Re: Off-Topic Question
On Wed, 16 Feb 2005 17:49:59 +0000, Eduardo Tongson <propolice@gmail.com> wrote:
> > (a) softdep is complex. *very* complex.
> > (b) softdep doesn't solve the problems of fast reboots after crash
> > (c) softdep doesn't allow replaying changes for off-site synchronisation
> >
> > I consider softdep a useful concept for a filesystem, it can really
> > help for those filesystems we can't add a journaling later on top.
> > But having native journaling support can be much better.
> >
> > Joerg
> >
>
> So that's the problem with softdep.
> Can you pls similarly point out the flaws of background fsck
>
>
There are several problems with BG fsck. An example of one problem
with background fsck is that if you end up with real disk corruption,
e.g. as result of a power failure, a BG fsck will potentially not
correct it. The CG bitmaps that are set that shouldn't be are
effectively read-only, and the snapshot permits you to fsck in
background. This process assumes that the only thing broken is the CG
bitmap. If anything else is broken then it's likely your system will
panic when it tries to use corrupt data in some pointer math or some
other use in the implementation of the filesystem in the kernel.
Problems like this lead to a never ending problem as you crash, then
reboot, then BG fsck starts up(not correcting the problem), and you
crash again. This process happens over and over until you run a full
fsck.
--
-David
Steven David Rhodus
<drhodus@machdep.com>
討論串 (同標題文章)
完整討論串 (本文為第 15 之 27 篇)：