Date: Wed, 15 Sep 2004 16:14:50 GMT From: Michel Gravey <michel.gravey@7ici.biz> To: freebsd-gnats-submit@FreeBSD.org Subject: kern/71771: Hang during heavy load with amr raid controller (466 series / dell perc 2 SC) Message-ID: <200409151614.i8FGEoCE076983@www.freebsd.org> Resent-Message-ID: <200409151620.i8FGKIdY048371@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 71771
>Category: kern
>Synopsis: Hang during heavy load with amr raid controller (466 series / dell perc 2 SC)
>Confidential: no
>Severity: serious
>Priority: low
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Wed Sep 15 16:20:17 GMT 2004
>Closed-Date:
>Last-Modified:
>Originator: Michel Gravey
>Release: 4.10-RELEASE--p2
>Organization:
7ici
>Environment:
FreeBSD proliant 4.10-RELEASE-p2 FreeBSD 4.10-RELEASE-p2 #19: Thu Aug 5 21:55:46 CEST 2004 root@proliant:/usr/src/sys/compile/PROLIANT i386
>Description:
The hang (no panic) comes under heavy load after 1-5 hours of make world running at -j4(or -j6 or -j12). The system is a proliant 1850R(P3SMP) with 4 drives on a hardware raid 5 controller, a amr 466 (dell perc2sc).
make.conf: COPTFLAGS= -O2 -pipe -march=pentiumpro or GENERIC kernel from fresh install
>How-To-Repeat:
Probably very hardware specific but running several make buildworld/installworld during several hours (1-5, maybe less, it depends) with SMP turned on, with a amr controller 466 series should repeat the problem. A big dbench doesn't reproduce the problem (maybe not tried enought time).
>Fix:
Here is a patch from cognet@freebsd.org whitch correct the problem
Index: amr.c
===================================================================
RCS file: /home/ncvs/src/sys/dev/amr/amr.c,v
retrieving revision 1.7.2.15
diff -u -p -r1.7.2.15 amr.c
--- amr.c 22 Jul 2004 16:35:18 -0000 1.7.2.15
+++ amr.c 5 Aug 2004 11:41:30 -0000
@@ -326,7 +326,7 @@ amr_startup(void *arg)
/*
* Start the timeout routine.
*/
-/* sc->amr_timeout = timeout(amr_periodic, sc, hz);*/
+ sc->amr_timeout = timeout(amr_periodic, sc, hz);
return;
}
@@ -542,14 +542,16 @@ static void
amr_periodic(void *data)
{
struct amr_softc *sc = (struct amr_softc *)data;
-
+ int s;
+
debug_called(2);
/* XXX perform periodic status checks here */
/* compensate for missed interrupts */
+ s = splbio();
amr_done(sc);
-
+ splx(s);
/* reschedule */
sc->amr_timeout = timeout(amr_periodic, sc, hz);
}
>Release-Note:
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200409151614.i8FGEoCE076983>