Date: Tue, 6 Sep 2005 03:50:36 -0700 (PDT) From: Don Lewis <truckman@FreeBSD.org> To: phk@haven.freebsd.dk Cc: current@FreeBSD.org Subject: Re: patch for ext2fs unmount problem at shutdown Message-ID: <200509061050.j86AoauO032564@gw.catspoiler.org> In-Reply-To: <64186.1125996351@phk.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 6 Sep, Poul-Henning Kamp wrote: > In message <200509060841.j868fj1I032057@gw.catspoiler.org>, Don Lewis writes: > >>I suspect both that is one possible reason, and another reason would be >>to avoid marking the file system clean if any writes timed out. > > In that case the unmount should fail, otherwise the filesystem is > buggy. ext2fs is definitely buggy because it writes the superblock to mark the file system clean before it writes some of the file system meta-data (without error checking), but this is a different bug that won't get caught by the nbusy check. Both ext2fs and ufs rely on vflush(), which essentially does a vgonel() on all the vnodes, which I don't think will catch any stuck writes. Ufs does quite a bit more error checking than ext2fs, but I don't know if it is enough to avoid marking the file system clean when it shouldn't be. > Considering that our kernel presently tend to explode violently on > disk errors I would say that the nbusy check should just be commented > out for now and vfs_unmountall() always tried. vfs_umountall() really needs to be made smart enough to properly unwind all the dependencies between mounted file systems, swap devices, file systems mounted on swap-backed or vnode-backed md's, and vnode-backed swap devices as was discussed at length a while ago. Perhaps the error checking on unmount should be combined with the vfs_unmountall() redesign.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200509061050.j86AoauO032564>