When browsing databases with phpPgAdmin, httpd randomly crashes, with : [core:notice] [pid 778] AH00052: child pid 803 exit signal Segmentation fault (11) Using mpm_prefork or mpm_itk, the result is the same. It happened on random pages, after clicking around. Mostly coming back to a page already visited. After some fiddling, I managed to get a core of httpd and a backtrace : (gdb) bt #0 0x000000081f15c000 in ?? () #1 0x000000080a20a08d in accel_shutdown () from /usr/local/lib/php/20131226/opcache.so #2 0x000000080a206efb in validate_timestamp_and_record () from /usr/local/lib/php/20131226/opcache.so #3 0x000000080a207d14 in persistent_compile_file () from /usr/local/lib/php/20131226/opcache.so #4 0x0000000808841772 in zend_execute_scripts () from /usr/local/libexec/apache24/libphp5.so #5 0x00000008087e39d9 in php_execute_script () from /usr/local/libexec/apache24/libphp5.so #6 0x00000008088cd5bb in zend_get_zval_ptr_ptr () from /usr/local/libexec/apache24/libphp5.so #7 0x00000000004449f6 in ap_pass_brigade (next=0x81f15bcb0, bb=0x81f159f80) at util_filter.c:590 #8 0x000000000045c053 in default_handler (r=0x81f1580a0) at core.c:4746 #9 0x000000000046b4b2 in ap_run_handler (r=0x81f1580a0) at config.c:170 #10 0x000000000046bd0c in ap_invoke_handler (r=0x81f1580a0) at config.c:434 #11 0x000000000048fcbc in ap_process_async_request (r=0x81f1580a0) at http_request.c:436 #12 0x000000000048fd91 in ap_process_request (r=0x81f1580a0) at http_request.c:471 #13 0x000000000048b0d0 in ap_process_http_sync_connection (c=0x81aef5290) at http_core.c:210 #14 0x000000000048ac6d in ap_process_http_connection (c=0x81aef5290) at http_core.c:251 #15 0x000000000047c122 in ap_run_process_connection (c=0x81aef5290) at connection.c:42 #16 0x0000000809c91c64 in itk_fork_process (c=0x81aef5290) at mpm_itk.c:220 #17 0x000000000047c122 in ap_run_process_connection (c=0x81aef5290) at connection.c:42 #18 0x000000000047c905 in ap_process_connection (c=0x81aef5290, csd=0x81aef50a0) at connection.c:226 #19 0x000000080607222a in child_main (child_num_arg=0, child_bucket=0) at prefork.c:612 #20 0x00000008060712a3 in make_child (s=0x802454110, slot=0, bucket=0) at prefork.c:706 #21 0x000000080607136d in startup_children (number_to_start=5) at prefork.c:725 #22 0x0000000806070288 in prefork_run (_pconf=0x802429028, plog=0x802458028, s=0x802454110) at prefork.c:894 #23 0x0000000000441402 in ap_run_mpm (pconf=0x802429028, plog=0x802458028, s=0x802454110) at mpm_common.c:94 #24 0x0000000000434f30 in main (argc=1, argv=0x7fffffffed28) at main.c:792 It pointed to opcache.so, so in php.ini I put opcache.enable=0 Problem disappeared. I would like to get the opcache back, eventually.
Added debug symbols to opcache. I narrowed it down to the schemas.php page : (gdb) bt #0 0x000000081f35c000 in ?? () #1 0x000000080a40a772 in zend_get_file_handle_timestamp (file_handle=0x7fffffffe1a8, size=0x0) at ZendAccelerator.c:830 #2 0x000000080a4092b1 in compile_and_cache_file (file_handle=0x7fffffffe1a8, type=8, key=0x80a636d54 "/usr/local/www/phpPgAdmin/schemas.php", key_length=37, op_array_p=0x7fffffffc690, from_shared_memory=0x7fffffffc6b0) at ZendAccelerator.c:1367 #3 0x000000080a408910 in persistent_compile_file (file_handle=0x7fffffffe1a8, type=8) at ZendAccelerator.c:1680 #4 0x000000080914cfa0 in zend_execute_scripts (type=8, retval=0x0, file_count=3) at zend.c:1333 #5 0x00000008090ad1c4 in php_execute_script (primary_file=0x7fffffffe1a8) at main.c:2613 #6 0x0000000809292011 in php_output_filter (f=0x81f35bcb0, bb=0x81f359f80) at sapi_apache2.c:548 #7 0x00000000004449f6 in ap_pass_brigade (next=0x81f35bcb0, bb=0x81f359f80) at util_filter.c:590 #8 0x000000000045c053 in default_handler (r=0x81f3580a0) at core.c:4746 #9 0x000000000046b4b2 in ap_run_handler (r=0x81f3580a0) at config.c:170 #10 0x000000000046bd0c in ap_invoke_handler (r=0x81f3580a0) at config.c:434 #11 0x000000000048fcbc in ap_process_async_request (r=0x81f3580a0) at http_request.c:436 #12 0x000000000048fd91 in ap_process_request (r=0x81f3580a0) at http_request.c:471 #13 0x000000000048b0d0 in ap_process_http_sync_connection (c=0x81b0f5290) at http_core.c:210 #14 0x000000000048ac6d in ap_process_http_connection (c=0x81b0f5290) at http_core.c:251 #15 0x000000000047c122 in ap_run_process_connection (c=0x81b0f5290) at connection.c:42 #16 0x0000000809e35c64 in itk_fork_process (c=0x81b0f5290) at mpm_itk.c:220 #17 0x000000000047c122 in ap_run_process_connection (c=0x81b0f5290) at connection.c:42 #18 0x000000000047c905 in ap_process_connection (c=0x81b0f5290, csd=0x81b0f50a0) at connection.c:226 #19 0x000000080607222a in child_main (child_num_arg=2, child_bucket=0) at prefork.c:612 #20 0x00000008060712a3 in make_child (s=0x802454110, slot=2, bucket=0) at prefork.c:706 #21 0x000000080607136d in startup_children (number_to_start=3) at prefork.c:725 #22 0x0000000806070288 in prefork_run (_pconf=0x802429028, plog=0x802458028, s=0x802454110) at prefork.c:894 #23 0x0000000000441402 in ap_run_mpm (pconf=0x802429028, plog=0x802458028, s=0x802454110) at mpm_common.c:94 #24 0x0000000000434f30 in main (argc=1, argv=0x7fffffffed28) at main.c:792 This page makes httpd crash consistently. More often than not.
Page seems to be random. Depending on the cache content I guess. I changed browser to test with chrome and it started to crash on different pages. Tried with roundcube : doesn't crash at all. No idea where the problem comes from and if it happens only with phpPgAdmin. I'll stay with opcache disabled for now.
Changing opcache.validate_timestamps to 0 in php.ini stops the crash. It can only be temporary because if the users change their pages, php will not notice it and keep using the cache.
I upgraded my system to FreeBSD 11.1-RELEASE and can't reproduce it anymore. Since I recompiled all ports, I'm not sure what "solved" the problem, or if it won't come back. If that's the case, I will probably open a bug at php.net, since it seems linked to php and not specifically to FreeBSD.