The udev-rules must be modified to make it site-specific. The site
specific part of the rule should be executed only on the sites where the
devices belong. Change the generic udev rule to support the site-specific
udev-rule.
During boot, the current site-id is read from the LOADPARM which is
located at /sys/firmware/ipl. The zdev_id command reads this LOADPARM
and prints the SITE_ID based on the value. This SITE_ID is the
current site where the udev-rule is executing. Based on this assumption,
the rule is divided in to multiple blocks where each site block is executed
only for the corresponding sites.
The rule executes based on the value read from the LOADPARM.
1. When the LOADPARM specifies a site-id and if the current device has
an associated configuration settings, it will be used.
2. When the LOADPARM specifies a site, and the current device does not
have any associated configuration set, the udev rule uses the
fallback configuration settings.
3. If the device does not have a fallback-configuration settings or any
site-specific settings, no configurations will be used.
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The currently active site can be found from the loadparm attribute which
is located at /sys/firmware/ipl/loadparm. Read the value of loadparm and
extract the site_id.
The loadparm is an 8 byte alphanumerical IPL parameter. The content of
loadparm can be of different forms to indicate different boot menu
options, which are used by Linux for selecting the respective boot menu
entries. To indicate the site-id, loadparm adds a 'S' character. Where,
the following character after 'S' indicates either the site-id or
indivcation to derive the site-id from SSID.
For example,
If the loadparm value is Sn,'n' is the integer which could be one of the
valid site_ids from 0 to 9; i.e when booted on site-1, loadparm value will
be S1.
If loadparm value is "SS", zdev_id extracts the site_id from the SSID of
the current ipl device. For ccw and zfcp devices, the current ipl device-id
can be found at /sys/firmware/ipl/device.
In any case, if the loadparm contains invalid value, or empty, the
site_id will be default to the common-site.
After the first invocation of the zdev_id from udev-rule, the site-id
information is stored in /run/zdev_id.env file. This will reduce the
number of further invocation of zdev_id from within udev-rules.
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Currently zdev supports the configuration and maintenance of one single
persistent device state. With the introduction of sites, zdev should be
able to store, configure and maintaine NUM_SITES number of states in
the persistent configuration. This patch modifies the struct device and
adds a new site_specific device_state array, which stores the
persistent configuration for all the available sites.
Total number of available sites are configured with NUM_SITES. Where,
10 site-specific configurations and 1 common configuration.
1. during read, all site data is always read and update the
site_specific[] array.
2. dev->persistent is a copy of dev->site_specific[global_site_id]
3. modifications are only done on dev->persistent, and during write
dev->persistent overlays dev->site_specific[global_site_id]
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
In some high-availability setups, root and boot disks of a Linux
installation are copied to remote sites that can take over execution
of a Linux workload in case of an outage of the original site.
A site in this context is an execution environment such as an LPAR
or z/VM guest.
Each site may provide a different set of devices, or require different
parameters to be applied per device. chzdev supports site-specific
device configuration for up to 10 sites. Each site is identified by
a number in the range 0 to 9. You can provide a separate set of persistent
device configuration for each site. The IPL Load parameter value specified
during boot determines the currently active site. Only the device
configuration of the active site is applied during boot and when new
devices become available. A common configuration can be provided that is
applied when no site-specific configuration is available for a device in
the active site.
User can use the --site parameter to configure devices for a specific
site only. Configuration actions without a site parameter apply to the
common configuration.
During the boot, the active site configuration is chosed based on the
IPL LOADPARM variable. If the LOADPARM contains no valid site-id (0 to 9),
the fallback-site id will be used. The fallback-site-id can also be
configured with the zdev tools by not specifying the --site parameter.
Some of the new chzdev/lszdev commands with site-parameter can be as
follows
1. chzdev -ep f001 --site 3
This command is used to configure the device f001 for the site 3. During
boot, if the LOADPARM value is S3, this device configuration will be
used.
2. lszdev --site 3
This command lists all the devices which are configured for site 3.
3. chzdev -ep f001
Above command does not have a --site parameter. This device settings
will be used for the following conditions
1. When the loadparm specifies a site, (e.g S3), and the device does not
have an associated config-set, zdev uses fallback config set.
2. When the loadparm does not specify any site information, then the
fallback config set will be used.
When the loadparm specifies a site, and the device does not have an
associated config set, nor a fallback site, nothing is configured for this
device.
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
While at it, add a list with words to ignore. This suppresses the following codespell finding:
$ codespell libcpumf/libcpumf_cpuset.c
libcpumf/libcpumf_cpuset.c:15: parm ==> param, pram, parma
libcpumf/libcpumf_cpuset.c:17: parm ==> param, pram, parma
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Provie a CheckPatch configuration file and mention it in the coding style
guidelines (this feature was introduced with CheckPatch version 0.32 ~11 years
ago).
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add a .clang-format configuration and mention it in the coding guidelines.
Usage example:
$ clang-format -i **/*.[ch]
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When configuring a vfio-ap device, specifying only control domains
without any usage domains results in a useless configuration in that
both adapters and usage domains are required to make cryptographic
resources available to the guest. Without that, the control domains
are useless. While the kernel allows this, let's reject this
combination with a message.
Suggested-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This patch enhances the zipl stage3 loader image adding a trailer as
required for secure boot by future firmware versions.
Note: with the change in this patch the padding via objcopy command line
options is replaced by padding via linker script directives with the
same effect.
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
lszcrypt allows the explicit definition of a list of devices. Continue
the parameter processing if multiple devices are specified.
Signed-off-by: Holger Dengler <dengler@linux.ibm.com>
Reviewed-By: Harald Freudenberger <freude@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
In an upcoming patch the refactored function `pt_load_add` is reused. While at
it, add a wrap-around check to `pt_load_add`, replace `g.opts.device` access
with `fh->path` (this makes this function easier testable) and improve the
documentation.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Remove `static` attribute from `print_usage_exit` and declare it in the header
file `opts.h`. The function will be reused in a later patch.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Caller of `dfi_vmcoreinfo_get` must not modify the returned value, therefore
return a `const char *` value for `dfi_vmcoreinfo_get`.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add guarded storage registers support for the ELF input and the ELF output
format. See `man 2 s390_guarded_storage` for details about guarded storage.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The maximum size of all supported ELF note entries for a CPU can be 0x4a4 bytes
and not 0x4a0. Use a function for the calculation so it's easier to maintain.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Currently, only the ELF version `1` is defined by the ELF specification. See
`man 5 elf` for details. Therefore of course, we do only support this version.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The functions are refactored so the `struct zg_fh` value is now passed as
argument to the `nt_*` functions instead of accessing the global variable `g.fh`
from within of the functions. This makes the functions reusable and easier to
test. While at it, improve the documentation of `nt_read`.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Introduce variable `phdrs` and shrink the scope of `phdr`. This removes
duplicated code `&phdr[i]` and makes the code easier to read. While at it,
constify the parameter of `pt_notes_add`.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
`ehdr->e_phoff` has the type `Elf64_Off` and this maps to `uint64_t`, but `off_t` is signed. Let's
add a check for this.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
generalize the collection step name from "docker" to "container host"
initial version of adding data collection for OCP.
all log file content might be too much for most installations
Reviewed-by: Alexander Klein <alkl@de.ibm.com>
Reviewed-by: Mario Held <mario.held@de.ibm.com>
Reviewed-by: Mike Storzer <mstorzer@de.ibm.com>
Signed-off-by: Joern Siglen <siglen@de.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
GUINT32_TO_BE and GUINT32_FROM_BE do exactly the same in case of glib2, just the
naming differs. However, at this point the conversion is from be to system
endianness therefore `GUINT32_FROM_BE` is the proper macro.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
It's often useful to know how much data was actually copied, therefore let's
introduce an nullable parameter `@copied` to `pv_gbytes_memcpy`.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
...by adding a compile time assertion about the buffer size. While at it, report
the actual user data size in case of an error.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The endianness handling was only partially implemented for uvio.c. Let's remove
it since doesn't make much sense here, because the IOCTL must be build on the
same system where they're actually executed.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
the buffer sizing did not work any longer
add explaining comments and rework the variable names for claritiy
refering commit f7cc14753cc9d930af804e2ac3ce108a12145db6
Reported-by: Mike Storzer <MSTORZER@de.ibm.com>
Reviewed-by: Mario Held <mario.held@de.ibm.com>
Signed-off-by: Joern Siglen <siglen@de.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Use `PV_CERT_ERROR` error domain whenever useful. Currently, it's only used
internally in libpv and therefore there is no problem in changing it.
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
There should be no problem with this change since the changed value was never
used. Therefore, let's remove the assignment and declare the parameter as const
to prevent such a behavior in the future.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The machine types are not strictly increasing anymore since Z16, therefore,
we cannot use numerical comparison to find out the correct HSA size
of a machine.
Fixes: 2515832469 ("util_arch: Add IBM z16 as known machine")
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
`afl-clang` requires a white space between a option name and the option value.
Therefore let's fix this for `-xc` by adding a white space.
How to reproduce:
$ make -C zdump CC=afl-clang-fast LINK=afl-clang-fast
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
some versions of /bin/dash do interpreted our commands different, which
causes a variable failure and termination of the script.
e.g. a new line char "\n" will split the cmd_type to a multiple line variable
not working with the followup commands. adding the head command does avoid
multiline variables now.
Problem found in Ubuntu 20.4 and some older Ubuntu versions
Reviewed-by: Mario Held <mario.held@de.ibm.com>
Signed-off-by: Joern Siglen <siglen@de.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
