Fwd: NPTL Thread lockup issue on ARMv7

Nitin Garg nitingarg98 at gmail.com
Tue Feb 1 16:27:13 UTC 2011


I think we should remove atomic.h file from
libpthread/nptl/sysdeps/unix/sysv/linux/arm/bits/ and add it to
libc/sysdeps/linux/arm/bits/ (as submitted in the patch).

This will be similar to other architectures too.

Regards,
Nitin

On Tue, Feb 1, 2011 at 10:21 AM, Khem Raj <raj.khem at gmail.com> wrote:
> On Tue, Feb 1, 2011 at 8:17 AM, Carmelo AMOROSO <carmelo.amoroso at st.com> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 2/1/2011 5:14 PM, Nitin Garg wrote:
>>> Hope I created the patch correctly. Pls let me know if this is
>>> incorrect, otherwise pls add it to repository.
>>>
>>> Thanks,
>>> Nitin
>>>
>>
>> Nitin,
>> according to the comment by Thomas, I've understood that we already have
>> a good implementation in the nptl path, it should be just the case to
>> fix the buildsys to properly pick-up the right one.
>>
>> I will check... Khem ?
>
> yes thats fine too it will fix nptl I thought moving it to common arm
> would make it LT use it too but probably fixing just for nptl it
> better
>>
>> Carmelo
>>
>>> On Tue, Feb 1, 2011 at 1:13 AM, Khem Raj <raj.khem at gmail.com> wrote:
>>>  > On (01/02/11 00:32), Nitin Garg wrote:
>>>  >> Recently we came accross an issue on ARMv7 processors while running
>>>  >> multiple threads. If say 3 threads are running continuously, 1 or 2
>>>  >> might get locked somewhere. If we attach gdb and run again, all 3
>>>  >> threads start executing for a while and after some time 1 or 2 threads
>>>  >> gets locked-up again. The backtrace showed the threads are stuck in
>>>  >> __pthread_mutex_lock (atomic_compare_and_exchange_val_acq).
>>>  >>
>>>  >> We noticed that the atomic compare and exchange functions are not
>>>  >> atomic for ARM. Once we added the Atomic compare and exchange
>>>  >> function, the problem was resolved.
>>>  >>
>>>  >> Pls see below patch for review and kindly add it to next release.
>>>  >
>>>  > This looks ok to me but you have to send a properly formatted and
>>>  > signed-off patch so it can be tested and included. Using git format-patch
>>>  > and git send-email is your best bet
>>>  >
>>>  > -Khem
>>>  >
>>>
>>
>> -----BEGIN PGP SIGNATURE-----
>> Version: GnuPG v1.4.10 (MingW32)
>> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>>
>> iEYEARECAAYFAk1IMi8ACgkQoRq/3BrK1s+NfgCdHXAXGJO4C/zGXQ2QuaNgMGfn
>> xzAAoLBQaoVqxQu5idKTxE/BxHDP4Tf+
>> =SYRu
>> -----END PGP SIGNATURE-----
>> _______________________________________________
>> uClibc mailing list
>> uClibc at uclibc.org
>> http://lists.busybox.net/mailman/listinfo/uclibc
>>
> _______________________________________________
> uClibc mailing list
> uClibc at uclibc.org
> http://lists.busybox.net/mailman/listinfo/uclibc


More information about the uClibc mailing list