Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 260977 - net-fs/openafs: File Corruption from my Gentoo Client
Summary: net-fs/openafs: File Corruption from my Gentoo Client
Status: RESOLVED UPSTREAM
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: New packages (show other bugs)
Hardware: x86 Linux
: High normal (vote)
Assignee: Stefaan De Roeck (RETIRED)
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-03-02 17:55 UTC by Troy Telford
Modified: 2010-06-21 23:11 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Troy Telford 2009-03-02 17:55:16 UTC
I've got an AFS server set up on an x86_64 gentoo system.  As far as I can tell, it works fine with most clients that I attach to it.

However, there are a few clients which can attach to it, but when they write files, the files are not written correctly - entire sections of the file are cut out.  It doesn't happen on every file, but it does happen fairly often.

For example, if I unpack a tarball containing the linux kernel on one of the 'problem' systems, and compare it to the same tarball unpacked to a non-AFS filesystem, there are a substantial number of files which have sections removed.  Of course, on one of the 'good' clients, there is no difference.

It has been suggested (in the openafs-info mailing list) that the FS that the client uses for a cache is important - I've also read this, but I've read that ext3 is 'tolerated'.  Either way, for a test I created a filesystem image that uses ext2, and loop-mounted it so the cache is in fact on an ext2 filesystem.  The behavior is the same, the files are corrupted.

Now for my setup:
Server OS:  Gentoo 2008.0 x86_64, Kernel 2.6.28-gentoo (gentoo-sources ebuild).
OpenAFS ebuild version: 1.48-r1

My 'problem' client:
Client OS:  Gentoo 2008.0 x86 Kernel 2.6.28-gentoo-r2 (gentoo-sources ebuild)
OpenAFS ebuild:  1.4.8-r1
openafs-kernel ebuild:  1.4.8-r1

I also have seen very similar behavior on SLES 10 (SP1, I think) - kernel 2.6.16.46-0.12-default, running OpenAFS 1.4.8.

I have no idea if the problem is in the client(s), or if it's in the server, or what.  I'm not sure what additional information you'd need at this time, and I certainly have no idea how to debug the problem.
Comment 1 Troy Telford 2009-03-03 18:38:00 UTC
Additional:  I've tested it using memcache (instead of the on-disk cache).  The error is the same.
Comment 2 Troy Telford 2009-03-03 18:38:27 UTC
Changing "AFS" to the more useful "OpenAFS"
Comment 3 Stefaan De Roeck (RETIRED) gentoo-dev 2009-03-06 21:48:40 UTC
I'm afraid this is hardly a problem that I'm capable of tracking down.
May I encourage you to take your problem to the upstream mailing list openafs-info?  Your report seems thorough, and I think they would appreciate it.
We can keep this bug open as to document what happens upstream. 

Stefaan 
Comment 4 Troy Telford 2009-03-07 01:06:43 UTC
Since I package stuff professionally, I think I understand about not being qualified to debug it; but I figured it would be worth a shot.

I'll give openafs-info another try, however, other than comments to the extent of "yeah, you should get that fixed", there was no additional interest or help.
Comment 5 SpanKY gentoo-dev 2010-06-21 23:11:41 UTC
openafs 1.4.12.1 is in the tree