Avoid that checkpatch reports the following message:
ERROR: space required before the open parenthesis '('
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5421 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Since the block layer has no way to specify bio as HQ, there's no
choice, but to use every measure to approximate it as close as possible.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5416 d57e44dd-8a1f-0410-8b47-8ef2f437770f
The COMPARE AND WRITE implementation has been tested as follows against
an SCST vdisk_blockio device:
lba=7
bdev=$bdev
for ((i=0;i<4;i++)); do dd if=/dev/urandom of=b$i bs=8k count=1; done
for p in "0 1" "1 2" "2 3" "3 0"; do set $p; cat b$1 b$2 >b$1$2; done
dd if=/dev/urandom of=$bdev
dd if=b0 of=$bdev seek=$lba
md5sum $bdev
for ((i=0;i<10000;i++)); do for f in b01 b12 b23 b30; do sg_compare_and_write -l $lba -i $f -n 16 -x 16384 $bdev; done; done
md5sum $bdev
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5412 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Using READ_SYNC instead of READ increases the priority of read requests.
Using WRITE_SYNC instead of REQ_WRITE increases the priority of write
requests and avoids that the CFQ scheduler queues such writes waiting
for further write requests.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5411 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Avoid that checkpatch reports the following error message:
ERROR: return is not a function, parentheses are not required
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5398 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Avoid that checkpatch reports the following error message:
ERROR: "(foo*)" should be "(foo *)"
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5397 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Complain if srpt_abort_cmd() is called for an I/O context that is
being processed by SCST and not by the HCA.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5393 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Only complain about a missing completion for I/O contexts that are
in a state where the ib_srpt driver is waiting for the HCA.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5392 d57e44dd-8a1f-0410-8b47-8ef2f437770f
The comment above scst_nexus_loss() is somewhat confusing so change
it into something that is more clear.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5387 d57e44dd-8a1f-0410-8b47-8ef2f437770f
The to_scst_lcl_sess() macro is based on container_of() and hence never
returns NULL. Hence remove the two tests that compare the result of that
macro against NULL.
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5385 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Scst_mutex intentially used directly in the sysfs handler, because
comming sysfs improvements will allow that.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5382 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Apparently with mlx4 firmware up to and including 2.30.8000 it is not
guaranteed that for a QP associated with an SRQ error completions are
generated for all pending work requests. Avoid triggering
srpt_pending_cmd_timeout() for missing error completions.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5370 d57e44dd-8a1f-0410-8b47-8ef2f437770f
Invoking srpt_zerolength_write() before the queue pair has reached
the error state is a bug, so complain loudly if that happens.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5369 d57e44dd-8a1f-0410-8b47-8ef2f437770f
The overhead of atomic_add_return() is lower than that of a
spin_lock() / spin_unlock() pair, hence switch to the former.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5367 d57e44dd-8a1f-0410-8b47-8ef2f437770f
All callers guarantee that the completion queue is empty so it is
not necessary to invoke ib_poll_cq() from inside this function.
git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@5366 d57e44dd-8a1f-0410-8b47-8ef2f437770f
