Various panics while using HWPMC on ARM64

Zbigniew Bodek zbb at semihalf.com
Mon Dec 14 15:24:42 UTC 2015


Hello,

Did you have time to look into that? Do you have any clues what could
be wrong here?
We would like to use hwpmc for profiling so your help will be very
much appreciated.

Best regards
zbb

2015-12-09 13:06 GMT+01:00 Zbigniew Bodek <zbb at semihalf.com>:
> Hello Ed,
>
> Done. I also check what happens when SMP is disabled and the kassert
> is triggered:
>
> root at thunderx_crb4:~ # pmcstat -S CPU_CYCLES -O cpu_cycles.pmc
> ^Cpanic: [pmc,4256] cpu 0 didn't find a sample to collect
> KDB: stack backtrace:
> db_trace_self() at db_trace_self_wrapper+0x28
> pc = 0xffffff80004e9aac  lr = 0xffffff800006d8b4
> sp = 0xffffff87cba976e0  fp = 0xffffff87cba97800
>
> db_trace_self_wrapper() at vpanic+0x9c
> pc = 0xffffff800006d8b4  lr = 0xffffff800027136c
> sp = 0xffffff87cba97810  fp = 0xffffff87cba97880
>
> vpanic() at kassert_panic+0x160
> pc = 0xffffff800027136c  lr = 0xffffff80002712cc
> sp = 0xffffff87cba97890  fp = 0xffffff87cba97950
>
> kassert_panic() at pmc_capture_user_callchain+0x1a4
> pc = 0xffffff80002712cc  lr = 0xffffff80000e1444
> sp = 0xffffff87cba97960  fp = 0xffffff87cba979c0
>
> pmc_capture_user_callchain() at pmc_hook_handler+0x7c0
> pc = 0xffffff80000e1444  lr = 0xffffff80000dfb78
> sp = 0xffffff87cba979d0  fp = 0xffffff87cba97a50
>
> pmc_hook_handler() at ast+0x14c
> pc = 0xffffff80000dfb78  lr = 0xffffff80002b976c
> sp = 0xffffff87cba97a60  fp = 0xffffff87cba97a90
>
> ast() at handle_el0_sync+0x90
> pc = 0xffffff80002b976c  lr = 0xffffff80004eb224
> sp = 0xffffff87cba97aa0  fp = 0xffffff87cba97bb0
>
> handle_el0_sync() at 0x406d60
> pc = 0xffffff80004eb224  lr = 0x0000000000406d60
> sp = 0xffffff87cba97bc0  fp = 0x0000007ffffff540
>
> KDB: enter: panic
> [ thread pid 679 tid 100061 ]
> Stopped at      kdb_enter+0x40:
> db>
>
>
> when invariants options is disabled I only get:
>
> root at thunderx_crb4:~ # pmcstat -S CPU_CYCLES -O cpu_cycles.pmc
> ^Cpmcstat: WARNING: sampling was paused at least 1 time.
> Please consider tuning the "kern.hwpmc.nsamples" tunable.
>
>
> Best regards
> zbb
>
> 2015-12-08 20:59 GMT+01:00 Ed Maste <emaste at freebsd.org>:
>> On 8 December 2015 at 14:34, Zbigniew Bodek <zbb at semihalf.com> wrote:
>>> Hello,
>>>
>>> I encountered some problems with FreeBSD on ARM64 while using hwpmc.
>>> Some of the errors that I found are listed below:
>>>
>>> * panic: Unknown kernel exception 0 esr_el1 2000000
>>> * panic: data abort in critical section or under mutex
>>> * panic: VFP exception in the kernel
>>> * panic: Unknown kernel exception 21 esr_el1 86000006
>>
>> Can you add these notes to PR 204686? I think there are SMP issues in
>> arm64 hwpmc that need to be resolved.


More information about the freebsd-arm mailing list