We should call the device handler prep_fn for all TYPE_FS requests,
not just simple read/write calls that are handled by the disk driver.

Restructure the common I/O code to call the prep_fn handler and zero
out the CDB, and just leave the call to scsi_init_io to the ULDs.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.de>

Until now the per-command transfer length has exclusively been gated by
the max_sectors parameter in the scsi_host template. Given that the size
of this parameter has been bumped to an unsigned int we have to be
careful not to exceed the target device's capabilities.

If the if the device specifies a Maximum Transfer Length in the Block
Limits VPD we'll use that value. Otherwise we'll use 0xffffffff for
devices that have use_16_for_rw set and 0xffff for the rest. We then
combine the chosen disk limit with max_sectors in the host template. The
smaller of the two will be used to set the max_hw_sectors queue limit.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NEwan D. Milne <emilne@redhat.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

In init_sd function, if kmem_cache_create or mempool_create_slab_pools
calls fail, the error will not be correclty reported because
class_register previously set the value of err to 0.
Signed-off-by: NClément Calmels <clement.calmels@free.fr>
Reviewed-by: NEwan D. Milne <emilne@redhat.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

This is a fix for commit 39c60a09

  "sd: fix array cache flushing bug causing performance problems"

We must notify the block layer via q->flush_flags after a temporary change
of the cache_type to write through.  Without this, a SYNCHRONIZE CACHE
command will still be generated.  This patch factors out a helper that
can be called from sd_revalidate_disk and cache_type_store.
Signed-off-by: NVaughan Cao <vaughan.cao@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

This change makes the scsi disk driver handle the requests whose
transfer length is greater than 0xffff with READ_16 or WRITE_16.

However, this is a preparation for extending the data type of
max_sectors in struct Scsi_Host and scsi_host_template.  So, it is
impossible to happen this condition for now, because SCSI low-level
drivers can not specify max_sectors greater than 0xffff due to the
data type limitation.
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Some buggy JMicron USB-ATA bridges don't know how to translate the FUA
bit in READs or WRITEs.  This patch adds an entry in unusual_devs.h
and a blacklist flag to tell the sd driver not to use FUA.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Reported-by: NMichael Büsch <m@bues.ch>
Tested-by: NMichael Büsch <m@bues.ch>
Acked-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>
CC: Matthew Dharm <mdharm-usb@one-eyed-alien.net>
CC: <stable@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

There is an error with the medium access timeout feature of the sd driver. The
sdkp->medium_access_timed_out value is reset to zero in sd_done() in the wrong
place.  Currently it is reset to zero only when a command returns sense data.
This can result in cases where the medium access check falsely triggers from
timed out commands which are hours or days apart.

For example, an I/O command times out and is aborted.  It then retries and
succeeds.  But with no sense data generated and returned, the
medium_access_timed_out value is not reset.  If no sd command returns sense
data, then the next command to time out (however far in time from the first
failure) will trigger the medium access timeout and put the device offline.

The resetting of sdkp->medium_access_timed_out should occur before the check
for sense data.

To reproduce using scsi_debug, use SCSI_DEBUG_OPT_TIMEOUT or
SCSI_DEBUG_OPT_MAC_TIMEOUT to force an I/O command to timeout.  Then, remove
the opt value so the I/O will succeed on retry.  Perform more I/O as desired.
Finally, repeat the process to make a new I/O command time out.  Without the
patch, the device will be marked offline even though many I/O commands have
succeeded between the 2 instances of timed out commands.
Signed-off-by: NDavid Jeffery <djeffery@redhat.com>
Reviewed-by: NEwan D. Milne <emilne@redhat.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Instead of letting the ULD play games with the prep_fn move back to
the model of a central prep_fn with a callback to the ULD.  This
already cleans up and shortens the code by itself, and will be required
to properly support blk-mq in the SCSI midlayer.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NNicholas Bellinger <nab@linux-iscsi.org>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NHannes Reinecke <hare@suse.de>

Store the pointer to the page there, so we can always safely
reference it from end_io context where ->bio may have been
cleared.
Signed-off-by: NJens Axboe <axboe@fb.com>

This was used in the olden days, back when onions were proper
yellow. Basically it mapped to the current buffer to be
transferred. With highmem being added more than a decade ago,
most drivers map pages out of a bio, and rq->buffer isn't
pointing at anything valid.

Convert old style drivers to just use bio_data().

For the discard payload use case, just reference the page
in the bio.
Signed-off-by: NJens Axboe <axboe@fb.com>

async_schedule() sd resume work to allow disks and other devices to
resume in parallel.

This moves the entirety of scsi_device resume to an async context to
ensure that scsi_device_resume() remains ordered with respect to the
completion of the start/stop command.  For the duration of the resume,
new command submissions (that do not originate from the scsi-core) will
be deferred (BLKPREP_DEFER).

It adds a new ASYNC_DOMAIN_EXCLUSIVE(scsi_sd_pm_domain) as a container
of these operations.  Like scsi_sd_probe_domain it is flushed at
sd_remove() time to ensure async ops do not continue past the
end-of-life of the sdev.  The implementation explicitly refrains from
reusing scsi_sd_probe_domain directly for this purpose as it is flushed
at the end of dpm_resume(), potentially defeating some of the benefit.
Given sdevs are quiesced it is permissible for these resume operations
to bleed past the async_synchronize_full() calls made by the driver
core.

We defer the resolution of which pm callback to call until
scsi_dev_type_{suspend|resume} time and guarantee that the callback
parameter is never NULL.  With this in place the type of resume
operation is encoded in the async function identifier.

There is a concern that async resume could trigger PSU overload.  In the
enterprise, storage enclosures enforce staggered spin-up regardless of
what the kernel does making async scanning safe by default.  Outside of
that context a user can disable asynchronous scanning via a kernel
command line or CONFIG_SCSI_SCAN_ASYNC.  Honor that setting when
deciding whether to do resume asynchronously.

Inspired by Todd's analysis and initial proposal [2]:
https://01.org/suspendresume/blogs/tebrandt/2013/hard-disk-resume-optimization-simpler-approach

Cc: Len Brown <len.brown@intel.com>
Cc: Phillip Susi <psusi@ubuntu.com>
[alan: bug fix and clean up suggestion]
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Suggested-by: NTodd Brandt <todd.e.brandt@linux.intel.com>
[djbw: kick all resume work to the async queue]
Signed-off-by: NDan Williams <dan.j.williams@intel.com>

Messages about discovered disk properties are only printed once unless
they are found to have changed. Errors encountered during mode sense,
however, are printed every time we revalidate.

Quiesce mode sense errors so they are only printed during the first
scan.

[jejb: checkpatch fixes]
Bugzilla: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=733565Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Evidently some wacky USB-ATA bridges don't recognize the SYNCHRONIZE
CACHE command, as shown in this email thread:

	http://marc.info/?t=138978356200002&r=1&w=2

The fact that we can't tell them to drain their caches shouldn't
prevent the system from going into suspend.  Therefore sd_sync_cache()
shouldn't return an error if the device replies with an Invalid
Command ASC.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Reported-by: NSven Neumann <s.neumann@raumfeld.com>
Tested-by: NDaniel Mack <zonque@gmail.com>
CC: <stable@vger.kernel.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

do_div() is meant for divisions of 64-bit number by 32-bit numbers.
Passing 64-bit divisor types caused issues in the past on 32-bit platforms,
cfr. commit ea077b1b ("m68k: Truncate base
in do_div()").

As scsi_device.sector_size is unsigned (int), factor should be unsigned
int, too.
Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Commit 18a4d0a2
(Handle disk devices which can not process medium access commands)
was introduced to offline any device which cannot process medium
access commands.
However, commit 3eef6257
(Reduce error recovery time by reducing use of TURs) reduced
the number of TURs by sending it only on the first failing
command, which might or might not be a medium access command.
So in combination this results in an erratic device offlining
during EH; if the command where the TUR was sent upon happens
to be a medium access command the device will be set offline,
if not everything proceeds as normal.

This patch moves the check to the final test, eliminating
this problem.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Some host adapters do not pass commands through to the target disk
directly. Instead they provide an emulated target which may or may not
accurately report its capabilities. In some cases the physical device
characteristics are reported even when the host adapter is processing
commands on the device's behalf. This can lead to adapter firmware hangs
or excessive I/O errors.

This patch disables WRITE SAME for devices connected to host adapters
that provide an emulated target. Driver writers can disable WRITE SAME
by setting the no_write_same flag in the host adapter template.

[jejb: fix up rejections due to eh_deadline patch]
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Cc: stable@kernel.org
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

For immutable biovecs, we'll be introducing a new bio_iovec() that uses
our new bvec iterator to construct a biovec, taking into account
bvec_iter->bi_bvec_done - this patch updates existing users for the new
usage.

Some of the existing users really do need a pointer into the bvec array
- those uses are all going to be removed, but we'll need the
functionality from immutable to remove them - so for now rename the
existing bio_iovec() -> __bio_iovec(), and it'll be removed in a couple
patches.
Signed-off-by: NKent Overstreet <kmo@daterainc.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Ed L. Cashin" <ecashin@coraid.com>
Cc: Alasdair Kergon <agk@redhat.com>
Cc: dm-devel@redhat.com
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>

We have officially run out of flags in a 32-bit space. Extend it
to 64-bit even on 32-bit archs.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Rather than having a separate constant for specifying the timeout on FLUSH
operations, use the basic I/O timeout value that is already configurable
on a per target basis to derive the FLUSH timeout. Looking at the current
definitions of these timeout values, the FLUSH operation is supposed to have
a value that is twice the normal timeout value. This patch preserves this
relationship while leveraging the flexibility of specifying the I/O timeout.

Based on a prior patch by KY Srinivasan <kys@microsoft.com>
Reviewed-by: NKY Srinivasan <kys@microsoft.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

It makes no sense to flush the cache of a device without medium.
Errors during suspend must be handled according to their causes.
Errors due to missing media or unplugged devices must be ignored.
Errors due to devices being offlined must also be ignored.
The error returns must be modified so that the generic layer
understands them.

[jejb: fix up whitespace and other formatting problems]
Signed-off-by: NOliver Neukum <oneukum@suse.de>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Somehow older areca firmware versions have issues with
scsi_get_vpd_page() and a large buffer, the firmware
seems to crash and the scsi error-handler will start endless
recovery retries.
Limiting the buf-size to 64-bytes fixes this issue with older
firmware versions (<1.49 for my controller).

Fixes a regression with areca controllers and older firmware versions
introduced by commit: 66c28f97Reported-by: NNix <nix@esperi.org.uk>
Tested-by: NNix <nix@esperi.org.uk>
Signed-off-by: NBernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Cc: stable@vger.kernel.org # delay inclusion for 2 months for testing
Acked-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Sujit has found a race condition that would make q->nr_pending
unbalanced, it occurs as Sujit explained:

"
sd_probe_async() ->
	add_disk() ->
		disk_add_event() ->
			schedule(disk_events_workfn)
	sd_revalidate_disk()
	blk_pm_runtime_init()
return;

Let's say the disk_events_workfn() calls sd_check_events() which tries
to send test_unit_ready() and because of sd_revalidate_disk() trying to
send another commands the test_unit_ready() might be re-queued as the
tagged command queuing is disabled.

So the race condition is -

Thread 1 			  |		Thread 2
sd_revalidate_disk()		  |	sd_check_events()
...nr_pending = 0 as q->dev = NULL|	scsi_queue_insert()
blk_runtime_pm_init()		  | 	blk_pm_requeue_request() ->
				  |	nr_pending = -1 since
				  |	q->dev != NULL
"

The problem is, the test_unit_ready request doesn't get counted the
first time it is queued, so the later decrement of q->nr_pending in
blk_pm_requeue_request makes it unbalanced.

Fix this by calling blk_pm_runtime_init before add_disk so that all
requests initiated there will all be counted.
Signed-off-by: NAaron Lu <aaron.lu@intel.com>
Reported-and-tested-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Cc: stable@vger.kernel.org
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

This patch fixes an out-of-bounds error in sd_read_cache_type(), found
by Google's AddressSanitizer tool.  When the loop ends, we know that
"offset" lies beyond the end of the data in the buffer, so no Caching
mode page was found.  In theory it may be present, but the buffer size
is limited to 512 bytes.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Reported-by: NDmitry Vyukov <dvyukov@google.com>
CC: <stable@vger.kernel.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

The dev_attrs field of struct class is going away soon, dev_groups
should be used instead.  This converts the scsi disk class code to use
the correct field.

It required some functions to be moved around to place the show and
store functions next to each other, the old order seemed to make no
sense at all.
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

sd_prep_fn will allocate a larger CDB for the command via mempool_alloc
for devices using DIF type 2 protection.  This CDB was being freed
in sd_done, which results in a kernel crash if the command is retried
due to a UNIT ATTENTION.  This change moves the code to free the larger
CDB into sd_unprep_fn instead, which is invoked after the request is
complete.

It is no longer necessary to call scsi_print_command separately for
this case as the ->cmnd will no longer be NULL in the normal code path.

Also removed conditional test for DIF type 2 when freeing the larger
CDB because the protection_type could have been changed via sysfs while
the command was executing.
Signed-off-by: NEwan D. Milne <emilne@redhat.com>
Acked-by: NMartin K. Petersen <martin.petersen@oracle.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Calling dev_set_name with a single paramter causes it to be handled as a
format string.  Many callers are passing potentially dynamic string
content, so use "%s" in those cases to avoid any potential accidents,
including wrappers like device_create*() and bdi_register().
Signed-off-by: NKees Cook <keescook@chromium.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

SATA drives located behind a SAS controller would incorrectly receive
WRITE SAME commands. Tweak the heuristics so that:

 - If REPORT SUPPORTED OPERATION CODES is provided we will use that to
   choose between WRITE SAME(16), WRITE SAME(10) and disabled. This also
   fixes an issue with the old code which would issue WRITE SAME(10)
   despite the command not being whitelisted in REPORT SUPPORTED
   OPERATION CODES.

 - If REPORT SUPPORTED OPERATION CODES is not provided we will fall back
   to WRITE SAME(10) unless the device has an ATA Information VPD page.
   The assumption is that a SATL which is smart enough to implement
   WRITE SAME would also provide REPORT SUPPORTED OPERATION CODES.

To facilitate the new heuristics scsi_report_opcode() has been modified
to so we can distinguish between "operation not supported" and "RSOC not
supported".
Reported-by: NH. Peter Anvin <hpa@zytor.com>
Tested-by: NBernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Commit 39c60a09 '[SCSI] sd: fix array cache flushing bug causing
performance problems' added temp as a pointer to "temporary " and used
sizeof(temp) - 1 as its length.  But sizeof(temp) is the size of the
pointer, not the size of the string constant.  Change temp to a static
array so that sizeof() does what was intended.
Signed-off-by: NBen Hutchings <ben@decadent.org.uk>
Cc: stable@vger.kernel.org
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

When multipathed systems run into an all-paths-down scenario
all devices might be dropped, too. This causes 'del_gendisk'
to be called, which will unregister the kobj_map->probe()
function for all disk device numbers.
When the device comes back the default ->probe() function
is run which will call __request_module(), which will
deadlock.
As 'del_gendisk' typically does _not_ trigger a module unload
the default ->probe() function is pointless anyway.
This patch implements a dummy ->probe() function, which will
just return NULL if the disk is not registered.
This will avoid the deadlock. Plus it'll speed up device
scanning.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

The value passed is 0 in all but "it can never happen" cases (and those
only in a couple of drivers) *and* it would've been lost on the way
out anyway, even if something tried to pass something meaningful.
Just don't bother.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Uses block layer runtime pm helper functions in
scsi_runtime_suspend/resume for devices that take advantage of it.

Remove scsi_autopm_* from sd open/release path and check_events path.
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Signed-off-by: NAaron Lu <aaron.lu@intel.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

With the introduction of REQ_PM, modify sd's runtime suspend operation
functions to use that flag so that the operations to put the device into
runtime suspended state(i.e. sync cache and stop device) will not affect
its runtime PM status.
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Signed-off-by: NAaron Lu <aaron.lu@intel.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Some arrays synchronize their full non volatile cache when the sd driver sends
a SYNCHRONIZE CACHE command.  Unfortunately, they can have Terrabytes of this
and we send a SYNCHRONIZE CACHE for every barrier if an array reports it has a
writeback cache.  This leads to massive slowdowns on journalled filesystems.

The fix is to allow userspace to turn off the writeback cache setting as a
temporary measure (i.e. without doing the MODE SELECT to write it back to the
device), so even though the device reported it has a writeback cache, the
user, knowing that the cache is non volatile and all they care about is
filesystem correctness, can turn that bit off in the kernel and avoid the
performance ruinous (and safety irrelevant) SYNCHRONIZE CACHE commands.

The way you do this is add a 'temporary' prefix when performing the usual
cache setting operations, so

echo temporary write through > /sys/class/scsi_disk/<disk>/cache_type
Reported-by: NRic Wheeler <rwheeler@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Update sd driver to use the callbacks defined in dev_pm_ops.

sd_freeze is NULL, the bus level callback has taken care of quiescing
the device so there should be nothing needs to be done here.
Consequently, sd_thaw is not needed here either.

suspend, poweroff and runtime suspend share the same routine sd_suspend,
which will sync flush and then stop the drive, this is the same as before.

resume, restore and runtime resume share the same routine sd_resume,
which will start the drive by putting it into active power state, this
is also the same as before.
Signed-off-by: NAaron Lu <aaron.lu@intel.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Acked-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

When device is runtime suspended, put it to stopped power state to save
some power.

This will also make the behaviour consistent with what the scsi_pm.c
thinks about sd as the comment says:
sd treats runtime suspend, system suspend and system hibernate identical.
With this patch, it is now identical.
And sd_shutdown will also do nothing when it finds the device has been
runtime suspended, if we do not spin down the disk in runtime suspend
by putting it into stopped power state, the disk will be shut down
incorrectly.
And the the same problem can be solved for runtime power off after
runtime suspended case by this change.

With the current runtime scheme for disk, it will only be runtime
suspended when no process opens the disk, so this shouldn't happen a
lot, which makes it acceptable to spin down the disk when runtime
suspended. If some day a more aggressive runtime scheme is used, like
the 'request based runtime pm for disk' that Alan Stern and Lin Ming
has been working, we can introduce some policy to control this. But for
now, make it simple and correct by spinning down the disk.
Signed-off-by: NAaron Lu <aaron.lu@intel.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Acked-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Force large capacity (> 0xFFFFFFFF blocks) drives to use READ/WRITE(16) instead
of READ/WRITE(10). Some(most/all?) USB enclosures do not like READ(10) commands
when a large capacity drive is installed. This issue was reported and discussed
here: http://marc.info/?l=linux-usb&m=135247705222324Signed-off-by: NJason J. Herne <hernejj@gmail.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

scsi_register_driver will register a prep_fn() function, which
in turn migh need to use the sd_cdp_pool for DIF.
Which hasn't been initialised at this point, leading to
a crash. So reshuffle the init_sd() and exit_sd() paths
to have the driver registered last.
Signed-off-by: NJoel D. Diaz <joeldiaz@us.ibm.com>
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Implement support for WRITE SAME(10) and WRITE SAME(16) in the SCSI disk
driver.

 - We set the default maximum to 0xFFFF because there are several
   devices out there that only support two-byte block counts even with
   WRITE SAME(16). We only enable transfers bigger than 0xFFFF if the
   device explicitly reports MAXIMUM WRITE SAME LENGTH in the BLOCK
   LIMITS VPD.

 - max_write_same_blocks can be overriden per-device basis in sysfs.

 - The UNMAP discovery heuristics remain unchanged but the discard
   limits are tweaked to match the "real" WRITE SAME commands.

 - In the error handling logic we now distinguish between WRITE SAME
   with and without UNMAP set.

The discovery process heuristics are:

 - If the device reports a SCSI level of SPC-3 or greater we'll issue
   READ SUPPORTED OPERATION CODES to find out whether WRITE SAME(16) is
   supported. If that's the case we will use it.

 - If the device supports the block limits VPD and reports a MAXIMUM
   WRITE SAME LENGTH bigger than 0xFFFF we will use WRITE SAME(16).

 - Otherwise we will use WRITE SAME(10) unless the target LBA is beyond
   0xFFFFFFFF or the block count exceeds 0xFFFF.

 - no_write_same is set for ATA, FireWire and USB.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NMike Snitzer <snitzer@redhat.com>
Reviewed-by: NJeff Garzik <jgarzik@redhat.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

Support requests with more than one bio payload for discards. The total
number of bytes to be discarded is stored in req->__data_len and used in
sd_done() to complete the I/O.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NMike Snitzer <snitzer@redhat.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

We set the capacity to zero when we discovered a device formatted with
an unknown DIF protection type. However, the read_capacity code would
override the capacity and cause the device to be enabled regardless.

Make sd_read_protection_type() return an error if the protection type is
unknown. Also prevent duplicate printk lines when the device is being
revalidated.
Reported-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>