httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Eissing <stefan.eiss...@greenbytes.de>
Subject Re: mod_http2 and Frequent wake-ups for mpm_event
Date Sun, 22 Jan 2017 16:17:02 GMT

> 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