lkml.org 
[lkml]   [2014]   [Oct]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 2/7] wait: Reimplement wait_event_freezable()
On 10/31, Peter Zijlstra wrote:
>
> Provide better implementations of wait_event_freezable() APIs.
>
> The problem is with freezer_do_not_count(), it hides the thread from
> the freezer, even though this thread might not actually freeze/sleep
> at all.

I agree, wait_event_freezable() is awful. But could you clarify "at all" ?

Sure, the task can be preempted right after it sets, it can do a lot
of things before it calls schedule(), it can be woken after that and
it can run again and do something else before freezer_count() calls
try_to_freeze(), etc.

Is this what you meant?


> +#define __wait_event_freezable(wq, condition) \
> + (void)___wait_event(wq, condition, TASK_INTERRUPTIBLE, 0, 0, \
> + schedule(); try_to_freeze())

I don't think this can work. wait_event_freezable() should be used by
kernel threads and thus we can't rely on TIF_SIGPENDING, freeze_task()
simply does wake_up_state(TASK_INTERRUPTIBLE) in this case.

Just for example, suppose that try_to_freeze_tasks() calls freeze_task()
before this kthread sets current->state = TASK_INTERRUPTIBLE. In this
case __wait_event_freezable()->schedule() will happily sleep and
try_to_freeze_tasks() will fail.

That is why I tried to suggest cmd == freezable_schedule(). Still not
good, but at least this narrows the window and (perhaps) we can improve
this freezable_schedule() later.

But on a second thought... Probably cmd => try_to_freeze(); schedule();
should work. Or just

#define __wait_event_freezable(wq, condition) \
__wait_event_interruptible(wq, ({ try_to_freeze(); (condition); }))

which looks simpler.

Oleg.



\
 
 \ /
  Last update: 2014-10-31 22:41    [W:0.094 / U:0.140 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site