Subject : BUG at perf_ctx_adjust_freq (kernel/perf_event.c:1582)
Submitter : Chris Wilson <email@example.com>
Date : 2010-12-10 12:00
Message-ID : firstname.lastname@example.org
References : http://marc.info/?l=linux-kernel&m=129198247531612&w=2
This entry is being used for tracking a regression from 2.6.36. Please don't
close it until the problem is fixed in the mainline.
I'm closing this as unreproducible for now. If it happens again, please shout.
Got same issue with plain 2.6.37. Can't really reproduce it as hunted this down for the first time but I suspect that same have happened a day before on two different hosts. Both are under huge load so this might be a key. Got a screenshot from iLO a bit mixed with top output.
Created attachment 51742 [details]
screenshot from iLO
Got the screenshot from iLO.
Another two hosts running vanilla 2.6.37 died in the night due to same issue. Please reopen.
And another one. top was running and the last thing it ,,recorded'' was:
top - 09:52:52 up 17:39, 14 users, load average: 3.25, 2.45, 2.70
Tasks: 472 total, 9 running, 458 sleeping, 1 stopped, 4 zombie
Cpu(s): 58.8%us, 6.1%sy, 5.1%ni, 29.2%id, 0.6%wa, 0.0%hi, 0.1%si, 0.0%st
Mem: 49553716k total, 49115852k used, 437864k free, 279460k buffers
Swap: 4095996k total, 0k used, 4095996k free, 39703932k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
23840 userr 20 0 706m 107m 9128 S 200.7 0.2 0:08.15 java
23544 userr 20 0 276m 16m 5328 S 140.7 0.0 0:05.63 wmp
Created attachment 51882 [details]
Another crash screenshot
Clean screenshot from last crash.
Is this still an issue in 2.6.38?
Did not tested that on 2.6.38. Might be hard as this is triggered by production CI system. For sure I can confirm the problem is in huge number of processes running which cause a load above 100.