[SCSI] libfcoe: add protocol description of FIP VN2VN mode

The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode.  It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services.  This is point-to-multipoint.  There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.

This patch defines the new message type and subtypes as well as
one new descriptor type used by VN2VN mode.

These are all still at the proposed stage and subject to change.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: track FIP exchanges

When an exchange is received with a FIP encapsulation, we need
to know that the response must be sent via FIP and what the original
ELS opcode was.  This becomes important for VN2VN mode, where we may
receive FLOGI or LOGO from several peer VN_ports, and the LS_ACC or
LS_RJT must be sent FIP-encapsulated with the correct sub-type.

Add a field to the struct fc_frame, fr_encaps, to indicate the
encapsulation values.  That term is chosen to be neutral and
LLD-agnostic in case non-FCoE/FIP LLDs might find it useful.

The frame fr_encaps is transferred from the ingress frame to the
exchange by fc_exch_recv_req(), and back to the outgoing frame
by fc_seq_send().

This is taking the last byte in the skb->cb array.  If needed,
we could combine the info in sof, eof, flags, and encaps
together into one field, but it'd be better to do that if
and when its needed.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: add FLOGI state to rport for VN2VN

The FIP proposal for VN_port to VN_port point-to-multipoint
operation requires a FLOGI be sent to each remote port.
The FLOGI is sent with the assigned S_ID and D_IDs of the
local and remote ports.  This and the response get
FIP-encapsulated for Ethernet.

Add FLOGI state to the remote port state machine.
This will be skipped if not in point-to-multipoint mode.

To reduce a little duplication between PLOGI and FLOGI
response handling, added fc_rport_login_complete(), which
handles the parameters for the rdata struct.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: Add local port point-to-multipoint flag

For VN_port to VN_port mode, the transport sets the port_id and
there's no lport FLOGI.  This is similar to FC loop mode.

Add a point_to_multipoint flag that indicates the local port is in
point-to-multipoint mode.  This skips FLOGI and discovery.
It also skips resetting the port_id on resets other than link down.

Add function fc_lport_set_local_id() that sets the local port_id.
This is called by libfcoe on behalf of the low-level driver
to set the port_id when the link comes up.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: fcoe: fnic: change fcoe_ctlr_init interface to specify mode

There are three modes that libfcoe currently supports, and a new one
is coming.  Change the fcoe_ctlr_init() interface to add the mode
desired.  This should not change any functionality.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: add discovery-private pointer for LLD

For VN_port to VN_port mode, FIP will do discovery and needs a
way to find its state from the local port or discovery structure.
It seems that any other LLD that implements its own discovery
would also need something like this.

Replace disc->lport with disc->priv, and use container_of to
find the lport.  We could use disc->priv for that, but
container_of is smaller and faster.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: convert FIP to lock with mutex instead of spin lock

It turns out most of the FIP work is now done from worker threads
or process context now, so there's no need to use a spin lock.

Change to use mutex instead of spin lock and delayed_work instead
of a timer.

This will make it nicer for the VN_port to VN_port feature that
will interact more with the libfc layers requiring that
spinlocks not be held.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: provide space for LLD after remote port structure

Add pre-zeroed space after the allocation for fc_rport_priv
for use by the lower-level driver.

This is primarily for VN2VN FIP mode, but could be used in
other ways someday.

The space required is specified in lport->rport_priv_size.

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: convert rport lookup to be RCU safe

To allow LLD to do lookups on rports without grabbing a mutex,
make them RCU-safe.  The caller of lport->tt.rport_lookup will
have the choice of holding disc_mutex or the rcu_read_lock().

Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] fcoe: adds src and dest mac address checking for fcoe frames

This is  per FC-BB-5 Annex-D recommendation and per that
if address checking fails then drop the frame.

FIP code paths are already doing this so only needed for fcoe
frames.

The src address checking is limited to only fip mode since
this might break non-fip mode used in p2p due to used OUI
based addressing in some p2p code paths, going forward FIP
will be the only mode, therefore limited this to only FIP
mode so that it won't break non-fip p2p mode for now.

-v2
Removes FCOE packet type checking since fcoe_rcv is
registered to receive only FCoE type packets from netdev
and it is already checked by netdev.

Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] fcoe: cleans up fcoe_disable and fcoe_enable

The fc_fabric_logoff and fc_fabric_login are redundant
here after recently added fcoe_ctlr_link_down/up to
these functions, therefore this patch removes logoff
and login to only use link down and up here. This works
best for their current usages with fcoe DCB link down or up.

This also works well to avoid EIO errors when fcoe DCB link
goes down as lport state moves out of ready quickly from
fcoe_ctlr_link_down and that allows re-queuing timed out IOs
for this case also.

Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: IO errors on link down due to cable unplug

In this case, sync IO fails with EIO(5) errors as:-

"Thread:1 System call error:5 - Input/output error (::pwrite() failed)".

This is due to IO time out while libfc doing link down processing
to block all rports and if timed out IO was at last retry
attempt then it fails to user with EIO error followed by
these log messages.

[77848.612169] host2: rport bf0015: Delete port
[77848.612221] host2: rport e10aef: work delete
[77848.612232] host2: rport e10002: work event 3
[77848.612422] sd 2:0:1:1: [sdi] Unhandled error code
[77848.612426] sd 2:0:1:1: [sdi] Result: hostbyte=DID_ERROR
driverbyte=DRIVER_OK
[77848.612431] sd 2:0:1:1: [sdi] CDB: Write(10): 2a 00 00 00 11 20 00 00 20 00
[77848.612445] end_request: I/O error, dev sdi, sector 4384
[77848.612553] sd 2:0:1:2: [sdj] Unhandled error code

To fix these EIO errors, such timed out incomplete IOs needs
to be re-queued without counting retry attempt and this patch
does that using DID_REQUEUE scsi code.

Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] fcoe: make it possible to verify fcoe with sparse

Analyzing fcoe with sparse currently fails. This is because struct
fcoe_rcv_info contains two enum members that have been declared with
__attribute__((packed)). Apparently gcc honors this attribute while sparse
ignores it. The result is that sizeof(struct fcoe_rcv_info)
== sizeof(struct sk_buff::cb) == 48 on a 64-bit system according to gcc, but
not according to sparse. The patch below modifies the definition of
struct fcoe_rcv_info such that gcc and sparse interpret this structure
definition in the same way. The current sparse output is as follows:

$ cd linux-2.6.34
$ make C=2 M=drivers/scsi/fcoe modules
 CHECK   drivers/scsi/fcoe/fcoe.c

include/scsi/fc_frame.h:81:9: error: invalid bitfield width, -1.
 CC [M]  drivers/scsi/fcoe/fcoe.o
 CHECK   drivers/scsi/fcoe/libfcoe.c

include/scsi/fc_frame.h:81:9: error: invalid bitfield width, -1.
drivers/scsi/fcoe/libfcoe.c:56:37: error: invalid initializer

Signed-off-by: Bart Van Assche <bart.vanassche@gmail.com>
Cc: jeykholt@cisco.com
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: fix slowpath error from WARN_ON in fc_fcp_send_data

This is exposed by a mpio test using EMC CLARiiON targets when LUN
tresspassing happens, the burst length from the XFER_READY for the
MODE SELECT(10) is 19 bytes, much smaller than FC_MIN_MAX_PAYLOAD as
256 bytes. This patch removes the related two WARN_ON()s.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] scsi_dh_rdac: Add Dell MD36xxi controller into RDAC device list

This patch is to add next generation of Dell iSCSI PowerVault
controller MD36xxi into RDAC device list.

Signed-off-by: Yanqing Liu <Yanqing_Liu@Dell.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] dpt_i2o: move range check forward

The check to test that "bus_no" was valid came after we had already used
it as an array offset.  This patch moves it forward.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] lpfc 8.3.15: Update driver version 8.3.15

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] lpfc 8.3.15: Add target queue depth throttling

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] lpfc 8.3.15: FCoE Related Fixes

FCoE Related Fixes
- Correct find-next-FCF routine so that it searches at next FCF rather
  than current one.
- Enhanced round-robin FCF failover algorithm to re-start on "New FCF"
  async event
- Update the manner in which we look at FCFs while they may be in
  their discovery state.
- Use LPFC_FCOE_NULL_VID macro when checkinf for valid vlan_id for FCF

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] lpfc 8.3.15: BSG, Discovery, and Misc fixes

- BSG interface related:
  - Fix node reference count if node is active
  - Warn if we're overwriting an active CT context

- Discovery related:
  - Clear "Ignore Reg Login" flag when purging mailbox queue
  - Pay attention to return code for fc_block_scsi_eh()
  - Stall device loss code if we're almost done when it fires
    (we're logged in, but PRLI is outstanding)

- Bugs
  - Correct DIF code for endianness issues
  - Correct where we had missed points to check txq on i/o
    completion/cleanup

Signed-off-by: Alex Iannicelli <alex.iannicelli@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: fix resource type update and add sdev and shost attributes

Setting the resource type in the ipr_update_res_entry function was incorrect in
that the top 4 bits were masked off.  The assignment has been updated to no
longer mask those bits.

Then, two new attributes were added to allow the user space utilities to more
easily get information.  The resource_type sdev attribute is set for all devices
in the adapter's configuration table and indicates the type of device.  The
fw_type shost attribute indicates the firmware type supported by the adapter.

Finally, the resource_path attribute was changed to be mode S_IRUGO.

Signed-off-by: Wayne Boyer <wayneb@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: fix transition to operational for new adapters

The method of transitioning to operational for new adapters includes using
initialization stages.  The current stage is indicated via a register read.
The final good stage in the sequence is "operational" but does not necessarily
indicate that the driver can proceed.  There is another bit that gets set in the
adapter->host interrupt register when the adapter has completed enough of its
bringup such that it can accept commands.  The driver was not checking that
bit before proceeding which led to intermittent errors and adapter resets.

The fix is to check the "transition to operational" bit in the interrupt
register after detecting that the initialization stage is "operational" and
only proceed if both are set.

Signed-off-by: Wayne Boyer <wayneb@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] scsi: add Kconfig dependency on NET

be2iscsi driver should #include linux/if_ether.h since it uses
sysfs_format_mac().

It should also depend on NET since it selects SCSI_ISCSI_ATTRS,
which depends on NET.

These changes fix a build error when CONFIG_NET is not enabled:
ERROR: "sysfs_format_mac" [drivers/scsi/be2iscsi/be2iscsi.ko] undefined!

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: The extended shift must be 16

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fix for premature buffer free

This patch fixes a bug where the buffer was being freed as soon as
 submission to HW is done.

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Remove debug print in IO path

This patch removes a Debug Print in the IO path

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Limit max_xmit_length

This patch limits max_xmit_length to 64K incase older
utilities are used

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Maintain same ITT across login

This patch ensures that the same ITT is maintained across
all login pdu's

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Adding crashdump support

These changes allow the driver to support crashdump. We need to reset the
chip incase of a crashdump

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Free tags allocated

This patch  frees tags that are already allocated in case of
failure

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fix to handle request_irq failure

This patch handles request_irq failures  by properly cleaning up

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: No return value for hwi_enable_intr

hwi_enable_intr need not return any value. This patch fixes the
that and removes code designed to handle a failure return value

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fix for freeing cid

This patch frees up the allocated cid and returns error if allocation
of tag fails.

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: pass the return from beiscsi_open_conn

This patch passes on  the value returned by
beiscsi_open_conn

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fixing the return type of functions

Fixing some functions return values that did not match with
the possible return values

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fixing return values

This patch fixes the return values as per comment from Mike Christie

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: Fix warnings from new checkpatch.pl

The latest checkpatch.pl throws some new warnings. Fixing it
to get rid of a bunch of warnings

Signed-off-by: Jayamohan Kallickal <jayamohank@serverengines.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] arcmsr: fix up bin_attr functions

Commit

commit 2c3c8bea608866d8bd9dcf92657d57fdcac011c5
Author: Chris Wright <chrisw@sous-sol.org>
Date:   Wed May 12 18:28:57 2010 -0700

    sysfs: add struct file* to bin_attr callbacks

Added an extra struct file * parameter at the beginning, which the
arcmsr binary attribute additions didn't have.  Fix this to prevent
nasty crashes.

Cc: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] aic7xxx: Remove OS utility wrappers

This patch removes malloc(), free(), and printf() wrappers from the aic7xxx
SCSI driver. I didn't use pr_debug for printf because of some 'clever' uses of
printf don't compile with the pr_debug. I didn't fix the overeager uses of
GFP_ATOMIC either because I wanted to keep this patch as simple as possible.

[jejb:fixed up checkpatch errors and fixed up missed conversion]
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] pmcraid : Remove unnecessary casts for void * pointers

Signed-off-by: Cyril Jayaprakash <cyril.jayaprakash@gmail.com>
Acked-by: Anil Ravindranath <anil_ravindranath@pmc-sierra.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: Update driver version to 5.02.00-k2

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] iscsi_transport: added new iscsi_param to display target alias in sysfs

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: wait for device_ready before device discovery

Signed-off-by: Karen Higgins <karen.higgins@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: replace all dev_info, dev_warn, dev_err with ql4_printk

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: Added support for ISP82XX

Signed-off-by: Vikas Chaudhary <Vikas Chaudhary@qlogic.com>
Signed-off-by: Karen Higgins <karen.higgins@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: Handle one H/W Interrupt at a time

Handle one H/W Interrupt at a time to prevent holding off H/W lock
for longer period of time.

Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: Fix the freeing of the buffer allocated for DMA

Fixed the DMA allocated memory freeing which wasn't taken care
in many cases.

Signed-off-by: Prasanna Mumbai <prasanna.mumbai@qlogic.com>
Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: correct return status in function qla4xxx_fw_ready

Handle fw_state "auto discovery in progress" correctly
to avoid marking adapter as offline.

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: unblock iscsi session after setting ddb state online.

Once the device goes *missing*, driver blocks the session
ie iscsi_block_session() to stall the I/O.  So after device
comes back *online*, driver needs to unblock the session ie
iscsi_unblock_session(), else I/Os will fail even if
ddb_state is ONLINE.

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla4xxx: set driver ddb state correctly in process_ddb_changed

If fw ddb state is ACTIVE mark driver ddb stat as ONLINE and
unblock iscsi session.

Signed-off-by: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Signed-off-by: Ravi Anand <ravi.anand@qlogic.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: add debugfs support

- Add debugfs support to obtain firmware trace, driver trace
  and read/write to registers.

- debugfs hierarchy:
  /sys/kernel/debug/bfa/host#
   where the host number corresponds to the one under /sys/class/scsi_host/host#

- Following are the new debugfs entries added:
  drvtrc: collect current driver trace
  fwtrc: collect current firmware trace.
  fwsave: collect last saved fw trace as a result of firmware crash.
  regwr: write one word to chip register
  regrd: read one or more words from chip register.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: update driver version string

Update driver version to 2.2.2.1

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: vport fixes

This patch fixes 3 bugs in vport create/delete.
1) Replace scsi_add_host() with scsi_add_host_with_dma()
2) Fix rmmod hang when there are vports configured. This is due to a race
condition between the workqueue destroy in pci remove context and the vport
delete works being handled. The fix is to use a counter to track the
vport delete work, so that workqueue destroy will not be called until all
configured vports are deleted from workqueue.
3) Fix rmmmod crash when there are PBC vport configured. PBC is not allowed
to be deleted dynamically. However, if someone try to delete it, it leaves the
vport is wrong state. The fix is to restore the vport back to original state
when the attempt to delete pbc vport delete is failed.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix wrong arg to callback

This patch fixes the issue of passing wrong argument to callback function.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: add ioc state checking

This patch adds ioc state checking while enabling a port.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: add description for module parameters

Add description for bfa driver module parameters.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: update to support BOFM

Update bfa driver API and data structure to support BOFM (IBM BladeCenter
Open Fabric Manager).

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix possible IO double completion

While processing the ioim in callback functions, the ioim is still in io_q.
During this time, if the itnim goes offline, the ioim is requeued from
itnim->io_q into itnim->delay_comp_q although the request is already completed.
This results in requeing the ioim into the callback queue if the ioim is not
freed by the time the ioim is requeued. This results in double completion of
the ioim. To fix this, whenever a response is received from firmware for an
ioim, deque it from io_q and enque to fcpim->comp_q. This will eliminate any
possibility of itnim picking any ioim for which the response is already
received.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix link state structure

When the FCoE Linkup event is sent to the host, the link_state
(struct bfa_pport_link_s) structure is copied to the RME buf to be sent to
the host. But the size of this structure(164 bytes) is larger than the
reserved RME buffer size(128 byes). The following changes reduce the size
of the structure to be less than RME buffer size(128 bytes):
- Remove the trunk and loop info from link_state structure, because both trunk
  and loop are not supported.
- Combine qos_vc_attr and fcf into an union.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: add dynamic queue selection

Add new bfa functionality to support dynamic queue selection (IO redirection).
IO redirection can only be enabled when QoS is disabled.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix uf post and rport fcpim state machine

BFA UF module did not hold lock when seding uf post buffer message to firmware
causing CPE-Q corruption. Fix is to check present of FCS and if FCS present
hold lock while posting UF buffers.

Handle PRLO with sending acc to it and relogin with rport. Discard fcxp
before any state change.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix chip and memory initialization

Clear PSS memory reset that is set as part of power-on-reset (pci reset).
Complete PMM memory reset before BISTR start. Clear EDRAM BISTR start bit
after fixed delay. BISTR DONE bit status is not getting set. Use a fixed
1ms delay for BISTR now. Expose PMM IT memory definitions to host.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: update to support firmware configuation

Update related data structures to support firmeare configuration.
Add AEN events related to firmware configuation.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: add PBC port disable handling

Add PBC port disable handling in BFA and return the appropriate status from
BFA APIs. In bfa_fcs_lport.c, handle OFFLINE event to avoid BFA_ASSERT.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix prli retry issues

Add a max retry limit for PRLI retries. Max retry limit (5) is same as used
in rport PLOGI. Once the retries are exhausted, invoke rport offline so that
existing logic of rport re-discovery can kick-in. Also fixed a bug in rport.c
where one less retry was happening.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix rport speed setting

When a rport goes offline, its speed setting was not reset. Subsequently, if
the rport was not deleted due to it coming back online within rport del
timeout, previously discovered speed would continue to show up. The fix is to
reset the speed when processing rport offline transition.

In rport attributes, rport's with unknown speed were indicated as TRL
enforced.  The right thing do to would be to use TRL default speed to
determine if TRL is enforced, when TRL is enabled.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: fix interrupt coalescing setting

Do not update the coalesce flag of the intr_attr struct in driver config area
on config response.  This is to prevent the coalesce flag being reported as on
after an ioc disable/enable even if it was set to off before disable.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: use standards defined timeout for ELS/CT

Use standards defined 2 * RA_TOV as a timeout for ELS Request retries.
And standards defined 3 * RA_TOV as a timeout for FC-CT Request retries.
Also, added a check to send RPSC2 to a Brocade Fabric only.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: statistics and typo fix

- Added time stamp for fcport stats reset
- Added new fileds to the statistics data structures.
- Typo removal and minor cleanup.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: ioc attributes fix

This patch fixes the APIs to obtain ioc attributes
- fix API to obtain wwpn, wwnn, and mac.
- add API to get mfg wwpn, wwnn, and mac.
- fix API to obtain wwn of boot target.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: vport state machine fix

Vport state machine does not cleanup associated lport in some states: while
waiting for fdisc response or fdisc failure state. The fixe is to cleanup lport
on vport delete in all states.  In fdisc state, discard fdisc response and
delete lport and wait for lport deletecompletion. in error state, delete lport
and wait for delete completion.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: PBC vport create

This patch enables creating PBC vport.
During fcs init, fcs will read PBC vport using bfa iocfc API and invoke fcb
callback to add the pbc vport entries into a list. The pbc vport list will be
traversed in the subsequent pci probe process and vport will be created using
fc transport provided vport create function.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: enable basic PBC support

The patch includes the driver side changes to enable basic PBC (PreBoot
Configuration) feature.
- Data structure changes and new definitions for PBC.
- APIs to access PBC info.
- Remove unused code.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bfa: enable new hardware

This patch enables support of new mezzanine cards for HP and IBM blade server.

- Add new pciids for HP and IBM mezzanine card.
- Add a new firmware image for HP mezzanine card, which is running in
  FC only mode. Rename firmware image to reflect the difference. Change the
  firmware download code accordingly for the above changes.

Signed-off-by: Jing Huang <huangj@brocade.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: driver fails to recover from injected PCIe bus errors

fixes surrounding PCIe enhanced error handling:

(1) We need to reject all request generated internaly inside the driver as well
as request arriving from the scsi mid layer when PCIe EEH is active. The fix is
to add a per adapter flag called pci_error_recovery which is checked thru out
the driver when request are generated.

(2) We don't need to call the pci_driver->remove directly from the PCIe
callbacks becuase its already called from the PCIe EEH code. In its place we are
shutting down the watchdog timer, and flushing back all pending IO.

(3) We need to save and restore the pci state across PCIe EEH handling.

Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Updated version from 2.1.1 to 2.1.2

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Added host param ISCSI_HOST_PARAM_IPADDRESS

This sysfs attribute is proven to be useful during pivot_root.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Fixed the TCP graceful termination initiation

In compliance to RFC793, a TCP graceful termination will be used
instead of an abortive termination for the case where the remote
has initiated the close of the connection.
Additionally, a TCP abortive termination will be used to close the
connection when a logout response is not received in time after a
logout request has been initiated.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Fine tuned conn destroy and context destroy timeout values

Added variables to separate the fine tuned timeout values for
connection destroy and context destroy for both 1g and 10g devices.

v2: Extended the 5771X disconnect timeout from 10s to 20s as the firmware
has a retransmission timeout of 16s.  This fixes one of the iscsi_endpoint
leak issues when the target is slow or non-responsive to our TCP FIN.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Optimized the bnx2i_stop connection clean up procedure

For cases where the iSCSI disconnection procedure times out due to
the iSCSI daemon being slow or unresponsive, the bnx2i_stop routine
will now perform hardware cleanup via bnx2i_hw_ep_disconnect on all
active endpoints so that subsequent operations will perform properly.
Also moved the mutex locks inside ep_connect and ep_disconnect so
that proper exclusivity can resolve simultaneous calls to the
ep_disconnect routine.

v2: Removed the unnecessary read lock in the bnx2i_stop

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Created an active linklist which holds bnx2i endpoints

This introduces a new active linklist which would link up all active
bnx2i_endpoints.  This will be used by subsequent patches that
follows.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] bnx2i: Separated the hardware's cleanup procedure from ep_disconnect

This patch introduces a new bnx2i_hw_ep_disconnect routine which
contains all chip related disconnect and clean up procedure of
iSCSI offload connections.  This separation is intended as a
preparation for the subsequent bnx2i_stop patch.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: Anil Veerabhadrappa <anilgv@broadcom.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] scsi_debug: fix map_region and unmap_region oops

map_region and unmap_region could access to invalid memory area since
they don't check the size boundary.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Douglas Gilbert <dgilbert@interlog.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: change endian swap key to match hardware spec change

The value used to change the endian representation on the new adapters has
changed.  This patch updates that value.

Signed-off-by: Wayne Boyer <wayneb@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: add support for new Obsidian-E embedded adapter

This patch allows the driver to recognize a new Obsidian-E based adapter that
uses a new subsystem ID.

This patch also fixes a few tab/space problems.

Signed-off-by: Wayne Boyer <wayneb@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mvsas: fix potential NULL dereference

Stanse found that in mvs_abort_task, mvi_dev is dereferenced earlier
than tested for being NULL. Move the assignment below the test.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mptfusion: print Doorbell register in a case of hard reset and timeout

Printing Doorbell register in a case of hard reset and timeout
should be useful for figuring out the state of the system.

Signed-off-by: Kei Tokunaga <tokunaga.keiich@jp.fujitsu.com>
Acked-by: "Desai, Kashyap" <Kashyap.Desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] scsi:hosts.c Fix warning: variable 'rval' set but not used

The below patch fixes a warning message generated by gcc 4.6.0
  CC      drivers/scsi/hosts.o
drivers/scsi/hosts.c: In function 'scsi_host_alloc':
drivers/scsi/hosts.c:328:6: warning: variable 'rval' set but not used

Fix this by removing the rval but placing a printk warning where it
would have been set.

Signed-off-by: Justin P. Mattock <justinmattock@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] Log msg when getting Unit Attention

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

We are working on adding support for handling of these errors,
but I think until then we should at least log a message so
tracking down problems as a result of one of these changes
is a little easier.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mptsas: fixed hot-removal processing

This patch fixes mptsas disk hot-removal processing.  The
hot-removal processing doesn't complete because of this condition.

  drivers/message/fusion/mptsas.c:
  mptsas_taskmgmt_complete()

  if ((mptsas_find_vtarget(ioc, channel, id)) && !ioc->fw_events_off)
    mptsas_queue_device_delete(...);

mptsas_queue_device_delete(), which must be called for
hot-removal, never gets called because mptsas_find_vtarget()
always returns 0 here.  At that time, the vtarget has already
been freed in mptsas_target_destroy(), and also the scsi_device
has been marked as SDEV_DEL.

As a result of the issue, port deletion functions won't get
called and the device ends up being in an incomplete state.
(Some data structures and sysfs entries, which should be
removed in hot-removal, remain.)  One side effect of this is
that a hot-addition of the device (bringing the device back
on) fails.

This patch just removes mptsas_find_vtarget() from the if-state
condition.

Signed-off-by: Kei Tokunaga <tokunaga.keiich@jp.fujitsu.com>
Acked-by: "Desai, Kashyap" <Kashyap.Desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: add MMIO write to perform BIST for 64 bit adapters

The 64 bit chip used in new adapters does not properly support the BIST register
in PCI config space.  This patch implements an alternative MMIO write reset
method.

Signed-off-by: Wayne Boyer <wayneb@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] SCSI: Support Type C RAID controller

1. To support Type C RAID controller, ACB_ADAPTER_TYPE_C, i.e. PCI device
ID: 0x1880.
Signed-off-by: Nick Cheng< nick.cheng@areca.com.tw >
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] enclosure: fix error path - actually return ERR_PTR() on error

we also need to clean up and free the cdev.

Reported-by: Jani Nikula <ext-jani.1.nikula@nokia.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvscsi: Driver version 1.5.9

Bump driver version

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvscsi: Fix possible request_limit issue

If we encounter an error when sending a management datagram (i.e. non
SCSI command, such as virtual adapter initialization command), we
end up incrementing the request_limit, even though we don't decrement
it for these commands. Fix this up by doing this increment in
the error path for SRP commands only.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvscsi: Fix error path deadlock

Fixes a deadlock that can occur if we hit a command timeout
during the virtual adapter initialization. The event done
functions are written with the assumption that no locks are held,
however, when purging requests this is not true. Fix up the
purge function to drop the lock so that the done function
is not called with the lock held, which can cause a deadlock.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvscsi: Fix softlockup on resume

This fixes a softlockup seen on resume. During resume, the CRQ
must be reenabled. However, the H_ENABLE_CRQ hcall used to do
this may return H_BUSY or H_LONG_BUSY. When this happens, the
caller is expected to retry later. This patch changes a simple
loop, which was causing the softlockup, to a loop at task level
which sleeps between retries rather than simply spinning.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvfc: Driver version 1.0.8

Bump driver version.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvfc: Add support for fc_block_scsi_eh

Adds support for fc_block_scsi_eh to block the EH handlers if
the target device is in the blocked state to ensure we don't
take devices offline.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvfc: Fix soft lockup on resume

This fixes a softlockup seen on resume. During resume, the CRQ
must be reenabled. However, the H_ENABLE_CRQ hcall used to do
this may return H_BUSY or H_LONG_BUSY. When this happens, the
caller is expected to retry later. Normally the H_ENABLE_CRQ
succeeds relatively soon. However, we have seen cases where
this can take long enough to see softlockup warnings.
This patch changes a simple loop, which was causing the
softlockup, to a loop at task level which sleeps between
retries rather than simply spinning.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>