:On the server I downgraded vfs_bio.c to rev 1.187 & rebooted; no luck. I
:then installed the same kernel (with the downgraded vfs_bio.c) to the
:client. Bingo. With both NFS client & server machine running rev 1.187,
:the problem so far as building XFree86-contrib from an NFS mounted
:/usr/ports disappears.
:
:As Chuck Robey noted, it seems like the client's writes are not completely
:being committed to the server, which results in partially baked files
:which are truncated.
:
:Unfortunately -r1.188 -r1.187 doesn't apply cleanly, so there's some work
:to be done by Eivind to adapt his subsequent commits if we were to say,
:back out 1.188 prior to the branch.
:
:-Chris
Hmm. r1.88 are Luoqi's fixes to the handling of misaligned buffers. It is
quite possible that there is a bug in there or with assumptions made in
the NFS code in regards to how buffers are handled, but most of those
changes are theoretically critical to the proper operation of the VFS code
on other platforms (aka alpha, with its larger page size), and pretty
important to the msdos code too under certain circumstances. And it was
reviewed pretty well, too. I'd rather we not lose the work.
I think it is important that we find the bug and fix it without having
to back-out that entire patch! I was hoping to avoid updating my
codebase until after the split, but I'll go ahead and try to sync it up
tonight and see if I can reproduce the NFS problem. If I can do that,
I should be able to locate the bug and fix it.
So, nobody backout 1.188 yet, please.
-Matt
Matthew Dillon
<dil...@backplane.com>
To Unsubscribe: send mail to majord...@freebsd.org
with "unsubscribe freebsd-current" in the body of the message
