Re: [Ext2-devel] Re: Filesystem can be marked clear when it is not

[Andreas Dilger <adilger@turbolinux.com>]

Daniel writes:
> I weighed in on that one too:
> 
>     http://marc.theaimsgroup.com/?l=ext2-devel&m=99090670900520&w=2
> 
> with a very simple sort-of patch, which I just made into a real patch.  

Ok, your patch works in this case (it is leaving sb->s_dirt = 1, however, so
the superblock will be written out again shortly).  In fact, the whole
thing can be replaced with a call to ext2_write_super() in my code, which
also turns off EXT2_VALID_FS and sets s_mtime, and writes it synchronously
to disk.  This means we can remove 2! lines from ext2_setup_super().

It makes me wonder, though, if we clear EXT2_VALID_FS synchronously in
ext2_setup_super() if we also need it in ext2_write_super().  If people
mount their root fs read-write e2fsck will clear EXT2_VALID_FS and we
may never hit ext2_setup_super() again to clear it, so I guess it needs
to stay there.

One of the other changes from my patch is that errors are also written
out synchronously to disk, for the same reason - in case we crash shortly
after having an error, and before dirty buffers are flushed to disk.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert