[SCSI] mpt2sas: Added expander phy control support

Added support to send link resets, hard resets, enable/disable phys, and
changing link rates for for expanders.  This will be exported to
attributes within the sas transport layer.  A new wrapper function was
added for sending SMP passthru to expanders for phy control.

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: Added expander phy counter support

Added support to retrieve the invalid_dword_count,
running_disparity_error_count, loss_of_dword_sync_count, and
phy_reset_problem_count for expanders.  This will be exported to
attributes within the sas transport layer.  A new wrapper function was
added for sending SMP passthru to retrieve the expander phy error log.

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: staged device discovery. disable_discovery module parameter is added.

Added command line option called disable_discovery.  When enabled
on the command line, the driver will not send a port_enable when loaded
for the first time.   If port_enable is not called, then there is
no discovery of devices, as well as the sas topology.  Then later if one
desires to invoke discovery, then they will need to issue a diagnostic reset.
A diagnostic reset can be issued various ways. One of the way is throught
sysfs.

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: Hold Controller reset when another reset is in progress

Driver should not allow multiple host reset when already host reset is in
progress. It is possible that host reset was sent by scsi mid layer while there was already an host reset active,
either issued via IOCTL interface or internaly, like a config page timeout.
Since there was a host reset active, the driver would return a FAILED response
to the scsi mid layer. The solution is make sure pending host resets will
wait for the active host reset to complete before returning control
back up the call stack.

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: Fix to use sas device list instead of enclosure list for _transpor_get_enclosure_identifier.

Enclosure_identifier not being returned by mpt2sas
The driver exports callback function to the sas transport layer
for obtaining the enclosure logical id.  This function is called
_transport_get_enclosure_identifier.  The driver was searching
the wrong list for the enclosure_identifier.  The driver should be
searching the sas device list instead of enclosure list.  The
sas address that is passed to the driver is for the end device, not
enclosure.

Signed-off-by: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] cxgb3i: zero out reserved or un-used fields.

Zero out the reserved or un-used CPL message fields to prevent any garbage
value.

Signed-off-by: Karen Xie <kxie@chelsio.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] hpsa: sanitize max commands

Some controllers might try to tell us they support 0 commands
in performant mode.  This is a lie told by buggy firmware.
We have to be wary of this lest we try to allocate a negative
number of command blocks, which will be treated as unsigned,
and get an out of memory condition.

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

[SCSI] hpsa: separate intx and msi/msix interrupt handlers

There are things which need to be done in the intx
interrupt handler which do not need to be done in
the msi/msix interrupt handler, like checking that
the interrupt is actually for us, and checking that the
interrupt pending bit on the hardware is set (which we
weren't previously doing at all, which means old controllers
wouldn't work), so it makes sense to separate these into
two functions.

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

[SCSI] hpsa: forbid hard reset of 640x boards

The 6402/6404 are two PCI devices -- two Smart Array controllers
-- that fit into one slot.  It is possible to reset them independently,
however, they share a battery backed cache module.  One of the pair
controls the cache and the 2nd one access the cache through the first
one.  If you reset the one controlling the cache, the other one will
not be a happy camper.  So we just forbid resetting this conjoined
mess.

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

[SCSI] hpsa: Fix hard reset code.

Smart Array controllers newer than the P600 do not honor the
PCI power state method of resetting the controllers.  Instead,
in these cases we can get them to reset via the "doorbell" register.

This escaped notice until we began using "performant" mode because
the fact that the controllers did not reset did not normally
impede subsequent operation, and so things generally appeared to
"work".  Once the performant mode code was added, if the controller
does not reset, it remains in performant mode.  The code immediately
after the reset presumes the controller is in "simple" mode
(which previously, it had remained in simple mode the whole time).
If the controller remains in performant mode any code which presumes
it is in simple mode will not work.  So the reset needs to be fixed.

Unfortunately there are some controllers which cannot be reset by
either method. (eg. p800).  We detect these cases by noticing that
the controller seems to remain in performant mode even after a
reset has been attempted.  In those case, we proceed anyway,
as if the reset has happened (and skip the step of waiting for
the controller to become ready -- which is expecting it to be in
"simple" mode.)  To sum up, we try to do a better job of resetting
the controller if "reset_devices" is set, and if it doesn't work,
we print a message and try to continue anyway.

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

[SCSI] hpsa: factor out the code to reset controllers on driver load

for kdump support

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

[SCSI] hpsa: factor out hpsa_find_cfg_addrs.

Rationale for this is that I will also need to use this code
in fixing kdump host reset code prior to having the hba structure.

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

[SCSI] hpsa: make hpsa_find_memory_BAR not require the per HBA structure.

Rationale for this is that in order to fix the hard reset code used
by kdump, we need to use this function before we even have the per
HBA structure.

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

[SCSI] hpsa: Make "hpsa_allow_any=1" boot param enable Compaq Smart Arrays.

We were previously only accepting HP boards.

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

[SCSI] hpsa: add new controllers

Add 5 CCISSE smart array controllers

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

[SCSI] arcmsr: Support 1024 scatter-gather list entries and improve AP while FW trapped and behaviors of EHs

1. To support 4M/1024 scatter-gather list entry, reorganize struct
   ARCMSR_CDB and struct CommandControlBlock
2. To modify arcmsr_probe
3. In order to help fix F/W issue, add the driver mode for type B card
4. To improve AP's behavior while F/W resets
5. To unify struct MessageUnit_B's members' naming in all OS drivers'
6. To improve error handlers, arcmsr_bus_reset(), arcmsr_abort()
7. To fix the arcmsr_queue_command() in bus reset stage, just let the
   commands pass down to FW, don't block

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

[SCSI] libfc: fix indefinite rport restart

Remote ports were restarting indefinitely after getting
rejects in PRLI.

Fix by adding a counter of restarts and limiting that with
the port login retry limit as well.

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: Fix remote port restart problem

This patch somewhat combines two fixes to remote port handing in libfc.

The first problem was that rport work could be queued on a deleted
and freed rport.  This is handled by not resetting rdata->event
ton NONE if the rdata is about to be deleted.

However, that fix led to the second problem, described by
Bhanu Gollapudi, as follows:
> Here is the sequence of events. T1 is first LOGO receive thread, T2 is
> fc_rport_work() scheduled by T1 and T3 is second LOGO receive thread and
> T4 is fc_rport_work scheduled by T3.
>
> 1. (T1)Received 1st LOGO in state Ready
> 2. (T1)Delete port & enter to RESTART state.
> 3. (T1)schdule event_work, since event is RPORT_EV_NONE.
> 4. (T1)set event = RPORT_EV_LOGO
> 5. (T1)Enter RESTART state as disc_id is set.
> 6. (T2)remember to PLOGI, and set event = RPORT_EV_NONE
> 6. (T3)Received 2nd LOGO
> 7. (T3)Delete Port & enter to RESTART state.
> 8. (T3)schedule event_work, since event is RPORT_EV_NONE.
> 9. (T3)Enter RESTART state as disc_id is set.
> 9. (T3)set event = RPORT_EV_LOGO
> 10.(T2)work restart, enter PLOGI state and issues PLOGI
> 11.(T4)Since state is not RESTART anymore, restart is not set, and the
> event is not reset to RPORT_EV_NONE. (current event is RPORT_EV_LOGO).
> 12. Now, PLOGI succeeds and fc_rport_enter_ready() will not schedule
> event_work, and hence the rport will never be created, eventually losing
> the target after dev_loss_tmo.

So, the problem here is that we were tracking the desire for
the rport be restarted by state RESTART, which was otherwise
equivalent to DELETE.  A contributing factor is that we dropped
the lock between steps 6 and 10 in thread T2, which allows the
state to change, and we didn't completely re-evaluate then.

This is hopefully corrected by the following minor redesign:

Simplify the rport restart logic by making the decision to
restart after deleting the transport rport.  That decision
is based on a new STARTED flag that indicates fc_rport_login()
has been called and fc_rport_logoff() has not been called
since then.  This replaces the need for the RESTART state.

Only restart if the rdata is still in DELETED state
and only if it still has the STARTED flag set.

Also now, since we clear the event code much later in the
work thread, allow for the possibility that the rport may
have become READY again via incoming PLOGI, and if so,
queue another event to handle that.

In the problem scenario, the second LOGO received will
cause the LOGO event to occur again.

Reported-by: Bhanu Gollapudi <bprakash@broadcom.com>
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] fnic: fnic_scsi.c: clean up

In fnic_abort_cmd() and fnic_device_reset() assign `rport' earlier to make
FNIC_SCSI_DBG() calls cleaner.

In fnic_clean_pending_aborts() `rport' is not used.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: lport state is enum not bit mask

lport state is enum not bit mask.

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] libfcoe: Check for order and missing critical descriptors for FIP ELS requests

As per FC-BB-5 rev.2, section 7.8.7.1, strict ordering of FIP descriptors
is required for ELS requests. Also, look for missing and duplicate critical
descriptors.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: Host doesnt handle CVL to NPIV ports

Clear virtual link for NPIV ports is now handled by resetting
the matching vnport.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: Handle duplicate critical descriptors

As per FC-BB-5 rev 2, section 7.8.6.2, malformed FIP frame shall be
discarded. Drop discovery adv, ELS and CLV's with duplicate critical
descriptors.

[Resending after incorporating Joe's review comments]

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: update FIP FCF D flag from advertisments

Allow the D flag (indicating that keep-alives are not needed) to
be updated dynamically from received FIP advertisements.

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: Use fka_period as periodic timeouts to age out fcf if

keep alives are disabled due to fd_flags set and also
stop updating keep alive values in that case.

Update select fcf time only if fcf is not already selected or
select time is not already determined from parse adv, and then
have select time cleared only once after fcf is selected.

Changed deadline check to time_after_eq() from time_after()
since now next timeout will be on exact 2.5 times FKA followed
by first advertisement.

Signed-off-by: Vasu Dev <vasu.dev@intel.com>
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: fix lenient aging of FCF advertisements

[This patch has several improvements to the code in
the fip timers.  It hasn't been tested yet.
I'm sending it out for review.  Vasu, perhaps you can
merge this with your patch and test it together.]

The current code allows an advertisement to be used
even if it has been 3 times the FCF keep-alive
advertisement period (FKA) since one was received from
that FCF.  The spec. calls for 2.5 times FKA.

Fix this and make sure we detect missed keep-alives promptly.

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: Handle unsolicited PRLO request

Resubmitting after incorporating Joe's review comment.

Unsolicited PRLO request is now handled by sending LS_ACC,
and then relogin to the remote port if an N-port login
session exists for that remote port.

Note that this patch should be applied on top of Joe Eykholt's
"Fix remote port restart problem" patch.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] fcoe: clean up TBD comments in FCoE prototype header

Some old comments in fc_fcoe.h say TBD long after the
standard has been passed by T11.  Clean them 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] libfc: Honor LS_ACC response codes for PRLI

As per FC-LS Rev 1.62 table 46, response codes are handled as follows:

1. If the Req executed is true, PRLI is accepted.
2. If Req executed is not set,  if resp code is 5,
   PRLI is not retried and port is logged out.
3. If resp code is anything apart from 1 or 5, PRLI is retired
   upto max retry count.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfc: Retry a rejected PRLI request

Retry upto max_rport_retry_count when a target responds with
LS_RJT for a PRLI request.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: No solicitation if adv is dropped

Host does not send discovery solicitation messages if Disc. Adv
from FCF are dropped. It restarts sending solicitation only
after receiving a Discovery Adv. from FCF. Fix is to restart
solicitation immediately after CVL processing.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] libfcoe: Avoid hang when receiving non-critical descriptors

Avoid infinite loop while processing FIP ELS or discovery
advertisement with non-critical descriptors.

Signed-off-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Acked-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: FIP link keep-alive should continue while logged off

A check in fcoe_ctlr_send_keep_alive() returns if there's no
port_id for the local port.  This could miss a keep alive if
we just did a host reset and have logged off and will log back in.

Return only if we are doing the port keep alive, in which case
we need to be logged in.

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] ipr: move setting of the allow_restart flag for vsets and disks

A problem was found where the call to scsi_add_device() fails intermittently
for an adapter.  This is caused when __scsi_add_device() returns -ENODEV as
a result of not calling scsi_probe_and_add_lun() since the call to
scsi_host_scan_allowed() fails.  scsi_host_scan_allowed() fails because the
adapter state is set to SHOST_RECOVERY instead of SHOST_RUNNING. The state of
the adapter is being set to SHOST_RECOVERY by scsi_eh_scmd_add() during
error handling.

This problem is avoided by moving the setting of the allow_restart flag to
later in the device initialization sequence.  This prevents further error
handling if we get a NOT_READY response from a TUR command by causing
scsi_check_sense() to return SUCCESS.  Therefore, scsi_eh_scmd_add() will
not run and the adapter state will remain as SHOST_RUNNING.

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] be2iscsi: fix null dereference on error path

"phba" is always null here so we can't dereference it.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] be2iscsi: fix memory leak on error path

I added a kfree(pwrb_arr) in front of the return.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: add writeq definition if needed

Compiling the driver will fail on 32 bit powerpc and other
architectures where writeq is not defined.  This patch adds a
definition for writeq.

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] be2iscsi: fix disconnection cleanup

This patch fixes 4 bugs in the connection connect/disconnect
cleanup path.

1. If beiscsi_open_conn fails beiscsi_free_ep was always being
called, and if beiscsi_open_conn failed because beiscsi_get_cid
failed then we would free an unallocated cid.

2. If beiscsi_ep_connect failed due to a beiscsi_open_conn failure
it was leaking iscsi_endpoints.

3. beiscsi_ep_disconnect was leaking iscsi_endpoints.
beiscsi_ep_disconnect should free the iscsi_endpoint. We cannot
do it in beiscsi_conn_stop because that is only called for
iscsi connection cleanup. If beiscsi_ep_connect returns
success, but then the poll function fails or the connect
times out then beiscsi_ep_disconnect will be called to clean
up the ep. The conn_stop callout will not be called in that path.

4. beiscsi_conn_stop was freeing the iscsi_endpoint then accessing
it a couple lines later.

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

[SCSI] lpfc 8.3.14: Update Driver version to 8.3.14

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 8.3.14: SCSI and SLI API fixes

- Fixed accounting of allocated SCSI buffers when post sgl fails.
- Restrict scsi buffer allocation based on LUN count (sdev_cnt).
- Create __lpfc_sli_free_rpi that doesn't take out the hbalock.
- Modify lpfc_sli_free_rpi to call __lpfc_sli_free_rpi.
- Call __lpfc_sli_free_rpi in lpfc_cleanup_pending_mbox.
- Do not swap the strings returned in mailbox commands and do
  not swap byte aligned data in VPD.

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 8.3.14: FCoE Discovery Fixes

- Prevent unregistring of unused FCF when FLOGI is pending.
- Prevent point to point discovery on a FCoE HBA.
- Fixed FCF discovery failure after swapping FCoE port by
  switching over to fast failover method when no FCF matches in-use 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 8.3.14: PCI fixes and enhancements

- Allow enabling MSI-X intterupts with fewer vectors than requested
  by looking at the return value from pci_enable_msix.
- Implemented driver PCI AER error handling routines for supporting
  AER error recovering on SLI4 devices.
- Remove redundant SLI_ACTIVE checks

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] pmcraid: MSI-X support and other changes

1. MSI-X interrupt support
2. Driver changes to support new maxRAID controller FW version. The
   changes are mainly done to handle async notification changes done in
   newer controller FW version.
3. Added state change notifications to notify applications of controller
   states.

Signed-off-by: Anil Ravindranath <anil_ravindranath@pmc-sierra.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] lpfc 8.3.13: Update Driver Version to 8.3.13

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 8.3.13: Add TX Queue Support for SLI4 ELS commands.

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 8.3.13: Misc fixes

- Change the Max receive size on CIN FCFs to 0x800
- (From linux community) Check boundary before checking for NULL.
- Update last completion time for completed I/O to prevent heartbeat.
- Add Balius PCI Device IDs

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 8.3.13: SCSI specific changes

- Fix hba_queue_depth to reflect actual available XRIs
- Add support for new SLER specific firmware status codes.
- Free SCSI buffer when iotag allocation fails.

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 8.3.13: Initialization code clean up and fixes.

- Add poll or wait flag parameter to hba_init_link and hba_down_link.
- (From Linux Community) Make return with ENXIO negative.
- Remove unused INB code from driver.
- Prevent block_magmt_io from returning until mailbox is inactive.

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 8.3.13: FC Discovery Fixes and enhancements.

- Retry PLOGI up to 48 times when LS_RJT reason is
  "Unable to supply requested data."
- When dev loss timeout occures do not change state if there
  is an outstanding REG_LOGIN.
- Add logic to ignore REG_LOGIN completion if discovery is
  restarted while waiting for REG_LOGIN.
- Only change state on REG_LOGIN completion if still in
  state waiting for REG_LOGIN completion.
- Only send ADISCs to FCP-2 Targets (not Initiators).

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: add endian swap enablement for 64 bit adapters

A change in the hardware design of the chip for the new adapters changes the
default endianness of MMIO operations.  This patch adds a register definition
which when written to with a predefined value will change the endianness
back to what the driver expects.

This patch also fixes two problems found during testing.

First, the first reserved field in the ipr_hostrcb64_fabirc_desc structure only
reserved one byte.  The correct amount to reserve is 2 bytes.

Second, the reserved field of the ipr_hostrcb64_error structure only reserved
2 bytes.  The correct amount to reserve is 16 bytes.

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] dpt_i2o: Use GFP_ATOMIC when a lock is held

The function adpt_i2o_post_wait is called from several places, in some of
which, such as adpt_abort, a lock may be held.

The functions adpt_i2o_reparse_lct and adpt_i2o_lct_get are called from
several places, including adpt_rescan where a lock may be held.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@gfp exists@
identifier fn;
position p;
@@

fn(...) {
 ... when != spin_unlock_irqrestore
     when any
 GFP_KERNEL@p
 ... when any
}

@locked@
identifier gfp.fn;
@@

spin_lock_irqsave(...)
...  when != spin_unlock_irqrestore
fn(...)

@depends on locked@
position gfp.p;
@@

- GFP_KERNEL@p
+ GFP_ATOMIC
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Updated version number to 8.03.03-k0.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: For ISP 23xx, select user specified login timeout value if greater than minuimum value(4 secs).

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Removed redundant check for ISP 84xx.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Enable CRB based doorbell posting for request queue as default for ISP 82xx.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: T10 DIF enablement for 81XX

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Fix flash write failure on ISP82xx.

Driver was not unprotecting correctly, use correct bits
to unprotect the flash on ISP 82xx.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Handle outstanding mbx cmds on hung f/w scenarios.

Outstanding mailbox commands, have no way to recover on f/w hung, and we
timeout on waiting for mbx response. This in turn affects the recovery process
as follows:
- We might already be in dpc while waiting for mbx to complete, so recovery for
that pci function will never get invoked. Reset Timeout (10 sec) is far less
than mbx timeout (30 sec).
- Other mbx cmds will get stuck due to serial mbx access.

Solution is to identify fw-hung scenario and handle outstanding mbx commands to
have an early-exit instead of waiting for response.
Other mbx commands waiting for access will also do an early-exit if fw-hung is
still applicable.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Do not enable VP in non fabric topology.

After topology change ISP is reset and VPs are re-enabled.
If the topology is not fabric, VPs could falsely get enabled.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Support for loading Unified ROM Image (URI) format firmware file.

Used bootloder address from FLT while loading FW from flash as well.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Make the FC port capability mutual exclusive.

In case of both target and initiator capabilities reported by fc port,
the fc port port capability is made mutualy exclusive with priority given
for target capabilities.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Add internal loopback support for ISP81xx.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Limit rport-flaps during link-disruptions.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Correct async-srb issues.

* hold the hardware_lock throughout the duration of ctx-sp
  timeout handling -- could result in use-after-free oops.

* retry a timed-out login-request.

* done() routines are called with the hardware-lock held, issue
qla2x00_mark_device_lost() with proper 'defer' flag.

* FCP2 capabilities are only relevant to target devices.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Correct use-after-free oops seen during EH-abort.

Hold a reference to the srb (sp) while aborting an I/O -- as the
I/O can/will complete from within the interrupt-context.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Fix cpu-affinity usage for non-capable ISPs.

The TMFs used for pre-24xx ISPs incorrectly assumed 'cpu' tag
data could be valid.  These chips have no multi-q/cpu-affinity
support.  This corrects an oops seen on ISP23xx parts.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Add portid to async-request messages.

This helps to correlate submission/completion messages during
triaging.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] qla2xxx: Clear drive active CRB register when not in use.

The CRB drive active register is cleared when driver is unloaded
or when driver enters failed state.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] hpsa: add entry to MAINTAINERS

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

[SCSI] hpsa: fix block fetch table problem.

We have 32 (MAXSGENTRIES) scatter gather elements embedded
in the command.  With all these, the total command size is
about 576 bytes.  However, the last entry in the block fetch table
is 35.  (the block fetch table contains the number of 16-byte chunks
the firmware needs to fetch for a given number of scatter gather
elements.)  35 * 16 = 560 bytes, which isn't enough.  It needs to be
36. (36 * 16 == 576) or, MAXSGENTRIES + 4.  (plus 4 because there's a
bunch of stuff at the front of the command before the first scatter
gather element that takes up 4 * 16 bytes.)  Without this fix, the
controller may have to perform two DMA operations to fetch the
command since the first one may not get the whole thing.

Signed-off-by: Don Brace <brace@beardog.cce.hp.com>
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] hpsa: expose controller firmware revision via /sys.

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

[SCSI] hpsa: Add hpsa.txt to Documentation/scsi

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

[SCSI] hpsa: remove unused firm_ver member of the per-hba structure

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

[SCSI] hpsa: factor out hpsa_enter_performant_mode

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

[SCSI] hpsa: remove unused variable trans_offset

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

[SCSI] hpsa: factor out hpsa_wait_for_mode_change_ack

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

[SCSI] hpsa: mark hpsa_mark_hpsa_put_ctlr_into_performant_mode as __devinit

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

[SCSI] hpsa: clean up debug ifdefs

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

[SCSI] hpsa: check that simple mode is supported

before trying to enter simple mode transport method.

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

[SCSI] hpsa: factor out hpsa_enter_simple_mode

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

[SCSI] hpsa: factor out hpsa_p600_dma_prefetch_quirk

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

[SCSI] hpsa: factor out hpsa_enable_scsi_prefetch

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

[SCSI] hpsa: factor out hpsa-CISS-signature-present

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

[SCSI] hpsa: hpsa factor out hpsa_find_board_params

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

[SCSI] hpsa: fix leak of ioremapped memory in hpsa_pci_init error path.

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

[SCSI] hpsa: factor out hpsa_find_cfgtables

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

[SCSI] hpsa: factor out hpsa_wait_for_board_ready

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

[SCSI] hpsa: factor out hpsa_find_memory_BAR

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

[SCSI] hpsa: remove redundant board_id parameter from hpsa_interrupt_mode

and delete duplicated comment

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

[SCSI] hpsa: factor out hpsa_board_disabled

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

[SCSI] hpsa: factor out hpsa_lookup_board_id

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

[SCSI] hpsa: save pdev pointer in per hba structure early to avoid passing it around so much.

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

[SCSI] hptiop: Eliminate a NULL pointer dereference

The end of the function is reachable both when host is and is not NULL.

A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
@r exists@
expression E,E1;
identifier f;
statement S1,S2,S3;
@@

if ((E == NULL && ...) || ...)
{
  ... when != if (...) S1 else S2
      when != E = E1
* E->f
  ... when any
  return ...;
}
else S3
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mptsas: fix hangs caused by ATA pass-through

I may have an explanation for the LSI 1068 HBA hangs provoked by ATA
pass-through commands, in particular by smartctl.

First, my version of the symptoms.  On an LSI SAS1068E B3 HBA running
01.29.00.00 firmware, with SATA disks, and with smartd running, I'm seeing
occasional task, bus, and host resets, some of which lead to hard faults of
the HBA requiring a reboot.  Abusively looping the smartctl command,

    # while true; do smartctl -a /dev/sdb > /dev/null; done

dramatically increases the frequency of these failures to nearly one per
minute.  A high IO load through the HBA while looping smartctl seems to
improve the chance of a full scsi host reset or a non-recoverable hang.

I reduced what smartctl was doing down to a simple test case which
causes the hang with a single IO when pointed at the sd interface.  See
the code at the bottom of this e-mail.  It uses an SG_IO ioctl to issue
a single pass-through ATA identify device command.  If the buffer
userspace gives for the read data has certain alignments, the task is
issued to the HBA but the HBA fails to respond.  If run against the sg
interface, neither the test code nor smartctl causes a hang.

sd and sg handle the SG_IO ioctl slightly differently.  Unless you
specifically set a flag to do direct IO, sg passes a buffer of its own,
which is page-aligned, to the block layer and later copies the result
into the userspace buffer regardless of its alignment.  sd, on the other
hand, always does direct IO unless the userspace buffer fails an
alignment test at block/blk-map.c line 57, in which case a page-aligned
buffer is created and used for the transfer.

The alignment test currently checks for word-alignment, the default
setup by scsi_lib.c; therefore, userspace buffers of almost any
alignment are given directly to the HBA as DMA targets.  The LSI 1068
hardware doesn't seem to like at least a couple of the alignments which
cross a page boundary (see the test code below).  Curiously, many
page-boundary-crossing alignments do work just fine.

So, either the hardware has an bug handling certain alignments or the
hardware has a stricter alignment requirement than the driver is
advertising.  If stricter alignment is required, then in no case should
misaligned buffers from userspace be allowed through without being
bounced or at least causing an error to be returned.

It seems the mptsas driver could use blk_queue_dma_alignment() to advertise
a stricter alignment requirement.  If it does, sd does the right thing and
bounces misaligned buffers (see block/blk-map.c line 57).  The following
patch to 2.6.34-rc5 makes my symptoms go away.  I'm sure this is the wrong
place for this code, but it gets my idea across.

Acked-by: "Desai, Kashyap" <Kashyap.Desai@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] mpt2sas: DIF Type 2 Protection Support

Adding DIF Type 2 protection support, as well as turning on 32 byte cdb's,
and setting the cdb length for > 16 byte in the SCSI_IO->control parameter.

Signed-off-by: Martin Petersen <martin.petersen@oracle.com>
Signed-off-by: Eric Moore <eric.moore@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ibmvscsi: Fix oops when an interrupt is pending during probe

A driver needs to be ready to take an interrupt as soon as it registers
an interrupt handler. I noticed the following oops when testing kdump:

ipr: IBM Power RAID SCSI Device Driver version: 2.5.0 (February 11, 2010)
ibmvscsi 30000002: SRP_VERSION: 16.a
ibmvscsi 30000002: SRP_VERSION: 16.a
Unable to handle kernel paging request for data at address 0x00000000
...
pc: c000000004085e34: .tasklet_action+0xf4/0x1dc
...
c000000004086fe4 .__do_softirq+0x16c/0x2c0
c00000000403138c .call_do_softirq+0x14/0x24
c00000000400ee14 .do_softirq+0xa0/0x104
c00000000408690c .irq_exit+0x70/0xd0
c00000000400f190 .do_IRQ+0x214/0x2a8
c000000004004804 hardware_interrupt_entry+0x1c/0x98
--- Exception: 501 (Hardware Interrupt) at c00000000400c544 .raw_local_irq_restore+0x48/0x54
c00000000465d2a8 ._raw_spin_unlock_irqrestore+0x74/0xa0
c0000000040e7f00 .__setup_irq+0x2ec/0x3f0
c0000000040e8198 .request_threaded_irq+0x194/0x22c
c00000000446d854 .rpavscsi_init_crq_queue+0x284/0x3f0
c00000000446c764 .ibmvscsi_probe+0x688/0x710
c00000000402903c .vio_bus_probe+0x37c/0x3e4
c000000004403f10 .driver_probe_device+0xec/0x1b8
c000000004404088 .__driver_attach+0xac/0xf4
c000000004403184 .bus_for_each_dev+0x98/0x104
c000000004403c98 .driver_attach+0x40/0x60
c0000000044026f0 .bus_add_driver+0x154/0x324
c0000000044045d0 .driver_register+0xe8/0x1ac
c00000000402b2a8 .vio_register_driver+0x54/0x74
c000000004933ea4 .ibmvscsi_module_init+0x80/0xc0
c000000004009834 .do_one_initcall+0x98/0x1d8
c0000000049005b4 .kernel_init+0x27c/0x33c
c000000004031550 .kernel_thread+0x54/0x70

srp_task needs to be setup before request_irq. The patch below fixes the oops.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] zfcp: Update status read mempool

Commit 64deb6efdc5504ce97b5c1c6f281fffbc150bd93 changed the way status
read buffers are handled but forgot to adjust the mempool to the new
size. Add the call to resize the mempool after the exchange config
data. Also use the define instead of the hard coded number in the fsf
callback for consistency.

Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] zfcp: Do not wait for SBALs on stopped queue

Trying to read the FC host statistics on an offline adapter results in
a 5 seconds wait. Reading the statistics tries to issue an exchange
port data request which first waits up to 5 seconds for an entry in
the request queue.

Change the strategy for getting a free SBAL to exit when the queue is
stopped. Reading the statistics will then fail without the wait.

Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] zfcp: Fix check whether unchained ct_els is possible

A false check was performed whether an unchained ct_els
is possible or not.

Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

[SCSI] ipr: fix resource path display and formatting

It was possible to overflow the buffer used to print out the formatted
version of the resource path.  The fix is to limit the number of
bytes that get formatted.

This patch also updates the ipr_show_resource_path function to display the
resource address for devices that are attached to adapters that don't
support resource paths.

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>