Because of the terrible structuring of scsi-bidi-commands
it breaks some of the life time rules of a scsi-command.
It is now not allowed to free up the block-request before
cleanup and partial deallocation of the scsi-command. (Which
is not so for none bidi commands)

The right fix to this problem would be to make bidi command
a first citizen by allocating a scsi_sdb pointer at scsi command
just like cmd->prot_sdb. The bidi sdb should be allocated/deallocated
as part of the get/put_command (Again like the prot_sdb) and the
current decoupling of scsi_cmnd and blk-request should be kept.

For now make sure scsi_release_buffers() is called before the
call to blk_end_request_all() which might cause the suicide of
the block requests. At best the leak of bidi buffers, at worse
a crash, as there is a race between the existence of the bidi_request
and the free of the associated bidi_sdb.

The reason this was never hit before is because only OSD has the potential
of doing asynchronous bidi commands. (So does bsg but it is never used)
And OSD clients just happen to do all their bidi commands synchronously, up
until recently.

CC: Stable Tree <stable@kernel.org>
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

A thin provisioned device may temporarily be out of sufficient
allocation units to fulfill a write request.  In that case it will
return a space allocation in progress error.  Wait a bit and retry the
write.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

Stanse found a potential NULL dereference in scsi_kill_request.

Instead of triggering BUG() in 'if (unlikely(cmd == NULL))' branch,
the kernel will Oops earlier on cmd dereference.

Move the dereferences after the if.
Signed-off-by: NJiri Slaby <jirislaby@gmail.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

When the Integrity check is done in scsi_io_completion it will
set error to -EILSEQ. However, at this point error is no longer
used, and blk_end_request_err has -EIO hardcoded.

It looks like there was just porting mistake with this patch
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=3e695f89c5debb735e4ff051e9e58d8fb4e95110
and we meant to send error upwards, so this patch changes the hard
coded EIO to the error variable.

I have only boot tested this patch.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Acked-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

Update scsi_io_completion() such that it only fails requests till the
next error boundary and retry the leftover.  This enables block layer
to merge requests with different failfast settings and still behave
correctly on errors.  Allow merge of requests of different failfast
settings.

As SCSI is currently the only subsystem which follows failfast status,
there's no need to worry about other block drivers for now.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Niel Lambrechts <niel.lambrechts@gmail.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

When a request fails we print the sense data but not the actual command
that failed.  Add a printout of the operation + CDB for failed commands.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

If a SCSI ULD driver sets blk_queue_prep_rq(), it should clean it
up itself on remove(), and not from the bus callbacks. This
removes the need to hook into bus->remove(), which should not
be used at the same time as driver->remove().

[jejb: fix sdkp initialisation problem due to mismerge]
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NKay Sievers <kay.sievers@vrfy.org>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

scsi timeout on two or more devices may cause extremely long execution
time for user applications because SDEV_OFFLINE state is changed to
SDEV_RUNNING state during scsi error recovery procedures triggered by
a bus reset or a host reset of scsi LLD, and scsi timeout can happens
on the same devices many times.

This happens because scsi_internal_device_unblock() changes device's
state to SDEV_RUNNING even if a device in other states than SDEV_BLOCK,
while the following two transitions are required in this function.

  SDEV_BLOCK -> SDEV_RUNNING
  SDEV_CREATED_BLOCK -> SDEV_CREATED

Otherwise, it returns -EINVAL.
Signed-off-by: NTakahiro Yasui <tyasui@redhat.com>
[matthew@wil.cx: supplied rewritten base for patch]
Signed-off-by: NMatthew Wilcox <matthew@wil.cx>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

The last request completion cleanup in scsi_lib left an unused
this_count variable in scsi_io_completion().
(It was used before in a code segment that now uses blk_end_request_all())
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Acked-by: NTejun Heo <tj@kernel.org>
Acked-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Commit c3a4d78c introduced
rq->data_len and converted residual count users to it.  While
converting, it mistakenly converted scsi_end_request() to finish
requests with residual count when it wants to do is fully complete the
request.  Fix it by using blk_end_request_all() instead.

This bug was spotted by Boaz Harrosh.
Signed-off-by: NTejun Heo <tj@kernel.org>
Spotted-by: NBoaz Harrosh <bharrosh@panasas.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Let's use blk_end_request_all() instead of blk_end_bidi_request().
Signed-off-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Till now block layer allowed two separate modes of request execution.
A request is always acquired from the request queue via
elv_next_request().  After that, drivers are free to either dequeue it
or process it without dequeueing.  Dequeue allows elv_next_request()
to return the next request so that multiple requests can be in flight.

Executing requests without dequeueing has its merits mostly in
allowing drivers for simpler devices which can't do sg to deal with
segments only without considering request boundary.  However, the
benefit this brings is dubious and declining while the cost of the API
ambiguity is increasing.  Segment based drivers are usually for very
old or limited devices and as converting to dequeueing model isn't
difficult, it doesn't justify the API overhead it puts on block layer
and its more modern users.

Previous patches converted all block low level drivers to dequeueing
model.  This patch completes the API transition by...

* renaming elv_next_request() to blk_peek_request()

* renaming blkdev_dequeue_request() to blk_start_request()

* adding blk_fetch_request() which is combination of peek and start

* disallowing completion of queued (not started) requests

* applying new API to all LLDs

Renamings are for consistency and to break out of tree code so that
it's apparent that out of tree drivers need updating.

[ Impact: block request issue API cleanup, no functional change ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Mike Miller <mike.miller@hp.com>
Cc: unsik Kim <donari75@gmail.com>
Cc: Paul Clements <paul.clements@steeleye.com>
Cc: Tim Waugh <tim@cyberelk.net>
Cc: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Laurent Vivier <Laurent@lvivier.info>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: Adrian McMenamin <adrian@mcmen.demon.co.uk>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Cc: Borislav Petkov <petkovbb@googlemail.com>
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: Pierre Ossman <drzeus@drzeus.cx>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Markus Lidel <Markus.Lidel@shadowconnect.com>
Cc: Stefan Weinhuber <wein@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

With the previous changes, the followings are now guaranteed for all
requests in any valid state.

* blk_rq_sectors() == blk_rq_bytes() >> 9
* blk_rq_cur_sectors() == blk_rq_cur_bytes() >> 9

Clean up accessor usages.  Notable changes are

* nbd,i2o_block: end_all used instead of explicit byte count
* scsi_lib: unnecessary conditional on request type removed

[ Impact: cleanup ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Paul Clements <paul.clements@steeleye.com>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: Markus Lidel <Markus.Lidel@shadowconnect.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

With recent unification of fields, it's now guaranteed that
rq->data_len always equals blk_rq_bytes().  Convert all non-IDE direct
users to accessors.  IDE will be converted in a separate patch.

Boaz: spotted incorrect data_len/resid_len conversion in osd.

[ Impact: convert direct rq->data_len usages to blk_rq_bytes() ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NSergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: Markus Lidel <Markus.Lidel@shadowconnect.com>
Cc: Darrick J. Wong <djwong@us.ibm.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: Boaz Harrosh <bharrosh@panasas.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

With recent cleanups, there is no place where low level driver
directly manipulates request fields.  This means that the 'hard'
request fields always equal the !hard fields.  Convert all
rq->sectors, nr_sectors and current_nr_sectors references to
accessors.

While at it, drop superflous blk_rq_pos() < 0 test in swim.c.

[ Impact: use pos and nr_sectors accessors ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NGeert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Tested-by: NGrant Likely <grant.likely@secretlab.ca>
Acked-by: NGrant Likely <grant.likely@secretlab.ca>
Tested-by: NAdrian McMenamin <adrian@mcmen.demon.co.uk>
Acked-by: NAdrian McMenamin <adrian@mcmen.demon.co.uk>
Acked-by: NMike Miller <mike.miller@hp.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Cc: Borislav Petkov <petkovbb@googlemail.com>
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Paul Clements <paul.clements@steeleye.com>
Cc: Tim Waugh <tim@cyberelk.net>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Dario Ballabio <ballabio_dario@emc.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: unsik Kim <donari75@gmail.com>
Cc: Laurent Vivier <Laurent@lvivier.info>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Implement accessors - blk_rq_pos(), blk_rq_sectors() and
blk_rq_cur_sectors() which return rq->hard_sector, rq->hard_nr_sectors
and rq->hard_cur_sectors respectively and convert direct references of
the said fields to the accessors.

This is in preparation of request data length handling cleanup.

Geert	: suggested adding const to struct request * parameter to accessors
Sergei	: spotted error in patch description

[ Impact: cleanup ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NGeert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Acked-by: NStephen Rothwell <sfr@canb.auug.org.au>
Tested-by: NGrant Likely <grant.likely@secretlab.ca>
Acked-by: NGrant Likely <grant.likely@secretlab.ca>
Ackec-by: NSergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Cc: Borislav Petkov <petkovbb@googlemail.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

rq->data_len served two purposes - the length of data buffer on issue
and the residual count on completion.  This duality creates some
headaches.

First of all, block layer and low level drivers can't really determine
what rq->data_len contains while a request is executing.  It could be
the total request length or it coulde be anything else one of the
lower layers is using to keep track of residual count.  This
complicates things because blk_rq_bytes() and thus
[__]blk_end_request_all() relies on rq->data_len for PC commands.
Drivers which want to report residual count should first cache the
total request length, update rq->data_len and then complete the
request with the cached data length.

Secondly, it makes requests default to reporting full residual count,
ie. reporting that no data transfer occurred.  The residual count is
an exception not the norm; however, the driver should clear
rq->data_len to zero to signify the normal cases while leaving it
alone means no data transfer occurred at all.  This reverse default
behavior complicates code unnecessarily and renders block PC on some
drivers (ide-tape/floppy) unuseable.

This patch adds rq->resid_len which is used only for residual count.

While at it, remove now unnecessasry blk_rq_bytes() caching in
ide_pc_intr() as rq->data_len is not changed anymore.

Boaz	: spotted missing conversion in osd
Sergei	: spotted too early conversion to blk_rq_bytes() in ide-tape

[ Impact: cleanup residual count handling, report 0 resid by default ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Cc: Borislav Petkov <petkovbb@googlemail.com>
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: Mike Miller <mike.miller@hp.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Doug Gilbert <dgilbert@interlog.com>
Cc: Mike Miller <mike.miller@hp.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: Darrick J. Wong <djwong@us.ibm.com>
Cc: Pete Zaitcev <zaitcev@redhat.com>
Cc: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Now that all block request data transfer is done via bio, rq->data
isn't used.  Kill it.

While at it, make the roles of rq->special and buffer clear.

[ Impact: drop now unncessary field from struct request ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Boaz Harrosh <bharrosh@panasas.com>

There are many [__]blk_end_request() call sites which call it with
full request length and expect full completion.  Many of them ensure
that the request actually completes by doing BUG_ON() the return
value, which is awkward and error-prone.

This patch adds [__]blk_end_request_all() which takes @rq and @error
and fully completes the request.  BUG_ON() is added to to ensure that
this actually happens.

Most conversions are simple but there are a few noteworthy ones.

* cdrom/viocd: viocd_end_request() replaced with direct calls to
  __blk_end_request_all().

* s390/block/dasd: dasd_end_request() replaced with direct calls to
  __blk_end_request_all().

* s390/char/tape_block: tapeblock_end_request() replaced with direct
  calls to blk_end_request_all().

[ Impact: cleanup ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Mike Miller <mike.miller@hp.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>

We cannot call blk_plug_device from scsi_target_queue_ready
because the q lock is not held. And we do not need to call
it from there because when we return 0, the scsi_request_fn
not_ready handling will plug the queue for us if needed.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

We have a problem with recovered error handling in that any command
which goes down as BLOCK_PC but which returns a sense code of RECOVERED
ERROR gets completed with -EIO.  For actual SG_IO commands, this doesn't
matter at all, since the error return code gets dropped in favour of
req->errors which contain the SCSI completion code.

However, if this command is part of the block system, then it will pay
attention to the returned error code.  In particularly if a SYNCHRONIZE
CACHE from a barrier command completes with RECOVERED ERROR, the
resulting -EIO on the barrier causes block to error the request and
return it to the filesystem.  Fix this by converting the -EIO for
recovered error to zero, plus remove the printing of this from sd and sr
so the message isn't double printed.
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

No one uses scsi_execute_async with data transfer now. We can remove
scsi_req_map_sg.

Only scsi_eh_lock_door uses scsi_execute_async. scsi_eh_lock_door
doesn't handle sense and the callback. So we can remove
scsi_io_context too.
Signed-off-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Instead of terminating after five retries, commands terminated by
ABORTED_COMMAND sense are retrying forever.  The problem was
introduced by:

commit b60af5b0
Author: Alan Stern <stern@rowland.harvard.edu>
Date:   Mon Nov 3 15:56:47 2008 -0500

    [SCSI] simplify scsi_io_completion()

Which introduced an error whereby ABORTED_COMMAND now gets erroneously
retried in scsi_io_completion.  Fix this by returning the behaviour
back to the default no retry.
Reported-by: NSitsofe Wheeler <sitsofe@yahoo.com>
Tested-by: NSitsofe Wheeler <sitsofe@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Andrew Vaszquez said:
> There's a problem that is causing commands returned by the LLD with
> a DID_RESET status to be reissued with cleared cmd->sdb data which
> in our tests are manifesting in firmware detected overruns.  Here's
> a snippet of a READ_10 scsi_cmnd upon completion by the storage

The problem is caused by:

commit b60af5b0
Author: Alan Stern <stern@rowland.harvard.edu>
Date:   Mon Nov 3 15:56:47 2008 -0500

    [SCSI] simplify scsi_io_completion()

Because scsi_release_buffers() is called before commands that go
through the ACTION_RETRY and ACTION_DELAYED_RETRY legs are requeued.
However, they're not re-prepared, so nothing ever reallocates the
buffer resources to them.  Fix this by releasing the buffers only if
we're not going to go down these legs (but scsi_release_buffers() on
all legs including two in scsi_end_request(); this latter needs a
special version __scsi_release_buffers() because the final one can be
called after the request has been freed, so the bidi test in
scsi_release_buffers(), which touches the request has to be skipped).
Reported-by: NAndrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

patch

commit b60af5b0
Author: Alan Stern <stern@rowland.harvard.edu>
Date:   Mon Nov 3 15:56:47 2008 -0500

    [SCSI] simplify scsi_io_completion()

broke DIX error handling.  Also, we are now using EILSEQ to indicate
integrity errors to the upper layers (as opposed to regular EIO
failures).  This allows filesystems to inspect buffers and decide
whether to retry the I/O.  Update scsi_io_completion() accordingly.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

A bug was introduced by

commit b60af5b0
Author: Alan Stern <stern@rowland.harvard.edu>
Date:   Mon Nov 3 15:56:47 2008 -0500

    [SCSI] simplify scsi_io_completion()
 
because the simplification uses scsi_queue_insert().  The problem with
this function is that it expects to be called from the completion path
while the command is still outstanding, so it decrements the device
and host busy counts to do the requeue.  The problem is that
scsi_io_completion() is a path executed well after these counts have
*already* been decremented, leading to a double decrement if the
command goes down any error path leading to ACTION_DELAYED_RETRY.

The fix is to allow a private function __scsi_queue_insert() with a
flag to say whether the busy counters should be decremented.  This is
made static to scsi_lib.c to discourage other use.
Reported-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This patch (as1191) adds a missing "default" case in
scsi_io_completion(), thereby fixing an "uninitialized variable"
error.  It also adds a missing newline to a log entry.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

scsi_execute() and scsi_execute_req() discard the residual length
information. Some callers need it. This adds residual argument
(optional) to scsi_execute and scsi_execute_req.
Signed-off-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This patch (as1142b) consolidates a lot of repetitious code in
scsi_io_completion().  It also fixes a few comments.  Most
importantly, however, it clearly distinguishes among the three sorts
of retries that can be done when a command fails to complete:

	Unprepare the request and resubmit it, so that a new
	command will be created for it.

	Requeue the request directly so that it will be retried
	immediately using the same command.

	Requeue the request so that it will be retried following
	a short delay.

	Complete the remainder of the request with an I/O error.

[jejb: Updates
     1. For several error conditions, we would now print the sense twice
        in slightly different ways, so unify the location of sense
        printing.
     2. I added more descriptions to actual failure conditions for
        better debugging
     3. according to spec, ABORTED_COMMAND is supposed to be retried
        (except on DIF failure).  Our old behaviour of erroring it looks
        to be a bug.
     4. I'd prefer not to default initialise the action variable because
        that ensures that every leg of the error handler has an
        associated action and the compiler will warn if someone later
        accidentally misses one or removes one.
]
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

It's called under that lock everywhere else and it does alter the
request state, so it should be.

This one occurance in scsi_requeue_command() could open a window where
req->special is set to NULL while the requests is going through either
timeout or completion processing leading to NULL pointer derefs of the
sort complained of in bugzillas 12020 and 12195.
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Close possible infinite loop with interrupts off when devices are
added back to the starved list.

Fixes: http://bugzilla.kernel.org/show_bug.cgi?id=11898

Reported-by: <alex.shi@intel.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This patch implements q->lld_busy_fn() for scsi mid layer to export
its busy state for request stacking drivers.

For efficiency, no lock is taken to check the busy state of
shost/starget/sdev, since the returned value is not guaranteed and
may be changed after request stacking drivers call the function,
regardless of taking lock or not.

When scsi can't dispatch I/Os anymore and needs to kill I/Os
(e.g. !sdev), scsi needs to return 'not busy'.
Otherwise, request stacking drivers may hold requests forever.
Signed-off-by: NKiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: NJun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This patch refactors the busy checking codes of scsi_device,
Scsi_Host and scsi_target.  There should be no functional change.

This is a preparation for another patch which exports scsi's busy
state to the block layer for request stacking drivers.
Signed-off-by: NKiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: NJun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

On Tue, 12 Aug 2008 15:08:14 +0200
Giuliano Pochini <pochini@shiny.it> wrote:

> Fujitsu magneto-optical drive, Adaptec 29160 and
> Linux Jay 2.6.26 #7 SMP Sun Aug 10 18:34:22 CEST 2008 ppc 7455, altivec supported PowerMac3,6 GNU/Linux
>
> When I insert a disk and I mount it, scsi_test_unit_ready() is called and
> the do-while loop gets sshdr->sense_key == UNIT_ATTENTION in the first
> cycle and 0 in the second one. So the if below misses the UNIT_ATTENTION
> and sdev->changed = 1 is not executed. At this point bad things can
> happen... I'm not sure how to fix this. Any clue ?

The problem is essentially caused by us eating UNIT_ATTENTION
conditions in scsi_test_unit_ready().  Fix by updating the ->changed
flag when this happens if the media is removable.

[pochini@shiny.it: updates to tidy up patch]
Signed-off-by: NGiuliano Pochini <pochini@shiny.it>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This checks the errors the scsi-ml determined were retryable
and returns if we should fast fail it based on the request
fail fast flags.

Without the patch, drivers like lpfc, qla2xxx and fcoe would return
DID_ERROR for what it determines is a temporary communication problem.
There is no loss of connectivity at that time and the driver thinks
that it would be fast to retry at the driver level. SCSI-ml will however
sees fast fail on the request and DID_ERROR and will fast fail the io.
This will then cause dm-multipath to fail the path and possibley switch
target controllers when we should be retrying at the scsi layer.

We also were fast failing device errors to dm multiapth when
unless the scsi_dh modules think otherwis we want to retry at
the scsi layer because multipath can only retry the IO like scsi
should have done. multipath is a little dumber though because it
does not what the error was for and assumes that it should fail
the paths.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

SCSI-ml manages the queueing limits for the device and host, but
does not do so at the target level. However something something similar
can come in userful when a driver is transitioning a transport object to
the the blocked state, becuase at that time we do not want to queue
io and we do not want the queuecommand to be called again.

The patch adds code similar to the exisiting SCSI_ML_*BUSY handlers.
You can now return SCSI_MLQUEUE_TARGET_BUSY when we hit
a transport level queueing issue like the hw cannot allocate some
resource at the iscsi session/connection level, or the target has temporarily
closed or shrunk the queueing window, or if we are transitioning
to the blocked state.

bnx2i, when they rework their firmware according to netdev
developers requests, will also need to be able to limit queueing at this
level. bnx2i will hook into libiscsi, but will allocate a scsi host per
netdevice/hba, so unlike pure software iscsi/iser which is allocating
a host per session, it cannot set the scsi_host->can_queue and return
SCSI_MLQUEUE_HOST_BUSY to reflect queueing limits on the transport.

The iscsi class/driver can also set a scsi_target->can_queue value which
reflects the max commands the driver/class can support. For iscsi this
reflects the number of commands we can support for each session due to
session/connection hw limits, driver limits, and to also reflect the
session/targets's queueing window.

Changes:
v1 - initial patch.
v2 - Fix scsi_run_queue handling of multiple blocked targets.
Previously we would break from the main loop if a device was added back on
the starved list. We now run over the list and check if any target is
blocked.
v3 - Rediff for scsi-misc.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Right now SCSI and others do their own command timeout handling.
Move those bits to the block layer.

Instead of having a timer per command, we try to be a bit more clever
and simply have one per-queue. This avoids the overhead of having to
tear down and setup a timer for each command, so it will result in a lot
less timer fiddling.
Signed-off-by: NMike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>

Brian King <brking@linux.vnet.ibm.com> reported that fibre channel
devices can oops during scanning if their ports block (because the
device goes from CREATED -> BLOCK -> RUNNING rather than CREATED ->
BLOCK -> CREATED).

Fix this by adding a new state: CREATED_BLOCK which can only transition
back to CREATED and disallow the CREATED -> BLOCK transition.  Now both
the created and blocked states that the mid-layer recognises can include
CREATED_BLOCK.
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Sometimes, particularly for USB devices with the last sector bug,
requests get completed in chunks.  There's a bug in this in that if
one of the chunks gets an error, we complete that chunk with an error
but never move on to the remaining ones, leading to the request
hanging (because it's not fully completed).

Fix this by completing all remaining chunks if an error is encountered.

Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

[jejb: fixed up a ton of missed conversions.

 All of you are on notice this has happened, driver trees will now
 need to be rebased]
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Cc: SCSI List <linux-scsi@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>