Currently there are two issues with unit-testing the functions defined
in kdumpctl and other shell scripts after sourcing them,
- kdumpctl would call main which requires root permission and would
create single instance lock (/var/lock/kdump)
- kdumpctl and other shell scripts directly source files under /usr/lib/kdump/
When ShellSpec load a script via "Include", it defines the__SOURCED__
variable. By making use of __SOURCED__, we can
1. let kdumpctl not call main when kdumpctl is "Include"d by ShellSpec
2. instruct kdumpctl and kdump-lib.sh to source the files in the repo
when running ShelSpec tests
Note coverage/ is added to .gitignore because ShellSpec generates code
coverage results in this folder.
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Make use of the new ${OPT[]} array and simplify local_fs_dump_target to
remove one more file operations.
While at it rename the local_fs_dump_target to is_local_target
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
With the introduction of ${OPT[fstype]} this call to kdump_get_conf_val
can be removed now as well.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
The variable is only used for ssh dump targets. Furthermore it is
identical to the value stored in ${OPT[_target]}. Thus drop DUMP_TARGET and
use ${OPT[_target]} instead.
In order to be able to distinguish between the different target types
introduce the internal ${OPT[_fstype]}.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
The variable is only used for ssh dump targets. Furthermore it is
identical to the value stored in ${OPT[sshkey]}. Thus drop
SSH_KEY_LOCATION and use ${OPT[sshkey]} instead.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
The variable is only used for ssh dump targets. Furthermore it is
identical to the value stored in ${OPT[path]}. Thus drop SAVE_PATH and
use ${OPT[path]} instead.
Also make sure that ${OPT[path]} is always set to the default value when
no entry in kdump.conf is found.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
Every call to kdump_get_conf_val parses kdump.conf although the file has
already been parsed in check_config. Thus store the values parsed in
check_config in an array and use them later instead of re-parsing the
file over and over again.
While at it rename check_config to parse_config.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
check_config and check_ssh_config both parse /etc/kdump.conf and are
usually used together. The difference between both is that
check_ssh_config does some extra checks on the format of the provided
ssh destination but ignores invalid or deprecated options in the config.
Thus merge check_ssh_config into check_config. Leave the additional
checks on the ssh destination in check_ssh_config but treat it like the
checks done for e.g. the failure_action.
This slightly changes the behavior of 'kdumpctl propagate', which now
fails if kdump.conf contains an invalid value unrelated to ssh. This
change in behavior isn't problematic because 'kdumpctl propagate' always
needs to be followed by a 'kdumpctl start' to have a working kdump
environment. For the situations where 'propagate' fails now the 'start'
would have failed in the past. So the failure only moved one step ahead
in the sequence.
While at it drop check_ssh_target and call check_and_wait_network_ready
directly.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
The function has multiple problems:
1) SSH_{USER,SERVER} aren't defined local
2) Weird use of cut and sed to parse the DUMP_TARGET for the user and
host although check_ssh_config guarantees that it has the format
<user>@<host>.
3) Unnecessary use of a variable for the return value
4) Weird behavior to first unpack the DUMP_TARGET to SSH_USER and
SSH_SERVER and then putting it back together again
5) Definition of variable errmsg that is only used once but breaks
grep-ability of error message.
6) Wrong order when redirecting output of ssh-keygen, see SC2069 [1]
Fix them now.
While at it also improve the error messages in the function.
[1] https://www.shellcheck.net/wiki/SC2069
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
For ssh targets kdumpctl only verifies that the config value has the
correct <user>@<host> format itself. For all other tests, e.g. if the
destination can be reached, it relies on ssh. This allows users to
provide a <host> that isn't the proper hostname but an alias defined in
the ssh_config without failing the tests. If this is done
dracut-module-setup.sh:kdump_get_remote_ip will fail to obtain the
targets ip address. This failure is not detected and thus will not fail
the initramfs creation. The resulting initramfs however doesn't have the
necessary information for setting up the network and thus will fail to
boot.
Prevent the use of alias hostnames by verifying that the given hostname
is the same one ssh would use after parsing the ssh_config.
Note: Don't use getent ahosts to verify that the given host can be
resolved as this requires the network to be up which cannot be
guaranteed when the kdump.conf is parsed.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
The time out was increased to 180 seconds in 680c0d3 ("kdumpctl:
distinguish the failed reason of ssh"). Update the comment to reflect
that change.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
There are currently three identical definitions for the default ssh key.
Combine them into one in kdump-lib-initramfs.sh.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
in prepare_kdump_bootinfo s/defaut/default/.
While at it declare it with the other local variables as local.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
Using syslog for StandardOutput in a service file was deprecated in
systemd v246 with commit f3dc6af20f ("core: automatically update
StandardOuput=syslog to =journal (and similar for StandardError=)").
Thus the following warnings are printed in the crash kernel when
creating a dump.
systemd[1]: /usr/lib/systemd/system/kdump-capture.service:23: Standard output type syslog is obsolete, automatically updating to journal. Please update your unit file, and consider removing the setting altogether.
systemd[1]: /usr/lib/systemd/system/kdump-capture.service:24: Standard output type syslog+console is obsolete, automatically updating to journal+console. Please update your unit file, and consider removing the setting altogether.
Fix this by redirecting the stdout and stderr to the journal.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Coiby Xu <coxu@redhat.com>
do_estimate prints the warning that the reserved crashkernel is lower
than the recommended one even then when both values are identical. This
might cause confusion. So omit printing the warning when both values are
equal.
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
There is a system-wide sync call at the end of mkdumprd, move it to
kdumpctl after rebuild initrd and add another one for mkfadumprd.
Sync only the $TARGET_INITRD to avoid a system-wide sync taking too
long on a system with high disk activity.
Also update the sync in kdumpctl:restore_default_initrd which will
mv the $DEFAULT_INITRD_BAK to $DEFAULT_INITRD.
Signed-off-by: Lichen Liu <lichliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
If GRUB_ETC_DEFAULT use crashkernel=auto or
crashkernel=OLD_DEFAULT_CRASHKERNEL, it should be updated as well.
Add a helper function to read kernel cmdline parameter from
GRUB_ETC_DEFAULT. This function is used to read kernel cmdline
parameter like fadump or crashkernel.
Reviewed-by: Philipp Rudo <prudo@redhat.com>
There is the case where there are multiple entries of the same parameter on
the command line, e.g.
GRUB_CMDLINE_LINUX="crashkernel=110M crashkernel=220M fadump=on crashkernel=330M".
In such an situation _update_kernel_cmdline_in_grub_etc_default only
updates/removes the last entry which is usually not what you want as the
kernel (for crashkernel) takes the last entry it can find.
Thus make sure the case with multiple entries of the same parameter is
handled properly by removing all occurrences of given parameter first.
Note
1. sed command group and conditional control has been used to get rid of
grep.
2. Fully supporting kernel cmdline as documented in
Documentation/admin-guide/kernel-parameters.rst is complex and in
foreseeable future a full implementation is not needed. So simply
document the unsupported cases instead.
Fixes: 140da74 ("rewrite reset_crashkernel to support fadump and to used by RPM scriptlet")
Reported-by: Philipp Rudo <prudo@redhat.com>
Suggested-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
When doing in-place upgrading using leapp on x86_64, kdumpcl can't
acquire instance lock when running in %post RPM scriplet on x86_64,
localhost upgrade[1306]: /bin/kdumpctl: line 49: /var/lock/kdump: No such file or directory
localhost upgrade[1306]: kdump: Create file lock failed
and running "touch /var/lock/dkump" also fails with
"No such file or directory". Thus kdumpctl can't be run in %post
scriptlet. But kdumpctl can be run in %posttrans RPM scriplet.
Besides, it's better to update crashkernel after the kernel has been
updated. So let's update kernel crashkernel in the %posttrans
scriptlet which will be run in the end of a transaction i.e. after
the kernel has been updated.
Note for %posttrans scriptlet, "$1 == 1" means both installing a new
package and upgrading a package.
[1] https://github.com/apptainer/singularity/issues/2386#issuecomment-474747054
Reported-by: Jie Li <jieli@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Previously the output of blkid is not checked. If the output
is empty, the eval will report the following error message:
/lib/kdump/kdump-lib.sh: eval: line 925: syntax error near unexpected token `;'
/lib/kdump/kdump-lib.sh: eval: line 925: `; echo $TYPE'
For example, we can observe such a failing when blkid is invoked
against a lvm thinpool block device:
$ blkid -u filesystem,crypto -o export -- "/dev/block/253\:2"
$ echo $?
2
$ udevadm info /dev/block/253\:2|grep S\:
S: mapper/vg00-thinpoll_tmeta
In this patch, we will use sed instead of eval, to output the
fstype of block device if any.
Signed-off-by: Tao Liu <ltao@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
It's found that the kernel cmdline crashkernel=auto doesn't get updated
when upgrading kexec-tools. This happens because _get_all_kernels_from_grubby
is called with no argument by reset_crashkernel_after_update. When retrieving
all kernel paths on the system, "grubby --info ALL" should be used. Fix this
error by passing "ALL" argument.
Fixes: 0adb0f4 ("try to reset kernel crashkernel when kexec-tools updates the default crashkernel value")
Reported-by: Jie Li <jieli@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Tao Liu <ltao@redhat.com>
_is_osbuild fails because it expects the 1st and 2nd function parameter
to be the environment variable and environ file path respectively. Fix
it by swapping the parameters in read_proc_environ_var.
Note the osbuild environ file path is defined in _OSBUILD_ENVIRON_PATH
so _is_osbuild can be unit-tested by overwriting _OSBUILD_ENVIRON_PATH.
Fixes: 6a3ce83 ("fix the error of parsing the container environ variable for osbuild")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Pingfan Liu <piliu@redhat.com>
shellcheck finds the following problem,
$ shellcheck kdump-lib.sh
In kdump-lib.sh line 876:
get_recommend_size "$sys_mem" "$ck_cmdline"
^---------^ SC2154: ck_cmdline is referenced but not assigned (did you mean '_ck_cmdline'?).
s/ck_cmdline/_ck_cmdline to fix kdump_get_arch_recommend_size.
Note s/sys_mem/_sys_mem as well to make the changes consistent.
Fixes: 105c016 ("factor out kdump_get_arch_recommend_crashkernel")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Acked-by: Tao Liu <ltao@redhat.com>
This patch makes the default crashkernel value consistent with previous
one.
Fixes: 105c016 ("factor out kdump_get_arch_recommend_crashkernel")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
The environment variable entries in /proc/[pid]/environ are separated by
null bytes instead of by spaces. Update the sed regex to fix this issue.
Note that,
1. this patch also fixes a issue which is kdumpctl would try to reset
crashkernel even osbuild has provided custom crashkernel value.
2. kernel hook 92-crashkernel.install installed by kexec-tools is
guaranteed to be ran by kernel-install. kexec-tools doesn't recommend
kernel so there is no guarantee kernel is installed after kexec-tools.
But dnf invokes kernel-install in the posttrans scriptlet (of kernel-core)
which is always ran after all packages including kexec-tools and kernel
in a dnf transaction.
3. To be able to do unit tests, the logic of reading environment variable
has been extracted as a separate function.
Fixes: ddd428a ("set up kernel crashkernel for osbuild in kernel hook")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Resolves: bz2025860
Upstream: git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git
commit 186e7b0752d8fce1618fa37519671c834c46340e
Author: Alexander Egorenkov <egorenar@linux.ibm.com>
Date: Wed Dec 15 18:48:53 2021 +0100
s390: handle R_390_PLT32DBL reloc entries in machine_apply_elf_rel()
Starting with gcc 11.3, the C compiler will generate PLT-relative function
calls even if they are local and do not require it. Later on during linking,
the linker will replace all PLT-relative calls to local functions with
PC-relative ones. Unfortunately, the purgatory code of kexec/kdump is
not being linked as a regular executable or shared library would have been,
and therefore, all PLT-relative addresses remain in the generated purgatory
object code unresolved. This in turn lets kexec-tools fail with
"Unknown rela relocation: 0x14 0x73c0901c" for such relocation types.
Furthermore, the clang C compiler has always behaved like described above
and this commit should fix the purgatory code built with the latter.
Because the purgatory code is no regular executable or shared library,
contains only calls to local functions and has no PLT, all R_390_PLT32DBL
relocation entries can be resolved just like a R_390_PC32DBL one.
* https://refspecs.linuxfoundation.org/ELF/zSeries/lzsabi0_zSeries/x1633.html#AEN1699
Relocation entries of purgatory code generated with gcc 11.3
------------------------------------------------------------
$ readelf -r purgatory/purgatory.o
Relocation section '.rela.text' at offset 0x6e8 contains 27 entries:
Offset Info Type Sym. Value Sym. Name + Addend
00000000000c 000300000013 R_390_PC32DBL 0000000000000000 .data + 2
00000000001a 001000000014 R_390_PLT32DBL 0000000000000000 sha256_starts + 2
000000000030 001100000014 R_390_PLT32DBL 0000000000000000 sha256_update + 2
000000000046 001200000014 R_390_PLT32DBL 0000000000000000 sha256_finish + 2
000000000050 000300000013 R_390_PC32DBL 0000000000000000 .data + 102
00000000005a 001300000014 R_390_PLT32DBL 0000000000000000 memcmp + 2
...
000000000118 001600000014 R_390_PLT32DBL 0000000000000000 setup_arch + 2
00000000011e 000300000013 R_390_PC32DBL 0000000000000000 .data + 2
00000000012c 000f00000014 R_390_PLT32DBL 0000000000000000 verify_sha256_digest + 2
000000000142 001700000014 R_390_PLT32DBL 0000000000000000
post_verification[...] + 2
Relocation entries of purgatory code generated with gcc 11.2
------------------------------------------------------------
$ readelf -r purgatory/purgatory.o
Relocation section '.rela.text' at offset 0x6e8 contains 27 entries:
Offset Info Type Sym. Value Sym. Name + Addend
00000000000e 000300000013 R_390_PC32DBL 0000000000000000 .data + 2
00000000001c 001000000013 R_390_PC32DBL 0000000000000000 sha256_starts + 2
000000000036 001100000013 R_390_PC32DBL 0000000000000000 sha256_update + 2
000000000048 001200000013 R_390_PC32DBL 0000000000000000 sha256_finish + 2
000000000052 000300000013 R_390_PC32DBL 0000000000000000 .data + 102
00000000005c 001300000013 R_390_PC32DBL 0000000000000000 memcmp + 2
...
00000000011a 001600000013 R_390_PC32DBL 0000000000000000 setup_arch + 2
000000000120 000300000013 R_390_PC32DBL 0000000000000000 .data + 122
000000000130 000f00000013 R_390_PC32DBL 0000000000000000 verify_sha256_digest + 2
000000000146 001700000013 R_390_PC32DBL 0000000000000000 post_verification[...] + 2
Corresponding s390 kernel discussion:
* https://lore.kernel.org/linux-s390/20211208105801.188140-1-egorenar@linux.ibm.com/T/#u
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reported-by: Tao Liu <ltao@redhat.com>
Suggested-by: Philipp Rudo <prudo@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
[hca@linux.ibm.com: changed commit message as requested by Philipp Rudo]
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
v2:
- Moved patch 601 -> 401
Signed-off-by: Philipp Rudo <prudo@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
There is a mechanism to keep memory consumption minimum, i.e. equal
to trace_buf_size=1, until tracing by ftrace is actually started:
tracing: keep ring buffer to minimum size till used
73c5162aa3
Since ftrace is usually never used in the kdump 2nd kernel, the kdump
2nd kernel behaves in the same way with or without trace_buf_size=1.
So the issue which the patch want to solve never exists. Let's revert
the patch for better maintainance and avoid confusion.
ref link: https://bugzilla.redhat.com/show_bug.cgi?id=2034501#c20
This reverts commit f39000f.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
fadump is a ppc only feature, mkfadumprd is only needed for fadump, drop
it for other arch.
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Kairui Song <kasong@tencent.com>
This patch will introduce early kdump test.
It reuses the code of nfs kdump test, in order to setup 2 seperated VMs,
one(the client) for trigger the early kdump and crash, the other(the server)
for saving vmcore and check if vmcore exists. In order to minimize the
repetted code, a soft link is made to copy the same server side code.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
When run tests with 2 VMs, for example nfs/ssh kdump tests, client VM will do the
crash and dump, server VM will do vmcore saving and if-vmcore-exists
check.
Previously, when client VM finishes running, run-test.sh will kill the lead background
process, and then check if server VM has outputted "TEST PASSED" or "TEST FAILED" string.
However it didn't wait for server VM to finish. As a result, the server VM's final
outputs are not collected and checked, leaving the test result as "TEST RESULT NOT FOUND"
sometimes.
For example, the following is the pstree status of $(jobs -p) before it
gets killed. We can see the server VM is still running:
run-test.sh,172455 /root/kexec-tools/tests/scripts/run-test.sh --console nfs-early-kdump
└─run-test.sh,172457 /root/kexec-tools/tests/scripts/run-test.sh --console...
└─timeout,172480 --foreground 10m /root/kexec-tools/tests/scripts/run-qemu...
└─qemu-system-x86,172481 -enable-kvm -cpu host -nodefaults...
├─{qemu-system-x86},172489
├─{qemu-system-x86},172492
├─{qemu-system-x86},172493
├─{qemu-system-x86},172628
└─{qemu-system-x86},172629
In this patch, we will wait for $(jobs -p) to finish, in order to get
the complete output of test results.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
kexec-tools runs hostname binary in the case of fence_kdump. Since this
is a trival dependency and should not block the kexec-tools installation
if non-existent, using weak-dependency to resolve it.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Tao Liu <ltao@redhat.com>
Since kdump-lib-initramfs.sh is included by kdump-lib.sh, and
FENCE_KDUMP_SEND is used by both 1st and 2nd kernel, moving
FENCE_KDUMP_SEND from kdump-lib.sh to kdump-lib-initramfs.sh.
Signed-off-by: Pingfan Liu <piliu@redhat.com>
Acked-by: Tao Liu <ltao@redhat.com>
This patch will make zstd as recommended instead of required for
kexec-tools. If zstd command/package is unavaliable, it can failback to invoke
gzip when making kdump initramfs.
Fixes: 0311f6e ("Set zstd as the default compression method for kdump initrd")
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
kdumpctl only accepts a specified set of options. Add
auto_reset_crashkernel to this set.
Fixes: 73ced7f ("introduce the auto_reset_crashkernel option to kdump.conf")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Acked-by: Tao Liu <ltao@redhat.com>
When a file doesn't exist or isn't readable, grep complains as follows,
grep: /proc/cmdline: No such file or directory
grep: /etc/kernel/cmdline: No such file or directory
/proc/cmdline doesn't exist when installing package for an OS image and
/etc/kernel/cmdline may not exist if osbuild doesn't want set custom
kernel cmdline.
Use "-s" to suppress the error messages.
Fixes: 0adb0f4 ("try to reset kernel crashkernel when kexec-tools updates the default crashkernel value")
Fixes: ddd428a ("set up kernel crashkernel for osbuild in kernel hook")
Signed-off-by: Coiby Xu <coxu@redhat.com>
Acked-by: Tao Liu <ltao@redhat.com>
The '|' in 'failure_action|default' should be replaced with '\|' when
passed to kdump_get_conf_val function. Because '|' needs to be escaped
to mean OR operation in sed regex, otherwise it will consider
'failure_action|default' as a whole string.
Fixes: ab1ef78 ("kdump-lib.sh: use kdump_get_conf_val to read config values")
v1 -> v2:
Rephased the commit message.
Replaced " with '.
Signed-off-by: Tao Liu <ltao@redhat.com>
Acked-by: Coiby Xu <coxu@redhat.com>
zstd has better compression ratio and time consumption balance.
When no customized compression method specified in kdump.conf,
we will use zstd as the default compression method.
**The test method:
I installed kexec-tools with and without the patch, executing the following
command for 4 times, and calculate the averange time:
$ rm -f /boot/initramfs-*kdump.img && time kdumpctl rebuild && \
ls -ail /boot/initramfs-*kdump.img
**The test result:
Bare metal x86_64 machine:
dracut with squash module
zlib lzo xz lz4 zstd
real 10.6282 11.0398 11.395 8.6424 10.1676
user 9.8932 11.9072 14.2304 2.8286 8.6468
sys 3.523 3.4626 3.6028 3.5 3.4942
size of
kdump.img 30575616 31419392 27102208 36666368 29236224
dracut without squash module
zlib lzo xz lz4 zstd
real 9.509 19.4876 11.6724 9.0338 10.267
user 10.6028 14.516 17.8662 4.0476 9.0936
sys 2.942 2.9184 3.0662 2.9232 3.0662
size of
kdump.img 19247949 19958120 14505056 21112544 17007764
PowerVM hosted ppc64le VM:
dracut with squash module | dracut without sqaush module
zlib zstd | zlib zstd
real 10.6742 10.7572 | 9.7676 10.5722
user 18.754 19.8338 | 20.7932 13.179
sys 1.8358 1.864 | 1.637 1.663
|
size of |
kdump.img 36917248 35467264 | 21441323 19007108
**discussion
zstd has a better compression ratio and time consumption balance.
v1 -> v2:
Use kdump_get_conf_val() to get dracut_args values of kdump.conf
v2 -> v3:
Attached testing benchmark
v3 -> v4:
Re-measured and re-attached the testing benchmark of x86_64 and ppc64le.
Changed regex '.*[[:space:]]' to '(^|[[:space:]])'
v4 -> v5:
Attacked lzo/xz/lz4 testing benchmark.
v5 -> v6:
Add zstd as required in kexec-tools.spec
Hello Coiby, you may use "RELEASE=34 make test-run", for
CONFIG_RD_ZSTD is enabled since fc-cloud-34
Acked-by: Coiby Xu <coxu@redhat.com>
Signed-off-by: Tao Liu <ltao@redhat.com>
Update crashkernel-howto since crashkernel.default has been removed. The
documentation is also simplified as a result.
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
osbuild is a tool to build OS images. It uses bwrap to install packages
inside a sandbox/container. Since the kernel package recommends
kexec-tools which in turn recommends grubby, the installation order would
be grubby -> kexec-tools -> kernel. So we can use the kernel hook
92-crashkernel.install provided by kexec-tools to set up kernel
crashkernel for the target OS image. But in osbuild's case, there is no
current running kernel and running `uname -r` in the container/sandbox
actually returns the host kernel release. To set up kernel crashkernel for
the OS image built by osbuild, a different logic is needed.
We will check if kernel hook is running inside the osbuild container
then set up kernel crashkernel only if osbuild hasn't specified a
custome value. osbuild exposes [1] the container=bwrap-osbuild environment
variable. According to [2], the environment variable is not inherited down
the process tree, so we need to check /proc/1/environ to detect this
environment variable to tell if the kernel hook is running inside a
bwrap-osbuild container. After that we need to know if osbuild wants to use
custom crashkernel value. This is done by checking if /etc/kernel/cmdline
has crashkernel set [3]. /etc/kernel/cmdline is written before packages
are installed.
[1] https://github.com/osbuild/osbuild/pull/926
[2] https://systemd.io/CONTAINER_INTERFACE/
[3] https://bugzilla.redhat.com/show_bug.cgi?id=2024976#c5
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
When kexec-tools updates the default crashkernel value, it will try to
reset the existing installed kernels including the currently running
kernel. So the running kernel could have different kernel cmdline
parameters from /proc/cmdline. When installing a kernel after updating
kexec-tools, /usr/lib/kernel/install.d/20-grub.install would be called
by kernel-install [1] which would use /proc/cmdline to set up new kernel's
cmdline. To address this special case, reset the new kernel's crashkernel
and fadump value to the value that would be used by running kernel after
rebooting by the installation hook. One side effect of this commit is it
would reset the installed kernel's crashkernel even currently running kernel
don't use the default crashkernel value after rebooting. But I think this
side effect is a benefit for the user.
The implementation depends on kernel-install which run the scripts in
/usr/lib/kernel/install.d passing the following arguments,
add KERNEL-VERSION $BOOT/MACHINE-ID/KERNEL-VERSION/ KERNEL-IMAGE [INITRD-FILE ...]
An concrete example is given as follows,
add 5.11.12-300.fc34.x86_64 /boot/e986846f63134c7295458cf36300ba5b/5.11.12-300.fc34.x86_64 /lib/modules/5.11.12-300.fc34.x86_64/vmlinuz
kernel-install could be started by the kernel package's RPM scriplet [2].
As mentioned in previous commit "try to reset kernel crashkernel when
kexec-tools updates the default crashkernel value", kdumpctl has difficulty
running in RPM scriptlet fore CoreOS. But rpm-ostree ignores all kernel hooks,
there is no need to disable the kernel hook for CoreOS/Atomic/Silverblue. But a
collaboration between rpm-ostree and kexec-tools is needed [3] to take care
of this special case.
Note the crashkernel.default support is dropped.
[1] https://www.freedesktop.org/software/systemd/man/kernel-install.html
[2] https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel.spec#_2680
[3] https://github.com/coreos/rpm-ostree/issues/2894
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
kexec-tools could update the default crashkernel value.
When auto_reset_crashkernel=yes, reset kernel to new crashkernel
value in the following two cases,
- crashkernel=auto is found in the kernel cmdline
- the kernel crashkernel was previously set by kexec-tools i.e.
the kernel is using old default crashkernel value
To tell if the user is using a custom value for the kernel crashkernel
or not, we assume the user would never use the default crashkernel value
as custom value. When kexec-tools gets updated,
1. save the default crashkernel value of the older package to
/tmp/crashkernel (for POWER system, /tmp/crashkernel_fadump is saved
as well).
2. If auto_reset_crashkernel=yes, iterate all installed kernels.
For each kernel, compare its crashkernel value with the old
default crashkernel and reset it if yes
The implementation makes use of two RPM scriptlets [2],
- %pre is run before a package is installed so we can use it to save
old default crashkernel value
- %post is run after a package installed so we can use it to try to reset
kernel crashkernel
There are several problems when running kdumpctl in the RPM scripts
for CoreOS/Atomic/Silverblue, for example, the lock can't be acquired by
kdumpctl, "rpm-ostree kargs" can't be run and etc.. So don't enable this
feature for CoreOS/Atomic/Silverblue.
Note latest shellcheck (0.8.0) gives false positives about the
associative array as of this commit. And Fedora's shellcheck is 0.7.2
and can't even correctly parse the shell code because of the associative
array.
[1] https://github.com/koalaman/shellcheck/issues/2399
[2] https://docs.fedoraproject.org/en-US/packaging-guidelines/Scriptlets/
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
This option will determine whether to reset kernel crashkernel
to new default value or not when kexec-tools updates the default
crashkernel value and existing kernels using the old default kernel
crashkernel value. Default to yes.
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Rewrite kdumpctl reset-crashkernel KERNEL_PATH as
kdumpctl reset-crashkernel [--fadump=[on|off|nocma]] [--kernel=path_to_kernel] [--reboot]
This interface would reset a specific kernel to the default crashkernel value
given the kernel path. And it also supports grubby's syntax so there are the
following special cases,
- if --kernel not specified,
- use KDUMP_KERNELVER if it's defined in /etc/sysconfig/kdump
- otherwise use current running kernel, i.e. `uname -r`
- if --kernel=DEFAULT, the default boot kernel is chosen
- if --kernel=ALL, all kernels would have its crashkernel reset to the
default value and the /etc/default/grub is updated as well
--fadump=[on|off|nocma] toggles fadump on/off for the kernel provided
in KERNEL_PATH. If --fadump is omitted, the dump mode is determined by
parsing the kernel command line for the kernel(s) to update.
CoreOS/Atomic/Silverblue needs to be treated as a special case because,
- "rpm-ostree kargs" is used to manage kernel command line parameters
so --kernel doesn't make sense and there is no need to find current
running kernel
- "rpm-ostree kargs" itself would prompt the user to reboot the system
after modify the kernel command line parameter
- POWER is not supported so we can assume the dump mode is always kdump
This interface will also be called by kexec-tools RPM scriptlets [1]
to reset crashkernel.
Note the support of crashkenrel.default is dropped.
[1] https://docs.fedoraproject.org/en-US/packaging-guidelines/Scriptlets/
Reviewed-by: Pingfan Liu <piliu@redhat.com>
Reviewed-by: Philipp Rudo <prudo@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Coiby Xu
Philipp Rudo
Lichen Liu
Tao Liu
Kairui Song
Fedora Release Engineering
Pingfan Liu