Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.

The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.

The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.

Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.

Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.

For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.

libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.

The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.

That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.

Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.

In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Restructure the iscsi_tcp_r2t_rsp routine in order to avoid allocating
r2t from r2tpool.queue and returning it back in case the parameters
rhdr->data_length and or rhdr->data_offset prohibit the requing.

Since the values of these parameters are known prior to the allocation,
we can pre-check and thus avoid futile allocations.

[jejb: checkpatch fixes]
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Fixed calculates wrong tuple values on T5 adapter: switch to use the exported
API cxgb4_select_ntuple() from cxgb4 base driver.
Signed-off-by: Karen Xie <kxie@chelsio.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

When megasas_get_pd_list() fails we cannot detect any drives,
so we should be checking the return value accordingly.

[jejb: checkpatch fix]
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

When no HBA is found we should be returning '-ENXIO' to be consistent
with the other return values.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The pointer to the queue info structure is potentially
a 64-bit value, so we should be using the correct macros
to set the values in the init frame.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The version of isci driver has not been updated for 2 years.
It was 83 isci commits ago. Suspend/resume support has been implemented
and many bugs have been fixed since 1.1. Now update the version to 1.2.
Signed-off-by: Lukasz Dorau <lukasz.dorau@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Maciej Patelczyk <maciej.patelczyk@intel.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

For various error conditions the bfa driver just returns
'DID_ERROR', which carries no information at all about the
actual source of error.
This patch updates the error handling to return a correct
error code, depending on the type of error occurred.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Acked-by: Vijaya Mohan Guvva <vmohan@brocade.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Fix following sparse warnings:-
 drivers/scsi/qla4xxx/ql4_os.c:2109:33: warning: cast truncates bits from constant value (ffff7fff becomes 7fff)
 drivers/scsi/qla4xxx/ql4_os.c:2306:33: warning: cast truncates bits from constant value (ffff7fff becomes 7fff)
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Nilesh Javali <nilesh.javali@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Removing unused code as FW does not need any value in mbox-5.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
While unloading driver MBOX 0x31 fail as DDB logout (MBOX 0x56)
operation is not completed.

Fix:
Wait for DDB Logout completion before MBOX 0x31
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
Driver holds rom-lock for too long during reset recovery.

During adapter reset testing, it was found that the driver
holds the rom-lock for too long, because of which other
drivers fail to acquire the rom-lock, leading to reset
failures.
The primary cause is, in the bootstrap code, while
holding the rom-lock, the driver checks if the peg is
halted, causing a 2 second contention.

Fix:
When a reset recovery starts, the driver deduces the cause, and
sets appropriate flags in watchdog & recover_adapter routines.
This flag should be used to determine if bootstrap is invoked
from probe or reset context, reducing the rom-lock footprint of
the drivers.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
minidump data collection fails as driver reports data mismatch

Fix:
When the driver encounters a new entry type that it cannot process,
it should just skip the entry and adjust the total buffer size by
subtracting the skipped bytes from it. This is to ensure that there
is no data mismatch because of the new entries.
Signed-off-by: Tej Parkash <tej.parkash@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
Pending IO wait does not complete after triggering Graceful reset,
causing ack timeout and call traces.

Fix:
1. Reducing the IO command wait timeout before triggering reset,
   as logically also timeout should be less than reset timeout (10sec).
2. Moving the abort IO after chip reset, because only after
   chip reset, driver owns the IO otherwise it is with firmware and can
   still revert back with response.
Signed-off-by: Tej Parkash <tej.parkash@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
While booting with kdump kernel, driver receive IOCB interrupts
for which it is not ready which results in processing them
before init_firmware during driver probe

Fix:
Two steps solution
1. Make driver ready to process the interrupt before interupts
   handlers is registered.
2. Stop driver processing iocb interrupts if not generated as per
   firmware protocol i.e R2H bit set
Signed-off-by: Tej Parkash <tej.parkash@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
Mailbox command 0x102 (IDC Time Extend) failure seen while applying
changes to iface using iscsiadm

Fix:
Added fix to extend IDC timeout only for ISP84xx when IDC ACK
needs to be posted and disable ACB mailbox command completion
is in intermediate state.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:
qla4xxx Unable to clear DDB indices when logout fails due to
failure of connection close mbox command.

Root cause:
If login to session fail, iscsiadm make call to destroy_session.
qla4xxx driver does not free ddb index map before free_ddb()

Fix:
Clear DDB Index map before free_ddb in "destroy_session"
in case of connection close mailbox command failure with 4005h.
Signed-off-by: Nilesh Javali <nilesh.javali@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

If waiting for signals was interrupted then the device was put to
FAILED state. Use msleep instead of msleep_interruptible to handle
this correctly.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Nilesh Javali <nilesh.javali@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Removed AF_83XX_NO_FWDUMP flag and directly checking IDC_CTRL bit1
while taking minidump, to check for graceful reset.
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Issue:

For ISP8xxx, adapter initialization is not retried if
qla4xxx_initialize_adapter fails.

Fix:

If qla4xxx_initialize_adapter fails, first check if failure is due to IRQs not
attached in order to skip retrial, then free the IRQs and then retry
initializing the adapter.
Signed-off-by: Nilesh Javali <nilesh.javali@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Setting DMA bit mask 64 and roll back to 32 if not supported.
Signed-off-by: Minh Tran <minhduc.tran@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

scsi_cmnd n io_task was not NULL when
 - Link goes down while IO was happening and session is closed.
 - Task for which TMF was sent.
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

In iSCSI Boot scenario, when machine is reboot/shutdown phase
the active sessions are not closed. Driver queue cleanup is
done as part of unload and device is disabled.

Sessions are still active, iSCSI commands are issued from
session which comes to driver, as driver cleanup and device
disabled there is kernel stack dump with errors.

Fix is invoking iscsi_session_failure with ISCSI_ERR_INVALID_HOST
on all the active sessions when shutdown routine is called.
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The doorbel format has been updated to support additonal functionalities
of SKH-R adapter. These changes are made such that older FW also works fine.
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

The 100Mbps port speed macro used was not proper.
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

When an MBX command timeout happens,the resources associated with
the MBX command were freed. If FW were to give the response to
host after the timeout value set by driver then driver crashes as the MBX Cmd
resources were already freed.

This patch fixes this issue by maintaing a state flag for each of
the MBX command posted/timedout/completed.
Signed-off-by: John Soni Jose <sony.john-n@emulex.com>
Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>