Error handling in UFS driver is broken and resets the host controller
for fatal errors without re-initialization. Correct the fatal error
handling sequence according to UFS Host Controller Interface (HCI)
v1.1 specification.

o Processed requests which are completed w/wo error are reported to
  SCSI layer and any pending commands that are not started are aborted
  in the controller and re-queued into scsi mid-layer queue.

o Upon determining fatal error condition the host controller may hang
  forever until a reset is applied. Block SCSI layer for sending new
  requests and apply reset in a separate error handling work.

o SCSI is informed about the expected Unit-Attention exception from the
  device for the immediate command after a reset so that the SCSI layer
  take necessary steps to establish communication with the device.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Reviewed-by: NYaniv Gardi <ygardi@codeaurora.org>
Tested-by: NDolev Raviv <draviv@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

As of now SCSI initiated error handling is broken because,
the reset APIs don't try to bring back the device initialized and
ready for further transfers.

In case of timeouts, the scsi error handler takes care of handling aborts
and resets. Improve the error handling in such scenario by resetting the
device and host and re-initializing them in proper manner.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Reviewed-by: NYaniv Gardi <ygardi@codeaurora.org>
Tested-by: NDolev Raviv <draviv@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

There is a possible race condition in the hardware when the abort
command is issued to terminate the ongoing SCSI command as described
below:

- A bit in the door-bell register is set in the controller for a
  new SCSI command.
- In some rare situations, before controller get a chance to issue
  the command to the device, the software issued an abort command.
- If the device recieves abort command first then it returns success
  because the command itself is not present.
- Now if the controller commits the command to device it will be
  processed.
- Software thinks that command is aborted and proceed while still
  the device is processing it.
- The software, controller and device may go out of sync because of
  this race condition.

To avoid this, query task presence in the device before sending abort
task command so that after the abort operation, the command is guaranteed
to be non-existent in both controller and the device.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Reviewed-by: NYaniv Gardi <ygardi@codeaurora.org>
Tested-by: NDolev Raviv <draviv@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Currently, sending Task Management (TM) command to the card might
be broken in some scenarios as listed below:

Problem: If there are more than 8 TM commands the implementation
         returns error to the caller.
Fix:     Wait for one of the slots to be emptied and send the command.

Problem: Sometimes it is necessary for the caller to know the TM service
         response code to determine the task status.
Fix:     Propogate the service response to the caller.

Problem: If the TM command times out no proper error recovery is
         implemented.
Fix:     Clear the command in the controller door-bell register, so that
         further commands for the same slot don't fail.

Problem: While preparing the TM command descriptor, the task tag used
         should be unique across SCSI/NOP/QUERY/TM commands and not the
	 task tag of the command which the TM command is trying to manage.
Fix:     Use a unique task tag instead of task tag of SCSI command.

Problem: Since the TM command involves H/W communication, abruptly ending
         the request on kill interrupt signal might cause h/w malfunction.
Fix:     Wait for hardware completion interrupt with TASK_UNINTERRUPTIBLE
         set.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Reviewed-by: NYaniv Gardi <ygardi@codeaurora.org>
Tested-by: NDolev Raviv <draviv@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Make undeclared functions static and declare exported symbols
to suppress warnings from sparse tool.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Fix many warnings with incorrect endian assumptions
which makes the code unportable to new architectures.

The UFS specification defines the byte order as big-endian
for UPIU structure and little-endian for the host controller
transfer/task management descriptors.
Signed-off-by: NSujit Reddy Thumma <sthumma@codeaurora.org>
Acked-by: NVinayak Holikatti <vinholikatti@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

rescan_hba_mode was defined as a u8 so could never be less than zero:

        rescan_hba_mode = hpsa_hba_mode_enabled(h);
        if (rescan_hba_mode < 0)
                goto out;
Signed-off-by: NJoe Handzik <joseph.t.handzik@hp.com>
Signed-off-by: NStephen M. Cameron <scameron@beardog.cce.hp.com>
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The sun3 drivers suffer from a whole bunch of duplicated code. Fix this
by following the g_NCR5380_mmio example. (Notionally, sun3_scsi relates to
sun3_scsi_vme in the same way that g_NCR5380 relates to g_NCR5380_mmio.)

Dead code is also removed: we now have working debug macros so
SUN3_SCSI_DEBUG is undesirable. Dead code within #ifdef OLD_DMA is also
dropped, consistent with sun3_scsi_vme.c.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Move the #include "NCR5380.h" out of the sun3_scsi.h header file and into
the driver .c files, like all the other NCR5380 drivers in the tree.

This improves uniformity and reduces the depth of nested includes. The
sequence of #include's, #define's and #if's no longer does my head in.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Remove the unused (and divergent) debugging macro definitions from
the sun3_NCR5380 and atari_NCR5380 drivers. These drivers have been
converted to use the common macros in NCR5380.h.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Acked-by: NMichael Schmitz <schmitz@debian.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

All three NCR5380 core driver implementations share the same NCR5380.h
header file so they need to agree on certain macro definitions.

The flag bit used by the NDEBUG_MERGING macro in atari_NCR5380 and
sun3_NCR5380 collides with the bit used by NDEBUG_LISTS.

Moreover, NDEBUG_ABORT appears in NCR5380.c so it should be defined in
NCR5380.h rather than in each of the many drivers using that core.

An undefined NDEBUG_ABORT macro caused compiler errors and led to dodgy
workarounds in the core driver that can now be removed.
(See commits f566a576 and
185a7a1c.)

Move all of the NDEBUG_ABORT, NDEBUG_TAGS and NDEBUG_MERGING macro
definitions into NCR5380.h where all the other NDEBUG macros live.

Also, incorrect "#ifdef NDEBUG" becomes "#if NDEBUG" to fix the warning:
drivers/scsi/mac_scsi.c: At top level:
drivers/scsi/NCR5380.c:418: warning: 'NCR5380_print' defined but not used
drivers/scsi/NCR5380.c:459: warning: 'NCR5380_print_phase' defined but not used

The debugging code is now enabled when NDEBUG != 0.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

All NCR5380 drivers already include the NCR5380.h header. Better to
adopt those macros rather than have three variations on them.

Moreover, the macros in NCR5380.h are preferable because the atari_NCR5380
and sun3_NCR5380 versions are inflexible. For example, they can't accomodate
dprintk(NDEBUG_MAIN | NDEBUG_QUEUES, ...)

Replace the *_PRINTK macros from atari_NCR5380.h and sun3_NCR5380.h with
the equivalent macros from NCR5380.h.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Acked-by: NMichael Schmitz <schmitz@debian.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

All NCR5380 drivers already include the NCR5380.h header. Better to
adopt those macros rather than have three variations on them.

Moreover, the macros in NCR5380.h are preferable anyway: the atari_NCR5380
and sun3_NCR5380 versions are inflexible. For example, they can't accomodate
NCR5380_dprint(NDEBUG_MAIN | NDEBUG_QUEUES, ...)

Replace the NCR_PRINT* macros from atari_NCR5380.h and sun3_NCR5380.h with
the equivalent macros from NCR5380.h.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Acked-by: NMichael Schmitz <schmitz@debian.org>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

There are three implementations of the core NCR5380 driver and three sets
of debugging macro definitions. And all three implementations use the
NCR5380.h header as well.

Two of the definitions of the dprintk macro accept a variable argument list
whereas the third does not. Standardize on the variable argument list.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The change from cmd->target to cmd->device->id was apparently the purpose of
commit a7f251228390e87d86c5e3846f99a455517fdd8e in
kernel/git/tglx/history.git but some instances have been missed.

Also fix the "NDEBUG_LAST_WRITE_SENT" and "NDEBUG_ALL" typo's.

Also fix some format strings (%ul becomes %lu) that caused compiler warnings.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Only the NCR5380_dprint() macro should invoke the NCR5380_print() function.
That's why NCR5380.c only defines the function #if NDEBUG. Use the standard
macro.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

HOSTS_C is always undefined. There is no hosts.c anymore.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

BOARD_NORMAL is completely unused and BOARD_NCR53C400 is used only by
g_NCR5380 internally. Remove the unused definitions.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Add an option to only transfer half the data for every n-th command.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NDouglas Gilbert <dgilbert@interlog.com>
Reviewed-by: NNicholas Bellinger <nab@linux-iscsi.org>

Taken almost entirely from Nicholas Bellinger's scsi-mq conversion.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NNicholas Bellinger <nab@linux-iscsi.org>

In case of error, the bnx2fc_allocate_hash_table() didn't free
all the memory it allocated.
Signed-off-by: NMaurizio Lombardi <mlombard@redhat.com>
Acked-by: NEddie Wai <eddie.wai@broadcom.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

If bnx2fc_allocate_hash_table() for some reasons fails, it is possible that the
hash_tbl_segments or the hash_tbl_pbl pointers are NULL.
In this case bnx2fc_free_hash_table() will panic the system.

this patch also fixes a memory leak, the hash_tbl_segments pointer was never
freed.
Signed-off-by: NMaurizio Lombardi <mlombard@redhat.com>
Acked-by: NEddie Wai <eddie.wai@broadcom.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

hash_table_size is not used by the bnx2fc_free_hash_table() function.
Signed-off-by: NMaurizio Lombardi <mlombard@redhat.com>
Acked-by: NEddie Wai <eddie.wai@broadcom.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NJoe Handzik <joseph.t.handzik@hp.com>
Signed-off-by: NStephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

And while we're at it fix a magic number
Signed-off-by: NJoe Handzik <joseph.t.handzik@hp.com>
Signed-off-by: NStephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NScott Teel <scott.teel@hp.com>
Signed-off-by: NJoe Handzik <joseph.t.handzik@hp.com>
Signed-off-by: NStephen M. Cameron <scameron@beardog.cce.hp.com>
Cc: stable@vger.kernel.org
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Checking for a NULL return from a kzalloc call in hpsa_get_pdisk_of_ioaccel2.
Signed-off-by: NScott Teel <scott.teel@hp.com>
Signed-off-by: NJoe Handzik <joseph.t.handzik@hp.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Checking return value for the memory allocattion and freeing it
while exiting the function
Signed-off-by: NViswas G <Viswas.G@pmcs.com>
Signed-off-by: NSuresh Thiagarajan <Suresh.Thiagarajan@pmcs.com>
Reviewed-by: NTomas Henzl <thenzl@redhat.com>
Acked-by: NJack Wang <xjtuwjp@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Access to tgt->req_vq is strictly serialized by spin_lock
of tgt->tgt_lock, so the ACCESS_ONCE() isn't necessary.

smp_read_barrier_depends() in virtscsi_req_done was introduced
to order reading req_vq and decreasing tgt->reqs, but it isn't
needed now because req_vq is read from
scsi->req_vqs[vq->index - VIRTIO_SCSI_VQ_BASE] instead of
tgt->req_vq, so remove the unnecessary barrier.

Also remove related comment about the barrier.
Signed-off-by: NMing Lei <tom.leiming@gmail.com>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Add a memory barrier to ensure the valid bit is read before
any of the cqe payload is read. This fixes an issue seen
on Power where the cqe payload was getting loaded before
the valid bit. When this occurred, we saw an iotag out of
range error when a command completed, but since the iotag
looked invalid the command didn't get completed to scsi core.
Later we hit the command timeout, attempted to abort the command,
then waited for the aborted command to get returned. Since the
adapter already returned the command, we timeout waiting,
and end up escalating EEH all the way to host reset. This
patch fixes this issue.
Signed-off-by: NBrian King <brking@linux.vnet.ibm.com>
Signed-off-by: NJames Smart <james.smart@emulex.com>

 ---

 lpfc_sli.c |   21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NMichael Schmitz <schmitzmic@gmail.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

[Resend of earlier patch - added equivalent changes to sun3 NCR5380 code]

The abort/reset lowlevel return codes had changed with the new
error SCSI handling - update Atari and Sun3 NCR5380 drivers to reflect this.

Change reset handling for Atari to clear queues only, do not attempt
to call done() on each command aborted by the reset. The EH code
should do that for us. Queues _must_ be cleared, otherwise
atari_scsi_bus_reset will not release the ST-DMA lock, deadlocking
further error recovery.

Update the Sun3 NCR5380 driver as well - the Sun3 driver was
derived from the Atari one. Kudos to Finn Thain for the Sun3 part
and cleaning up the header files. After the header cleanup, the
initio.h include (!) can be dropped from sun3_scsi.h now.
Signed-off-by: NMichael Schmitz <schmitz@debian.org>
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Acked-by: NSam Creasey <sammy@sammy.net>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: James E.J. Bottomley <JBottomley@parallels.com>
Cc: linux-scsi@vger.kernel.org
Signed-off-by: NChristoph Hellwig <hch@lst.de>

The format strings for various printk()s make use of a temporary
variable that is declared 'static'. This is probably not intended,
so fix those.

Found in the PaX patch, written by the PaX Team.
Signed-off-by: NMathias Krause <minipli@googlemail.com>
Acked-by: NHannes Reinecke <hare@suse.de>
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>
Cc: PaX Team <pageexec@freemail.hu>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Minor fix for a message in the driver so that it matches the function name.
Signed-off-by: NMarek Vasut <marex@denx.de>
Cc: James Bottomley <JBottomley@Parallels.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-by: NBradley Grove <bgrove@attotech.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

bfa_fcb_pbc_vport_create() is called only from bfa_fcs_pbc_vport_init(),
that is called only from bfad_drv_start() with bfad_lock spinlock held.
So the patch replaces GFP_KERNEL with GFP_ATOMIC to avoid
sleeping in atomic spinlock context.

Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: NAlexey Khoroshilov <khoroshilov@ispras.ru>
Acked-by: NAnil Gurumurthy <anil.gurumurthy@qlogic.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NVikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NVikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Fix following smatch warning:-
 drivers/scsi/qla4xxx/ql4_os.c:1752 qla4xxx_get_ep_param() warn: variable dereferenced before check 'qla_ep' (see line 1745)
Signed-off-by: NVikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NChristoph Hellwig <hch@lst.de>

Signed-off-by: NVikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NChristoph Hellwig <hch@lst.de>