[Btrfs-devel] performance fixes pushed out

[Bron Gondwana]

On Wed, 20 Jun 2007 10:58:35 -0400, "Chris Mason" <chris.mason at oracle.com> said:
> On Thu, Jun 21, 2007 at 12:56:14AM +1000, Bron Gondwana wrote:
> > On Wed, Jun 20, 2007 at 08:57:11AM -0400, Chris Mason wrote:
> > > Hello everyone,
> > > 
> > > While working on enospc, I hit some low hanging performance fixes in the
> > > allocator.  For inserts into the extent tree, the allocator has to
> > > search for a worst case group of available blocks.  The fix I pushed out
> > > caches that search for reuse later.
> > > 
> > > Also, I added some btree readahead for the snapshot dropping code.
> > > 
> > > You can pull the latest from:
> > > 
> > > http://oss.oracle.com/mercurial/mason/btrfs
> > 
> > My maildirs died almost immediately this time, with one big folder
> > coming up with files full of NULLs rather than the actual contents.
> > I'm going to try to recreate it tomorrow - it's getting crazy late
> > here and I had to much actual work[tm] (Cyrus patches) to spend too
> > much time playing.
> > 
> > This is with a hg clone from about 16 hours ago, morning Australian
> > time.
> 
> If you still have the bad FS, please try running btrfsck on it.

I tried that and it didn't find anything.  I also tried unmounting and
remounting in case something was corrupted in memory and I tried loading
the same mailbox from the reiserfs I'd just copied it from a few minutes
earlier, where it loaded fine.

That said, there was one corrupted file on the reiserfs in one of the other
maildirs which I found while copying.  The laptop lives a pretty rough life
(it's missing two keys and has been dropped pretty hard at least once - but
it's on it second hard drive, second motherboard and 4th screen, so it's 
being kept fixed-up)

I'm going to run 'badblocks' over the device first this time and see if
anything shows up!

Bron.
-- 
  Bron Gondwana
  brong at fastmail.fm