httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Priebe - Profihost AG <s.pri...@profihost.ag>
Subject Re: mod_http2 and Frequent wake-ups for mpm_event
Date Sun, 22 Jan 2017 16:06:52 GMT
Looks like others have the same crashes too:
https://bz.apache.org/bugzilla/show_bug.cgi?id=60071
and
https://github.com/apache/httpd/commit/8e63c3c9372cd398f57357099aa941cbba695758

So it looks like mod_http2 is running fine now. Thanks a lot Stefan.

Yann i think i can start testing your mpm patch again after the
segfaults in 2.4 branch are fixed.

Greets,
Stefan

Am 22.01.2017 um 13:16 schrieb Stefan Priebe:
> Hi,
> 
> and a new one but also in ap_start_lingering_close:
> 
> Program terminated with signal SIGSEGV, Segmentation fault.
> #0  apr_palloc (pool=pool@entry=0x7f455805e138, in_size=in_size@entry=32)
>     at memory/unix/apr_pools.c:684
> #0  apr_palloc (pool=pool@entry=0x7f455805e138, in_size=in_size@entry=32)
>     at memory/unix/apr_pools.c:684
> #1  0x00007f456bc5d8b4 in apr_brigade_create (p=0x7f455805e138,
>     list=0x7f45040034e8) at buckets/apr_brigade.c:61
> #2  0x000055e165efa319 in ap_shutdown_conn (c=c@entry=0x7f455805e458,
>     flush=flush@entry=1) at connection.c:76
> #3  0x000055e165efa40d in ap_flush_conn (c=0x7f455805e458) at
> connection.c:95
> #4  ap_start_lingering_close (c=0x7f455805e458) at connection.c:145
> #5  0x000055e165f942dd in start_lingering_close_blocking (cs=<optimized
> out>)
>     at event.c:876
> #6  process_socket (my_thread_num=<optimized out>,
>     my_child_num=<optimized out>, cs=0x7f455805e3c8, sock=<optimized out>,
>     p=<optimized out>, thd=<optimized out>) at event.c:1153
> #7  worker_thread (thd=0x7f455805e138, dummy=0x20) at event.c:2001
> #8  0x00007f456b80a0a4 in start_thread ()
>    from /lib/x86_64-linux-gnu/libpthread.so.0
> #9  0x00007f456b53f62d in clone () from /lib/x86_64-linux-gnu/libc.so.6
> 
> Stefan
> 
> Am 21.01.2017 um 19:31 schrieb Stefan Priebe:
>> All last traces come from event, proces_longering_close ap_push_pool but
>> end in different functions. It looks like a race somewhere and it just
>> races at different function in the event of close and pool clear.
>>
>> Might there be two places where the same pool gets cleared?
>>
>> Stefan
>>
>> Am 21.01.2017 um 19:07 schrieb Stefan Priebe:
>>> Hi Stefan,
>>>
>>> thanks. No crashes where h2 comes up. But i still have these and no idea
>>> how to find out who and why they're crashing.
>>>
>>> Using host libthread_db library
>>> "/lib/x86_64-linux-gnu/libthread_db.so.1".
>>> Core was generated by `/usr/local/apache2/bin/httpd -k start'.
>>> Program terminated with signal SIGSEGV, Segmentation fault.
>>> #0  allocator_free (node=0x0, allocator=0x7f6e08066540)
>>>     at memory/unix/apr_pools.c:381
>>> #0  allocator_free (node=0x0, allocator=0x7f6e08066540)
>>>     at memory/unix/apr_pools.c:381
>>> #1  apr_pool_clear (pool=0x7f6e0808d238) at memory/unix/apr_pools.c:793
>>> #2  0x00000000004fe528 in ap_push_pool (queue_info=0x0,
>>>     pool_to_recycle=0x7f6e08066548) at fdqueue.c:234
>>> #3  0x00000000004fa2c8 in process_lingering_close (cs=0x7f6e0808d4c8,
>>>     pfd=0x1d3bf98) at event.c:1439
>>> #4  0x00000000004fd410 in listener_thread (thd=0x1d3cb70,
>>> dummy=0x7f6e0808d4c8)
>>>     at event.c:1704
>>> #5  0x00007f6e1aed20a4 in start_thread ()
>>>    from /lib/x86_64-linux-gnu/libpthread.so.0
>>> #6  0x00007f6e1aa0362d in clone () from /lib/x86_64-linux-gnu/libc.so.6
>>> (gdb) (gdb) quit
>>>
>>> Reading symbols from /usr/local/apache/bin/httpd...Reading symbols from
>>> /usr/lib/debug//usr/local/apache2/bin/httpd...done.
>>> done.
>>> [Thread debugging using libthread_db enabled]
>>> Using host libthread_db library
>>> "/lib/x86_64-linux-gnu/libthread_db.so.1".
>>> Core was generated by `/usr/local/apache2/bin/httpd -k start'.
>>> Program terminated with signal SIGSEGV, Segmentation fault.
>>> #0  allocator_free (node=0x0, allocator=0x7f6e08053ae0)
>>>     at memory/unix/apr_pools.c:381
>>> #0  allocator_free (node=0x0, allocator=0x7f6e08053ae0)
>>>     at memory/unix/apr_pools.c:381
>>> #1  apr_pool_clear (pool=0x7f6e08076bb8) at memory/unix/apr_pools.c:793
>>> #2  0x00000000004fe528 in ap_push_pool (queue_info=0x0,
>>>     pool_to_recycle=0x7f6e08053ae8) at fdqueue.c:234
>>> #3  0x00000000004fa2c8 in process_lingering_close (cs=0x7f6e08076e48,
>>>     pfd=0x1d3bf98) at event.c:1439
>>> #4  0x00000000004fd410 in listener_thread (thd=0x1d3cb70,
>>> dummy=0x7f6e08076e48)
>>>     at event.c:1704
>>> #5  0x00007f6e1aed20a4 in start_thread ()
>>>    from /lib/x86_64-linux-gnu/libpthread.so.0
>>> #6  0x00007f6e1aa0362d in clone () from /lib/x86_64-linux-gnu/libc.so.6
>>> (gdb) (gdb) quit
>>>
>>> Stefan
>>>
>>> Am 21.01.2017 um 17:03 schrieb Stefan Eissing:
>>>> Stefan,
>>>>
>>>> made a release at https://github.com/icing/mod_h2/releases/tag/v1.8.9
>>>> with all patches and improved (hopefully) on them a bit. If you dare
>>>> to drop that into your installation, that'd be great.
>>>>
>>>> Cheers,
>>>>
>>>> Stefan
>>>>
>>>>> Am 21.01.2017 um 15:25 schrieb Stefan Priebe <s.priebe@profihost.ag>:
>>>>>
>>>>> and i got another crash here:
>>>>>
>>>>> 2346 static void run_cleanups(cleanup_t **cref)
>>>>> 2347 {
>>>>> 2348     cleanup_t *c = *cref;
>>>>> 2349
>>>>> 2350     while (c) {
>>>>> 2351         *cref = c->next;
>>>>> 2352         (*c->plain_cleanup_fn)((void *)c->data);   <==
here
>>>>> 2353         c = *cref;
>>>>> 2354
>>>>>
>>>>> which looks similar to the other crash.
>>>>>
>>>>> #0  0x00007fe4bbd33e1b in run_cleanups (cref=<optimized out>) at
>>>>> memory/unix/apr_pools.c:2352
>>>>> #1  apr_pool_clear (pool=0x7fe4a804dac8) at
>>>>> memory/unix/apr_pools.c:772
>>>>> #2  0x00000000004feb38 in ap_push_pool
>>>>> (queue_info=0x6d616e79642d3733, pool_to_recycle=0x2) at fdqueue.c:234
>>>>> #3  0x00000000004fa8d8 in process_lingering_close (cs=0x7fe4a804dd58,
>>>>> pfd=0x25d3f98) at event.c:1439
>>>>>
>>>>> Details:
>>>>> (gdb) print c
>>>>> $1 = (cleanup_t *) 0x7fe4a804e9f0
>>>>> (gdb) print *c
>>>>> $2 = {next = 0x7fe4a804e870, data = 0x6d616e79642d3733,
>>>>> plain_cleanup_fn = 0x392d3734322e6369,
>>>>>  child_cleanup_fn = 0x617465722e722d35}
>>>>> (gdb) print *c->data
>>>>> Attempt to dereference a generic pointer.
>>>>> (gdb) print *c->plain_cleanup_fn
>>>>> Cannot access memory at address 0x392d3734322e6369
>>>>> (gdb)
>>>>>
>>>>> Stefan
>>>>>
>>>>> Am 21.01.2017 um 15:18 schrieb Stefan Priebe:
>>>>>> Hi,
>>>>>>
>>>>>> #0  apr_pool_cleanup_kill (p=0x7fe4a8072358,
>>>>>> data=data@entry=0x7fe4a80723e0,
>>>>>>    cleanup_fn=cleanup_fn@entry=0x7fe4bbd38a40 <socket_cleanup>)
at
>>>>>> memory/unix/apr_pools.c:2276
>>>>>>
>>>>>> it crashes here in apr:
>>>>>> 2276         if (c->data == data && c->plain_cleanup_fn
==
>>>>>> cleanup_fn) {
>>>>>>
>>>>>> some lines before c becomes this
>>>>>> 2264     c = p->cleanups;
>>>>>>
>>>>>> p is:
>>>>>> (gdb) print *p
>>>>>> $1 = {parent = 0x256f138, child = 0x7fe46c0751c8, sibling =
>>>>>> 0x7fe4a8096888, ref = 0x7fe4a8069fe8, cleanups = 0x7fe478159748,
>>>>>>  free_cleanups = 0x7fe478159788, allocator = 0x7fe4a803b490,
>>>>>> subprocesses = 0x0, abort_fn = 0x43da00 <abort_on_oom>,
>>>>>>  user_data = 0x0, tag = 0x502285 "transaction", active =
>>>>>> 0x7fe478158d70, self = 0x7fe4a8072330,
>>>>>>  self_first_avail = 0x7fe4a80723d0 "X#\a\250\344\177", pre_cleanups
=
>>>>>> 0x7fe4a8072ab8}
>>>>>>
>>>>>> wouldn't the error mean that p->cleanups is NULL?
>>>>>>
>>>>>> (gdb) print *p->cleanups
>>>>>> $2 = {next = 0x7fe478159628, data = 0x7fe478159648,
>>>>>> plain_cleanup_fn =
>>>>>> 0x7fe4bbd2ffd0 <apr_unix_file_cleanup>,
>>>>>>  child_cleanup_fn = 0x7fe4bbd2ff70 <apr_unix_child_file_cleanup>}
>>>>>>
>>>>>> So p->cleanups->data is 0x7fe478159648 and data is 0x7fe4a80723e0?
>>>>>>
>>>>>> I don't get why it's segfaulting
>>>>>>
>>>>>> Stefan
>>>>>> Am 21.01.2017 um 09:50 schrieb Yann Ylavic:
>>>>>>> Hi Stefan,
>>>>>>>
>>>>>>> On Sat, Jan 21, 2017 at 9:45 AM, Stefan Priebe
>>>>>>> <s.priebe@profihost.ag>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>> after running the whole night. These are the only ones still
>>>>>>>> happening.
>>>>>>>> Should i revert the mpm patch to check whether it's the source?
>>>>>>>
>>>>>>> Yes please, we need to determine...
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Yann.
>>>>>>>
>>>>
>>>> Stefan Eissing
>>>>
>>>> <green/>bytes GmbH
>>>> Hafenstrasse 16
>>>> 48155 M√ľnster
>>>> www.greenbytes.de
>>>>

Mime
View raw message