Re: [RFC PATCH] perf: fix panic by mark recursion inside perf_log_throttle

[王贇]

On 2021/9/13 上午11:00, 王贇 wrote:
> 
> 
> On 2021/9/10 下午11:38, Peter Zijlstra wrote:
>> On Thu, Sep 09, 2021 at 11:13:21AM +0800, 王贇 wrote:
>>> When running with ftrace function enabled, we observed panic
>>> as below:
>>>
>>>   traps: PANIC: double fault, error_code: 0x0
>>>   [snip]
>>>   RIP: 0010:perf_swevent_get_recursion_context+0x0/0x70
>>>   [snip]
>>>   Call Trace:
>>>    <NMI>
>>>    perf_trace_buf_alloc+0x26/0xd0
>>>    perf_ftrace_function_call+0x18f/0x2e0
>>>    kernelmode_fixup_or_oops+0x5/0x120
>>>    __bad_area_nosemaphore+0x1b8/0x280
>>>    do_user_addr_fault+0x410/0x920
>>>    exc_page_fault+0x92/0x300
>>>    asm_exc_page_fault+0x1e/0x30
>>>   RIP: 0010:__get_user_nocheck_8+0x6/0x13
>>>    perf_callchain_user+0x266/0x2f0
>>>    get_perf_callchain+0x194/0x210
>>>    perf_callchain+0xa3/0xc0
>>>    perf_prepare_sample+0xa5/0xa60
>>>    perf_event_output_forward+0x7b/0x1b0
>>>    __perf_event_overflow+0x67/0x120
>>>    perf_swevent_overflow+0xcb/0x110
>>>    perf_swevent_event+0xb0/0xf0
>>>    perf_tp_event+0x292/0x410
>>>    perf_trace_run_bpf_submit+0x87/0xc0
>>>    perf_trace_lock_acquire+0x12b/0x170
>>>    lock_acquire+0x1bf/0x2e0
>>>    perf_output_begin+0x70/0x4b0
>>>    perf_log_throttle+0xe2/0x1a0
>>>    perf_event_nmi_handler+0x30/0x50
>>>    nmi_handle+0xba/0x2a0
>>>    default_do_nmi+0x45/0xf0
>>>    exc_nmi+0x155/0x170
>>>    end_repeat_nmi+0x16/0x55
>>
>> kernel/events/Makefile has:
>>
>> ifdef CONFIG_FUNCTION_TRACER
>> CFLAGS_REMOVE_core.o = $(CC_FLAGS_FTRACE)
>> endif
>>
>> Which, afaict, should avoid the above, no?
> 
> I'm afraid it's not working for this case, the
> start point of tracing is at lock_acquire() which
> is not from 'kernel/events/core', the following PF
> related function are also not from 'core', prevent
> ftrace on 'core' can't prevent this from happen...

By a second thinking, I think you're right about the
way it should be fixed, since disabling ftrace on
'arch/x86/mm/fault.c' could also fix the problem.

Will send a formal patch later :-)

Regards,
Michael Wang

> 
> Regards,
> Michael Wang
> 
>>