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 16:21:18 GMT
Hi Stefan,

no i was mistaken customer isn't using mod_proxy - but I think this is
the patch causing me problems:
https://github.com/apache/httpd/commit/a61a4bd02e483fb45d433343740d0130ee3d8d5d

What do you think?

Greets,
Stefan

Am 22.01.2017 um 17:17 schrieb Stefan Eissing:
> 
>> Am 22.01.2017 um 17:14 schrieb Stefan Priebe - Profihost AG <s.priebe@profihost.ag>:
>>
>> *arg* it's just mod_proxy - just saw thread safety and apr bucket aloc.
> 
> ??? Can you elaborate? Is your finding the known hcheck bug or something else?
> 
>> Stefan
>>
>> Am 22.01.2017 um 17:06 schrieb Stefan Priebe - Profihost AG:
>>> 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 Münster
>>>>>>> www.greenbytes.de
>>>>>>>
> 
> Stefan Eissing
> 
> <green/>bytes GmbH
> Hafenstrasse 16
> 48155 Münster
> www.greenbytes.de
> 

Mime
View raw message