Date: Thu, 22 Mar 2007 06:27:31 GMT From: Paul P<pprocacci@datapipe.com> To: freebsd-gnats-submit@FreeBSD.org Subject: i386/110658: vmspace mutex lock? Message-ID: <200703220627.l2M6RVeh090489@www.freebsd.org> Resent-Message-ID: <200703220640.l2M6e3Uh052194@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 110658 >Category: i386 >Synopsis: vmspace mutex lock? >Confidential: no >Severity: non-critical >Priority: low >Responsible: freebsd-i386 >State: open >Quarter: >Keywords: >Date-Required: >Class: sw-bug >Submitter-Id: current-users >Arrival-Date: Thu Mar 22 06:40:02 GMT 2007 >Closed-Date: >Last-Modified: >Originator: Paul P >Release: 6.2 >Organization: DataPipe >Environment: smtp-proxy03# uname -a FreeBSD smtp-proxy03.datapipe.net 6.1-RELEASE-p3 FreeBSD 6.1-RELEASE-p3 #0: Sat Aug 12 03:46:51 EDT 2006 root@smtp-proxy02.datapipe.net:/usr/obj/usr/src/sys/CUSTOM i386 >Description: Looking for some guidance on a problem that myself as well as a few others have been experiencing with FBSD 6.2 and spamassassin. At first I thought this was a spamassassin bug, but now after reading through the FBSD 6.2 source code a bit, I'm beginning to think this is a BSD bug. I'm currently running spamassassin setup in several mail gateways. There are 5 machines total, all being load balanced. The first machine is a FBSD 4.11 machine where the problem doesn't exist, while the other 4 machines are FBSD 6.2 where the problem exists on all those machines. The machines have been configured as identically as possible. http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5313 This is one of the ongoing bug reports in relation to the problem being experienced. The short of it, is that after several hours of running spamassasin on FBSD 6.2, spamassassin suddenly fails on a select returning -1, and fails to select any new socket for reading/writing. Pay particular attention to comment #14 in the above referenced url. That is mine. It shows resources are at a minimum usage with processes *waiting* for spamassassin to get it's act back in gear. Now I'm not the greatest kernel hacker, but I did come across the following couple of lines in the source code and believe this *potentially* can cause the problem, but certainly please do not take my word for it: /* The line which I *believe* could be causing this problem is:*/ Lines 173-197:sys/kern/kern_exit.c while (p->p_lock > 0) msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0); "exithold" is the current state of the processes as described in the url given above. Can anyone provide any logical insight to this? Being that this has first cropped up in FBSD 6.2 and not 4.11, I believe somewhere this bug is being triggered. I can provide whatever information necessary to help aid in debugging this problem which again, I believe currently to be isolated to FBSD 6.2. >How-To-Repeat: My setup: spamassassin setup to scan/filter any/all messages going through this gateway. Messages are roughly 150000 per day. Concurrency doesn't seem to be a factor. Even though I use the above, I imagine any application that utilizes this part of the kernel has the potential of runninging into this problem. >Fix: Unknown >Release-Note: >Audit-Trail: >Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200703220627.l2M6RVeh090489>