[Bug nptl/16410] New: pthread_cond_timedwait has 70us latency when abstime already expired

[Bug nptl/16410] New: pthread_cond_timedwait has 70us latency when abstime already expired

[weiqi at kylinos dot com.cn]

https://sourceware.org/bugzilla/show_bug.cgi?id=16410

            Bug ID: 16410
           Summary: pthread_cond_timedwait has 70us latency when abstime
                    already expired
           Product: glibc
           Version: 2.18
            Status: NEW
          Severity: normal
          Priority: P2
         Component: nptl
          Assignee: unassigned at sourceware dot org
          Reporter: weiqi at kylinos dot com.cn
                CC: drepper.fsp at gmail dot com

Created attachment 7340
  --> https://sourceware.org/bugzilla/attachment.cgi?id=7340&action=edit
the test.c for this problem

Since this patch
(http://repo.or.cz/w/glibc.git/commit/e88726b483a275824e852f64476087568dbae7bb),
it seems that  pthread_cond_timedwait.S does not care the arg "abstime" which
may already expired. Pass it to syscall futex, if the futex have the ability
FUTEX_CLOCK_REALTIME. 

    Before the futex_clock_realtime patch, pthread_cond_timedwait.S check the
abstime arg firstly by syscall clock_gettime, so if abstime already expired,
pthread_cond_timedwait would return -ETIMEDOUT immediatly.

    I have reported a bug to redhat bugzilla described this
case(https://bugzilla.redhat.com/show_bug.cgi?id=1048731),and added a test.c
show this.

    The kernel futex_wait use hrtimer to calculate timeout, and do not special
considerations such cases.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug nptl/16410] pthread_cond_timedwait has 70us latency when abstime already expired

[weiqi at kylinos dot com.cn]

https://sourceware.org/bugzilla/show_bug.cgi?id=16410

weiqi <weiqi at kylinos dot com.cn> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |weiqi at kylinos dot com.cn

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug nptl/16410] pthread_cond_timedwait has 70us latency when abstime already expired

[neleai at seznam dot cz]

https://sourceware.org/bugzilla/show_bug.cgi?id=16410

Ondrej Bilka <neleai at seznam dot cz> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |neleai at seznam dot cz
            Version|2.18                        |2.20

--- Comment #1 from Ondrej Bilka <neleai at seznam dot cz> ---
This could be handled after freeze.


It looks that patch would be putting a check back which should be relatively
straightforward.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug nptl/16410] pthread_cond_timedwait has 70us latency when abstime already expired

[weiqi at kylinos dot com.cn]

https://sourceware.org/bugzilla/show_bug.cgi?id=16410

--- Comment #2 from weiqi <weiqi at kylinos dot com.cn> ---
(In reply to Ondrej Bilka from comment #1)
> This could be handled after freeze.
> 
Is there any reason futex_wait do not check the abstime at first?

> 
> It looks that patch would be putting a check back which should be relatively
> straightforward.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug nptl/16410] pthread_cond_timedwait has 70us latency when abstime already expired

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16410

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.