[OpenAFS-devel] openafs cvs broke on >=2.4.22, but 1.2.10a works fine

Nathan Neulinger nneul@umr.edu
Wed, 03 Dec 2003 12:56:58 -0600


On Tue, 2003-12-02 at 13:02, Derrick J Brashear wrote:
> On Tue, 2 Dec 2003, Nathan Neulinger wrote:
> 
> > Symptom is that /afs is mounted, but any access to it yields ENOTDIR/Not
> > a directory.
> >
> > Current cvs works fine against 2.4.21 but not 2.4.22 or newer. 1.2.10a
> > works fine against 2.4.21, .22, .23. From previous testing, it stopped
> > working with the first 2.4.22-pre patch.
> >
> > Any ideas on what might be causing this?
> 
> retabify src/afs/LINUX and diff; i bet it's in osi_vnodeops.c


osi_file - large space instead of small space
osi_modile - which symbol is looked for

misc changes in signal handling and sleeping

in vfsops - a PutVCache for afs_globalVp was added if afs_globalVp was
non null. 

vattr2inode call was changes from AFSTOI from AFSTOV

file read buffering changes for AFS_64BIT_CLIENT in osi_vnodeops.c,
related to a patch from hartmut for large file support. 

changes to do logs on tdc instead of avc...

some changes for @sys handling...



I sure don't see anything jumping out at me that screams problem, but
I'm also not really in tune with kernel internals. 



printf'ing the hell out of the afs/LINUX/ code yields information that
it's never even getting to anything in vnodeops... Really the only thing
that appears to get touched is afs_root in osi_vfsops.c, and it all
looks pretty normal. 

I'm set up to easily test stuff for this at the moment, so if you've got
any place you want me to look at, holler.

-- Nathan

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-6679
UMR Information Technology             Fax: (573) 341-4216