linux.git/block, branch v3.17-rc4

cfq-iosched: Add comments on update timing of weight

2014-08-28T14:16:29+00:00

Explain that weight has to be updated on activation.
This complements previous fix e15693ef18e1 ("cfq-iosched: Fix wrong
children_weight calculation").

Signed-off-by: Toshiaki Makita 
Signed-off-by: Jens Axboe

cfq-iosched: Fix wrong children_weight calculation

2014-08-26T16:17:30+00:00

cfq_group_service_tree_add() is applying new_weight at the beginning of
the function via cfq_update_group_weight().
This actually allows weight to change between adding it to and subtracting
it from children_weight, and triggers WARN_ON_ONCE() in
cfq_group_service_tree_del(), or even causes oops by divide error during
vfr calculation in cfq_group_service_tree_add().

The detailed scenario is as follows:
1. Create blkio cgroups X and Y as a child of X.
   Set X's weight to 500 and perform some I/O to apply new_weight.
   This X's I/O completes before starting Y's I/O.
2. Y starts I/O and cfq_group_service_tree_add() is called with Y.
3. cfq_group_service_tree_add() walks up the tree during children_weight
   calculation and adds parent X's weight (500) to children_weight of root.
   children_weight becomes 500.
4. Set X's weight to 1000.
5. X starts I/O and cfq_group_service_tree_add() is called with X.
6. cfq_group_service_tree_add() applies its new_weight (1000).
7. I/O of Y completes and cfq_group_service_tree_del() is called with Y.
8. I/O of X completes and cfq_group_service_tree_del() is called with X.
9. cfq_group_service_tree_del() subtracts X's weight (1000) from
   children_weight of root. children_weight becomes -500.
   This triggers WARN_ON_ONCE().
10. Set X's weight to 500.
11. X starts I/O and cfq_group_service_tree_add() is called with X.
12. cfq_group_service_tree_add() applies its new_weight (500) and adds it
    to children_weight of root. children_weight becomes 0. Calcularion of
    vfr triggers oops by divide error.

weight should be updated right before adding it to children_weight.

Reported-by: Ruki Sekiya 
Signed-off-by: Toshiaki Makita 
Acked-by: Tejun Heo 
Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe

block: fix error handling in sg_io

2014-08-26T14:20:01+00:00

Before commit 2cada584b200 ("block: cleanup error handling in sg_io"),
we had ret = 0 before entering the last big if block of sg_io.

Since 2cada584b200, ret = -EFAULT, which breaks hdparm:

/dev/sda:
 setting Advanced Power Management level to 0xc8 (200)
 HDIO_DRIVE_CMD failed: Bad address
 APM_level      = 128

Signed-off-by: Sabrina Dubroca 
Fixes: 2cada584b200 ("block: cleanup error handling in sg_io")
Reviewed-by: Christoph Hellwig 
Signed-off-by: Jens Axboe

fix regression in SCSI_IOCTL_SEND_COMMAND

2014-08-22T20:04:33+00:00

blk_rq_set_block_pc() memsets rq->cmd to 0, so it should come
immediately after blk_get_request() to avoid overwriting the
user-supplied CDB.  Also check for failure to allocate rq.

Fixes: f27b087b81b7 ("block: add blk_rq_set_block_pc()")
Cc:  # 3.16.x
Signed-off-by: Tony Battersby 
Signed-off-by: Jens Axboe

scsi-mq: fix requests that use a separate CDB buffer

2014-08-22T20:04:31+00:00

This patch fixes code such as the following with scsi-mq enabled:

    rq = blk_get_request(...);
    blk_rq_set_block_pc(rq);

    rq->cmd = my_cmd_buffer; /* separate CDB buffer */

    blk_execute_rq_nowait(...);

Code like this appears in e.g. sg_start_req() in drivers/scsi/sg.c (for
large CDBs only).  Without this patch, scsi_mq_prep_fn() will set
rq->cmd back to rq->__cmd, causing the wrong CDB to be sent to the device.

Signed-off-by: Tony Battersby 
Signed-off-by: Jens Axboe

block: support > 16 byte CDBs for SG_IO

2014-08-22T01:42:01+00:00

Signed-off-by: Christoph Hellwig 
Reviewed-by: Boaz Harrosh 
Signed-off-by: Jens Axboe

block: cleanup error handling in sg_io

2014-08-22T01:42:01+00:00

Make sure we always clean up through the out label and just have
a single place to put the request.

Signed-off-by: Christoph Hellwig 
Signed-off-by: Jens Axboe

blk-mq: blk_mq_freeze_queue() should allow nesting

2014-08-22T01:37:51+00:00

While converting to percpu_ref for freezing, add703fda981 ("blk-mq:
use percpu_ref for mq usage count") incorrectly made
blk_mq_freeze_queue() misbehave when freezing is nested due to
percpu_ref_kill() being invoked on an already killed ref.

Fix it by making blk_mq_freeze_queue() kill and kick the queue only
for the outermost freeze attempt.  All the nested ones can simply wait
for the ref to reach zero.

While at it, remove unnecessary @wake initialization from
blk_mq_unfreeze_queue().

Signed-off-by: Tejun Heo 
Reported-by: Ming Lei 
Signed-off-by: Jens Axboe

blk-mq: correct a few wrong/bad comments

2014-08-22T01:37:49+00:00

Just grammar or spelling errors, nothing major.

Signed-off-by: Jens Axboe

block: Fix BUG_ON when pi errors occur

2014-08-22T01:37:47+00:00

When getting a pi error we get to bio_integrity_end_io with
bi_remaining already decremented to 0 where we will eventually
need to call bio_endio with restored original bio completion handler.
Calling bio_endio invokes a BUG_ON(). We should call bio_endio_nodec
instead, like what is done in bio_integrity_verify_fn.

Signed-off-by: Sagi Grimberg 
Signed-off-by: Jens Axboe