The existing code can already determine the required devices for the
actively running system. In case users would have a need to configure
additional devices for early boot in kdump initrd, import those devices
marked with "zdev:early=1" for completeness. Assuming this is a small
number of devices and could even overlap between the root-fs and the kdump
case, do not further distinguish those two sets of devices.
Github-ID: https://github.com/ibm-s390-linux/s390-tools/pull/158
Reviewed-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Acked-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Acked-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Additional debugging output can be generated with e.g. dracut option
"--stdlog 5" (or short -L5). It shows the chzdev export result, the output
of chzdev imports, and an overview of the resulting persistent config
within the initrd.
On systems, which default to using dracut option "--quiet", you might need
an additional "--verbose" to counter "--quiet" so -L5 has effect.
Typically combined with "--debug" to get a shell trace from building an
initrd (Note: --debug does not increase the log levels).
Github-ID: https://github.com/ibm-s390-linux/s390-tools/pull/158
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Acked-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Acked-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Otherwise, dracut skips modules that are "not hostonly", i.e. not used /
loaded on the system when running dracut to build an initrd.
Without the fix, zdev device auto configuration only works for device
types for which a device driver happened to be loaded when building an
initrd. Likewise for specifying rd.* boot options.
Hostonly mode is often the default in Linux distributions.
Make zdev device auto configuration work nonetheless without users
having to know about and tweak dracut modes for initrd generation.
On the typical sloppy hostonly mode, the code disables hostonly for
installing kernel modules. This covers building regular initrds, where zdev
device auto configuration should even work for device types, for which
device drivers were not loaded when an initrd was generated. This can
happen when new devices of new types are configured for a DPM logical
partition. Also, users could want to start using dracut cmdline options
rd.{dasd,zfcp,znet} for a device type that was not used when the initrd was
built.
The special strict hostonly mode is used by some kdump implementations. In
that case, hostonly remains in effect intentionally because only support
for the really required devices as determined by dracut module 95zdev-kdump
should be included in a kdump initrd due to the memory-constrained kdump
environment. Cf. commit 73c46a3056 ("zdev/dracut: fix kdump by only
activating required devices"), which also provides more references on
strict hostonly mode. Even for non-kdump cases, let strict hostonly mode be
effective here in 95zdev in case some future use case appears for this
special mode beyond kdump.
Github-ID: https://github.com/ibm-s390-linux/s390-tools/pull/158
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Acked-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Acked-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The sscanf() format specifiers for signed and unsigned int mistakenly
used "%d"/"%u" prefix analogous to "%l" for long but those do not exist.
Fixes: 37348ef662 ("libutil: add util_file_read_i()/util_file_read_ui()")
Acked-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
The crate dependencies were a bit to slack. Due to the rust dependency
resolver's strategy of always selecting the latest version this never
lead to any issues.
This has no impact on the workspaces Cargo.lock
Reviewed-by: Marc Hartmayer <marc@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
With the last patch introducing the rust workspace the location of
Cargo.lock has changed. Therefore, remove all crate level lock-files and
add rust/Cargo.lock as the only lock-file.
Steps to reproduce:
```
cd rust
mv pvsecret/Cargo.lock .
cargo build
cargo update -p openssl
cargo update -p curl-sys
cargo update -p rustix
```
While at it update some dependencies to get fixes for security issues.
Reviewed-by: Marc Hartmayer <marc@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
A workspaces simplifies the build and packaging process significantly.
All build artifacts and binaries are now built in a single location
(e.g., rust/target/release/*), and a unified dependency resolution is
used. Hence one Cargo.lock for all crates at rust/Cargo.lock.
Closes: https://github.com/ibm-s390-linux/s390-tools/issues/156
Reviewed-by: Marc Hartmayer <marc@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Since mdevctl commit acf78c1ff6c9 it is now possible for the
get-attributes event to occur between a pre-define and post-define.
This is done in order to obtain the active attributes for the device
before writing them to the config file, and implies that the
get-attributes cannot re-obtain the file lock. For other cases
where mdevctl calls get-attributes, the file lock is not already
held and must be obtained by ap-check before reading attributes from
active devices.
To solve this, let's use the knowledge that mdevctl is a single-threaded
tool and add a test to detect this scenario. If the file lock is
already held by the parent during a get-attributes, don't attempt to
re-acquire it.
Reported-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Add a check to the stage3b that the kernel cmdline is always
null-terminated. While at it, ensure the coding style is consistent.
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Make sure that the kernel command line used for the Secure Execution
boot image is always null-terminated. Before this change, users had to
ensure that the provided kernel cmdline was null-terminated, which was
error-prone. But since the default s390x Linux kernel command line is
set to `root=/dev/ram0 ro` the remaining reserved memory for the kernel
command line is zeroed out. Therefore, the problem only shows up if the
used kernel command line is shorter than the default kernel command
line.
Fixes: 65b9fc442c ("genprotimg: introduce new tool for the creation of PV images")
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Fix a valgrind finding. Fix an invalid read/write of one byte after the
actual struct to clear. Not fixing this may result in a illegal write or
memory corruption of the program. Fortunately, for the actual only user,
pvsecret this is not the case.
Fixes: c6f621d0 ("rust: Add library for pv tools")
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Pai collects data from per CPU ring buffers and stores them in the
memory mapped output file. When data is collected from many CPUs at
the same time, writing data to output file can be slow.
Improve this and allow the pai recording to run with higher
real time priority. This is the same approach as done by the perf tool.
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
in customer situation we found lsqeth listing devices like:
Device name : (unnamed net_device)
Device name : enc2000
Device name : enc3000
- the braket around the "unnamed" device is braking the function call and
leads to stop the dbginfo.sh script.
- this patch removes brakets > the functions call works and call of osaoat
will report an unknown device instead of braking the dbginfo.sh script
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>
Make the description of the --keep_volser option more generic and avoid
mentioning specific tooling.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The zipl bootloaders are s390x specific, so only build them if the
`HOST_ARCH' is set to s390x.
While at it, rename `INC_FILES` to `EMBEDDED_BOOTLOADERS`. Also
introduce `EXTERNAL_BOOTLOADERS` variable and use it in the `install`
Makefile target.
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When checking if a card or an APQN is online, not only check the 'online'
sysfs attribute, but also check the 'config' and 'chkstop' attribute.
Cards and APQNs in check-stopped or deconfigured state can still be reported
as online via the sysfs attribute, although they are not available to be
used for zkey.
In case the 2 additional sysfs attributes are not available in sysfs, then
don't fail, but rely on the 'online' attribute only.
Signed-off-by: Ingo Franzki <ifranzki@linux.ibm.com>
Reviewed-by: Harald Freudenberger <freude@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Move the code responsible for installing stage3.bin to the
boot/Makefile. In addition, remove the stage3.bin from the Makefile
`all` target prerequisites in src/Makefile, as zipl can be built without
it. While at it, use $(INSTALL) instead of $(CP) for the bootloader
installation.
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Linux stable kernels can reach greater than 256 sublevels [1],
which can cause the cpi tooling to generate an invalid string
that gets passed to the firmware and causes unusual responses:
$ uname -r
5.4.255
$ cat /sys/firmware/cpi/system_level
0x04260000000504ff
--reboot--
$ uname -r
5.4.256
$ cat /sys/firmware/cpi/system_level
0x4260000000504100
The first sublevel field is defined as one byte, so ensure that
a value larger than that isn't included.
[1] https://lore.kernel.org/lkml/1612534196241236@kroah.com/
Signed-off-by: Eric Farman <farman@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Currently ZDEV_SITE_ID is derived with the help of an additional
udev-rule, 40-zdev-id.rules. The sole purpose of this rule is to
determine the ZDEV_SITE_ID environment value with the help of zdev_id
binary. This solution is minimal, but this has some unwanted side-
effects. The zdev_id logic get executed for all the events, even
those completely unrelated to zdev/or site, and imports the unneeded
envionment values to the udev-db.
Instead of having an additional rule file, add this logic as part of
the udev-rule of those devices which are configured with site-support.
The logic will then be available on all those rules with the
site-supported devices only.
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Introduce dev_site_configured macro,which can be used to find the
availability of site configurations for the device during udev rule
creation.
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Previously SITE_FALLBACK and other site-specific configuration support
macros were defined in device.h. Instead, move them to a relatively
smaller header file which is exclusive for site-related definitions.
This way, light-weight zdev_id also can use the same header file.
Reported-by: Steffen Maier<maier@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Behaviour of the qeth performance_stats sysfs attribute has changed
with kernel commit
b0abc4f5df76 ("s390/qeth: overhaul ethtool statistics")
that went into kernel v5.1.
Before the kernel commit
- collection of statistics was turned on and off by writing 1 or 0
- default after device activation was 0
- statistics were reset by writing 0
After the kernel commit:
- collection is always on
- attribute always reads 1
- statistics is reset by writing 1; writing 0 is a no-op
Problems of chzdev on new kernels:
chzdev cannot reset statistics ('performance_stats=1' does nothing).
'chzdev --export' always lists performance_stats.
'chzdev qeth --help-attribute performance_stats' reflects old behaviour.
This patch will do the following:
'chzdev qeth --help-attribute performance_stats' reflects new behaviour.
'chzdev --export' does not list performance_stats on new kernels.
'chzdev performance_stats=1' resets statistics on new kernels.
'chzdev performance_stats=0' still resets statistics on old kernels,
does nothing on new kernels.
Suggested-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
With many CPUs, such as triple digit counts, the by default many empty
sysfs directories are prohibitive to collect, especially if the number
of SCSI disk devices is also large, such as 4-digit counts.
Excluding them from being collected from sysfs saves hundreds times
thousands of archive entries and inodes on expansion.
Since the number device-mapper devices (multipath and other target types
such as LVM) is smaller and can include devices not backed by zfcp, keep
collecting
/sys/devices/virtual/block/dm-[0-9]*/mq/0/cpu[0-9]*/
Definitely keep collecting
/sys/kernel/debug/block/{sd,dm-}*/hctx0/cpu[0-9]*/
as it contains actual statistics files:
completed default_rq_list dispatched merged poll_rq_list read_rq_list
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
