Compare commits

..

24 Commits

Author SHA1 Message Date
Peng Tao
3852d7c180 Merge pull request #8 from liubin/fix/7-add-sync-marker-to-error-cause
errors: add sync marker to error cause
2020-09-14 16:15:12 +08:00
Peng Tao
0aaf7dba0b Merge pull request #6 from liubin/fix/cfs_quota-to-i64
cpu: change cfs_quota from u64 to i64
2020-09-14 16:14:22 +08:00
bin liu
42685bbfe9 errors: add sync marker to error cause
Errors cause should be shared to let anyhow to wrap it.

Fixes: #7

Signed-off-by: bin liu <bin@hyper.sh>
2020-09-14 15:22:55 +08:00
bin liu
f8d653e987 cpu: change cfs_quota from u64 to i64
cfs_quota can be set to -1 to indicate no limit

Fixes: #5

Signed-off-by: bin liu <bin@hyper.sh>
2020-09-14 13:00:43 +08:00
Peng Tao
d20e6a5383 Merge pull request #4 from liubin/master
add cgroup v2 support
2020-09-09 20:44:28 +08:00
bin liu
be617190c7 hugetlb: update test to avoid depending on test env
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-09 17:09:12 +08:00
bin liu
250ada183a cargo fmt
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-09 16:42:55 +08:00
bin liu
ff6a0ea82a add SPDX-License-Identifier for modified files
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-09 13:35:46 +08:00
bin liu
bcbf438823 Merge remote-tracking branch 'upstream/master' 2020-09-09 13:27:12 +08:00
bin liu
c3912223d0 use let Some() instead of is_some() to make codes clear
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-09 13:26:18 +08:00
bin liu
5d51e50bec get relative paths at new/load func default.
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-08 10:30:29 +08:00
Peng Tao
a1bc5868d6 Merge pull request #3 from liubin/fix/2
add SPDX-License-Identifier for all source
2020-09-08 09:16:47 +08:00
bin liu
86b245076c add SPDX-License-Identifier for all source
Add SPDX-License-Identifier for all existing codes.

Fixes: #2

Signed-off-by: bin liu <bin@hyper.sh>
2020-09-07 20:29:40 +08:00
bin liu
cd2c748a74 add systemd controller; add more tests; update blkio stat for v2
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-07 14:18:41 +08:00
bin liu
c623dc3fba add basic v2 cpu/memory functions
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-02 22:03:11 +08:00
bin liu
704db324ae fix: create all dirs
Signed-off-by: bin liu <bin@hyper.sh>
2020-09-01 00:01:28 +08:00
bin liu
9fe6cb58e4 change *limit* in memory from u64 to i64
Signed-off-by: bin liu <bin@hyper.sh>
2020-08-27 23:06:44 +08:00
bin liu
c702852fd7 add disable oom_control function
Signed-off-by: bin liu <bin@hyper.sh>
2020-08-25 11:35:01 +08:00
bin liu
9d70467327 add hugetlb functions
Signed-off-by: bin liu <bin@hyper.sh>
2020-08-25 10:58:56 +08:00
bin liu
6b1b26b1fe copy cpuset after created
Signed-off-by: bin liu <bin@hyper.sh>
2020-08-25 10:51:29 +08:00
bin liu
9ce206f6bc probe cgroup v1 mount path
Signed-off-by: bin liu <bin@hyper.sh>
2020-08-25 10:50:52 +08:00
Levente Kurusa
ed1e8162d5 rust2018: remove extern crate
Signed-off-by: Levente Kurusa <lkurusa@acm.org>
2018-12-23 16:57:41 +01:00
Levente Kurusa
932a6e770f Cargo: bump Rust version to 2018
Signed-off-by: Levente Kurusa <lkurusa@acm.org>
2018-12-23 16:32:00 +01:00
Levente Kurusa
01a3c22829 Transition to Rust 2018
Signed-off-by: Levente Kurusa <lkurusa@acm.org>
2018-12-23 16:31:05 +01:00
32 changed files with 2555 additions and 625 deletions

View File

@@ -7,10 +7,13 @@ categories = ["os", "api-bindings", "os::unix-apis"]
license = "MIT OR Apache-2.0"
version = "0.1.1-alpha.0"
authors = ["Levente Kurusa <lkurusa@acm.org>", "Sam Wilson <tecywiz121@hotmail.com>"]
edition = "2018"
[dependencies]
log = "0.4"
regex = "1.1"
nix = "0.18.0"
libc = "0.2"
[dev-dependencies]
nix = "0.11.0"
libc = "0.2.43"
libc = "0.2.76"

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `blkio` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,10 +12,10 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {
use crate::{
BlkIoResources, ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem,
};
@@ -21,6 +27,7 @@ use {
pub struct BlkIoController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
#[derive(Eq, PartialEq, Debug)]
@@ -48,11 +55,33 @@ pub struct IoService {
/// How many items were synchronously transferred.
pub sync: u64,
/// How many items were asynchronously transferred.
pub async: u64,
pub r#async: u64,
/// Total number of items transferred.
pub total: u64,
}
#[derive(Eq, PartialEq, Debug)]
/// Per-device activity from the control group.
/// Only for cgroup v2
pub struct IoStat {
/// The major number of the device.
pub major: i16,
/// The minor number of the device.
pub minor: i16,
/// How many bytes were read from the device.
pub rbytes: u64,
/// How many bytes were written to the device.
pub wbytes: u64,
/// How many iops were read from the device.
pub rios: u64,
/// How many iops were written to the device.
pub wios: u64,
/// How many discard bytes were read from the device.
pub dbytes: u64,
/// How many discard iops were written to the device.
pub dios: u64,
}
fn parse_io_service(s: String) -> Result<Vec<IoService>> {
s.lines()
.filter(|x| x.split_whitespace().collect::<Vec<_>>().len() == 3)
@@ -77,7 +106,7 @@ fn parse_io_service(s: String) -> Result<Vec<IoService>> {
read: read_val.parse::<u64>().unwrap(),
write: write_val.parse::<u64>().unwrap(),
sync: sync_val.parse::<u64>().unwrap(),
async: async_val.parse::<u64>().unwrap(),
r#async: async_val.parse::<u64>().unwrap(),
total: total_val.parse::<u64>().unwrap(),
}),
_ => None,
@@ -94,6 +123,41 @@ fn parse_io_service(s: String) -> Result<Vec<IoService>> {
})
}
fn get_value(s: &str) -> String {
let arr = s.split(':').collect::<Vec<&str>>();
if arr.len() != 2 {
return "0".to_string();
}
arr[1].to_string()
}
fn parse_io_stat(s: String) -> Result<Vec<IoStat>> {
// line:
// 8:0 rbytes=180224 wbytes=0 rios=3 wios=0 dbytes=0 dios=0
let v = s
.lines()
.filter(|x| x.split_whitespace().collect::<Vec<_>>().len() == 7)
.map(|x| {
let arr = x.split_whitespace().collect::<Vec<&str>>();
let device = arr[0].split(":").collect::<Vec<&str>>();
let (major, minor) = (device[0], device[1]);
IoStat {
major: major.parse::<i16>().unwrap(),
minor: minor.parse::<i16>().unwrap(),
rbytes: get_value(arr[1]).parse::<u64>().unwrap(),
wbytes: get_value(arr[2]).parse::<u64>().unwrap(),
rios: get_value(arr[3]).parse::<u64>().unwrap(),
wios: get_value(arr[4]).parse::<u64>().unwrap(),
dbytes: get_value(arr[5]).parse::<u64>().unwrap(),
dios: get_value(arr[6]).parse::<u64>().unwrap(),
}
})
.collect::<Vec<IoStat>>();
Ok(v)
}
fn parse_io_service_total(s: String) -> Result<u64> {
s.lines()
.filter(|x| x.split_whitespace().collect::<Vec<_>>().len() == 2)
@@ -141,7 +205,7 @@ fn parse_blkio_data(s: String) -> Result<Vec<BlkIoData>> {
/// Current state and statistics about how throttled are the block devices when accessed from the
/// controller's control group.
#[derive(Debug)]
#[derive(Default, Debug)]
pub struct BlkIoThrottle {
/// Statistics about the bytes transferred between the block devices by the tasks in this
/// control group.
@@ -176,7 +240,7 @@ pub struct BlkIoThrottle {
}
/// Statistics and state of the block devices.
#[derive(Debug)]
#[derive(Default, Debug)]
pub struct BlkIo {
/// The number of BIOS requests merged into I/O requests by the control group's tasks.
pub io_merged: Vec<IoService>,
@@ -253,6 +317,9 @@ pub struct BlkIo {
pub weight: u64,
/// Same as `weight`, but per-block-device.
pub weight_device: Vec<BlkIoData>,
/// IoStat for cgroup v2
pub io_stat: Vec<IoStat>,
}
impl ControllerInternal for BlkIoController {
@@ -269,17 +336,30 @@ impl ControllerInternal for BlkIoController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let res: &BlkIoResources = &res.blkio;
if res.update_values {
let _ = self.set_weight(res.weight as u64);
let _ = self.set_leaf_weight(res.leaf_weight as u64);
if let Some(weight) = res.weight {
let _ = self.set_weight(weight as u64);
}
if let Some(leaf_weight) = res.leaf_weight {
let _ = self.set_leaf_weight(leaf_weight as u64);
}
for dev in &res.weight_device {
let _ = self.set_weight_for_device(dev.major, dev.minor, dev.weight as u64);
let _ = self.set_leaf_weight_for_device(dev.major, dev.minor, dev.leaf_weight as u64);
if let Some(weight) = dev.weight {
let _ = self.set_weight_for_device(dev.major, dev.minor, weight as u64);
}
if let Some(leaf_weight) = dev.leaf_weight {
let _ =
self.set_leaf_weight_for_device(dev.major, dev.minor, leaf_weight as u64);
}
}
for dev in &res.throttle_read_bps_device {
@@ -334,25 +414,45 @@ fn read_string_from(mut file: File) -> Result<String> {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
impl BlkIoController {
/// Constructs a new `BlkIoController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
fn blkio_v2(&self) -> BlkIo {
let mut blkio: BlkIo = Default::default();
blkio.io_stat = self
.open_path("io.stat", false)
.and_then(read_string_from)
.and_then(parse_io_stat)
.unwrap_or(Vec::new());
blkio
}
/// Gathers statistics about and reports the state of the block devices used by the control
/// group's tasks.
pub fn blkio(&self) -> BlkIo {
if self.v2 {
return self.blkio_v2();
}
BlkIo {
io_merged: self
.open_path("blkio.io_merged", false)
@@ -574,6 +674,7 @@ impl BlkIoController {
.and_then(read_string_from)
.and_then(parse_blkio_data)
.unwrap_or(Vec::new()),
io_stat: Vec::new(),
}
}
@@ -588,12 +689,7 @@ impl BlkIoController {
}
/// Same as `set_leaf_weight()`, but settable per each block device.
pub fn set_leaf_weight_for_device(
&self,
major: u64,
minor: u64,
weight: u64,
) -> Result<()> {
pub fn set_leaf_weight_for_device(&self, major: u64, minor: u64, weight: u64) -> Result<()> {
self.open_path("blkio.leaf_weight_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, weight).as_ref())
@@ -612,92 +708,97 @@ impl BlkIoController {
/// Throttle the bytes per second rate of read operation affecting the block device
/// `major:minor` to `bps`.
pub fn throttle_read_bps_for_device(
&self,
major: u64,
minor: u64,
bps: u64,
) -> Result<()> {
self.open_path("blkio.throttle.read_bps_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, bps).to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn throttle_read_bps_for_device(&self, major: u64, minor: u64, bps: u64) -> Result<()> {
let mut file = "blkio.throttle.read_bps_device";
let mut content = format!("{}:{} {}", major, minor, bps);
if self.v2 {
file = "io.max";
content = format!("{}:{} rbps={}", major, minor, bps);
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Throttle the I/O operations per second rate of read operation affecting the block device
/// `major:minor` to `bps`.
pub fn throttle_read_iops_for_device(
&self,
major: u64,
minor: u64,
iops: u64,
) -> Result<()> {
self.open_path("blkio.throttle.read_iops_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, iops).to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn throttle_read_iops_for_device(&self, major: u64, minor: u64, iops: u64) -> Result<()> {
let mut file = "blkio.throttle.read_iops_device";
let mut content = format!("{}:{} {}", major, minor, iops);
if self.v2 {
file = "io.max";
content = format!("{}:{} riops={}", major, minor, iops);
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Throttle the bytes per second rate of write operation affecting the block device
/// `major:minor` to `bps`.
pub fn throttle_write_bps_for_device(
&self,
major: u64,
minor: u64,
bps: u64,
) -> Result<()> {
self.open_path("blkio.throttle.write_bps_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, bps).to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn throttle_write_bps_for_device(&self, major: u64, minor: u64, bps: u64) -> Result<()> {
let mut file = "blkio.throttle.write_bps_device";
let mut content = format!("{}:{} {}", major, minor, bps);
if self.v2 {
file = "io.max";
content = format!("{}:{} wbps={}", major, minor, bps);
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Throttle the I/O operations per second rate of write operation affecting the block device
/// `major:minor` to `bps`.
pub fn throttle_write_iops_for_device(
&self,
major: u64,
minor: u64,
iops: u64,
) -> Result<()> {
self.open_path("blkio.throttle.write_iops_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, iops).to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn throttle_write_iops_for_device(&self, major: u64, minor: u64, iops: u64) -> Result<()> {
let mut file = "blkio.throttle.write_iops_device";
let mut content = format!("{}:{} {}", major, minor, iops);
if self.v2 {
file = "io.max";
content = format!("{}:{} wiops={}", major, minor, iops);
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Set the weight of the control group's tasks.
pub fn set_weight(&self, w: u64) -> Result<()> {
self.open_path("blkio.weight", true)
.and_then(|mut file| {
file.write_all(w.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
// Attation: may not find in high kernel version.
let mut file = "blkio.weight";
if self.v2 {
file = "io.bfq.weight";
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(w.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Same as `set_weight()`, but settable per each block device.
pub fn set_weight_for_device(
&self,
major: u64,
minor: u64,
weight: u64,
) -> Result<()> {
self.open_path("blkio.weight_device", true)
.and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, weight).as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn set_weight_for_device(&self, major: u64, minor: u64, weight: u64) -> Result<()> {
let mut file = "blkio.weight_device";
if self.v2 {
// Attation: there is no weight for device in runc
// https://github.com/opencontainers/runc/blob/46be7b612e2533c494e6a251111de46d8e286ed5/libcontainer/cgroups/fs2/io.go#L30
// may depends on IO schedulers https://wiki.ubuntu.com/Kernel/Reference/IOSchedulers
file = "io.bfq.weight";
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(format!("{}:{} {}", major, minor, weight).as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
}
#[cfg(test)]
mod test {
use blkio::{parse_blkio_data, BlkIoData};
use blkio::{parse_io_service, parse_io_service_total, IoService};
use error::*;
use crate::blkio::{parse_blkio_data, BlkIoData};
use crate::blkio::{parse_io_service, parse_io_service_total, IoService};
use crate::error::*;
static TEST_VALUE: &str = "\
8:32 Read 4280320
@@ -755,10 +856,7 @@ Total 61823067136
#[test]
fn test_parse_io_service_total() {
let ok = parse_io_service_total(TEST_VALUE.to_string()).unwrap();
assert_eq!(
ok,
61823067136
);
assert_eq!(ok, 61823067136);
}
#[test]
@@ -773,7 +871,7 @@ Total 61823067136
read: 4280320,
write: 0,
sync: 4280320,
async: 0,
r#async: 0,
total: 4280320,
},
IoService {
@@ -782,7 +880,7 @@ Total 61823067136
read: 5705479168,
write: 56096055296,
sync: 11213923328,
async: 50587611136,
r#async: 50587611136,
total: 61801534464,
},
IoService {
@@ -791,7 +889,7 @@ Total 61823067136
read: 10059776,
write: 0,
sync: 10059776,
async: 0,
r#async: 0,
total: 10059776,
},
IoService {
@@ -800,16 +898,13 @@ Total 61823067136
read: 7192576,
write: 0,
sync: 7192576,
async: 0,
r#async: 0,
total: 7192576,
}
]
);
let err = parse_io_service(TEST_WRONG_VALUE.to_string()).unwrap_err();
assert_eq!(
err.kind(),
&ErrorKind::ParseError,
);
assert_eq!(err.kind(), &ErrorKind::ParseError,);
}
#[test]

View File

@@ -1,11 +1,22 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module handles cgroup operations. Start here!
use error::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {CgroupPid, ControllIdentifier, Controller, Hierarchy, Resources, Subsystem};
use crate::libc_rmdir;
use crate::{CgroupPid, ControllIdentifier, Controller, Hierarchy, Resources, Subsystem};
use std::collections::HashMap;
use std::convert::From;
use std::path::Path;
use std::fs;
use std::path::{Path, PathBuf};
/// A control group is the central structure to this crate.
///
@@ -24,27 +35,35 @@ pub struct Cgroup<'b> {
subsystems: Vec<Subsystem>,
/// The hierarchy.
hier: &'b Hierarchy,
hier: Box<&'b dyn Hierarchy>,
path: String,
}
impl<'b> Cgroup<'b> {
/// Create this control group.
fn create(&self) {
for subsystem in &self.subsystems {
subsystem.to_controller().create();
if self.hier.v2() {
create_v2_cgroup(self.hier.root().clone(), &self.path);
} else {
for subsystem in &self.subsystems {
subsystem.to_controller().create();
}
}
}
pub fn v2(&self) -> bool {
self.hier.v2()
}
/// Create a new control group in the hierarchy `hier`, with name `path`.
///
/// Returns a handle to the control group that can be used to manipulate it.
///
/// Note that if the handle goes out of scope and is dropped, the control group is _not_
/// destroyed.
pub fn new<P: AsRef<Path>>(hier: &Hierarchy, path: P) -> Cgroup {
let cg = Cgroup::load(hier, path);
cg.create();
cg
pub fn new<P: AsRef<Path>>(hier: Box<&'b dyn Hierarchy>, path: P) -> Cgroup<'b> {
let relative_paths = get_cgroups_relative_paths().unwrap();
Cgroup::new_with_relative_paths(hier, path, relative_paths)
}
/// Create a handle for a control group in the hierarchy `hier`, with name `path`.
@@ -54,19 +73,65 @@ impl<'b> Cgroup<'b> {
///
/// Note that if the handle goes out of scope and is dropped, the control group is _not_
/// destroyed.
pub fn load<P: AsRef<Path>>(hier: &Hierarchy, path: P) -> Cgroup {
pub fn load<P: AsRef<Path>>(hier: Box<&'b dyn Hierarchy>, path: P) -> Cgroup<'b> {
let relative_paths = get_cgroups_relative_paths().unwrap();
Cgroup::load_with_relative_paths(hier, path, relative_paths)
}
/// Create a new control group in the hierarchy `hier`, with name `path`.
/// and relative paths from `/proc/self/cgroup`
///
/// Returns a handle to the control group that can be used to manipulate it.
///
/// Note that if the handle goes out of scope and is dropped, the control group is _not_
/// destroyed.
pub fn new_with_relative_paths<P: AsRef<Path>>(
hier: Box<&'b dyn Hierarchy>,
path: P,
relative_paths: HashMap<String, String>,
) -> Cgroup<'b> {
let cg = Cgroup::load_with_relative_paths(hier, path, relative_paths);
cg.create();
cg
}
/// Create a handle for a control group in the hierarchy `hier`, with name `path`,
/// and relative paths from `/proc/self/cgroup`
///
/// Returns a handle to the control group (that possibly does not exist until `create()` has
/// been called on the cgroup.
///
/// Note that if the handle goes out of scope and is dropped, the control group is _not_
/// destroyed.
pub fn load_with_relative_paths<P: AsRef<Path>>(
hier: Box<&'b dyn Hierarchy>,
path: P,
relative_paths: HashMap<String, String>,
) -> Cgroup<'b> {
let path = path.as_ref();
let mut subsystems = hier.subsystems();
if path.as_os_str() != "" {
subsystems = subsystems
.into_iter()
.map(|x| x.enter(path))
.map(|x| {
let cn = x.controller_name();
if relative_paths.contains_key(&cn) {
let rp = relative_paths.get(&cn).unwrap();
let valid_path = rp.trim_start_matches("/").to_string();
let mut p = PathBuf::from(valid_path);
p.push(path);
x.enter(p.as_ref())
} else {
x.enter(path)
}
})
.collect::<Vec<_>>();
}
let cg = Cgroup {
subsystems: subsystems,
hier: hier,
path: path.to_str().unwrap().to_string(),
};
cg
@@ -84,6 +149,15 @@ impl<'b> Cgroup<'b> {
/// actually removed, and remove the descendants first if not. In the future, this behavior
/// will change.
pub fn delete(self) {
if self.v2() {
if self.path != "" {
let mut p = self.hier.root().clone();
p.push(self.path);
libc_rmdir(p.to_str().unwrap());
}
return;
}
self.subsystems.into_iter().for_each(|sub| match sub {
Subsystem::Pid(pidc) => pidc.delete(),
Subsystem::Mem(c) => c.delete(),
@@ -98,6 +172,7 @@ impl<'b> Cgroup<'b> {
Subsystem::NetPrio(c) => c.delete(),
Subsystem::HugeTlb(c) => c.delete(),
Subsystem::Rdma(c) => c.delete(),
Subsystem::Systemd(c) => c.delete(),
});
}
@@ -143,25 +218,120 @@ impl<'b> Cgroup<'b> {
/// Attach a task to the control group.
pub fn add_task(&self, pid: CgroupPid) -> Result<()> {
self.subsystems()
.iter()
.try_for_each(|sub| sub.to_controller().add_task(&pid))
if self.v2() {
let subsystems = self.subsystems();
if subsystems.len() > 0 {
let c = subsystems[0].to_controller();
c.add_task(&pid)
} else {
Ok(())
}
} else {
self.subsystems()
.iter()
.try_for_each(|sub| sub.to_controller().add_task(&pid))
}
}
/// Returns an Iterator that can be used to iterate over the tasks that are currently in the
/// control group.
pub fn tasks(&self) -> Vec<CgroupPid> {
// Collect the tasks from all subsystems
let mut v = self
.subsystems()
.iter()
.map(|x| x.to_controller().tasks())
.fold(vec![], |mut acc, mut x| {
acc.append(&mut x);
acc
});
let mut v = if self.v2() {
let subsystems = self.subsystems();
if subsystems.len() > 0 {
let c = subsystems[0].to_controller();
c.tasks()
} else {
vec![]
}
} else {
self.subsystems()
.iter()
.map(|x| x.to_controller().tasks())
.fold(vec![], |mut acc, mut x| {
acc.append(&mut x);
acc
})
};
v.sort();
v.dedup();
v
}
}
pub const UNIFIED_MOUNTPOINT: &'static str = "/sys/fs/cgroup";
fn enable_controllers(controllers: &Vec<String>, path: &PathBuf) {
let mut f = path.clone();
f.push("cgroup.subtree_control");
for c in controllers {
let body = format!("+{}", c);
let _rest = fs::write(f.as_path(), body.as_bytes());
}
}
fn supported_controllers(p: &PathBuf) -> Vec<String> {
let p = format!("{}/{}", UNIFIED_MOUNTPOINT, "cgroup.controllers");
let ret = fs::read_to_string(p.as_str());
ret.unwrap_or(String::new())
.split(" ")
.map(|x| x.to_string())
.collect::<Vec<String>>()
}
fn create_v2_cgroup(root: PathBuf, path: &str) -> Result<()> {
// controler list ["memory", "cpu"]
let controllers = supported_controllers(&root);
let mut fp = root;
// enable for root
enable_controllers(&controllers, &fp);
// path: "a/b/c"
let elements = path.split("/").collect::<Vec<&str>>();
let last_index = elements.len() - 1;
for (i, ele) in elements.iter().enumerate() {
// ROOT/a
fp.push(ele);
// create dir, need not check if is a file or directory
if !fp.exists() {
match ::std::fs::create_dir(fp.clone()) {
Err(e) => return Err(Error::with_cause(ErrorKind::FsError, e)),
Ok(_) => {}
}
}
if i < last_index {
// enable controllers for substree
enable_controllers(&controllers, &fp);
}
}
Ok(())
}
pub fn get_cgroups_relative_paths() -> Result<HashMap<String, String>> {
let mut m = HashMap::new();
let content =
fs::read_to_string("/proc/self/cgroup").map_err(|e| Error::with_cause(ReadFailed, e))?;
for l in content.lines() {
let fl: Vec<&str> = l.split(':').collect();
if fl.len() != 3 {
continue;
}
let keys: Vec<&str> = fl[1].split(',').collect();
for key in &keys {
// this is a workaround, cgroup file are using `name=systemd`,
// but if file system the name is `systemd`
if *key == "name=systemd" {
m.insert("systemd".to_string(), fl[2].to_string());
} else {
m.insert(key.to_string(), fl[2].to_string());
}
}
}
Ok(m)
}

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module allows the user to create a control group using the Builder pattern.
//! # Example
//!
@@ -13,8 +19,9 @@
//! # use cgroups::*;
//! # use cgroups::devices::*;
//! # use cgroups::cgroup_builder::*;
//! let v1 = cgroups::hierarchies::V1::new();
//! let cgroup: Cgroup = CgroupBuilder::new("hello", &v1)
//! let h = cgroups::hierarchies::auto();
//! let h = Box::new(&*h);
//! let cgroup: Cgroup = CgroupBuilder::new("hello", h)
//! .memory()
//! .kernel_memory_limit(1024 * 1024)
//! .memory_hard_limit(1024 * 1024)
@@ -40,10 +47,10 @@
//! .limit("2G".to_string(), 2 * 1024 * 1024 * 1024)
//! .done()
//! .blkio()
//! .weight(123)
//! .leaf_weight(99)
//! .weight_device(6, 1, 100, 55)
//! .weight_device(6, 1, 100, 55)
//! .weight(Some(123))
//! .leaf_weight(Some(99))
//! .weight_device(6, 1, Some(100), Some(55))
//! .weight_device(6, 1, Some(100), Some(55))
//! .throttle_iops()
//! .read(6, 1, 10)
//! .write(11, 1, 100)
@@ -53,9 +60,12 @@
//! .done()
//! .build();
//! ```
use error::*;
use crate::error::*;
use {pid, BlkIoDeviceResource, BlkIoDeviceThrottleResource, Cgroup, DeviceResource, Hierarchy, HugePageResource, NetworkPriority, Resources};
use crate::{
pid, BlkIoDeviceResource, BlkIoDeviceThrottleResource, Cgroup, DeviceResource, Hierarchy,
HugePageResource, MaxValue, NetworkPriority, Resources,
};
macro_rules! gen_setter {
($res:ident, $cont:ident, $func:ident, $name:ident, $ty:ty) => {
@@ -65,13 +75,13 @@ macro_rules! gen_setter {
self.cgroup.resources.$res.$name = $name;
self
}
}
};
}
/// A control group builder instance
pub struct CgroupBuilder<'a> {
name: String,
hierarchy: &'a Hierarchy,
hierarchy: Box<&'a dyn Hierarchy>,
/// Internal, unsupported field: use the associated builders instead.
resources: Resources,
}
@@ -80,7 +90,7 @@ impl<'a> CgroupBuilder<'a> {
/// Start building a control group with the supplied hierarchy and name pair.
///
/// Note that this does not actually create the control group until `build()` is called.
pub fn new(name: &'a str, hierarchy: &'a Hierarchy) -> CgroupBuilder<'a> {
pub fn new(name: &'a str, hierarchy: Box<&'a dyn Hierarchy>) -> CgroupBuilder<'a> {
CgroupBuilder {
name: name.to_owned(),
hierarchy: hierarchy,
@@ -90,46 +100,34 @@ impl<'a> CgroupBuilder<'a> {
/// Builds the memory resources of the control group.
pub fn memory(self) -> MemoryResourceBuilder<'a> {
MemoryResourceBuilder {
cgroup: self,
}
MemoryResourceBuilder { cgroup: self }
}
/// Builds the pid resources of the control group.
pub fn pid(self) -> PidResourceBuilder<'a> {
PidResourceBuilder {
cgroup: self,
}
PidResourceBuilder { cgroup: self }
}
/// Builds the cpu resources of the control group.
pub fn cpu(self) -> CpuResourceBuilder<'a> {
CpuResourceBuilder {
cgroup: self,
}
CpuResourceBuilder { cgroup: self }
}
/// Builds the devices resources of the control group, disallowing or
/// allowing access to certain devices in the system.
pub fn devices(self) -> DeviceResourceBuilder<'a> {
DeviceResourceBuilder {
cgroup: self,
}
DeviceResourceBuilder { cgroup: self }
}
/// Builds the network resources of the control group, setting class id, or
/// various priorities on networking interfaces.
pub fn network(self) -> NetworkResourceBuilder<'a> {
NetworkResourceBuilder {
cgroup: self,
}
NetworkResourceBuilder { cgroup: self }
}
/// Builds the hugepage/hugetlb resources available to the control group.
pub fn hugepages(self) -> HugepagesResourceBuilder<'a> {
HugepagesResourceBuilder {
cgroup: self,
}
HugepagesResourceBuilder { cgroup: self }
}
/// Builds the block I/O resources available for the control group.
@@ -154,12 +152,35 @@ pub struct MemoryResourceBuilder<'a> {
}
impl<'a> MemoryResourceBuilder<'a> {
gen_setter!(memory, MemController, set_kmem_limit, kernel_memory_limit, u64);
gen_setter!(memory, MemController, set_limit, memory_hard_limit, u64);
gen_setter!(memory, MemController, set_soft_limit, memory_soft_limit, u64);
gen_setter!(memory, MemController, set_tcp_limit, kernel_tcp_memory_limit, u64);
gen_setter!(memory, MemController, set_memswap_limit, memory_swap_limit, u64);
gen_setter!(
memory,
MemController,
set_kmem_limit,
kernel_memory_limit,
i64
);
gen_setter!(memory, MemController, set_limit, memory_hard_limit, i64);
gen_setter!(
memory,
MemController,
set_soft_limit,
memory_soft_limit,
i64
);
gen_setter!(
memory,
MemController,
set_tcp_limit,
kernel_tcp_memory_limit,
i64
);
gen_setter!(
memory,
MemController,
set_memswap_limit,
memory_swap_limit,
i64
);
gen_setter!(memory, MemController, set_swappiness, swappiness, u64);
/// Finish the construction of the memory resources of a control group.
@@ -174,8 +195,13 @@ pub struct PidResourceBuilder<'a> {
}
impl<'a> PidResourceBuilder<'a> {
gen_setter!(pid, PidController, set_pid_max, maximum_number_of_processes, pid::PidMax);
gen_setter!(
pid,
PidController,
set_pid_max,
maximum_number_of_processes,
MaxValue
);
/// Finish the construction of the pid resources of a control group.
pub fn done(self) -> CgroupBuilder<'a> {
@@ -189,8 +215,8 @@ pub struct CpuResourceBuilder<'a> {
}
impl<'a> CpuResourceBuilder<'a> {
gen_setter!(cpu, CpuSetController, set_cpus, cpus, String);
// FIXME this should all changed to options.
gen_setter!(cpu, CpuSetController, set_cpus, cpus, Option<String>);
gen_setter!(cpu, CpuSetController, set_mems, mems, String);
gen_setter!(cpu, CpuController, set_shares, shares, u64);
gen_setter!(cpu, CpuController, set_cfs_quota, quota, i64);
@@ -210,22 +236,22 @@ pub struct DeviceResourceBuilder<'a> {
}
impl<'a> DeviceResourceBuilder<'a> {
/// Restrict (or allow) a device to the tasks inside the control group.
pub fn device(mut self,
major: i64,
minor: i64,
devtype: ::devices::DeviceType,
allow: bool,
access: Vec<::devices::DevicePermissions>)
-> DeviceResourceBuilder<'a> {
pub fn device(
mut self,
major: i64,
minor: i64,
devtype: crate::devices::DeviceType,
allow: bool,
access: Vec<crate::devices::DevicePermissions>,
) -> DeviceResourceBuilder<'a> {
self.cgroup.resources.devices.update_values = true;
self.cgroup.resources.devices.devices.push(DeviceResource {
major,
minor,
devtype,
allow,
access
access,
});
self
}
@@ -242,18 +268,17 @@ pub struct NetworkResourceBuilder<'a> {
}
impl<'a> NetworkResourceBuilder<'a> {
gen_setter!(network, NetclsController, set_class, class_id, u64);
/// Set the priority of the tasks when operating on a networking device defined by `name` to be
/// `priority`.
pub fn priority(mut self, name: String, priority: u64)
-> NetworkResourceBuilder<'a> {
pub fn priority(mut self, name: String, priority: u64) -> NetworkResourceBuilder<'a> {
self.cgroup.resources.network.update_values = true;
self.cgroup.resources.network.priorities.push(NetworkPriority {
name,
priority,
});
self.cgroup
.resources
.network
.priorities
.push(NetworkPriority { name, priority });
self
}
@@ -269,15 +294,14 @@ pub struct HugepagesResourceBuilder<'a> {
}
impl<'a> HugepagesResourceBuilder<'a> {
/// Limit the usage of certain hugepages (determined by `size`) to be at most `limit` bytes.
pub fn limit(mut self, size: String, limit: u64)
-> HugepagesResourceBuilder<'a> {
pub fn limit(mut self, size: String, limit: u64) -> HugepagesResourceBuilder<'a> {
self.cgroup.resources.hugepages.update_values = true;
self.cgroup.resources.hugepages.limits.push(HugePageResource {
size,
limit,
});
self.cgroup
.resources
.hugepages
.limits
.push(HugePageResource { size, limit });
self
}
@@ -294,24 +318,34 @@ pub struct BlkIoResourcesBuilder<'a> {
}
impl<'a> BlkIoResourcesBuilder<'a> {
gen_setter!(blkio, BlkIoController, set_weight, weight, u16);
gen_setter!(blkio, BlkIoController, set_leaf_weight, leaf_weight, u16);
gen_setter!(blkio, BlkIoController, set_weight, weight, Option<u16>);
gen_setter!(
blkio,
BlkIoController,
set_leaf_weight,
leaf_weight,
Option<u16>
);
/// Set the weight of a certain device.
pub fn weight_device(mut self,
major: u64,
minor: u64,
weight: u16,
leaf_weight: u16)
-> BlkIoResourcesBuilder<'a> {
pub fn weight_device(
mut self,
major: u64,
minor: u64,
weight: Option<u16>,
leaf_weight: Option<u16>,
) -> BlkIoResourcesBuilder<'a> {
self.cgroup.resources.blkio.update_values = true;
self.cgroup.resources.blkio.weight_device.push(BlkIoDeviceResource {
major,
minor,
weight,
leaf_weight,
});
self.cgroup
.resources
.blkio
.weight_device
.push(BlkIoDeviceResource {
major,
minor,
weight,
leaf_weight,
});
self
}
@@ -328,35 +362,41 @@ impl<'a> BlkIoResourcesBuilder<'a> {
}
/// Limit the read rate of the current metric for a certain device.
pub fn read(mut self, major: u64, minor: u64, rate: u64)
-> BlkIoResourcesBuilder<'a> {
pub fn read(mut self, major: u64, minor: u64, rate: u64) -> BlkIoResourcesBuilder<'a> {
self.cgroup.resources.blkio.update_values = true;
let throttle = BlkIoDeviceThrottleResource {
major,
minor,
rate,
};
let throttle = BlkIoDeviceThrottleResource { major, minor, rate };
if self.throttling_iops {
self.cgroup.resources.blkio.throttle_read_iops_device.push(throttle);
self.cgroup
.resources
.blkio
.throttle_read_iops_device
.push(throttle);
} else {
self.cgroup.resources.blkio.throttle_read_bps_device.push(throttle);
self.cgroup
.resources
.blkio
.throttle_read_bps_device
.push(throttle);
}
self
}
/// Limit the write rate of the current metric for a certain device.
pub fn write(mut self, major: u64, minor: u64, rate: u64)
-> BlkIoResourcesBuilder<'a> {
pub fn write(mut self, major: u64, minor: u64, rate: u64) -> BlkIoResourcesBuilder<'a> {
self.cgroup.resources.blkio.update_values = true;
let throttle = BlkIoDeviceThrottleResource {
major,
minor,
rate,
};
let throttle = BlkIoDeviceThrottleResource { major, minor, rate };
if self.throttling_iops {
self.cgroup.resources.blkio.throttle_write_iops_device.push(throttle);
self.cgroup
.resources
.blkio
.throttle_write_iops_device
.push(throttle);
} else {
self.cgroup.resources.blkio.throttle_write_bps_device.push(throttle);
self.cgroup
.resources
.blkio
.throttle_write_bps_device
.push(throttle);
}
self
}

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `cpu` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -7,11 +13,13 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::{parse_max_value, read_i64_from};
use {
ControllIdentifier, ControllerInternal, Controllers, CpuResources, Resources, Subsystem,
use crate::{
ControllIdentifier, ControllerInternal, Controllers, CpuResources, MaxValue, Resources,
Subsystem,
};
/// A controller that allows controlling the `cpu` subsystem of a Cgroup.
@@ -23,6 +31,7 @@ use {
pub struct CpuController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
/// The current state of the control group and its processes.
@@ -34,6 +43,13 @@ pub struct Cpu {
pub stat: String,
}
/// The current state of the control group and its processes.
#[derive(Debug)]
struct CFSQuotaAndPeriod {
quota: MaxValue,
period: u64,
}
impl ControllerInternal for CpuController {
fn control_type(&self) -> Controllers {
Controllers::Cpu
@@ -51,12 +67,15 @@ impl ControllerInternal for CpuController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let res: &CpuResources = &res.cpu;
if res.update_values {
// apply pid_max
let _ = self.set_shares(res.shares);
if self.shares()? != res.shares as u64 {
return Err(Error::new(ErrorKind::Other));
@@ -67,8 +86,8 @@ impl ControllerInternal for CpuController {
return Err(Error::new(ErrorKind::Other));
}
let _ = self.set_cfs_quota(res.quota as u64);
if self.cfs_quota()? != res.quota as u64 {
let _ = self.set_cfs_quota(res.quota);
if self.cfs_quota()? != res.quota {
return Err(Error::new(ErrorKind::Other));
}
@@ -102,19 +121,25 @@ impl<'a> From<&'a Subsystem> for &'a CpuController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
impl CpuController {
/// Contructs a new `CpuController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
@@ -130,7 +155,8 @@ impl CpuController {
Ok(_) => Ok(s),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}).unwrap_or("".to_string()),
})
.unwrap_or("".to_string()),
}
}
@@ -141,7 +167,12 @@ impl CpuController {
/// `shares` to `200` ensures that control group `B` receives twice as much as CPU bandwidth.
/// (Assuming both `A` and `B` are of the same parent)
pub fn set_shares(&self, shares: u64) -> Result<()> {
self.open_path("cpu.shares", true).and_then(|mut file| {
let mut file = "cpu.shares";
if self.v2 {
file = "cpu.weight";
}
// NOTE: .CpuShares is not used here. Conversion is the caller's responsibility.
self.open_path(file, true).and_then(|mut file| {
file.write_all(shares.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
@@ -150,12 +181,19 @@ impl CpuController {
/// Retrieve the CPU bandwidth that this control group (relative to other control groups and
/// this control group's parent) can use.
pub fn shares(&self) -> Result<u64> {
self.open_path("cpu.shares", false).and_then(read_u64_from)
let mut file = "cpu.shares";
if self.v2 {
file = "cpu.weight";
}
self.open_path(file, false).and_then(read_u64_from)
}
/// Specify a period (when using the CFS scheduler) of time in microseconds for how often this
/// control group's access to the CPU should be reallocated.
pub fn set_cfs_period(&self, us: u64) -> Result<()> {
if self.v2 {
return self.set_cfs_quota_and_period(None, Some(us));
}
self.open_path("cpu.cfs_period_us", true)
.and_then(|mut file| {
file.write_all(us.to_string().as_ref())
@@ -166,13 +204,22 @@ impl CpuController {
/// Retrieve the period of time of how often this cgroup's access to the CPU should be
/// reallocated in microseconds.
pub fn cfs_period(&self) -> Result<u64> {
if self.v2 {
let current_value = self
.open_path("cpu.max", false)
.and_then(parse_cfs_quota_and_period)?;
return Ok(current_value.period);
}
self.open_path("cpu.cfs_period_us", false)
.and_then(read_u64_from)
}
/// Specify a quota (when using the CFS scheduler) of time in microseconds for which all tasks
/// in this control group can run during one period (see: `set_cfs_period()`).
pub fn set_cfs_quota(&self, us: u64) -> Result<()> {
pub fn set_cfs_quota(&self, us: i64) -> Result<()> {
if self.v2 {
return self.set_cfs_quota_and_period(Some(us), None);
}
self.open_path("cpu.cfs_quota_us", true)
.and_then(|mut file| {
file.write_all(us.to_string().as_ref())
@@ -182,8 +229,99 @@ impl CpuController {
/// Retrieve the quota of time for which all tasks in this cgroup can run during one period, in
/// microseconds.
pub fn cfs_quota(&self) -> Result<u64> {
pub fn cfs_quota(&self) -> Result<i64> {
if self.v2 {
let current_value = self
.open_path("cpu.max", false)
.and_then(parse_cfs_quota_and_period)?;
return Ok(current_value.quota.to_i64());
}
self.open_path("cpu.cfs_quota_us", false)
.and_then(read_u64_from)
.and_then(read_i64_from)
}
pub fn set_cfs_quota_and_period(&self, quota: Option<i64>, period: Option<u64>) -> Result<()> {
if !self.v2 {
if let Some(q) = quota {
self.set_cfs_quota(q)?;
}
if let Some(p) = period {
self.set_cfs_period(p)?;
}
return Ok(());
}
// https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html
// cpu.max
// A read-write two value file which exists on non-root cgroups. The default is “max 100000”.
// The maximum bandwidth limit. Its in the following format:
// $MAX $PERIOD
// which indicates that the group may consume upto $MAX in each $PERIOD duration.
// “max” for $MAX indicates no limit. If only one number is written, $MAX is updated.
let current_value = self
.open_path("cpu.max", false)
.and_then(parse_cfs_quota_and_period)?;
let new_quota = if let Some(q) = quota {
if q > 0 {
q.to_string()
} else {
"max".to_string()
}
} else {
current_value.quota.to_string()
};
let new_period = if let Some(p) = period {
p.to_string()
} else {
current_value.period.to_string()
};
let line = format!("{} {}", new_quota, new_period);
self.open_path("cpu.max", true).and_then(|mut file| {
file.write_all(line.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
pub fn set_rt_runtime(&self, us: i64) -> Result<()> {
self.open_path("cpu.rt_runtime_us", true)
.and_then(|mut file| {
file.write_all(us.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
pub fn set_rt_period_us(&self, us: u64) -> Result<()> {
self.open_path("cpu.rt_period_us", true)
.and_then(|mut file| {
file.write_all(us.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
}
fn parse_cfs_quota_and_period(mut file: File) -> Result<CFSQuotaAndPeriod> {
let mut content = String::new();
file.read_to_string(&mut content)
.map_err(|e| Error::with_cause(ReadFailed, e))?;
let fields = content.trim().split(' ').collect::<Vec<&str>>();
if fields.len() != 2 {
return Err(Error::from_string(format!("invaild format: {}", content)));
}
let quota = parse_max_value(&fields[0].to_string())?;
let period = fields[1]
.parse::<u64>()
.map_err(|e| Error::with_cause(ParseError, e))?;
Ok(CFSQuotaAndPeriod {
quota: quota,
period: period,
})
}

View File

@@ -1,3 +1,8 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `cpuacct` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,10 +11,10 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
/// A controller that allows controlling the `cpuacct` subsystem of a Cgroup.
///
@@ -162,7 +167,9 @@ impl CpuAcctController {
/// Reset the statistics the kernel has gathered about the control group.
pub fn reset(&self) -> Result<()> {
self.open_path("cpuacct.usage", true)
.and_then(|mut file| file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e)))
self.open_path("cpuacct.usage", true).and_then(|mut file| {
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
}

View File

@@ -1,15 +1,23 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `cpuset` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
//! [Documentation/cgroup-v1/cpusets.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/cpusets.txt)
use log::*;
use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {
use crate::{
ControllIdentifier, ControllerInternal, Controllers, CpuResources, Resources, Subsystem,
};
@@ -21,6 +29,7 @@ use {
pub struct CpuSetController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
/// The current state of the `cpuset` controller for this control group.
@@ -93,17 +102,93 @@ impl ControllerInternal for CpuSetController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let res: &CpuResources = &res.cpu;
if res.update_values {
let _ = self.set_cpus(&res.cpus);
if res.cpus.is_some() {
let _ = self.set_cpus(res.cpus.as_ref().unwrap().as_str());
}
let _ = self.set_mems(&res.mems);
}
Ok(())
}
fn post_create(&self) {
if self.is_v2() {
return;
}
let current = self.get_path();
let parent = match current.parent() {
Some(p) => p,
None => return,
};
if current != self.get_base() {
match copy_from_parent(current.to_str().unwrap(), "cpuset.cpus") {
Ok(_) => (),
Err(err) => error!("error create_dir for cpuset.cpus {:?}", err),
}
match copy_from_parent(current.to_str().unwrap(), "cpuset.mems") {
Ok(_) => (),
Err(err) => error!("error create_dir for cpuset.mems {:?}", err),
}
}
}
}
fn find_no_empty_parent(from: &str, file: &str) -> Result<(String, Vec<PathBuf>)> {
let mut current_path = ::std::path::Path::new(from).to_path_buf();
let mut v = vec![];
loop {
let current_value =
match ::std::fs::read_to_string(current_path.clone().join(file).to_str().unwrap()) {
Ok(cpus) => String::from(cpus.trim()),
Err(e) => return Err(Error::with_cause(ReadFailed, e)),
};
if current_value != "" {
return Ok((current_value, v));
}
v.push(current_path.clone());
let parent = match current_path.parent() {
Some(p) => p,
None => return Ok(("".to_string(), v)),
};
// next loop, find parent
current_path = parent.to_path_buf();
}
}
/// copy_from_parent copy the cpuset.cpus and cpuset.mems from the parent
/// directory to the current directory if the file's contents are 0
fn copy_from_parent(current: &str, file: &str) -> Result<()> {
// find not empty cpus/memes from current directory.
let (value, parents) = find_no_empty_parent(current, file)?;
if value == "" || parents.len() == 0 {
return Ok(());
}
for p in parents.iter().rev() {
let mut pb = p.clone();
pb.push(file);
match ::std::fs::write(pb.to_str().unwrap(), value.as_bytes()) {
Ok(_) => (),
Err(e) => return Err(Error::with_cause(WriteFailed, e)),
}
}
Ok(())
}
impl ControllIdentifier for CpuSetController {
@@ -137,7 +222,10 @@ fn read_string_from(mut file: File) -> Result<String> {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
@@ -181,12 +269,15 @@ fn parse_range(s: String) -> Result<Vec<(u64, u64)>> {
impl CpuSetController {
/// Contructs a new `CpuSetController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
@@ -285,9 +376,11 @@ impl CpuSetController {
self.open_path("cpuset.cpu_exclusive", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -298,9 +391,11 @@ impl CpuSetController {
self.open_path("cpuset.mem_exclusive", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -335,9 +430,11 @@ impl CpuSetController {
self.open_path("cpuset.mem_hardwall", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -348,9 +445,11 @@ impl CpuSetController {
self.open_path("cpuset.sched_load_balance", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -372,9 +471,11 @@ impl CpuSetController {
self.open_path("cpuset.memory_migrate", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -385,9 +486,11 @@ impl CpuSetController {
self.open_path("cpuset.memory_spread_page", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -398,9 +501,11 @@ impl CpuSetController {
self.open_path("cpuset.memory_spread_slab", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -417,9 +522,11 @@ impl CpuSetController {
self.open_path("cpuset.memory_pressure_enabled", true)
.and_then(|mut file| {
if b {
file.write_all(b"1").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"1")
.map_err(|e| Error::with_cause(WriteFailed, e))
} else {
file.write_all(b"0").map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(b"0")
.map_err(|e| Error::with_cause(WriteFailed, e))
}
})
}
@@ -427,7 +534,7 @@ impl CpuSetController {
#[cfg(test)]
mod tests {
use cpuset;
use crate::cpuset;
#[test]
fn test_parse_range() {
let test_cases = vec![

View File

@@ -1,3 +1,8 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `devices` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -5,10 +10,12 @@
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use log::*;
use {
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::{
ControllIdentifier, ControllerInternal, Controllers, DeviceResource, DeviceResources,
Resources, Subsystem,
};
@@ -122,8 +129,7 @@ impl DevicePermissions {
return Ok(v);
}
for e in s.chars() {
let perm = DevicePermissions::from_char(e)
.ok_or_else(|| Error::new(ParseError))?;
let perm = DevicePermissions::from_char(e).ok_or_else(|| Error::new(ParseError))?;
v.push(perm);
}

View File

@@ -1,9 +1,18 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
use std::error::Error as StdError;
use std::fmt;
/// The different types of errors that can occur while manipulating control groups.
#[derive(Debug, Eq, PartialEq)]
pub enum ErrorKind {
FsError,
Common(String),
/// An error occured while writing to a control group file.
WriteFailed,
@@ -27,6 +36,8 @@ pub enum ErrorKind {
/// This crate checks against this and operations will fail with this error.
InvalidPath,
InvalidBytesSize,
/// An unknown error has occured.
Other,
}
@@ -34,18 +45,21 @@ pub enum ErrorKind {
#[derive(Debug)]
pub struct Error {
kind: ErrorKind,
cause: Option<Box<StdError + Send>>,
cause: Option<Box<StdError + Send + Sync>>,
}
impl fmt::Display for Error {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
let msg = match self.kind {
ErrorKind::WriteFailed => "unable to write to a control group file",
ErrorKind::ReadFailed => "unable to read a control group file",
ErrorKind::ParseError => "unable to parse control group file",
ErrorKind::InvalidOperation => "the requested operation is invalid",
ErrorKind::InvalidPath => "the given path is invalid",
ErrorKind::Other => "an unknown error",
let msg = match &self.kind {
ErrorKind::FsError => "fs error".to_string(),
ErrorKind::Common(s) => s.clone(),
ErrorKind::WriteFailed => "unable to write to a control group file".to_string(),
ErrorKind::ReadFailed => "unable to read a control group file".to_string(),
ErrorKind::ParseError => "unable to parse control group file".to_string(),
ErrorKind::InvalidOperation => "the requested operation is invalid".to_string(),
ErrorKind::InvalidPath => "the given path is invalid".to_string(),
ErrorKind::InvalidBytesSize => "invalid bytes size".to_string(),
ErrorKind::Other => "an unknown error".to_string(),
};
write!(f, "{}", msg)
@@ -62,16 +76,19 @@ impl StdError for Error {
}
impl Error {
pub(crate) fn new(kind: ErrorKind) -> Self {
pub(crate) fn from_string(s: String) -> Self {
Self {
kind,
kind: ErrorKind::Common(s),
cause: None,
}
}
pub(crate) fn new(kind: ErrorKind) -> Self {
Self { kind, cause: None }
}
pub(crate) fn with_cause<E>(kind: ErrorKind, cause: E) -> Self
where
E: 'static + Send + StdError,
E: 'static + Send + Sync + StdError,
{
Self {
kind,

90
src/events.rs Normal file
View File

@@ -0,0 +1,90 @@
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
use eventfd::{eventfd, EfdFlags};
use nix::sys::eventfd;
use std::fs::{self, File};
use std::io::Read;
use std::os::unix::io::{AsRawFd, FromRawFd};
use std::path::{Path, PathBuf};
use std::sync::mpsc::{self, Receiver};
use std::thread;
use crate::error::ErrorKind::*;
use crate::error::*;
// notify_on_oom returns channel on which you can expect event about OOM,
// if process died without OOM this channel will be closed.
pub fn notify_on_oom_v2(key: &str, dir: &PathBuf) -> Result<Receiver<String>> {
register_memory_event(key, dir, "memory.oom_control", "")
}
// notify_on_oom returns channel on which you can expect event about OOM,
// if process died without OOM this channel will be closed.
pub fn notify_on_oom_v1(key: &str, dir: &PathBuf) -> Result<Receiver<String>> {
register_memory_event(key, dir, "memory.oom_control", "")
}
// level is one of "low", "medium", or "critical"
pub fn notify_memory_pressure(key: &str, dir: &PathBuf, level: &str) -> Result<Receiver<String>> {
if level != "low" && level != "medium" && level != "critical" {
return Err(Error::from_string(format!(
"invalid pressure level {}",
level
)));
}
register_memory_event(key, dir, "memory.pressure_level", level)
}
fn register_memory_event(
key: &str,
cg_dir: &PathBuf,
event_name: &str,
arg: &str,
) -> Result<Receiver<String>> {
let path = cg_dir.join(event_name);
let event_file = File::open(path).map_err(|e| Error::with_cause(ReadFailed, e))?;
let eventfd =
eventfd(0, EfdFlags::EFD_CLOEXEC).map_err(|e| Error::with_cause(ReadFailed, e))?;
let event_control_path = cg_dir.join("cgroup.event_control");
let data;
if arg == "" {
data = format!("{} {}", eventfd, event_file.as_raw_fd());
} else {
data = format!("{} {} {}", eventfd, event_file.as_raw_fd(), arg);
}
// write to file and set mode to 0700(FIXME)
fs::write(&event_control_path, data).map_err(|e| Error::with_cause(WriteFailed, e));
let mut eventfd_file = unsafe { File::from_raw_fd(eventfd) };
let (sender, receiver) = mpsc::channel();
let key = key.to_string();
thread::spawn(move || {
loop {
let mut buf = [0; 8];
match eventfd_file.read(&mut buf) {
Err(err) => {
return;
}
Ok(_) => {}
}
// When a cgroup is destroyed, an event is sent to eventfd.
// So if the control path is gone, return instead of notifying.
if !Path::new(&event_control_path).exists() {
return;
}
sender.send(key.clone()).unwrap();
}
});
Ok(receiver)
}

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `freezer` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -5,10 +11,10 @@
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
/// A controller that allows controlling the `freezer` subsystem of a Cgroup.
///
@@ -22,6 +28,7 @@ use {ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
pub struct FreezerController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
/// The current state of the control group
@@ -75,40 +82,62 @@ impl<'a> From<&'a Subsystem> for &'a FreezerController {
impl FreezerController {
/// Contructs a new `FreezerController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
/// Freezes the processes in the control group.
pub fn freeze(&self) -> Result<()> {
self.open_path("freezer.state", true).and_then(|mut file| {
file.write_all("FROZEN".to_string().as_ref())
let mut file = "freezer.state";
let mut content = "FROZEN".to_string();
if self.v2 {
file = "cgroup.freeze";
content = "1".to_string();
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Thaws, that is, unfreezes the processes in the control group.
pub fn thaw(&self) -> Result<()> {
self.open_path("freezer.state", true).and_then(|mut file| {
file.write_all("THAWED".to_string().as_ref())
let mut file = "freezer.state";
let mut content = "THAWED".to_string();
if self.v2 {
file = "cgroup.freeze";
content = "0".to_string();
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(content.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Retrieve the state of processes in the control group.
pub fn state(&self) -> Result<FreezerState> {
self.open_path("freezer.state", false).and_then(|mut file| {
let mut file = "freezer.state";
if self.v2 {
file = "cgroup.freeze";
}
self.open_path(file, false).and_then(|mut file| {
let mut s = String::new();
let res = file.read_to_string(&mut s);
match res {
Ok(_) => match s.as_ref() {
"FROZEN" => Ok(FreezerState::Frozen),
"THAWED" => Ok(FreezerState::Thawed),
"1" => Ok(FreezerState::Frozen),
"0" => Ok(FreezerState::Thawed),
"FREEZING" => Ok(FreezerState::Freezing),
_ => Err(Error::new(ParseError)),
},

View File

@@ -1,64 +1,85 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module represents the various control group hierarchies the Linux kernel supports.
//!
//! Currently, we only support the cgroupv1 hierarchy, but in the future we will add support for
//! the Unified Hierarchy.
use nix::sys::statfs;
use std::fs::File;
use std::fs::{self, File};
use std::io::BufRead;
use std::io::BufReader;
use std::path::{Path, PathBuf};
use blkio::BlkIoController;
use cpu::CpuController;
use cpuacct::CpuAcctController;
use cpuset::CpuSetController;
use devices::DevicesController;
use freezer::FreezerController;
use hugetlb::HugeTlbController;
use memory::MemController;
use net_cls::NetClsController;
use net_prio::NetPrioController;
use perf_event::PerfEventController;
use pid::PidController;
use rdma::RdmaController;
use {Controllers, Hierarchy, Subsystem};
use log::*;
use cgroup::Cgroup;
use crate::blkio::BlkIoController;
use crate::cpu::CpuController;
use crate::cpuacct::CpuAcctController;
use crate::cpuset::CpuSetController;
use crate::devices::DevicesController;
use crate::freezer::FreezerController;
use crate::hugetlb::HugeTlbController;
use crate::memory::MemController;
use crate::net_cls::NetClsController;
use crate::net_prio::NetPrioController;
use crate::perf_event::PerfEventController;
use crate::pid::PidController;
use crate::rdma::RdmaController;
use crate::systemd::SystemdController;
use crate::{Controllers, Hierarchy, Subsystem};
use crate::cgroup::Cgroup;
/// The standard, original cgroup implementation. Often referred to as "cgroupv1".
pub struct V1 {
mount_point: String,
}
pub struct V2 {
root: String,
}
impl Hierarchy for V1 {
fn v2(&self) -> bool {
false
}
fn subsystems(&self) -> Vec<Subsystem> {
let mut subs = vec![];
if self.check_support(Controllers::Pids) {
subs.push(Subsystem::Pid(PidController::new(self.root())));
subs.push(Subsystem::Pid(PidController::new(self.root(), false)));
}
if self.check_support(Controllers::Mem) {
subs.push(Subsystem::Mem(MemController::new(self.root())));
subs.push(Subsystem::Mem(MemController::new(self.root(), false)));
}
if self.check_support(Controllers::CpuSet) {
subs.push(Subsystem::CpuSet(CpuSetController::new(self.root())));
subs.push(Subsystem::CpuSet(CpuSetController::new(self.root(), false)));
}
if self.check_support(Controllers::CpuAcct) {
subs.push(Subsystem::CpuAcct(CpuAcctController::new(self.root())));
}
if self.check_support(Controllers::Cpu) {
subs.push(Subsystem::Cpu(CpuController::new(self.root())));
subs.push(Subsystem::Cpu(CpuController::new(self.root(), false)));
}
if self.check_support(Controllers::Devices) {
subs.push(Subsystem::Devices(DevicesController::new(self.root())));
}
if self.check_support(Controllers::Freezer) {
subs.push(Subsystem::Freezer(FreezerController::new(self.root())));
subs.push(Subsystem::Freezer(FreezerController::new(
self.root(),
false,
)));
}
if self.check_support(Controllers::NetCls) {
subs.push(Subsystem::NetCls(NetClsController::new(self.root())));
}
if self.check_support(Controllers::BlkIo) {
subs.push(Subsystem::BlkIo(BlkIoController::new(self.root())));
subs.push(Subsystem::BlkIo(BlkIoController::new(self.root(), false)));
}
if self.check_support(Controllers::PerfEvent) {
subs.push(Subsystem::PerfEvent(PerfEventController::new(self.root())));
@@ -67,17 +88,27 @@ impl Hierarchy for V1 {
subs.push(Subsystem::NetPrio(NetPrioController::new(self.root())));
}
if self.check_support(Controllers::HugeTlb) {
subs.push(Subsystem::HugeTlb(HugeTlbController::new(self.root())));
subs.push(Subsystem::HugeTlb(HugeTlbController::new(
self.root(),
false,
)));
}
if self.check_support(Controllers::Rdma) {
subs.push(Subsystem::Rdma(RdmaController::new(self.root())));
}
if self.check_support(Controllers::Systemd) {
subs.push(Subsystem::Systemd(SystemdController::new(
self.root(),
false,
)));
}
subs
}
fn root_control_group(&self) -> Cgroup {
Cgroup::load(self, "".to_string())
let b: &Hierarchy = self as &Hierarchy;
Cgroup::load(Box::new(&*b), "".to_string())
}
fn check_support(&self, sub: Controllers) -> bool {
@@ -97,10 +128,77 @@ impl Hierarchy for V1 {
}
}
impl Hierarchy for V2 {
fn v2(&self) -> bool {
true
}
fn subsystems(&self) -> Vec<Subsystem> {
let p = format!("{}/{}", UNIFIED_MOUNTPOINT, "cgroup.controllers");
let ret = fs::read_to_string(p.as_str());
if ret.is_err() {
return vec![];
}
let mut subs = vec![];
let controllers = ret.unwrap().trim().to_string();
let controller_list: Vec<&str> = controllers.split(' ').collect();
for s in controller_list {
match s {
"cpu" => {
subs.push(Subsystem::Cpu(CpuController::new(self.root(), true)));
}
"io" => {
subs.push(Subsystem::BlkIo(BlkIoController::new(self.root(), true)));
}
"cpuset" => {
subs.push(Subsystem::CpuSet(CpuSetController::new(self.root(), true)));
}
"memory" => {
subs.push(Subsystem::Mem(MemController::new(self.root(), true)));
}
"pids" => {
subs.push(Subsystem::Pid(PidController::new(self.root(), true)));
}
"freezer" => {
subs.push(Subsystem::Freezer(FreezerController::new(
self.root(),
true,
)));
}
"hugetlb" => {
subs.push(Subsystem::HugeTlb(HugeTlbController::new(
self.root(),
true,
)));
}
_ => {}
}
}
subs
}
fn root_control_group(&self) -> Cgroup {
let b: &Hierarchy = self as &Hierarchy;
Cgroup::load(Box::new(&*b), "".to_string())
}
fn check_support(&self, _sub: Controllers) -> bool {
return false;
}
fn root(&self) -> PathBuf {
PathBuf::from(self.root.clone())
}
}
impl V1 {
/// Finds where control groups are mounted to and returns a hierarchy in which control groups
/// can be created.
pub fn new() -> Self {
pub fn new() -> V1 {
let mount_point = find_v1_mount().unwrap();
V1 {
mount_point: mount_point,
@@ -108,6 +206,60 @@ impl V1 {
}
}
impl V2 {
/// Finds where control groups are mounted to and returns a hierarchy in which control groups
/// can be created.
pub fn new() -> V2 {
V2 {
root: String::from(UNIFIED_MOUNTPOINT),
}
}
}
pub const UNIFIED_MOUNTPOINT: &'static str = "/sys/fs/cgroup";
#[cfg(all(target_os = "linux", not(target_env = "musl")))]
pub fn is_cgroup2_unified_mode() -> bool {
let path = Path::new(UNIFIED_MOUNTPOINT);
let fs_stat = statfs::statfs(path);
if fs_stat.is_err() {
return false;
}
// FIXME notwork, nix will not compile CGROUP2_SUPER_MAGIC because not(target_env = "musl")
fs_stat.unwrap().filesystem_type() == statfs::CGROUP2_SUPER_MAGIC
}
pub const INIT_CGROUP_PATHS: &'static str = "/proc/1/cgroup";
#[cfg(all(target_os = "linux", target_env = "musl"))]
pub fn is_cgroup2_unified_mode() -> bool {
let lines = fs::read_to_string(INIT_CGROUP_PATHS);
if lines.is_err() {
return false;
}
for line in lines.unwrap().lines() {
let fields: Vec<&str> = line.split(':').collect();
if fields.len() != 3 {
continue;
}
if fields[0] != "0" {
return false;
}
}
true
}
pub fn auto() -> Box<dyn Hierarchy> {
if is_cgroup2_unified_mode() {
Box::new(V2::new())
} else {
Box::new(V1::new())
}
}
fn find_v1_mount() -> Option<String> {
// Open mountinfo so we can get a parseable mount list
let mountinfo_path = Path::new("/proc/self/mountinfo");
@@ -123,12 +275,22 @@ fn find_v1_mount() -> Option<String> {
let line = _line.unwrap();
let mut fields = line.split_whitespace();
let index = line.find(" - ").unwrap();
let mut more_fields = line[index + 3..].split_whitespace().collect::<Vec<_>>();
let fstype = more_fields[0];
if fstype == "tmpfs" && more_fields[2].contains("ro") {
let more_fields = line[index + 3..].split_whitespace().collect::<Vec<_>>();
if more_fields.len() == 0 {
continue;
}
if more_fields[0] == "cgroup" {
if more_fields.len() < 3 {
continue;
}
let cgroups_mount = fields.nth(4).unwrap();
info!("found cgroups at {:?}", cgroups_mount);
return Some(cgroups_mount.to_string());
if let Some(parent) = std::path::Path::new(cgroups_mount).parent() {
if let Some(path) = parent.as_os_str().to_str() {
debug!("found cgroups {:?} from {:?}", path, cgroups_mount);
return Some(path.to_string());
}
}
continue;
}
}

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `hugetlb` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,12 +12,12 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::flat_keyed_to_vec;
use {
ControllIdentifier, ControllerInternal, Controllers, HugePageResources, Resources,
Subsystem,
use crate::{
ControllIdentifier, ControllerInternal, Controllers, HugePageResources, Resources, Subsystem,
};
/// A controller that allows controlling the `hugetlb` subsystem of a Cgroup.
@@ -22,6 +28,8 @@ use {
pub struct HugeTlbController {
base: PathBuf,
path: PathBuf,
sizes: Vec<String>,
v2: bool,
}
impl ControllerInternal for HugeTlbController {
@@ -38,6 +46,10 @@ impl ControllerInternal for HugeTlbController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let res: &HugePageResources = &res.hugepages;
@@ -77,30 +89,63 @@ impl<'a> From<&'a Subsystem> for &'a HugeTlbController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
impl HugeTlbController {
/// Constructs a new `HugeTlbController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
let sizes = get_hugepage_sizes().unwrap();
Self {
base: root.clone(),
path: root,
sizes: sizes,
v2: v2,
}
}
/// Whether the system supports `hugetlb_size` hugepages.
pub fn size_supported(&self, _hugetlb_size: &str) -> bool {
// TODO
true
pub fn size_supported(&self, hugetlb_size: &str) -> bool {
for s in &self.sizes {
if s == hugetlb_size {
return true;
}
}
false
}
pub fn get_sizes(&self) -> Vec<String> {
self.sizes.clone()
}
fn failcnt_v2(&self, hugetlb_size: &str) -> Result<u64> {
self.open_path(&format!("hugetlb.{}.events", hugetlb_size), false)
.and_then(flat_keyed_to_vec)
.and_then(|x| {
if x.len() == 0 {
return Err(Error::from_string(format!(
"get empty from hugetlb.{}.events",
hugetlb_size
)));
}
Ok(x[0].1 as u64)
})
}
/// Check how many times has the limit of `hugetlb_size` hugepages been hit.
pub fn failcnt(&self, hugetlb_size: &str) -> Result<u64> {
if self.v2 {
return self.failcnt_v2(hugetlb_size);
}
self.open_path(&format!("hugetlb.{}.failcnt", hugetlb_size), false)
.and_then(read_u64_from)
}
@@ -115,8 +160,11 @@ impl HugeTlbController {
/// Get the current usage of memory that is backed by hugepages of a certain size
/// (`hugetlb_size`).
pub fn usage_in_bytes(&self, hugetlb_size: &str) -> Result<u64> {
self.open_path(&format!("hugetlb.{}.usage_in_bytes", hugetlb_size), false)
.and_then(read_u64_from)
let mut file = format!("hugetlb.{}.usage_in_bytes", hugetlb_size);
if self.v2 {
file = format!("hugetlb.{}.current", hugetlb_size);
}
self.open_path(&file, false).and_then(read_u64_from)
}
/// Get the maximum observed usage of memory that is backed by hugepages of a certain size
@@ -125,16 +173,139 @@ impl HugeTlbController {
self.open_path(
&format!("hugetlb.{}.max_usage_in_bytes", hugetlb_size),
false,
).and_then(read_u64_from)
)
.and_then(read_u64_from)
}
/// Set the limit (in bytes) of how much memory can be backed by hugepages of a certain size
/// (`hugetlb_size`).
pub fn set_limit_in_bytes(&self, hugetlb_size: &str, limit: u64) -> Result<()> {
self.open_path(&format!("hugetlb.{}.limit_in_bytes", hugetlb_size), true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
let mut file = format!("hugetlb.{}.limit_in_bytes", hugetlb_size);
if self.v2 {
file = format!("hugetlb.{}.max", hugetlb_size);
}
self.open_path(&file, true).and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
}
pub const HUGEPAGESIZE_DIR: &'static str = "/sys/kernel/mm/hugepages";
use regex::Regex;
use std::collections::HashMap;
use std::fs;
fn get_hugepage_sizes() -> Result<Vec<String>> {
let mut m = Vec::new();
let dirs = fs::read_dir(HUGEPAGESIZE_DIR);
if dirs.is_err() {
return Ok(m);
}
for e in dirs.unwrap() {
let entry = e.unwrap();
let name = entry.file_name().into_string().unwrap();
let parts: Vec<&str> = name.split('-').collect();
if parts.len() != 2 {
continue;
}
let bmap = get_binary_size_map();
let size = parse_size(parts[1], &bmap)?;
let dabbrs = get_decimal_abbrs();
m.push(custom_size(size as f64, 1024.0, &dabbrs));
}
Ok(m)
}
pub const KB: u128 = 1000;
pub const MB: u128 = 1000 * KB;
pub const GB: u128 = 1000 * MB;
pub const TB: u128 = 1000 * GB;
pub const PB: u128 = 1000 * TB;
pub const KiB: u128 = 1024;
pub const MiB: u128 = 1024 * KiB;
pub const GiB: u128 = 1024 * MiB;
pub const TiB: u128 = 1024 * GiB;
pub const PiB: u128 = 1024 * TiB;
pub fn get_binary_size_map() -> HashMap<String, u128> {
let mut m = HashMap::new();
m.insert("k".to_string(), KiB);
m.insert("m".to_string(), MiB);
m.insert("g".to_string(), GiB);
m.insert("t".to_string(), TiB);
m.insert("p".to_string(), PiB);
m
}
pub fn get_decimal_size_map() -> HashMap<String, u128> {
let mut m = HashMap::new();
m.insert("k".to_string(), KB);
m.insert("m".to_string(), MB);
m.insert("g".to_string(), GB);
m.insert("t".to_string(), TB);
m.insert("p".to_string(), PB);
m
}
pub fn get_decimal_abbrs() -> Vec<String> {
let m = vec![
"B".to_string(),
"KB".to_string(),
"MB".to_string(),
"GB".to_string(),
"TB".to_string(),
"PB".to_string(),
"EB".to_string(),
"ZB".to_string(),
"YB".to_string(),
];
m
}
fn parse_size(s: &str, m: &HashMap<String, u128>) -> Result<u128> {
let re = Regex::new(r"(?P<num>\d+)(?P<mul>[kKmMgGtTpP]?)[bB]?$");
if re.is_err() {
return Err(Error::new(InvalidBytesSize));
}
let caps = re.unwrap().captures(s).unwrap();
let num = caps.name("num");
let size: u128 = if num.is_some() {
let n = num.unwrap().as_str().trim().parse::<u128>();
if n.is_err() {
return Err(Error::new(InvalidBytesSize));
}
n.unwrap()
} else {
return Err(Error::new(InvalidBytesSize));
};
let q = caps.name("mul");
let mul: u128 = if q.is_some() {
let t = m.get(q.unwrap().as_str());
if t.is_some() {
*t.unwrap()
} else {
return Err(Error::new(InvalidBytesSize));
}
} else {
return Err(Error::new(InvalidBytesSize));
};
Ok(size * mul)
}
fn custom_size(mut size: f64, base: f64, m: &Vec<String>) -> String {
let mut i = 0;
while size >= base && i < m.len() - 1 {
size /= base;
i += 1;
}
format!("{}{}", size, m[i].as_str())
}

View File

@@ -1,17 +1,25 @@
#[macro_use]
extern crate log;
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
use log::*;
use std::collections::HashMap;
use std::fs::File;
use std::io::{BufRead, BufReader, Write};
use std::io::{BufRead, BufReader, Read, Write};
use std::path::{Path, PathBuf};
pub mod blkio;
pub mod cgroup;
pub mod cgroup_builder;
pub mod cpu;
pub mod cpuacct;
pub mod cpuset;
pub mod devices;
pub mod error;
pub mod events;
pub mod freezer;
pub mod hierarchies;
pub mod hugetlb;
@@ -21,24 +29,26 @@ pub mod net_prio;
pub mod perf_event;
pub mod pid;
pub mod rdma;
pub mod cgroup_builder;
pub mod systemd;
use blkio::BlkIoController;
use cpu::CpuController;
use cpuacct::CpuAcctController;
use cpuset::CpuSetController;
use devices::DevicesController;
use error::*;
use freezer::FreezerController;
use hugetlb::HugeTlbController;
use memory::MemController;
use net_cls::NetClsController;
use net_prio::NetPrioController;
use perf_event::PerfEventController;
use pid::PidController;
use rdma::RdmaController;
use crate::blkio::BlkIoController;
use crate::cpu::CpuController;
use crate::cpuacct::CpuAcctController;
use crate::cpuset::CpuSetController;
use crate::devices::DevicesController;
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::freezer::FreezerController;
use crate::hugetlb::HugeTlbController;
use crate::memory::MemController;
use crate::net_cls::NetClsController;
use crate::net_prio::NetPrioController;
use crate::perf_event::PerfEventController;
use crate::pid::PidController;
use crate::rdma::RdmaController;
use crate::systemd::SystemdController;
pub use cgroup::Cgroup;
pub use crate::cgroup::Cgroup;
/// Contains all the subsystems that are available in this crate.
#[derive(Debug)]
@@ -69,6 +79,8 @@ pub enum Subsystem {
HugeTlb(HugeTlbController),
/// Controller for the `Rdma` subsystem, see `RdmaController` for more information.
Rdma(RdmaController),
/// Controller for the `Systemd` subsystem, see `SystemdController` for more information.
Systemd(SystemdController),
}
#[doc(hidden)]
@@ -87,6 +99,7 @@ pub enum Controllers {
NetPrio,
HugeTlb,
Rdma,
Systemd,
}
impl Controllers {
@@ -105,6 +118,7 @@ impl Controllers {
Controllers::NetPrio => return "net_prio".to_string(),
Controllers::HugeTlb => return "hugetlb".to_string(),
Controllers::Rdma => return "rdma".to_string(),
Controllers::Systemd => return "systemd".to_string(),
}
}
}
@@ -121,6 +135,13 @@ mod sealed {
fn get_path_mut(&mut self) -> &mut PathBuf;
fn get_base(&self) -> &PathBuf;
/// Hooks running after controller crated, if have
fn post_create(&self) {}
fn is_v2(&self) -> bool {
false
}
fn verify_path(&self) -> Result<()> {
if self.get_path().starts_with(self.get_base()) {
Ok(())
@@ -148,6 +169,17 @@ mod sealed {
}
}
fn get_max_value(&self, f: &str) -> Result<MaxValue> {
self.open_path(f, false).and_then(|mut file| {
let mut string = String::new();
let res = file.read_to_string(&mut string);
match res {
Ok(_) => parse_max_value(&string),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
})
}
#[doc(hidden)]
fn path_exists(&self, p: &str) -> bool {
if let Err(_) = self.verify_path() {
@@ -156,11 +188,10 @@ mod sealed {
std::path::Path::new(p).exists()
}
}
}
pub(crate) use sealed::ControllerInternal;
pub(crate) use crate::sealed::ControllerInternal;
/// A Controller is a subsystem attached to the control group.
///
@@ -190,9 +221,14 @@ pub trait Controller {
/// Get the list of tasks that this controller has.
fn tasks(&self) -> Vec<CgroupPid>;
fn v2(&self) -> bool;
}
impl<T> Controller for T where T: ControllerInternal {
impl<T> Controller for T
where
T: ControllerInternal,
{
fn control_type(&self) -> Controllers {
ControllerInternal::control_type(self)
}
@@ -209,10 +245,11 @@ impl<T> Controller for T where T: ControllerInternal {
/// Create this controller
fn create(&self) {
self.verify_path().expect("path should be valid");
self.verify_path()
.expect(format!("path should be valid: {:?}", self.path()).as_str());
match ::std::fs::create_dir(self.get_path()) {
Ok(_) => (),
match ::std::fs::create_dir_all(self.get_path()) {
Ok(_) => self.post_create(),
Err(e) => warn!("error create_dir {:?}", e),
}
}
@@ -225,13 +262,17 @@ impl<T> Controller for T where T: ControllerInternal {
/// Delete the controller.
fn delete(&self) {
if self.get_path().exists() {
let _ = ::std::fs::remove_dir(self.get_path());
libc_rmdir(self.get_path().to_str().unwrap());
}
}
/// Attach a task to this controller.
fn add_task(&self, pid: &CgroupPid) -> Result<()> {
self.open_path("tasks", true).and_then(|mut file| {
let mut file = "tasks";
if self.is_v2() {
file = "cgroup.procs";
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(pid.pid.to_string().as_ref())
.map_err(|e| Error::with_cause(ErrorKind::WriteFailed, e))
})
@@ -239,7 +280,11 @@ impl<T> Controller for T where T: ControllerInternal {
/// Get the list of tasks that this controller has.
fn tasks(&self) -> Vec<CgroupPid> {
self.open_path("tasks", false)
let mut file = "tasks";
if self.is_v2() {
file = "cgroup.procs";
}
self.open_path(file, false)
.and_then(|file| {
let bf = BufReader::new(file);
let mut v = Vec::new();
@@ -250,7 +295,12 @@ impl<T> Controller for T where T: ControllerInternal {
}
}
Ok(v.into_iter().map(CgroupPid::from).collect())
}).unwrap_or(vec![])
})
.unwrap_or(vec![])
}
fn v2(&self) -> bool {
self.is_v2()
}
}
@@ -271,6 +321,8 @@ pub trait Hierarchy {
/// Return a handle to the root control group in the hierarchy.
fn root_control_group(&self) -> Cgroup;
fn v2(&self) -> bool;
/// Checks whether a certain subsystem is supported in the hierarchy.
///
/// This is an internal function and should not be used.
@@ -284,16 +336,16 @@ pub struct MemoryResources {
/// Whether values should be applied to the controller.
pub update_values: bool,
/// How much memory (in bytes) can the kernel consume.
pub kernel_memory_limit: u64,
pub kernel_memory_limit: i64,
/// Upper limit of memory usage of the control group's tasks.
pub memory_hard_limit: u64,
pub memory_hard_limit: i64,
/// How much memory the tasks in the control group can use when the system is under memory
/// pressure.
pub memory_soft_limit: u64,
pub memory_soft_limit: i64,
/// How much of the kernel's memory (in bytes) can be used for TCP-related buffers.
pub kernel_tcp_memory_limit: u64,
pub kernel_tcp_memory_limit: i64,
/// How much memory and swap together can the tasks in the control group use.
pub memory_swap_limit: u64,
pub memory_swap_limit: i64,
/// Controls the tendency of the kernel to swap out parts of the address space of the tasks to
/// disk. Lower value implies less likely.
///
@@ -312,7 +364,7 @@ pub struct PidResources {
/// Note that attaching processes to the control group will still succeed _even_ if the limit
/// would be violated, however forks/clones inside the control group will have with `EAGAIN` if
/// they would violate the limit set here.
pub maximum_number_of_processes: pid::PidMax,
pub maximum_number_of_processes: MaxValue,
}
/// Resources limits about how the tasks can use the CPU.
@@ -323,7 +375,7 @@ pub struct CpuResources {
// cpuset
/// A comma-separated list of CPU IDs where the task in the control group can run. Dashes
/// between numbers indicate ranges.
pub cpus: String,
pub cpus: Option<String>,
/// Same syntax as the `cpus` field of this structure, but applies to memory nodes instead of
/// processors.
pub mems: String,
@@ -347,13 +399,13 @@ pub struct DeviceResource {
/// If true, access to the device is allowed, otherwise it's denied.
pub allow: bool,
/// `'c'` for character device, `'b'` for block device; or `'a'` for all devices.
pub devtype: ::devices::DeviceType,
pub devtype: crate::devices::DeviceType,
/// The major number of the device.
pub major: i64,
/// The minor number of the device.
pub minor: i64,
/// Sequence of `'r'`, `'w'` or `'m'`, each denoting read, write or mknod permissions.
pub access: Vec<::devices::DevicePermissions>,
pub access: Vec<crate::devices::DevicePermissions>,
}
/// Limit the usage of devices for the control group's tasks.
@@ -415,9 +467,9 @@ pub struct BlkIoDeviceResource {
/// The minor number of the device.
pub minor: u64,
/// The weight of the device against the descendant nodes.
pub weight: u16,
pub weight: Option<u16>,
/// The weight of the device against the sibling nodes.
pub leaf_weight: u16,
pub leaf_weight: Option<u16>,
}
/// Provides the ability to throttle a device (both byte/sec, and IO op/s)
@@ -437,9 +489,9 @@ pub struct BlkIoResources {
/// Whether values should be applied to the controller.
pub update_values: bool,
/// The weight of the control group against descendant nodes.
pub weight: u16,
pub weight: Option<u16>,
/// The weight of the control group against sibling nodes.
pub leaf_weight: u16,
pub leaf_weight: Option<u16>,
/// For each device, a separate weight (both normal and leaf) can be provided.
pub weight_device: Vec<BlkIoDeviceResource>,
/// Throttled read bytes/second can be provided for each device.
@@ -559,6 +611,11 @@ impl Subsystem {
c.get_path_mut().push(path);
c
}),
Subsystem::Systemd(cont) => Subsystem::Systemd({
let mut c = cont.clone();
c.get_path_mut().push(path);
c
}),
}
}
@@ -577,6 +634,149 @@ impl Subsystem {
Subsystem::NetPrio(cont) => cont,
Subsystem::HugeTlb(cont) => cont,
Subsystem::Rdma(cont) => cont,
Subsystem::Systemd(cont) => cont,
}
}
fn controller_name(&self) -> String {
self.to_controller().control_type().to_string()
}
}
/// The values for `memory.hight` or `pids.max`
#[derive(Eq, PartialEq, Copy, Clone, Debug)]
pub enum MaxValue {
/// This value is returned when the text is `"max"`.
Max,
/// When the value is a numerical value, they are returned via this enum field.
Value(i64),
}
impl Default for MaxValue {
fn default() -> Self {
MaxValue::Max
}
}
impl MaxValue {
fn to_i64(&self) -> i64 {
match self {
MaxValue::Max => -1,
MaxValue::Value(num) => *num,
}
}
fn to_string(&self) -> String {
match self {
MaxValue::Max => "max".to_string(),
MaxValue::Value(num) => num.to_string(),
}
}
}
pub fn parse_max_value(s: &String) -> Result<MaxValue> {
if s.trim() == "max" {
return Ok(MaxValue::Max);
}
match s.trim().parse() {
Ok(val) => Ok(MaxValue::Value(val)),
Err(e) => Err(Error::with_cause(ParseError, e)),
}
}
// Flat keyed
// KEY0 VAL0\n
// KEY1 VAL1\n
pub fn flat_keyed_to_vec(mut file: File) -> Result<Vec<(String, i64)>> {
let mut content = String::new();
file.read_to_string(&mut content)
.map_err(|e| Error::with_cause(ReadFailed, e))?;
let mut v = Vec::new();
for line in content.lines() {
let parts: Vec<&str> = line.split(' ').collect();
if parts.len() == 2 {
match parts[1].parse::<i64>() {
Ok(i) => {
v.push((parts[0].to_string(), i));
}
Err(_) => {}
}
}
}
Ok(v)
}
// Flat keyed
// KEY0 VAL0\n
// KEY1 VAL1\n
pub fn flat_keyed_to_hashmap(mut file: File) -> Result<HashMap<String, i64>> {
let mut content = String::new();
file.read_to_string(&mut content)
.map_err(|e| Error::with_cause(ReadFailed, e))?;
let mut h = HashMap::new();
for line in content.lines() {
let parts: Vec<&str> = line.split(' ').collect();
if parts.len() == 2 {
match parts[1].parse::<i64>() {
Ok(i) => {
h.insert(parts[0].to_string(), i);
}
Err(_) => {}
}
}
}
Ok(h)
}
// Nested keyed
// KEY0 SUB_KEY0=VAL00 SUB_KEY1=VAL01...
// KEY1 SUB_KEY0=VAL10 SUB_KEY1=VAL11...
pub fn nested_keyed_to_hashmap(mut file: File) -> Result<HashMap<String, HashMap<String, i64>>> {
let mut content = String::new();
file.read_to_string(&mut content)
.map_err(|e| Error::with_cause(ReadFailed, e))?;
let mut h = HashMap::new();
for line in content.lines() {
let parts: Vec<&str> = line.split(' ').collect();
if parts.len() == 0 {
continue;
}
let mut th = HashMap::new();
for item in parts[1..].into_iter() {
let fields: Vec<&str> = item.split('=').collect();
if fields.len() == 2 {
match fields[1].parse::<i64>() {
Ok(i) => {
th.insert(fields[0].to_string(), i);
}
Err(_) => {}
}
}
}
h.insert(parts[0].to_string(), th);
}
Ok(h)
}
/// fs::remove_dir_all or fs::remove_dir can't work with cgroup directory sometimes.
/// with error: `Os { code: 1, kind: PermissionDenied, message: "Operation not permitted" }`
pub fn libc_rmdir(p: &str) {
// with int return value
let _ = unsafe { libc::rmdir(p.as_ptr() as *const i8) };
}
/// read and parse an i64 data
pub fn read_i64_from(mut file: File) -> Result<i64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}

View File

@@ -1,16 +1,28 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `memory` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
//! [Documentation/cgroup-v1/memory.txt](https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt)
use std::collections::HashMap;
use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use std::sync::mpsc::Receiver;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::events;
use {
ControllIdentifier, ControllerInternal, Controllers, MemoryResources, Resources, Subsystem,
use crate::flat_keyed_to_hashmap;
use crate::{
ControllIdentifier, ControllerInternal, Controllers, MaxValue, MemoryResources, Resources,
Subsystem,
};
/// A controller that allows controlling the `memory` subsystem of a Cgroup.
@@ -22,6 +34,15 @@ use {
pub struct MemController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
#[derive(Default, Debug, PartialEq, Eq)]
pub struct SetMemory {
pub low: Option<MaxValue>,
pub high: Option<MaxValue>,
pub min: Option<MaxValue>,
pub max: Option<MaxValue>,
}
/// Controls statistics and controls about the OOM killer operating in this control group.
@@ -85,7 +106,7 @@ pub struct NumaStat {
fn parse_numa_stat(s: String) -> Result<NumaStat> {
// Parse the number of nodes
let nodes = (s.split_whitespace().collect::<Vec<_>>().len() - 8) / 8;
let _nodes = (s.split_whitespace().collect::<Vec<_>>().len() - 8) / 8;
let mut ls = s.lines();
let total_line = ls.next().unwrap();
let file_line = ls.next().unwrap();
@@ -109,7 +130,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
file_pages: file_line
.split(|x| x == ' ' || x == '=')
@@ -123,7 +145,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
anon_pages: anon_line
.split(|x| x == ' ' || x == '=')
@@ -137,7 +160,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
unevictable_pages: unevict_line
.split(|x| x == ' ' || x == '=')
@@ -151,7 +175,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
hierarchical_total_pages: hier_total_line
.split(|x| x == ' ' || x == '=')
@@ -165,7 +190,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
hierarchical_file_pages: hier_file_line
.split(|x| x == ' ' || x == '=')
@@ -179,7 +205,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
hierarchical_anon_pages: hier_anon_line
.split(|x| x == ' ' || x == '=')
@@ -193,7 +220,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
hierarchical_unevictable_pages: hier_unevict_line
.split(|x| x == ' ' || x == '=')
@@ -207,7 +235,8 @@ fn parse_numa_stat(s: String) -> Result<NumaStat> {
x.split("=").collect::<Vec<_>>()[1]
.parse::<u64>()
.unwrap_or(0)
}).collect()
})
.collect()
},
})
}
@@ -231,8 +260,8 @@ pub struct MemoryStat {
pub inactive_file: u64,
pub active_file: u64,
pub unevictable: u64,
pub hierarchical_memory_limit: u64,
pub hierarchical_memsw_limit: u64,
pub hierarchical_memory_limit: i64,
pub hierarchical_memsw_limit: i64,
pub total_cache: u64,
pub total_rss: u64,
pub total_rss_huge: u64,
@@ -250,52 +279,63 @@ pub struct MemoryStat {
pub total_inactive_file: u64,
pub total_active_file: u64,
pub total_unevictable: u64,
pub raw: HashMap<String, u64>,
}
fn parse_memory_stat(s: String) -> Result<MemoryStat> {
let sp: Vec<&str> = s
.split_whitespace()
.filter(|x| x.parse::<u64>().is_ok())
.collect();
let mut raw = HashMap::new();
for l in s.lines() {
let t: Vec<&str> = l.split(' ').collect();
if t.len() != 2 {
continue;
}
let n = t[1].trim().parse::<u64>();
if n.is_err() {
continue;
}
raw.insert(t[0].to_string(), n.unwrap());
}
let mut spl = sp.iter();
Ok(MemoryStat {
cache: spl.next().unwrap().parse::<u64>().unwrap(),
rss: spl.next().unwrap().parse::<u64>().unwrap(),
rss_huge: spl.next().unwrap().parse::<u64>().unwrap(),
shmem: spl.next().unwrap().parse::<u64>().unwrap(),
mapped_file: spl.next().unwrap().parse::<u64>().unwrap(),
dirty: spl.next().unwrap().parse::<u64>().unwrap(),
writeback: spl.next().unwrap().parse::<u64>().unwrap(),
swap: spl.next().unwrap().parse::<u64>().unwrap(),
pgpgin: spl.next().unwrap().parse::<u64>().unwrap(),
pgpgout: spl.next().unwrap().parse::<u64>().unwrap(),
pgfault: spl.next().unwrap().parse::<u64>().unwrap(),
pgmajfault: spl.next().unwrap().parse::<u64>().unwrap(),
inactive_anon: spl.next().unwrap().parse::<u64>().unwrap(),
active_anon: spl.next().unwrap().parse::<u64>().unwrap(),
inactive_file: spl.next().unwrap().parse::<u64>().unwrap(),
active_file: spl.next().unwrap().parse::<u64>().unwrap(),
unevictable: spl.next().unwrap().parse::<u64>().unwrap(),
hierarchical_memory_limit: spl.next().unwrap().parse::<u64>().unwrap(),
hierarchical_memsw_limit: spl.next().unwrap().parse::<u64>().unwrap(),
total_cache: spl.next().unwrap().parse::<u64>().unwrap(),
total_rss: spl.next().unwrap().parse::<u64>().unwrap(),
total_rss_huge: spl.next().unwrap().parse::<u64>().unwrap(),
total_shmem: spl.next().unwrap().parse::<u64>().unwrap(),
total_mapped_file: spl.next().unwrap().parse::<u64>().unwrap(),
total_dirty: spl.next().unwrap().parse::<u64>().unwrap(),
total_writeback: spl.next().unwrap().parse::<u64>().unwrap(),
total_swap: spl.next().unwrap().parse::<u64>().unwrap(),
total_pgpgin: spl.next().unwrap().parse::<u64>().unwrap(),
total_pgpgout: spl.next().unwrap().parse::<u64>().unwrap(),
total_pgfault: spl.next().unwrap().parse::<u64>().unwrap(),
total_pgmajfault: spl.next().unwrap().parse::<u64>().unwrap(),
total_inactive_anon: spl.next().unwrap().parse::<u64>().unwrap(),
total_active_anon: spl.next().unwrap().parse::<u64>().unwrap(),
total_inactive_file: spl.next().unwrap().parse::<u64>().unwrap(),
total_active_file: spl.next().unwrap().parse::<u64>().unwrap(),
total_unevictable: spl.next().unwrap().parse::<u64>().unwrap(),
cache: *raw.get("cache").unwrap_or(&0),
rss: *raw.get("rss").unwrap_or(&0),
rss_huge: *raw.get("rss_huge").unwrap_or(&0),
shmem: *raw.get("shmem").unwrap_or(&0),
mapped_file: *raw.get("mapped_file").unwrap_or(&0),
dirty: *raw.get("dirty").unwrap_or(&0),
writeback: *raw.get("writeback").unwrap_or(&0),
swap: *raw.get("swap").unwrap_or(&0),
pgpgin: *raw.get("pgpgin").unwrap_or(&0),
pgpgout: *raw.get("pgpgout").unwrap_or(&0),
pgfault: *raw.get("pgfault").unwrap_or(&0),
pgmajfault: *raw.get("pgmajfault").unwrap_or(&0),
inactive_anon: *raw.get("inactive_anon").unwrap_or(&0),
active_anon: *raw.get("active_anon").unwrap_or(&0),
inactive_file: *raw.get("inactive_file").unwrap_or(&0),
active_file: *raw.get("active_file").unwrap_or(&0),
unevictable: *raw.get("unevictable").unwrap_or(&0),
hierarchical_memory_limit: *raw.get("hierarchical_memory_limit").unwrap_or(&0) as i64,
hierarchical_memsw_limit: *raw.get("hierarchical_memsw_limit").unwrap_or(&0) as i64,
total_cache: *raw.get("total_cache").unwrap_or(&0),
total_rss: *raw.get("total_rss").unwrap_or(&0),
total_rss_huge: *raw.get("total_rss_huge").unwrap_or(&0),
total_shmem: *raw.get("total_shmem").unwrap_or(&0),
total_mapped_file: *raw.get("total_mapped_file").unwrap_or(&0),
total_dirty: *raw.get("total_dirty").unwrap_or(&0),
total_writeback: *raw.get("total_writeback").unwrap_or(&0),
total_swap: *raw.get("total_swap").unwrap_or(&0),
total_pgpgin: *raw.get("total_pgpgin").unwrap_or(&0),
total_pgpgout: *raw.get("total_pgpgout").unwrap_or(&0),
total_pgfault: *raw.get("total_pgfault").unwrap_or(&0),
total_pgmajfault: *raw.get("total_pgmajfault").unwrap_or(&0),
total_inactive_anon: *raw.get("total_inactive_anon").unwrap_or(&0),
total_active_anon: *raw.get("total_active_anon").unwrap_or(&0),
total_inactive_file: *raw.get("total_inactive_file").unwrap_or(&0),
total_active_file: *raw.get("total_active_file").unwrap_or(&0),
total_unevictable: *raw.get("total_unevictable").unwrap_or(&0),
raw: raw,
})
}
@@ -306,7 +346,7 @@ pub struct MemSwap {
/// How many times the limit has been hit.
pub fail_cnt: u64,
/// Memory and swap usage limit in bytes.
pub limit_in_bytes: u64,
pub limit_in_bytes: i64,
/// Current usage of memory and swap in bytes.
pub usage_in_bytes: u64,
/// The maximum observed usage of memory and swap in bytes.
@@ -320,7 +360,7 @@ pub struct Memory {
/// How many times the limit has been hit.
pub fail_cnt: u64,
/// The limit in bytes of the memory usage of the control group's tasks.
pub limit_in_bytes: u64,
pub limit_in_bytes: i64,
/// The current usage of memory by the control group's tasks.
pub usage_in_bytes: u64,
/// The maximum observed usage of memory by the control group's tasks.
@@ -342,7 +382,7 @@ pub struct Memory {
pub oom_control: OomControl,
/// Allows setting a limit to memory usage which is enforced when the system (note, _not_ the
/// control group) detects memory pressure.
pub soft_limit_in_bytes: u64,
pub soft_limit_in_bytes: i64,
/// Contains a wide array of statistics about the memory usage of the tasks in the control
/// group.
pub stat: MemoryStat,
@@ -365,7 +405,7 @@ pub struct Tcp {
pub fail_cnt: u64,
/// The limit in bytes of the memory usage of the kernel's TCP buffers by control group's
/// tasks.
pub limit_in_bytes: u64,
pub limit_in_bytes: i64,
/// The current memory used by the kernel's TCP buffers related to these tasks.
pub usage_in_bytes: u64,
/// The observed maximum usage of memory by the kernel's TCP buffers (that originated from
@@ -382,7 +422,7 @@ pub struct Kmem {
/// How many times the limit has been hit.
pub fail_cnt: u64,
/// The limit in bytes of the kernel memory used by the control group's tasks.
pub limit_in_bytes: u64,
pub limit_in_bytes: i64,
/// The current usage of kernel memory used by the control group's tasks, in bytes.
pub usage_in_bytes: u64,
/// The maximum observed usage of kernel memory used by the control group's tasks, in bytes.
@@ -405,6 +445,10 @@ impl ControllerInternal for MemController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let memres: &MemoryResources = &res.memory;
@@ -424,12 +468,76 @@ impl ControllerInternal for MemController {
impl MemController {
/// Contructs a new `MemController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
// for v2
pub fn set_mem(&self, m: SetMemory) -> Result<()> {
let values = vec![
(m.high, "memory.high"),
(m.low, "memory.low"),
(m.max, "memory.max"),
(m.min, "memory.min"),
];
for value in values {
let v = value.0;
let f = value.1;
if v.is_some() {
let v = v.unwrap().to_string();
self.open_path(f, true).and_then(|mut file| {
file.write_all(v.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})?;
}
}
Ok(())
}
// for v2
pub fn get_mem(&self) -> Result<SetMemory> {
let mut m: SetMemory = Default::default();
self.get_max_value("memory.high").map(|x| m.high = Some(x));
self.get_max_value("memory.low").map(|x| m.low = Some(x));
self.get_max_value("memory.max").map(|x| m.max = Some(x));
self.get_max_value("memory.min").map(|x| m.min = Some(x));
Ok(m)
}
fn memory_stat_v2(&self) -> Memory {
let set = self.get_mem().unwrap();
Memory {
fail_cnt: 0,
limit_in_bytes: set.max.unwrap().to_i64(),
usage_in_bytes: self
.open_path("memory.current", false)
.and_then(read_u64_from)
.unwrap_or(0),
max_usage_in_bytes: 0,
move_charge_at_immigrate: 0,
numa_stat: NumaStat::default(),
oom_control: OomControl::default(),
soft_limit_in_bytes: set.low.unwrap().to_i64(),
stat: self
.open_path("memory.stat", false)
.and_then(read_string_from)
.and_then(parse_memory_stat)
.unwrap_or(MemoryStat::default()),
swappiness: self
.open_path("memory.swap.current", false)
.and_then(read_u64_from)
.unwrap_or(0),
use_hierarchy: 0,
}
}
@@ -439,6 +547,10 @@ impl MemController {
/// See the individual fields for more explanation, and as always, remember to consult the
/// kernel Documentation and/or sources.
pub fn memory_stat(&self) -> Memory {
if self.v2 {
return self.memory_stat_v2();
}
Memory {
fail_cnt: self
.open_path("memory.failcnt", false)
@@ -446,7 +558,7 @@ impl MemController {
.unwrap_or(0),
limit_in_bytes: self
.open_path("memory.limit_in_bytes", false)
.and_then(read_u64_from)
.and_then(read_i64_from)
.unwrap_or(0),
usage_in_bytes: self
.open_path("memory.usage_in_bytes", false)
@@ -472,7 +584,7 @@ impl MemController {
.unwrap_or(OomControl::default()),
soft_limit_in_bytes: self
.open_path("memory.soft_limit_in_bytes", false)
.and_then(read_u64_from)
.and_then(read_i64_from)
.unwrap_or(0),
stat: self
.open_path("memory.stat", false)
@@ -499,8 +611,8 @@ impl MemController {
.unwrap_or(0),
limit_in_bytes: self
.open_path("memory.kmem.limit_in_bytes", false)
.and_then(read_u64_from)
.unwrap_or(0),
.and_then(read_i64_from)
.unwrap_or(-1),
usage_in_bytes: self
.open_path("memory.kmem.usage_in_bytes", false)
.and_then(read_u64_from)
@@ -526,7 +638,7 @@ impl MemController {
.unwrap_or(0),
limit_in_bytes: self
.open_path("memory.kmem.tcp.limit_in_bytes", false)
.and_then(read_u64_from)
.and_then(read_i64_from)
.unwrap_or(0),
usage_in_bytes: self
.open_path("memory.kmem.tcp.usage_in_bytes", false)
@@ -539,9 +651,32 @@ impl MemController {
}
}
pub fn memswap_v2(&self) -> MemSwap {
MemSwap {
fail_cnt: self
.open_path("memory.swap.events", false)
.and_then(flat_keyed_to_hashmap)
.and_then(|x| Ok(*x.get("fail").unwrap_or(&0) as u64))
.unwrap(),
limit_in_bytes: self
.open_path("memory.swap.max", false)
.and_then(read_i64_from)
.unwrap_or(0),
usage_in_bytes: self
.open_path("memory.swap.current", false)
.and_then(read_u64_from)
.unwrap_or(0),
max_usage_in_bytes: 0,
}
}
/// Gathers information about the memory usage of the control group including the swap usage
/// (if any).
pub fn memswap(&self) -> MemSwap {
if self.v2 {
return self.memswap_v2();
}
MemSwap {
fail_cnt: self
.open_path("memory.memsw.failcnt", false)
@@ -549,7 +684,7 @@ impl MemController {
.unwrap_or(0),
limit_in_bytes: self
.open_path("memory.memsw.limit_in_bytes", false)
.and_then(read_u64_from)
.and_then(read_i64_from)
.unwrap_or(0),
usage_in_bytes: self
.open_path("memory.memsw.usage_in_bytes", false)
@@ -564,11 +699,10 @@ impl MemController {
/// Reset the fail counter
pub fn reset_fail_count(&self) -> Result<()> {
self.open_path("memory.failcnt", true)
.and_then(|mut file| {
file.write_all("0".to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
self.open_path("memory.failcnt", true).and_then(|mut file| {
file.write_all("0".to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Reset the kernel memory fail counter
@@ -599,16 +733,19 @@ impl MemController {
}
/// Set the memory usage limit of the control group, in bytes.
pub fn set_limit(&self, limit: u64) -> Result<()> {
self.open_path("memory.limit_in_bytes", true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn set_limit(&self, limit: i64) -> Result<()> {
let mut file = "memory.limit_in_bytes";
if self.v2 {
file = "memory.max";
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Set the kernel memory limit of the control group, in bytes.
pub fn set_kmem_limit(&self, limit: u64) -> Result<()> {
pub fn set_kmem_limit(&self, limit: i64) -> Result<()> {
self.open_path("memory.kmem.limit_in_bytes", true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
@@ -617,16 +754,19 @@ impl MemController {
}
/// Set the memory+swap limit of the control group, in bytes.
pub fn set_memswap_limit(&self, limit: u64) -> Result<()> {
self.open_path("memory.memsw.limit_in_bytes", true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn set_memswap_limit(&self, limit: i64) -> Result<()> {
let mut file = "memory.memsw.limit_in_bytes";
if self.v2 {
file = "memory.swap.max";
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Set how much kernel memory can be used for TCP-related buffers by the control group.
pub fn set_tcp_limit(&self, limit: u64) -> Result<()> {
pub fn set_tcp_limit(&self, limit: i64) -> Result<()> {
self.open_path("memory.kmem.tcp.limit_in_bytes", true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
@@ -638,12 +778,15 @@ impl MemController {
///
/// This limit is enforced when the system is nearing OOM conditions. Contrast this with the
/// hard limit, which is _always_ enforced.
pub fn set_soft_limit(&self, limit: u64) -> Result<()> {
self.open_path("memory.soft_limit_in_bytes", true)
.and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
pub fn set_soft_limit(&self, limit: i64) -> Result<()> {
let mut file = "memory.soft_limit_in_bytes";
if self.v2 {
file = "memory.low"
}
self.open_path(file, true).and_then(|mut file| {
file.write_all(limit.to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
/// Set how likely the kernel is to swap out parts of the address space used by the control
@@ -657,6 +800,22 @@ impl MemController {
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
pub fn disable_oom_killer(&self) -> Result<()> {
self.open_path("memory.oom_control", true)
.and_then(|mut file| {
file.write_all("1".to_string().as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}
pub fn register_oom_event(&self, key: &str) -> Result<Receiver<String>> {
if self.v2 {
events::notify_on_oom_v2(key, self.get_path())
} else {
events::notify_on_oom_v1(key, self.get_path())
}
}
}
impl ControllIdentifier for MemController {
@@ -682,7 +841,21 @@ impl<'a> From<&'a Subsystem> for &'a MemController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
fn read_i64_from(mut file: File) -> Result<i64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
@@ -697,9 +870,10 @@ fn read_string_from(mut file: File) -> Result<String> {
#[cfg(test)]
mod tests {
use memory::{
use crate::memory::{
parse_memory_stat, parse_numa_stat, parse_oom_control, MemoryStat, NumaStat, OomControl,
};
use std::collections::HashMap;
static GOOD_VALUE: &str = "\
total=51189 N0=51189 N1=123
@@ -800,6 +974,7 @@ total_unevictable 81920
#[test]
fn test_parse_memory_stat() {
let ok = parse_memory_stat(GOOD_MEMORYSTAT_VAL.to_string()).unwrap();
let raw = ok.raw.clone();
assert_eq!(
ok,
MemoryStat {
@@ -839,6 +1014,7 @@ total_unevictable 81920
total_inactive_file: 1272135680,
total_active_file: 2338816000,
total_unevictable: 81920,
raw: raw,
}
);
}

View File

@@ -1,3 +1,8 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `net_cls` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,12 +11,11 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {
ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources,
Subsystem,
use crate::{
ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, Subsystem,
};
/// A controller that allows controlling the `net_cls` subsystem of a Cgroup.
@@ -76,7 +80,10 @@ impl<'a> From<&'a Subsystem> for &'a NetClsController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
@@ -97,7 +104,8 @@ impl NetClsController {
self.open_path("net_cls.classid", true)
.and_then(|mut file| {
let s = format!("{:#08X}", class);
file.write_all(s.as_ref()).map_err(|e| Error::with_cause(WriteFailed, e))
file.write_all(s.as_ref())
.map_err(|e| Error::with_cause(WriteFailed, e))
})
}

View File

@@ -1,3 +1,8 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `net_prio` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -7,12 +12,11 @@ use std::fs::File;
use std::io::{BufRead, BufReader, Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {
ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources,
Subsystem,
use crate::{
ControllIdentifier, ControllerInternal, Controllers, NetworkResources, Resources, Subsystem,
};
/// A controller that allows controlling the `net_prio` subsystem of a Cgroup.
@@ -77,7 +81,10 @@ impl<'a> From<&'a Subsystem> for &'a NetPrioController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}

View File

@@ -1,12 +1,17 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `perf_event` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
//! [tools/perf/Documentation/perf-record.txt](https://raw.githubusercontent.com/torvalds/linux/master/tools/perf/Documentation/perf-record.txt)
use std::path::PathBuf;
use error::*;
use crate::error::*;
use {ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
/// A controller that allows controlling the `perf_event` subsystem of a Cgroup.
///

View File

@@ -1,3 +1,9 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `pids` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,11 +12,12 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {
ControllIdentifier, ControllerInternal, Controllers, PidResources, Resources, Subsystem,
use crate::{
parse_max_value, ControllIdentifier, ControllerInternal, Controllers, MaxValue, PidResources,
Resources, Subsystem,
};
/// A controller that allows controlling the `pids` subsystem of a Cgroup.
@@ -18,22 +25,7 @@ use {
pub struct PidController {
base: PathBuf,
path: PathBuf,
}
/// The values found in the `pids.max` file in a Cgroup's `pids` subsystem.
#[derive(Eq, PartialEq, Copy, Clone, Debug)]
pub enum PidMax {
/// This value is returned when the text found `pids.max` is `"max"`.
Max,
/// When the value in `pids.max` is a numerical value, they are returned via this enum field.
Value(i64),
}
impl Default for PidMax {
/// By default, (as per the kernel) `pids.max` should contain `"max"`.
fn default() -> Self {
PidMax::Max
}
v2: bool,
}
impl ControllerInternal for PidController {
@@ -50,6 +42,10 @@ impl ControllerInternal for PidController {
&self.base
}
fn is_v2(&self) -> bool {
self.v2
}
fn apply(&self, res: &Resources) -> Result<()> {
// get the resources that apply to this controller
let pidres: &PidResources = &res.pid;
@@ -99,7 +95,10 @@ impl<'a> From<&'a Subsystem> for &'a PidController {
fn read_u64_from(mut file: File) -> Result<u64> {
let mut string = String::new();
match file.read_to_string(&mut string) {
Ok(_) => string.trim().parse().map_err(|e| Error::with_cause(ParseError, e)),
Ok(_) => string
.trim()
.parse()
.map_err(|e| Error::with_cause(ParseError, e)),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
}
@@ -107,12 +106,15 @@ fn read_u64_from(mut file: File) -> Result<u64> {
impl PidController {
/// Constructors a new `PidController` instance, with `oroot` serving as the controller's root
/// directory.
pub fn new(oroot: PathBuf) -> Self {
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
root.push(Self::controller_type().to_string());
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
@@ -140,19 +142,12 @@ impl PidController {
}
/// The maximum number of processes that can exist at one time in the control group.
pub fn get_pid_max(&self) -> Result<PidMax> {
pub fn get_pid_max(&self) -> Result<MaxValue> {
self.open_path("pids.max", false).and_then(|mut file| {
let mut string = String::new();
let res = file.read_to_string(&mut string);
match res {
Ok(_) => if string.trim() == "max" {
Ok(PidMax::Max)
} else {
match string.trim().parse() {
Ok(val) => Ok(PidMax::Value(val)),
Err(e) => Err(Error::with_cause(ParseError, e)),
}
},
Ok(_) => parse_max_value(&string),
Err(e) => Err(Error::with_cause(ReadFailed, e)),
}
})
@@ -163,12 +158,9 @@ impl PidController {
/// Note that if `get_pid_current()` returns a higher number than what you
/// are about to set (`max_pid`), then no processess will be killed. Additonally, attaching
/// extra processes to a control group disregards the limit.
pub fn set_pid_max(&self, max_pid: PidMax) -> Result<()> {
pub fn set_pid_max(&self, max_pid: MaxValue) -> Result<()> {
self.open_path("pids.max", true).and_then(|mut file| {
let string_to_write = match max_pid {
PidMax::Max => "max".to_string(),
PidMax::Value(num) => num.to_string(),
};
let string_to_write = max_pid.to_string();
match file.write_all(string_to_write.as_ref()) {
Ok(_) => Ok(()),
Err(e) => Err(Error::with_cause(WriteFailed, e)),

View File

@@ -1,3 +1,8 @@
// Copyright (c) 2018 Levente Kurusa
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `rdma` cgroup subsystem.
//!
//! See the Kernel's documentation for more information about this subsystem, found at:
@@ -6,10 +11,10 @@ use std::fs::File;
use std::io::{Read, Write};
use std::path::PathBuf;
use error::*;
use error::ErrorKind::*;
use crate::error::ErrorKind::*;
use crate::error::*;
use {ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
/// A controller that allows controlling the `rdma` subsystem of a Cgroup.
///

76
src/systemd.rs Normal file
View File

@@ -0,0 +1,76 @@
// Copyright (c) 2020 Ant Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! This module contains the implementation of the `systemd` cgroup subsystem.
//!
use std::path::PathBuf;
use crate::error::ErrorKind::*;
use crate::error::*;
use crate::{ControllIdentifier, ControllerInternal, Controllers, Resources, Subsystem};
/// A controller that allows controlling the `systemd` subsystem of a Cgroup.
///
#[derive(Debug, Clone)]
pub struct SystemdController {
base: PathBuf,
path: PathBuf,
v2: bool,
}
impl ControllerInternal for SystemdController {
fn control_type(&self) -> Controllers {
Controllers::Systemd
}
fn get_path(&self) -> &PathBuf {
&self.path
}
fn get_path_mut(&mut self) -> &mut PathBuf {
&mut self.path
}
fn get_base(&self) -> &PathBuf {
&self.base
}
fn apply(&self, _res: &Resources) -> Result<()> {
Ok(())
}
}
impl ControllIdentifier for SystemdController {
fn controller_type() -> Controllers {
Controllers::Systemd
}
}
impl<'a> From<&'a Subsystem> for &'a SystemdController {
fn from(sub: &'a Subsystem) -> &'a SystemdController {
unsafe {
match sub {
Subsystem::Systemd(c) => c,
_ => {
assert_eq!(1, 0);
::std::mem::uninitialized()
}
}
}
}
}
impl SystemdController {
/// Constructs a new `SystemdController` with `oroot` serving as the root of the control group.
pub fn new(oroot: PathBuf, v2: bool) -> Self {
let mut root = oroot;
if !v2 {
root.push(Self::controller_type().to_string());
}
Self {
base: root.clone(),
path: root,
v2: v2,
}
}
}

View File

@@ -1,22 +1,28 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Some simple tests covering the builder pattern for control groups.
extern crate cgroups;
use cgroups::*;
use cgroups::cpu::*;
use cgroups::devices::*;
use cgroups::pid::*;
use cgroups::memory::*;
use cgroups::net_cls::*;
use cgroups::hugetlb::*;
use cgroups::blkio::*;
use cgroups::cgroup_builder::*;
use cgroups::cpu::*;
use cgroups::devices::*;
use cgroups::hugetlb::*;
use cgroups::memory::*;
use cgroups::net_cls::*;
use cgroups::pid::*;
use cgroups::*;
#[test]
pub fn test_cpu_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_cpu_res_build", &v1)
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_cpu_res_build", h)
.cpu()
.shares(85)
.done()
.shares(85)
.done()
.build();
{
@@ -30,19 +36,22 @@ pub fn test_cpu_res_build() {
#[test]
pub fn test_memory_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_memory_res_build", &v1)
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_memory_res_build", h)
.memory()
.kernel_memory_limit(128 * 1024 * 1024)
.swappiness(70)
.memory_hard_limit(1024 * 1024 * 1024)
.done()
.kernel_memory_limit(128 * 1024 * 1024)
.swappiness(70)
.memory_hard_limit(1024 * 1024 * 1024)
.done()
.build();
{
let c: &MemController = cg.controller_of().unwrap();
assert_eq!(c.kmem_stat().limit_in_bytes, 128 * 1024 * 1024);
assert_eq!(c.memory_stat().swappiness, 70);
if !c.v2() {
assert_eq!(c.kmem_stat().limit_in_bytes, 128 * 1024 * 1024);
assert_eq!(c.memory_stat().swappiness, 70);
}
assert_eq!(c.memory_stat().limit_in_bytes, 1024 * 1024 * 1024);
}
@@ -51,17 +60,18 @@ pub fn test_memory_res_build() {
#[test]
pub fn test_pid_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_pid_res_build", &v1)
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_pid_res_build", h)
.pid()
.maximum_number_of_processes(PidMax::Value(123))
.done()
.maximum_number_of_processes(MaxValue::Value(123))
.done()
.build();
{
let c: &PidController = cg.controller_of().unwrap();
assert!(c.get_pid_max().is_ok());
assert_eq!(c.get_pid_max().unwrap(), PidMax::Value(123));
assert_eq!(c.get_pid_max().unwrap(), MaxValue::Value(123));
}
cg.delete();
@@ -70,37 +80,43 @@ pub fn test_pid_res_build() {
#[test]
#[ignore] // ignore this test for now, not sure why my kernel doesn't like it
pub fn test_devices_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_devices_res_build", &v1)
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_devices_res_build", h)
.devices()
.device(1, 6, DeviceType::Char, true,
vec![DevicePermissions::Read])
.done()
.device(1, 6, DeviceType::Char, true, vec![DevicePermissions::Read])
.done()
.build();
{
let c: &DevicesController = cg.controller_of().unwrap();
assert!(c.allowed_devices().is_ok());
assert_eq!(c.allowed_devices().unwrap(), vec![
DeviceResource {
allow: true,
devtype: DeviceType::Char,
major: 1,
minor: 6,
access: vec![DevicePermissions::Read],
}
]);
assert_eq!(
c.allowed_devices().unwrap(),
vec![DeviceResource {
allow: true,
devtype: DeviceType::Char,
major: 1,
minor: 6,
access: vec![DevicePermissions::Read],
}]
);
}
cg.delete();
}
#[test]
pub fn test_network_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_network_res_build", &v1)
let h = cgroups::hierarchies::auto();
if h.v2() {
// FIXME add cases for v2
return;
}
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_network_res_build", h)
.network()
.class_id(1337)
.done()
.class_id(1337)
.done()
.build();
{
@@ -113,28 +129,38 @@ pub fn test_network_res_build() {
#[test]
pub fn test_hugepages_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_hugepages_res_build", &v1)
let h = cgroups::hierarchies::auto();
if h.v2() {
// FIXME add cases for v2
return;
}
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_hugepages_res_build", h)
.hugepages()
.limit("2MB".to_string(), 4 * 2 * 1024 * 1024)
.done()
.limit("2MB".to_string(), 4 * 2 * 1024 * 1024)
.done()
.build();
{
let c: &HugeTlbController = cg.controller_of().unwrap();
assert!(c.limit_in_bytes(&"2MB".to_string()).is_ok());
assert_eq!(c.limit_in_bytes(&"2MB".to_string()).unwrap(), 4 * 2 * 1024 * 1024);
assert_eq!(
c.limit_in_bytes(&"2MB".to_string()).unwrap(),
4 * 2 * 1024 * 1024
);
}
cg.delete();
}
#[test]
#[ignore] // high version kernel not support `blkio.weight`
pub fn test_blkio_res_build() {
let v1 = ::hierarchies::V1::new();
let cg: Cgroup = CgroupBuilder::new("test_blkio_res_build", &v1)
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg: Cgroup = CgroupBuilder::new("test_blkio_res_build", h)
.blkio()
.weight(100)
.done()
.weight(Some(100))
.done()
.build();
{

View File

@@ -1,18 +1,28 @@
//! Simple unit tests about the control groups system.
extern crate cgroups;
use cgroups::{Cgroup, CgroupPid};
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
extern crate libc;
extern crate nix;
//! Simple unit tests about the control groups system.
use cgroups::memory::{MemController, SetMemory};
use cgroups::Controller;
use cgroups::{Cgroup, CgroupPid, Hierarchy, Subsystem};
use std::collections::HashMap;
#[test]
fn test_tasks_iterator() {
let hier = cgroups::hierarchies::V1::new();
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let pid = libc::pid_t::from(nix::unistd::getpid()) as u64;
let cg = Cgroup::new(&hier, String::from("test_tasks_iterator"));
let cg = Cgroup::new(h, String::from("test_tasks_iterator"));
{
// Add a task to the control group.
cg.add_task(CgroupPid::from(pid));
cg.add_task(CgroupPid::from(pid)).unwrap();
use std::{thread, time};
thread::sleep(time::Duration::from_millis(100));
let mut tasks = cg.tasks().into_iter();
// Verify that the task is indeed in the control group
assert_eq!(tasks.next(), Some(CgroupPid::from(pid)));
@@ -27,3 +37,81 @@ fn test_tasks_iterator() {
}
cg.delete();
}
#[test]
fn test_cgroup_with_relative_paths() {
if cgroups::hierarchies::is_cgroup2_unified_mode() {
return;
}
let h = cgroups::hierarchies::auto();
let cgroup_root = h.root();
let h = Box::new(&*h);
let mut relative_paths = HashMap::new();
let mem_relative_path = "/mmm/abc/def";
relative_paths.insert("memory".to_string(), mem_relative_path.to_string());
let cgroup_name = "test_cgroup_with_relative_paths";
let cg = Cgroup::new_with_relative_paths(h, String::from(cgroup_name), relative_paths);
{
let subsystems = cg.subsystems();
subsystems.into_iter().for_each(|sub| match sub {
Subsystem::Pid(c) => {
let cgroup_path = c.path().to_str().unwrap();
let relative_path = "/pids/";
// cgroup_path = cgroup_root + relative_path + cgroup_name
assert_eq!(
cgroup_path,
format!(
"{}{}{}",
cgroup_root.to_str().unwrap(),
relative_path,
cgroup_name
)
);
}
Subsystem::Mem(c) => {
let cgroup_path = c.path().to_str().unwrap();
// cgroup_path = cgroup_root + relative_path + cgroup_name
assert_eq!(
cgroup_path,
format!(
"{}/memory{}/{}",
cgroup_root.to_str().unwrap(),
mem_relative_path,
cgroup_name
)
);
}
_ => {}
});
}
cg.delete();
}
#[test]
fn test_cgroup_v2() {
if !cgroups::hierarchies::is_cgroup2_unified_mode() {
return;
}
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new_with_relative_paths(h, String::from("test_v2"), HashMap::new());
let mem_controller: &MemController = cg.controller_of().unwrap();
let (mem, swp, rev) = (4 * 1024 * 1000, 2 * 1024 * 1000, 1024 * 1000);
let _ = mem_controller.set_limit(mem);
let _ = mem_controller.set_memswap_limit(swp);
let _ = mem_controller.set_soft_limit(rev);
let memory_stat = mem_controller.memory_stat();
println!("memory_stat {:?}", memory_stat);
assert_eq!(mem, memory_stat.limit_in_bytes);
assert_eq!(rev, memory_stat.soft_limit_in_bytes);
let memswap = mem_controller.memswap();
println!("memswap {:?}", memswap);
assert_eq!(swp, memswap.limit_in_bytes);
cg.delete();
}

61
tests/cpu.rs Normal file
View File

@@ -0,0 +1,61 @@
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Simple unit tests about the CPU control groups system.
use cgroups::cpu::CpuController;
use cgroups::error::ErrorKind;
use cgroups::{Cgroup, CgroupPid, CpuResources, Hierarchy, Resources};
use std::fs;
#[test]
fn test_cfs_quota_and_periods() {
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_cfs_quota_and_periods"));
let cpu_controller: &CpuController = cg.controller_of().unwrap();
let current_quota = cpu_controller.cfs_quota().unwrap();
let current_peroid = cpu_controller.cfs_period().unwrap();
// verify default value
// The default is “max 100000”.
assert_eq!(-1, current_quota);
assert_eq!(100000, current_peroid);
// case 1 set quota
let r = cpu_controller.set_cfs_quota(2000);
let current_quota = cpu_controller.cfs_quota().unwrap();
let current_peroid = cpu_controller.cfs_period().unwrap();
assert_eq!(2000, current_quota);
assert_eq!(100000, current_peroid);
// case 2 set period
cpu_controller.set_cfs_period(1000000);
let current_quota = cpu_controller.cfs_quota().unwrap();
let current_peroid = cpu_controller.cfs_period().unwrap();
assert_eq!(2000, current_quota);
assert_eq!(1000000, current_peroid);
// case 3 set both quota and period
cpu_controller.set_cfs_quota_and_period(Some(5000), Some(100000));
let current_quota = cpu_controller.cfs_quota().unwrap();
let current_peroid = cpu_controller.cfs_period().unwrap();
assert_eq!(5000, current_quota);
assert_eq!(100000, current_peroid);
// case 4 set both quota and period, set quota to -1
cpu_controller.set_cfs_quota_and_period(Some(-1), None);
let current_quota = cpu_controller.cfs_quota().unwrap();
let current_peroid = cpu_controller.cfs_period().unwrap();
assert_eq!(-1, current_quota);
assert_eq!(100000, current_peroid);
cg.delete();
}

View File

@@ -1,13 +1,20 @@
extern crate cgroups;
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
use cgroups::cpuset::CpuSetController;
use cgroups::error::ErrorKind;
use cgroups::Cgroup;
use cgroups::{Cgroup, CgroupPid, CpuResources, Hierarchy, Resources};
use std::fs;
#[test]
fn test_cpuset_memory_pressure_root_cg() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("test_cpuset_memory_pressure_root_cg"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_cpuset_memory_pressure_root_cg"));
{
let cpuset: &CpuSetController = cg.controller_of().unwrap();
@@ -17,3 +24,73 @@ fn test_cpuset_memory_pressure_root_cg() {
}
cg.delete();
}
#[test]
fn test_cpuset_set_cpus() {
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_cpuset_set_cpus"));
{
let cpuset: &CpuSetController = cg.controller_of().unwrap();
let set = cpuset.cpuset();
if cg.v2() {
assert_eq!(0, set.cpus.len());
} else {
// for cgroup v1, cpuset is copied from parent.
assert_eq!(true, set.cpus.len() > 0);
}
// 0
let r = cpuset.set_cpus("0");
assert_eq!(true, r.is_ok());
let set = cpuset.cpuset();
assert_eq!(1, set.cpus.len());
assert_eq!((0, 0), set.cpus[0]);
// all cpus in system
let cpus =
fs::read_to_string("/sys/fs/cgroup/cpuset.cpus.effective").unwrap_or("".to_string());
let cpus = cpus.trim();
if cpus != "" {
let r = cpuset.set_cpus(&cpus);
assert_eq!(true, r.is_ok());
let set = cpuset.cpuset();
assert_eq!(1, set.cpus.len());
assert_eq!(format!("{}-{}", set.cpus[0].0, set.cpus[0].1), cpus);
}
}
cg.delete();
}
#[test]
fn test_cpuset_set_cpus_add_task() {
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_cpuset_set_cpus_add_task/sub-dir"));
let cpuset: &CpuSetController = cg.controller_of().unwrap();
let set = cpuset.cpuset();
if cg.v2() {
assert_eq!(0, set.cpus.len());
} else {
// for cgroup v1, cpuset is copied from parent.
assert_eq!(true, set.cpus.len() > 0);
}
// Add a task to the control group.
let pid_i = libc::pid_t::from(nix::unistd::getpid()) as u64;
let _ = cg.add_task(CgroupPid::from(pid_i));
let tasks = cg.tasks();
assert_eq!(true, tasks.len() > 0);
println!("tasks after added: {:?}", tasks);
// remove task
let _ = cg.remove_task(CgroupPid::from(pid_i));
let tasks = cg.tasks();
println!("tasks after deleted: {:?}", tasks);
assert_eq!(0, tasks.len());
cg.delete();
}

View File

@@ -1,13 +1,24 @@
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Integration tests about the devices subsystem
extern crate cgroups;
use cgroups::devices::{DevicePermissions, DeviceType, DevicesController};
use cgroups::{Cgroup, DeviceResource};
use cgroups::{Cgroup, DeviceResource, Hierarchy};
#[test]
fn test_devices_parsing() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("test_devices_parsing"));
// now only v2
if cgroups::hierarchies::is_cgroup2_unified_mode() {
return;
}
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_devices_parsing"));
{
let devices: &DevicesController = cg.controller_of().unwrap();

48
tests/hugetlb.rs Normal file
View File

@@ -0,0 +1,48 @@
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Integration tests about the hugetlb subsystem
use cgroups::hugetlb::{self, HugeTlbController};
use cgroups::Controller;
use cgroups::{Cgroup, Hierarchy};
use cgroups::error::ErrorKind::*;
use cgroups::error::*;
use std::fs;
#[test]
fn test_hugetlb_sizes() {
// now only v2
if cgroups::hierarchies::is_cgroup2_unified_mode() {
return;
}
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_hugetlb_sizes"));
{
let hugetlb_controller: &HugeTlbController = cg.controller_of().unwrap();
let sizes = hugetlb_controller.get_sizes();
// test sizes count
let sizes = hugetlb_controller.get_sizes();
let sizes_count = fs::read_dir(hugetlb::HUGEPAGESIZE_DIR).unwrap().count();
assert_eq!(sizes.len(), sizes_count);
for size in sizes {
let supported = hugetlb_controller.size_supported(&size);
assert_eq!(supported, true);
assert_no_error(hugetlb_controller.failcnt(&size));
assert_no_error(hugetlb_controller.limit_in_bytes(&size));
assert_no_error(hugetlb_controller.usage_in_bytes(&size));
assert_no_error(hugetlb_controller.max_usage_in_bytes(&size));
}
}
cg.delete();
}
fn assert_no_error(r: Result<u64>) {
assert_eq!(!r.is_err(), true)
}

93
tests/memory.rs Normal file
View File

@@ -0,0 +1,93 @@
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Integration tests about the hugetlb subsystem
use cgroups::memory::{MemController, SetMemory};
use cgroups::Controller;
use cgroups::{Cgroup, MaxValue};
#[test]
fn test_disable_oom_killer() {
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_disable_oom_killer"));
{
let mem_controller: &MemController = cg.controller_of().unwrap();
// before disable
let m = mem_controller.memory_stat();
assert_eq!(m.oom_control.oom_kill_disable, false);
// now only v1
if !mem_controller.v2() {
// disable oom killer
let r = mem_controller.disable_oom_killer();
assert_eq!(r.is_err(), false);
// after disable
let m = mem_controller.memory_stat();
assert_eq!(m.oom_control.oom_kill_disable, true);
}
}
cg.delete();
}
#[test]
fn set_mem_v2() {
let h = cgroups::hierarchies::auto();
if !h.v2() {
return;
}
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("set_mem_v2"));
{
let mem_controller: &MemController = cg.controller_of().unwrap();
// before disable
let m = mem_controller.get_mem().unwrap();
// case 1: get default value
assert_eq!(m.low, Some(MaxValue::Value(0)));
assert_eq!(m.min, Some(MaxValue::Value(0)));
assert_eq!(m.high, Some(MaxValue::Max));
assert_eq!(m.max, Some(MaxValue::Max));
// case 2: set parts
let m = SetMemory {
low: Some(MaxValue::Value(1024 * 1024 * 2)),
high: Some(MaxValue::Value(1024 * 1024 * 1024 * 2)),
min: Some(MaxValue::Value(1024 * 1024 * 3)),
max: None,
};
let r = mem_controller.set_mem(m);
assert_eq!(true, r.is_ok());
let m = mem_controller.get_mem().unwrap();
// get
assert_eq!(m.low, Some(MaxValue::Value(1024 * 1024 * 2)));
assert_eq!(m.min, Some(MaxValue::Value(1024 * 1024 * 3)));
assert_eq!(m.high, Some(MaxValue::Value(1024 * 1024 * 1024 * 2)));
assert_eq!(m.max, Some(MaxValue::Max));
// case 3: set parts
let m = SetMemory {
max: Some(MaxValue::Value(1024 * 1024 * 1024 * 2)),
min: Some(MaxValue::Value(1024 * 1024 * 4)),
high: Some(MaxValue::Max),
low: None,
};
let r = mem_controller.set_mem(m);
assert_eq!(true, r.is_ok());
let m = mem_controller.get_mem().unwrap();
// get
assert_eq!(m.low, Some(MaxValue::Value(1024 * 1024 * 2)));
assert_eq!(m.min, Some(MaxValue::Value(1024 * 1024 * 4)));
assert_eq!(m.max, Some(MaxValue::Value(1024 * 1024 * 1024 * 2)));
assert_eq!(m.high, Some(MaxValue::Max));
}
cg.delete();
}

View File

@@ -1,36 +1,41 @@
//! Integration tests about the pids subsystem
extern crate cgroups;
use cgroups::pid::{PidController, PidMax};
use cgroups::Controller;
use cgroups::{Cgroup, CgroupPid, PidResources, Resources};
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
//! Integration tests about the pids subsystem
use cgroups::pid::PidController;
use cgroups::Controller;
use cgroups::{Cgroup, CgroupPid, Hierarchy, MaxValue, PidResources, Resources};
extern crate nix;
use nix::sys::wait::{waitpid, WaitStatus};
use nix::unistd::{fork, ForkResult, Pid};
extern crate libc;
use libc::pid_t;
use std::thread;
#[test]
fn create_and_delete_cgroup() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("create_and_delete_cgroup"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("create_and_delete_cgroup"));
{
let pidcontroller: &PidController = cg.controller_of().unwrap();
pidcontroller.set_pid_max(PidMax::Value(1337));
pidcontroller.set_pid_max(MaxValue::Value(1337));
let max = pidcontroller.get_pid_max();
assert!(max.is_ok());
assert_eq!(max.unwrap(), PidMax::Value(1337));
assert_eq!(max.unwrap(), MaxValue::Value(1337));
}
cg.delete();
}
#[test]
fn test_pids_current_is_zero() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("test_pids_current_is_zero"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_pids_current_is_zero"));
{
let pidcontroller: &PidController = cg.controller_of().unwrap();
let current = pidcontroller.get_pid_current();
@@ -41,8 +46,9 @@ fn test_pids_current_is_zero() {
#[test]
fn test_pids_events_is_zero() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("test_pids_events_is_zero"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_pids_events_is_zero"));
{
let pidcontroller: &PidController = cg.controller_of().unwrap();
let events = pidcontroller.get_pid_events();
@@ -54,8 +60,9 @@ fn test_pids_events_is_zero() {
#[test]
fn test_pid_events_is_not_zero() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("test_pid_events_is_not_zero"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("test_pid_events_is_not_zero"));
{
let pids: &PidController = cg.controller_of().unwrap();
let before = pids.get_pid_events();
@@ -64,13 +71,13 @@ fn test_pid_events_is_not_zero() {
match fork() {
Ok(ForkResult::Parent { child, .. }) => {
// move the process into the control group
pids.add_task(&(pid_t::from(child) as u64).into());
let _ = pids.add_task(&(pid_t::from(child) as u64).into());
println!("added task to cg: {:?}", child);
// Set limit to one
pids.set_pid_max(PidMax::Value(1));
println!("err = {:?}", pids.get_pid_max());
let _ = pids.set_pid_max(MaxValue::Value(1));
println!("current pid.max = {:?}", pids.get_pid_max());
// wait on the child
let res = waitpid(child, None);
@@ -87,7 +94,7 @@ fn test_pid_events_is_not_zero() {
}
Ok(ForkResult::Child) => loop {
let pids_max = pids.get_pid_max();
if pids_max.is_ok() && pids_max.unwrap() == PidMax::Value(1) {
if pids_max.is_ok() && pids_max.unwrap() == MaxValue::Value(1) {
if let Err(_) = fork() {
unsafe { libc::exit(0) };
} else {

View File

@@ -1,18 +1,23 @@
//! Integration test about setting resources using `apply()`
extern crate cgroups;
// Copyright (c) 2018 Levente Kurusa
// Copyright (c) 2020 And Group
//
// SPDX-License-Identifier: Apache-2.0 or MIT
//
use cgroups::pid::{PidController, PidMax};
use cgroups::{Cgroup, PidResources, Resources};
//! Integration test about setting resources using `apply()`
use cgroups::pid::PidController;
use cgroups::{Cgroup, Hierarchy, MaxValue, PidResources, Resources};
#[test]
fn pid_resources() {
let hier = cgroups::hierarchies::V1::new();
let cg = Cgroup::new(&hier, String::from("pid_resources"));
let h = cgroups::hierarchies::auto();
let h = Box::new(&*h);
let cg = Cgroup::new(h, String::from("pid_resources"));
{
let res = Resources {
pid: PidResources {
update_values: true,
maximum_number_of_processes: PidMax::Value(512),
maximum_number_of_processes: MaxValue::Value(512),
},
..Default::default()
};
@@ -22,7 +27,7 @@ fn pid_resources() {
let pidcontroller: &PidController = cg.controller_of().unwrap();
let pid_max = pidcontroller.get_pid_max();
assert_eq!(pid_max.is_ok(), true);
assert_eq!(pid_max.unwrap(), PidMax::Value(512));
assert_eq!(pid_max.unwrap(), MaxValue::Value(512));
}
cg.delete();
}

View File

@@ -1,4 +1,9 @@
#!/bin/sh
#
# Copyright (c) 2018 Levente Kurusa
#
# SPDX-License-Identifier: Apache-2.0 or MIT
#
CONTROL_GROUPS=`cargo test -- --list 2>/dev/null | egrep 'test$' | egrep -v '^src' | cut -d':' -f1`

View File

@@ -1,4 +1,9 @@
#!/bin/sh
#
# Copyright (c) 2018 Levente Kurusa
#
# SPDX-License-Identifier: Apache-2.0 or MIT
#
CONTROL_GROUPS=`cargo test -- --list 2>/dev/null | egrep 'test$' | egrep -v '^src' | cut -d':' -f1`