public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
* [PATCH] nptl: pthread_rwlock_rdlock return in low priority
@ 2023-03-07  3:50 abushwang
  2023-03-07  8:40 ` Xi Ruoyao
  2023-03-07  8:56 ` Florian Weimer
  0 siblings, 2 replies; 6+ messages in thread
From: abushwang @ 2023-03-07  3:50 UTC (permalink / raw)
  To: libc-alpha, drepper.fsp; +Cc: abushwang

hi, I have noticed reader will return directly on fast-path in pthread_rwlock_common.c

>  /* We have registered as a reader, so if we are in a read phase, we have
>     acquired a read lock.  This is also the reader--reader fast-path.
>     Even if there is a primary writer, we just return.  If writers are to
>     be preferred and we are the only active reader, we could try to enter a
>     write phase to let the writer proceed.  This would be okay because we
>     cannot have acquired the lock previously as a reader (which could result
>     in deadlock if we would wait for the primary writer to run).  However,
>     this seems to be a corner case and handling it specially not be worth the
>     complexity.  */
>  if (__glibc_likely ((r & PTHREAD_RWLOCK_WRPHASE) == 0))
>    return 0;

However, there is a situation:
    main, thread_wr, thread_rd.

    SCHED_FIFO priority:
        main > thread_wr > thread_rd
    main first acquires read lock, then create thread_wr which will block on the lock.
    Next, main creates thread_rd. this thread will acquires read lock on fast-path even
    though it has a lower priority compared to thread_wr.

You can get demo from the following repository:
https://github.com/emscripten-core/posixtestsuite.git
./conformance/interfaces/pthread_rwlock_rdlock/2-1.c

According to "man -M man-pages-posix-2017/ 3p pthread_rwlock_rdlock"

> DESCRIPTION
> The pthread_rwlock_rdlock() function shall apply a read lock to the
> read-write lock referenced by rwlock.  The calling thread acquires the
> read lock  if  a writer does not hold the lock and there are no
> writers blocked on the lock.
>
> If  the  Thread  Execution  Scheduling  option  is supported,  and the
> threads involved in the lock are executing with the scheduling
> policies SCHED_FIFO or SCHED_RR, the calling thread shall not acquire
> the lock if a writer holds the lock or if writers of higher or equal
> priority are blocked on the lock;  other‐ wise, the calling thread
> shall acquire the lock.

I was wondering that whether this fast-path is reasonable, and whether
this posix standard should be enforced.

Thanks

Signed-off-by: abushwang <abushwangs@gmail.com>
---
 nptl/pthread_rwlock_common.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/nptl/pthread_rwlock_common.c b/nptl/pthread_rwlock_common.c
index 5266a00ed1..35b00fc14f 100644
--- a/nptl/pthread_rwlock_common.c
+++ b/nptl/pthread_rwlock_common.c
@@ -389,6 +389,7 @@ __pthread_rwlock_rdlock_full64 (pthread_rwlock_t *rwlock, clockid_t clockid,
      complexity.  */
   if (__glibc_likely ((r & PTHREAD_RWLOCK_WRPHASE) == 0))
     return 0;
+
   /* Otherwise, if we were in a write phase (states #6 or #8), we must wait
      for explicit hand-over of the read phase; the only exception is if we
      can start a read phase if there is no primary writer currently.  */
-- 
2.36.1


^ permalink raw reply	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2023-03-07  9:57 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-03-07  3:50 [PATCH] nptl: pthread_rwlock_rdlock return in low priority abushwang
2023-03-07  8:40 ` Xi Ruoyao
2023-03-07  8:53   ` abush wang
2023-03-07  8:56 ` Florian Weimer
2023-03-07  9:06   ` abush wang
2023-03-07  9:57     ` Xi Ruoyao

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 read-only IMAP folder(s) and NNTP newsgroup(s).