Ken Teh wrote:
We've discovered a bug in the NFSv4, SL4.4, 2.6.9-42.0.10. We have an
extremely simple piece of FORTRAN
program main
integer a,b,c
read(5,*) a,b,c
stop
end
If the input matches the data type, all is ok. If, instead of a
number, you input a character for any of the 3 variables, the code as
compiled by the Intel compiler will crash performing a traceback. If
the current directory where the traceback is triggered is an NFSv4
mounted directory, the kernel panics. /var/log/messages posts an
NFSv4 race condition before the kernel OOP's.
If you do the same in a non-NFSv4 mounted directory, everything is fine.
IF you do the same in a NFSv3 mounted directory, everything is also fine.
Any suggestions? How does one go about reporting this?
NFSv4 -- Discussion of the linux nfsv4 client and server
http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
USV - RH Bugzilla Info - If it's not in Bugzilla, it's not a bug
http://www.redhat.com/magazine/020jun06/features/bugzilla/
RH Bugzilla query pages -
https://bugzilla.redhat.com/bugzilla/query.cgi?format=simple
https://bugzilla.redhat.com/bugzilla/query.cgi?format=advanced
SL Bugs & Features - https://www.scientificlinux.org/bug.feature/
--
Art Wildman - NWS JAX FL. - http://www.srh.noaa.gov/jax
"If the thunder don't get you, then the lightning will..." - Grateful Dead
