Topic: php-pecl-mongo(db) segfaults
On my CentOS 6 server, x86_64, I had php-pecl-mongo-1.6.14-1.el6.remi.5.6 installed, and then a couple of days ago, something, I guess maybe the php-horde-Horde-mongo-1.0.3-3.el6.remi.noarch update, required php-pecl-mongodb-1.1.7-2.el6.remi.5.6.x86_64 to be installed, and suddenly I started seeing these messages repeatedly in my /var/log/messages:
kernel: php[XXXXX]: segfault at XXXXXXXXXXXX ip 00007XXXXXXXXXXX sp 00007ffXXXXXXXXX error 14 in mongo.so[7fXXXXXXX000+5d000]
where the 'X's are apparently arbitrary digits (first is a process number, I suppose, the rest maybe memory locations?), [s]and the curly braces should be square brackets but your BBCode won't let me[/s] (and is giving me fits trying to figure out what it's counting as links even after cleaning that up).
Found that php-pecl-mongo owns both /usr/lib64/php/modules/mongo.so and /usr/lib64/php-zts/modules/mongo.so, so I removed it, and haven't seen those messages since. But php-pecl-mongo wasn't updated at the time things started going haywire, php-pecl-mongodb was installed, and some others updated as well. If it helps, here's the full list from that time:
(This seems to be what's triggering as links for phpBB, so I'll try to reproduce it in a follow-up.)
The first segfault:
Jun 29 06:47:36 rs4 kernel: php: segfault at 7f834b92bad0 ip 00007f834b92bad0 sp 00007fff9c879838 error 14 in mongo.so[7f834bd8a000+5d000]
…one second after the php-*mongo* installs.
FWIW, I never saw any negative effects outside of the logs, but I didn't open up every bit of the system myself, either. I could compare to the httpd logs and see what was being requested at the moments of segfaults (I'd assume something was).