|| ||Tejun Heo <firstname.lastname@example.org>|
|| ||email@example.com, firstname.lastname@example.org, James.Bottomley@steeleye.com,
|| ||[PATCH linux-2.6-block:master 00/10] blk: reimplementation of I/O barrier|
|| ||Wed, 27 Jul 2005 00:45:50 +0900 (KST)|
Hello, Jens, James, Jeff and Bartlomiej.
This is the third posting of blk ordered reimplementation. Changes
since the scond posting are...
* Dispatch queue patchset is reordered in front of this patchset.
* Draining bug fix
* Proper requeueing of requests in an ordered sequence for TAG
* Fallback mechanism stripped out. This also makes -EOPNOTSUPP
changes in SCSI and IDE unnecessary.
* TAG ordering request issue fix
* SCSI/libata/IDE patches splitted as requested
* Hopefully, changes to libata and IDE are more acceptable
* Documentation/block/barrier.txt added
This patchset is part of the patch series described in the following mail.
As all previous patches don't really concern subsystems other than
block layer. They are only sent to Jens and LKML. If they're needed,
preceding patches are...
For general description please read barrier.txt and the previous posting.
==== Some excerpts from barrier.txt. ====
* SCSI layer currently can't use TAG ordering even if the drive,
controller and driver support it. The problem is that SCSI midlayer
request dispatch function is not atomic. It releases queue lock and
switch to SCSI host lock during issue and it's possible and likely
to happen in time that requests change their relative positions.
Once this problem is solved, TAG ordering can be enabled.
* Error handling. Currently, block layer will report error to upper
layer if any of requests in an ordered sequence fails.
Unfortunately, this doesn't seem to be enough. Look at the
following request flow. QUEUE_ORDERED_TAG_FLUSH is in use.
    [pre] [barrier] [post] <    ... >
still in elevator
Let's say request ,  are write requests to update file system
metadata (journal or whatever) and [barrier] is used to mark that
those updates are valid. Consider the following sequence.
i. Requests  ~ [post] leaves the request queue and enters
ii. After a while, unfortunately, something goes wrong and the
drive fails . Note that any of ,  and  could have
completed by this time, but [pre] couldn't have been finished
as the drive must process it in order and it failed before
processing that command.
iii. Error handling kicks in and determines that the error is
unrecoverable and fails , and resumes operation.
iv. [pre] [barrier] [post] gets processed.
v. *BOOM* power fails
The problem here is that the barrier request is *supposed* to
indicate that filesystem update requests  and  made it safely
to the physical medium and, if the machine crashes after the barrier
is written, filesystem recovery code can depend on that. Sadly,
that isn't true in this case anymore. IOW, the success of a I/O
barrier should also be dependent on success of some of the preceding
requests, where only upper layer (filesystem) knows what 'some' is.
This can be solved by implementing a way to tell the block layer
which requests affect the success of the following barrier request
and making lower lever drivers to resume operation on error only
after block layer tells it to do so.
As the probability of this happening is very low and the drive
should be faulty, implementing the fix is probably an overkill.
But, still, it's there.
[ Start of patch descriptions ]
: add @uptodate to end_that_request_last() and @error to rq_end_io_fn()
Add @uptodate argument to end_that_request_last() and @error
to rq_end_io_fn(). There's no generic way to pass error code
to request completion function, making generic error handling
of non-fs request difficult (rq->errors is driver-specific and
each driver uses it differently). This patch adds @uptodate
to end_that_request_last() and @error to rq_end_io_fn().
For fs requests, this doesn't really matter, so just using the
same uptodate argument used in the last call to
end_that_request_first() should suffice. IMHO, this can also
help the generic command-carrying request Jens is working on.
: separate out bio init part from __make_request
Separate out bio initialization part from __make_request. It
will be used by the following blk_ordered_reimpl.
: reimplement handling of barrier request
Reimplement handling of barrier requests.
* Flexible handling to deal with various capabilities of
* Retry support for falling back.
* Tagged queues which don't support ordered tag can do ordered.
: update SCSI to use new blk_ordered
All ordered request related stuff delegated to HLD. Midlayer
now doens't deal with ordered setting or prepare_flush
callback. sd.c updated to deal with blk_queue_ordered
setting. Currently, ordered tag isn't used as SCSI midlayer
cannot guarantee request ordering.
: add FUA support to SCSI disk
Add FUA support to SCSI disk.
: update libata to use new blk_ordered
Reflect changes in SCSI midlayer and updated to use new
ordered request implementation
: add FUA support to libata
Add FUA support to libata.
: update IDE to use new blk_ordered
Update IDE to use new blk_ordered.
: add FUA support to IDE
Add FUA support to IDE
: I/O barrier documentation
I/O barrier documentation
[ End of patch descriptions ]