Date: Wed, 26 May 2010 14:15:40 +0300 From: Andriy Gapon <avg@icyb.net.ua> To: Pawel Jakub Dawidek <pjd@FreeBSD.org> Cc: freebsd-fs@FreeBSD.org Subject: Re: ZFS panic: wrong length 131072 for sectorsize 2352 Message-ID: <4BFD02DC.1050203@icyb.net.ua> In-Reply-To: <20100525192721.GE1659@garage.freebsd.pl> References: <4BFA9AEC.1070608@jrv.org> <4BFB884A.3010804@jrv.org> <20100525192721.GE1659@garage.freebsd.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
on 25/05/2010 22:27 Pawel Jakub Dawidek said the following: > On Tue, May 25, 2010 at 03:20:26AM -0500, James R. Van Artsdalen wrote: >> On 5/24/2010 10:27 AM, James R. Van Artsdalen wrote: >>> panic: wrong length 131072 for sectorsize 2352 >> This fixes the bug. > > I don't think so:) size should be properly calculated at this point and > should be multiple of sectorsize. The problem is that vdev_geom_io() > splits request into MAXPHYS chunks if it is too big, which is wrong, > because MAXPHYS doesn't have to be multiple of sectorsize. But do we really want to support here sector sizes that are not power of two? And/or MAXPHYS which is not power of two? > Could you try this patch instead: > > http://people.freebsd.org/~pjd/patches/vdev_geom.c.4.patch Otherwise MAXPHYS % cp->provider->sectorsize would always be zero. >> Index: sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c >> =================================================================== >> --- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c (revision 208373) >> +++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c (working copy) >> @@ -250,6 +250,9 @@ >> if ((offset % pp->sectorsize) != 0) >> continue; >> >> + if ((size % pp->sectorsize) != 0) >> + continue; >> + >> if (vdev_geom_io(cp, BIO_READ, label, offset, size) != 0) >> continue; >> buf = label->vl_vdev_phys.vp_nvlist; > -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4BFD02DC.1050203>