Bug 1339412

Description of problem:

Upgrading my home server to Fedora 23 causing it to consume 100 % CPU after a while just running apache with php, gallery3 and an owncloud instance. Debugging into it shows all apache processes to hang at some mod_gnutls operation.

Version-Release number of selected component (if applicable):

Fedora 23 fully upt-to-date:
kernel-4.4.9-300.fc23.x86_64
httpd-2.4.18-1.fc23.x86_64
mod_gnutls-0.7-2.fc23.x86_64
gnutls-3.4.11-1.fc23.x86_64
gallery3-3.0.9-5.fc23.noarch
owncloud-8.2.4-1.fc23.noarch
php-5.6.21-1.fc23.x86_64

How reproducible:

restart httpd, wait a few minutes, do some https accesses and the first few apache processes will consume 100 % CPU

Steps to Reproduce:
1. sudo systemctl restart httpd
2. wget --no-check-certificate https://localhost
3. top

Actual results:
Tasks: 379 total,   4 running, 375 sleeping,   0 stopped,   0 zombie
%Cpu(s):  4.2 us,  0.1 sy,  0.0 ni, 95.7 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem : 98994728 total, 82720880 free,  1201508 used, 15072340 buff/cache
KiB Swap: 24780796 total, 24780796 free,        0 used. 97535424 avail Mem
 
  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
10109 apache    20   0  804756  25956  13136 R 100.0  0.0   0:09.43 httpd

Expected results:
apache should not consume 100 % CPU

Additional info:
[user@host ~]$ sudo pstack 10109
Thread 2 (Thread 0x7efef88c2700 (LWP 10110)):
#0  0x00007eff20abfb10 in pthread_cond_wait@@GLIBC_2.3.2 () from target:/lib64/libpthread.so.0
#1  0x00007eff19cbf07a in get_mplx_next () from target:/etc/httpd/modules/mod_http2.so
#2  0x00007eff19cbe8ac in execute () from target:/etc/httpd/modules/mod_http2.so
#3  0x00007eff20aba60a in start_thread () from target:/lib64/libpthread.so.0
#4  0x00007eff205f059d in clone () from target:/lib64/libc.so.6
Thread 1 (Thread 0x7eff21f508c0 (LWP 10109)):
#0  0x00007eff1093c0bb in _gnutls_recv_in_buffers () from target:/lib64/libgnutls.so.30
#1  0x00007eff1093da38 in _gnutls_recv_int () from target:/lib64/libgnutls.so.30
#2  0x00007eff1093df64 in gnutls_record_recv () from target:/lib64/libgnutls.so.30
#3  0x00007eff10c4af2a in gnutls_io_input_read () from target:/etc/httpd/modules/mod_gnutls.so
#4  0x00007eff10c4bb40 in mgs_filter_input () from target:/etc/httpd/modules/mod_gnutls.so
#5  0x0000559be5e5efc7 in ap_process_request_after_handler ()
#6  0x0000559be5e604b0 in ap_process_request ()
#7  0x0000559be5e5c59e in ap_process_http_connection ()
#8  0x0000559be5e53950 in ap_run_process_connection ()
#9  0x00007eff15964b21 in child_main () from target:/etc/httpd/modules/mod_mpm_prefork.so
#10 0x00007eff15964db3 in make_child () from target:/etc/httpd/modules/mod_mpm_prefork.so
#11 0x00007eff15964e2b in startup_children () from target:/etc/httpd/modules/mod_mpm_prefork.so
#12 0x00007eff15965e07 in prefork_run () from target:/etc/httpd/modules/mod_mpm_prefork.so
#13 0x0000559be5e2ddae in ap_run_mpm ()
#14 0x0000559be5e27437 in main ()

I found a report of a similar problem upstream unfortunately at the time of this writing their OpenID Login did not work for me:
https://mod.gnutls.org/ticket/136