LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] signal: reorder struct sighand_struct
@ 2019-05-03 19:28 Alexey Dobriyan
  2019-05-04 23:33 ` Michal Hocko
  0 siblings, 1 reply; 3+ messages in thread
From: Alexey Dobriyan @ 2019-05-03 19:28 UTC (permalink / raw)
  To: akpm; +Cc: linux-kernel

struct sighand_struct::siglock field is the most used field by far,
put it first so that is can be accessed without IMM8 or IMM32 encoding
on x86_64.

Space savings (on trimmed down VM test config):

add/remove: 0/0 grow/shrink: 8/68 up/down: 49/-1147 (-1098)
Function                                     old     new   delta
complete_signal                              512     533     +21
do_signalfd4                                 335     346     +11
__cleanup_sighand                             39      43      +4
unhandled_signal                              49      52      +3
prepare_signal                               692     695      +3
ignore_signals                                37      40      +3
__tty_check_change.part                      248     251      +3
ksys_unshare                                 780     781      +1
sighand_ctor                                  33      29      -4
ptrace_trap_notify                            60      56      -4
sigqueue_free                                 98      91      -7
run_posix_cpu_timers                        1389    1382      -7
proc_pid_status                             2448    2441      -7
proc_pid_limits                              344     337      -7
posix_cpu_timer_rearm                        222     215      -7
posix_cpu_timer_get                          249     242      -7
kill_pid_info_as_cred                        243     236      -7
freeze_task                                  197     190      -7
flush_old_exec                              1873    1866      -7
do_task_stat                                3363    3356      -7
do_send_sig_info                              98      91      -7
do_group_exit                                147     140      -7
init_sighand                                2088    2080      -8
do_notify_parent_cldstop                     399     391      -8
signalfd_cleanup                              50      41      -9
do_notify_parent                             557     545     -12
__send_signal                               1029    1017     -12
ptrace_stop                                  590     577     -13
get_signal                                  1576    1563     -13
__lock_task_sighand                          112      99     -13
zap_pid_ns_processes                         391     377     -14
update_rlimit_cpu                             78      64     -14
tty_signal_session_leader                    413     399     -14
tty_open_proc_set_tty                        149     135     -14
tty_jobctrl_ioctl                            936     922     -14
set_cpu_itimer                               339     325     -14
ptrace_resume                                226     212     -14
ptrace_notify                                110      96     -14
proc_clear_tty                                81      67     -14
posix_cpu_timer_del                          229     215     -14
kernel_sigaction                             156     142     -14
getrusage                                    977     963     -14
get_current_tty                               98      84     -14
force_sigsegv                                 89      75     -14
force_sig_info                               205     191     -14
flush_signals                                 83      69     -14
flush_itimer_signals                          85      71     -14
do_timer_create                             1120    1106     -14
do_sigpending                                 88      74     -14
do_signal_stop                               537     523     -14
cgroup_init_fs_context                       644     630     -14
call_usermodehelper_exec_async               402     388     -14
calculate_sigpending                          58      44     -14
__x64_sys_timer_delete                       248     234     -14
__set_current_blocked                         80      66     -14
__ptrace_unlink                              310     296     -14
__ptrace_detach.part                         187     173     -14
send_sigqueue                                362     347     -15
get_cpu_itimer                               214     199     -15
signalfd_poll                                175     159     -16
dequeue_signal                               340     323     -17
do_getitimer                                 192     174     -18
release_task.part                           1060    1040     -20
ptrace_peek_siginfo                          408     387     -21
posix_cpu_timer_set                          827     806     -21
exit_signals                                 437     416     -21
do_sigaction                                 541     520     -21
do_setitimer                                 485     464     -21
disassociate_ctty.part                       545     517     -28
__x64_sys_rt_sigtimedwait                    721     679     -42
__x64_sys_ptrace                            1319    1277     -42
ptrace_request                              1828    1782     -46
signalfd_read                                507     459     -48
wait_consider_task                          2027    1971     -56
do_coredump                                 3672    3616     -56
copy_process.part                           6936    6871     -65

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
---

 include/linux/sched/signal.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/include/linux/sched/signal.h
+++ b/include/linux/sched/signal.h
@@ -15,10 +15,10 @@
  */
 
 struct sighand_struct {
-	refcount_t		count;
-	struct k_sigaction	action[_NSIG];
 	spinlock_t		siglock;
+	refcount_t		count;
 	wait_queue_head_t	signalfd_wqh;
+	struct k_sigaction	action[_NSIG];
 };
 
 /*

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] signal: reorder struct sighand_struct
  2019-05-03 19:28 [PATCH] signal: reorder struct sighand_struct Alexey Dobriyan
@ 2019-05-04 23:33 ` Michal Hocko
  2019-05-24  2:11   ` Andrew Morton
  0 siblings, 1 reply; 3+ messages in thread
From: Michal Hocko @ 2019-05-04 23:33 UTC (permalink / raw)
  To: Alexey Dobriyan; +Cc: akpm, linux-kernel, Oleg Nesterov

CCing Oleg.

On Fri 03-05-19 22:28:00, Alexey Dobriyan wrote:
[...]
> add/remove: 0/0 grow/shrink: 8/68 up/down: 49/-1147 (-1098)
[...]
> --- a/include/linux/sched/signal.h
> +++ b/include/linux/sched/signal.h
> @@ -15,10 +15,10 @@
>   */
>  
>  struct sighand_struct {
> -	refcount_t		count;
> -	struct k_sigaction	action[_NSIG];
>  	spinlock_t		siglock;
> +	refcount_t		count;
>  	wait_queue_head_t	signalfd_wqh;
> +	struct k_sigaction	action[_NSIG];
>  };

Is it possible that this would cause false sharing of the cache line
that would have performance implications now?

-- 
Michal Hocko
SUSE Labs

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] signal: reorder struct sighand_struct
  2019-05-04 23:33 ` Michal Hocko
@ 2019-05-24  2:11   ` Andrew Morton
  0 siblings, 0 replies; 3+ messages in thread
From: Andrew Morton @ 2019-05-24  2:11 UTC (permalink / raw)
  To: Michal Hocko; +Cc: Alexey Dobriyan, linux-kernel, Oleg Nesterov

On Sat, 4 May 2019 19:33:02 -0400 Michal Hocko <mhocko@kernel.org> wrote:

> CCing Oleg.
> 
> On Fri 03-05-19 22:28:00, Alexey Dobriyan wrote:
> [...]
> > add/remove: 0/0 grow/shrink: 8/68 up/down: 49/-1147 (-1098)
> [...]
> > --- a/include/linux/sched/signal.h
> > +++ b/include/linux/sched/signal.h
> > @@ -15,10 +15,10 @@
> >   */
> >  
> >  struct sighand_struct {
> > -	refcount_t		count;
> > -	struct k_sigaction	action[_NSIG];
> >  	spinlock_t		siglock;
> > +	refcount_t		count;
> >  	wait_queue_head_t	signalfd_wqh;
> > +	struct k_sigaction	action[_NSIG];
> >  };
> 
> Is it possible that this would cause false sharing of the cache line
> that would have performance implications now?

Doesn't seem likely.  Possible .count vs .siglock, but .count only gets
altered by fork/exec-style code, so it's pretty low bandwidth.


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2019-05-24  2:11 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-03 19:28 [PATCH] signal: reorder struct sighand_struct Alexey Dobriyan
2019-05-04 23:33 ` Michal Hocko
2019-05-24  2:11   ` Andrew Morton

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).