*arg* it's just mod_proxy - just saw thread safety and apr bucket aloc. 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.pri...@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.pri...@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 >>>>>
