[odb-users] Deadlock in apache module that uses odb-mysql?

Val Hendrix vchendrix at lbl.gov
Wed Apr 29 16:39:26 EDT 2015


Thanks for the speedy feedback!  I have successfully tested our apache
module using MariaDB 10.0.17.


Val Hendrix
vchendrix at lbl.gov
Lawrence Berkeley National Lab

Mail Stop: 50B-2239
Room:      50B-2258E
Phone:     (510) 495-2905
Skype:     val.hendrix

On Wed, Apr 29, 2015 at 9:37 AM, Boris Kolpackov <boris at codesynthesis.com>

> Hi Val,
> I did some digging and it appears something is going wrong with
> MySQL threading initialization when done from an apache module.
> The code in question (i.e., nptl_pthread_exit_hack_handler() in
> the stack trace) itself is dealing with (allegedly) a bug in
> NPTL. Here is the comment from my_thr_init.c:
>   /*
>     BUG#24507: Race conditions inside current NPTL pthread_exit()
>     implementation.
>     To avoid a possible segmentation fault during concurrent
>     executions of pthread_exit(), a dummy thread is spawned which
>     initializes internal variables of pthread lib. See bug description
>     for a full explanation.
>     TODO: Remove this code when fixed versions of glibc6 are in common
>     use.
>   */
> Also, generally, from our experience, MySQL multi-threading support
> is a huge mess. See the comment at the beginning of libodb-mysql/odb/
> mysql/connection-factory.cxx for the kind of hacks we have to resort
> to make it work.
> Note also that while we could reproduce this problem with 5.5.37-
> MariaDB, it seem to have been fixed in 10.0.17-MariaDB.
> Boris

More information about the odb-users mailing list