]> asedeno.scripts.mit.edu Git - linux.git/commit
scsi: csiostor: Fix completion usage
authorDaniel Wagner <daniel.wagner@bmw-carit.de>
Tue, 13 Sep 2016 08:58:48 +0000 (10:58 +0200)
committerMartin K. Petersen <martin.petersen@oracle.com>
Wed, 14 Sep 2016 17:19:15 +0000 (13:19 -0400)
commit3e3f5a8a0f03e5af8a243f6b04cb007068481358
tree2330613954349d3d074f5dcbde398affaf07ec2a
parent4086eae52beab7808b8aa1a9f7d4095eebc247c7
scsi: csiostor: Fix completion usage

The (re)initializing of the completion object should be done before we
trigger the transfer. Doing this after triggering the hardware opens up
a race window. Without the timeout we would problaly even deadlock. Use
also reinit_completion because we initalize the whole data structure in
csio_scscim_init().

There is only one waiter for the completion, therefore there is no need
to use complete_all(). Let's make that clear by using complete() instead
of complete_all().

The usage pattern of the completion is:

waiter context                          waker context

csio_eh_abort_handler()
  reinit_completion()
  wait_for_completion_timeout()

                                        csio_scsi_err_handler()
                                          complete()

[mkp: fix typo]

Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/csiostor/csio_scsi.c