On Monday 14 May 2001 20:33, Andreas Dilger wrote:
> Daniel, you write:
> > Now, if the check routine tells us how much good data it found we
> > could use that to set a limit for the dirent scan, thus keeping the
> > same robustness as the old code but without having all the checks
> > in the inner loop. Or. We could have separate loops for good
> > blocks and bad blocks, it's just a very small amount of code.
>
> Yes, I was thinking about both of those as well. I think the latter
> would be easiest, because we only need to keep a single error bit per
> buffer.
Today's patch has the first part of that fix:
http://nl.linux.org/~phillips/htree/dx.pcache-2.4.4-6
I broke up Al's check_page routine into the page-specific part and the
dirent-specific part, which I call every time a buffer is brought
uptodate in ext2_bread. This is roughly as efficient as Al's
page-oriented check. I could get rid of the code in Al's check_page
that initializes sets the rec_lens of a new dir page to blocksize
because I do that explicitly in ext2_add_entry now. This will make it
a little cleaner.
The next step is to try and incorporate the intelligence about the good
parts of a bad dirent block into the entry lookup code cleanly.
I moved ext2_bread and ext2_append into dir.c because of their
dir-specific nature. (I have some plans for ext2_getblk that have
nothing to do with directories, which is why I'm trying to keep it
generic.)
-- Daniel - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue May 15 2001 - 21:00:37 EST