Date: Mon, 23 Jun 2008 19:34:24 +0400 From: Lev Serebryakov <lev@FreeBSD.org> To: John Baldwin <jhb@freebsd.org> Cc: Lev Serebryakov <lev@freebsd.org>, freebsd-stable@freebsd.org Subject: Re[2]: FreeBSD 7-STABLE deadlock! Message-ID: <168637165.20080623193424@serebryakov.spb.ru> In-Reply-To: <200806231047.33494.jhb@freebsd.org> References: <372651116.20080615162317@serebryakov.spb.ru> <485612AD.2000301@FreeBSD.org> <48564CAB.4030102@FreeBSD.org> <200806231047.33494.jhb@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hello, John. You wrote 23 =E8=FE=ED=FF 2008 =E3., 18:47:33: > On Monday 16 June 2008 07:21:15 am Lev Serebryakov wrote: >> Lev Serebryakov wrote: >> > It seems to be ATA/SATA or UFS2 problem: now I have computer in state, >> > when 4 iozone processes are hanged in "Disk wait" state, and I can not >> > "cd" to filesystem, which is tested by "iozone". >> > But I can create processes, work on system, etc., if I don't touch th= is >> > filesystem. >> >> I can reproduce it, creating gmirror on 5 disks (yes, not very useful >> configuration, but I've started from non-base-system RAID5 and need to >> exclude it), FS with 64Kb blocks, and 4 threads of iozone with mixed >> workload (-i 8 -+p 70). >> >> All 5 disks are ICH9DO-based, SATA-II WD5000AAKS HDDs. > Try getting the 'ps' output from ddb. Also, get a crash dump if you can. It was tracked douwn to known deadlock in buffer allocator when buffer map is fragmented (thnx to kib@!). Workaround is known: don't use FSes with 16Kb and 64Kb blocks on same system in one time. 16/32 mixture works well :) --=20 // Black Lion AKA Lev Serebryakov <lev@FreeBSD.org>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?168637165.20080623193424>