[OpenAFS-devel] Odd Behavior with 1.4.2fc3

[chas williams - CONTRACTOR]

In message <4506CEFD.8070105@cert.org>,Jason McCormick writes:
>The second problem is an intermittent problem with CVS repositories
>stored in AFS.  This has worked fine since time immemorial but with
>1.4.2fc3 I get periodic problems with repository locks.  CVS reports
>that it can't write the lock file but then accepts the transaction just
>fine.

i think i have seen this problem as well with an svn repository
running 1.4.2fc2.  access to a particular file fails with EACCESS,
but this issue is resolved after i issue an 'fs flush' on the file.
btw, i was running memcache and the servers are 1.4.1.

the fstrace looked like:

time 252.943837, pid 916: Access vp 0xdff10680 mode 0x40 len (0x0, 0x1000)
time 252.943843, pid 916: Access vp 0xec304120 mode 0x40 len (0x0, 0x800)
time 252.943849, pid 916: Access vp 0xc73a77a0 mode 0x40 len (0x0, 0x800)
time 252.943855, pid 916: Access vp 0xce459360 mode 0x40 len (0x0, 0x800)
time 252.943860, pid 916: Access vp 0xd2177bc0 mode 0x40 len (0x0, 0x800)
time 252.944987, pid 916: Analyze RPC op 2 conn 0xf400b2e0 code 0x0 user 0x41f767c5
time 252.944992, pid 916: ProcessFS vp 0xd2177bc0 old len (0x0, 0x800) new len (0x0, 0x800)
time 252.945016, pid 916: Access vp 0xc53d8700 mode 0x100 len (0x0, 0x2a8)
time 252.946067, pid 916: Analyze RPC op 2 conn 0xf400b2e0 code 0x66 user 0x41f767c5
time 252.946074, pid 916: Returning code 13 from 17
time 252.987257, pid 916: Analyze RPC op 15 conn 0xf400b2e0 code 0x0 user 0x41f767c5

what puzzles me here is the call to afs_Analyze() after we are
in afs_Access().  it seems to have a code 0x66 which is ENETRESET.
this seems to imply that RXAFS_FetchStatus() was returning this code.