scsi: ufs: Use the resource-managed function to add devfreq device

This patch uses the resource-managed to add the devfreq device.  This
function will make it easy to handle the devfreq device.

- struct devfreq *devm_devfreq_add_device(struct device *dev,
  struct devfreq_dev_profile *profile,
  const char *governor_name,
  void *data);
Cc: Vinayak Holikatti <vinholikatti@gmail.com>
Cc: James E.J. Bottomley <jejb@linux.vnet.ibm.com>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: linux-scsi@vger.kernel.org
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fix spelling mistake in error message

Trivial fix to spelling mistake "operatio" to "operation" in critical
error message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: fix some spelling mistakes in message and comments

Trivial fixes, minor spelling mistakes in comments and in a KERN_INFO
message.

[mkp: fixed spelling mistake in patch description]

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: qcom: Don't free resource-managed kmalloc element

Host is allocated by managed kmalloc (devm_kmalloc). The
memory allocated with this function is automatically
freed on driver detach.
So, no need to make an exclusive free call over it.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: qcom: Add phy_exit call in hcd exit path

Do a phy_exit() over the ufs phy in the ufs qcom exit path
to de-initialize the phy.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Remove common layer phy exit callback

The common layer phy exit callback ufs_qcom_phy_exit()
calls phy_power_off() that has no meaning when phy_power_off()
callback is already registered with the phy provider and
the consumer makes use of the same.
Instead, add a no-op specific phy_exit() callback for now
to add the exit sequence at a later point.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs-qcom: phy/hcd: Refactoring phy clock handling

Add phy clock enable code to phy_power_on/off callbacks, and
remove explicit calls to enable these phy clocks from the
ufs-qcom hcd driver.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs-qmp-xx: Move clock and regulator init out of phy init

The phy init is meant to do phy initialization rather than
just getting the clock and regulator. Move these clock and
regulator get to probe(), to make room for actual phy
initialization sequence.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Remove unnecessary function declarations

Move the functions' definitions to remove unnecessary
declarations.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs-qmp-xx: Discard remove callback for drivers.

remove() callback does a phy_power_off() only over the phy,
and nothing else now.
The phy_power_off() over the generic phy is called from the phy
consumer, and phy provider driver should not explicitly need to
call any phy ops.
So discard the remove callback for qcom-ufs phy platform drivers.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Skip obtaining rx/tx_iface_clk for msm8996 based phy

The tx_iface_clk and rx_iface_clk no longer exist with UFS Phy
present on msm8996. So skip obtaining these clocks using
compatible match.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs-14nm: Add new compatible for msm8996 based phy

Add a new compatible string for 14nm ufs phy present on msm8996
chipset. This phy is bit different from the legacy 14nm ufs phy
in terms of the clocks that are needed to be handled in the driver.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Cleanup clock and regulator initialization

Different menthods pass around generic phy pointer to
extract device pointer. Instead, pass the device pointer
directly between function calls.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Use devm sibling of kstrdup for regulator names

This helps us in avoiding any requirement for kfree() operation
to be called exclusively over the allocated string pointer.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

phy: qcom-ufs: Remove unnecessary BUG_ON

BUG_ON() are not preferred in the driver, plus the variable
on which BUG_ON is asserted is already checked in the code
before passing.

Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: driver version upgrade

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: Do not set MPI2_TYPE_CUDA for JBOD FP path for FW which does not support JBOD sequence map

CC: stable@vger.kernel.org
Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

MAINTAINERS: Update megaraid maintainers list

Update MEGARAID drivers maintainers list.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: Send SYNCHRONIZE_CACHE for VD to firmware

Until now the megaraid_sas driver has reported successful completion on
SYNCHRONIZE_CACHE commands without sending them down to the controller.
The controller firmware has been responsible for taking care of flushing
disk caches for all drives that belong to a Virtual Disk at the time of
system reboot/shutdown.

There may have been a reason to avoid sending SYNCHRONIZE_CACHE to a VD
in the past but that no longer appears to be valid.

Older versions of MegaRaid firmware (Gen2 and Gen2.5) set the WCE bit
for Virtual Disks but the firmware does not report correct completion
status for a SYNCHRONIZE_CACHE command. As a result, we must use another
method to identify whether it is safe to send the command to the
controller. We use the canHandleSyncCache firmware flag in the scratch
pad register at offset 0xB4.

New SYNCHRONIZE_CACHE behavior:

IF 'JBOD'

Driver sends SYNCHRONIZE_CACHE command to the firmware
Firmware sends SYNCHRONIZE_CACHE to drive
Firmware obtains status from drive and returns same status back to driver

ELSEIF 'VirtualDisk'

IF firmware supports new API bit called canHandleSyncCache
Driver sends SYNCHRONIZE_CACHE command to the firmware
Firmware does not send SYNCHRONIZE_CACHE to drives
Firmware returns SUCCESS
ELSE
Driver does not send SYNCHRONIZE_CACHE command to the firmware
Driver return SUCCESS for that command
ENDIF
ENDIF

[mkp: edited patch description]

Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: smartpqi: switch to pci_alloc_irq_vectors

Which cleans up a lot of the MSI-X handling, and allows us to use the
PCI IRQ layer provided vector mapping, which we can then expose to
blk-mq.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Don Brace <don.brace@microsemi.com>
Tested-by: Don Brace <don.brace@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: allow LLDDs to expose the queue mapping to blk-mq

Just hand through the blk-mq map_queues method in the host template.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

blk-mq: export blk_mq_map_queues

This will allow SCSI to have a single blk_mq_ops structure that either
lets the LLDD map the queues to PCIe MSIx vectors or use the default.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: change device rails hpm mode ramp up sequence

When we are resuming the UFS device rails in HPM mode, we are first
powering on the VCC rail while VCCQ and VCCQ2 rails still being in LPM
mode. Some UFS devices may take VCC on event as hint that host wants UFS
device to be resumed and may start drawing more power from the
VCCQ/VCCQ2 rails (while they are still in LPM mode) causing voltage drop
on these rails. This change fixes this issue by bringing VCCQ & VCCQ2
rails out of LPM before powering on VCC rail.

Reviewed-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: suspend clock scaling at the start of suspend

Currently clock scaling is suspended only after the host and device are
put in low power mode but we should avoid clock scaling running after
UFS link is put in low power mode (hibern8). This change suspends clock
scaling before putting host/device in low power mode.

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufshcd: release resources if probe fails

If ufshcd pltfrm/pci driver's probe fails for some reason then ensure
that scsi host is released to avoid memory leak but managed memory
allocations (via devm_* calls) need not to be freed explicitly on probe
failure as memory allocated with these functions is automatically freed
on driver detach.

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: fix race between clock gating and devfreq scaling work

UFS devfreq clock scaling work may require clocks to be ON if it need to
execute some UFS commands hence it may request for clock hold before
issuing the command. But if UFS clock gating work is already running in
parallel, ungate work would end up waiting for the clock gating work to
finish and as clock gating work would also wait for the clock scaling
work to finish, we would enter in deadlock state. Here is the call trace
during this deadlock state:

Workqueue: devfreq_wq devfreq_monitor
__switch_to
__schedule
schedule
schedule_timeout
wait_for_common
wait_for_completion
flush_work
ufshcd_hold
ufshcd_send_uic_cmd
ufshcd_dme_get_attr
ufs_qcom_set_dme_vs_core_clk_ctrl_clear_div
ufs_qcom_clk_scale_notify
ufshcd_scale_clks
ufshcd_devfreq_target
update_devfreq
devfreq_monitor
process_one_work
worker_thread
kthread
ret_from_fork

Workqueue: events ufshcd_gate_work
__switch_to
__schedule
schedule
schedule_preempt_disabled
__mutex_lock_slowpath
mutex_lock
devfreq_monitor_suspend
devfreq_simple_ondemand_handler
devfreq_suspend_device
ufshcd_gate_work
process_one_work
worker_thread
kthread
ret_from_fork

Workqueue: events ufshcd_ungate_work
__switch_to
__schedule
schedule
schedule_timeout
wait_for_common
wait_for_completion
flush_work
__cancel_work_timer
cancel_delayed_work_sync
ufshcd_ungate_work
process_one_work
worker_thread
kthread
ret_from_fork

This change fixes this deadlock by doing this in devfreq work (devfreq_wq):
Try cancelling clock gating work. If we are able to cancel gating work
or it wasn't scheduled, hold the clock reference count until scaling is
in progress. If gate work is already running in parallel, let's skip
the frequecy scaling at this time and it will be retried once next scaling
window expires.

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Bump driver version as "14.101.00.00"

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Fix for Endianness issue.

Use le16_to_cpu only for accessing two byte data provided by controller.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Use the new MPI 2.6 32-bit Atomic Request Descriptors for SAS35 devices.

Support Atomic Request Descriptors for Ventura/SAS35 devices.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: set EEDP-escape-flags for SAS35 devices.

An UNMAP command on a PI formatted device will leave the Logical Block
Application Tag and Logical Block Reference Tag as all F's (for those LBAs
that are unmapped). To avoid IO errors if those LBAs are subsequently read
before they are written with valid tag fields, the MPI SCSI IO requests
need to set the EEDPFlags element EEDP Escape Mode field, Bits [7:6]
appropriately. A value of 2 should be set to disable all PI checks if the
Logical Block Application Tag is 0xFFFF for PI types 1 and 2.  A value
of 3 should be set to disable all PI checks if the Logical Block
Application Tag is 0xFFFF and the Logical Block Reference Tag is
0xFFFFFFFF for PI type 3.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Increased/Additional MSIX support for SAS35 devices.

For SAS35 devices MSIX vectors are inceased to 128 from 96. To support this
Reply post host index register count is increased to 16. Also variable
msix96_vector is replaced with combined_reply_queue and variable
combined_reply_index_count is added to set different values for SAS3 and
SAS35 devices.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Added Device ID's for SAS35 devices and updated MPI header.

Added Device ID's for SAS35 devices (Ventura, Crusader, Harpoon &
Tomcat) and updated mpi header file for the same. Also added
"is_gen35_ioc" to MPT3SAS_ADAPTER structure for identifying SAS35 adapters.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Bump driver version as "14.100.00.00"

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Remove unused macro "MPT_DEVICE_TLR_ON"

Removing macro "MPT_DEVICE_TLR_ON" defined in header file as its unused

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Implement device_remove_in_progress check in IOCTL path

When device missing event arrives, device_remove_in_progress bit will be
set and hence driver has to stop sending IOCTL commands.Now the check has
been added in IOCTL path to test device_remove_in_progress bit is set, if
so then IOCTL will be failed printing failure message.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Fix for incorrect numbers for MSIX vectors enabled when non RDPQ card is enumerated first.

No. of MSIX vectors supported = min (Total no. of CPU cores,
MSIX vectors supported by card)

when RDPQ is disabled "max_msix_vectors" module parameter which was
declared as global was set to '8' and hence if there are more than one card
in system among which if RDPQ disabled card is enumerated first then only 8
MSIX vectors was getting enabled for all the cards(including RDPQ enabled
card,which can support more than 8 MSIX vectors).

Used local variable instead of global variable ,if RDPQ is disabled this
local variable is set to '8' else it is set to "max_msix_vectors" (by
default this is set to -1, whose value can be set by user during driver
load time).So now regardless of whether RDPQ disabled card is enumerated
first or RDPQ enabled card is enumerated first , MSIX vectors enabled
depends on the cards capability.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: mpt3sas: Fix for improper info displayed in var log, while blocking or unblocking the device.

Return value and Device_handle Arguments passed in correct order
 to match with its format string.

Signed-off-by: Chaitra P B <chaitra.basappa@broadcom.com>
Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: advansys: fix build warning for PCI=n

The advansys probe function tries to handle both ISA and PCI cases, each
hidden in an #ifdef when unused. This leads to a warning indicating that
when PCI is disabled we could be using uninitialized data:

drivers/scsi/advansys.c: In function  advansys_board_found :
drivers/scsi/advansys.c:11036:5: error:  ret  may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/advansys.c:10928:28: note:  ret  was declared here
drivers/scsi/advansys.c:11309:8: error:  share_irq  may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/scsi/advansys.c:10928:6: note:  share_irq  was declared here

This cannot happen in practice because the hardware in question only
exists for PCI, but changing the code to just error out here is better
for consistency and avoids the warning.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Don't have fc_exch_find log errors on a new exchange

With the error message I added in "libfc: sanity check cpu number
extracted from xid" I didn't account for the fact that fc_exch_find is
called with FC_XID_UNKNOWN at the start of a new exchange if we are the
responder.

It doesn't come up with the initiator much, but that's basically every
exchange for a target.  By checking the xid for FC_XID_UNKNOWN first, we
not only prevent the erroneous error message, but skip the unnecessary
lookup attempt as well.

[mkp: applied by hand due to conflict with Hannes' libfc patch series]

Signed-off-by: Chris Leech <cleech@redhat.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: optimize clock gate work

In a case where gate work is called as part of cancel work from ungate
path the clk state would be marked as REQ_CLKS_ON.  There is no point
gating the clocks and then end up turning them ON immediately in ungate
work, save time by skipping the gate work and change the clk state to
CLKS_ON as they are not turned off yet.

Signed-off-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufshcd: Fix race between clk scaling and ungate work

The ungate work turns on the clock before it exits hibern8, if the link
was put in hibern8 during clock gating work.  There occurs a race
condition when clock scaling work calls ufshcd_hold() to make sure low
power states cannot be entered, but that returns by checking only
whether the clocks are on.  This causes the clock scaling work to issue
UIC commands when the link is in hibern8 causing failures. Make sure we
exit hibern8 state before returning from ufshcd_hold().

Callstacks for race condition:

 ufshcd_scale_gear
 ufshcd_devfreq_scale
 ufshcd_devfreq_target
 update_devfreq
 devfreq_monitor
 process_one_work
 worker_thread
 kthread
 ret_from_fork

 ufshcd_uic_hibern8_exit
 ufshcd_ungate_work
 process_one_work
 worker_thread
 kthread
 ret_from_fork

Signed-off-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: fail ufshcd_probe_hba() if power configuration fails

In case the power configuration fails, skip further processing of the
probing function and return immediately. This has 2 reasons:

1. Don't allow the UFS to continue running in PWM

2. Avoid multiple calls to pm_runtime_put_sync() when not in error
   handling or power management contexts

Signed-off-by: Dov Levenglick <dovl@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: suspend clock scaling for failed runtime_resume

During runtime resume operation, clock scaling may get indirectly
resumed via call to ufshcd_set_dev_pwr_mode(): Start/Stop Unit command
times out and SCSI error handling ultimately calls the host reset
handler to recover, during which clock scaling is resumed.  Error case
exit path of runtime resume will disable clocks.  As clock scaling was
already resumed, it will get scheduled later on and try to access UFS
registers while clocks are disabled, resulting in unclocked register
access.

Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: fix sense buffer size to 18 bytes

According to UFS device specification, sense data can be only 18 bytes
long, this change makes the changes accordingly.

Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: commit descriptors before setting the doorbell

Add a write memory barrier to make sure descriptors prepared are
actually written to memory before ringing the doorbell. We have also
added the write memory barrier after ringing the doorbell register so
that controller sees the new request immediately.

Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: fix bugs related to null pointer access and array size

In this change there are a few fixes of possible NULL pointer access and
possible access to index that exceeds array boundaries.

Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: cciss: replace custom function to hexdump

For small buffers we may use %*ph[N] specifier, for the bigger blocks
print_hex_dump() call.

Cc: Don Brace <don.brace@microsemi.com>
Cc: esc.storagedev@microsemi.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Don Brace <don.brace@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: replace custom approach to hexdump small buffers

In kernel we have defined specifier (%*ph[C]) to dump small buffers in a
hex format. Replace custom approach by a generic one.

Cc: Jon Mason <jonmason@broadcom.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ips: don't use custom hex_asc_upper[] table

We have table of the HEX characters in the kernel. Replace custom by a
generic one.

Cc: Adaptec OEM Raid Solutions <aacraid@adaptec.com>
Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qla4xxx: print MAC and SID via %p[mM][R]

In the kernel we have nice specifier to print MAC by given pointer to
the address in a binary form.

Signed-off-by: Oleksandr Khoshaba <Oleksandr.Khoshaba@gmail.com>
Acked-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Cc: QLogic-Storage-Upstream@qlogic.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fusion: print lan address via %pMR

LAN MAC addresses can be printed directly using %pMR specifier.

Cc: Sathya Prakash <sathya.prakash@broadcom.com>
Cc: Chaitra P B <chaitra.basappa@broadcom.com>
Cc: Suganath Prabu Subramani <suganath-prabu.subramani@broadcom.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fnic: use kernel's '%pM' format option to print MAC

Instead of supplying each byte through stack let's use %pM specifier.

Cc: Hiral Patel <hiralpat@cisco.com>
Cc: Suma Ramars <sramars@cisco.com>
Acked-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: Do not fire DCMDs during PCI shutdown/detach

This patch addresses the issue of driver firing DCMDs in PCI
shutdown/detach path irrespective of firmware state.  Driver will now
check whether firmware is in operational state or not before firing
DCMDs. If firmware is in unrecoverable state or does not become
operational within specfied time, driver will skip firing DCMDs.

[mkp: fixed typos]

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan Srikanteshwara <shivasharan.srikanteshwara@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: Send correct PhysArm to FW for R1 VD downgrade

This patch fixes the issue of wrong PhysArm was sent to firmware for R1
VD downgrade.

Signed-off-by: Kiran Kumar Kasturi <kiran-kumar.kasturi@broadcom.com>
Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid_sas: For SRIOV enabled firmware, ensure VF driver waits for 30secs before reset

For SRIOV enabled firmware, if there is a OCR(online controller reset)
possibility driver set the convert flag to 1, which is not happening if
there are outstanding commands even after 180 seconds.  As driver does
not set convert flag to 1 and still making the OCR to run, VF(Virtual
function) driver is directly writing on to the register instead of
waiting for 30 seconds. Setting convert flag to 1 will cause VF driver
will wait for 30 secs before going for reset.

CC: stable@vger.kernel.org
Signed-off-by: Kiran Kumar Kasturi <kiran-kumar.kasturi@broadcom.com>
Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

MAINTAINERS: Revise lpfc maintainers to reflect Broadcom

Avago is now known as Broadcom. Revise the emails and website for lpfc
accordingly.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_release callback with function call

The ->seq_release callback only ever had one implementation,
so call the function directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_assign callback with function call

The ->seq_assign callback only ever had one implementation,
so call the function directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_set_resp callback with direct function call

The ->seq_set_resp callback only ever had one implementation,
so call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_start_next callback with function call

The ->seq_start_next callback only ever had one implementation,
so call the function directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->exch_done callback with function call

The ->exch_done callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_exch_abort callback with function call

The ->seq_exch_abort callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_send callback with function call

The ->seq_send callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Remove fc_rport_init()

Function is empty now and can be removed.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Reviewed-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_flush_queue callback with function call

The ->rport_flush_queue callback only ever had a single
implementation, so we can as well call it directly and
drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_recv_req callback with function call

The ->rport_recv_req callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_logoff callback with function call

The ->rport_logoff callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Reviewed-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_login callback with function call

The ->rport_login callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_create callback with function call

The ->rport_create callback only ever had a single implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: scsi_transport_fc: rename 'fc_rport_create' to 'fc_remote_port_create'

Required for the next patch.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_lookup callback with function call

The ->rport_lookup callback only ever had a single implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->rport_destroy callback with function call

The ->rport_destroy callback only ever had one implementation,
so we can as well call it directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->exch_seq_send callback with function call

The ->exch_seq_send callback only ever had one implementation,
so we can call the function directly and drop the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->lport_recv with function call

The ->lport_recv callback only ever had one implementation,
so call the function directly and remove the callback.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->lport_reset callback with function call

The ->lport_reset callback only ever had one implementation,
which already is exported. So remove it and use the function
directly.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Replace ->seq_els_rsp_send callback with function call

The 'seq_els_rsp_send' callback only ever had one implementation,
so we might as well drop it and use the function directly.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Issues from Dan Carpenter/Smatch

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Return correct partition name/# to client

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Clean up properly if target_submit_cmd/tmr fails

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Synchronize cmds at remove time

This patch adds code to disconnect from the client, which will make sure
any outstanding commands have been completed, before continuing on with
the remove operation.

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Synchronize cmds at tpg_enable_store time

This patch changes the way the IBM vSCSI server driver manages its
Command/Response Queue (CRQ).  We used to register the CRQ with phyp at
probe time.  Now we wait until tpg_enable_store.  Similarly, when
tpg_enable_store is called to "disable" (i.e. the stored value is 0),
we unregister the queue with phyp.

One consquence to this is that we have no need for the PART_UP_WAIT_ENAB
state, since we can't get an Init Message from the client in our CRQ if
we're waiting to be enabled, since we haven't registered the queue yet.

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ibmvscsis: Rearrange functions for future patches

This patch reorders functions in a manner necessary for a follow-on
patch.  It also makes some minor styling changes (mostly removing extra
spaces) and fixes some typos.

There are no code changes in this patch, with one exception: due to the
reordering of the functions, I needed to explicitly declare a function
at the top of the file.  However, this will be removed in the next patch,
since the code requiring the predeclaration will be removed.

Signed-off-by: Michael Cyr <mikecyr@us.ibm.com>
Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Tested-by: Steven Royer <seroyer@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: g_NCR5380: add HAS_IOPORT_MAP dependency

The driver was changed to call ioport_map, which breaks platforms that
cannot provide this function:

drivers/scsi/g_NCR5380.o: In function `generic_NCR5380_init_one.constprop.0':
g_NCR5380.c:(.text.generic_NCR5380_init_one.constprop.0+0x388): undefined reference to `ioport_map'

This adds a Kconfig dependency.

Fixes: 04c40f82ccc5 ("scsi: g_NCR5380: Merge g_NCR5380 and g_NCR5380_mmio drivers")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: Fixup missing initialisation in fcoe_dcb_create()

Found by 0-day robot.

Fixes: a99ac6e715bc ("scsi: fcoe: set default TC priority")
Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Use %zd format string for size_t

A recent bugfix introduced a harmless warning in the lpfc driver:

drivers/scsi/lpfc/lpfc_init.c: In function 'lpfc_write_firmware':
drivers/scsi/lpfc/lpfc_logmsg.h:56:45: error: format '%ld' expects argument of type 'long int', but argument 9 has type 'size_t {aka const unsigned int}' [-Werror=format=]

'size_t' is always the same width as 'long' in the kernel, but the
compiler doesn't know that. The %z modifier is what the standard expects
to be used here, and this shuts up the warning.

Fixes: 679053c651fb ("scsi: lpfc: Fix fw download on SLI-4 FC adapters")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: filter out frames from invalid vlans

Any multicase address is set on all interfaces, the base interface
and any VLAN interfaces on top of this. So we might receive frames
which are not destined for us.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: correct sending FIP VLAN packets on VLAN 0

The FIP VLAN frame consists of an ethernet header followed
by the FIP VLAN frame, so we need to skip the ethernet header
if we want to check the FIP opcode.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: FIP debugging

Add additional statements for debugging FIP frames.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: catch invalid values for the 'enabled' attribute

The 'enabled' sysfs attribute only accepts the values '0' and '1',
so we should error out any other values.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fcoe: set default TC priority

If DCB is not enabled or compiled in we still should be setting
a sane default priority. So put FCoE frames in priority class
'interactive' and FIP frames in priority class 'besteffort'.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Check xid when looking up REC exchanges

We currently can only lookup the local xid, so we need
to reject REC with empty rxid.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: wait for E_D_TOV when out-of-order sequence is received

When detecting an out-of-order sequence we should be waiting for
E_D_TOV before trying to abort the sequence.
The response might still be stuck in the queue somewhere.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: reset timeout on queue full

When we're receiving a timeout we should be checking for queue
full status; if there are still some packets pending we should
be resetting the counter to ensure we're not missing out any
packets which are still queued.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Do not drop out-of-order frames

When receiving packets from the network we cannot guarantee any
frame ordering, so we should be receiving all valid frames and
let the upper layers deal with it.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: don't fail sequence abort for completed exchanges

If a sequence should be aborted the exchange might already
be completed (eg if the response is still queued in the rx
queue), so this shouldn't considered as an error.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: quarantine timed out xids

When a sequence times out we have no idea what happened to the
frame. And we do not know if we will ever receive the frame.
Hence we cannot re-use the xid as we would risk data corruption
if the xid had been re-used and the timed out frame would be
received after that.
So we need to quarantine the xid until the lport is reset.
Yes, I know this will (eventually) deplete the xid pool.
But for now it's the safest method.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: safeguard against invalid exchange index

The cached exchange index might be invalid, in which case
we should drop down to allocate a new one.
And we should not try to access an invalid exchange when
responding to a BA_ABTS.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Clarify ramp-down messages

When the queue depth is reduced we should print out the reason
for this; it might be due to a queue full condition.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Return LS_RJT_BUSY for PRLI in status PLOGI

Occasionally it might happen that we receive a PRLI while we're still
waiting for our PLOGI response. In that case we should return
'busy' LS status instead of 'plogi required' LS status.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: libfc: Rework PRLI handling

PRLI is only required if the port is acting as an initiator; ports
which support target functionality only do not need to send PRLI.
At the same time the PRLI state is only used if the port initiated
a PRLI transfer; if we received a PRLI request we should _not_
change the state as this would cause our PRLI response to be dropped.
And when we receive a PRLI response we need to check if an image
pair has been established; if not the remote port cannot act as a
target for us and we need to disable target functionality.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Acked-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>