On Fri, 6 Jul 2001, Jeffrey Hutzelman wrote:
> I haven't read the code to verify this, but one would hope that whoever
> wrote the AFS code in question realized that they might have to give up
> the read lock in order to get the write lock, and thus the write-locked
> portion of the code doesn't depend on any data gathered during the
> read-locked portion. If this is the case, then the AFS_TRYUP that is
> there now is perfectly fine, and the call to rw_tryupgrade() is merely an
> optimization to avoid yielding. If not, then AFS_TRYUP is not OK, and I
> can think of no possible behaviour of rw_tryupgrade() that would make AFS
> happy.
Nothing calls AFS_TRYUP, so it doesn't matter.
-D
_______________________________________________
OpenAFS-devel mailing list
[EMAIL PROTECTED]
https://lists.openafs.org/mailman/listinfo/openafs-devel
