lkml.org 
[lkml]   [2010]   [Dec]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: buggy perf callgraph output
On Thu, Dec 09, 2010 at 03:40:15AM +1100, Nick Piggin wrote:
> Hi,
>
> I get strange looking perf output with the callchain hits not sorted
> properly for graph and fractal:
>
> 7.34% open [kernel.kallsyms] [k] kmem_cache_alloc
> |
> --- kmem_cache_alloc
> |
> |---0.05%-- get_empty_filp
> | do_filp_open
> | do_sys_open
> | sys_open
> | system_call
> | __open_nocancel
> |
> ---0.16%-- getname
> do_sys_open
> sys_open
> system_call
> __open_nocancel
>
> 6.17% open [kernel.kallsyms] [k] _raw_spin_lock
> |
> --- _raw_spin_lock
> |
> |---0.01%-- add_partial
> | __slab_free
> | kmem_cache_free
> | file_free_rcu
> | __rcu_process_callbacks
> | rcu_process_callbacks
> | __do_softirq
> | call_softirq
> | do_softirq
> | irq_exit
> | smp_apic_timer_interrupt
> | apic_timer_interrupt
> |
> |---0.16%-- dput
> | |
> | --0.57%-- fput
> | filp_close
> | sys_close
> | system_call
> | __libc_close
> |
> |--0.60%-- link_path_walk
> | do_path_lookup
> | do_filp_open
> | do_sys_open
> | sys_open
> | system_call
> | __open_nocancel
> |
> |--0.57%-- __slab_free
> | kmem_cache_free
> | file_free_rcu
> | __rcu_process_callbacks
> | rcu_process_callbacks
> | __do_softirq
>
>
> With flat it looks a bit better:
> 7.34% open [kernel.kallsyms] [k] kmem_cache_alloc
> 6.09%
> kmem_cache_alloc
> get_empty_filp
> do_filp_open
> do_sys_open
> sys_open
> system_call
> __open_nocancel
>
> 1.07%
> kmem_cache_alloc
> getname
> do_sys_open
> sys_open
> system_call
> __open_nocancel
>
> 6.17% open [kernel.kallsyms] [k] _raw_spin_lock
> 0.60%
> _raw_spin_lock
> link_path_walk
> do_path_lookup
> do_filp_open
> do_sys_open
> sys_open
> system_call
> __open_nocancel
>
> 0.57%
> _raw_spin_lock
> fd_install
> do_sys_open
> sys_open
> system_call
> __open_nocancel
>
> 0.57%
> _raw_spin_lock
> dput
> fput
> filp_close
> sys_close
> system_call
> __libc_close


I can not reproduce it. Could you please try to reproduce,
run perf archive and send me your perf.data.tar.bz2 ?

Thanks.


\
 
 \ /
  Last update: 2010-12-08 22:51    [W:0.957 / U:0.604 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site