On 2019/11/12 22:30, Steven Rostedt wrote:
> On Tue, 12 Nov 2019 14:18:50 +0800
> Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
>
>> From: "Steven Rostedt (Red Hat)" <rostedt@xxxxxxxxxxx>
>>
>> Commit 33fddff24d05d71f97722cb7deec4964d39d10dc upstream
>>
>> There's no real difference between trace_buffer_unlock_commit() and
>> trace_buffer_unlock_commit_regs() except that the former passes NULL to
>> ftrace_stack_trace() instead of regs. Have the former be a static inline of
>> the latter which passes NULL for regs.
>>
>> Signed-off-by: Steven Rostedt <rostedt@xxxxxxxxxxx>
>> [backport: modify trace_buffer_unlock_commit() in include/linux/trace_events.h]
>> Fixes: c5e0535fe67b ("tracing: Skip more functions when doing stack tracing of events")
^^^^^^^^^^^^
should be 70b3d6c5aa71 in v4.4.x, I will resend a new one if no objection.
>> Signed-off-by: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
>> ---
>
> I'm not against this backport, but I'd like to know more why you feel
> this is a stable fix. Just to get a better backtrace? Can you show some
> examples of the issue you have and how this fixes it?
PATCH1: "tracing: Skip more functions when doing stack tracing of events" from v4.8-rc1~111^2~9
PATCH2: "tracing: Have trace_buffer_unlock_commit() call the _regs version with NULL" from tags/v4.7-rc1~72^2~8
PATCH1 is after PATCH2, the "skip" of ftrace_trace_stack() called by trace_buffer_unlock_commit_regs()
and trace_buffer_unlock_commit() is all changed. but we only have PATCH1 in 4.4.x, I found this when
using stacktrace on arm64,
cd /sys/kernel/debug/tracing
echo 1 > options/stacktrace
echo 1 > events/kmem/mm_page_alloc/enable
cat trace
before
---
sh-781 [004] ...1 31.300106: mm_page_alloc: page=ffffffbdc0246380 pfn=299406 order=0 migratetype=1 gfp_flags=GFP_HIGHUSER_MOVABLE
sh-781 [004] ...1 31.301373: <stack trace>
=> do_page_fault
=> do_translation_fault
=> do_mem_abort
=> el0_da
sh-781 [004] ...1 31.303494: mm_page_alloc: page=ffffffbdc021cf00 pfn=296764 order=2 migratetype=0 gfp_flags=GFP_USER|GFP_ZERO|GFP_NOTRACK
sh-781 [004] ...1 31.303508: <stack trace>
=> copy_process.isra.8
=> _do_fork
=> SyS_clone
=> el0_svc_naked
sh-781 [004] ...1 31.303975: mm_page_alloc: page=ffffffbdc021cd40 pfn=296757 order=0 migratetype=0 gfp_flags=GFP_USER|GFP_REPEAT|GFP_ZERO|GFP_NOTRACK
sh-781 [004] ...1 31.303987: <stack trace>
=> pgd_alloc
=> mm_init.isra.4
=> copy_process.isra.8
=> _do_fork
=> SyS_clone
=> el0_svc_naked
after
---
sh-780 [002] ...1 14.728783: mm_page_alloc: page=ffffffbdc2e86200 pfn=1024392 order=0 migratetype=1 gfp_flags=GFP_HIGHUSER_MOVABLE
sh-780 [002] ...1 14.729985: <stack trace>
=> __alloc_pages_nodemask
=> handle_mm_fault
=> do_page_fault
=> do_translation_fault
=> do_mem_abort
=> el0_da
sh-780 [002] ...1 14.731854: mm_page_alloc: page=ffffffbdc021dc00 pfn=296816 order=2 migratetype=0 gfp_flags=GFP_USER|GFP_ZERO|GFP_NOTRACK
sh-780 [002] ...1 14.731867: <stack trace>
=> __alloc_pages_nodemask
=> alloc_kmem_pages_node
=> copy_process.isra.8
=> _do_fork
=> SyS_clone
=> el0_svc_naked
sh-780 [002] ...1 14.732069: mm_page_alloc: page=ffffffbdc021d600 pfn=296792 order=0 migratetype=0 gfp_flags=GFP_USER|GFP_REPEAT|GFP_ZERO|GFP_NOTRACK
sh-780 [002] ...1 14.732079: <stack trace>
=> __alloc_pages_nodemask
=> __get_free_pages
=> pgd_alloc
=> mm_init.isra.4
=> copy_process.isra.8
=> _do_fork
=> SyS_clone
=> el0_svc_naked
>
> Thanks!
>
> -- Steve
>
> .
>
