[PATCH] Document a behavior of an elided pthread_rwlock_unlock

[PATCH] Document a behavior of an elided pthread_rwlock_unlock

[Tulio Magno Quites Machado Filho]

Explain that pthread_rwlock_unlock may crash if called on a lock not
held by the current thread.

2016-10-21  Tulio Magno Quites Machado Filho  <tuliom@linux.vnet.ibm.com>

	* nptl/pthread_rwlock_unlock.c: Add a comment explaining its
	behavior when eliding a lock not held by the current thread.
	* sysdeps/powerpc/nptl/elide.h: Likewise.
---
 nptl/pthread_rwlock_unlock.c | 4 ++++
 sysdeps/powerpc/nptl/elide.h | 2 ++
 2 files changed, 6 insertions(+)

diff --git a/nptl/pthread_rwlock_unlock.c b/nptl/pthread_rwlock_unlock.c
index a6cadd4..112f748 100644
--- a/nptl/pthread_rwlock_unlock.c
+++ b/nptl/pthread_rwlock_unlock.c
@@ -35,6 +35,10 @@ __pthread_rwlock_unlock (pthread_rwlock_t *rwlock)
 
   LIBC_PROBE (rwlock_unlock, 1, rwlock);
 
+  /* Trying to elide an unlocked lock may crash the process.  This
+     is expected and is compatible with POSIX.1-2008: "results are
+     undefined if the read-write lock rwlock is not held by the
+     calling thread".  */
   if (ELIDE_UNLOCK (rwlock->__data.__writer == 0
 		    && rwlock->__data.__nr_readers == 0))
     return 0;
diff --git a/sysdeps/powerpc/nptl/elide.h b/sysdeps/powerpc/nptl/elide.h
index 77bd82e..c573981 100644
--- a/sysdeps/powerpc/nptl/elide.h
+++ b/sysdeps/powerpc/nptl/elide.h
@@ -102,6 +102,8 @@ __elide_unlock (int is_lock_free)
 {
   if (is_lock_free)
     {
+      /* Intentionally crashes when trying to unlock a lock not held by this
+	 thread.  */
       __libc_tend (0);
       return true;
     }
-- 
2.1.0

Re: [PATCH] Document a behavior of an elided pthread_rwlock_unlock

[Torvald Riegel]

On Fri, 2016-10-21 at 19:06 -0200, Tulio Magno Quites Machado Filho
wrote:
> Explain that pthread_rwlock_unlock may crash if called on a lock not
> held by the current thread.
> 
> 2016-10-21  Tulio Magno Quites Machado Filho  <tuliom@linux.vnet.ibm.com>
> 
> 	* nptl/pthread_rwlock_unlock.c: Add a comment explaining its
> 	behavior when eliding a lock not held by the current thread.
> 	* sysdeps/powerpc/nptl/elide.h: Likewise.
> ---
>  nptl/pthread_rwlock_unlock.c | 4 ++++
>  sysdeps/powerpc/nptl/elide.h | 2 ++
>  2 files changed, 6 insertions(+)
> 
> diff --git a/nptl/pthread_rwlock_unlock.c b/nptl/pthread_rwlock_unlock.c
> index a6cadd4..112f748 100644
> --- a/nptl/pthread_rwlock_unlock.c
> +++ b/nptl/pthread_rwlock_unlock.c
> @@ -35,6 +35,10 @@ __pthread_rwlock_unlock (pthread_rwlock_t *rwlock)
>  
>    LIBC_PROBE (rwlock_unlock, 1, rwlock);
>  
> +  /* Trying to elide an unlocked lock may crash the process.  This
> +     is expected and is compatible with POSIX.1-2008: "results are
> +     undefined if the read-write lock rwlock is not held by the
> +     calling thread".  */
>    if (ELIDE_UNLOCK (rwlock->__data.__writer == 0
>  		    && rwlock->__data.__nr_readers == 0))
>      return 0;
> diff --git a/sysdeps/powerpc/nptl/elide.h b/sysdeps/powerpc/nptl/elide.h
> index 77bd82e..c573981 100644
> --- a/sysdeps/powerpc/nptl/elide.h
> +++ b/sysdeps/powerpc/nptl/elide.h
> @@ -102,6 +102,8 @@ __elide_unlock (int is_lock_free)
>  {
>    if (is_lock_free)
>      {
> +      /* Intentionally crashes when trying to unlock a lock not held by this
> +	 thread.  */

I'd rather say that this can happen, and is okay (refering to the
comment in pthread_rwlock_unlock.c).  Otherwise, LGTM.

>        __libc_tend (0);
>        return true;
>      }

Re: [PATCH] Document a behavior of an elided pthread_rwlock_unlock

[Tulio Magno Quites Machado Filho]

Torvald Riegel <triegel@redhat.com> writes:

> On Fri, 2016-10-21 at 19:06 -0200, Tulio Magno Quites Machado Filho
> wrote:
>> diff --git a/sysdeps/powerpc/nptl/elide.h b/sysdeps/powerpc/nptl/elide.h
>> index 77bd82e..c573981 100644
>> --- a/sysdeps/powerpc/nptl/elide.h
>> +++ b/sysdeps/powerpc/nptl/elide.h
>> @@ -102,6 +102,8 @@ __elide_unlock (int is_lock_free)
>>  {
>>    if (is_lock_free)
>>      {
>> +      /* Intentionally crashes when trying to unlock a lock not held by this
>> +	 thread.  */
>
> I'd rather say that this can happen, and is okay (refering to the
> comment in pthread_rwlock_unlock.c).  Otherwise, LGTM.

Good point.  I'll make this change before pushing it.

Thanks!

-- 
Tulio Magno