On Mon, 2015-04-06 at 23:33 -0700, Christoph Hellwig wrote:
Hi Dan,
it seems like the MISCOMPARE case of COMPARE AND WRITE has been broken
for quite a while.
See my post from March 30 ("hang during COMPARE AND SWAP", which shows a
similar issues using a synthetic test suite.
Not exactly. The one you reported earlier is a bug specific to loopback
and vhost with COMPARE_AND_WRITE, and AFAICT has always been broken.
I'm working on a patch for that now..
From what I can gather from Dan's log with qla2xxx, a MISCOMPARE
occurs
in compare_and_write_callback(), followed up by an ABORT_TASK. This is
not what's happening with loopback.
Dan, can you confirm which kernel version you where previously using
qla2xxx target mode with Ubuntu 11.04..?
Thanks,
--nab