On 09/01/2017 07:09 AM, William Allen Simpson wrote:
On 8/30/17 1:34 PM, William Allen Simpson wrote:
On 8/28/17 1:23 AM, Frank Filz wrote:
WRT14 is the test that failed that made me kick Bill’s patch out of
dev.5, then I couldn’t get it to fail again, so I included the patch
in dev.6.
Since it turned out not to be a dev.5 issue. Malahal is reporting dev.3.
We should start a new thread, as this isn't about dev.6.
I think there is something that is timing sensitive that is exposed,
and maybe c291141 is the trigger.
I need someone who can reliably re-create to dive into it… Maybe that
will be me this week…
Malahal says he's able to reliably reproduce on GPFS, was going to
test VFS too.
He also was going to send us his config, but we haven't seen that yet.
DanG was able to reproduce. Turns out the key was not using loopback to
test; required sending over an actual network to induce the timing. Then
logging worked.
It only occurs where WRT14 sends a minimal 4 byte XDR fragment after a
series of long ones, and the TCP segmentation happens to align.
Although I'm not yet able to reproduce myself, the problem was obvious
walking through the code. An accidental line deletion (setting a local
flags variable) in 1 of 3 paths. (That line was present in earlier
versions.) Because it is set properly in the most common path, the
compiler didn't give an uninitialized warning.
Hopefully DanG will be able to verify, and we'll get it in today!
I've verified the fix.
Daniel
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Nfs-ganesha-devel mailing list
Nfs-ganesha-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
