Hi everyone,
I fixed 2 cases where my 2.3 kernel `oopsed' in Coda.
- Long standing problem in coda_rename, only happens if the file being
renamed hasn't been accessed yet. This is hard to trigger because most
applications actually seem to stat(2) the objects of a rename before
calling rename. In my case the target inode was purged due to memory
pressure.
- New one in coda_lookup, 2 fids happen to hash to the same inode
number, coda_iget returns failure, but the test for the error return
was a couple of lines too low.
Both are reasonably hard to trigger, and I'm happy to have managed to
capture these.
Updated patch against 2.3.99-pre6-7 is at
ftp.coda.cs.cmu.edu:/pub/coda/linux/kernel/patch-coda-2.3.99-pre6-7.4.gz
It should apply to just about every 2.3.99-pre kernel.
Jan
