[OpenAFS-devel] Bug with OpenAFS memory-mapped file I/O under Linux

[chas williams]

>We've run into a rather annoying issue here while running tests on Openafs
>1.2.2 on Redhat 7.0 (using the 2.4.9-12 kernel).  Maybe someone on the list
>is familiar with the problem or has a suggestion.

any chance you could do a 'fstrace setset -active', run your test code, 
'fstrace dump' and send the result output?  that would be somewhat
useful.

>BTW, earlier Redhat kernels had a similar problem under NFS, but it's fixed
>in 2.4.9-12 and later.

a quick search on the net shows that the following patch fixed the nfs problem:

http://www.fys.uio.no/~trondmy/src/2.4.7/linux-2.4.7-mmap.dif

it looks the big change was to cache the rpc credentials with the file.
i suppose its possible that the write/close for mmap is being handled by an
anonymous thread.  (irix does this somewhat)