Messages in this thread | | | Date | Thu, 17 Sep 2015 14:54:45 +0900 | From | Hiraku Toyooka <> | Subject | Re: [PATCH 2/2] selftests/pstore: add pstore test scripts going with reboot |
| |
Hello,
>> +prlog "Causing kernel crash ..." >> + >> +# enable all functions triggered by sysrq >> +echo 1 > /proc/sys/kernel/sysrq >> +# setting to reboot in 3 seconds after panic >> +echo 3 > /proc/sys/kernel/panic >> +# setting to cause panic when oops occurs >> +echo 1 > /proc/sys/kernel/panic_on_oops >> + >> +# create a file as reboot flag >> +touch $REBOOT_FILE >> +sync >> + >> +# cause crash >> +echo c > /proc/sysrq-trigger > > Do you need to stop kdump service before the sysrq?
Yes, I should check /sys/kernel/kexec_crash_loaded. If the value is 1, this script should try to unload kexec kernel.
> Or, does it cover oops and kdump case?
No, not yet. I think we should support oops case at first.
Best regards, Hiraku Toyooka
阿口誠司 / AGUCHI,SEIJI wrote: > >> +prlog "Causing kernel crash ..." >> + >> +# enable all functions triggered by sysrq >> +echo 1 > /proc/sys/kernel/sysrq >> +# setting to reboot in 3 seconds after panic >> +echo 3 > /proc/sys/kernel/panic >> +# setting to cause panic when oops occurs >> +echo 1 > /proc/sys/kernel/panic_on_oops >> + >> +# create a file as reboot flag >> +touch $REBOOT_FILE >> +sync >> + >> +# cause crash >> +echo c > /proc/sysrq-trigger > > Do you need to stop kdump service before the sysrq? > Or, does it cover oops and kdump case? > > Seiji > >> -----Original Message----- >> From: 豊岡拓 / Toyooka,Hiraku >> Sent: Tuesday, September 15, 2015 11:42 AM >> To: Kees Cook >> Cc: LKML; Tony Luck; Linux API; Anton Vorontsov; Shuah Khan; Mark Salyzyn; Colin Cross; 阿口誠司 / AGUCHI,SEIJI >> Subject: Re: [PATCH 2/2] selftests/pstore: add pstore test scripts going with reboot >> >> Hello Kees, >> >> >> +run_crash: >> >> + @sh pstore_crash_test || echo "pstore_crash_test: [FAIL]" >> > >> > This is probably better written to exit 1 on failure, otherwise it >> > just _says_ it fails. (Though lots of selftests in the tree already >> > have this problem, it's best to avoid the pattern for new stuff.) >> > Maybe something like: >> > >> > @sh pstore_crash_test || { echo "pstore_crash_test: [FAIL]"; >> exit 1; } >> >> OK. I'll add the "exit 1". >> >> >> +prlog -n "Checking dmesg files exist in pstore filesystem ... " >> >> +if [ -e dmesg-${backend}-0 ]; then >> >> + prlog "ok" >> >> + for f in `ls dmesg-${backend}-*`; do >> >> + prlog -e "\t${f}" >> >> + done >> >> +else >> >> + prlog "FAIL" >> >> + rc=1 >> >> +fi >> > >> > This test pattern is repeated a lot. Maybe better to create a helper >> > function instead? It could make the tests much more readable. >> >> Yes, I should make a helper function in v2. >> >> Best regards, >> Hiraku Toyooka
-- Hiraku Toyooka Systems Productivity Research Dept. / Linux Technology Center Center for Technology Innovation - Systems Engineering, Hitachi Ltd.
| |