Instead of just passing 'EIO' for any I/O error we should be
notifying the upper layers with more details about the cause
of this error.

Update the possible I/O errors to:

- ENOLINK: Link failure between host and target
- EIO: Retryable I/O error
- EREMOTEIO: Non-retryable I/O error
- EBADE: I/O error restricted to the I_T_L nexus

'Retryable' in this context means that an I/O error _might_ be
restricted to the I_T_L nexus (vulgo: path), so retrying on another
nexus / path might succeed.

'Non-retryable' in general refers to a target failure, so this
error will always be generated regardless of the I_T_L nexus
it was send on.

I/O errors restricted to the I_T_L nexus might be retried
on another nexus / path, but they should _not_ be queued
if no paths are available.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NMike Snitzer <snitzer@redhat.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

The current code in scsi_eh_target_reset() has an off by one error
that actually sends spurious extra resets.  Since there's no real need
to reset the targets in numerical order, simply chunk up the command
recovery list doing target resets and pulling matching targets out of
the list (that also makes the loop O(N) instead of O(N^2).

[mike christie found and fixed a list_splice -> list_splice_init problem]

Reported-by: Hillf Danton<dhillf@gmail.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

The error handler is using the test cmd->serial_number == 0 in the
abort routines to signal that the command to be aborted has already
completed normally.  This design was to close a race window in the
original error handler where a command could go through the normal
completion routines after it timed out but before error handling was
started.

Mike Anderson pointed out that when we converted our timeout and
softirq completions, we picked up atomicity here because the block
layer now mediates this with the REQ_ATOM_COMPLETE flag and guarantees
that *either* the command times out or our done routine is called, but
ensures we can't get both occurring.  That makes the serial number
zero check redundant and it can be removed.
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

Move the mid-layer's ->queuecommand() invocation from being locked
with the host lock to being unlocked to facilitate speeding up the
critical path for drivers who don't need this lock taken anyway.

The patch below presents a simple SCSI host lock push-down as an
equivalent transformation.  No locking or other behavior should change
with this patch.  All existing bugs and locking orders are preserved.

Additionally, add one parameter to queuecommand,
	struct Scsi_Host *
and remove one parameter from queuecommand,
	void (*done)(struct scsi_cmnd *)

Scsi_Host* is a convenient pointer that most host drivers need anyway,
and 'done' is redundant to struct scsi_cmnd->scsi_done.

Minimal code disturbance was attempted with this change.  Most drivers
needed only two one-line modifications for their host lock push-down.
Signed-off-by: NJeff Garzik <jgarzik@redhat.com>
Acked-by: NJames Bottomley <James.Bottomley@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

REQ_HARDBARRIER is dead now, so remove the leftovers.  What's left
at this point is:

 - various checks inside the block layer.
 - sanity checks in bio based drivers.
 - now unused bio_empty_barrier helper.
 - Xen blockfront use of BLKIF_OP_WRITE_BARRIER - it's dead for a while,
   but Xen really needs to sort out it's barrier situaton.
 - setting of ordered tags in uas - dead code copied from old scsi
   drivers.
 - scsi different retry for barriers - it's dead and should have been
   removed when flushes were converted to FS requests.
 - blktrace handling of barriers - removed.  Someone who knows blktrace
   better should add support for REQ_FLUSH and REQ_FUA, though.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

commit 5f91bb05
Author: Michael Reed <mdr@sgi.com>
Date:   Mon Aug 10 11:59:28 2009 -0500

    [SCSI] reservation conflict after timeout causes device to be taken offline

Flipped us from always returning failed to always returning success in
the name of fixing the problem where reservation conflict returns from
test unit ready cause the device always to be taken offline.
Unfortuantely, it also introduced a problem whereby for commands other
than test unit ready, the eh dispatcher thinks they succeeded when
reservation conflict is returned, whereas in reality they failed.  Fix
this by only returning success for the test unit ready case.
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

When the transport is busy and we're sending an EH command drivers
occasionally return 'BUSY'. As this in most cases is the TUR
command sent as part of the error recovery this is a sure way
to make the error recovery escalate. Returning 'NEEDS_RETRY'
here will just retry the TUR command and eventually abort the
original command, thus making error handling far smoother.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

scsi-ml uses REQ_TYPE_BLOCK_PC for flush requests from file
systems. The definition of REQ_TYPE_BLOCK_PC is that we don't retry
requests even when we can (e.g. UNIT ATTENTION) and we send the
response to the callers (then the callers can decide what they want).
We need a workaround such as the commit
77a42297 to retry BLOCK_PC flush
requests. We will need the similar workaround for discard requests too
since SCSI-ml handle them as BLOCK_PC internally.

This uses REQ_TYPE_FS for flush requests from file systems instead of
REQ_TYPE_BLOCK_PC.

scsi-ml retries only REQ_TYPE_FS requests that have data to
transfer when we can retry them (e.g. UNIT_ATTENTION). However, we
also need to retry REQ_TYPE_FS requests without data because the
callers don't.

This also changes scsi_check_sense() to retry all the REQ_TYPE_FS
requests when appropriate. Thanks to scsi_noretry_cmd(),
REQ_TYPE_BLOCK_PC requests don't be retried as before.

Note that basically, this reverts the commit
77a42297 since now we use REQ_TYPE_FS
for flush requests.
Signed-off-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

Remove all the trivial wrappers for the cmd_type and cmd_flags fields in
struct requests.  This allows much easier grepping for different request
types instead of unwinding through macros.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

This patch (as1398b) adds runtime PM support to the SCSI layer.  Only
the machanism is provided; use of it is up to the various high-level
drivers, and the patch doesn't change any of them.  Except for sg --
the patch expicitly prevents a device from being runtime-suspended
while its sg device file is open.

The implementation is simplistic.  In general, hosts and targets are
automatically suspended when all their children are asleep, but for
them the runtime-suspend code doesn't actually do anything.  (A host's
runtime PM status is propagated up the device tree, though, so a
runtime-PM-aware lower-level driver could power down the host adapter
hardware at the appropriate times.)  There are comments indicating
where a transport class might be notified or some other hooks added.

LUNs are runtime-suspended by calling the drivers' existing suspend
handlers (and likewise for runtime-resume).  Somewhat arbitrarily, the
implementation delays for 100 ms before suspending an eligible LUN.
This is because there typically are occasions during bootup when the
same device file is opened and closed several times in quick
succession.

The way this all works is that the SCSI core increments a device's
PM-usage count when it is registered.  If a high-level driver does
nothing then the device will not be eligible for runtime-suspend
because of the elevated usage count.  If a high-level driver wants to
use runtime PM then it can call scsi_autopm_put_device() in its probe
routine to decrement the usage count and scsi_autopm_get_device() in
its remove routine to restore the original count.

Hosts, targets, and LUNs are not suspended while they are being probed
or removed, or while the error handler is running.  In fact, a fairly
large part of the patch consists of code to make sure that things
aren't suspended at such times.

[jejb: fix up compile issues in PM config variations]
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

If the user accidentally changes LUN mappings or it occurs
due to a bug, then it can cause data corruption that can take
months and months to track down. This patch adds a log
message when getting REPORT_LUNS_DATA_CHANGED and it adds
a generic message for other Unit Attentions with asc == 0x3f.

We are working on adding support for handling of these errors,
but I think until then we should at least log a message so
tracking down problems as a result of one of these changes
is a little easier.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

There's nastyness in the way we currently handle barriers (and
discards): They're effectively filesystem commands, but they get
processed as BLOCK_PC commands.  Unfortunately BLOCK_PC commands are
taken by SCSI to be SG_IO commands and the issuer expects to see and
handle any returned errors, however trivial.  This leads to a huge
problem, because the block layer doesn't expect this to happen and any
trivially retryable error on a barrier causes an immediate I/O error
to the filesystem.

The only real way to hack around this is to take the usual class of
offending errors (unit attentions) and make them all retryable in the
case of a REQ_HARDBARRIER.  A correct fix would involve a rework of
the entire block and SCSI submit system, and so is out of scope for a
quick fix.

Cc: Hannes Reinecke <hare@suse.de>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

Signed-off-by: NXiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Signed-off-by: NTomohiro Kusumi <kusumi.tomohiro@jp.fujitsu.com>
Signed-off-by: NKei Tokunaga <tokunaga.keiich@jp.fujitsu.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

If the scsi eh is running and then a FC LLD calls
fc_remote_port_delete, the SCSI commands sent from the eh will fail.
To prevent this, a FC LLD can call fc_block_scsi_eh from the eh
callback, blocking the eh thread until the dev_loss_tmo fires or the
remote port is available again.

If (e.g. for a multipathing setup) the dev_loss_tmo is set to a very
large value, thus preventing the scsi device removal , the scsi eh can
block for a long time. For multipathing, the fast_io_fail_tmo is then
set to a low value to detect path problems sooner.

This patch introduces a new return code FAST_IO_FAIL. The function
fc_block_scsi_eh now returns FAST_IO_FAIL when the fast_io_fail_tmo
fires. This indicates that the LLD terminated all pending I/O requests
and there are no more pending SCSI commands for the scsi eh to wait
for. This return code can be passed back to the scsi eh to stop the
escalation and finish the recovery process for this device.
Signed-off-by: NChristof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h

percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files.  percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.

percpu.h -> slab.h dependency is about to be removed.  Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability.  As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.

  http://userweb.kernel.org/~tj/misc/slabh-sweep.py

The script does the followings.

* Scan files for gfp and slab usages and update includes such that
  only the necessary includes are there.  ie. if only gfp is used,
  gfp.h, if slab is used, slab.h.

* When the script inserts a new include, it looks at the include
  blocks and try to put the new include such that its order conforms
  to its surrounding.  It's put in the include block which contains
  core kernel includes, in the same order that the rest are ordered -
  alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
  doesn't seem to be any matching order.

* If the script can't find a place to put a new include (mostly
  because the file doesn't have fitting include block), it prints out
  an error message indicating which .h file needs to be added to the
  file.

The conversion was done in the following steps.

1. The initial automatic conversion of all .c files updated slightly
   over 4000 files, deleting around 700 includes and adding ~480 gfp.h
   and ~3000 slab.h inclusions.  The script emitted errors for ~400
   files.

2. Each error was manually checked.  Some didn't need the inclusion,
   some needed manual addition while adding it to implementation .h or
   embedding .c file was more appropriate for others.  This step added
   inclusions to around 150 files.

3. The script was run again and the output was compared to the edits
   from #2 to make sure no file was left behind.

4. Several build tests were done and a couple of problems were fixed.
   e.g. lib/decompress_*.c used malloc/free() wrappers around slab
   APIs requiring slab.h to be added manually.

5. The script was run on all .h files but without automatically
   editing them as sprinkling gfp.h and slab.h inclusions around .h
   files could easily lead to inclusion dependency hell.  Most gfp.h
   inclusion directives were ignored as stuff from gfp.h was usually
   wildly available and often used in preprocessor macros.  Each
   slab.h inclusion directive was examined and added manually as
   necessary.

6. percpu.h was updated not to include slab.h.

7. Build test were done on the following configurations and failures
   were fixed.  CONFIG_GCOV_KERNEL was turned off for all tests (as my
   distributed build env didn't work with gcov compiles) and a few
   more options had to be turned off depending on archs to make things
   build (like ipr on powerpc/64 which failed due to missing writeq).

   * x86 and x86_64 UP and SMP allmodconfig and a custom test config.
   * powerpc and powerpc64 SMP allmodconfig
   * sparc and sparc64 SMP allmodconfig
   * ia64 SMP allmodconfig
   * s390 SMP allmodconfig
   * alpha SMP allmodconfig
   * um on x86_64 SMP allmodconfig

8. percpu.h modifications were reverted so that it could be applied as
   a separate patch and serve as bisection point.

Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: NTejun Heo <tj@kernel.org>
Guess-its-ok-by: NChristoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>

Current FC HBA queue_depth ramp up code depends on last queue
full time. The sdev already  has last_queue_full_time field to
track last queue full time but stored value is truncated by
last four bits.

So this patch updates last_queue_full_time without truncating
last 4 bits to store full value and then updates its only
current usages in scsi_track_queue_full to ignore last four bits
to keep current usages same while also use this field
in added ramp up code.

Adds scsi_handle_queue_ramp_up to ramp up queue_depth on
successful completion of IO. The scsi_handle_queue_ramp_up will
do ramp up on all luns of a target, just same as ramp down done
on all luns on a target.

The ramp up is skipped in case the change_queue_depth is not
supported by LLD or already reached to added max_queue_depth.

Updates added max_queue_depth on every new update to default
queue_depth value.

The ramp up is also skipped if lapsed time since either last
queue ramp up or down is less than LLD specified
queue_ramp_up_period.

Adds queue_ramp_up_period to sysfs but only if change_queue_depth
is supported since ramp up and queue_ramp_up_period is needed only
in case change_queue_depth is supported first.

Initializes queue_ramp_up_period to 120HZ jiffies as initial
default value, it is same as used in existing lpfc and qla2xxx.

-v2
 Combined all ramp code into this single patch.

-v3
 Moves max_queue_depth initialization after slave_configure is
called from after slave_alloc calling done. Also adjusted
max_queue_depth check to skip ramp up if current queue_depth
is >= max_queue_depth.

-v4
 Changes sdev->queue_ramp_up_period unit to ms when using sysfs i/f
to store or show its value.
Signed-off-by: NVasu Dev <vasu.dev@intel.com>
Tested-by: NChristof Schmitt <christof.schmitt@de.ibm.com>
Tested-by: NGiridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

This has scsi-ml call the change_queue_depth functions when
we get a QUEUE_FULL. It will only change the queue depth if
change_queue_depth is set because the LLD may have to
modify some internal resources, so I thought this would
be the safest route.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>

-v2
Limits change_queue_depth to only all luns of target by adding
channel check while iterating for all luns of Scsi_Host. This is
same as currently qla2xxx FC HBA does on QUEUE_FULL event.
Signed-off-by: NVasu Dev <vasu.dev@intel.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

A target reset when I/O is ongoing might result
an eventual device offline, as scsi_eh_completed_normally()
might return ADD_TO_MLQUEUE in addition to the
advertised SUCCESS, FAILED, and NEEDS_RETRY.

Which is unfortunate as scsi_send_eh_cmnd() will
therefore map ADD_TO_MLQUEUE to FAILED instead of
the more appropriate NEEDS_RETRY.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

An IBM tape drive failed to complete a PERSISTENT RESERVE IN within the scsi
cmd timeout.  Error recovery was initiated and it sequenced from abort through
taking the tape drive offline.

The device was taken offline because it repeatedly responded to the TUR command
issued by error recovery with a RESERVATION CONFLICT status.  The tape drive
was reserved to another system.  This is perfectly legitimate response to TUR,
and is one that an escalation of recovery is unlikely to clear.  Further,
escalation of recovery can have undesirable side effects on the operation of
tape drives shared with other initiators.

Instead of escalating recovery, error recovery should treat the RESERVATION
CONFLICT response to the TUR as a good status, giving the issuer of the
command the opportunity to handle the timeout and reservation conflict.
Signed-off-by: NMichael reed <mdr@sgi.com>
Signed-off-by: NJames Bottomley <James.Bottomley@suse.de>

The Documentation is incorrect (we removed some functions referred to), and
none of the bug warnings now apply.  Additionally remove the spurious check on
the return from blk_get_request() which can't fail if __GFP_WAIT is passed in.
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Signed-off-by: NBartlomiej Zolnierkiewicz <bzolnier@gmail.com>
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>

Typo fix.
Signed-off-by: NFrederik Schwarzer <schwarzerf@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

Make sure the control flow in scsi_times_out makes sense.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

This patch improves handling of TASK ABORTED status by Linux SCSI
mid-layer. Currently, command returned with this status considered
failed and returned to upper layers. It leads to additional error
recovery load on file systems and block layer, which sometimes can
cause undesired side effects, like I/O errors and file systems
corruptions. See http://lkml.org/lkml/2008/11/1/38, for instance.

From other side, TASK ABORTED status is returned by SCSI target if the
corresponding command was aborted by another initiator and the target
has TAS bit set in the control mode page. So, in the majority of cases
commands with TASK ABORTED status should be simply retried. In other
cases, maybe_retry path will not retry if no retries are allowed.

This patch implement suggestion by James Bottomley from
http://marc.info/?l=linux-scsi&m=121932916906009&w=2.
Signed-off-by: NVladislav Bolkhovitin <vst@vlnb.net>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

...and the list of recent breakage goes on and on, this time
it's 242f9dcb (block: unify request timeout handling)
which broke it.
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: NJens Axboe <jens.axboe@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

scsi_eh_try_stu() was still using the timeout parameter in the device
which is now not set (i.e. zero filled) meaning that it waited no time
at all for the start unit command to complete (leading the routine to
conclude failure every time).  This lead to a 2.6.27 regression:

http://bugzilla.kernel.org/show_bug.cgi?id=12120

Where firewire devices that were non spec compliant wouldn't spin up.

Fix this by using the block queue timeout value instead.
Reported-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Drivers want to be able to return DID_TRANSPORT_DISRUPTED and
have it do the right thing for commands like tape and passthrouh
as far as retries go. The LLDs previously used DID_BUS_BUSY or DID_ERROR
which followed the cmd->retries limit, but DID_TRANSPORT_DISRUPTED
was skipping that check so it could have caused a problem with tape
commands.

This patch has DID_TRANSPORT_DISRUPTED check the cmd->retries/cmd->allowed.
Signed-off-by: NMike Christie <mchristi@redhat.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

There's a target reset bug.

This loop:

	for (id = 0; id <= shost->max_id; id++) {

Never terminates if shost->max_id is set to ~0, like aic94xx does.

It's also pretty inefficient since you mostly have compact target
numbers, but the max_id can be very high.  The best way would be to
sort the recovery list by target id and skip them if they're equal,
but even a worst case O(N^2) traversal is probably OK here, so fix it
by finding the next highest target number (assuming n+1) and
terminating when there isn't one.

Cc: Mike Christie <michaelc@cs.wisc.edu>
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>

Currently, if there is a transport problem the iscsi drivers will return
outstanding commands (commands being exeucted by the driver/fw/hw) with
DID_BUS_BUSY and block the session so no new commands can be queued.
Commands that are caught between the failure handling and blocking are
failed with DID_IMM_RETRY or one of the scsi ml queuecommand return values.
When the recovery_timeout fires, the iscsi drivers then fail IO with
DID_NO_CONNECT.

For fcp, some drivers will fail some outstanding IO (disk but possibly not
tape) with DID_BUS_BUSY or DID_ERROR or some other value that causes a retry
and hits the scsi_error.c failfast check, block the rport, and commands
caught in the race are failed with DID_IMM_RETRY. Other drivers, may
hold onto all IO and wait for the terminate_rport_io or dev_loss_tmo_callbk
to be called.

The following patches attempt to unify what upper layers will see drivers
like multipath can make a good guess. This relies on drivers being
hooked into their transport class.

This first patch just defines two new host byte errors so drivers can
return the same value for when a rport/session is blocked and for
when the fast_io_fail_tmo fires.

The idea is that if the LLD/class detects a problem and is going to block
a rport/session, then if the LLD wants or must return the command to scsi-ml,
then it can return it with DID_TRANSPORT_DISRUPTED. This will requeue
the IO into the same scsi queue it came from, until the fast io fail timer
fires and the class decides what to do.

When using multipath and the fast_io_fail_tmo fires then the class
can fail commands with DID_TRANSPORT_FAILFAST or drivers can use
DID_TRANSPORT_FAILFAST in their terminate_rport_io callbacks or
the equivlent in iscsi if we ever implement more advanced recovery methods.
A LLD, like lpfc, could continue to return DID_ERROR and then it will hit
the normal failfast path, so drivers do not have fully be ported to
work better. The point of the patches is that upper layers will
not see a failure that could be recovered from while the rport/session is
blocked until fast_io_fail_tmo/recovery_timeout fires.

V3
Remove some comments.
V2
Fixed patch/diff errors and renamed DID_TRANSPORT_BLOCKED to
DID_TRANSPORT_DISRUPTED.
V1
initial patch.
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>

Change scsi_check_sense HARDWARE_ERROR check to return ADD_TO_MLQUEUE
if device->retry_hwerror is set to allow retries to occur without
restriction of blk_noretry_request check.
Signed-off-by: NMike Anderson <andmike@linux.vnet.ibm.com>
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>

This patch (as1116) fixes a bug in scsi_eh_prep_cmnd() and
scsi_eh_restore_cmnd().  These routines are supposed to save any
values they change and restore them later, but someone forgot to
save & restore scmd->underflow.

This fixes part of the problem reported in Bugzilla #9638.

[jejb: fix up rejections around DIF/DIX]
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

If initiator or target reject the I/O due to DIF errors there is no
point in retrying.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Controllers that support DMA of protection information must be told
explicitly how to handle the I/O.  The controller has no knowledge of
the protection capabilities of the target device so this information
must be passed in the scsi_cmnd.

 - The protection operation tells the HBA whether to generate, strip or
   verify protection information.

 - The protection type tells the HBA which layout the target is
   formatted with.  This is necessary because the controller must be
   able to correctly interpret the included protection information in
   order to verify it.

 - When a scsi_cmnd is reused for error handling the protection
   operation must be cleared and saved while error handling is in
   progress.

 - prot_op and prot_type are placed in an existing hole in scsi_cmnd
   and don't cause the structure to grow.
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Some of the storage devices (that can be accessed through multiple paths),
do need some special handling for
	1. Activating the passive path of the storage access.
	2. Decode and handle the special sense codes returned by the devices.
	3. Handle the I/Os being sent to the passive path, especially
           during the device probe time.
when accessed through multiple paths.

As of today this special device handling is done at the dm-multipath
layer using dm-handlers. That works well for (1); for (2) to be handled
at dm layer, scsi sense information need to be exported from SCSI to dm-layer,
which is not very attractive; (3) cannot be done at all at the dm layer.

Device handler has been moved to SCSI mainly to handle (2) and (3) properly.
Signed-off-by: NChandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: NMike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

 - struct scsi_cmnd had a 16 bytes command buffer of its own.
   This is an unnecessary duplication and copy of request's
   cmd. It is probably left overs from the time that scsi_cmnd
   could function without a request attached. So clean that up.

 - Once above is done, few places, apart from scsi-ml, needed
   adjustments due to changing the data type of scsi_cmnd->cmnd.

 - Lots of drivers still use MAX_COMMAND_SIZE. So I have left
   that #define but equate it to BLK_MAX_CDB. The way I see it
   and is reflected in the patch below is.
   MAX_COMMAND_SIZE - means: The longest fixed-length (*) SCSI CDB
                      as per the SCSI standard and is not related
                      to the implementation.
   BLK_MAX_CDB.     - The allocated space at the request level

 - I have audit all ISA drivers and made sure none use ->cmnd in a DMA
   Operation. Same audit was done by Andi Kleen.

(*)fixed-length here means commands that their size can be determined
   by their opcode and the CDB does not carry a length specifier, (unlike
   the VARIABLE_LENGTH_CMD(0x7f) command). This is actually not exactly
   true and the SCSI standard also defines extended commands and
   vendor specific commands that can be bigger than 16 bytes. The kernel
   will support these using the same infrastructure used for VARLEN CDB's.
   So in effect MAX_COMMAND_SIZE means the maximum size command
   scsi-ml supports without specifying a cmd_len by ULD's
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJames Bottomley <James.Bottomley@HansenPartnership.com>

Any path needs to call it to initialize the request.

This is a preparation for large command support, which needs to
initialize the request in a proper way (that is, just doing a memset()
will not work).
Signed-off-by: NFUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: NJens Axboe <jens.axboe@oracle.com>