Date: Mon, 18 Jan 1999 20:46:32 -0500 (EST) From: Matt Behrens <matt@megaweapon.zigg.com> To: FreeBSD-gnats-submit@FreeBSD.ORG Subject: kern/9560: softupdates freezes system under heavy load Message-ID: <199901190146.UAA46096@megaweapon.zigg.com>
next in thread | raw e-mail | index | archive | help
>Number: 9560 >Category: kern >Synopsis: softupdates freezes system under heavy load >Confidential: no >Severity: critical >Priority: high >Responsible: freebsd-bugs >State: open >Quarter: >Keywords: >Date-Required: >Class: sw-bug >Submitter-Id: current-users >Arrival-Date: Mon Jan 18 18:00:01 PST 1999 >Closed-Date: >Last-Modified: >Originator: Matt Behrens >Release: FreeBSD 3.0.0-19990112-SNAP i386 >Organization: zigg.com >Environment: 486 DX2-50 Diamond Speedstar Pro SVGA 1MB ISA Generic (?) VLB IO/FDD/IDE controller - Conner 1275 MB CFS1275A - Maxtor 3663 MB 83840 A6 Adaptec AHA-1542 ISA - Conner tape drive (not used during this problem) AT Keyboard Soundblaster compatible (not configured) Kingston EtheRx ISA NE2000 compatible 32MB (4 4MB 4x9 30-pin SIMMs) >Description: Under heavy load, when using softupdates, the system will occasionally stop processing, except for syscons, which seems to still be working. The drive will thrash violently until the system is powered down. A trace from DDB gave the following information: Debugger(f021f865) at Debugger+0x37 sc_alloc_history_buffer(f025267c,2,0,f025267c,7011d) at sc_alloc_history_buffer+ 0xcd0 scdevtotty(f025267c,0,0,0,f0612c00) at scdevtotty+0x35c atkbd_attach_unit(f025267c,0,0,f32b7e24,f01effce) at atkbd_attach_unit+0x4e7 is_physical_memory(0,c0084040,f0610010,f0130010,7011d) at is_physical_memory+0x2 cc Xintr1(0,0,f01a7188,c0000000,0) at Xintr1+0x5e wdintr(0,c0000000,f0610010,10,c0000000) at wdintr+0x5b8 Xintr14(80000000,f32b0010,f01d0010,f32b7f88,f32a6200) at Xintr14+0x61 doreti_popl_es_fault(f32b7f88) at doreti_popl_es_fault+0x49 vn_syncer_add_to_worklist(f32abbf7,f02192a5,f0230588,f01f30fs,f01ef5c0) at vn_sy ncer_add_to_worklist+0x13c kproc_start(f0230588) at kproc_start+0x32 fork_trampoline(e4ec1589,c766f025,25e4ee05,800008f0,25e4f025) at fork_trampoline +0x30 Attempting to panic the system to sync it from DDB got stuck on the sync phase due to wd error 58. Upon rebooting, the system can successfully be fsck -p'd back into an intact filesystem, reporting INCORRECT BLOCK COUNT and UNREF FILE errors. When finished, several of the files that were last created are missing from the disk. This system works fine if drives are not mounted with softupdates, performing the same tasks (one of these instances was caused by untarring XFree86 in the ports collection, just minutes after a reboot) very reliably. No known solid method of repeating the problem. It seems to occur under heavy load but cannot necessarily be repeated with certaincy. >How-To-Repeat: >Fix: None known. >Release-Note: >Audit-Trail: >Unformatted: To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901190146.UAA46096>