[PVFS2-developers] Determining meaning of error code on I/O

Dean Hildebrand dhildebz at eecs.umich.edu
Wed Jul 21 12:14:22 EDT 2004


Thanks Neill, sorry for my delayed response.

> > getattr_acache state: fs_id:925967295 handle:2147483649
> > PINT_acache_lookup entered
> > PINT_acache_lookup exiting
> > getattr_acache state: Found!
>
> You've modified the getattr_acache state machine to grab attrs from
> the NFSv4 client?  I'm curious how your pNFS work is going, so if you
> want to describe what you're doing a bit more or join us on IRC, I'm
> all ears :-)

Sounds good!  What is your IRC server?  Do I need a username/password or
anything?

> Also, does it matter that most of the code path you've pasted in has
> been heavily modified somewhat recently?  Are you planning to keep
> your changes up to date alongside the pvfs2 source tree?

Well, for this summer I don't think it is an issue that I'm on the 0.5.1
source.  At this point it would be quite a big chore to integrate my
changes to 0.6.0, so hopefully I can get some performance results out of
the 0.5.1 base. I'm interested in keeping my changes up to date mainly for
the purpose of utilizing bug fixes (in PVFS2 and NFSv4).  There is also
the issue of functionality as it seems 0.5.1 doesn't use the page cache
and so I'm not either.  (aio_read/aio_write)  It seems the code exists in
inode.c but I can't see where it is called.  Has pvfs2_file_read/write in
file.c been replaced with the inode.c functions in 0.6.0?  I can ask more
questions about this on irc.  I'm also interested in if you guys are
planning to have a user level and kernel level cache, etc.

Dean



More information about the PVFS2-developers mailing list