| Message ID | 1367271946-7239-10-git-send-email-ccross@android.com (mailing list archive) |
|---|---|
| State | RFC, archived |
| Headers | show |
On 04/29, Colin Cross wrote: > > Avoid waking up every thread sleeping in a sigtimedwait call during > suspend and resume by calling a freezable blocking call. This doesn't explain why do want this change... OK, probably to avoid -EAGAIN from sigtimedwait() if the freezer wakes up the caller. > --- a/kernel/signal.c > +++ b/kernel/signal.c > @@ -2845,7 +2845,7 @@ int do_sigtimedwait(const sigset_t *which, siginfo_t *info, > recalc_sigpending(); > spin_unlock_irq(&tsk->sighand->siglock); > > - timeout = schedule_timeout_interruptible(timeout); > + timeout = freezable_schedule_timeout_interruptible(timeout); And I guess freezable_schedule_timeout_interruptible() is added by http://marc.info/?l=linux-kernel&m=136727195719575 ... +#define freezable_schedule_timeout_interruptible(timeout) \ +({ \ + long __retval; \ + freezer_do_not_count(); \ + __retval = schedule_timeout_interruptible(timeout); \ + freezer_count(); \ + __retval; \ +}) How this can help? The task will be interrupted anyway and the syscall will return -EAGAIN, this only changes the time when try_to_freeze() is called. For what? The task will call do_signal/try_to_freeze really "soon". Confused... Oleg. -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 04/30, Oleg Nesterov wrote: > > On 04/29, Colin Cross wrote: > > > > Avoid waking up every thread sleeping in a sigtimedwait call during > > suspend and resume by calling a freezable blocking call. > > This doesn't explain why do want this change... > > OK, probably to avoid -EAGAIN from sigtimedwait() if the freezer wakes > up the caller. > > > --- a/kernel/signal.c > > +++ b/kernel/signal.c > > @@ -2845,7 +2845,7 @@ int do_sigtimedwait(const sigset_t *which, siginfo_t *info, > > recalc_sigpending(); > > spin_unlock_irq(&tsk->sighand->siglock); > > > > - timeout = schedule_timeout_interruptible(timeout); > > + timeout = freezable_schedule_timeout_interruptible(timeout); > > And I guess freezable_schedule_timeout_interruptible() is added by > http://marc.info/?l=linux-kernel&m=136727195719575 ... > > +#define freezable_schedule_timeout_interruptible(timeout) \ > +({ \ > + long __retval; \ > + freezer_do_not_count(); \ > + __retval = schedule_timeout_interruptible(timeout); \ > + freezer_count(); \ > + __retval; \ > +}) > > How this can help? > > The task will be interrupted anyway and the syscall will return > -EAGAIN, this only changes the time when try_to_freeze() is called. OK, I wasn't cced, I have found another patch http://marc.info/?l=linux-kernel&m=136727200519602 which should make a difference, it won't be woken if PF_FREEZER_SKIP was already set. This is racy, but it seems that "avoid -EAGAIN" was not your goal... > For what? The task will call do_signal/try_to_freeze really "soon". It seems that you want to speed up the freezing. Oleg. -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Apr 30, 2013 at 9:38 AM, Oleg Nesterov <oleg@redhat.com> wrote: > On 04/29, Colin Cross wrote: >> >> Avoid waking up every thread sleeping in a sigtimedwait call during >> suspend and resume by calling a freezable blocking call. > > This doesn't explain why do want this change... > > OK, probably to avoid -EAGAIN from sigtimedwait() if the freezer wakes > up the caller. See http://marc.info/?l=linux-pm&m=136727197819593&w=2 for the full justification. I will include a fuller description of the reason for this patch in the next version. >> --- a/kernel/signal.c >> +++ b/kernel/signal.c >> @@ -2845,7 +2845,7 @@ int do_sigtimedwait(const sigset_t *which, siginfo_t *info, >> recalc_sigpending(); >> spin_unlock_irq(&tsk->sighand->siglock); >> >> - timeout = schedule_timeout_interruptible(timeout); >> + timeout = freezable_schedule_timeout_interruptible(timeout); > > And I guess freezable_schedule_timeout_interruptible() is added by > http://marc.info/?l=linux-kernel&m=136727195719575 ... > > +#define freezable_schedule_timeout_interruptible(timeout) \ > +({ \ > + long __retval; \ > + freezer_do_not_count(); \ > + __retval = schedule_timeout_interruptible(timeout); \ > + freezer_count(); \ > + __retval; \ > +}) > > How this can help? > > The task will be interrupted anyway and the syscall will return > -EAGAIN, this only changes the time when try_to_freeze() is called. > > For what? The task will call do_signal/try_to_freeze really "soon". See http://marc.info/?l=linux-pm&m=136727204919622&w=2, which removes the wakeup sent to skipped tasks, so schedule_timeout_interruptible() will only return if the timeout finishes or another task (not the freezer) calls wake_up on the task. If that happens during freezing or while frozen freezer_count() will freeze. If another task does not wake up this task while frozen, this task will not need to run at all during suspend or resume, saving cpu cycles, context switches, and power. > Confused... > > Oleg. > -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 04/30, Colin Cross wrote: > On Tue, Apr 30, 2013 at 9:38 AM, Oleg Nesterov <oleg@redhat.com> wrote: > > On 04/29, Colin Cross wrote: > >> > >> Avoid waking up every thread sleeping in a sigtimedwait call during > >> suspend and resume by calling a freezable blocking call. > > > > This doesn't explain why do want this change... > > > > OK, probably to avoid -EAGAIN from sigtimedwait() if the freezer wakes > > up the caller. > > See http://marc.info/?l=linux-pm&m=136727197819593&w=2 for the full > justification. I will include a fuller description of the reason for > this patch in the next version. Yes, thanks, I already realized what are you trying to do. > > The task will be interrupted anyway and the syscall will return > > -EAGAIN, this only changes the time when try_to_freeze() is called. > > > > For what? The task will call do_signal/try_to_freeze really "soon". > > See http://marc.info/?l=linux-pm&m=136727204919622&w=2, which removes > the wakeup sent to skipped tasks, so schedule_timeout_interruptible() > will only return if the timeout finishes or another task Or if freeze_task() was already called. but I guess you do not care. Oleg. -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/kernel/signal.c b/kernel/signal.c index 598dc06..10a70a0 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -2845,7 +2845,7 @@ int do_sigtimedwait(const sigset_t *which, siginfo_t *info, recalc_sigpending(); spin_unlock_irq(&tsk->sighand->siglock); - timeout = schedule_timeout_interruptible(timeout); + timeout = freezable_schedule_timeout_interruptible(timeout); spin_lock_irq(&tsk->sighand->siglock); __set_task_blocked(tsk, &tsk->real_blocked);
Avoid waking up every thread sleeping in a sigtimedwait call during suspend and resume by calling a freezable blocking call. Signed-off-by: Colin Cross <ccross@android.com> --- kernel/signal.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)