From nobody@hyperreal.com Sun Apr 13 11:53:43 1997 Received: (from nobody@localhost) by hyperreal.com (8.8.4/8.8.4) id LAA28354; Sun, 13 Apr 1997 11:53:43 -0700 (PDT) Message-Id: <199704131853.LAA28354@hyperreal.com> Date: Sun, 13 Apr 1997 11:53:43 -0700 (PDT) From: Brian Moore Reply-To: bem@cmc.net To: apbugs@hyperreal.com Subject: mod_proxy(?) seems to alarm(0) somewhere X-Send-Pr-Version: 3.2 >Number: 374 >Category: mod_proxy >Synopsis: mod_proxy(?) seems to alarm(0) somewhere >Confidential: no >Severity: serious >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Sun Apr 13 12:00:01 1997 >Last-Modified: Sat Jul 19 03:40:41 PDT 1997 >Originator: bem@cmc.net >Organization: >Release: 1.2b8 >Environment: Solaris 2.5, all recommended patches, gcc 2.7.2 >Description: Looks like there's one other problem in mod_proxy with alarms being turned off (not blocked via the block_alarms() call, but alarm(0)'d for some reason). I'm guessing on the module involved, since the three dead children this morning were all doing proxy stuff. The backtrace of a child that's been waiting for 110k seconds: #0 0xef67792c in _read () #1 0x29364 in saferead () #2 0x29480 in bread () #3 0x488b0 in proxy_send_fb () #4 0x47e78 in proxy_http_handler () #5 0x432c0 in proxy_handler () #6 0x1f040 in invoke_handler () #7 0x21dc0 in process_request_internal () #8 0x21df4 in process_request () #9 0x1bf30 in child_main () #10 0x1c0cc in make_child () #11 0x1c8c8 in standalone_main () #12 0x1cb88 in main () (gdb) up #1 0x29364 in saferead () (gdb) print alarms_blocked $1 = 0 So this seems to be something calling alarm(0) somewhere instead of a 'logical' alarms-off via the official mechanism. >How-To-Repeat: Not sure: virtually all of our proxy users are on a 10Mbps ethernet but behind a firewall. This usage may or may not be relevant. The children I found dead this morning were fetching files from cdrom.com via http, so it should be normal the only odd thing is that these were quake files so they were no doubt huge. >Fix: Will be looking at the code myself this week >Audit-Trail: State-Changed-From-To: open-feedback State-Changed-By: chuck State-Changed-When: Sun Apr 27 21:01:07 PDT 1997 State-Changed-Why: Try 1.2b10 (due out tonight). I believe the problem is fixed there. State-Changed-From-To: feedback-closed State-Changed-By: dgaudet State-Changed-When: Sat Jul 19 03:40:40 PDT 1997 State-Changed-Why: I'm pretty sure you and Chuck worked this one out. Dean >Unformatted: