Created attachment 727827 [details]
reproducer code
The testsuite of systemd version 233 contains a test of the machines watchdog.
When running it on a machine with hardware watchdog the kernel crashes:
-->
teviot login: root
Password:
Last login: Tue Jun 6 16:22:57 from 2620:113:80c0:8000:c::50a
Have a lot of fun...
teviot:~ #
teviot:~ # cd /systemd-testsuite/run
teviot:/systemd-testsuite/run # ./test-watchdog
Hardware watchdog 'HPE iLO2+ HW [ 185.386548] hpwdt: Unexpected close, not
stopping watchdog!
Watchdog Timer', version 0teviot:/systemd-testsuite/run #
teviot:/systemd-testsuite/run #
teviot:/systemd-testsuite/run #
teviot:/systemd-testsuite/run # [ 208.152002] Kernel panic - not syncing: An
NMI occurred. Depending on your system the reason for the NMI is logged in any
one of the following resources:
[ 208.152002] 1. Integrated Management Log (IML)
[ 208.152002] 2. OA Syslog
[ 208.152002] 3. OA Forward Progress Log
[ 208.152002] 4. iLO Event Log
[ 208.152002] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.11.3-1-default #1
[ 208.152002] Hardware name: HP ProLiant BL465c G1 , BIOS A13 05/02/2011
[ 208.152002] Call Trace:
[ 208.152002] <NMI>
[ 208.152002] dump_stack+0x5c/0x78
[ 208.152002] panic+0xd5/0x21e
[ 208.152002] nmi_panic+0x35/0x40
[ 208.152002] hpwdt_pretimeout+0x7f/0xe7 [hpwdt]
[ 208.152002] nmi_handle+0x60/0x120
[ 208.152002] unknown_nmi_error+0x16/0x80
[ 208.152002] do_nmi+0xe5/0x130
[ 208.152002] end_repeat_nmi+0x1a/0x1e
[ 208.152002] ? native_safe_halt+0x2/0x10
[ 208.152002] ? native_safe_halt+0x2/0x10
[ 208.152002] ? native_safe_halt+0x2/0x10
[ 208.152002] </NMI>
[ 208.152002] ? default_idle+0x1a/0x100
[ 208.152002] ? do_idle+0x161/0x1f0
[ 208.152002] ? cpu_startup_entry+0x5d/0x60
[ 208.152002] ? start_kernel+0x436/0x43e
[ 208.152002] ? early_idt_handler_array+0x120/0x120
[ 208.152002] ? x86_64_start_kernel+0x127/0x136
[ 208.152002] ? start_cpu+0x14/0x14
[ 208.152002] Kernel Offset: 0x3a000000 from 0xffffffff81000000 (relocation
range: 0xffffffff80000000-0xffffffffbfffffff)
--<
Unfortunatel, I couldn't find any more information about the NMI in the IML or
the iLO log.
I could, however reproduce the issue with a code snippet broken out of systemd.
The question is whether this is a kernel bug or a bug in the systemd code.
Attaching the reproducer.
Could the kernel maintainers please take a look and give a statement?