[PVFS2-developers] Re: smallio server death with old kernel

Pete Wyckoff pw at osc.edu
Thu Jan 5 11:02:09 EST 2006


(Murali: same mail as before, but with compressed attachments to
work around the -developers list 150 kB limit.)

vilayann at mcs.anl.gov wrote on Wed, 04 Jan 2006 21:26 -0600:
> Could you try the attached patch and let me know if it fixes the crash?
> (There are still some other bugs like the size of the file reported is
> off-by-one, but I haven't dug deep into the SMALL_IO code just yet..)

Yes, that fixes the crash, thanks!  But it is indeed off-by-one:

    piv001$ tar tvfz tarball.tgz 
    drwxr-xr-x pw/oscsys         0 2005-12-23 17:15:49 ./
    -rw-r--r-- pw/oscsys    180224 2005-12-23 17:15:40 ./data
    piv001$ tar xfzvC tarball.tgz /pvfs
    ./
    ./data
    piv001$ pvfs2-ls -la /pvfs
    drwxrwxrwx    1 pw       oscsys          4096 2006-01-05 10:05 .
    drwxrwxrwx    1 pw       oscsys          4096 2006-01-05 10:05 .. (faked)
    -rw-r--r--    1 pw       oscsys        180223 2005-12-23 17:15 data
    drwxrwxrwx    1 pw       oscsys          4096 2006-01-05 10:04 lost+found

But I don't see your debugging statements in the logs when I turn on
verbose on the client; weird.  Server logs attached too just in case those
are interesting.

> Also, could you look over and see if the encode_skip4 that I added will
> fix those pesky alignment warnings on 64 bit machines or if they are
> unnecessary?

Yours looks fine, but a bit more direct would be to remove the existing
skip4 and let the fs_id and enum nestle together in a single 64-bit
word.  It didn't break anything on ia64 (modified patch attached.)
There are other unrelated unaligned problems that have cropped up in
the past few months that I'll take a look at, though.

> Sam wrote the SMALL_IO protocol and he knows it best. I am pretty sure
> this is an ugly fix if at all it works :)
> I don't know how/if Rob/Sam want SMALL_IO protocol to be disabled since
> we need people to use the CVS head version and find the last remaining
> bugs (hopefully!). We could make an environment variable/mount-time option
> that would disable small i/o changes temporarily but I think that decision is upto
> them..

Agree, it should be fixed.  I was just under pressure here to get
past some old bugs and was hoping to do so by moving up to the CVS
head where I know at least one has been addressed (append bug).

		-- Pete
-------------- next part --------------
A non-text attachment was scrubbed...
Name: client.log.gz
Type: application/x-gunzip
Size: 34671 bytes
Desc: not available
Url : http://www.beowulf-underground.org/pipermail/pvfs2-developers/attachments/20060105/f4cc10bd/client.log-0001.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: server-41.log.gz
Type: application/x-gunzip
Size: 27427 bytes
Desc: not available
Url : http://www.beowulf-underground.org/pipermail/pvfs2-developers/attachments/20060105/f4cc10bd/server-41.log-0001.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: server-44.log.gz
Type: application/x-gunzip
Size: 11166 bytes
Desc: not available
Url : http://www.beowulf-underground.org/pipermail/pvfs2-developers/attachments/20060105/f4cc10bd/server-44.log-0001.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smallio2.patch.gz
Type: application/x-gunzip
Size: 1125 bytes
Desc: not available
Url : http://www.beowulf-underground.org/pipermail/pvfs2-developers/attachments/20060105/f4cc10bd/smallio2.patch-0001.bin


More information about the PVFS2-developers mailing list