[PATCH] misc/tst-clone3: Fix waiting for exited thread.

classic Classic list List threaded Threaded
8 messages Options
Reply | Threaded
Open this post in threaded view
|

[PATCH] misc/tst-clone3: Fix waiting for exited thread.

Stefan Liebler-2
Hi,

from time to time the test misc/tst-clone3 fails with an timeout.  Then
futex_wait is blocking.  Usually ctid should be set to zero due to
CLONE_CHILD_CLEARTID and the futex should be waken up.  But the fail
occures if the thread has already exited before ctid is set to the
return value of clone().  Then futex_wait() will block as there will be
nobody who wakes the futex up again.

This patch initializes ctid to a known value before calling clone and
the kernel is the only one who updates the value to zero after clone.
If futex_wait is called then it is either waked up due to the exited
thread or the futex syscall fails as *ctid_ptr is already zero instead
of the specified value 1.

Okay to commit?

Bye
Stefan

ChangeLog:

        * sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
        Initialize ctid with a known value and remove update of ctid
        after clone.
        (wait_tid): Adjust arguments and call futex_wait with ctid_val
        as assumed current value of ctid_ptr.

20190208_misc_tst-clone3.patch (2K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Adhemerval Zanella-2


On 08/02/2019 13:12, Stefan Liebler wrote:

> Hi,
>
> from time to time the test misc/tst-clone3 fails with an timeout.  Then futex_wait is blocking.  Usually ctid should be set to zero due to CLONE_CHILD_CLEARTID and the futex should be waken up.  But the fail occures if the thread has already exited before ctid is set to the return value of clone().  Then futex_wait() will block as there will be nobody who wakes the futex up again.
>
> This patch initializes ctid to a known value before calling clone and the kernel is the only one who updates the value to zero after clone. If futex_wait is called then it is either waked up due to the exited thread or the futex syscall fails as *ctid_ptr is already zero instead of the specified value 1.
>
> Okay to commit?
>
> Bye
> Stefan
>
> ChangeLog:
>
>     * sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
>     Initialize ctid with a known value and remove update of ctid
>     after clone.
>     (wait_tid): Adjust arguments and call futex_wait with ctid_val
>     as assumed current value of ctid_ptr.
>
> 20190208_misc_tst-clone3.patch
>
> commit 5a8f80973dbbf06a0eebc2064d2a14521c6a6131
> Author: Stefan Liebler <[hidden email]>
> Date:   Fri Feb 8 12:42:52 2019 +0100
>
>     misc/tst-clone3: Fix waiting for exited thread.
>    
>     From time to time the test misc/tst-clone3 fails with an timeout.
>     Then futex_wait is blocking.  Usually ctid should be set to zero
>     due to CLONE_CHILD_CLEARTID and the futex should be waken up.
>     But the fail occures if the thread has already exited before
>     ctid is set to the return value of clone().  Then futex_wait() will
>     block as there will be nobody who wakes the futex up again.
>    
>     This patch initializes ctid to a known value before calling clone
>     and the kernel is the only one who updates the value to zero after clone.
>     If futex_wait is called then it is either waked up due to the exited thread
>     or the futex syscall fails as *ctid_ptr is already zero instead of the
>     specified value 1.
>    
>     ChangeLog:
>    
>             * sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
>             Initialize ctid with a known value and remove update of ctid
>             after clone.
>             (wait_tid): Adjust arguments and call futex_wait with ctid_val
>             as assumed current value of ctid_ptr.

Thanks for catching it.

>
> diff --git a/sysdeps/unix/sysv/linux/tst-clone3.c b/sysdeps/unix/sysv/linux/tst-clone3.c
> index aa8e718afe..ffa2056eb6 100644
> --- a/sysdeps/unix/sysv/linux/tst-clone3.c
> +++ b/sysdeps/unix/sysv/linux/tst-clone3.c
> @@ -42,11 +42,11 @@ f (void *a)
>  
>  /* Futex wait for TID argument, similar to pthread_join internal
>     implementation.  */
> -#define wait_tid(tid) \
> +#define wait_tid(ctid_ptr, ctid_val) \
>    do { \
> -    __typeof (tid) __tid; \
> -    while ((__tid = (tid)) != 0) \
> -      futex_wait (&(tid), __tid); \
> +    __typeof (*(ctid_ptr)) __tid; \
> +    while ((__tid = *(ctid_ptr)) != 0) \
> +      futex_wait (ctid_ptr, ctid_val); \
>    } while (0)

lll_wait_tid uses an atomic load with acquire semantic, I think we should
use it as well. We can either include the atomic header or use c11 atomic.

>  
>  static inline int
> @@ -64,7 +64,11 @@ do_test (void)
>    clone_flags |= CLONE_VM | CLONE_SIGHAND;
>    /* We will used ctid to call on futex to wait for thread exit.  */
>    clone_flags |= CLONE_CHILD_CLEARTID;
> -  pid_t ctid, tid;
> +  /* Initialize with a known value.  ctid is set to zero by the kernel after the
> +     cloned thread has exited.  */
> +#define CTID_INIT_VAL 1
> +  pid_t ctid = CTID_INIT_VAL;
> +  pid_t tid;
>  
>  #ifdef __ia64__
>    extern int __clone2 (int (*__fn) (void *__arg), void *__child_stack_base,
> @@ -86,8 +90,7 @@ do_test (void)
>    if (tid == -1)
>      FAIL_EXIT1 ("clone failed: %m");
>  
> -  ctid = tid;
> -  wait_tid (ctid);
> +  wait_tid (&ctid, CTID_INIT_VAL);
>  
>    return 2;
>  }
>
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Stefan Liebler-2
Attached the updated patch.
See also notes below.

On 02/08/2019 07:37 PM, Adhemerval Zanella wrote:

>
>
> On 08/02/2019 13:12, Stefan Liebler wrote:
>> Hi,
>>
>> from time to time the test misc/tst-clone3 fails with an timeout.  Then futex_wait is blocking.  Usually ctid should be set to zero due to CLONE_CHILD_CLEARTID and the futex should be waken up.  But the fail occures if the thread has already exited before ctid is set to the return value of clone().  Then futex_wait() will block as there will be nobody who wakes the futex up again.
>>
>> This patch initializes ctid to a known value before calling clone and the kernel is the only one who updates the value to zero after clone. If futex_wait is called then it is either waked up due to the exited thread or the futex syscall fails as *ctid_ptr is already zero instead of the specified value 1.
>>
>> Okay to commit?
>>
>> Bye
>> Stefan
>>
>> ChangeLog:
>>
>>      * sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
>>      Initialize ctid with a known value and remove update of ctid
>>      after clone.
>>      (wait_tid): Adjust arguments and call futex_wait with ctid_val
>>      as assumed current value of ctid_ptr.
>>
>> 20190208_misc_tst-clone3.patch
>>
>> commit 5a8f80973dbbf06a0eebc2064d2a14521c6a6131
>> Author: Stefan Liebler <[hidden email]>
>> Date:   Fri Feb 8 12:42:52 2019 +0100
>>
>>      misc/tst-clone3: Fix waiting for exited thread.
>>      
>>      From time to time the test misc/tst-clone3 fails with an timeout.
>>      Then futex_wait is blocking.  Usually ctid should be set to zero
>>      due to CLONE_CHILD_CLEARTID and the futex should be waken up.
>>      But the fail occures if the thread has already exited before
>>      ctid is set to the return value of clone().  Then futex_wait() will
>>      block as there will be nobody who wakes the futex up again.
>>      
>>      This patch initializes ctid to a known value before calling clone
>>      and the kernel is the only one who updates the value to zero after clone.
>>      If futex_wait is called then it is either waked up due to the exited thread
>>      or the futex syscall fails as *ctid_ptr is already zero instead of the
>>      specified value 1.
>>      
>>      ChangeLog:
>>      
>>              * sysdeps/unix/sysv/linux/tst-clone3.c (do_test):
>>              Initialize ctid with a known value and remove update of ctid
>>              after clone.
>>              (wait_tid): Adjust arguments and call futex_wait with ctid_val
>>              as assumed current value of ctid_ptr.
>
> Thanks for catching it.
>
>>
>> diff --git a/sysdeps/unix/sysv/linux/tst-clone3.c b/sysdeps/unix/sysv/linux/tst-clone3.c
>> index aa8e718afe..ffa2056eb6 100644
>> --- a/sysdeps/unix/sysv/linux/tst-clone3.c
>> +++ b/sysdeps/unix/sysv/linux/tst-clone3.c
>> @@ -42,11 +42,11 @@ f (void *a)
>>  
>>   /* Futex wait for TID argument, similar to pthread_join internal
>>      implementation.  */
>> -#define wait_tid(tid) \
>> +#define wait_tid(ctid_ptr, ctid_val) \
>>     do { \
>> -    __typeof (tid) __tid; \
>> -    while ((__tid = (tid)) != 0) \
>> -      futex_wait (&(tid), __tid); \
>> +    __typeof (*(ctid_ptr)) __tid; \
>> +    while ((__tid = *(ctid_ptr)) != 0) \
>> +      futex_wait (ctid_ptr, ctid_val); \
>>     } while (0)
>
> lll_wait_tid uses an atomic load with acquire semantic, I think we should
> use it as well. We can either include the atomic header or use c11 atomic.
>
I've first tried to include atomic.h, but it failed building on x86_64.
Thus I'm using the c11 atomic load in the updated patch.
Okay to commit?


As information, I've observed those gcc errors on x86_64:
In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
                  from ../sysdeps/x86_64/nptl/tls.h:28,
                  from ../sysdeps/x86/atomic-machine.h:23,
                  from ../include/atomic.h:50,
                  from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
‘_dl_discover_osversion’:
../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
declaration specifiers before ‘attribute_hidden’
  extern int _dl_discover_osversion (void) attribute_hidden;
                                           ^~~~~~~~~~~~~~~~
In file included from ../sysdeps/x86_64/nptl/tls.h:31,
                  from ../sysdeps/x86/atomic-machine.h:23,
                  from ../include/atomic.h:50,
                  from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
../sysdeps/generic/dl-dtv.h:22:1: error: empty declaration [-Werror]
  struct dtv_pointer
  ^~~~~~
../sysdeps/generic/dl-dtv.h:33:3: error: storage class specified for
parameter ‘dtv_t’
  } dtv_t;
    ^~~~~
...
many many more errors
...

I've also tried to add tst-clone3 to tests_internal instead of tests in
the Makefile, but then I've got:
In file included from ../sysdeps/x86_64/nptl/tls.h:130,
                  from ../sysdeps/x86/atomic-machine.h:23,
                  from ../include/atomic.h:50,
                  from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
../nptl/descr.h:104:8: error: redefinition of ‘struct robust_list_head’
  struct robust_list_head
         ^~~~~~~~~~~~~~~~
In file included from ../sysdeps/unix/sysv/linux/tst-clone3.c:26:
/usr/include/linux/futex.h:70:8: note: originally defined here
  struct robust_list_head {
         ^~~~~~~~~~~~~~~~

>>  
>>   static inline int
>> @@ -64,7 +64,11 @@ do_test (void)
>>     clone_flags |= CLONE_VM | CLONE_SIGHAND;
>>     /* We will used ctid to call on futex to wait for thread exit.  */
>>     clone_flags |= CLONE_CHILD_CLEARTID;
>> -  pid_t ctid, tid;
>> +  /* Initialize with a known value.  ctid is set to zero by the kernel after the
>> +     cloned thread has exited.  */
>> +#define CTID_INIT_VAL 1
>> +  pid_t ctid = CTID_INIT_VAL;
>> +  pid_t tid;
>>  
>>   #ifdef __ia64__
>>     extern int __clone2 (int (*__fn) (void *__arg), void *__child_stack_base,
>> @@ -86,8 +90,7 @@ do_test (void)
>>     if (tid == -1)
>>       FAIL_EXIT1 ("clone failed: %m");
>>  
>> -  ctid = tid;
>> -  wait_tid (ctid);
>> +  wait_tid (&ctid, CTID_INIT_VAL);
>>  
>>     return 2;
>>   }
>>
>


20190211_1030_misc_tst-clone3.patch (2K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Florian Weimer-5
* Stefan Liebler:

> I've first tried to include atomic.h, but it failed building on
> x86_64. Thus I'm using the c11 atomic load in the updated patch.
> Okay to commit?
>
>
> As information, I've observed those gcc errors on x86_64:
> In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
>                  from ../sysdeps/x86_64/nptl/tls.h:28,
>                  from ../sysdeps/x86/atomic-machine.h:23,
>                  from ../include/atomic.h:50,
>                  from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
> ../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
> ‘_dl_discover_osversion’:
> ../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
> declaration specifiers before ‘attribute_hidden’
>  extern int _dl_discover_osversion (void) attribute_hidden;
>                                           ^~~~~~~~~~~~~~~~

That's because the test isn't in tests-internal.

> +    while ((__tid = __atomic_load_n (ctid_ptr, __ATOMIC_ACQUIRE)) != 0) \

Actually, that's not a C11 atomic construct, but I think it's okay to
use that here.  (The C11 stuff lives in <stdatomic.h> and should be
functionally equivalent.)

Sorry, this is a pet peeve of mine.  We have three different atomic
access facilities that people refer to as C11 atomics: Our own
<atomic.h>, the GCC __atomic builtins, and <stdatomic.h>.

I still think this contributes to cognitive load, and we should
eliminate all but one (leaving us with new-style atomics and the old
macros in <atomic.h>).  The GCC __atomic builtins have the best freely
available documentation, so they are a natural candidate IMHO.

Thanks,
Florian
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Adhemerval Zanella-2


On 11/02/2019 08:59, Florian Weimer wrote:

> * Stefan Liebler:
>
>> I've first tried to include atomic.h, but it failed building on
>> x86_64. Thus I'm using the c11 atomic load in the updated patch.
>> Okay to commit?
>>
>>
>> As information, I've observed those gcc errors on x86_64:
>> In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
>>                  from ../sysdeps/x86_64/nptl/tls.h:28,
>>                  from ../sysdeps/x86/atomic-machine.h:23,
>>                  from ../include/atomic.h:50,
>>                  from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
>> ../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
>> ‘_dl_discover_osversion’:
>> ../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
>> declaration specifiers before ‘attribute_hidden’
>>  extern int _dl_discover_osversion (void) attribute_hidden;
>>                                           ^~~~~~~~~~~~~~~~
>
> That's because the test isn't in tests-internal.
>
>> +    while ((__tid = __atomic_load_n (ctid_ptr, __ATOMIC_ACQUIRE)) != 0) \
>
> Actually, that's not a C11 atomic construct, but I think it's okay to
> use that here.  (The C11 stuff lives in <stdatomic.h> and should be
> functionally equivalent.)
>
> Sorry, this is a pet peeve of mine.  We have three different atomic
> access facilities that people refer to as C11 atomics: Our own
> <atomic.h>, the GCC __atomic builtins, and <stdatomic.h>.
>
> I still think this contributes to cognitive load, and we should
> eliminate all but one (leaving us with new-style atomics and the old
> macros in <atomic.h>).  The GCC __atomic builtins have the best freely
> available documentation, so they are a natural candidate IMHO.

It really annoying that C11 standard is not freely available from ISO,
although the working draft is still open [1]. I don't have a strong
opinion, but since there is support in the language itself and they
are fully supported by the compiler I also don't see why not prefer it.

[1] http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Stefan Liebler-2
On 02/11/2019 01:11 PM, Adhemerval Zanella wrote:

>
>
> On 11/02/2019 08:59, Florian Weimer wrote:
>> * Stefan Liebler:
>>
>>> I've first tried to include atomic.h, but it failed building on
>>> x86_64. Thus I'm using the c11 atomic load in the updated patch.
>>> Okay to commit?
>>>
>>>
>>> As information, I've observed those gcc errors on x86_64:
>>> In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
>>>                   from ../sysdeps/x86_64/nptl/tls.h:28,
>>>                   from ../sysdeps/x86/atomic-machine.h:23,
>>>                   from ../include/atomic.h:50,
>>>                   from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
>>> ‘_dl_discover_osversion’:
>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
>>> declaration specifiers before ‘attribute_hidden’
>>>   extern int _dl_discover_osversion (void) attribute_hidden;
>>>                                            ^~~~~~~~~~~~~~~~
>>
>> That's because the test isn't in tests-internal.
>>
>>> +    while ((__tid = __atomic_load_n (ctid_ptr, __ATOMIC_ACQUIRE)) != 0) \
>>
>> Actually, that's not a C11 atomic construct, but I think it's okay to
>> use that here.  (The C11 stuff lives in <stdatomic.h> and should be
>> functionally equivalent.)
>>
>> Sorry, this is a pet peeve of mine.  We have three different atomic
>> access facilities that people refer to as C11 atomics: Our own
>> <atomic.h>, the GCC __atomic builtins, and <stdatomic.h>.
>>
>> I still think this contributes to cognitive load, and we should
>> eliminate all but one (leaving us with new-style atomics and the old
>> macros in <atomic.h>).  The GCC __atomic builtins have the best freely
>> available documentation, so they are a natural candidate IMHO.
>
> It really annoying that C11 standard is not freely available from ISO,
> although the working draft is still open [1]. I don't have a strong
> opinion, but since there is support in the language itself and they
> are fully supported by the compiler I also don't see why not prefer it.
>
> [1] http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
>
Then I have a further update of the patch which uses stdatomic.h and
atomic_load_explicit.

20190212_1245_misc_tst-clone3.patch (3K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Stefan Liebler-2
On 02/12/2019 01:53 PM, Stefan Liebler wrote:

> On 02/11/2019 01:11 PM, Adhemerval Zanella wrote:
>>
>>
>> On 11/02/2019 08:59, Florian Weimer wrote:
>>> * Stefan Liebler:
>>>
>>>> I've first tried to include atomic.h, but it failed building on
>>>> x86_64. Thus I'm using the c11 atomic load in the updated patch.
>>>> Okay to commit?
>>>>
>>>>
>>>> As information, I've observed those gcc errors on x86_64:
>>>> In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
>>>>                   from ../sysdeps/x86_64/nptl/tls.h:28,
>>>>                   from ../sysdeps/x86/atomic-machine.h:23,
>>>>                   from ../include/atomic.h:50,
>>>>                   from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
>>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
>>>> ‘_dl_discover_osversion’:
>>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
>>>> declaration specifiers before ‘attribute_hidden’
>>>>   extern int _dl_discover_osversion (void) attribute_hidden;
>>>>                                            ^~~~~~~~~~~~~~~~
>>>
>>> That's because the test isn't in tests-internal.
>>>
>>>> +    while ((__tid = __atomic_load_n (ctid_ptr, __ATOMIC_ACQUIRE))
>>>> != 0)    \
>>>
>>> Actually, that's not a C11 atomic construct, but I think it's okay to
>>> use that here.  (The C11 stuff lives in <stdatomic.h> and should be
>>> functionally equivalent.)
>>>
>>> Sorry, this is a pet peeve of mine.  We have three different atomic
>>> access facilities that people refer to as C11 atomics: Our own
>>> <atomic.h>, the GCC __atomic builtins, and <stdatomic.h>.
>>>
>>> I still think this contributes to cognitive load, and we should
>>> eliminate all but one (leaving us with new-style atomics and the old
>>> macros in <atomic.h>).  The GCC __atomic builtins have the best freely
>>> available documentation, so they are a natural candidate IMHO.
>>
>> It really annoying that C11 standard is not freely available from ISO,
>> although the working draft is still open [1]. I don't have a strong
>> opinion, but since there is support in the language itself and they
>> are fully supported by the compiler I also don't see why not prefer it.
>>
>> [1] http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
>>
> Then I have a further update of the patch which uses stdatomic.h and
> atomic_load_explicit.


ping
okay to commit?

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] misc/tst-clone3: Fix waiting for exited thread.

Florian Weimer-5
* Stefan Liebler:

> On 02/12/2019 01:53 PM, Stefan Liebler wrote:
>> On 02/11/2019 01:11 PM, Adhemerval Zanella wrote:
>>>
>>>
>>> On 11/02/2019 08:59, Florian Weimer wrote:
>>>> * Stefan Liebler:
>>>>
>>>>> I've first tried to include atomic.h, but it failed building on
>>>>> x86_64. Thus I'm using the c11 atomic load in the updated patch.
>>>>> Okay to commit?
>>>>>
>>>>>
>>>>> As information, I've observed those gcc errors on x86_64:
>>>>> In file included from ../sysdeps/unix/sysv/linux/x86_64/sysdep.h:30,
>>>>>                   from ../sysdeps/x86_64/nptl/tls.h:28,
>>>>>                   from ../sysdeps/x86/atomic-machine.h:23,
>>>>>                   from ../include/atomic.h:50,
>>>>>                   from ../sysdeps/unix/sysv/linux/tst-clone3.c:29:
>>>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h: In function
>>>>> ‘_dl_discover_osversion’:
>>>>> ../sysdeps/unix/sysv/linux/dl-sysdep.h:31:42: error: expected
>>>>> declaration specifiers before ‘attribute_hidden’
>>>>>   extern int _dl_discover_osversion (void) attribute_hidden;
>>>>>                                            ^~~~~~~~~~~~~~~~
>>>>
>>>> That's because the test isn't in tests-internal.
>>>>
>>>>> +    while ((__tid = __atomic_load_n (ctid_ptr,
>>>>> __ATOMIC_ACQUIRE)) != 0)    \
>>>>
>>>> Actually, that's not a C11 atomic construct, but I think it's okay to
>>>> use that here.  (The C11 stuff lives in <stdatomic.h> and should be
>>>> functionally equivalent.)
>>>>
>>>> Sorry, this is a pet peeve of mine.  We have three different atomic
>>>> access facilities that people refer to as C11 atomics: Our own
>>>> <atomic.h>, the GCC __atomic builtins, and <stdatomic.h>.
>>>>
>>>> I still think this contributes to cognitive load, and we should
>>>> eliminate all but one (leaving us with new-style atomics and the old
>>>> macros in <atomic.h>).  The GCC __atomic builtins have the best freely
>>>> available documentation, so they are a natural candidate IMHO.
>>>
>>> It really annoying that C11 standard is not freely available from ISO,
>>> although the working draft is still open [1]. I don't have a strong
>>> opinion, but since there is support in the language itself and they
>>> are fully supported by the compiler I also don't see why not prefer it.
>>>
>>> [1] http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
>>>
>> Then I have a further update of the patch which uses stdatomic.h and
>> atomic_load_explicit.
>
>
> ping
> okay to commit?

I think so.  Patch looks reasonable to me.  Thanks.

Florian