.c.RemovePodSandbox(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) PortForward(ctx context.Context, r *runtime.PortForwardRequest) (res *runtime.PortForwardResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("Portforward for %q port %v", r.GetPodSandboxId(), r.GetPort())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("Portforward for %q failed", r.GetPodSandboxId())
} else {
log.G(ctx).Infof("Portforward for %q returns URL %q", r.GetPodSandboxId(), res.GetUrl())
}
}()
res, err = in.c.PortForward(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) CreateContainer(ctx context.Context, r *runtime.CreateContainerRequest) (res *runtime.CreateContainerResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("CreateContainer within sandbox %q for container %+v",
r.GetPodSandboxId(), r.GetConfig().GetMetadata())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("CreateContainer within sandbox %q for %+v failed",
r.GetPodSandboxId(), r.GetConfig().GetMetadata())
} else {
log.G(ctx).Infof("CreateContainer within sandbox %q for %+v returns container id %q",
r.GetPodSandboxId(), r.GetConfig().GetMetadata(), res.GetContainerId())
}
}()
res, err = in.c.CreateContainer(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) StartContainer(ctx context.Context, r *runtime.StartContainerRequest) (_ *runtime.StartContainerResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("StartContainer for %q", r.GetContainerId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("StartContainer for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("StartContainer for %q returns successfully", r.GetContainerId())
}
}()
res, err := in.c.StartContainer(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ListContainers(ctx context.Context, r *runtime.ListContainersRequest) (res *runtime.ListContainersResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ListContainers with filter %+v", r.GetFilter())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ListContainers with filter %+v failed", r.GetFilter())
} else {
log.G(ctx).Tracef("ListContainers with filter %+v returns containers %+v",
r.GetFilter(), res.GetContainers())
}
}()
res, err = in.c.ListContainers(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ContainerStatus(ctx context.Context, r *runtime.ContainerStatusRequest) (res *runtime.ContainerStatusResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ContainerStatus for %q", r.GetContainerId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ContainerStatus for %q failed", r.GetContainerId())
} else {
log.G(ctx).Tracef("ContainerStatus for %q returns status %+v", r.GetContainerId(), res.GetStatus())
}
}()
res, err = in.c.ContainerStatus(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) StopContainer(ctx context.Context, r *runtime.StopContainerRequest) (res *runtime.StopContainerResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("StopContainer for %q with timeout %d (s)", r.GetContainerId(), r.GetTimeout())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("StopContainer for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("StopContainer for %q returns successfully", r.GetContainerId())
}
}()
res, err = in.c.StopContainer(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) RemoveContainer(ctx context.Context, r *runtime.RemoveContainerRequest) (res *runtime.RemoveContainerResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("RemoveContainer for %q", r.GetContainerId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("RemoveContainer for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("RemoveContainer for %q returns successfully", r.GetContainerId())
}
}()
res, err = in.c.RemoveContainer(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ExecSync(ctx context.Context, r *runtime.ExecSyncRequest) (res *runtime.ExecSyncResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("ExecSync for %q with command %+v and timeout %d (s)", r.GetContainerId(), r.GetCmd(), r.GetTimeout())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ExecSync for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("ExecSync for %q returns with exit code %d", r.GetContainerId(), res.GetExitCode())
log.G(ctx).Debugf("ExecSync for %q outputs - stdout: %q, stderr: %q", r.GetContainerId(),
res.GetStdout(), res.GetStderr())
}
}()
res, err = in.c.ExecSync(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) Exec(ctx context.Context, r *runtime.ExecRequest) (res *runtime.ExecResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("Exec for %q with command %+v, tty %v and stdin %v",
r.GetContainerId(), r.GetCmd(), r.GetTty(), r.GetStdin())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("Exec for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("Exec for %q returns URL %q", r.GetContainerId(), res.GetUrl())
}
}()
res, err = in.c.Exec(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) Attach(ctx context.Context, r *runtime.AttachRequest) (res *runtime.AttachResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("Attach for %q with tty %v and stdin %v", r.GetContainerId(), r.GetTty(), r.GetStdin())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("Attach for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("Attach for %q returns URL %q", r.GetContainerId(), res.Url)
}
}()
res, err = in.c.Attach(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) UpdateContainerResources(ctx context.Context, r *runtime.UpdateContainerResourcesRequest) (res *runtime.UpdateContainerResourcesResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("UpdateContainerResources for %q with %+v", r.GetContainerId(), r.GetLinux())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("UpdateContainerResources for %q failed", r.GetContainerId())
} else {
log.G(ctx).Infof("UpdateContainerResources for %q returns successfully", r.GetContainerId())
}
}()
res, err = in.c.UpdateContainerResources(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) PullImage(ctx context.Context, r *runtime.PullImageRequest) (res *runtime.PullImageResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("PullImage %q", r.GetImage().GetImage())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("PullImage %q failed", r.GetImage().GetImage())
} else {
log.G(ctx).Infof("PullImage %q returns image reference %q",
r.GetImage().GetImage(), res.GetImageRef())
}
}()
res, err = in.c.PullImage(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ListImages(ctx context.Context, r *runtime.ListImagesRequest) (res *runtime.ListImagesResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ListImages with filter %+v", r.GetFilter())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ListImages with filter %+v failed", r.GetFilter())
} else {
log.G(ctx).Tracef("ListImages with filter %+v returns image list %+v",
r.GetFilter(), res.GetImages())
}
}()
res, err = in.c.ListImages(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ImageStatus(ctx context.Context, r *runtime.ImageStatusRequest) (res *runtime.ImageStatusResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ImageStatus for %q", r.GetImage().GetImage())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ImageStatus for %q failed", r.GetImage().GetImage())
} else {
log.G(ctx).Tracef("ImageStatus for %q returns image status %+v",
r.GetImage().GetImage(), res.GetImage())
}
}()
res, err = in.c.ImageStatus(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) RemoveImage(ctx context.Context, r *runtime.RemoveImageRequest) (_ *runtime.RemoveImageResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Infof("RemoveImage %q", r.GetImage().GetImage())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("RemoveImage %q failed", r.GetImage().GetImage())
} else {
log.G(ctx).Infof("RemoveImage %q returns successfully", r.GetImage().GetImage())
}
}()
res, err := in.c.RemoveImage(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ImageFsInfo(ctx context.Context, r *runtime.ImageFsInfoRequest) (res *runtime.ImageFsInfoResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Debugf("ImageFsInfo")
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("ImageFsInfo failed")
} else {
log.G(ctx).Debugf("ImageFsInfo returns filesystem info %+v", res.ImageFilesystems)
}
}()
res, err = in.c.ImageFsInfo(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ContainerStats(ctx context.Context, r *runtime.ContainerStatsRequest) (res *runtime.ContainerStatsResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Debugf("ContainerStats for %q", r.GetContainerId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ContainerStats for %q failed", r.GetContainerId())
} else {
log.G(ctx).Debugf("ContainerStats for %q returns stats %+v", r.GetContainerId(), res.GetStats())
}
}()
res, err = in.c.ContainerStats(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ListContainerStats(ctx context.Context, r *runtime.ListContainerStatsRequest) (res *runtime.ListContainerStatsResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("ListContainerStats with filter %+v", r.GetFilter())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("ListContainerStats failed")
} else {
log.G(ctx).Tracef("ListContainerStats returns stats %+v", res.GetStats())
}
}()
res, err = in.c.ListContainerStats(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) Status(ctx context.Context, r *runtime.StatusRequest) (res *runtime.StatusResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("Status")
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("Status failed")
} else {
log.G(ctx).Tracef("Status returns status %+v", res.GetStatus())
}
}()
res, err = in.c.Status(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) Version(ctx context.Context, r *runtime.VersionRequest) (res *runtime.VersionResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Tracef("Version with client side version %q", r.GetVersion())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("Version failed")
} else {
log.G(ctx).Tracef("Version returns %+v", res)
}
}()
res, err = in.c.Version(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) UpdateRuntimeConfig(ctx context.Context, r *runtime.UpdateRuntimeConfigRequest) (res *runtime.UpdateRuntimeConfigResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Debugf("UpdateRuntimeConfig with config %+v", r.GetRuntimeConfig())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Error("UpdateRuntimeConfig failed")
} else {
log.G(ctx).Debug("UpdateRuntimeConfig returns returns successfully")
}
}()
res, err = in.c.UpdateRuntimeConfig(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}
func (in *instrumentedService) ReopenContainerLog(ctx context.Context, r *runtime.ReopenContainerLogRequest) (res *runtime.ReopenContainerLogResponse, err error) {
if err := in.checkInitialized(); err != nil {
return nil, err
}
log.G(ctx).Debugf("ReopenContainerLog for %q", r.GetContainerId())
defer func() {
if err != nil {
log.G(ctx).WithError(err).Errorf("ReopenContainerLog for %q failed", r.GetContainerId())
} else {
log.G(ctx).Debugf("ReopenContainerLog for %q returns successfully", r.GetContainerId())
}
}()
res, err = in.c.ReopenContainerLog(ctrdutil.WithNamespace(ctx), r)
return res, errdefs.ToGRPC(err)
}

View File

@@ -1,236 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"errors"
"io"
"strings"
"sync"
"github.com/containerd/containerd/cio"
"github.com/sirupsen/logrus"
cioutil "github.com/containerd/cri/pkg/ioutil"
"github.com/containerd/cri/pkg/util"
)
// streamKey generates a key for the stream.
func streamKey(id, name string, stream StreamType) string {
return strings.Join([]string{id, name, string(stream)}, "-")
}
// ContainerIO holds the container io.
type ContainerIO struct {
id string
fifos *cio.FIFOSet
*stdioPipes
stdoutGroup *cioutil.WriterGroup
stderrGroup *cioutil.WriterGroup
closer *wgCloser
}
var _ cio.IO = &ContainerIO{}
// ContainerIOOpts sets specific information to newly created ContainerIO.
type ContainerIOOpts func(*ContainerIO) error
// WithFIFOs specifies existing fifos for the container io.
func WithFIFOs(fifos *cio.FIFOSet) ContainerIOOpts {
return func(c *ContainerIO) error {
c.fifos = fifos
return nil
}
}
// WithNewFIFOs creates new fifos for the container io.
func WithNewFIFOs(root string, tty, stdin bool) ContainerIOOpts {
return func(c *ContainerIO) error {
fifos, err := newFifos(root, c.id, tty, stdin)
if err != nil {
return err
}
return WithFIFOs(fifos)(c)
}
}
// NewContainerIO creates container io.
func NewContainerIO(id string, opts ...ContainerIOOpts) (_ *ContainerIO, err error) {
c := &ContainerIO{
id: id,
stdoutGroup: cioutil.NewWriterGroup(),
stderrGroup: cioutil.NewWriterGroup(),
}
for _, opt := range opts {
if err := opt(c); err != nil {
return nil, err
}
}
if c.fifos == nil {
return nil, errors.New("fifos are not set")
}
// Create actual fifos.
stdio, closer, err := newStdioPipes(c.fifos)
if err != nil {
return nil, err
}
c.stdioPipes = stdio
c.closer = closer
return c, nil
}
// Config returns io config.
func (c *ContainerIO) Config() cio.Config {
return c.fifos.Config
}
// Pipe creates container fifos and pipe container output
// to output stream.
func (c *ContainerIO) Pipe() {
wg := c.closer.wg
if c.stdout != nil {
wg.Add(1)
go func() {
if _, err := io.Copy(c.stdoutGroup, c.stdout); err != nil {
logrus.WithError(err).Errorf("Failed to pipe stdout of container %q", c.id)
}
c.stdout.Close()
c.stdoutGroup.Close()
wg.Done()
logrus.Infof("Finish piping stdout of container %q", c.id)
}()
}
if !c.fifos.Terminal && c.stderr != nil {
wg.Add(1)
go func() {
if _, err := io.Copy(c.stderrGroup, c.stderr); err != nil {
logrus.WithError(err).Errorf("Failed to pipe stderr of container %q", c.id)
}
c.stderr.Close()
c.stderrGroup.Close()
wg.Done()
logrus.Infof("Finish piping stderr of container %q", c.id)
}()
}
}
// Attach attaches container stdio.
// TODO(random-liu): Use pools.Copy in docker to reduce memory usage?
func (c *ContainerIO) Attach(opts AttachOptions) {
var wg sync.WaitGroup
key := util.GenerateID()
stdinKey := streamKey(c.id, "attach-"+key, Stdin)
stdoutKey := streamKey(c.id, "attach-"+key, Stdout)
stderrKey := streamKey(c.id, "attach-"+key, Stderr)
var stdinStreamRC io.ReadCloser
if c.stdin != nil && opts.Stdin != nil {
// Create a wrapper of stdin which could be closed. Note that the
// wrapper doesn't close the actual stdin, it only stops io.Copy.
// The actual stdin will be closed by stream server.
stdinStreamRC = cioutil.NewWrapReadCloser(opts.Stdin)
wg.Add(1)
go func() {
if _, err := io.Copy(c.stdin, stdinStreamRC); err != nil {
logrus.WithError(err).Errorf("Failed to pipe stdin for container attach %q", c.id)
}
logrus.Infof("Attach stream %q closed", stdinKey)
if opts.StdinOnce && !opts.Tty {
// Due to kubectl requirements and current docker behavior, when (opts.StdinOnce &&
// opts.Tty) we have to close container stdin and keep stdout and stderr open until
// container stops.
c.stdin.Close()
// Also closes the containerd side.
if err := opts.CloseStdin(); err != nil {
logrus.WithError(err).Errorf("Failed to close stdin for container %q", c.id)
}
} else {
if opts.Stdout != nil {
c.stdoutGroup.Remove(stdoutKey)
}
if opts.Stderr != nil {
c.stderrGroup.Remove(stderrKey)
}
}
wg.Done()
}()
}
attachStream := func(key string, close <-chan struct{}) {
<-close
logrus.Infof("Attach stream %q closed", key)
// Make sure stdin gets closed.
if stdinStreamRC != nil {
stdinStreamRC.Close()
}
wg.Done()
}
if opts.Stdout != nil {
wg.Add(1)
wc, close := cioutil.NewWriteCloseInformer(opts.Stdout)
c.stdoutGroup.Add(stdoutKey, wc)
go attachStream(stdoutKey, close)
}
if !opts.Tty && opts.Stderr != nil {
wg.Add(1)
wc, close := cioutil.NewWriteCloseInformer(opts.Stderr)
c.stderrGroup.Add(stderrKey, wc)
go attachStream(stderrKey, close)
}
wg.Wait()
}
// AddOutput adds new write closers to the container stream, and returns existing
// write closers if there are any.
func (c *ContainerIO) AddOutput(name string, stdout, stderr io.WriteCloser) (io.WriteCloser, io.WriteCloser) {
var oldStdout, oldStderr io.WriteCloser
if stdout != nil {
key := streamKey(c.id, name, Stdout)
oldStdout = c.stdoutGroup.Get(key)
c.stdoutGroup.Add(key, stdout)
}
if stderr != nil {
key := streamKey(c.id, name, Stderr)
oldStderr = c.stderrGroup.Get(key)
c.stderrGroup.Add(key, stderr)
}
return oldStdout, oldStderr
}
// Cancel cancels container io.
func (c *ContainerIO) Cancel() {
c.closer.Cancel()
}
// Wait waits container io to finish.
func (c *ContainerIO) Wait() {
c.closer.Wait()
}
// Close closes all FIFOs.
func (c *ContainerIO) Close() error {
c.closer.Close()
if c.fifos != nil {
return c.fifos.Close()
}
return nil
}

View File

@@ -1,146 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"io"
"sync"
"github.com/containerd/containerd/cio"
"github.com/sirupsen/logrus"
cioutil "github.com/containerd/cri/pkg/ioutil"
)
// ExecIO holds the exec io.
type ExecIO struct {
id string
fifos *cio.FIFOSet
*stdioPipes
closer *wgCloser
}
var _ cio.IO = &ExecIO{}
// NewExecIO creates exec io.
func NewExecIO(id, root string, tty, stdin bool) (*ExecIO, error) {
fifos, err := newFifos(root, id, tty, stdin)
if err != nil {
return nil, err
}
stdio, closer, err := newStdioPipes(fifos)
if err != nil {
return nil, err
}
return &ExecIO{
id: id,
fifos: fifos,
stdioPipes: stdio,
closer: closer,
}, nil
}
// Config returns io config.
func (e *ExecIO) Config() cio.Config {
return e.fifos.Config
}
// Attach attaches exec stdio. The logic is similar with container io attach.
func (e *ExecIO) Attach(opts AttachOptions) <-chan struct{} {
var wg sync.WaitGroup
var stdinStreamRC io.ReadCloser
if e.stdin != nil && opts.Stdin != nil {
stdinStreamRC = cioutil.NewWrapReadCloser(opts.Stdin)
wg.Add(1)
go func() {
if _, err := io.Copy(e.stdin, stdinStreamRC); err != nil {
logrus.WithError(err).Errorf("Failed to redirect stdin for container exec %q", e.id)
}
logrus.Infof("Container exec %q stdin closed", e.id)
if opts.StdinOnce && !opts.Tty {
e.stdin.Close()
if err := opts.CloseStdin(); err != nil {
logrus.WithError(err).Errorf("Failed to close stdin for container exec %q", e.id)
}
} else {
if e.stdout != nil {
e.stdout.Close()
}
if e.stderr != nil {
e.stderr.Close()
}
}
wg.Done()
}()
}
attachOutput := func(t StreamType, stream io.WriteCloser, out io.ReadCloser) {
if _, err := io.Copy(stream, out); err != nil {
logrus.WithError(err).Errorf("Failed to pipe %q for container exec %q", t, e.id)
}
out.Close()
stream.Close()
if stdinStreamRC != nil {
stdinStreamRC.Close()
}
e.closer.wg.Done()
wg.Done()
logrus.Infof("Finish piping %q of container exec %q", t, e.id)
}
if opts.Stdout != nil {
wg.Add(1)
// Closer should wait for this routine to be over.
e.closer.wg.Add(1)
go attachOutput(Stdout, opts.Stdout, e.stdout)
}
if !opts.Tty && opts.Stderr != nil {
wg.Add(1)
// Closer should wait for this routine to be over.
e.closer.wg.Add(1)
go attachOutput(Stderr, opts.Stderr, e.stderr)
}
done := make(chan struct{})
go func() {
wg.Wait()
close(done)
}()
return done
}
// Cancel cancels exec io.
func (e *ExecIO) Cancel() {
e.closer.Cancel()
}
// Wait waits exec io to finish.
func (e *ExecIO) Wait() {
e.closer.Wait()
}
// Close closes all FIFOs.
func (e *ExecIO) Close() error {
if e.closer != nil {
e.closer.Close()
}
if e.fifos != nil {
return e.fifos.Close()
}
return nil
}

View File

@@ -1,144 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"io"
"os"
"path/filepath"
"sync"
"syscall"
"github.com/containerd/containerd/cio"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// AttachOptions specifies how to attach to a container.
type AttachOptions struct {
Stdin io.Reader
Stdout io.WriteCloser
Stderr io.WriteCloser
Tty bool
StdinOnce bool
// CloseStdin is the function to close container stdin.
CloseStdin func() error
}
// StreamType is the type of the stream, stdout/stderr.
type StreamType string
const (
// Stdin stream type.
Stdin StreamType = "stdin"
// Stdout stream type.
Stdout StreamType = StreamType(runtime.Stdout)
// Stderr stream type.
Stderr StreamType = StreamType(runtime.Stderr)
)
type wgCloser struct {
ctx context.Context
wg *sync.WaitGroup
set []io.Closer
cancel context.CancelFunc
}
func (g *wgCloser) Wait() {
g.wg.Wait()
}
func (g *wgCloser) Close() {
for _, f := range g.set {
f.Close()
}
}
func (g *wgCloser) Cancel() {
g.cancel()
}
// newFifos creates fifos directory for a container.
func newFifos(root, id string, tty, stdin bool) (*cio.FIFOSet, error) {
root = filepath.Join(root, "io")
if err := os.MkdirAll(root, 0700); err != nil {
return nil, err
}
fifos, err := cio.NewFIFOSetInDir(root, id, tty)
if err != nil {
return nil, err
}
if !stdin {
fifos.Stdin = ""
}
return fifos, nil
}
type stdioPipes struct {
stdin io.WriteCloser
stdout io.ReadCloser
stderr io.ReadCloser
}
// newStdioPipes creates actual fifos for stdio.
func newStdioPipes(fifos *cio.FIFOSet) (_ *stdioPipes, _ *wgCloser, err error) {
var (
f io.ReadWriteCloser
set []io.Closer
ctx, cancel = context.WithCancel(context.Background())
p = &stdioPipes{}
)
defer func() {
if err != nil {
for _, f := range set {
f.Close()
}
cancel()
}
}()
if fifos.Stdin != "" {
if f, err = openPipe(ctx, fifos.Stdin, syscall.O_WRONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
return nil, nil, err
}
p.stdin = f
set = append(set, f)
}
if fifos.Stdout != "" {
if f, err = openPipe(ctx, fifos.Stdout, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
return nil, nil, err
}
p.stdout = f
set = append(set, f)
}
if fifos.Stderr != "" {
if f, err = openPipe(ctx, fifos.Stderr, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
return nil, nil, err
}
p.stderr = f
set = append(set, f)
}
return p, &wgCloser{
wg: &sync.WaitGroup{},
set: set,
ctx: ctx,
cancel: cancel,
}, nil
}

View File

@@ -1,31 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"io"
"os"
"github.com/containerd/fifo"
"golang.org/x/net/context"
)
func openPipe(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
return fifo.OpenFifo(ctx, fn, flag, perm)
}

View File

@@ -1,81 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"io"
"net"
"os"
"sync"
winio "github.com/Microsoft/go-winio"
"github.com/pkg/errors"
"golang.org/x/net/context"
)
type pipe struct {
l net.Listener
con net.Conn
conErr error
conWg sync.WaitGroup
}
func openPipe(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
l, err := winio.ListenPipe(fn, nil)
if err != nil {
return nil, err
}
p := &pipe{l: l}
p.conWg.Add(1)
go func() {
defer p.conWg.Done()
c, err := l.Accept()
if err != nil {
p.conErr = err
return
}
p.con = c
}()
return p, nil
}
func (p *pipe) Write(b []byte) (int, error) {
p.conWg.Wait()
if p.conErr != nil {
return 0, errors.Wrap(p.conErr, "connection error")
}
return p.con.Write(b)
}
func (p *pipe) Read(b []byte) (int, error) {
p.conWg.Wait()
if p.conErr != nil {
return 0, errors.Wrap(p.conErr, "connection error")
}
return p.con.Read(b)
}
func (p *pipe) Close() error {
p.l.Close()
p.conWg.Wait()
if p.con != nil {
return p.con.Close()
}
return p.conErr
}

View File

@@ -1,196 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package io
import (
"bufio"
"bytes"
"fmt"
"io"
"io/ioutil"
"time"
"github.com/sirupsen/logrus"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
cioutil "github.com/containerd/cri/pkg/ioutil"
)
const (
// delimiter used in CRI logging format.
delimiter = ' '
// eof is end-of-line.
eol = '\n'
// timestampFormat is the timestamp format used in CRI logging format.
timestampFormat = time.RFC3339Nano
// defaultBufSize is the default size of the read buffer in bytes.
defaultBufSize = 4096
)
// NewDiscardLogger creates logger which discards all the input.
func NewDiscardLogger() io.WriteCloser {
return cioutil.NewNopWriteCloser(ioutil.Discard)
}
// NewCRILogger returns a write closer which redirect container log into
// log file, and decorate the log line into CRI defined format. It also
// returns a channel which indicates whether the logger is stopped.
// maxLen is the max length limit of a line. A line longer than the
// limit will be cut into multiple lines.
func NewCRILogger(path string, w io.Writer, stream StreamType, maxLen int) (io.WriteCloser, <-chan struct{}) {
logrus.Debugf("Start writing stream %q to log file %q", stream, path)
prc, pwc := io.Pipe()
stop := make(chan struct{})
go func() {
redirectLogs(path, prc, w, stream, maxLen)
close(stop)
}()
return pwc, stop
}
// bufio.ReadLine in golang eats both read errors and tailing newlines
// (See https://golang.org/pkg/bufio/#Reader.ReadLine). When reading
// to io.EOF, it is impossible for the caller to figure out whether
// there is a newline at the end, for example:
// 1) When reading "CONTENT\n", it returns "CONTENT" without error;
// 2) When reading "CONTENT", it also returns "CONTENT" without error.
//
// To differentiate these 2 cases, we need to write a readLine function
// ourselves to not ignore the error.
//
// The code is similar with https://golang.org/src/bufio/bufio.go?s=9537:9604#L359.
// The only difference is that it returns all errors from `ReadSlice`.
//
// readLine returns err != nil if and only if line does not end with a new line.
func readLine(b *bufio.Reader) (line []byte, isPrefix bool, err error) {
line, err = b.ReadSlice('\n')
if err == bufio.ErrBufferFull {
// Handle the case where "\r\n" straddles the buffer.
if len(line) > 0 && line[len(line)-1] == '\r' {
// Unread the last '\r'
if err := b.UnreadByte(); err != nil {
panic(fmt.Sprintf("invalid unread %v", err))
}
line = line[:len(line)-1]
}
return line, true, nil
}
if len(line) == 0 {
if err != nil {
line = nil
}
return
}
if line[len(line)-1] == '\n' {
// "ReadSlice returns err != nil if and only if line does not end in delim"
// (See https://golang.org/pkg/bufio/#Reader.ReadSlice).
if err != nil {
panic(fmt.Sprintf("full read with unexpected error %v", err))
}
drop := 1
if len(line) > 1 && line[len(line)-2] == '\r' {
drop = 2
}
line = line[:len(line)-drop]
}
return
}
func redirectLogs(path string, rc io.ReadCloser, w io.Writer, s StreamType, maxLen int) {
defer rc.Close()
var (
stream = []byte(s)
delimiter = []byte{delimiter}
partial = []byte(runtime.LogTagPartial)
full = []byte(runtime.LogTagFull)
buf [][]byte
length int
bufSize = defaultBufSize
)
// Make sure bufSize <= maxLen
if maxLen > 0 && maxLen < bufSize {
bufSize = maxLen
}
r := bufio.NewReaderSize(rc, bufSize)
writeLine := func(tag, line []byte) {
timestamp := time.Now().AppendFormat(nil, timestampFormat)
data := bytes.Join([][]byte{timestamp, stream, tag, line}, delimiter)
data = append(data, eol)
if _, err := w.Write(data); err != nil {
logrus.WithError(err).Errorf("Fail to write %q log to log file %q", s, path)
// Continue on write error to drain the container output.
}
}
for {
var stop bool
newLine, isPrefix, err := readLine(r)
// NOTE(random-liu): readLine can return actual content even if there is an error.
if len(newLine) > 0 {
// Buffer returned by ReadLine will change after
// next read, copy it.
l := make([]byte, len(newLine))
copy(l, newLine)
buf = append(buf, l)
length += len(l)
}
if err != nil {
if err == io.EOF {
logrus.Debugf("Getting EOF from stream %q while redirecting to log file %q", s, path)
} else {
logrus.WithError(err).Errorf("An error occurred when redirecting stream %q to log file %q", s, path)
}
if length == 0 {
// No content left to write, break.
break
}
// Stop after writing the content left in buffer.
stop = true
}
if maxLen > 0 && length > maxLen {
exceedLen := length - maxLen
last := buf[len(buf)-1]
if exceedLen > len(last) {
// exceedLen must <= len(last), or else the buffer
// should have be written in the previous iteration.
panic("exceed length should <= last buffer size")
}
buf[len(buf)-1] = last[:len(last)-exceedLen]
writeLine(partial, bytes.Join(buf, nil))
buf = [][]byte{last[len(last)-exceedLen:]}
length = exceedLen
}
if isPrefix {
continue
}
if stop {
// readLine only returns error when the message doesn't
// end with a newline, in that case it should be treated
// as a partial line.
writeLine(partial, bytes.Join(buf, nil))
} else {
writeLine(full, bytes.Join(buf, nil))
}
buf = nil
length = 0
if stop {
break
}
}
logrus.Debugf("Finish redirecting stream %q to log file %q", s, path)
}

View File

@@ -1,51 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"context"
"github.com/containerd/containerd"
"github.com/containerd/containerd/log"
"github.com/containerd/nri"
v1 "github.com/containerd/nri/types/v1"
)
// WithNRISandboxDelete calls delete for a sandbox'd task
func WithNRISandboxDelete(sandboxID string) containerd.ProcessDeleteOpts {
return func(ctx context.Context, p containerd.Process) error {
task, ok := p.(containerd.Task)
if !ok {
return nil
}
nric, err := nri.New()
if err != nil {
log.G(ctx).WithError(err).Error("unable to create nri client")
return nil
}
if nric == nil {
return nil
}
sb := &nri.Sandbox{
ID: sandboxID,
}
if _, err := nric.InvokeWithSandbox(ctx, task, v1.Delete, sb); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to delete nri for %q", task.ID())
}
return nil
}
}

View File

@@ -1,483 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"io/ioutil"
"os"
"path/filepath"
goruntime "runtime"
"time"
"github.com/containerd/containerd"
containerdio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/errdefs"
containerdimages "github.com/containerd/containerd/images"
"github.com/containerd/containerd/log"
"github.com/containerd/containerd/platforms"
"github.com/containerd/typeurl"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
"github.com/containerd/cri/pkg/netns"
cio "github.com/containerd/cri/pkg/server/io"
containerstore "github.com/containerd/cri/pkg/store/container"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// NOTE: The recovery logic has following assumption: when the cri plugin is down:
// 1) Files (e.g. root directory, netns) and checkpoint maintained by the plugin MUST NOT be
// touched. Or else, recovery logic for those containers/sandboxes may return error.
// 2) Containerd containers may be deleted, but SHOULD NOT be added. Or else, recovery logic
// for the newly added container/sandbox will return error, because there is no corresponding root
// directory created.
// 3) Containerd container tasks may exit or be stoppped, deleted. Even though current logic could
// tolerant tasks being created or started, we prefer that not to happen.
// recover recovers system state from containerd and status checkpoint.
func (c *criService) recover(ctx context.Context) error {
// Recover all sandboxes.
sandboxes, err := c.client.Containers(ctx, filterLabel(containerKindLabel, containerKindSandbox))
if err != nil {
return errors.Wrap(err, "failed to list sandbox containers")
}
for _, sandbox := range sandboxes {
sb, err := c.loadSandbox(ctx, sandbox)
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to load sandbox %q", sandbox.ID())
continue
}
log.G(ctx).Debugf("Loaded sandbox %+v", sb)
if err := c.sandboxStore.Add(sb); err != nil {
return errors.Wrapf(err, "failed to add sandbox %q to store", sandbox.ID())
}
if err := c.sandboxNameIndex.Reserve(sb.Name, sb.ID); err != nil {
return errors.Wrapf(err, "failed to reserve sandbox name %q", sb.Name)
}
}
// Recover all containers.
containers, err := c.client.Containers(ctx, filterLabel(containerKindLabel, containerKindContainer))
if err != nil {
return errors.Wrap(err, "failed to list containers")
}
for _, container := range containers {
cntr, err := c.loadContainer(ctx, container)
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to load container %q", container.ID())
continue
}
log.G(ctx).Debugf("Loaded container %+v", cntr)
if err := c.containerStore.Add(cntr); err != nil {
return errors.Wrapf(err, "failed to add container %q to store", container.ID())
}
if err := c.containerNameIndex.Reserve(cntr.Name, cntr.ID); err != nil {
return errors.Wrapf(err, "failed to reserve container name %q", cntr.Name)
}
}
// Recover all images.
cImages, err := c.client.ListImages(ctx)
if err != nil {
return errors.Wrap(err, "failed to list images")
}
c.loadImages(ctx, cImages)
// It's possible that containerd containers are deleted unexpectedly. In that case,
// we can't even get metadata, we should cleanup orphaned sandbox/container directories
// with best effort.
// Cleanup orphaned sandbox and container directories without corresponding containerd container.
for _, cleanup := range []struct {
cntrs []containerd.Container
base string
errMsg string
}{
{
cntrs: sandboxes,
base: filepath.Join(c.config.RootDir, sandboxesDir),
errMsg: "failed to cleanup orphaned sandbox directories",
},
{
cntrs: sandboxes,
base: filepath.Join(c.config.StateDir, sandboxesDir),
errMsg: "failed to cleanup orphaned volatile sandbox directories",
},
{
cntrs: containers,
base: filepath.Join(c.config.RootDir, containersDir),
errMsg: "failed to cleanup orphaned container directories",
},
{
cntrs: containers,
base: filepath.Join(c.config.StateDir, containersDir),
errMsg: "failed to cleanup orphaned volatile container directories",
},
} {
if err := cleanupOrphanedIDDirs(ctx, cleanup.cntrs, cleanup.base); err != nil {
return errors.Wrap(err, cleanup.errMsg)
}
}
return nil
}
// loadContainerTimeout is the default timeout for loading a container/sandbox.
// One container/sandbox hangs (e.g. containerd#2438) should not affect other
// containers/sandboxes.
// Most CRI container/sandbox related operations are per container, the ones
// which handle multiple containers at a time are:
// * ListPodSandboxes: Don't talk with containerd services.
// * ListContainers: Don't talk with containerd services.
// * ListContainerStats: Not in critical code path, a default timeout will
// be applied at CRI level.
// * Recovery logic: We should set a time for each container/sandbox recovery.
// * Event monitor: We should set a timeout for each container/sandbox event handling.
const loadContainerTimeout = 10 * time.Second
// loadContainer loads container from containerd and status checkpoint.
func (c *criService) loadContainer(ctx context.Context, cntr containerd.Container) (containerstore.Container, error) {
ctx, cancel := context.WithTimeout(ctx, loadContainerTimeout)
defer cancel()
id := cntr.ID()
containerDir := c.getContainerRootDir(id)
volatileContainerDir := c.getVolatileContainerRootDir(id)
var container containerstore.Container
// Load container metadata.
exts, err := cntr.Extensions(ctx)
if err != nil {
return container, errors.Wrap(err, "failed to get container extensions")
}
ext, ok := exts[containerMetadataExtension]
if !ok {
return container, errors.Errorf("metadata extension %q not found", containerMetadataExtension)
}
data, err := typeurl.UnmarshalAny(&ext)
if err != nil {
return container, errors.Wrapf(err, "failed to unmarshal metadata extension %q", ext)
}
meta := data.(*containerstore.Metadata)
// Load status from checkpoint.
status, err := containerstore.LoadStatus(containerDir, id)
if err != nil {
log.G(ctx).WithError(err).Warnf("Failed to load container status for %q", id)
status = unknownContainerStatus()
}
var containerIO *cio.ContainerIO
err = func() error {
// Load up-to-date status from containerd.
t, err := cntr.Task(ctx, func(fifos *containerdio.FIFOSet) (_ containerdio.IO, err error) {
stdoutWC, stderrWC, err := c.createContainerLoggers(meta.LogPath, meta.Config.GetTty())
if err != nil {
return nil, err
}
defer func() {
if err != nil {
if stdoutWC != nil {
stdoutWC.Close()
}
if stderrWC != nil {
stderrWC.Close()
}
}
}()
containerIO, err = cio.NewContainerIO(id,
cio.WithFIFOs(fifos),
)
if err != nil {
return nil, err
}
containerIO.AddOutput("log", stdoutWC, stderrWC)
containerIO.Pipe()
return containerIO, nil
})
if err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to load task")
}
var s containerd.Status
var notFound bool
if errdefs.IsNotFound(err) {
// Task is not found.
notFound = true
} else {
// Task is found. Get task status.
s, err = t.Status(ctx)
if err != nil {
// It's still possible that task is deleted during this window.
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to get task status")
}
notFound = true
}
}
if notFound {
// Task is not created or has been deleted, use the checkpointed status
// to generate container status.
switch status.State() {
case runtime.ContainerState_CONTAINER_CREATED:
// NOTE: Another possibility is that we've tried to start the container, but
// containerd got restarted during that. In that case, we still
// treat the container as `CREATED`.
containerIO, err = cio.NewContainerIO(id,
cio.WithNewFIFOs(volatileContainerDir, meta.Config.GetTty(), meta.Config.GetStdin()),
)
if err != nil {
return errors.Wrap(err, "failed to create container io")
}
case runtime.ContainerState_CONTAINER_RUNNING:
// Container was in running state, but its task has been deleted,
// set unknown exited state. Container io is not needed in this case.
status.FinishedAt = time.Now().UnixNano()
status.ExitCode = unknownExitCode
status.Reason = unknownExitReason
default:
// Container is in exited/unknown state, return the status as it is.
}
} else {
// Task status is found. Update container status based on the up-to-date task status.
switch s.Status {
case containerd.Created:
// Task has been created, but not started yet. This could only happen if containerd
// gets restarted during container start.
// Container must be in `CREATED` state.
if _, err := t.Delete(ctx, containerd.WithProcessKill); err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to delete task")
}
if status.State() != runtime.ContainerState_CONTAINER_CREATED {
return errors.Errorf("unexpected container state for created task: %q", status.State())
}
case containerd.Running:
// Task is running. Container must be in `RUNNING` state, based on our assuption that
// "task should not be started when containerd is down".
switch status.State() {
case runtime.ContainerState_CONTAINER_EXITED:
return errors.Errorf("unexpected container state for running task: %q", status.State())
case runtime.ContainerState_CONTAINER_RUNNING:
default:
// This may happen if containerd gets restarted after task is started, but
// before status is checkpointed.
status.StartedAt = time.Now().UnixNano()
status.Pid = t.Pid()
}
// Wait for the task for exit monitor.
// wait is a long running background request, no timeout needed.
exitCh, err := t.Wait(ctrdutil.NamespacedContext())
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to wait for task")
}
// Container was in running state, but its task has been deleted,
// set unknown exited state.
status.FinishedAt = time.Now().UnixNano()
status.ExitCode = unknownExitCode
status.Reason = unknownExitReason
} else {
// Start exit monitor.
c.eventMonitor.startExitMonitor(context.Background(), id, status.Pid, exitCh)
}
case containerd.Stopped:
// Task is stopped. Updata status and delete the task.
if _, err := t.Delete(ctx, containerd.WithProcessKill); err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to delete task")
}
status.FinishedAt = s.ExitTime.UnixNano()
status.ExitCode = int32(s.ExitStatus)
default:
return errors.Errorf("unexpected task status %q", s.Status)
}
}
return nil
}()
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to load container status for %q", id)
// Only set the unknown field in this case, because other fields may
// contain useful information loaded from the checkpoint.
status.Unknown = true
}
opts := []containerstore.Opts{
containerstore.WithStatus(status, containerDir),
containerstore.WithContainer(cntr),
}
// containerIO could be nil for container in unknown state.
if containerIO != nil {
opts = append(opts, containerstore.WithContainerIO(containerIO))
}
return containerstore.NewContainer(*meta, opts...)
}
// loadSandbox loads sandbox from containerd.
func (c *criService) loadSandbox(ctx context.Context, cntr containerd.Container) (sandboxstore.Sandbox, error) {
ctx, cancel := context.WithTimeout(ctx, loadContainerTimeout)
defer cancel()
var sandbox sandboxstore.Sandbox
// Load sandbox metadata.
exts, err := cntr.Extensions(ctx)
if err != nil {
return sandbox, errors.Wrap(err, "failed to get sandbox container extensions")
}
ext, ok := exts[sandboxMetadataExtension]
if !ok {
return sandbox, errors.Errorf("metadata extension %q not found", sandboxMetadataExtension)
}
data, err := typeurl.UnmarshalAny(&ext)
if err != nil {
return sandbox, errors.Wrapf(err, "failed to unmarshal metadata extension %q", ext)
}
meta := data.(*sandboxstore.Metadata)
s, err := func() (sandboxstore.Status, error) {
status := unknownSandboxStatus()
// Load sandbox created timestamp.
info, err := cntr.Info(ctx)
if err != nil {
return status, errors.Wrap(err, "failed to get sandbox container info")
}
status.CreatedAt = info.CreatedAt
// Load sandbox state.
t, err := cntr.Task(ctx, nil)
if err != nil && !errdefs.IsNotFound(err) {
return status, errors.Wrap(err, "failed to load task")
}
var taskStatus containerd.Status
var notFound bool
if errdefs.IsNotFound(err) {
// Task is not found.
notFound = true
} else {
// Task is found. Get task status.
taskStatus, err = t.Status(ctx)
if err != nil {
// It's still possible that task is deleted during this window.
if !errdefs.IsNotFound(err) {
return status, errors.Wrap(err, "failed to get task status")
}
notFound = true
}
}
if notFound {
// Task does not exist, set sandbox state as NOTREADY.
status.State = sandboxstore.StateNotReady
} else {
if taskStatus.Status == containerd.Running {
// Wait for the task for sandbox monitor.
// wait is a long running background request, no timeout needed.
exitCh, err := t.Wait(ctrdutil.NamespacedContext())
if err != nil {
if !errdefs.IsNotFound(err) {
return status, errors.Wrap(err, "failed to wait for task")
}
status.State = sandboxstore.StateNotReady
} else {
// Task is running, set sandbox state as READY.
status.State = sandboxstore.StateReady
status.Pid = t.Pid()
c.eventMonitor.startExitMonitor(context.Background(), meta.ID, status.Pid, exitCh)
}
} else {
// Task is not running. Delete the task and set sandbox state as NOTREADY.
if _, err := t.Delete(ctx, containerd.WithProcessKill); err != nil && !errdefs.IsNotFound(err) {
return status, errors.Wrap(err, "failed to delete task")
}
status.State = sandboxstore.StateNotReady
}
}
return status, nil
}()
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to load sandbox status for %q", cntr.ID())
}
sandbox = sandboxstore.NewSandbox(*meta, s)
sandbox.Container = cntr
// Load network namespace.
if goruntime.GOOS != "windows" &&
meta.Config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
// Don't need to load netns for host network sandbox.
return sandbox, nil
}
sandbox.NetNS = netns.LoadNetNS(meta.NetNSPath)
// It doesn't matter whether task is running or not. If it is running, sandbox
// status will be `READY`; if it is not running, sandbox status will be `NOT_READY`,
// kubelet will stop the sandbox which will properly cleanup everything.
return sandbox, nil
}
// loadImages loads images from containerd.
func (c *criService) loadImages(ctx context.Context, cImages []containerd.Image) {
snapshotter := c.config.ContainerdConfig.Snapshotter
for _, i := range cImages {
ok, _, _, _, err := containerdimages.Check(ctx, i.ContentStore(), i.Target(), platforms.Default())
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to check image content readiness for %q", i.Name())
continue
}
if !ok {
log.G(ctx).Warnf("The image content readiness for %q is not ok", i.Name())
continue
}
// Checking existence of top-level snapshot for each image being recovered.
unpacked, err := i.IsUnpacked(ctx, snapshotter)
if err != nil {
log.G(ctx).WithError(err).Warnf("Failed to check whether image is unpacked for image %s", i.Name())
continue
}
if !unpacked {
log.G(ctx).Warnf("The image %s is not unpacked.", i.Name())
// TODO(random-liu): Consider whether we should try unpack here.
}
if err := c.updateImage(ctx, i.Name()); err != nil {
log.G(ctx).WithError(err).Warnf("Failed to update reference for image %q", i.Name())
continue
}
log.G(ctx).Debugf("Loaded image %q", i.Name())
}
}
func cleanupOrphanedIDDirs(ctx context.Context, cntrs []containerd.Container, base string) error {
// Cleanup orphaned id directories.
dirs, err := ioutil.ReadDir(base)
if err != nil && !os.IsNotExist(err) {
return errors.Wrap(err, "failed to read base directory")
}
idsMap := make(map[string]containerd.Container)
for _, cntr := range cntrs {
idsMap[cntr.ID()] = cntr
}
for _, d := range dirs {
if !d.IsDir() {
log.G(ctx).Warnf("Invalid file %q found in base directory %q", d.Name(), base)
continue
}
if _, ok := idsMap[d.Name()]; ok {
// Do not remove id directory if corresponding container is found.
continue
}
dir := filepath.Join(base, d.Name())
if err := ensureRemoveAll(ctx, dir); err != nil {
log.G(ctx).WithError(err).Warnf("Failed to remove id directory %q", dir)
} else {
log.G(ctx).Debugf("Cleanup orphaned id directory %q", dir)
}
}
return nil
}

View File

@@ -1,101 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// ListPodSandbox returns a list of Sandbox.
func (c *criService) ListPodSandbox(ctx context.Context, r *runtime.ListPodSandboxRequest) (*runtime.ListPodSandboxResponse, error) {
// List all sandboxes from store.
sandboxesInStore := c.sandboxStore.List()
var sandboxes []*runtime.PodSandbox
for _, sandboxInStore := range sandboxesInStore {
sandboxes = append(sandboxes, toCRISandbox(
sandboxInStore.Metadata,
sandboxInStore.Status.Get(),
))
}
sandboxes = c.filterCRISandboxes(sandboxes, r.GetFilter())
return &runtime.ListPodSandboxResponse{Items: sandboxes}, nil
}
// toCRISandbox converts sandbox metadata into CRI pod sandbox.
func toCRISandbox(meta sandboxstore.Metadata, status sandboxstore.Status) *runtime.PodSandbox {
// Set sandbox state to NOTREADY by default.
state := runtime.PodSandboxState_SANDBOX_NOTREADY
if status.State == sandboxstore.StateReady {
state = runtime.PodSandboxState_SANDBOX_READY
}
return &runtime.PodSandbox{
Id: meta.ID,
Metadata: meta.Config.GetMetadata(),
State: state,
CreatedAt: status.CreatedAt.UnixNano(),
Labels: meta.Config.GetLabels(),
Annotations: meta.Config.GetAnnotations(),
RuntimeHandler: meta.RuntimeHandler,
}
}
func (c *criService) normalizePodSandboxFilter(filter *runtime.PodSandboxFilter) {
if sb, err := c.sandboxStore.Get(filter.GetId()); err == nil {
filter.Id = sb.ID
}
}
// filterCRISandboxes filters CRISandboxes.
func (c *criService) filterCRISandboxes(sandboxes []*runtime.PodSandbox, filter *runtime.PodSandboxFilter) []*runtime.PodSandbox {
if filter == nil {
return sandboxes
}
c.normalizePodSandboxFilter(filter)
filtered := []*runtime.PodSandbox{}
for _, s := range sandboxes {
// Filter by id
if filter.GetId() != "" && filter.GetId() != s.Id {
continue
}
// Filter by state
if filter.GetState() != nil && filter.GetState().GetState() != s.State {
continue
}
// Filter by label
if filter.GetLabelSelector() != nil {
match := true
for k, v := range filter.GetLabelSelector() {
got, ok := s.Labels[k]
if !ok || got != v {
match = false
break
}
}
if !match {
continue
}
}
filtered = append(filtered, s)
}
return filtered
}

View File

@@ -1,38 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// PortForward prepares a streaming endpoint to forward ports from a PodSandbox, and returns the address.
func (c *criService) PortForward(ctx context.Context, r *runtime.PortForwardRequest) (retRes *runtime.PortForwardResponse, retErr error) {
sandbox, err := c.sandboxStore.Get(r.GetPodSandboxId())
if err != nil {
return nil, errors.Wrapf(err, "failed to find sandbox %q", r.GetPodSandboxId())
}
if sandbox.Status.Get().State != sandboxstore.StateReady {
return nil, errors.New("sandbox container is not running")
}
// TODO(random-liu): Verify that ports are exposed.
return c.streamServer.GetPortForward(r)
}

View File

@@ -1,128 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"fmt"
"io"
"net"
"time"
"github.com/containerd/containerd/log"
"github.com/containernetworking/plugins/pkg/ns"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// portForward uses netns to enter the sandbox namespace, and forwards a stream inside the
// the namespace to a specific port. It keeps forwarding until it exits or client disconnect.
func (c *criService) portForward(ctx context.Context, id string, port int32, stream io.ReadWriteCloser) error {
s, err := c.sandboxStore.Get(id)
if err != nil {
return errors.Wrapf(err, "failed to find sandbox %q in store", id)
}
var netNSDo func(func(ns.NetNS) error) error
// netNSPath is the network namespace path for logging.
var netNSPath string
securityContext := s.Config.GetLinux().GetSecurityContext()
hostNet := securityContext.GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE
if !hostNet {
if closed, err := s.NetNS.Closed(); err != nil {
return errors.Wrapf(err, "failed to check netwok namespace closed for sandbox %q", id)
} else if closed {
return errors.Errorf("network namespace for sandbox %q is closed", id)
}
netNSDo = s.NetNS.Do
netNSPath = s.NetNS.GetPath()
} else {
// Run the function directly for host network.
netNSDo = func(do func(_ ns.NetNS) error) error {
return do(nil)
}
netNSPath = "host"
}
log.G(ctx).Infof("Executing port forwarding in network namespace %q", netNSPath)
err = netNSDo(func(_ ns.NetNS) error {
defer stream.Close()
// TODO: hardcoded to tcp4 because localhost resolves to ::1 by default if the system has IPv6 enabled.
// Theoretically happy eyeballs will try IPv6 first and fallback to IPv4
// but resolving localhost doesn't seem to return and IPv4 address, thus failing the connection.
conn, err := net.Dial("tcp4", fmt.Sprintf("localhost:%d", port))
if err != nil {
return errors.Wrapf(err, "failed to dial %d", port)
}
defer conn.Close()
errCh := make(chan error, 2)
// Copy from the the namespace port connection to the client stream
go func() {
log.G(ctx).Debugf("PortForward copying data from namespace %q port %d to the client stream", id, port)
_, err := io.Copy(stream, conn)
errCh <- err
}()
// Copy from the client stream to the namespace port connection
go func() {
log.G(ctx).Debugf("PortForward copying data from client stream to namespace %q port %d", id, port)
_, err := io.Copy(conn, stream)
errCh <- err
}()
// Wait until the first error is returned by one of the connections
// we use errFwd to store the result of the port forwarding operation
// if the context is cancelled close everything and return
var errFwd error
select {
case errFwd = <-errCh:
log.G(ctx).Debugf("PortForward stop forwarding in one direction in network namespace %q port %d: %v", id, port, errFwd)
case <-ctx.Done():
log.G(ctx).Debugf("PortForward cancelled in network namespace %q port %d: %v", id, port, ctx.Err())
return ctx.Err()
}
// give a chance to terminate gracefully or timeout
// after 1s
// https://linux.die.net/man/1/socat
const timeout = time.Second
select {
case e := <-errCh:
if errFwd == nil {
errFwd = e
}
log.G(ctx).Debugf("PortForward stopped forwarding in both directions in network namespace %q port %d: %v", id, port, e)
case <-time.After(timeout):
log.G(ctx).Debugf("PortForward timed out waiting to close the connection in network namespace %q port %d", id, port)
case <-ctx.Done():
log.G(ctx).Debugf("PortForward cancelled in network namespace %q port %d: %v", id, port, ctx.Err())
errFwd = ctx.Err()
}
return errFwd
})
if err != nil {
return errors.Wrapf(err, "failed to execute portforward in network namespace %q", netNSPath)
}
log.G(ctx).Infof("Finish port forwarding for %q port %d", id, port)
return nil
}

View File

@@ -1,80 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"bytes"
"fmt"
"io"
"github.com/pkg/errors"
"golang.org/x/net/context"
"k8s.io/utils/exec"
"github.com/containerd/cri/pkg/ioutil"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
func (c *criService) portForward(ctx context.Context, id string, port int32, stream io.ReadWriter) error {
stdout := ioutil.NewNopWriteCloser(stream)
stderrBuffer := new(bytes.Buffer)
stderr := ioutil.NewNopWriteCloser(stderrBuffer)
// localhost is resolved to 127.0.0.1 in ipv4, and ::1 in ipv6.
// Explicitly using ipv4 IP address in here to avoid flakiness.
cmd := []string{"wincat.exe", "127.0.0.1", fmt.Sprint(port)}
err := c.execInSandbox(ctx, id, cmd, stream, stdout, stderr)
if err != nil {
return errors.Wrapf(err, "failed to execute port forward in sandbox: %s", stderrBuffer.String())
}
return nil
}
func (c *criService) execInSandbox(ctx context.Context, sandboxID string, cmd []string, stdin io.Reader, stdout, stderr io.WriteCloser) error {
// Get sandbox from our sandbox store.
sb, err := c.sandboxStore.Get(sandboxID)
if err != nil {
return errors.Wrapf(err, "failed to find sandbox %q in store", sandboxID)
}
// Check the sandbox state
state := sb.Status.Get().State
if state != sandboxstore.StateReady {
return errors.Errorf("sandbox is in %s state", fmt.Sprint(state))
}
opts := execOptions{
cmd: cmd,
stdin: stdin,
stdout: stdout,
stderr: stderr,
tty: false,
resize: nil,
}
exitCode, err := c.execInternal(ctx, sb.Container, sandboxID, opts)
if err != nil {
return errors.Wrap(err, "failed to exec in sandbox")
}
if *exitCode == 0 {
return nil
}
return &exec.CodeExitError{
Err: errors.Errorf("error executing command %v, exit code %d", cmd, *exitCode),
Code: int(*exitCode),
}
}

View File

@@ -1,115 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// RemovePodSandbox removes the sandbox. If there are running containers in the
// sandbox, they should be forcibly removed.
func (c *criService) RemovePodSandbox(ctx context.Context, r *runtime.RemovePodSandboxRequest) (*runtime.RemovePodSandboxResponse, error) {
sandbox, err := c.sandboxStore.Get(r.GetPodSandboxId())
if err != nil {
if err != store.ErrNotExist {
return nil, errors.Wrapf(err, "an error occurred when try to find sandbox %q",
r.GetPodSandboxId())
}
// Do not return error if the id doesn't exist.
log.G(ctx).Tracef("RemovePodSandbox called for sandbox %q that does not exist",
r.GetPodSandboxId())
return &runtime.RemovePodSandboxResponse{}, nil
}
// Use the full sandbox id.
id := sandbox.ID
// Return error if sandbox container is still running or unknown.
state := sandbox.Status.Get().State
if state == sandboxstore.StateReady || state == sandboxstore.StateUnknown {
logrus.Infof("Forcibly stopping sandbox %q", id)
if err := c.stopPodSandbox(ctx, sandbox); err != nil {
return nil, errors.Wrapf(err, "failed to forcibly stop sandbox %q", id)
}
}
// Return error if sandbox network namespace is not closed yet.
if sandbox.NetNS != nil {
nsPath := sandbox.NetNS.GetPath()
if closed, err := sandbox.NetNS.Closed(); err != nil {
return nil, errors.Wrapf(err, "failed to check sandbox network namespace %q closed", nsPath)
} else if !closed {
return nil, errors.Errorf("sandbox network namespace %q is not fully closed", nsPath)
}
}
// Remove all containers inside the sandbox.
// NOTE(random-liu): container could still be created after this point, Kubelet should
// not rely on this behavior.
// TODO(random-liu): Introduce an intermediate state to avoid container creation after
// this point.
cntrs := c.containerStore.List()
for _, cntr := range cntrs {
if cntr.SandboxID != id {
continue
}
_, err = c.RemoveContainer(ctx, &runtime.RemoveContainerRequest{ContainerId: cntr.ID})
if err != nil {
return nil, errors.Wrapf(err, "failed to remove container %q", cntr.ID)
}
}
// Cleanup the sandbox root directories.
sandboxRootDir := c.getSandboxRootDir(id)
if err := ensureRemoveAll(ctx, sandboxRootDir); err != nil {
return nil, errors.Wrapf(err, "failed to remove sandbox root directory %q",
sandboxRootDir)
}
volatileSandboxRootDir := c.getVolatileSandboxRootDir(id)
if err := ensureRemoveAll(ctx, volatileSandboxRootDir); err != nil {
return nil, errors.Wrapf(err, "failed to remove volatile sandbox root directory %q",
volatileSandboxRootDir)
}
// Delete sandbox container.
if err := sandbox.Container.Delete(ctx, containerd.WithSnapshotCleanup); err != nil {
if !errdefs.IsNotFound(err) {
return nil, errors.Wrapf(err, "failed to delete sandbox container %q", id)
}
log.G(ctx).Tracef("Remove called for sandbox container %q that does not exist", id)
}
// Remove sandbox from sandbox store. Note that once the sandbox is successfully
// deleted:
// 1) ListPodSandbox will not include this sandbox.
// 2) PodSandboxStatus and StopPodSandbox will return error.
// 3) On-going operations which have held the reference will not be affected.
c.sandboxStore.Delete(id)
// Release the sandbox name reserved for the sandbox.
c.sandboxNameIndex.ReleaseByKey(id)
return &runtime.RemovePodSandboxResponse{}, nil
}

View File

@@ -1,548 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
"math"
goruntime "runtime"
"strings"
"github.com/containerd/containerd"
containerdio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
cni "github.com/containerd/go-cni"
"github.com/containerd/nri"
v1 "github.com/containerd/nri/types/v1"
"github.com/containerd/typeurl"
"github.com/davecgh/go-spew/spew"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/annotations"
criconfig "github.com/containerd/cri/pkg/config"
customopts "github.com/containerd/cri/pkg/containerd/opts"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
"github.com/containerd/cri/pkg/netns"
"github.com/containerd/cri/pkg/server/bandwidth"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
"github.com/containerd/cri/pkg/util"
selinux "github.com/opencontainers/selinux/go-selinux"
)
func init() {
typeurl.Register(&sandboxstore.Metadata{},
"github.com/containerd/cri/pkg/store/sandbox", "Metadata")
}
// RunPodSandbox creates and starts a pod-level sandbox. Runtimes should ensure
// the sandbox is in ready state.
func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandboxRequest) (_ *runtime.RunPodSandboxResponse, retErr error) {
config := r.GetConfig()
log.G(ctx).Debugf("Sandbox config %+v", config)
// Generate unique id and name for the sandbox and reserve the name.
id := util.GenerateID()
metadata := config.GetMetadata()
if metadata == nil {
return nil, errors.New("sandbox config must include metadata")
}
name := makeSandboxName(metadata)
log.G(ctx).Debugf("Generated id %q for sandbox %q", id, name)
// Reserve the sandbox name to avoid concurrent `RunPodSandbox` request starting the
// same sandbox.
if err := c.sandboxNameIndex.Reserve(name, id); err != nil {
return nil, errors.Wrapf(err, "failed to reserve sandbox name %q", name)
}
defer func() {
// Release the name if the function returns with an error.
if retErr != nil {
c.sandboxNameIndex.ReleaseByName(name)
}
}()
// Create initial internal sandbox object.
sandbox := sandboxstore.NewSandbox(
sandboxstore.Metadata{
ID: id,
Name: name,
Config: config,
RuntimeHandler: r.GetRuntimeHandler(),
},
sandboxstore.Status{
State: sandboxstore.StateUnknown,
},
)
// Ensure sandbox container image snapshot.
image, err := c.ensureImageExists(ctx, c.config.SandboxImage, config)
if err != nil {
return nil, errors.Wrapf(err, "failed to get sandbox image %q", c.config.SandboxImage)
}
containerdImage, err := c.toContainerdImage(ctx, *image)
if err != nil {
return nil, errors.Wrapf(err, "failed to get image from containerd %q", image.ID)
}
ociRuntime, err := c.getSandboxRuntime(config, r.GetRuntimeHandler())
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox runtime")
}
log.G(ctx).Debugf("Use OCI %+v for sandbox %q", ociRuntime, id)
podNetwork := true
// Pod network is always needed on windows.
if goruntime.GOOS != "windows" &&
config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
// Pod network is not needed on linux with host network.
podNetwork = false
}
if podNetwork {
// If it is not in host network namespace then create a namespace and set the sandbox
// handle. NetNSPath in sandbox metadata and NetNS is non empty only for non host network
// namespaces. If the pod is in host network namespace then both are empty and should not
// be used.
sandbox.NetNS, err = netns.NewNetNS()
if err != nil {
return nil, errors.Wrapf(err, "failed to create network namespace for sandbox %q", id)
}
sandbox.NetNSPath = sandbox.NetNS.GetPath()
defer func() {
if retErr != nil {
// Teardown network if an error is returned.
if err := c.teardownPodNetwork(ctx, sandbox); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to destroy network for sandbox %q", id)
}
if err := sandbox.NetNS.Remove(); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to remove network namespace %s for sandbox %q", sandbox.NetNSPath, id)
}
sandbox.NetNSPath = ""
}
}()
// Setup network for sandbox.
// Certain VM based solutions like clear containers (Issue containerd/cri-containerd#524)
// rely on the assumption that CRI shim will not be querying the network namespace to check the
// network states such as IP.
// In future runtime implementation should avoid relying on CRI shim implementation details.
// In this case however caching the IP will add a subtle performance enhancement by avoiding
// calls to network namespace of the pod to query the IP of the veth interface on every
// SandboxStatus request.
if err := c.setupPodNetwork(ctx, &sandbox); err != nil {
return nil, errors.Wrapf(err, "failed to setup network for sandbox %q", id)
}
}
// Create sandbox container.
// NOTE: sandboxContainerSpec SHOULD NOT have side
// effect, e.g. accessing/creating files, so that we can test
// it safely.
spec, err := c.sandboxContainerSpec(id, config, &image.ImageSpec.Config, sandbox.NetNSPath, ociRuntime.PodAnnotations)
if err != nil {
return nil, errors.Wrap(err, "failed to generate sandbox container spec")
}
log.G(ctx).Debugf("Sandbox container %q spec: %#+v", id, spew.NewFormatter(spec))
sandbox.ProcessLabel = spec.Process.SelinuxLabel
defer func() {
if retErr != nil {
selinux.ReleaseLabel(sandbox.ProcessLabel)
}
}()
// handle any KVM based runtime
if err := modifyProcessLabel(ociRuntime.Type, spec); err != nil {
return nil, err
}
if config.GetLinux().GetSecurityContext().GetPrivileged() {
// If privileged don't set selinux label, but we still record the MCS label so that
// the unused label can be freed later.
spec.Process.SelinuxLabel = ""
}
// Generate spec options that will be applied to the spec later.
specOpts, err := c.sandboxContainerSpecOpts(config, &image.ImageSpec.Config)
if err != nil {
return nil, errors.Wrap(err, "failed to generate sanbdox container spec options")
}
sandboxLabels := buildLabels(config.Labels, containerKindSandbox)
runtimeOpts, err := generateRuntimeOptions(ociRuntime, c.config)
if err != nil {
return nil, errors.Wrap(err, "failed to generate runtime options")
}
opts := []containerd.NewContainerOpts{
containerd.WithSnapshotter(c.config.ContainerdConfig.Snapshotter),
customopts.WithNewSnapshot(id, containerdImage),
containerd.WithSpec(spec, specOpts...),
containerd.WithContainerLabels(sandboxLabels),
containerd.WithContainerExtension(sandboxMetadataExtension, &sandbox.Metadata),
containerd.WithRuntime(ociRuntime.Type, runtimeOpts)}
container, err := c.client.NewContainer(ctx, id, opts...)
if err != nil {
return nil, errors.Wrap(err, "failed to create containerd container")
}
defer func() {
if retErr != nil {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
if err := container.Delete(deferCtx, containerd.WithSnapshotCleanup); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to delete containerd container %q", id)
}
}
}()
// Create sandbox container root directories.
sandboxRootDir := c.getSandboxRootDir(id)
if err := c.os.MkdirAll(sandboxRootDir, 0755); err != nil {
return nil, errors.Wrapf(err, "failed to create sandbox root directory %q",
sandboxRootDir)
}
defer func() {
if retErr != nil {
// Cleanup the sandbox root directory.
if err := c.os.RemoveAll(sandboxRootDir); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to remove sandbox root directory %q",
sandboxRootDir)
}
}
}()
volatileSandboxRootDir := c.getVolatileSandboxRootDir(id)
if err := c.os.MkdirAll(volatileSandboxRootDir, 0755); err != nil {
return nil, errors.Wrapf(err, "failed to create volatile sandbox root directory %q",
volatileSandboxRootDir)
}
defer func() {
if retErr != nil {
// Cleanup the volatile sandbox root directory.
if err := c.os.RemoveAll(volatileSandboxRootDir); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to remove volatile sandbox root directory %q",
volatileSandboxRootDir)
}
}
}()
// Setup files required for the sandbox.
if err = c.setupSandboxFiles(id, config); err != nil {
return nil, errors.Wrapf(err, "failed to setup sandbox files")
}
defer func() {
if retErr != nil {
if err = c.cleanupSandboxFiles(id, config); err != nil {
log.G(ctx).WithError(err).Errorf("Failed to cleanup sandbox files in %q",
sandboxRootDir)
}
}
}()
// Update sandbox created timestamp.
info, err := container.Info(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox container info")
}
// Create sandbox task in containerd.
log.G(ctx).Tracef("Create sandbox container (id=%q, name=%q).",
id, name)
taskOpts := c.taskOpts(ociRuntime.Type)
// We don't need stdio for sandbox container.
task, err := container.NewTask(ctx, containerdio.NullIO, taskOpts...)
if err != nil {
return nil, errors.Wrap(err, "failed to create containerd task")
}
defer func() {
if retErr != nil {
deferCtx, deferCancel := ctrdutil.DeferContext()
defer deferCancel()
// Cleanup the sandbox container if an error is returned.
if _, err := task.Delete(deferCtx, WithNRISandboxDelete(id), containerd.WithProcessKill); err != nil && !errdefs.IsNotFound(err) {
log.G(ctx).WithError(err).Errorf("Failed to delete sandbox container %q", id)
}
}
}()
// wait is a long running background request, no timeout needed.
exitCh, err := task.Wait(ctrdutil.NamespacedContext())
if err != nil {
return nil, errors.Wrap(err, "failed to wait for sandbox container task")
}
nric, err := nri.New()
if err != nil {
return nil, errors.Wrap(err, "unable to create nri client")
}
if nric != nil {
nriSB := &nri.Sandbox{
ID: id,
Labels: config.Labels,
}
if _, err := nric.InvokeWithSandbox(ctx, task, v1.Create, nriSB); err != nil {
return nil, errors.Wrap(err, "nri invoke")
}
}
if err := task.Start(ctx); err != nil {
return nil, errors.Wrapf(err, "failed to start sandbox container task %q", id)
}
if err := sandbox.Status.Update(func(status sandboxstore.Status) (sandboxstore.Status, error) {
// Set the pod sandbox as ready after successfully start sandbox container.
status.Pid = task.Pid()
status.State = sandboxstore.StateReady
status.CreatedAt = info.CreatedAt
return status, nil
}); err != nil {
return nil, errors.Wrap(err, "failed to update sandbox status")
}
// Add sandbox into sandbox store in INIT state.
sandbox.Container = container
if err := c.sandboxStore.Add(sandbox); err != nil {
return nil, errors.Wrapf(err, "failed to add sandbox %+v into store", sandbox)
}
// start the monitor after adding sandbox into the store, this ensures
// that sandbox is in the store, when event monitor receives the TaskExit event.
//
// TaskOOM from containerd may come before sandbox is added to store,
// but we don't care about sandbox TaskOOM right now, so it is fine.
c.eventMonitor.startExitMonitor(context.Background(), id, task.Pid(), exitCh)
return &runtime.RunPodSandboxResponse{PodSandboxId: id}, nil
}
// setupPodNetwork setups up the network for a pod
func (c *criService) setupPodNetwork(ctx context.Context, sandbox *sandboxstore.Sandbox) error {
var (
id = sandbox.ID
config = sandbox.Config
path = sandbox.NetNSPath
)
if c.netPlugin == nil {
return errors.New("cni config not initialized")
}
opts, err := cniNamespaceOpts(id, config)
if err != nil {
return errors.Wrap(err, "get cni namespace options")
}
result, err := c.netPlugin.Setup(ctx, id, path, opts...)
if err != nil {
return err
}
logDebugCNIResult(ctx, id, result)
// Check if the default interface has IP config
if configs, ok := result.Interfaces[defaultIfName]; ok && len(configs.IPConfigs) > 0 {
sandbox.IP, sandbox.AdditionalIPs = selectPodIPs(configs.IPConfigs)
sandbox.CNIResult = result
return nil
}
return errors.Errorf("failed to find network info for sandbox %q", id)
}
// cniNamespaceOpts get CNI namespace options from sandbox config.
func cniNamespaceOpts(id string, config *runtime.PodSandboxConfig) ([]cni.NamespaceOpts, error) {
opts := []cni.NamespaceOpts{
cni.WithLabels(toCNILabels(id, config)),
}
portMappings := toCNIPortMappings(config.GetPortMappings())
if len(portMappings) > 0 {
opts = append(opts, cni.WithCapabilityPortMap(portMappings))
}
// Will return an error if the bandwidth limitation has the wrong unit
// or an unreasonable value see validateBandwidthIsReasonable()
bandWidth, err := toCNIBandWidth(config.Annotations)
if err != nil {
return nil, err
}
if bandWidth != nil {
opts = append(opts, cni.WithCapabilityBandWidth(*bandWidth))
}
dns := toCNIDNS(config.GetDnsConfig())
if dns != nil {
opts = append(opts, cni.WithCapabilityDNS(*dns))
}
return opts, nil
}
// toCNILabels adds pod metadata into CNI labels.
func toCNILabels(id string, config *runtime.PodSandboxConfig) map[string]string {
return map[string]string{
"K8S_POD_NAMESPACE": config.GetMetadata().GetNamespace(),
"K8S_POD_NAME": config.GetMetadata().GetName(),
"K8S_POD_INFRA_CONTAINER_ID": id,
"IgnoreUnknown": "1",
}
}
// toCNIBandWidth converts CRI annotations to CNI bandwidth.
func toCNIBandWidth(annotations map[string]string) (*cni.BandWidth, error) {
ingress, egress, err := bandwidth.ExtractPodBandwidthResources(annotations)
if err != nil {
return nil, errors.Wrap(err, "reading pod bandwidth annotations")
}
if ingress == nil && egress == nil {
return nil, nil
}
bandWidth := &cni.BandWidth{}
if ingress != nil {
bandWidth.IngressRate = uint64(ingress.Value())
bandWidth.IngressBurst = math.MaxUint32
}
if egress != nil {
bandWidth.EgressRate = uint64(egress.Value())
bandWidth.EgressBurst = math.MaxUint32
}
return bandWidth, nil
}
// toCNIPortMappings converts CRI port mappings to CNI.
func toCNIPortMappings(criPortMappings []*runtime.PortMapping) []cni.PortMapping {
var portMappings []cni.PortMapping
for _, mapping := range criPortMappings {
if mapping.HostPort <= 0 {
continue
}
portMappings = append(portMappings, cni.PortMapping{
HostPort: mapping.HostPort,
ContainerPort: mapping.ContainerPort,
Protocol: strings.ToLower(mapping.Protocol.String()),
HostIP: mapping.HostIp,
})
}
return portMappings
}
// toCNIDNS converts CRI DNSConfig to CNI.
func toCNIDNS(dns *runtime.DNSConfig) *cni.DNS {
if dns == nil {
return nil
}
return &cni.DNS{
Servers: dns.GetServers(),
Searches: dns.GetSearches(),
Options: dns.GetOptions(),
}
}
// selectPodIPs select an ip from the ip list. It prefers ipv4 more than ipv6
// and returns the additional ips
// TODO(random-liu): Revisit the ip order in the ipv6 beta stage. (cri#1278)
func selectPodIPs(ipConfigs []*cni.IPConfig) (string, []string) {
var (
additionalIPs []string
ip string
)
for _, c := range ipConfigs {
if c.IP.To4() != nil && ip == "" {
ip = c.IP.String()
} else {
additionalIPs = append(additionalIPs, c.IP.String())
}
}
if ip != "" {
return ip, additionalIPs
}
if len(ipConfigs) == 1 {
return additionalIPs[0], nil
}
return additionalIPs[0], additionalIPs[1:]
}
// untrustedWorkload returns true if the sandbox contains untrusted workload.
func untrustedWorkload(config *runtime.PodSandboxConfig) bool {
return config.GetAnnotations()[annotations.UntrustedWorkload] == "true"
}
// hostAccessingSandbox returns true if the sandbox configuration
// requires additional host access for the sandbox.
func hostAccessingSandbox(config *runtime.PodSandboxConfig) bool {
securityContext := config.GetLinux().GetSecurityContext()
namespaceOptions := securityContext.GetNamespaceOptions()
if namespaceOptions.GetNetwork() == runtime.NamespaceMode_NODE ||
namespaceOptions.GetPid() == runtime.NamespaceMode_NODE ||
namespaceOptions.GetIpc() == runtime.NamespaceMode_NODE {
return true
}
return false
}
// getSandboxRuntime returns the runtime configuration for sandbox.
// If the sandbox contains untrusted workload, runtime for untrusted workload will be returned,
// or else default runtime will be returned.
func (c *criService) getSandboxRuntime(config *runtime.PodSandboxConfig, runtimeHandler string) (criconfig.Runtime, error) {
if untrustedWorkload(config) {
// If the untrusted annotation is provided, runtimeHandler MUST be empty.
if runtimeHandler != "" && runtimeHandler != criconfig.RuntimeUntrusted {
return criconfig.Runtime{}, errors.New("untrusted workload with explicit runtime handler is not allowed")
}
// If the untrusted workload is requesting access to the host/node, this request will fail.
//
// Note: If the workload is marked untrusted but requests privileged, this can be granted, as the
// runtime may support this. For example, in a virtual-machine isolated runtime, privileged
// is a supported option, granting the workload to access the entire guest VM instead of host.
// TODO(windows): Deprecate this so that we don't need to handle it for windows.
if hostAccessingSandbox(config) {
return criconfig.Runtime{}, errors.New("untrusted workload with host access is not allowed")
}
runtimeHandler = criconfig.RuntimeUntrusted
}
if runtimeHandler == "" {
runtimeHandler = c.config.ContainerdConfig.DefaultRuntimeName
}
handler, ok := c.config.ContainerdConfig.Runtimes[runtimeHandler]
if !ok {
return criconfig.Runtime{}, errors.Errorf("no runtime for %q is configured", runtimeHandler)
}
return handler, nil
}
func logDebugCNIResult(ctx context.Context, sandboxID string, result *cni.CNIResult) {
if logrus.GetLevel() < logrus.DebugLevel {
return
}
cniResult, err := json.Marshal(result)
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to marshal CNI result for sandbox %q: %v", sandboxID, err)
return
}
log.G(ctx).Debugf("cni result for sandbox %q: %s", sandboxID, string(cniResult))
}

View File

@@ -1,310 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"fmt"
"os"
"strings"
"github.com/containerd/containerd"
"github.com/containerd/containerd/oci"
"github.com/containerd/containerd/plugin"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
selinux "github.com/opencontainers/selinux/go-selinux"
"github.com/pkg/errors"
"golang.org/x/sys/unix"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/annotations"
customopts "github.com/containerd/cri/pkg/containerd/opts"
osinterface "github.com/containerd/cri/pkg/os"
)
func (c *criService) sandboxContainerSpec(id string, config *runtime.PodSandboxConfig,
imageConfig *imagespec.ImageConfig, nsPath string, runtimePodAnnotations []string) (_ *runtimespec.Spec, retErr error) {
// Creates a spec Generator with the default spec.
// TODO(random-liu): [P1] Compare the default settings with docker and containerd default.
specOpts := []oci.SpecOpts{
customopts.WithoutRunMount,
customopts.WithoutDefaultSecuritySettings,
customopts.WithRelativeRoot(relativeRootfsPath),
oci.WithEnv(imageConfig.Env),
oci.WithRootFSReadonly(),
oci.WithHostname(config.GetHostname()),
}
if imageConfig.WorkingDir != "" {
specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
}
if len(imageConfig.Entrypoint) == 0 && len(imageConfig.Cmd) == 0 {
// Pause image must have entrypoint or cmd.
return nil, errors.Errorf("invalid empty entrypoint and cmd in image config %+v", imageConfig)
}
specOpts = append(specOpts, oci.WithProcessArgs(append(imageConfig.Entrypoint, imageConfig.Cmd...)...))
// Set cgroups parent.
if c.config.DisableCgroup {
specOpts = append(specOpts, customopts.WithDisabledCgroups)
} else {
if config.GetLinux().GetCgroupParent() != "" {
cgroupsPath := getCgroupsPath(config.GetLinux().GetCgroupParent(), id)
specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
}
}
// When cgroup parent is not set, containerd-shim will create container in a child cgroup
// of the cgroup itself is in.
// TODO(random-liu): [P2] Set default cgroup path if cgroup parent is not specified.
// Set namespace options.
var (
securityContext = config.GetLinux().GetSecurityContext()
nsOptions = securityContext.GetNamespaceOptions()
)
if nsOptions.GetNetwork() == runtime.NamespaceMode_NODE {
specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.NetworkNamespace))
specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.UTSNamespace))
} else {
specOpts = append(specOpts, oci.WithLinuxNamespace(
runtimespec.LinuxNamespace{
Type: runtimespec.NetworkNamespace,
Path: nsPath,
}))
}
if nsOptions.GetPid() == runtime.NamespaceMode_NODE {
specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.PIDNamespace))
}
if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.IPCNamespace))
}
// It's fine to generate the spec before the sandbox /dev/shm
// is actually created.
sandboxDevShm := c.getSandboxDevShm(id)
if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
sandboxDevShm = devShm
}
specOpts = append(specOpts, oci.WithMounts([]runtimespec.Mount{
{
Source: sandboxDevShm,
Destination: devShm,
Type: "bind",
Options: []string{"rbind", "ro"},
},
// Add resolv.conf for katacontainers to setup the DNS of pod VM properly.
{
Source: c.getResolvPath(id),
Destination: resolvConfPath,
Type: "bind",
Options: []string{"rbind", "ro"},
},
}))
processLabel, mountLabel, err := initLabelsFromOpt(securityContext.GetSelinuxOptions())
if err != nil {
return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
}
defer func() {
if retErr != nil {
selinux.ReleaseLabel(processLabel)
}
}()
supplementalGroups := securityContext.GetSupplementalGroups()
specOpts = append(specOpts,
customopts.WithSelinuxLabels(processLabel, mountLabel),
customopts.WithSupplementalGroups(supplementalGroups),
)
// Add sysctls
sysctls := config.GetLinux().GetSysctls()
specOpts = append(specOpts, customopts.WithSysctls(sysctls))
// Note: LinuxSandboxSecurityContext does not currently provide an apparmor profile
if !c.config.DisableCgroup {
specOpts = append(specOpts, customopts.WithDefaultSandboxShares)
}
specOpts = append(specOpts, customopts.WithPodOOMScoreAdj(int(defaultSandboxOOMAdj), c.config.RestrictOOMScoreAdj))
for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
runtimePodAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
specOpts = append(specOpts,
customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeSandbox),
customopts.WithAnnotation(annotations.SandboxID, id),
customopts.WithAnnotation(annotations.SandboxLogDir, config.GetLogDirectory()),
)
return c.runtimeSpec(id, "", specOpts...)
}
// sandboxContainerSpecOpts generates OCI spec options for
// the sandbox container.
func (c *criService) sandboxContainerSpecOpts(config *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
var (
securityContext = config.GetLinux().GetSecurityContext()
specOpts []oci.SpecOpts
)
seccompSpecOpts, err := c.generateSeccompSpecOpts(
securityContext.GetSeccompProfilePath(),
securityContext.GetPrivileged(),
c.seccompEnabled())
if err != nil {
return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
}
if seccompSpecOpts != nil {
specOpts = append(specOpts, seccompSpecOpts)
}
userstr, err := generateUserString(
"",
securityContext.GetRunAsUser(),
securityContext.GetRunAsGroup(),
)
if err != nil {
return nil, errors.Wrap(err, "failed to generate user string")
}
if userstr == "" {
// Lastly, since no user override was passed via CRI try to set via OCI
// Image
userstr = imageConfig.User
}
if userstr != "" {
specOpts = append(specOpts, oci.WithUser(userstr))
}
return specOpts, nil
}
// setupSandboxFiles sets up necessary sandbox files including /dev/shm, /etc/hosts,
// /etc/resolv.conf and /etc/hostname.
func (c *criService) setupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
sandboxEtcHostname := c.getSandboxHostname(id)
hostname := config.GetHostname()
if hostname == "" {
var err error
hostname, err = c.os.Hostname()
if err != nil {
return errors.Wrap(err, "failed to get hostname")
}
}
if err := c.os.WriteFile(sandboxEtcHostname, []byte(hostname+"\n"), 0644); err != nil {
return errors.Wrapf(err, "failed to write hostname to %q", sandboxEtcHostname)
}
// TODO(random-liu): Consider whether we should maintain /etc/hosts and /etc/resolv.conf in kubelet.
sandboxEtcHosts := c.getSandboxHosts(id)
if err := c.os.CopyFile(etcHosts, sandboxEtcHosts, 0644); err != nil {
return errors.Wrapf(err, "failed to generate sandbox hosts file %q", sandboxEtcHosts)
}
// Set DNS options. Maintain a resolv.conf for the sandbox.
var err error
resolvContent := ""
if dnsConfig := config.GetDnsConfig(); dnsConfig != nil {
resolvContent, err = parseDNSOptions(dnsConfig.Servers, dnsConfig.Searches, dnsConfig.Options)
if err != nil {
return errors.Wrapf(err, "failed to parse sandbox DNSConfig %+v", dnsConfig)
}
}
resolvPath := c.getResolvPath(id)
if resolvContent == "" {
// copy host's resolv.conf to resolvPath
err = c.os.CopyFile(resolvConfPath, resolvPath, 0644)
if err != nil {
return errors.Wrapf(err, "failed to copy host's resolv.conf to %q", resolvPath)
}
} else {
err = c.os.WriteFile(resolvPath, []byte(resolvContent), 0644)
if err != nil {
return errors.Wrapf(err, "failed to write resolv content to %q", resolvPath)
}
}
// Setup sandbox /dev/shm.
if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
if _, err := c.os.Stat(devShm); err != nil {
return errors.Wrapf(err, "host %q is not available for host ipc", devShm)
}
} else {
sandboxDevShm := c.getSandboxDevShm(id)
if err := c.os.MkdirAll(sandboxDevShm, 0700); err != nil {
return errors.Wrap(err, "failed to create sandbox shm")
}
shmproperty := fmt.Sprintf("mode=1777,size=%d", defaultShmSize)
if err := c.os.(osinterface.UNIX).Mount("shm", sandboxDevShm, "tmpfs", uintptr(unix.MS_NOEXEC|unix.MS_NOSUID|unix.MS_NODEV), shmproperty); err != nil {
return errors.Wrap(err, "failed to mount sandbox shm")
}
}
return nil
}
// parseDNSOptions parse DNS options into resolv.conf format content,
// if none option is specified, will return empty with no error.
func parseDNSOptions(servers, searches, options []string) (string, error) {
resolvContent := ""
if len(searches) > maxDNSSearches {
return "", errors.Errorf("DNSOption.Searches has more than %d domains", maxDNSSearches)
}
if len(searches) > 0 {
resolvContent += fmt.Sprintf("search %s\n", strings.Join(searches, " "))
}
if len(servers) > 0 {
resolvContent += fmt.Sprintf("nameserver %s\n", strings.Join(servers, "\nnameserver "))
}
if len(options) > 0 {
resolvContent += fmt.Sprintf("options %s\n", strings.Join(options, " "))
}
return resolvContent, nil
}
// cleanupSandboxFiles unmount some sandbox files, we rely on the removal of sandbox root directory to
// remove these files. Unmount should *NOT* return error if the mount point is already unmounted.
func (c *criService) cleanupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() != runtime.NamespaceMode_NODE {
path, err := c.os.FollowSymlinkInScope(c.getSandboxDevShm(id), "/")
if err != nil {
return errors.Wrap(err, "failed to follow symlink")
}
if err := c.os.(osinterface.UNIX).Unmount(path); err != nil && !os.IsNotExist(err) {
return errors.Wrapf(err, "failed to unmount %q", path)
}
}
return nil
}
// taskOpts generates task options for a (sandbox) container.
func (c *criService) taskOpts(runtimeType string) []containerd.NewTaskOpts {
// TODO(random-liu): Remove this after shim v1 is deprecated.
var taskOpts []containerd.NewTaskOpts
if c.config.NoPivot && (runtimeType == plugin.RuntimeRuncV1 || runtimeType == plugin.RuntimeRuncV2) {
taskOpts = append(taskOpts, containerd.WithNoPivotRoot)
}
return taskOpts
}

View File

@@ -1,91 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd"
"github.com/containerd/containerd/oci"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/annotations"
customopts "github.com/containerd/cri/pkg/containerd/opts"
)
func (c *criService) sandboxContainerSpec(id string, config *runtime.PodSandboxConfig,
imageConfig *imagespec.ImageConfig, nsPath string, runtimePodAnnotations []string) (*runtimespec.Spec, error) {
// Creates a spec Generator with the default spec.
specOpts := []oci.SpecOpts{
oci.WithEnv(imageConfig.Env),
oci.WithHostname(config.GetHostname()),
}
if imageConfig.WorkingDir != "" {
specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
}
if len(imageConfig.Entrypoint) == 0 && len(imageConfig.Cmd) == 0 {
// Pause image must have entrypoint or cmd.
return nil, errors.Errorf("invalid empty entrypoint and cmd in image config %+v", imageConfig)
}
specOpts = append(specOpts, oci.WithProcessArgs(append(imageConfig.Entrypoint, imageConfig.Cmd...)...))
specOpts = append(specOpts,
// Clear the root location since hcsshim expects it.
// NOTE: readonly rootfs doesn't work on windows.
customopts.WithoutRoot,
customopts.WithWindowsNetworkNamespace(nsPath),
)
specOpts = append(specOpts, customopts.WithWindowsDefaultSandboxShares)
for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
runtimePodAnnotations) {
specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
}
specOpts = append(specOpts,
customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeSandbox),
customopts.WithAnnotation(annotations.SandboxID, id),
customopts.WithAnnotation(annotations.SandboxLogDir, config.GetLogDirectory()),
)
return c.runtimeSpec(id, "", specOpts...)
}
// No sandbox container spec options for windows yet.
func (c *criService) sandboxContainerSpecOpts(config *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
return nil, nil
}
// No sandbox files needed for windows.
func (c *criService) setupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
return nil
}
// No sandbox files needed for windows.
func (c *criService) cleanupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
return nil
}
// No task options needed for windows.
func (c *criService) taskOpts(runtimeType string) []containerd.NewTaskOpts {
return nil
}

View File

@@ -1,217 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
goruntime "runtime"
"github.com/containerd/containerd"
"github.com/containerd/containerd/errdefs"
cni "github.com/containerd/go-cni"
runtimespec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// PodSandboxStatus returns the status of the PodSandbox.
func (c *criService) PodSandboxStatus(ctx context.Context, r *runtime.PodSandboxStatusRequest) (*runtime.PodSandboxStatusResponse, error) {
sandbox, err := c.sandboxStore.Get(r.GetPodSandboxId())
if err != nil {
return nil, errors.Wrap(err, "an error occurred when try to find sandbox")
}
ip, additionalIPs, err := c.getIPs(sandbox)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox ip")
}
status := toCRISandboxStatus(sandbox.Metadata, sandbox.Status.Get(), ip, additionalIPs)
if status.GetCreatedAt() == 0 {
// CRI doesn't allow CreatedAt == 0.
info, err := sandbox.Container.Info(ctx)
if err != nil {
return nil, errors.Wrapf(err, "failed to get CreatedAt for sandbox container in %q state", status.State)
}
status.CreatedAt = info.CreatedAt.UnixNano()
}
if !r.GetVerbose() {
return &runtime.PodSandboxStatusResponse{Status: status}, nil
}
// Generate verbose information.
info, err := toCRISandboxInfo(ctx, sandbox)
if err != nil {
return nil, errors.Wrap(err, "failed to get verbose sandbox container info")
}
return &runtime.PodSandboxStatusResponse{
Status: status,
Info: info,
}, nil
}
func (c *criService) getIPs(sandbox sandboxstore.Sandbox) (string, []string, error) {
config := sandbox.Config
if goruntime.GOOS != "windows" &&
config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
// For sandboxes using the node network we are not
// responsible for reporting the IP.
return "", nil, nil
}
if closed, err := sandbox.NetNS.Closed(); err != nil {
return "", nil, errors.Wrap(err, "check network namespace closed")
} else if closed {
return "", nil, nil
}
return sandbox.IP, sandbox.AdditionalIPs, nil
}
// toCRISandboxStatus converts sandbox metadata into CRI pod sandbox status.
func toCRISandboxStatus(meta sandboxstore.Metadata, status sandboxstore.Status, ip string, additionalIPs []string) *runtime.PodSandboxStatus {
// Set sandbox state to NOTREADY by default.
state := runtime.PodSandboxState_SANDBOX_NOTREADY
if status.State == sandboxstore.StateReady {
state = runtime.PodSandboxState_SANDBOX_READY
}
nsOpts := meta.Config.GetLinux().GetSecurityContext().GetNamespaceOptions()
var ips []*runtime.PodIP
for _, additionalIP := range additionalIPs {
ips = append(ips, &runtime.PodIP{Ip: additionalIP})
}
return &runtime.PodSandboxStatus{
Id: meta.ID,
Metadata: meta.Config.GetMetadata(),
State: state,
CreatedAt: status.CreatedAt.UnixNano(),
Network: &runtime.PodSandboxNetworkStatus{
Ip: ip,
AdditionalIps: ips,
},
Linux: &runtime.LinuxPodSandboxStatus{
Namespaces: &runtime.Namespace{
Options: &runtime.NamespaceOption{
Network: nsOpts.GetNetwork(),
Pid: nsOpts.GetPid(),
Ipc: nsOpts.GetIpc(),
},
},
},
Labels: meta.Config.GetLabels(),
Annotations: meta.Config.GetAnnotations(),
RuntimeHandler: meta.RuntimeHandler,
}
}
// SandboxInfo is extra information for sandbox.
// TODO (mikebrow): discuss predefining constants structures for some or all of these field names in CRI
type SandboxInfo struct {
Pid uint32 `json:"pid"`
Status string `json:"processStatus"`
NetNSClosed bool `json:"netNamespaceClosed"`
Image string `json:"image"`
SnapshotKey string `json:"snapshotKey"`
Snapshotter string `json:"snapshotter"`
// Note: a new field `RuntimeHandler` has been added into the CRI PodSandboxStatus struct, and
// should be set. This `RuntimeHandler` field will be deprecated after containerd 1.3 (tracked
// in https://github.com/containerd/cri/issues/1064).
RuntimeHandler string `json:"runtimeHandler"` // see the Note above
RuntimeType string `json:"runtimeType"`
RuntimeOptions interface{} `json:"runtimeOptions"`
Config *runtime.PodSandboxConfig `json:"config"`
RuntimeSpec *runtimespec.Spec `json:"runtimeSpec"`
CNIResult *cni.CNIResult `json:"cniResult"`
}
// toCRISandboxInfo converts internal container object information to CRI sandbox status response info map.
func toCRISandboxInfo(ctx context.Context, sandbox sandboxstore.Sandbox) (map[string]string, error) {
container := sandbox.Container
task, err := container.Task(ctx, nil)
if err != nil && !errdefs.IsNotFound(err) {
return nil, errors.Wrap(err, "failed to get sandbox container task")
}
var processStatus containerd.ProcessStatus
if task != nil {
taskStatus, err := task.Status(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get task status")
}
processStatus = taskStatus.Status
}
si := &SandboxInfo{
Pid: sandbox.Status.Get().Pid,
RuntimeHandler: sandbox.RuntimeHandler,
Status: string(processStatus),
Config: sandbox.Config,
CNIResult: sandbox.CNIResult,
}
if si.Status == "" {
// If processStatus is empty, it means that the task is deleted. Apply "deleted"
// status which does not exist in containerd.
si.Status = "deleted"
}
if sandbox.NetNS != nil {
// Add network closed information if sandbox is not using host network.
closed, err := sandbox.NetNS.Closed()
if err != nil {
return nil, errors.Wrap(err, "failed to check network namespace closed")
}
si.NetNSClosed = closed
}
spec, err := container.Spec(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox container runtime spec")
}
si.RuntimeSpec = spec
ctrInfo, err := container.Info(ctx)
if err != nil {
return nil, errors.Wrap(err, "failed to get sandbox container info")
}
// Do not use config.SandboxImage because the configuration might
// be changed during restart. It may not reflect the actual image
// used by the sandbox container.
si.Image = ctrInfo.Image
si.SnapshotKey = ctrInfo.SnapshotKey
si.Snapshotter = ctrInfo.Snapshotter
runtimeOptions, err := getRuntimeOptions(ctrInfo)
if err != nil {
return nil, errors.Wrap(err, "failed to get runtime options")
}
si.RuntimeType = ctrInfo.Runtime.Name
si.RuntimeOptions = runtimeOptions
infoBytes, err := json.Marshal(si)
if err != nil {
return nil, errors.Wrapf(err, "failed to marshal info %v", si)
}
return map[string]string{
"info": string(infoBytes),
}, nil
}

View File

@@ -1,195 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"syscall"
"time"
eventtypes "github.com/containerd/containerd/api/events"
"github.com/containerd/containerd/errdefs"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
)
// StopPodSandbox stops the sandbox. If there are any running containers in the
// sandbox, they should be forcibly terminated.
func (c *criService) StopPodSandbox(ctx context.Context, r *runtime.StopPodSandboxRequest) (*runtime.StopPodSandboxResponse, error) {
sandbox, err := c.sandboxStore.Get(r.GetPodSandboxId())
if err != nil {
return nil, errors.Wrapf(err, "an error occurred when try to find sandbox %q",
r.GetPodSandboxId())
}
if err := c.stopPodSandbox(ctx, sandbox); err != nil {
return nil, err
}
return &runtime.StopPodSandboxResponse{}, nil
}
func (c *criService) stopPodSandbox(ctx context.Context, sandbox sandboxstore.Sandbox) error {
// Use the full sandbox id.
id := sandbox.ID
// Stop all containers inside the sandbox. This terminates the container forcibly,
// and container may still be created, so production should not rely on this behavior.
// TODO(random-liu): Introduce a state in sandbox to avoid future container creation.
containers := c.containerStore.List()
for _, container := range containers {
if container.SandboxID != id {
continue
}
// Forcibly stop the container. Do not use `StopContainer`, because it introduces a race
// if a container is removed after list.
if err := c.stopContainer(ctx, container, 0); err != nil {
return errors.Wrapf(err, "failed to stop container %q", container.ID)
}
}
if err := c.cleanupSandboxFiles(id, sandbox.Config); err != nil {
return errors.Wrap(err, "failed to cleanup sandbox files")
}
// Only stop sandbox container when it's running or unknown.
state := sandbox.Status.Get().State
if state == sandboxstore.StateReady || state == sandboxstore.StateUnknown {
if err := c.stopSandboxContainer(ctx, sandbox); err != nil {
return errors.Wrapf(err, "failed to stop sandbox container %q in %q state", id, state)
}
}
// Teardown network for sandbox.
if sandbox.NetNS != nil {
// Use empty netns path if netns is not available. This is defined in:
// https://github.com/containernetworking/cni/blob/v0.7.0-alpha1/SPEC.md
if closed, err := sandbox.NetNS.Closed(); err != nil {
return errors.Wrap(err, "failed to check network namespace closed")
} else if closed {
sandbox.NetNSPath = ""
}
if err := c.teardownPodNetwork(ctx, sandbox); err != nil {
return errors.Wrapf(err, "failed to destroy network for sandbox %q", id)
}
if err := sandbox.NetNS.Remove(); err != nil {
return errors.Wrapf(err, "failed to remove network namespace for sandbox %q", id)
}
}
log.G(ctx).Infof("TearDown network for sandbox %q successfully", id)
return nil
}
// stopSandboxContainer kills the sandbox container.
// `task.Delete` is not called here because it will be called when
// the event monitor handles the `TaskExit` event.
func (c *criService) stopSandboxContainer(ctx context.Context, sandbox sandboxstore.Sandbox) error {
id := sandbox.ID
container := sandbox.Container
state := sandbox.Status.Get().State
task, err := container.Task(ctx, nil)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to get sandbox container")
}
// Don't return for unknown state, some cleanup needs to be done.
if state == sandboxstore.StateUnknown {
return cleanupUnknownSandbox(ctx, id, sandbox)
}
return nil
}
// Handle unknown state.
// The cleanup logic is the same with container unknown state.
if state == sandboxstore.StateUnknown {
// Start an exit handler for containers in unknown state.
waitCtx, waitCancel := context.WithCancel(ctrdutil.NamespacedContext())
defer waitCancel()
exitCh, err := task.Wait(waitCtx)
if err != nil {
if !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to wait for task")
}
return cleanupUnknownSandbox(ctx, id, sandbox)
}
exitCtx, exitCancel := context.WithCancel(context.Background())
stopCh := c.eventMonitor.startExitMonitor(exitCtx, id, task.Pid(), exitCh)
defer func() {
exitCancel()
// This ensures that exit monitor is stopped before
// `Wait` is cancelled, so no exit event is generated
// because of the `Wait` cancellation.
<-stopCh
}()
}
// Kill the sandbox container.
if err = task.Kill(ctx, syscall.SIGKILL); err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "failed to kill sandbox container")
}
return c.waitSandboxStop(ctx, sandbox)
}
// waitSandboxStop waits for sandbox to be stopped until context is cancelled or
// the context deadline is exceeded.
func (c *criService) waitSandboxStop(ctx context.Context, sandbox sandboxstore.Sandbox) error {
select {
case <-ctx.Done():
return errors.Wrapf(ctx.Err(), "wait sandbox container %q", sandbox.ID)
case <-sandbox.Stopped():
return nil
}
}
// teardownPodNetwork removes the network from the pod
func (c *criService) teardownPodNetwork(ctx context.Context, sandbox sandboxstore.Sandbox) error {
if c.netPlugin == nil {
return errors.New("cni config not initialized")
}
var (
id = sandbox.ID
path = sandbox.NetNSPath
config = sandbox.Config
)
opts, err := cniNamespaceOpts(id, config)
if err != nil {
return errors.Wrap(err, "get cni namespace options")
}
return c.netPlugin.Remove(ctx, id, path, opts...)
}
// cleanupUnknownSandbox cleanup stopped sandbox in unknown state.
func cleanupUnknownSandbox(ctx context.Context, id string, sandbox sandboxstore.Sandbox) error {
// Reuse handleSandboxExit to do the cleanup.
return handleSandboxExit(ctx, &eventtypes.TaskExit{
ContainerID: id,
ID: id,
Pid: 0,
ExitStatus: unknownExitCode,
ExitedAt: time.Now(),
}, sandbox)
}

View File

@@ -1,325 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
"fmt"
"io"
"net/http"
"os"
"path/filepath"
"time"
"github.com/containerd/containerd"
"github.com/containerd/containerd/oci"
"github.com/containerd/containerd/plugin"
"github.com/containerd/cri/pkg/streaming"
cni "github.com/containerd/go-cni"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"google.golang.org/grpc"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/store/label"
"github.com/containerd/cri/pkg/atomic"
criconfig "github.com/containerd/cri/pkg/config"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
osinterface "github.com/containerd/cri/pkg/os"
"github.com/containerd/cri/pkg/registrar"
containerstore "github.com/containerd/cri/pkg/store/container"
imagestore "github.com/containerd/cri/pkg/store/image"
sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
snapshotstore "github.com/containerd/cri/pkg/store/snapshot"
)
// grpcServices are all the grpc services provided by cri containerd.
type grpcServices interface {
runtime.RuntimeServiceServer
runtime.ImageServiceServer
}
// CRIService is the interface implement CRI remote service server.
type CRIService interface {
Run() error
// io.Closer is used by containerd to gracefully stop cri service.
io.Closer
plugin.Service
grpcServices
}
// criService implements CRIService.
type criService struct {
// config contains all configurations.
config criconfig.Config
// imageFSPath is the path to image filesystem.
imageFSPath string
// os is an interface for all required os operations.
os osinterface.OS
// sandboxStore stores all resources associated with sandboxes.
sandboxStore *sandboxstore.Store
// sandboxNameIndex stores all sandbox names and make sure each name
// is unique.
sandboxNameIndex *registrar.Registrar
// containerStore stores all resources associated with containers.
containerStore *containerstore.Store
// containerNameIndex stores all container names and make sure each
// name is unique.
containerNameIndex *registrar.Registrar
// imageStore stores all resources associated with images.
imageStore *imagestore.Store
// snapshotStore stores information of all snapshots.
snapshotStore *snapshotstore.Store
// netPlugin is used to setup and teardown network when run/stop pod sandbox.
netPlugin cni.CNI
// client is an instance of the containerd client
client *containerd.Client
// streamServer is the streaming server serves container streaming request.
streamServer streaming.Server
// eventMonitor is the monitor monitors containerd events.
eventMonitor *eventMonitor
// initialized indicates whether the server is initialized. All GRPC services
// should return error before the server is initialized.
initialized atomic.Bool
// cniNetConfMonitor is used to reload cni network conf if there is
// any valid fs change events from cni network conf dir.
cniNetConfMonitor *cniNetConfSyncer
// baseOCISpecs contains cached OCI specs loaded via `Runtime.BaseRuntimeSpec`
baseOCISpecs map[string]*oci.Spec
}
// NewCRIService returns a new instance of CRIService
func NewCRIService(config criconfig.Config, client *containerd.Client) (CRIService, error) {
var err error
labels := label.NewStore()
c := &criService{
config: config,
client: client,
os: osinterface.RealOS{},
sandboxStore: sandboxstore.NewStore(labels),
containerStore: containerstore.NewStore(labels),
imageStore: imagestore.NewStore(client),
snapshotStore: snapshotstore.NewStore(),
sandboxNameIndex: registrar.NewRegistrar(),
containerNameIndex: registrar.NewRegistrar(),
initialized: atomic.NewBool(false),
}
if client.SnapshotService(c.config.ContainerdConfig.Snapshotter) == nil {
return nil, errors.Errorf("failed to find snapshotter %q", c.config.ContainerdConfig.Snapshotter)
}
c.imageFSPath = imageFSPath(config.ContainerdRootDir, config.ContainerdConfig.Snapshotter)
logrus.Infof("Get image filesystem path %q", c.imageFSPath)
if err := c.initPlatform(); err != nil {
return nil, errors.Wrap(err, "initialize platform")
}
// prepare streaming server
c.streamServer, err = newStreamServer(c, config.StreamServerAddress, config.StreamServerPort, config.StreamIdleTimeout)
if err != nil {
return nil, errors.Wrap(err, "failed to create stream server")
}
c.eventMonitor = newEventMonitor(c)
c.cniNetConfMonitor, err = newCNINetConfSyncer(c.config.NetworkPluginConfDir, c.netPlugin, c.cniLoadOptions())
if err != nil {
return nil, errors.Wrap(err, "failed to create cni conf monitor")
}
// Preload base OCI specs
c.baseOCISpecs, err = loadBaseOCISpecs(&config)
if err != nil {
return nil, err
}
return c, nil
}
// Register registers all required services onto a specific grpc server.
// This is used by containerd cri plugin.
func (c *criService) Register(s *grpc.Server) error {
return c.register(s)
}
// RegisterTCP register all required services onto a GRPC server on TCP.
// This is used by containerd CRI plugin.
func (c *criService) RegisterTCP(s *grpc.Server) error {
if !c.config.DisableTCPService {
return c.register(s)
}
return nil
}
// Run starts the CRI service.
func (c *criService) Run() error {
logrus.Info("Start subscribing containerd event")
c.eventMonitor.subscribe(c.client)
logrus.Infof("Start recovering state")
if err := c.recover(ctrdutil.NamespacedContext()); err != nil {
return errors.Wrap(err, "failed to recover state")
}
// Start event handler.
logrus.Info("Start event monitor")
eventMonitorErrCh := c.eventMonitor.start()
// Start snapshot stats syncer, it doesn't need to be stopped.
logrus.Info("Start snapshots syncer")
snapshotsSyncer := newSnapshotsSyncer(
c.snapshotStore,
c.client.SnapshotService(c.config.ContainerdConfig.Snapshotter),
time.Duration(c.config.StatsCollectPeriod)*time.Second,
)
snapshotsSyncer.start()
// Start CNI network conf syncer
logrus.Info("Start cni network conf syncer")
cniNetConfMonitorErrCh := make(chan error, 1)
go func() {
defer close(cniNetConfMonitorErrCh)
cniNetConfMonitorErrCh <- c.cniNetConfMonitor.syncLoop()
}()
// Start streaming server.
logrus.Info("Start streaming server")
streamServerErrCh := make(chan error)
go func() {
defer close(streamServerErrCh)
if err := c.streamServer.Start(true); err != nil && err != http.ErrServerClosed {
logrus.WithError(err).Error("Failed to start streaming server")
streamServerErrCh <- err
}
}()
// Set the server as initialized. GRPC services could start serving traffic.
c.initialized.Set()
var eventMonitorErr, streamServerErr, cniNetConfMonitorErr error
// Stop the whole CRI service if any of the critical service exits.
select {
case eventMonitorErr = <-eventMonitorErrCh:
case streamServerErr = <-streamServerErrCh:
case cniNetConfMonitorErr = <-cniNetConfMonitorErrCh:
}
if err := c.Close(); err != nil {
return errors.Wrap(err, "failed to stop cri service")
}
// If the error is set above, err from channel must be nil here, because
// the channel is supposed to be closed. Or else, we wait and set it.
if err := <-eventMonitorErrCh; err != nil {
eventMonitorErr = err
}
logrus.Info("Event monitor stopped")
// There is a race condition with http.Server.Serve.
// When `Close` is called at the same time with `Serve`, `Close`
// may finish first, and `Serve` may still block.
// See https://github.com/golang/go/issues/20239.
// Here we set a 2 second timeout for the stream server wait,
// if it timeout, an error log is generated.
// TODO(random-liu): Get rid of this after https://github.com/golang/go/issues/20239
// is fixed.
const streamServerStopTimeout = 2 * time.Second
select {
case err := <-streamServerErrCh:
if err != nil {
streamServerErr = err
}
logrus.Info("Stream server stopped")
case <-time.After(streamServerStopTimeout):
logrus.Errorf("Stream server is not stopped in %q", streamServerStopTimeout)
}
if eventMonitorErr != nil {
return errors.Wrap(eventMonitorErr, "event monitor error")
}
if streamServerErr != nil {
return errors.Wrap(streamServerErr, "stream server error")
}
if cniNetConfMonitorErr != nil {
return errors.Wrap(cniNetConfMonitorErr, "cni network conf monitor error")
}
return nil
}
// Close stops the CRI service.
// TODO(random-liu): Make close synchronous.
func (c *criService) Close() error {
logrus.Info("Stop CRI service")
if err := c.cniNetConfMonitor.stop(); err != nil {
logrus.WithError(err).Error("failed to stop cni network conf monitor")
}
c.eventMonitor.stop()
if err := c.streamServer.Stop(); err != nil {
return errors.Wrap(err, "failed to stop stream server")
}
return nil
}
func (c *criService) register(s *grpc.Server) error {
instrumented := newInstrumentedService(c)
runtime.RegisterRuntimeServiceServer(s, instrumented)
runtime.RegisterImageServiceServer(s, instrumented)
return nil
}
// imageFSPath returns containerd image filesystem path.
// Note that if containerd changes directory layout, we also needs to change this.
func imageFSPath(rootDir, snapshotter string) string {
return filepath.Join(rootDir, fmt.Sprintf("%s.%s", plugin.SnapshotPlugin, snapshotter))
}
func loadOCISpec(filename string) (*oci.Spec, error) {
file, err := os.Open(filename)
if err != nil {
return nil, errors.Wrapf(err, "failed to open base OCI spec: %s", filename)
}
defer file.Close()
spec := oci.Spec{}
if err := json.NewDecoder(file).Decode(&spec); err != nil {
return nil, errors.Wrap(err, "failed to parse base OCI spec file")
}
return &spec, nil
}
func loadBaseOCISpecs(config *criconfig.Config) (map[string]*oci.Spec, error) {
specs := map[string]*oci.Spec{}
for _, cfg := range config.Runtimes {
if cfg.BaseRuntimeSpec == "" {
continue
}
// Don't load same file twice
if _, ok := specs[cfg.BaseRuntimeSpec]; ok {
continue
}
spec, err := loadOCISpec(cfg.BaseRuntimeSpec)
if err != nil {
return nil, errors.Wrapf(err, "failed to load base OCI spec from file: %s", cfg.BaseRuntimeSpec)
}
specs[cfg.BaseRuntimeSpec] = spec
}
return specs, nil
}

View File

@@ -1,72 +0,0 @@
// +build !windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd/sys"
cni "github.com/containerd/go-cni"
"github.com/opencontainers/selinux/go-selinux"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
)
// networkAttachCount is the minimum number of networks the PodSandbox
// attaches to
const networkAttachCount = 2
// initPlatform handles linux specific initialization for the CRI service.
func (c *criService) initPlatform() error {
var err error
if sys.RunningInUserNS() {
if !(c.config.DisableCgroup && !c.apparmorEnabled() && c.config.RestrictOOMScoreAdj) {
logrus.Warn("Running containerd in a user namespace typically requires disable_cgroup, disable_apparmor, restrict_oom_score_adj set to be true")
}
}
if c.config.EnableSelinux {
if !selinux.GetEnabled() {
logrus.Warn("Selinux is not supported")
}
if r := c.config.SelinuxCategoryRange; r > 0 {
selinux.CategoryRange = uint32(r)
}
} else {
selinux.SetDisabled()
}
// Pod needs to attach to at least loopback network and a non host network,
// hence networkAttachCount is 2. If there are more network configs the
// pod will be attached to all the networks but we will only use the ip
// of the default network interface as the pod IP.
c.netPlugin, err = cni.New(cni.WithMinNetworkCount(networkAttachCount),
cni.WithPluginConfDir(c.config.NetworkPluginConfDir),
cni.WithPluginMaxConfNum(c.config.NetworkPluginMaxConfNum),
cni.WithPluginDir([]string{c.config.NetworkPluginBinDir}))
if err != nil {
return errors.Wrap(err, "failed to initialize cni")
}
return nil
}
// cniLoadOptions returns cni load options for the linux.
func (c *criService) cniLoadOptions() []cni.CNIOpt {
return []cni.CNIOpt{cni.WithLoNetwork, cni.WithDefaultConf}
}

View File

@@ -1,52 +0,0 @@
// +build windows
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
cni "github.com/containerd/go-cni"
"github.com/pkg/errors"
)
// windowsNetworkAttachCount is the minimum number of networks the PodSandbox
// attaches to
const windowsNetworkAttachCount = 1
// initPlatform handles linux specific initialization for the CRI service.
func (c *criService) initPlatform() error {
var err error
// For windows, the loopback network is added as default.
// There is no need to explicitly add one hence networkAttachCount is 1.
// If there are more network configs the pod will be attached to all the
// networks but we will only use the ip of the default network interface
// as the pod IP.
c.netPlugin, err = cni.New(cni.WithMinNetworkCount(windowsNetworkAttachCount),
cni.WithPluginConfDir(c.config.NetworkPluginConfDir),
cni.WithPluginMaxConfNum(c.config.NetworkPluginMaxConfNum),
cni.WithPluginDir([]string{c.config.NetworkPluginBinDir}))
if err != nil {
return errors.Wrap(err, "failed to initialize cni")
}
return nil
}
// cniLoadOptions returns cni load options for the windows.
func (c *criService) cniLoadOptions() []cni.CNIOpt {
return []cni.CNIOpt{cni.WithDefaultConf}
}

View File

@@ -1,120 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"context"
"time"
"github.com/containerd/containerd/errdefs"
snapshot "github.com/containerd/containerd/snapshots"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
snapshotstore "github.com/containerd/cri/pkg/store/snapshot"
)
// snapshotsSyncer syncs snapshot stats periodically. imagefs info and container stats
// should both use cached result here.
// TODO(random-liu): Benchmark with high workload. We may need a statsSyncer instead if
// benchmark result shows that container cpu/memory stats also need to be cached.
type snapshotsSyncer struct {
store *snapshotstore.Store
snapshotter snapshot.Snapshotter
syncPeriod time.Duration
}
// newSnapshotsSyncer creates a snapshot syncer.
func newSnapshotsSyncer(store *snapshotstore.Store, snapshotter snapshot.Snapshotter,
period time.Duration) *snapshotsSyncer {
return &snapshotsSyncer{
store: store,
snapshotter: snapshotter,
syncPeriod: period,
}
}
// start starts the snapshots syncer. No stop function is needed because
// the syncer doesn't update any persistent states, it's fine to let it
// exit with the process.
func (s *snapshotsSyncer) start() {
tick := time.NewTicker(s.syncPeriod)
go func() {
defer tick.Stop()
// TODO(random-liu): This is expensive. We should do benchmark to
// check the resource usage and optimize this.
for {
if err := s.sync(); err != nil {
logrus.WithError(err).Error("Failed to sync snapshot stats")
}
<-tick.C
}
}()
}
// sync updates all snapshots stats.
func (s *snapshotsSyncer) sync() error {
ctx := ctrdutil.NamespacedContext()
start := time.Now().UnixNano()
var snapshots []snapshot.Info
// Do not call `Usage` directly in collect function, because
// `Usage` takes time, we don't want `Walk` to hold read lock
// of snapshot metadata store for too long time.
// TODO(random-liu): Set timeout for the following 2 contexts.
if err := s.snapshotter.Walk(ctx, func(ctx context.Context, info snapshot.Info) error {
snapshots = append(snapshots, info)
return nil
}); err != nil {
return errors.Wrap(err, "walk all snapshots failed")
}
for _, info := range snapshots {
sn, err := s.store.Get(info.Name)
if err == nil {
// Only update timestamp for non-active snapshot.
if sn.Kind == info.Kind && sn.Kind != snapshot.KindActive {
sn.Timestamp = time.Now().UnixNano()
s.store.Add(sn)
continue
}
}
// Get newest stats if the snapshot is new or active.
sn = snapshotstore.Snapshot{
Key: info.Name,
Kind: info.Kind,
Timestamp: time.Now().UnixNano(),
}
usage, err := s.snapshotter.Usage(ctx, info.Name)
if err != nil {
if !errdefs.IsNotFound(err) {
logrus.WithError(err).Errorf("Failed to get usage for snapshot %q", info.Name)
}
continue
}
sn.Size = uint64(usage.Size)
sn.Inodes = uint64(usage.Inodes)
s.store.Add(sn)
}
for _, sn := range s.store.List() {
if sn.Timestamp >= start {
continue
}
// Delete the snapshot stats if it's not updated this time.
s.store.Delete(sn.Key)
}
return nil
}

View File

@@ -1,83 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"encoding/json"
"fmt"
goruntime "runtime"
"github.com/containerd/containerd/log"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// networkNotReadyReason is the reason reported when network is not ready.
const networkNotReadyReason = "NetworkPluginNotReady"
// Status returns the status of the runtime.
func (c *criService) Status(ctx context.Context, r *runtime.StatusRequest) (*runtime.StatusResponse, error) {
// As a containerd plugin, if CRI plugin is serving request,
// containerd must be ready.
runtimeCondition := &runtime.RuntimeCondition{
Type: runtime.RuntimeReady,
Status: true,
}
networkCondition := &runtime.RuntimeCondition{
Type: runtime.NetworkReady,
Status: true,
}
// Check the status of the cni initialization
if err := c.netPlugin.Status(); err != nil {
networkCondition.Status = false
networkCondition.Reason = networkNotReadyReason
networkCondition.Message = fmt.Sprintf("Network plugin returns error: %v", err)
}
resp := &runtime.StatusResponse{
Status: &runtime.RuntimeStatus{Conditions: []*runtime.RuntimeCondition{
runtimeCondition,
networkCondition,
}},
}
if r.Verbose {
configByt, err := json.Marshal(c.config)
if err != nil {
return nil, err
}
resp.Info = make(map[string]string)
resp.Info["config"] = string(configByt)
versionByt, err := json.Marshal(goruntime.Version())
if err != nil {
return nil, err
}
resp.Info["golang"] = string(versionByt)
cniConfig, err := json.Marshal(c.netPlugin.GetConfig())
if err != nil {
log.G(ctx).WithError(err).Errorf("Failed to marshal CNI config %v", err)
}
resp.Info["cniconfig"] = string(cniConfig)
lastCNILoadStatus := "OK"
if lerr := c.cniNetConfMonitor.lastStatus(); lerr != nil {
lastCNILoadStatus = lerr.Error()
}
resp.Info["lastCNILoadStatus"] = lastCNILoadStatus
}
return resp, nil
}

View File

@@ -1,239 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"context"
"crypto/tls"
"io"
"math"
"net"
"os"
"time"
"github.com/pkg/errors"
k8snet "k8s.io/apimachinery/pkg/util/net"
"k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/client-go/tools/remotecommand"
k8scert "k8s.io/client-go/util/cert"
"k8s.io/utils/exec"
ctrdutil "github.com/containerd/cri/pkg/containerd/util"
"github.com/containerd/cri/pkg/streaming"
)
type streamListenerMode int
const (
x509KeyPairTLS streamListenerMode = iota
selfSignTLS
withoutTLS
)
func getStreamListenerMode(c *criService) (streamListenerMode, error) {
if c.config.EnableTLSStreaming {
if c.config.X509KeyPairStreaming.TLSCertFile != "" && c.config.X509KeyPairStreaming.TLSKeyFile != "" {
return x509KeyPairTLS, nil
}
if c.config.X509KeyPairStreaming.TLSCertFile != "" && c.config.X509KeyPairStreaming.TLSKeyFile == "" {
return -1, errors.New("must set X509KeyPairStreaming.TLSKeyFile")
}
if c.config.X509KeyPairStreaming.TLSCertFile == "" && c.config.X509KeyPairStreaming.TLSKeyFile != "" {
return -1, errors.New("must set X509KeyPairStreaming.TLSCertFile")
}
return selfSignTLS, nil
}
if c.config.X509KeyPairStreaming.TLSCertFile != "" {
return -1, errors.New("X509KeyPairStreaming.TLSCertFile is set but EnableTLSStreaming is not set")
}
if c.config.X509KeyPairStreaming.TLSKeyFile != "" {
return -1, errors.New("X509KeyPairStreaming.TLSKeyFile is set but EnableTLSStreaming is not set")
}
return withoutTLS, nil
}
func newStreamServer(c *criService, addr, port, streamIdleTimeout string) (streaming.Server, error) {
if addr == "" {
a, err := k8snet.ResolveBindAddress(nil)
if err != nil {
return nil, errors.Wrap(err, "failed to get stream server address")
}
addr = a.String()
}
config := streaming.DefaultConfig
if streamIdleTimeout != "" {
var err error
config.StreamIdleTimeout, err = time.ParseDuration(streamIdleTimeout)
if err != nil {
return nil, errors.Wrap(err, "invalid stream idle timeout")
}
}
config.Addr = net.JoinHostPort(addr, port)
run := newStreamRuntime(c)
tlsMode, err := getStreamListenerMode(c)
if err != nil {
return nil, errors.Wrapf(err, "invalid stream server configuration")
}
switch tlsMode {
case x509KeyPairTLS:
tlsCert, err := tls.LoadX509KeyPair(c.config.X509KeyPairStreaming.TLSCertFile, c.config.X509KeyPairStreaming.TLSKeyFile)
if err != nil {
return nil, errors.Wrap(err, "failed to load x509 key pair for stream server")
}
config.TLSConfig = &tls.Config{
Certificates: []tls.Certificate{tlsCert},
}
return streaming.NewServer(config, run)
case selfSignTLS:
tlsCert, err := newTLSCert()
if err != nil {
return nil, errors.Wrap(err, "failed to generate tls certificate for stream server")
}
config.TLSConfig = &tls.Config{
Certificates: []tls.Certificate{tlsCert},
InsecureSkipVerify: true,
}
return streaming.NewServer(config, run)
case withoutTLS:
return streaming.NewServer(config, run)
default:
return nil, errors.New("invalid configuration for the stream listener")
}
}
type streamRuntime struct {
c *criService
}
func newStreamRuntime(c *criService) streaming.Runtime {
return &streamRuntime{c: c}
}
// Exec executes a command inside the container. exec.ExitError is returned if the command
// returns non-zero exit code.
func (s *streamRuntime) Exec(containerID string, cmd []string, stdin io.Reader, stdout, stderr io.WriteCloser,
tty bool, resize <-chan remotecommand.TerminalSize) error {
exitCode, err := s.c.execInContainer(ctrdutil.NamespacedContext(), containerID, execOptions{
cmd: cmd,
stdin: stdin,
stdout: stdout,
stderr: stderr,
tty: tty,
resize: resize,
})
if err != nil {
return errors.Wrap(err, "failed to exec in container")
}
if *exitCode == 0 {
return nil
}
return &exec.CodeExitError{
Err: errors.Errorf("error executing command %v, exit code %d", cmd, *exitCode),
Code: int(*exitCode),
}
}
func (s *streamRuntime) Attach(containerID string, in io.Reader, out, err io.WriteCloser, tty bool,
resize <-chan remotecommand.TerminalSize) error {
return s.c.attachContainer(ctrdutil.NamespacedContext(), containerID, in, out, err, tty, resize)
}
func (s *streamRuntime) PortForward(podSandboxID string, port int32, stream io.ReadWriteCloser) error {
if port <= 0 || port > math.MaxUint16 {
return errors.Errorf("invalid port %d", port)
}
ctx := ctrdutil.NamespacedContext()
return s.c.portForward(ctx, podSandboxID, port, stream)
}
// handleResizing spawns a goroutine that processes the resize channel, calling resizeFunc for each
// remotecommand.TerminalSize received from the channel.
func handleResizing(ctx context.Context, resize <-chan remotecommand.TerminalSize, resizeFunc func(size remotecommand.TerminalSize)) {
if resize == nil {
return
}
go func() {
defer runtime.HandleCrash()
for {
select {
case <-ctx.Done():
return
case size, ok := <-resize:
if !ok {
return
}
if size.Height < 1 || size.Width < 1 {
continue
}
resizeFunc(size)
}
}
}()
}
// newTLSCert returns a self CA signed tls.certificate.
// TODO (mikebrow): replace / rewrite this function to support using CA
// signing of the certificate. Requires a security plan for kubernetes regarding
// CRI connections / streaming, etc. For example, kubernetes could configure or
// require a CA service and pass a configuration down through CRI.
func newTLSCert() (tls.Certificate, error) {
fail := func(err error) (tls.Certificate, error) { return tls.Certificate{}, err }
hostName, err := os.Hostname()
if err != nil {
return fail(errors.Wrap(err, "failed to get hostname"))
}
addrs, err := net.InterfaceAddrs()
if err != nil {
return fail(errors.Wrap(err, "failed to get host IP addresses"))
}
var alternateIPs []net.IP
var alternateDNS []string
for _, addr := range addrs {
var ip net.IP
switch v := addr.(type) {
case *net.IPNet:
ip = v.IP
case *net.IPAddr:
ip = v.IP
default:
continue
}
alternateIPs = append(alternateIPs, ip)
alternateDNS = append(alternateDNS, ip.String())
}
// Generate a self signed certificate key (CA is self)
certPem, keyPem, err := k8scert.GenerateSelfSignedCertKey(hostName, alternateIPs, alternateDNS)
if err != nil {
return fail(errors.Wrap(err, "certificate key could not be created"))
}
// Load the tls certificate
tlsCert, err := tls.X509KeyPair(certPem, keyPem)
if err != nil {
return fail(errors.Wrap(err, "certificate could not be loaded"))
}
return tlsCert, nil
}

View File

@@ -1,128 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"net"
"os"
"path/filepath"
"strings"
"text/template"
"github.com/containerd/containerd/log"
"github.com/pkg/errors"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// cniConfigTemplate contains the values containerd will overwrite
// in the cni config template.
type cniConfigTemplate struct {
// PodCIDR is the cidr for pods on the node.
PodCIDR string
// PodCIDRRanges is the cidr ranges for pods on the node.
PodCIDRRanges []string
// Routes is a list of routes configured.
Routes []string
}
const (
// cniConfigFileName is the name of cni config file generated by containerd.
cniConfigFileName = "10-containerd-net.conflist"
// zeroCIDRv6 is the null route for IPv6.
zeroCIDRv6 = "::/0"
// zeroCIDRv4 is the null route for IPv4.
zeroCIDRv4 = "0.0.0.0/0"
)
// UpdateRuntimeConfig updates the runtime config. Currently only handles podCIDR updates.
func (c *criService) UpdateRuntimeConfig(ctx context.Context, r *runtime.UpdateRuntimeConfigRequest) (*runtime.UpdateRuntimeConfigResponse, error) {
podCIDRs := r.GetRuntimeConfig().GetNetworkConfig().GetPodCidr()
if podCIDRs == "" {
return &runtime.UpdateRuntimeConfigResponse{}, nil
}
cidrs := strings.Split(podCIDRs, ",")
for i := range cidrs {
cidrs[i] = strings.TrimSpace(cidrs[i])
}
routes, err := getRoutes(cidrs)
if err != nil {
return nil, errors.Wrap(err, "get routes")
}
confTemplate := c.config.NetworkPluginConfTemplate
if confTemplate == "" {
log.G(ctx).Info("No cni config template is specified, wait for other system components to drop the config.")
return &runtime.UpdateRuntimeConfigResponse{}, nil
}
if err := c.netPlugin.Status(); err == nil {
log.G(ctx).Infof("Network plugin is ready, skip generating cni config from template %q", confTemplate)
return &runtime.UpdateRuntimeConfigResponse{}, nil
} else if err := c.netPlugin.Load(c.cniLoadOptions()...); err == nil {
log.G(ctx).Infof("CNI config is successfully loaded, skip generating cni config from template %q", confTemplate)
return &runtime.UpdateRuntimeConfigResponse{}, nil
}
log.G(ctx).Infof("Generating cni config from template %q", confTemplate)
// generate cni config file from the template with updated pod cidr.
t, err := template.ParseFiles(confTemplate)
if err != nil {
return nil, errors.Wrapf(err, "failed to parse cni config template %q", confTemplate)
}
if err := os.MkdirAll(c.config.NetworkPluginConfDir, 0755); err != nil {
return nil, errors.Wrapf(err, "failed to create cni config directory: %q", c.config.NetworkPluginConfDir)
}
confFile := filepath.Join(c.config.NetworkPluginConfDir, cniConfigFileName)
f, err := os.OpenFile(confFile, os.O_WRONLY|os.O_CREATE, 0644)
if err != nil {
return nil, errors.Wrapf(err, "failed to open cni config file %q", confFile)
}
defer f.Close()
if err := t.Execute(f, cniConfigTemplate{
PodCIDR: cidrs[0],
PodCIDRRanges: cidrs,
Routes: routes,
}); err != nil {
return nil, errors.Wrapf(err, "failed to generate cni config file %q", confFile)
}
return &runtime.UpdateRuntimeConfigResponse{}, nil
}
// getRoutes generates required routes for the passed in cidrs.
func getRoutes(cidrs []string) ([]string, error) {
var (
routes []string
hasV4, hasV6 bool
)
for _, c := range cidrs {
_, cidr, err := net.ParseCIDR(c)
if err != nil {
return nil, err
}
if cidr.IP.To4() != nil {
hasV4 = true
} else {
hasV6 = true
}
}
if hasV4 {
routes = append(routes, zeroCIDRv4)
}
if hasV6 {
routes = append(routes, zeroCIDRv6)
}
return routes, nil
}

View File

@@ -1,42 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package server
import (
"github.com/containerd/containerd/version"
"golang.org/x/net/context"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
"github.com/containerd/cri/pkg/constants"
)
const (
containerName = "containerd"
// kubeAPIVersion is the api version of kubernetes.
// TODO(random-liu): Change this to actual CRI version.
kubeAPIVersion = "0.1.0"
)
// Version returns the runtime name, runtime version and runtime API version.
func (c *criService) Version(ctx context.Context, r *runtime.VersionRequest) (*runtime.VersionResponse, error) {
return &runtime.VersionResponse{
Version: kubeAPIVersion,
RuntimeName: containerName,
RuntimeVersion: version.Version,
RuntimeApiVersion: constants.CRIVersion,
}, nil
}

View File

@@ -1,71 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package seutil
import (
"bufio"
"os"
"github.com/opencontainers/selinux/go-selinux"
)
var seTypes map[string]struct{}
const typePath = "/etc/selinux/targeted/contexts/customizable_types"
func init() {
seTypes = make(map[string]struct{})
if !selinux.GetEnabled() {
return
}
f, err := os.Open(typePath)
if err != nil {
return
}
defer f.Close()
s := bufio.NewScanner(f)
for s.Scan() {
seTypes[s.Text()] = struct{}{}
}
}
// HasType returns true if the underlying system has the
// provided selinux type enabled.
func HasType(name string) bool {
_, ok := seTypes[name]
return ok
}
// ChangeToKVM process label
func ChangeToKVM(l string) (string, error) {
if l == "" || !selinux.GetEnabled() {
return "", nil
}
proc, _ := selinux.KVMContainerLabels()
selinux.ReleaseLabel(proc)
current, err := selinux.NewContext(l)
if err != nil {
return "", err
}
next, err := selinux.NewContext(proc)
if err != nil {
return "", err
}
current["type"] = next["type"]
return current.Get(), nil
}

View File

@@ -1,177 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package container
import (
"sync"
"github.com/containerd/containerd"
"github.com/containerd/cri/pkg/store/label"
"github.com/docker/docker/pkg/truncindex"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
cio "github.com/containerd/cri/pkg/server/io"
"github.com/containerd/cri/pkg/store"
)
// Container contains all resources associated with the container. All methods to
// mutate the internal state are thread-safe.
type Container struct {
// Metadata is the metadata of the container, it is **immutable** after created.
Metadata
// Status stores the status of the container.
Status StatusStorage
// Container is the containerd container client.
Container containerd.Container
// Container IO.
// IO could only be nil when the container is in unknown state.
IO *cio.ContainerIO
// StopCh is used to propagate the stop information of the container.
*store.StopCh
}
// Opts sets specific information to newly created Container.
type Opts func(*Container) error
// WithContainer adds the containerd Container to the internal data store.
func WithContainer(cntr containerd.Container) Opts {
return func(c *Container) error {
c.Container = cntr
return nil
}
}
// WithContainerIO adds IO into the container.
func WithContainerIO(io *cio.ContainerIO) Opts {
return func(c *Container) error {
c.IO = io
return nil
}
}
// WithStatus adds status to the container.
func WithStatus(status Status, root string) Opts {
return func(c *Container) error {
s, err := StoreStatus(root, c.ID, status)
if err != nil {
return err
}
c.Status = s
if s.Get().State() == runtime.ContainerState_CONTAINER_EXITED {
c.Stop()
}
return nil
}
}
// NewContainer creates an internally used container type.
func NewContainer(metadata Metadata, opts ...Opts) (Container, error) {
c := Container{
Metadata: metadata,
StopCh: store.NewStopCh(),
}
for _, o := range opts {
if err := o(&c); err != nil {
return Container{}, err
}
}
return c, nil
}
// Delete deletes checkpoint for the container.
func (c *Container) Delete() error {
return c.Status.Delete()
}
// Store stores all Containers.
type Store struct {
lock sync.RWMutex
containers map[string]Container
idIndex *truncindex.TruncIndex
labels *label.Store
}
// NewStore creates a container store.
func NewStore(labels *label.Store) *Store {
return &Store{
containers: make(map[string]Container),
idIndex: truncindex.NewTruncIndex([]string{}),
labels: labels,
}
}
// Add a container into the store. Returns store.ErrAlreadyExist if the
// container already exists.
func (s *Store) Add(c Container) error {
s.lock.Lock()
defer s.lock.Unlock()
if _, ok := s.containers[c.ID]; ok {
return store.ErrAlreadyExist
}
if err := s.labels.Reserve(c.ProcessLabel); err != nil {
return err
}
if err := s.idIndex.Add(c.ID); err != nil {
return err
}
s.containers[c.ID] = c
return nil
}
// Get returns the container with specified id. Returns store.ErrNotExist
// if the container doesn't exist.
func (s *Store) Get(id string) (Container, error) {
s.lock.RLock()
defer s.lock.RUnlock()
id, err := s.idIndex.Get(id)
if err != nil {
if err == truncindex.ErrNotExist {
err = store.ErrNotExist
}
return Container{}, err
}
if c, ok := s.containers[id]; ok {
return c, nil
}
return Container{}, store.ErrNotExist
}
// List lists all containers.
func (s *Store) List() []Container {
s.lock.RLock()
defer s.lock.RUnlock()
var containers []Container
for _, c := range s.containers {
containers = append(containers, c)
}
return containers
}
// Delete deletes the container from store with specified id.
func (s *Store) Delete(id string) {
s.lock.Lock()
defer s.lock.Unlock()
id, err := s.idIndex.Get(id)
if err != nil {
// Note: The idIndex.Delete and delete doesn't handle truncated index.
// So we need to return if there are error.
return
}
s.labels.Release(s.containers[id].ProcessLabel)
s.idIndex.Delete(id) // nolint: errcheck
delete(s.containers, id)
}

View File

@@ -1,62 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package container
import "sync"
// WithFakeStatus adds fake status to the container.
func WithFakeStatus(status Status) Opts {
return func(c *Container) error {
c.Status = &fakeStatusStorage{status: status}
if status.FinishedAt != 0 {
// Fake the TaskExit event
c.Stop()
}
return nil
}
}
// fakeStatusStorage is a fake status storage for testing.
type fakeStatusStorage struct {
sync.RWMutex
status Status
}
func (f *fakeStatusStorage) Get() Status {
f.RLock()
defer f.RUnlock()
return f.status
}
func (f *fakeStatusStorage) UpdateSync(u UpdateFunc) error {
return f.Update(u)
}
func (f *fakeStatusStorage) Update(u UpdateFunc) error {
f.Lock()
defer f.Unlock()
newStatus, err := u(f.status)
if err != nil {
return err
}
f.status = newStatus
return nil
}
func (f *fakeStatusStorage) Delete() error {
return nil
}

View File

@@ -1,89 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package container
import (
"encoding/json"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// NOTE(random-liu):
// 1) Metadata is immutable after created.
// 2) Metadata is checkpointed as containerd container label.
// metadataVersion is current version of container metadata.
const metadataVersion = "v1" // nolint
// versionedMetadata is the internal versioned container metadata.
// nolint
type versionedMetadata struct {
// Version indicates the version of the versioned container metadata.
Version string
// Metadata's type is metadataInternal. If not there will be a recursive call in MarshalJSON.
Metadata metadataInternal
}
// metadataInternal is for internal use.
type metadataInternal Metadata
// Metadata is the unversioned container metadata.
type Metadata struct {
// ID is the container id.
ID string
// Name is the container name.
Name string
// SandboxID is the sandbox id the container belongs to.
SandboxID string
// Config is the CRI container config.
// NOTE(random-liu): Resource limits are updatable, the source
// of truth for resource limits are in containerd.
Config *runtime.ContainerConfig
// ImageRef is the reference of image used by the container.
ImageRef string
// LogPath is the container log path.
LogPath string
// StopSignal is the system call signal that will be sent to the container to exit.
// TODO(random-liu): Add integration test for stop signal.
StopSignal string
// ProcessLabel is the SELinux process label for the container
ProcessLabel string
}
// MarshalJSON encodes Metadata into bytes in json format.
func (c *Metadata) MarshalJSON() ([]byte, error) {
return json.Marshal(&versionedMetadata{
Version: metadataVersion,
Metadata: metadataInternal(*c),
})
}
// UnmarshalJSON decodes Metadata from bytes.
func (c *Metadata) UnmarshalJSON(data []byte) error {
versioned := &versionedMetadata{}
if err := json.Unmarshal(data, versioned); err != nil {
return err
}
// Handle old version after upgrade.
switch versioned.Version {
case metadataVersion:
*c = Metadata(versioned.Metadata)
return nil
}
return errors.Errorf("unsupported version: %q", versioned.Version)
}

View File

@@ -1,247 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package container
import (
"encoding/json"
"io/ioutil"
"os"
"path/filepath"
"sync"
"github.com/containerd/continuity"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// The container state machine in the CRI plugin:
//
// + +
// | |
// | Create | Load
// | |
// +----v----+ |
// | | |
// | CREATED <---------+-----------+
// | | | |
// +----+----- | |
// | | |
// | Start | |
// | | |
// +----v----+ | |
// Exec +--------+ | | |
// Attach | | RUNNING <---------+ |
// LogReopen +--------> | | |
// +----+----+ | |
// | | |
// | Stop/Exit | |
// | | |
// +----v----+ | |
// | <---------+ +----v----+
// | EXITED | | |
// | <----------------+ UNKNOWN |
// +----+----+ Stop | |
// | +---------+
// | Remove
// v
// DELETED
// statusVersion is current version of container status.
const statusVersion = "v1" // nolint
// versionedStatus is the internal used versioned container status.
// nolint
type versionedStatus struct {
// Version indicates the version of the versioned container status.
Version string
Status
}
// Status is the status of a container.
type Status struct {
// Pid is the init process id of the container.
Pid uint32
// CreatedAt is the created timestamp.
CreatedAt int64
// StartedAt is the started timestamp.
StartedAt int64
// FinishedAt is the finished timestamp.
FinishedAt int64
// ExitCode is the container exit code.
ExitCode int32
// CamelCase string explaining why container is in its current state.
Reason string
// Human-readable message indicating details about why container is in its
// current state.
Message string
// Starting indicates that the container is in starting state.
// This field doesn't need to be checkpointed.
Starting bool `json:"-"`
// Removing indicates that the container is in removing state.
// This field doesn't need to be checkpointed.
Removing bool `json:"-"`
// Unknown indicates that the container status is not fully loaded.
// This field doesn't need to be checkpointed.
Unknown bool `json:"-"`
}
// State returns current state of the container based on the container status.
func (s Status) State() runtime.ContainerState {
if s.Unknown {
return runtime.ContainerState_CONTAINER_UNKNOWN
}
if s.FinishedAt != 0 {
return runtime.ContainerState_CONTAINER_EXITED
}
if s.StartedAt != 0 {
return runtime.ContainerState_CONTAINER_RUNNING
}
if s.CreatedAt != 0 {
return runtime.ContainerState_CONTAINER_CREATED
}
return runtime.ContainerState_CONTAINER_UNKNOWN
}
// encode encodes Status into bytes in json format.
func (s *Status) encode() ([]byte, error) {
return json.Marshal(&versionedStatus{
Version: statusVersion,
Status: *s,
})
}
// decode decodes Status from bytes.
func (s *Status) decode(data []byte) error {
versioned := &versionedStatus{}
if err := json.Unmarshal(data, versioned); err != nil {
return err
}
// Handle old version after upgrade.
switch versioned.Version {
case statusVersion:
*s = versioned.Status
return nil
}
return errors.New("unsupported version")
}
// UpdateFunc is function used to update the container status. If there
// is an error, the update will be rolled back.
type UpdateFunc func(Status) (Status, error)
// StatusStorage manages the container status with a storage backend.
type StatusStorage interface {
// Get a container status.
Get() Status
// UpdateSync updates the container status and the on disk checkpoint.
// Note that the update MUST be applied in one transaction.
UpdateSync(UpdateFunc) error
// Update the container status. Note that the update MUST be applied
// in one transaction.
Update(UpdateFunc) error
// Delete the container status.
// Note:
// * Delete should be idempotent.
// * The status must be deleted in one trasaction.
Delete() error
}
// StoreStatus creates the storage containing the passed in container status with the
// specified id.
// The status MUST be created in one transaction.
func StoreStatus(root, id string, status Status) (StatusStorage, error) {
data, err := status.encode()
if err != nil {
return nil, errors.Wrap(err, "failed to encode status")
}
path := filepath.Join(root, "status")
if err := continuity.AtomicWriteFile(path, data, 0600); err != nil {
return nil, errors.Wrapf(err, "failed to checkpoint status to %q", path)
}
return &statusStorage{
path: path,
status: status,
}, nil
}
// LoadStatus loads container status from checkpoint. There shouldn't be threads
// writing to the file during loading.
func LoadStatus(root, id string) (Status, error) {
path := filepath.Join(root, "status")
data, err := ioutil.ReadFile(path)
if err != nil {
return Status{}, errors.Wrapf(err, "failed to read status from %q", path)
}
var status Status
if err := status.decode(data); err != nil {
return Status{}, errors.Wrapf(err, "failed to decode status %q", data)
}
return status, nil
}
type statusStorage struct {
sync.RWMutex
path string
status Status
}
// Get a copy of container status.
func (s *statusStorage) Get() Status {
s.RLock()
defer s.RUnlock()
return s.status
}
// UpdateSync updates the container status and the on disk checkpoint.
func (s *statusStorage) UpdateSync(u UpdateFunc) error {
s.Lock()
defer s.Unlock()
newStatus, err := u(s.status)
if err != nil {
return err
}
data, err := newStatus.encode()
if err != nil {
return errors.Wrap(err, "failed to encode status")
}
if err := continuity.AtomicWriteFile(s.path, data, 0600); err != nil {
return errors.Wrapf(err, "failed to checkpoint status to %q", s.path)
}
s.status = newStatus
return nil
}
// Update the container status.
func (s *statusStorage) Update(u UpdateFunc) error {
s.Lock()
defer s.Unlock()
newStatus, err := u(s.status)
if err != nil {
return err
}
s.status = newStatus
return nil
}
// Delete deletes the container status from disk atomically.
func (s *statusStorage) Delete() error {
temp := filepath.Dir(s.path) + ".del-" + filepath.Base(s.path)
if err := os.Rename(s.path, temp); err != nil && !os.IsNotExist(err) {
return err
}
return os.RemoveAll(temp)
}

View File

@@ -1,33 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package store
import "github.com/containerd/containerd/errdefs"
var (
// ErrAlreadyExist is the error returned when data added in the store
// already exists.
//
// This error has been DEPRECATED and will be removed in 1.5. Please switch
// usage directly to `errdefs.ErrAlreadyExists`.
ErrAlreadyExist = errdefs.ErrAlreadyExists
// ErrNotExist is the error returned when data is not in the store.
//
// This error has been DEPRECATED and will be removed in 1.5. Please switch
// usage directly to `errdefs.ErrNotFound`.
ErrNotExist = errdefs.ErrNotFound
)

View File

@@ -1,34 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package image
import "github.com/pkg/errors"
// NewFakeStore returns an image store with predefined images.
// Update is not allowed for this fake store.
func NewFakeStore(images []Image) (*Store, error) {
s := NewStore(nil)
for _, i := range images {
for _, ref := range i.References {
s.refCache[ref] = i.ID
}
if err := s.store.add(i); err != nil {
return nil, errors.Wrapf(err, "add image %+v", i)
}
}
return s, nil
}

View File

@@ -1,256 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package image
import (
"context"
"encoding/json"
"sync"
"github.com/containerd/containerd"
"github.com/containerd/containerd/content"
"github.com/containerd/containerd/errdefs"
imagedigest "github.com/opencontainers/go-digest"
"github.com/opencontainers/go-digest/digestset"
imageidentity "github.com/opencontainers/image-spec/identity"
imagespec "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/pkg/errors"
storeutil "github.com/containerd/cri/pkg/store"
"github.com/containerd/cri/pkg/util"
)
// Image contains all resources associated with the image. All fields
// MUST not be mutated directly after created.
type Image struct {
// Id of the image. Normally the digest of image config.
ID string
// References are references to the image, e.g. RepoTag and RepoDigest.
References []string
// ChainID is the chainID of the image.
ChainID string
// Size is the compressed size of the image.
Size int64
// ImageSpec is the oci image structure which describes basic information about the image.
ImageSpec imagespec.Image
}
// Store stores all images.
type Store struct {
lock sync.RWMutex
// refCache is a containerd image reference to image id cache.
refCache map[string]string
// client is the containerd client.
client *containerd.Client
// store is the internal image store indexed by image id.
store *store
}
// NewStore creates an image store.
func NewStore(client *containerd.Client) *Store {
return &Store{
refCache: make(map[string]string),
client: client,
store: &store{
images: make(map[string]Image),
digestSet: digestset.NewSet(),
},
}
}
// Update updates cache for a reference.
func (s *Store) Update(ctx context.Context, ref string) error {
s.lock.Lock()
defer s.lock.Unlock()
i, err := s.client.GetImage(ctx, ref)
if err != nil && !errdefs.IsNotFound(err) {
return errors.Wrap(err, "get image from containerd")
}
var img *Image
if err == nil {
img, err = getImage(ctx, i)
if err != nil {
return errors.Wrap(err, "get image info from containerd")
}
}
return s.update(ref, img)
}
// update updates the internal cache. img == nil means that
// the image does not exist in containerd.
func (s *Store) update(ref string, img *Image) error {
oldID, oldExist := s.refCache[ref]
if img == nil {
// The image reference doesn't exist in containerd.
if oldExist {
// Remove the reference from the store.
s.store.delete(oldID, ref)
delete(s.refCache, ref)
}
return nil
}
if oldExist {
if oldID == img.ID {
return nil
}
// Updated. Remove tag from old image.
s.store.delete(oldID, ref)
}
// New image. Add new image.
s.refCache[ref] = img.ID
return s.store.add(*img)
}
// getImage gets image information from containerd.
func getImage(ctx context.Context, i containerd.Image) (*Image, error) {
// Get image information.
diffIDs, err := i.RootFS(ctx)
if err != nil {
return nil, errors.Wrap(err, "get image diffIDs")
}
chainID := imageidentity.ChainID(diffIDs)
size, err := i.Size(ctx)
if err != nil {
return nil, errors.Wrap(err, "get image compressed resource size")
}
desc, err := i.Config(ctx)
if err != nil {
return nil, errors.Wrap(err, "get image config descriptor")
}
id := desc.Digest.String()
rb, err := content.ReadBlob(ctx, i.ContentStore(), desc)
if err != nil {
return nil, errors.Wrap(err, "read image config from content store")
}
var ociimage imagespec.Image
if err := json.Unmarshal(rb, &ociimage); err != nil {
return nil, errors.Wrapf(err, "unmarshal image config %s", rb)
}
return &Image{
ID: id,
References: []string{i.Name()},
ChainID: chainID.String(),
Size: size,
ImageSpec: ociimage,
}, nil
}
// Resolve resolves a image reference to image id.
func (s *Store) Resolve(ref string) (string, error) {
s.lock.RLock()
defer s.lock.RUnlock()
id, ok := s.refCache[ref]
if !ok {
return "", storeutil.ErrNotExist
}
return id, nil
}
// Get gets image metadata by image id. The id can be truncated.
// Returns various validation errors if the image id is invalid.
// Returns storeutil.ErrNotExist if the image doesn't exist.
func (s *Store) Get(id string) (Image, error) {
return s.store.get(id)
}
// List lists all images.
func (s *Store) List() []Image {
return s.store.list()
}
type store struct {
lock sync.RWMutex
images map[string]Image
digestSet *digestset.Set
}
func (s *store) list() []Image {
s.lock.RLock()
defer s.lock.RUnlock()
var images []Image
for _, i := range s.images {
images = append(images, i)
}
return images
}
func (s *store) add(img Image) error {
s.lock.Lock()
defer s.lock.Unlock()
if _, err := s.digestSet.Lookup(img.ID); err != nil {
if err != digestset.ErrDigestNotFound {
return err
}
if err := s.digestSet.Add(imagedigest.Digest(img.ID)); err != nil {
return err
}
}
i, ok := s.images[img.ID]
if !ok {
// If the image doesn't exist, add it.
s.images[img.ID] = img
return nil
}
// Or else, merge the references.
i.References = util.MergeStringSlices(i.References, img.References)
s.images[img.ID] = i
return nil
}
func (s *store) get(id string) (Image, error) {
s.lock.RLock()
defer s.lock.RUnlock()
digest, err := s.digestSet.Lookup(id)
if err != nil {
if err == digestset.ErrDigestNotFound {
err = storeutil.ErrNotExist
}
return Image{}, err
}
if i, ok := s.images[digest.String()]; ok {
return i, nil
}
return Image{}, storeutil.ErrNotExist
}
func (s *store) delete(id, ref string) {
s.lock.Lock()
defer s.lock.Unlock()
digest, err := s.digestSet.Lookup(id)
if err != nil {
// Note: The idIndex.Delete and delete doesn't handle truncated index.
// So we need to return if there are error.
return
}
i, ok := s.images[digest.String()]
if !ok {
return
}
i.References = util.SubtractStringSlice(i.References, ref)
if len(i.References) != 0 {
s.images[digest.String()] = i
return
}
// Remove the image if it is not referenced any more.
s.digestSet.Remove(digest) // nolint: errcheck
delete(s.images, digest.String())
}

View File

@@ -1,90 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package label
import (
"sync"
"github.com/opencontainers/selinux/go-selinux"
)
type Store struct {
sync.Mutex
levels map[string]int
Releaser func(string)
Reserver func(string)
}
func NewStore() *Store {
return &Store{
levels: map[string]int{},
Releaser: selinux.ReleaseLabel,
Reserver: selinux.ReserveLabel,
}
}
func (s *Store) Reserve(label string) error {
s.Lock()
defer s.Unlock()
context, err := selinux.NewContext(label)
if err != nil {
return err
}
level := context["level"]
// no reason to count empty
if level == "" {
return nil
}
if _, ok := s.levels[level]; !ok {
s.Reserver(label)
}
s.levels[level]++
return nil
}
func (s *Store) Release(label string) {
s.Lock()
defer s.Unlock()
context, err := selinux.NewContext(label)
if err != nil {
return
}
level := context["level"]
if level == "" {
return
}
count, ok := s.levels[level]
if !ok {
return
}
switch {
case count == 1:
s.Releaser(label)
delete(s.levels, level)
case count < 1:
delete(s.levels, level)
case count > 1:
s.levels[level] = count - 1
}
}

View File

@@ -1,89 +0,0 @@
/*
Copyright The containerd Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package sandbox
import (
"encoding/json"
cni "github.com/containerd/go-cni"
"github.com/pkg/errors"
runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
)
// NOTE(random-liu):
// 1) Metadata is immutable after created.
// 2) Metadata is checkpointed as containerd container label.
// metadataVersion is current version of sandbox metadata.
const metadataVersion = "v1" // nolint
// versionedMetadata is the internal versioned sandbox metadata.
// nolint
type versionedMetadata struct {
// Version indicates the version of the versioned sandbox metadata.
Version string
// Metadata's type is metadataInternal. If not there will be a recursive call in MarshalJSON.
Metadata metadataInternal
}
// metadataInternal is for internal use.
type metadataInternal Metadata
// Metadata is the unversioned sandbox metadata.
type Metadata struct {
// ID is the sandbox id.
ID string
// Name is the sandbox name.
Name string
// Config is the CRI sandbox config.
Config *runtime.PodSandboxConfig
// NetNSPath is the network namespace used by the sandbox.
NetNSPath string
// IP of Pod if it is attached to non host network
IP string
// AdditionalIPs of the Pod if it is attached to non host network
AdditionalIPs []string
// RuntimeHandler is the runtime handler name of the pod.
RuntimeHandler string
// CNIresult resulting configuration for attached network namespace interfaces
CNIResult *cni.CNIResult
// ProcessLabel is the SELinux process label for the container
ProcessLabel string
}
// MarshalJSON encodes Metadata into bytes in json format.
func (c *Metadata) MarshalJSON() ([]byte, error) {
return json.Marshal(&versionedMetadata{
Version: metadataVersion,
Metadata: metadataInternal(*c),
})
}
// UnmarshalJSON decodes Metadata from bytes.
func (c *Metadata) UnmarshalJSON(data []byte) error {
versioned := &versionedMetadata{}
if err := json.Unmarshal(data, versioned); err != nil {
return err
}
// Handle old version after upgrade.
switch versioned.Version {
case metadataVersion:
*c = Metadata(versioned.Metadata)
return nil
}
return errors.Errorf("unsupported version: %q", versioned.Version)
}

Some files were not shown because too many files have changed in this diff Show More