scsi: ufs: fix error recovery after the hibern8 exit failure

Hibern8 exit can be called from 3 different contexts:
    - ufshcd_hibern8_exit_work
    - ufshcd_ungate_work
    - runtime/system resume

If hibern8 exit fails for some reason then we try to bring the link to
active state by link startup but this recovery mechanism results into
deadlock or errors from first 2 context listed above. This change fixes
the recovery by adding proper error handling mechanism.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: add retries for hibern8 enter

If hibern8 enter command fails then UFS link state may be unknown which
may result into timeout of all the commands issued after failure.

This change does 2 things (for pre-defined number of retry counts) after
hibern8 enter failure:

1. Recovers the UFS link to active state

2. If link is recovered to active state, tries to put the UFS link in
   hibern8 enter again until retry count expires.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: add retries to dme_peer get and set attribute

The dme_peer get/set attribute commands are prone to errors, therefore
we add three retries for the UIC command sending.  Error code returned
from ufshcd_send_uic_cmd() is checked, and unless it was successful or
the retries have finished, another command will be sent.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Lee Susman <lsusman@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: avoid exception event handler racing with PM callbacks

If device raises the exception event in the response to the commands
sent during the runtime/system PM callbacks, exception event handler
might run in parallel with PM callbacks and may see unclocked register
accesses. This change fixes this issue by not scheduling the exception
event handler while PM callbacks are running.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: increase fDeviceInit query response timeout

fDeviceInit query response time for some devices is too long that
default query request timeout of 100ms may not be enough. Experiments
show that fDeviceInit response sometimes takes 500ms so to be on safer
side this change sets the timeout to 600ms. Without this change, we
might unnecessarily have to retry fDeviceInit query requests multiple
times and each query request timeout prints one error message.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: clear outstanding_request bit in case query timeout

When sending a query to the device returns with a timeout error, we
clear the corresponding bit in the DOORBELL register but we don't clear
the outstanding_request field as we should.  This patch fixes this bug.

Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: verify command tag validity

A race condition appear to exist between request completion when
scsi_done() is called to end the request and set the tag back to -1 (at
blk_queue_end_tag() scsi_end_request), and scsi layer error handling
which aborts the command and reuses it to request sense data. Sending
the request sense is done with tag which was set to -1 and so it is
invalid.  Assert command tag passed from scsi layer is valid.

Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: clear fields UTRD, UPIU req and rsp before new transfers

Some of the data structures (like response UPIU) and/or its elements
(unused fields) should be cleared before sending out the respective
command to UFS device.

This change clears the UPIU response data structure for query commands
and NOP command before sending out the command. We also initialize the
PRDT table length to zero which should take care of commands which
doesn't have any data associated with it. We are also clearing the
unused fields in request UPIU for NOP command.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Tested-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: clear UTRD, UPIU req and rsp before new transfers

Clear the UFS data structures before sending new request.

The SCSI command is sent to the device within the UFS UPIU request.
As part of the transfer UPIU preparation, the SCSI command is copied
to the UPIU structure according to the SCSI command size.
As different SCSI commands differ in size from each other, we need
to clear the whole SCSI command field to prevent sending uninitialized
data to the device.

The UPIU response doesn't always include the sense data and can differ
in size.
Hence, the UPIU response should also be cleared before the transfer.

Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi_dh: force modular build if SCSI is a module

When the scsi_dh core was moved into the scsi core module,
CONFIG_SCSI_DH became a 'bool' option, and now anything depending on it
can be built-in even when CONFIG_SCSI=m. This of course cannot link
successfully:

drivers/scsi/built-in.o: In function `rdac_init':
scsi_dh_alua.c:(.init.text+0x14): undefined reference to `scsi_register_device_handler'
scsi_dh_alua.c:(.init.text+0x64): undefined reference to `scsi_unregister_device_handler'
drivers/scsi/built-in.o: In function `alua_init':
scsi_dh_alua.c:(.init.text+0xb0): undefined reference to `scsi_register_device_handler'

As a workaround, this adds an extra dependency on CONFIG_SCSI, so
Kconfig can figure out whether built-in is allowed or not.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 086b91d052eb ("scsi_dh: integrate into the core SCSI code")
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: aha1542: avoid uninitialized variable warnings

Gcc incorrectly detects that two variables in aha1542_queuecommand might
be used without an initialization:

scsi/aha1542.c: In function 'aha1542_queuecommand':
scsi/aha1542.c:382:16: error: 'cptr' may be used uninitialized in this function [-Werror=maybe-uninitialized]
scsi/aha1542.c:379:11: error: 'sg_count' may be used uninitialized in this function [-Werror=maybe-uninitialized]

The only user of these is doing the same check that the assigment has,
so it is actually guaranteed to work. Adding an "else" clause with a
fake initialization shuts up the warning.

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

scsi: qla4xxx: shut up warning for rd_reg_indirect

The qla4_83xx_rd_reg_indirect() function can fail when it is unable to
read a register, but not all callers check its return value before using
the register data, and gcc correctly warns about this:

qla4xxx/ql4_83xx.c: In function 'qla4_83xx_process_reset_template':
qla4xxx/ql4_83xx.c:1073:36: warning: 'value' may be used uninitialized in this function
     ha->reset_tmplt.array[index++] = value;
                                    ^
qla4xxx/ql4_83xx.c:1050:11: note: 'value' was declared here
  uint32_t value;
           ^
qla4xxx/ql4_83xx.c:902:8: warning: 'value' may be used uninitialized in this function
  value &= p_rmw_hdr->test_mask;
        ^
qla4xxx/ql4_83xx.c:895:11: note: 'value' was declared here
  uint32_t value;
           ^
In file included from ../include/linux/io.h:25:0,
                 from ../include/linux/pci.h:31,
                 from ../drivers/scsi/qla4xxx/ql4_def.h:16,
                 from ../drivers/scsi/qla4xxx/ql4_83xx.c:10:
asm/io.h:101:2: warning: 'value' may be used uninitialized in this function
  asm volatile("str %1, %0"
  ^
qla4xxx/ql4_83xx.c:874:11: note: 'value' was declared here
  uint32_t value;
           ^

Unfortunately, I don't see any helpful way to add proper error handling
for this case, and the failure scenario for rd_reg seems rather obscure,
so this bails out and makes the rd_reg accessor set the result to
0xffffffff so we at least get a predictable value.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Nilesh Javali <nilesh.javali@qlogic.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fas216: avoid fas216_log_setup for loadable module

We get a warning for the fas216 driver when it is compiled as a loadable
module, as the __setup() functions are never called then:

scsi/arm/fas216.c:101:19: warning: 'fas216_log_setup' defined but not used [-Wunused-function]
 static int __init fas216_log_setup(char *str)

This adds an #ifndef MODULE around the definition to shut up the warning
and clarify for the reader when it is used or not.

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

scsi: acornscsi: mark calc_sync_xfer as __maybe_unused

The calc_sync_xfer function is only used if CONFIG_SCSI_ACORNSCSI_SYNC
is set, otherwise we get a compiler warning:

scsi/arm/acornscsi.c:680:15: warning: 'calc_sync_xfer' defined but not used [-Wunused-function]
 unsigned char calc_sync_xfer(unsigned int period, unsigned int offset)

This marks the function as __maybe_unused to shut up the warning and
silently drop the function in the object code when there is no caller.

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

mptfusion: hide unused seq_mpt_print_ioc_summary function

The seq_mpt_print_ioc_summary function is used for the
/proc/mpt/iocN/summary implementation and never gets called when
CONFIG_PROC_FS is disabled:

drivers/message/fusion/mptbase.c:6851:13: warning: 'seq_mpt_print_ioc_summary' defined but not used [-Wunused-function]
 static void seq_mpt_print_ioc_summary(MPT_ADAPTER *ioc, struct seq_file *m, int showlan)

This adds an #ifdef to hide the function definition in that case and
avoid the warning.

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

scsi: fdomain: drop fdomain_pci_tbl when built-in

The fdomain SCSI host driver is one of the last remaining drivers that
manually search the PCI bus using pci_get_device rather than registering
a pci_driver instance.

This means the module device table is unused when the driver is
built-in, and we get a warning about it:

drivers/scsi/fdomain.c:1773:29: warning: 'fdomain_pci_tbl' defined but not used [-Wunused-variable]

To avoid the warning, this adds another #ifdef around the table
definition.

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

megaraid_sas: Add an i/o barrier

A barrier should be added to ensure proper ordering of memory mapped
writes.

Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Reviewed-by: Kashyap Desai <kashyap.desai@broadcom.com>
Acked-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

lpfc: Remove redundant code block in lpfc_scsi_cmd_iocb_cmpl

This removes a redundant code block that will either be executed if the
ENABLE_FCP_RING_POLLING flag is set in phba->cfg_poll or not. The code
is just duplicated in both cases, hence we unify it again.

This probably is a left over from some sort of refactoring.

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Reviewed-by: Sebastian Herbszt <herbszt@gmx.de>
Acked-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Updating mpt3sas driver version to 12.100.00.00

Bump mpt3sas driver version from 09.102.00.00 to 12.100.00.00

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Fix for Asynchronous completion of timedout IO and task abort of timedout IO.

Track msix of each IO and use the same msix for issuing abort to timed
out IO. With this driver will process IO's reply first followed by TM.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Updated MPI Header to 2.00.42

Updated MPI version and MPI header files.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Add support for configurable Chain Frame Size

Added support for configurable Chain Frame Size. Calculate the
Chain Message Frame size from the IOCMaxChainSegementSize (iocfacts).
Applicable only for mpt3sas/SAS3.0 HBA's.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Added smp_affinity_enable module parameter.

Module parameter to enable/disable configuring affinity hint for msix
vector.  SMP affinity feature can be enabled/disabled by setting module
parameter "smp_affinity_enable" to 1/0.  By default this feature is
enabled. (smp_affinity_enable = 1 enabled).

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Make use of additional HighPriority credit message frames for sending SCSI IO's

Driver assumes HighPriority credit as part of Global credit. But,
Firmware treats HighPriority credit value and global cedits as two
different values. Changed host queue algorithm to treat global credits
and highPriority credits as two different values.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Never block the Enclosure device

Never block the SEP device (i.e. Never invoke the
scsi_internal_device_block() API for SEP device) even for the delay not
responding events. Blocking the SEP device will create a deadlock while
adding any device to the OS.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Fix static analyzer(coverity) tool identified defects

1.Wrong size of argument is being passed
 The size of struct being passed as an argument to memset func and area of
 memory being pointed by an instance of struct in memset func should be of
 same structure type.
2.Dereference null return value
3.Array compared against '0'
 Check whether value pointed by particular index of an array is null or not
 in "if" statement.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Used IEEE SGL instead of MPI SGL while framing a SMP Passthrough request message.

As driver was using MPI SGL while framing the SMP Passthrough request
message due to which firmware unable to post the Reply Data in the host
memory and timeout is observed for this SMP Passthrough request message
and so unable to perform phy disable operation.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

mpt3sas: Added support for high port count HBA variants.

Updated hardware description headers with MPI v2.6 and
mpt3sas_pci_table[] with vendor_ids, device_ids of Cutlass and Intruder
HBA which have support for 4 ports.

Signed-off-by: Suganath prabu Subramani <suganath-prabu.subramani@avagotech.com>
Signed-off-by: Chaitra P B <chaitra.basappa@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

st: Fix MTMKPART to work with newer drives

Change the MTMKPART operation of the MTIOCTOP ioctl so that it works
also with current drives (LTO-5/6, etc.). Send a separate FORMAT MEDIUM
command if the partition mode page indicates that this is required. Use
LOAD to position the tape at the beginning of tape.

The operation is extended so that if the argument is negative, its
absolute value specifies the size of partition 0, which is the
physically first partition of the tape.

Signed-off-by: Kai Mäkisara <kai.makisara@kolumbus.fi>
Tested-by: Shane M Seymour <shane.seymour@hpe.com>
Tested-by: Laurence Oberman <loberman@redhat.com>
Tested-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Fix SMAP issue

Inside compat IOCTL hook of driver, driver was using wrong address of
ioc->frame.raw which leads sense_ioc_ptr to be calculated wrongly and
failing IOCTL.

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

megaraid_sas: Fix for IO failing post OCR in SRIOV environment

Driver assumes that VFs always have peers present whenever they have
same LD IDs. But this is not the case.  This patch handles the above
mentioned by explicitly checking for a peer before making HA/non-HA path
decision.

Signed-off-by: Uday Lingala <uday.lingala@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: update driver version to 1.2

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v1 hw ACPI support

Add support in v1 hw driver for ACPI.

A check on whether an ACPI handle is available for the device is used to
decide on whether to use ACPI reset handler or syscon for hw reset.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: use Unified Device Properties API

The hisi_sas driver is required to support both device tree and
ACPI. The scanning of the device properties now uses the Unified Device
Properties API, which serves both OF and ACPI.

Since syscon is not supported by ACPI, syscon is only used in the driver
when device tree is used.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Add lock to protect WRB alloc and free

FW got into UE after running IO stress test

With kernel change to split session lock in frwd_lock and back_lock for
tx and rx path correspondingly, in the IO path, common resource used in
driver such as WRB was left unprotected.

Add wrb_lock spinlock to protect allocation and freeing of WRB.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: _bh for io_sgl_lock and mgmt_sgl_lock

Processing of mgmt and IO tasks are done in process context and
softirqs.

Allocation and freeing of sgl_handles needs to be done under
spin_lock_bh/spin_unlock_bh and move the locks to the routines.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Fix ExpStatSn in management tasks

Connection resets observed from some targets when NOP-Out with wrong
ExpStatSn is sent.

FW keeps track of StatSn and fills up ExpStatSn accordingly.  The header
filled up by the stack needs to be modified by driver to clear
ExpStatSn. If the field is not cleared, FW recalculates ExpStatSn and
wrong offset'ed ExpStatSn is seen in the wire trace.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Couple MCC tag and WRB alloc and free

WARN_ON(atomic_read(&mccq->used) >= mccq->len) seen when FW gets into
UE.

MCCQ overflow is happening because driver discards any new request and
frees up the tag. The tag allocation controls the number of MCC WRB
posted.  It is being replenished but WRBs are not hence the WARN_ON.

Allocation and freeing of WRB and tags for MCC is now done in one place.
This helps to achieve proper accounting of WRB indices and MCC tags.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Fix MCC WRB leak in open_connection

In open with IP of unknown address family, only tag is freed and error
returned. MCC WRB allocated for the operation is not freed.

Added check for supported family of IP in the beginning before
allocating the tag and WRB.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Cleanup processing of BMBX completion

Remove confusingly named be_mcc_compl_is_new and be_mcc_compl_use
functions in processing of BMBX. Rearrange beiscsi_process_mbox_compl
function.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Fix be_mcc_compl_poll to use tag_state

be_mcc_compl_poll waits till 'used' count of MCC WRBQ is zero. This is
to determine the completion of an MCC sent.

Change function to poll for the tag of MCC sent, instead, and wait till
its tag_state is cleared.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Remove be_mbox_notify_wait function

be_mbox_notify_wait does exactly same thing as be_mbox_notify.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Rename MCC and BMBX processing functions

beiscsi_mccq_compl -> beiscsi_mccq_compl_wait - indicate blocking call.
be_mcc_wait_compl -> be_mcc_compl_poll - indicate polling for completion.
be_mbox_db_ready_wait -> be_mbox_db_ready_poll - indicate polling for RDY.
be_mcc_compl_process -> beiscsi_process_mbox_compl - indicate BMBX compl.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Remove redundant MCC processing code

be_mcc_compl_process_isr is removed. MCC CQ processing is done only in
beiscsi_process_mcc_cq and MCC CQE processing is done only in
beiscsi_process_mcc_compl.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Use macros for MCC WRB and CQE fields

Rename mcc_numtag to mcc_tag_status.  MCC CQE status is processed using
macros already defined in be_cmds.h.

Add MCC_Q_WRB_ and MCC_Q_CMD_TAG_MASK macros to map to already defined
CQE_STATUS_ macros to be consistent when posting MCC.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

be2iscsi: Remove unused mcc_cq_lock

mcc_cq_lock spin_lock is used only in beiscsi_process_mcc which is
called only when all interrupts are disabled from mgmt_epfw_cleanup
during unloading of driver. There is no other context where there can be
contention for the processing of CQ.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

bfa: use strncpy() instead of memcpy()

BFA_MFG_NAME is "QLogic" which is only 7 bytes, but we are copying 8
bytes.  It's harmless because the badding byte is likely zero but it
makes static checkers complain.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Anil Gurumurthy <anil.gurumurthy@qlogic.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid: fix null pointer check in megasas_detach_one().

The pd_seq_sync pointer can't be NULL, we have to check its entries
instead.

Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Acked-by: Sumit Saxena <sumit.saxena@broadcom.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Update driver version

Updated diver version to 41052

Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Fix character device re-initialization

During EEH PCI hotplug activity kernel unloads and loads the driver,
causing character device to be unregistered(aac_remove_one).When the
driver is loaded back using aac_probe_one the character device needs
to be registered again for the AIF management tools to work.

Fixed by adding code to register character device in aac_probe_one if
it is unregistered in aac_remove_one.

Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Shane Seymour <shane.seymour@hpe.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Fix AIF triggered IOP_RESET

while driver removal is in progress or PCI shutdown is invoked, driver
kills AIF aacraid thread, but IOCTL requests from the management tools
re-start AIF thread leading to IOP_RESET.

Fixed by setting adapter_shutdown flag when PCI shutdown is invoked.

Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Shane Seymour <shane.seymour@hpe.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Created new mutex for ioctl path

aac_mutex was used to create protect the ioctl path for only the compat
path, it would be make more sense to place mutex in aac_do_ioctl, which
is the main ioctl function call that handles all ioctl commands.

Created new mutex ioctl_mutex in struct aac_dev to protect switch case
in aac_do_ioctl and removed aac_mutex from aac_cfg_ioctl and
aac_compat_do_ioctl

Signed-off-by: Raghava Aditya Renukunta <RaghavaAditya.Renukunta@pmcs.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Fundamental reset support for Series 7

Series 7 does not support PCI hot reset used by EEH.

Enabled fundamental reset only for Series 7

Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Set correct msix count for EEH recovery

During EEH recovery number of online CPU's might change thereby changing
the number of MSIx vectors. Since each fib is allocated to a vector,
changes in the number of vectors causes fib to be sent thru invalid
vectors.In addition the correct number of MSIx vectors is not updated in
the INIT struct sent to the controller, when it is reinitialized.

Fixed by reassigning vectors to fibs based on the updated number of MSIx
vectors and updating the INIT structure before sending to controller.

Fixes: MSI-X vector calculation for suspend/resume
Cc: stable@vger.kernel.org
Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Shane Seymour <shane.seymour@hpe.com>
Reviewed-by: Johannes Thumshirn <jthushirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Fix memory leak in aac_fib_map_free

aac_fib_map_free() calls pci_free_consistent() without checking that
dev->hw_fib_va is not NULL and dev->max_fib_size is not zero.If they are
indeed NULL/0, this will result in a hang as pci_free_consistent() will
attempt to invalidate cache for the entire 64-bit address space
(which would take a very long time).

Fixed by adding a check to make sure that dev->hw_fib_va and
dev->max_fib_size are not NULL and 0 respectively.

Fixes: 9ad5204d6 - "[SCSI]aacraid: incorrect dma mapping mask during blinked recover or user initiated reset"
Cc: stable@vger.kernel.org
Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Added EEH support

Added support for PCI EEH (extended error handling).

Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: Fix RRQ overload

The driver utilizes an array of atomic variables to keep track of IO
submissions to each vector. To submit an IO multiple threads iterate
through the array to find a vector which has empty slots to send an
IO. The reading and updating of the variable is not atomic, causing race
conditions when a thread uses a full vector to submit an IO.

Fixed by mapping each FIB to a vector, the submission path then uses
said vector to submit IO thereby removing the possibly of a race
condition.The vector assignment is started from 1 since vector 0 is
reserved for the use of AIF management FIBS.If the number of MSIx
vectors is 1 (MSI or INTx mode) then all the fibs are allocated to
vector 0.

Fixes: 495c0217 "aacraid: MSI-x support"
Cc: stable@vger.kernel.org # v4.1
Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

aacraid: SCSI blk tag support

The method to allocate and free FIB's in the present code utilizes
spinlocks. Multiple IO's have to wait on the spinlock to acquire or free
fibs creating a performance bottleneck.

An alternative solution would be to use block layer tags to keep track
of the fibs allocated and freed. To this end aac_fib_alloc_tag was
created to utilize the blk layer tags to plug into the Fib pool.These
functions are used exclusively in the IO path. 8 fibs are reserved for
the use of AIF management software and utilize the previous spinlock
based implementations.

Signed-off-by: Raghava Aditya Renukunta <RaghavaAditya.Renukunta@pmcs.com>
Reviewed-by: Shane Seymour <shane.seymour@hpe.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: update driver version to 1.1

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 tmf functions

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 slot error handler

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 path to send ATA command

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 code for itct setup and free

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 code to send smp command

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 path to send ssp frame

Include code to prep ssp frame and deliver to hardware.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 cq interrupt handler

Also include slot_complete_v2_hw handler

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 SATA interrupt handler

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 channel interrupt handler

This also includes broadcast handler. Unlike v1 hw, broadcast does not
have its own dedicated interrupt.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 phy down handler

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 int init and phy up handler

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 phy init code

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add init_id_frame_v2_hw()

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 hw init

Add code to initialise the hardware.

Support is also added to deal with the "am-max-transmissions" (amt)
limitation in hip06 controller #1. This is how many connection requests
we can send on the system bus before waiting for a response.  Due to
chip bus design, controller #1 is limited to 32 amt, while, by design, a
controller supports 64.  The default value for the nibbles in the
relevant registers is 0x40; these need to be programmed with 0x20.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add v2 register definitions

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add bare v2 hw driver

Just add enough to build and init the module.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: rename some fields in hisi_sas_itct

Since hisi_sas_itct format is different between v1 and v2 hw, give more
general names for some fields.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: add hisi_sas_err_record_v1

Since the error record structure is different for v2 hw, make
hisi_sas_err_record opaque and add hisi_sas_err_record_v1.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: reduce max itct entries

Since v2 hw only supports 2048 itct entries, as opposed to 4096 for v1
hw, set the max itct entries to the lower of the two.

It is not anticipated that any device with v1 will ever require to
connect > 2048 devices.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: set max commands as configurable

Since v2 hardware permits different numbers of commands to v1, set this
as configurable in hisi_sas_hw.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

hisi_sas: relocate DEV_IS_EXPANDER

Relocate DEV_IS_EXPANDER to hisi_sas.h as it will be required for v2 hw
support.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

devicetree: bindings: hisi_sas: add v2 HW bindings

Add the dt bindings for HiSi SAS controller v2 HW.

The main difference in the controller from dt perspective is
interrupts. The v2 controller does not have dedicated fatal and
broadcast interrupts - they are multiplexed on the channel interrupt.

Each SAS v2 controller can issue upto 64 commands (or connection
requests) on the system bus without waiting for a response - this is
know as am-max-transmissions.  In hip06, sas controller #1 has a
limitation that it has to limit am-max-transmissions to 32 - this
limitation is due to chip system bus design. It is not anticipated that
any future chip incorporating v2 controller will have such a limitation.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

bnx2fc: bnx2fc_eh_abort(): fix wrong return code.

If the link is not ready, the bnx2fc_eh_abort() function should return
FAILED.

Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Ewan Milne <emilne@redhat.com>
Acked-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

bnx2fc: Show information about log levels in 'modinfo'

This patch adds the information of the different values that can be used
in the module parameter 'debug_logging', as it is shown below:

$ modinfo bnx2fc
[...]
parm:           debug_logging:Option to enable extended logging,
Default is 0 - no logging.
0x01 - SCSI cmd error, cleanup.
0x02 - Session setup, cleanup, etc.
0x04 - lport events, link, mtu, etc.
0x08 - ELS logs.
0x10 - fcoe L2 fame related logs.
0xff - LOG all messages. (int)

Signed-off-by: Jose Castillo <jcastillo@redhat.com>
Acked-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Use ida for host number management

Update the SCSI hosts module to use ida to manage its host_no index
instead of an atomic integer. This means that the SCSI host number will
now be reclaimable.

Use the ida "simple" mechanism, since there should be no need for a
separate spin lock current usage. Ida was chosen over idr because the
hosts module already has its own instance and locking mechanisms that
aren't easily changed.

Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: driver version upgrade

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: SPERC OCR changes

This patch fixes online controller resets on SRIOV-enabled series of
Avago controllers.

1) Remove late detection heartbeat.

2) Change in the behavior if the FW found in READY/OPERATIONAL state.

Signed-off-by: Uday Lingala <uday.lingala@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Introduce module parameter for SCSI command timeout

This patch will introduce module-parameter for SCSI command timeout
value and fix setting of resetwaittime beyond a value.

Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: MFI adapter OCR changes

Optimized MFI adapters' OCR path, particularly
megasas_wait_for_outstanding() function.

Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Make adprecovery variable atomic

Make instance->adprecovery variable atomic and removes hba_lock spinlock
while accessing instance->adprecovery.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: IO throttling support

This patch will add capability in driver to tell firmware that it can
throttle IOs in case controller's queue depth is downgraded post OFU
(online firmware upgrade). This feature will ensure firmware can be
downgraded from higher queue depth to lower queue depth without needing
system reboot. Added throttling code in IO path of driver, in case OS
tries to send more IOs than post OFU firmware's queue depth.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Dual queue depth support

1. For iMR controllers, firmware will report two queue depths:

   - Controller-wide queue depth
   - LDIO queue depth (240)

Controller-wide queue depth will be greater among the two. Using this
new feature, iMR can provide larger Queue depth(QD) for JBOD and limited
QD for Virtual Disk(VD).

2. megaraid_sas driver will throttle read/write LDIOs based on "LDIO
Queue Depth".

3. Dual queue depth can be enabled/disabled via module parameter. It is
enabled by default if the firmware supports it. Only specific firmware
builds will enable the feature.

4. Added sysfs parameter "ldio_outstanding" which permits querying the
number of outstanding LDIO requests at runtime.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Code optimization build_and_issue_cmd return-type

build_and_issue_cmd should return SCSI_MLQUEUE_HOST_BUSY for a few error
cases instead of returning 1.

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

megaraid_sas: Reply Descriptor Post Queue (RDPQ) support

This patch will create a reply queue pool for each MSI-X index and will
provide an array of base addresses instead of the single address of
legacy mode. Using this new interface the driver can support higher
queue depths through scattered DMA pools.

If array mode is not supported driver will fall back to the legacy
method of reply pool allocation. This limits controller queue depth to
1K max. To enable a queue depth of more than 1K driver requires firmware
to support array mode and scratch_pad3 will provide the new queue depth
value.

When RDPQ is used, downgrading to an older firmware release should not
be permitted. This may cause firmware fault and is not supported.

Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Fastpath region lock bypass

Firmware will fill out per-LD data to tell driver whether a particular
LD supports region lock bypass. If yes, then driver will send non-FP
LDIO to region lock bypass FIFO. With this change in driver, firmware
will optimize certain code to improve performance.

Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Update device queue depth based on interface type

This patch will update device Queue depth based on interface type(SAS,
SATA..) for sysPDs.  For Virtual disks(VDs), there will be no change in
queue depth (will remain 256).  To fetch interface type (SAS or SATA or
FC..) of syspD, driver will send DCMD MR_DCMD_PD_GET_INFO.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Task management support

This patch adds task management for SCSI commands. Added functions are
task abort and target reset.

1. Currently, megaraid_sas driver performs controller reset when any IO
times out.  With task management support added, task abort and target
reset will be tried to recover timed out IO. If task management fails,
then controller reset will be performaned. If the task management
request times out, fail the request and escalate to the next
level (controller reset).

2. mr_device_priv_data will be allocated for all generations of
controller, but is_tm_capable flag will never be set for
controllers (prior to Invader series) as firmware support is not
available for task management.

3. Task management capable firmware will set is_tm_capable flag in
firmware API.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Syncing request flags macro names with firmware

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

megaraid_sas: MFI IO timeout handling

This patch will do proper error handling for DCMD timeout failure cases
for Fusion adapters:

1. For MFI adapters, in case of DCMD timeout (DCMD which must return
SUCCESS) driver will call kill adapter.

2. What action needs to be taken in case of DCMD timeout is decided by
function dcmd_timeout_ocr_possible().  DCMD timeout causing OCR is
applicable to the following commands:

MR_DCMD_PD_LIST_QUERY
MR_DCMD_LD_GET_LIST
MR_DCMD_LD_LIST_QUERY
MR_DCMD_CTRL_SET_CRASH_DUMP_PARAMS
MR_DCMD_SYSTEM_PD_MAP_GET_INFO
MR_DCMD_LD_MAP_GET_INFO

3. If DCMD fails from driver init path there are certain DCMDs which
must return SUCCESS. If those DCMDs fail, driver bails out. For optional
DCMDs like pd_info etc., driver continues without executing certain
functionality.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

megaraid_sas: Do not allow PCI access during OCR

This patch will do synhronization between OCR function and AEN function
using "reset_mutex" lock.  reset_mutex will be acquired only in the
first half of the AEN function which issues a DCMD. Second half of the
function which calls SCSI API (scsi_add_device/scsi_remove_device)
should be out of reset_mutex to avoid deadlock between scsi_eh thread
and driver.

During chip reset (inside OCR function), there should not be any PCI
access and AEN function (which is called in delayed context) may be
firing DCMDs (doing PCI writes) when chip reset is happening in parallel
which will cause FW fault. This patch will solve the problem by making
AEN thread and OCR thread mutually exclusive.

Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Reviewed-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Export function scsi_scan.c:sanitize_inquiry_string

The hpsa driver uses this function to cleanup inquiry data. Our new pqi
driver will also use this function. This function was copied into both
drivers.

This patch exports sanitize_inquiry_string so the hpsa and the pqi
drivers can use this function directly.

Suggested-by: Hannes Reinecke <hare@suse.de>
Suggested-by: Matthew R. Ochs mrochs@linux.vnet.ibm.com
Reviewed-by: Kevin Barnett <kevin.barnett@pmcs.com>
Reviewed-by: Justin Lindley <justin.lindley@pmcs.com>
Reviewed-by: Scott Teel <scott.teel@pmcs.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Don Brace <don.brace@pmcs.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>