platform/kernel/linux-rpi.git
5 years agoscsi: megaraid_sas: Add support for MegaRAID Aero controllers
Shivasharan S [Fri, 9 Nov 2018 17:47:20 +0000 (09:47 -0800)]
scsi: megaraid_sas: Add support for MegaRAID Aero controllers

This patch adds support for MegaRAID Aero controller PCI IDs.  Print a
message when a configurable secure type controller is encountered.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: mpt3sas: Display message on Configurable secure HBA
Sreekanth Reddy [Fri, 9 Nov 2018 09:34:12 +0000 (15:04 +0530)]
scsi: mpt3sas: Display message on Configurable secure HBA

Display following warning message only upon detection of configurable
secure type controllers.

"HBA is in Configurable Secure mode"

[mkp: typos]

Signed-off-by: Sreekanth Reddy <sreekanth.reddy@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libfc: Remove set but not used variable 'disc'
Yue Haibing [Thu, 8 Nov 2018 06:31:21 +0000 (06:31 +0000)]
scsi: libfc: Remove set but not used variable 'disc'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/scsi/libfc/fc_rport.c: In function 'fc_rport_recv_flogi_req':
drivers/scsi/libfc/fc_rport.c:866:18: warning:
 variable 'disc' set but not used [-Wunused-but-set-variable]

It no used any more after
commit baa6719f902a ("libfc: Update rport reference counting")

Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
Reviewed-by: Johannes Thumshirn <jth@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: target: add emulate_pr backstore attr to toggle PR support
David Disseldorp [Wed, 7 Nov 2018 13:11:07 +0000 (14:11 +0100)]
scsi: target: add emulate_pr backstore attr to toggle PR support

The new emulate_pr backstore attribute allows for Persistent Reservation
and SCSI2 RESERVE/RELEASE support to be completely disabled. This can be
useful for scenarios such as:

- Ensuring ATS (Compare & Write) usage on recent VMware ESXi initiators.

- Allowing clustered (e.g. tcm-user) backends to block such requests,
  avoiding the multi-node reservation state propagation.

When explicitly disabled, PR and RESERVE/RELEASE requests receive Invalid
Command Operation Code response sense data.

Signed-off-by: David Disseldorp <ddiss@suse.de>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: cxgb4i: fix thermal configuration dependencies
Arnd Bergmann [Fri, 2 Nov 2018 15:35:23 +0000 (16:35 +0100)]
scsi: cxgb4i: fix thermal configuration dependencies

I fixed a bug by adding a dependency in the network driver, but that fix
caused a related bug in the SCSI driver:

WARNING: unmet direct dependencies detected for CHELSIO_T4
  Depends on [m]: NETDEVICES [=y] && ETHERNET [=y] && NET_VENDOR_CHELSIO [=y] && PCI [=y] && (IPV6 [=y] || IPV6 [=y]=n) && (THERMAL [=m] || !THERMAL [=m])
  Selected by [y]:
  - SCSI_CXGB4_ISCSI [=y] && SCSI_LOWLEVEL [=y] && SCSI [=y] && PCI [=y] && INET [=y] && (IPV6 [=y] || IPV6 [=y]=n)
drivers/net/ethernet/chelsio/cxgb4/cxgb4_thermal.o: In function `cxgb4_thermal_init':
cxgb4_thermal.c:(.text+0x158): undefined reference to `thermal_zone_device_register'
drivers/net/ethernet/chelsio/cxgb4/cxgb4_thermal.o: In function `cxgb4_thermal_remove':
cxgb4_thermal.c:(.text+0x1d8): undefined reference to `thermal_zone_device_unregister'
/git/arm-soc/Makefile:1042: recipe for target 'vmlinux' failed

The same dependency needs to be propagated here to make it work correctly
with CONFIG_THERMAL=m and SCSI_CXGB4_ISCSI=y. That change by itself causes
another problem with a circular dependency, as we use 'select NETDEVICES'.
This is something we really should not do anyway, as a driver symbol should
never select another major subsystem, so let's turn that into a 'depends
on'. I don't see any downsides of that, as NETDEVICES is only disabled in
rather obscure cases that are not relevant to the users of cxgb4i.

Fixes: e70a57fa59bb ("cxgb4: fix thermal configuration dependencies")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: aic94xx: Use dma_pool_zalloc
Souptick Joarder [Thu, 1 Nov 2018 04:19:09 +0000 (09:49 +0530)]
scsi: aic94xx: Use dma_pool_zalloc

Replaced dma_pool_alloc + memset with dma_pool_zalloc.

Signed-off-by: Brajeswar Ghosh <brajeswar.linux@gmail.com>
Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop old default switch case which might paper over missing case
Steffen Maier [Thu, 8 Nov 2018 14:44:58 +0000 (15:44 +0100)]
scsi: zfcp: drop old default switch case which might paper over missing case

This was introduced with v2.6.27 commit 287ac01acf22 ("[SCSI] zfcp: Cleanup
code in zfcp_erp.c") but would now suppress helpful -Wswitch compiler
warnings when building with W=1 such as the following forced example:

drivers/s390/scsi/zfcp_erp.c: In function 'zfcp_erp_setup_act':
drivers/s390/scsi/zfcp_erp.c:220:2: warning: enumeration value 'ZFCP_ERP_ACTION_REOPEN_PORT' not handled in switch [-Wswitch]
  switch (need) {
  ^~~~~~

But then again, only with W=1 we would notice unhandled enum cases.
Without the default cases and a missed unhandled enum case, the code might
perform unforeseen things we might not want...

As of today, we never run through the removed default case, so removing it
is no functional change.  In the future, we never should run through a
default case but introduce the necessary specific case(s) to handle new
functionality.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop default switch case which might paper over missing case
Steffen Maier [Thu, 8 Nov 2018 14:44:57 +0000 (15:44 +0100)]
scsi: zfcp: drop default switch case which might paper over missing case

This was introduced with v4.18 commit 8c3d20aada70 ("scsi: zfcp: fix
missing REC trigger trace for all objects in ERP_FAILED") but would now
suppress helpful -Wswitch compiler warnings when building with W=1 such as
the following forced example:

drivers/s390/scsi/zfcp_erp.c: In function 'zfcp_erp_handle_failed':
drivers/s390/scsi/zfcp_erp.c:126:2: warning: enumeration value 'ZFCP_ERP_ACTION_REOPEN_PORT_FORCED' not handled in switch [-Wswitch]
  switch (want) {
  ^~~~~~

But then again, only with W=1 we would notice unhandled enum cases.
Without the default cases and a missed unhandled enum case, the code might
perform unforeseen things we might not want...

As of today, we never run through the removed default case, so removing it
is no functional change.  In the future, we never should run through a
default case but introduce the necessary specific case(s) to handle new
functionality.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: silence -Wimplicit-fallthrough in zfcp_erp_lun_strategy()
Steffen Maier [Thu, 8 Nov 2018 14:44:56 +0000 (15:44 +0100)]
scsi: zfcp: silence -Wimplicit-fallthrough in zfcp_erp_lun_strategy()

For some reason the already existing substring "fall through" in the
comment is not sufficient for GCC to silence -Wimplicit-fallthrough.

  CC [M]  drivers/s390/scsi/zfcp_erp.o
drivers/s390/scsi/zfcp_erp.c: In function 'zfcp_erp_lun_strategy':
drivers/s390/scsi/zfcp_erp.c:1065:6: warning: this statement may fall through [-Wimplicit-fallthrough=]
  if (atomic_read(&zfcp_sdev->status) & ZFCP_STATUS_COMMON_OPEN)
      ^
drivers/s390/scsi/zfcp_erp.c:1068:2: note: here
  case ZFCP_ERP_STEP_LUN_CLOSING:
  ^~~~

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: silence remaining kdoc warnings in header files
Steffen Maier [Thu, 8 Nov 2018 14:44:55 +0000 (15:44 +0100)]
scsi: zfcp: silence remaining kdoc warnings in header files

Improve whatever the following simple invocation reported:
$ ./scripts/kernel-doc -none drivers/s390/scsi/*.h

While at it, improve some related kdoc,
including struct zfcp_fsf_ct_els in zfcp_fsf.h.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: silence all W=1 build warnings for existing kdoc
Steffen Maier [Thu, 8 Nov 2018 14:44:54 +0000 (15:44 +0100)]
scsi: zfcp: silence all W=1 build warnings for existing kdoc

While at it also improve some copy & paste kdoc mistakes.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: properly format LUN (and WWPN) for LUN sharing violation kmsg
Steffen Maier [Thu, 8 Nov 2018 14:44:53 +0000 (15:44 +0100)]
scsi: zfcp: properly format LUN (and WWPN) for LUN sharing violation kmsg

zfcp: <devbusid>: LUN 0x0 on port 0x5005076......... ...
zfcp: <devbusid>: LUN 0x1000000000000 on port 0x5005076......... ...

should be

zfcp: <devbusid>: LUN 0x0000000000000000 on port 0x5005076......... ...
zfcp: <devbusid>: LUN 0x0001000000000000 on port 0x5005076.........
                  is already in use by CSS., MIF Image ID .

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: use enum zfcp_erp_act_result for argument/return of affected functions
Steffen Maier [Thu, 8 Nov 2018 14:44:52 +0000 (15:44 +0100)]
scsi: zfcp: use enum zfcp_erp_act_result for argument/return of affected functions

With that instead of just "int" it becomes clear which functions return
this type and which ones also accept it as argument they just pass through
in some cases or modify in other cases.  v2.6.27 commit 287ac01acf22
("[SCSI] zfcp: Cleanup code in zfcp_erp.c") introduced the enum which was
cpp defines previously.

Silence some false -Wswitch compiler warning cases with individual
NOP cases. When adding more enum values and building with W=1 we
would get compiler warnings about missed new cases.

Consistently use the variable name "result", so change "retval" in
zfcp_erp_strategy() to "result". This avoids confusion with other compile
unit variables "retval" having different semantics and type.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: use enum zfcp_erp_steps for struct zfcp_erp_action.step
Steffen Maier [Thu, 8 Nov 2018 14:44:51 +0000 (15:44 +0100)]
scsi: zfcp: use enum zfcp_erp_steps for struct zfcp_erp_action.step

Use the already defined enum for this purpose to get at least some build
checking (even though an enum is type equivalent to an int in C).  v2.6.27
commit 287ac01acf22 ("[SCSI] zfcp: Cleanup code in zfcp_erp.c") introduced
the enum which was cpp defines previously.

Since struct zfcp_erp_action type is embedded into other structures living
in zfcp_def.h, we have to move enum zfcp_erp_act_type from its private
definition in zfcp_erp.c to the zfcp-global zfcp_def.h

Silence some false -Wswitch compiler warning cases with individual NOP
cases. When adding more enum values and building with W=1 we would get
compiler warnings about missed new cases.

Add missing break statements in some of the above switch cases.  No
functional change, but making it future-proof.  I think all of these should
have had a break statement ever since, even if these switch cases happened
to be the last ones in the switch statement body.

"Fall through" in the context of switch case usually means not to have a
break and fall through to the subsequent switch case. However, I think this
old comment meant that here we do not have an _early return_ in the switch
case but the code path continues after the switch case body.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: the action field of zfcp_erp_action is actually the type
Steffen Maier [Thu, 8 Nov 2018 14:44:50 +0000 (15:44 +0100)]
scsi: zfcp: the action field of zfcp_erp_action is actually the type

&zfcp_erp_action.action ==> &zfcp_erp_action.type

While at it, make use of the already defined enum for this purpose to get
at least some build checking (even though an enum is type equivalent to an
int in C). v2.6.27 commit 287ac01acf22 ("[SCSI] zfcp: Cleanup code in
zfcp_erp.c") introduced the enum which was cpp defines previously.

To prevent compiler warnings with the switch(act->type), we have to
separate the recently added eyecatchers from enum zfcp_erp_act_type.

Since struct zfcp_erp_action type is embedded into other structures living
in zfcp_def.h, we have to move enum zfcp_erp_act_type from its private
definition in zfcp_erp.c to the zfcp-global zfcp_def.h.

Silence one false -Wswitch compiler warning case: LUNs as the leaves in our
object tree do not have any follow-up success recovery.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: clarify function argument name for trace tag string
Steffen Maier [Thu, 8 Nov 2018 14:44:49 +0000 (15:44 +0100)]
scsi: zfcp: clarify function argument name for trace tag string

v2.6.30 commit 5ffd51a5e495 ("[SCSI] zfcp: replace current ERP logging with
a more convenient version") changed trace record distinguishing from a
numerical ID to a 7 character string called "trace tag". While starting to
use function arguments with different type and semantics, it did not change
the argument name accordingly.

v2.6.38 commit ae0904f60fab ("[SCSI] zfcp: Redesign of the debug tracing
for recovery actions.") renamed variable names "id" into "tag" but only
within zfcp_dbf.*, not within zfcp_erp.c.

This was a bit confusing since the remainder of zfcp does use the term
"trace tag". Also "id" is quite generic and it's not obvious for what.
Just unify it consistently and use the "dbf" prefix to relate the arguments
to the code in zfcp_dbf.*.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: ERP thread setup kdoc update
Steffen Maier [Thu, 8 Nov 2018 14:44:48 +0000 (15:44 +0100)]
scsi: zfcp: ERP thread setup kdoc update

zfcp_erp_thread_setup() update complements v2.6.32 commit 347c6a965dc1
("[SCSI] zfcp: Use kthread API for zfcp erp thread").

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: update kernel message for invalid FCP_CMND length, it's not the CDB
Steffen Maier [Thu, 8 Nov 2018 14:44:47 +0000 (15:44 +0100)]
scsi: zfcp: update kernel message for invalid FCP_CMND length, it's not the CDB

The CDB is just a part inside of FCP_CMND, see zfcp_fc_scsi_to_fcp().
While at it, fix the device driver reaction: adapter not LUN shutdown.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop duplicate seq_no from zfcp_fsf_req which is also in QTCB header
Steffen Maier [Thu, 8 Nov 2018 14:44:46 +0000 (15:44 +0100)]
scsi: zfcp: drop duplicate seq_no from zfcp_fsf_req which is also in QTCB header

There is no point for double bookkeeping especially just for tracing.  The
trace can take it from the QTCB which always exists for non-SRB responses
traced with zfcp_dbf_hba_fsf_res().

As a side effect, this removes an alignment hole and reduces the size of
struct zfcp_fsf_req, and thus of each pending request, by 8 bytes.

Before:
$ pahole -C zfcp_fsf_req drivers/s390/scsi/zfcp.ko
...
struct fsf_qtcb *          qtcb;                 /*   144     8 */
u32                        seq_no;               /*   152     4 */
/* XXX 4 bytes hole, try to pack */
void *                     data;                 /*   160     8 */
...
/* size: 296, cachelines: 2, members: 14 */
/* sum members: 288, holes: 2, sum holes: 8 */
/* last cacheline: 40 bytes */
After:
$ pahole -C zfcp_fsf_req drivers/s390/scsi/zfcp.ko
...
struct fsf_qtcb *          qtcb;                 /*   144     8 */
void *                     data;                 /*   152     8 */
...
/* size: 288, cachelines: 2, members: 13 */
        /* sum members: 284, holes: 1, sum holes: 4 */

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop duplicate fsf_command from zfcp_fsf_req which is also in QTCB header
Steffen Maier [Thu, 8 Nov 2018 14:44:45 +0000 (15:44 +0100)]
scsi: zfcp: drop duplicate fsf_command from zfcp_fsf_req which is also in QTCB header

Status read buffers (SRBs, unsolicited notifications) never use a QTCB
[zfcp_fsf_req_create()]. zfcp_fsf_req_send() already uses this to
distinguish SRBs from other FSF request types. We can re-use this method in
zfcp_fsf_req_complete(). Introduce a helper function to make the check for
req->qtcb less magic.

SRBs always are FSF_QTCB_UNSOLICITED_STATUS, so we can hard-code this for
the two trace functions dealing with SRBs.

All other FSF request types have a QTCB and we can get the fsf_command from
there.

zfcp_dbf_hba_fsf_response() and thus zfcp_dbf_hba_fsf_res() are only called
for non-SRB requests so it's safe to dereference the QTCB
[zfcp_fsf_req_complete() returns early on SRB, else calls
zfcp_fsf_protstatus_eval() which calls zfcp_dbf_hba_fsf_response()].  In
zfcp_scsi_forget_cmnd() we guard the QTCB dereference with a preceding NULL
check and rely on boolean shortcut evaluation.

As a side effect, this causes an alignment hole which we can close in
a later patch after having cleaned up all fields of struct zfcp_fsf_req.
Before:
$ pahole -C zfcp_fsf_req drivers/s390/scsi/zfcp.ko
...
u32                        status;               /*   136     4 */
u32                        fsf_command;          /*   140     4 */
struct fsf_qtcb *          qtcb;                 /*   144     8 */
...
After:
$ pahole -C zfcp_fsf_req drivers/s390/scsi/zfcp.ko
...
u32                        status;               /*   136     4 */
/* XXX 4 bytes hole, try to pack */
struct fsf_qtcb *          qtcb;                 /*   144     8 */
...

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop unnecessary forward prototype for struct zfcp_fsf_req
Steffen Maier [Thu, 8 Nov 2018 14:44:44 +0000 (15:44 +0100)]
scsi: zfcp: drop unnecessary forward prototype for struct zfcp_fsf_req

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: group sort internal structure definitions for proximity
Steffen Maier [Thu, 8 Nov 2018 14:44:43 +0000 (15:44 +0100)]
scsi: zfcp: group sort internal structure definitions for proximity

Have structures just before the structures that use them (without
disrupting sequences of using structures such as zfcp_unit and
zfcp_scsi_dev):

 - zfcp_adapter_mempool embedded in zfcp_adapter,

 - zfcp_latenc... embedded in zfcp_scsi_dev.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: namespace prefix for internal latency data structures
Steffen Maier [Thu, 8 Nov 2018 14:44:42 +0000 (15:44 +0100)]
scsi: zfcp: namespace prefix for internal latency data structures

In contrast to struct fsf_qual_latency_info, the ones here are not FSF but
software defined zfcp-internal.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: update width in comment for ZFCP_COMMON_FLAGS mask
Steffen Maier [Thu, 8 Nov 2018 14:44:41 +0000 (15:44 +0100)]
scsi: zfcp: update width in comment for ZFCP_COMMON_FLAGS mask

v2.6.10 history commit 4062e12b2ba2 ("[PATCH] s390: zfcp act enhancements")
extended this mask by one nibble with the introduction of
ZFCP_STATUS_COMMON_ACCESS_DENIED == 0x00800000 for ACT (access control
table).

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: move scsi_eh & non-ERP timeout defines owned by and local to zfcp_fsf.c
Steffen Maier [Thu, 8 Nov 2018 14:44:40 +0000 (15:44 +0100)]
scsi: zfcp: move scsi_eh & non-ERP timeout defines owned by and local to zfcp_fsf.c

Also clarify namespace prefix for the timeout used for FSF requests on
behalf of SCSI error recovery: It is zfcp_fsf_ not zfcp_scsi_.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: drop unnecessary forward prototype for struct zfcp_reqlist
Steffen Maier [Thu, 8 Nov 2018 14:44:39 +0000 (15:44 +0100)]
scsi: zfcp: drop unnecessary forward prototype for struct zfcp_reqlist

While struct zfcp_adapter contains a pointer to zfcp_reqlist, the pointer
field does not need to know the structure or even a prototype.

The prototype was introduced with v2.6.34 commit b6bd2fb92a7b ("[SCSI]
zfcp: Move FSF request tracking code to new file").

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: move SG table helper from aux to fc and make them static
Steffen Maier [Thu, 8 Nov 2018 14:44:38 +0000 (15:44 +0100)]
scsi: zfcp: move SG table helper from aux to fc and make them static

Since commit 663e0890e31c ("[SCSI] zfcp: remove access control tables
interface") these helper functions are only used for auto port scan in
zfcp_fc.c. Also change them to the corresponding namespace prefix.

This is a small cleanup for the miscellaneous catchall compile unit
zfcp_aux.c.

Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: zfcp: remove unnecessary null pointer check before mempool_destroy
zhong jiang [Thu, 8 Nov 2018 14:44:37 +0000 (15:44 +0100)]
scsi: zfcp: remove unnecessary null pointer check before mempool_destroy

mempool_destroy has taken null pointer check into account. so remove the
redundant check.

Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Acked-by: Benjamin Block <bblock@linux.ibm.com>
[maier@linux.ibm.com: depends on v4.3 4e3ca3e033d1 ("mm/mempool: allow NULL `pool' pointer in mempool_destroy()")]
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libsas: Remove pcidev reference
John Garry [Thu, 15 Nov 2018 10:20:32 +0000 (18:20 +0800)]
scsi: libsas: Remove pcidev reference

Not all host drivers are PCI drivers - like hisi_sas, which supports a
platform driver - so remove reference to "pcidev".

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libsas: Drop SAS_DPRINTK() and revise logs levels
John Garry [Thu, 15 Nov 2018 10:20:31 +0000 (18:20 +0800)]
scsi: libsas: Drop SAS_DPRINTK() and revise logs levels

Like sas_printk() did previously, SAS_DPRINTK() offers little value now
that libsas logs already have the "sas" prefix through pr_fmt(fmt). So it
can be dropped.

However, after reviewing some logs in libsas, it is noticed that debug
level is too low in many instances.

So this change drops SAS_DPRINTK() and revises some logs to a more
appropriate level. However many stay at debug level, although some
are significantly promoted.

We add -DDEBUG for compilation so that we keep the debug messages by
default, as before.

All the pre-existing checkpatch errors for spanning messages across
multiple lines are also fixed.

Finally, all other references to printk() [apart from special formatting
in sas_ata.c] are removed and replaced with appropriate pr_xxx().

Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libsas: Drop sas_printk()
John Garry [Thu, 15 Nov 2018 10:20:30 +0000 (18:20 +0800)]
scsi: libsas: Drop sas_printk()

The printk wrapper sas_printk() adds little value now that libsas logs
already have the "sas" prefix through pr_fmt(fmt), so just use pr_notice()
directly.

In addition, strings which span multiple lines are reunited.

Originally-from: Joe Perches <joe@perches.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libsas: Use pr_fmt(fmt)
John Garry [Thu, 15 Nov 2018 10:20:29 +0000 (18:20 +0800)]
scsi: libsas: Use pr_fmt(fmt)

In preparation for dropping the libsas printk wrappers, use pr_fmt(fmt)
declaration to add the framework log prefix - "sas".

Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: libsas: Delete sas_dump.{c, h}
John Garry [Thu, 15 Nov 2018 10:20:28 +0000 (18:20 +0800)]
scsi: libsas: Delete sas_dump.{c, h}

The code in these files is not longer referenced, so delete them.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: change the time of SAS SSP connection
Xiang Chen [Fri, 9 Nov 2018 14:06:36 +0000 (22:06 +0800)]
scsi: hisi_sas: change the time of SAS SSP connection

Currently the time of SAS SSP connection is 1ms, which means the link
connection will fail if no IO response after this period.

For some disks handling large IO (such as 512k), 1ms is not enough, so
change it to 5ms.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: Relocate some codes to avoid an unused check
Xiang Chen [Fri, 9 Nov 2018 14:06:35 +0000 (22:06 +0800)]
scsi: hisi_sas: Relocate some codes to avoid an unused check

In function hisi_sas_task_prep(), we check asd_sas_port, but in function
hisi_sas_task_exec(), we already refer to asd_sas_port by using function
dev_to_hisi_hba() implicitly. So to avoid this possible invalid
dereference, relocate the check to function hisi_sas_task_prep().

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: Add support for interrupt coalescing for v3 hw
Xiang Chen [Fri, 9 Nov 2018 14:06:34 +0000 (22:06 +0800)]
scsi: hisi_sas: Add support for interrupt coalescing for v3 hw

If INT_COAL_EN is enabled, configure time and count of interrupt
coalescing.  Then if CQ collects count of CQ entries in time, it will
report the interrupt. Or if CQ doesn't collect enough CQ entries in time,
it will report the interrupt at timeout.

As all the registers are not supported to be changed dynamically, we need
to config those register between disable and enable PHYs.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: Add support for interrupt converge for v3 hw
Xiang Chen [Fri, 9 Nov 2018 14:06:33 +0000 (22:06 +0800)]
scsi: hisi_sas: Add support for interrupt converge for v3 hw

If CQ_INT_CONVERGE_EN is enabled, the interrupts of all the 16 CQ queues
will be reported by CQ0.

So we need to change the process of CQ tasklet for this situation.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: Create separate host attributes per HBA
Xiang Chen [Fri, 9 Nov 2018 14:06:32 +0000 (22:06 +0800)]
scsi: hisi_sas: Create separate host attributes per HBA

Currently all the three HBA (v1/v2/v3 HW) share the same host attributes.

To support each HBA having separate attributes in future, create per-HBA
attributes.

Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: wd719x: always use generic DMA API
Christoph Hellwig [Fri, 9 Nov 2018 15:54:06 +0000 (16:54 +0100)]
scsi: wd719x: always use generic DMA API

The wd719x driver currently uses a mix of the legacy PCI DMA and the
generic DMA APIs.  Switch it over to the generic DMA API entirely.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: wd719x: use per-command private data
Christoph Hellwig [Fri, 9 Nov 2018 15:54:05 +0000 (16:54 +0100)]
scsi: wd719x: use per-command private data

Add the SCB onto the scsi command allocation and use dma streaming mappings
for it only when in use.  This avoid possibly calling dma_alloc_coherent
under a lock or even in irq context, while also making the code simpler.

Thanks to Ondrej Zary for testing and various bug fixes.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: wd719x: there should be no active SCBs on removal
Christoph Hellwig [Fri, 9 Nov 2018 15:54:04 +0000 (16:54 +0100)]
scsi: wd719x: there should be no active SCBs on removal

Warn on that case instead of trying to free them which would be fatal in
case we actually had active ones.

[mkp: typos]

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: sym53c8xx: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:24 +0000 (15:10 +0200)]
scsi: sym53c8xx: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: stex: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:23 +0000 (15:10 +0200)]
scsi: stex: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Switch it over to the better generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: mvumi: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:22 +0000 (15:10 +0200)]
scsi: mvumi: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: lpfc: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:21 +0000 (15:10 +0200)]
scsi: lpfc: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Switch it over to the better generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: isci: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:20 +0000 (15:10 +0200)]
scsi: isci: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Switch it over to the better generic DMA API.

[mkp: s/iscsi/isci/]

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: initio: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:19 +0000 (15:10 +0200)]
scsi: initio: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hptiop: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:18 +0000 (15:10 +0200)]
scsi: hptiop: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: hisi_sas: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:17 +0000 (15:10 +0200)]
scsi: hisi_sas: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Switch it over to the better generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: esas2r: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:16 +0000 (15:10 +0200)]
scsi: esas2r: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Also move the dma_get_required_mask check before actually
setting the dma mask so that we don't end up with inconsistent settings in
corner cases.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: dpt_i2o: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:15 +0000 (15:10 +0200)]
scsi: dpt_i2o: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Also move the dma_get_required_mask check before actually
setting the dma mask, so that we don't end up with inconsistent settings in
corner cases.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: bfa: use dma_set_mask_and_coherent
Christoph Hellwig [Thu, 18 Oct 2018 13:10:14 +0000 (15:10 +0200)]
scsi: bfa: use dma_set_mask_and_coherent

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.  Switch it over to the better generic DMA API helper and
also ensure we set the coherent mask as well in the resume path.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: arcmsr: use dma_set_mask
Christoph Hellwig [Thu, 18 Oct 2018 13:10:13 +0000 (15:10 +0200)]
scsi: arcmsr: use dma_set_mask

The driver currently uses pci_set_dma_mask despite otherwise using the
generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: ufs: Add UFS platform driver for Cadence UFS
Janek Kotas [Thu, 20 Sep 2018 13:08:30 +0000 (13:08 +0000)]
scsi: ufs: Add UFS platform driver for Cadence UFS

This patch adds a device tree platform driver for Cadence UFS Host
Controller.  It can be enabled with SCSI_UFS_CDNS_PLATFORM Kconfig option.

Signed-off-by: Jan Kotas <jank@cadence.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: dt-bindings: ufs: Add bindings for Cadence UFS
Janek Kotas [Thu, 20 Sep 2018 13:08:22 +0000 (13:08 +0000)]
scsi: dt-bindings: ufs: Add bindings for Cadence UFS

This patch adds a DT binding documentation for Cadence UFS Host Controller.

Signed-off-by: Jan Kotas <jank@cadence.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: ufs: set the device reference clock setting
Subhash Jadavani [Tue, 16 Oct 2018 08:59:41 +0000 (14:29 +0530)]
scsi: ufs: set the device reference clock setting

UFS host supplies the reference clock to UFS device and UFS device
specification allows host to provide one of the 4 frequencies (19.2 MHz, 26
MHz, 38.4 MHz, 52 MHz) for reference clock. Host should set the device
reference clock frequency setting in the device based on what frequency it
is supplying to UFS device.

Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Sayali Lokhande <sayalil@codeaurora.org>
Reviewed-by: Evan Green <evgreen@chromium.org>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: pmcraid: use generic DMA API
Christoph Hellwig [Thu, 18 Oct 2018 13:05:37 +0000 (15:05 +0200)]
scsi: pmcraid: use generic DMA API

Switch from the legacy PCI DMA API to the generic DMA API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: pmcraid: don't allocate a dma coherent buffer for sense data
Christoph Hellwig [Thu, 18 Oct 2018 13:05:36 +0000 (15:05 +0200)]
scsi: pmcraid: don't allocate a dma coherent buffer for sense data

We can just dma map the sense buffer passed with the scsi command,
and that gets us out of the nasty business of doing dma coherent
allocations from irq context.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: pmcraid: simplify pmcraid_cancel_all a bit
Christoph Hellwig [Thu, 18 Oct 2018 13:05:35 +0000 (15:05 +0200)]
scsi: pmcraid: simplify pmcraid_cancel_all a bit

No need for a local cmd_done variable, and pass boolean values as bool
type instead of u32.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: qla2xxx: use lower_32_bits and upper_32_bits instead of reinventing them
Christoph Hellwig [Thu, 18 Oct 2018 13:03:37 +0000 (15:03 +0200)]
scsi: qla2xxx: use lower_32_bits and upper_32_bits instead of reinventing them

This also moves the optimization for builds with 32-bit dma_addr_t to
the compiler (where it belongs) instead of opencoding it based on
incorrect assumptions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: qla1280: properly handle 64-bit DMA
Christoph Hellwig [Thu, 18 Oct 2018 13:03:36 +0000 (15:03 +0200)]
scsi: qla1280: properly handle 64-bit DMA

CONFIG_HIGHMEM is not in fact an indicator for > 32-bit dma addressing
Given that the driver is a bit weird and wants a compile time selection
switch to checking CONFIG_ARCH_DMA_ADDR_T_64BIT instead.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: qla1280: use lower_32_bits and upper_32_bits instead of reinventing them
Christoph Hellwig [Thu, 18 Oct 2018 13:03:35 +0000 (15:03 +0200)]
scsi: qla1280: use lower_32_bits and upper_32_bits instead of reinventing them

This also moves the optimization for builds with 32-bit dma_addr_t to
the compiler (where it belongs) instead of opencoding it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: ips: properly handle 64-bit DMA
Christoph Hellwig [Thu, 18 Oct 2018 13:03:34 +0000 (15:03 +0200)]
scsi: ips: properly handle 64-bit DMA

CONFIG_HIGHMEM64 is only one (and these days unusual) way to indicate
that > 32-bit dma address are possible.  Replace it with a check of the
dma_addr_t size.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: ips: use lower_32_bits and upper_32_bits instead of reinventing them
Christoph Hellwig [Thu, 18 Oct 2018 13:03:33 +0000 (15:03 +0200)]
scsi: ips: use lower_32_bits and upper_32_bits instead of reinventing them

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: qla4xxx: remvoe unnecessary condition check for mempool_destroy()
Chengguang Xu [Sun, 4 Nov 2018 13:57:50 +0000 (21:57 +0800)]
scsi: qla4xxx: remvoe unnecessary condition check for mempool_destroy()

mempool_destroy() can handle NULL pointer correctly, so there is no
need to check NULL pointer before calling mempool_destroy().

Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Acked-by: Manish Rangankar <Manish.Rangankar@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: qla4xxx: remove unnecessary condition check for dma_pool_destroy()
Chengguang Xu [Sun, 4 Nov 2018 13:57:49 +0000 (21:57 +0800)]
scsi: qla4xxx: remove unnecessary condition check for dma_pool_destroy()

dma_pool_destroy() can handle NULL pointer correctly, so there is no
need to check NULL pointer before calling dma_pool_destroy().

Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Acked-by: Manish Rangankar <Manish.Rangankar@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: update driver version to 12.0.0.8
James Smart [Tue, 23 Oct 2018 20:41:12 +0000 (13:41 -0700)]
scsi: lpfc: update driver version to 12.0.0.8

Update the driver version to 12.0.0.8

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: add Trunking support
James Smart [Tue, 23 Oct 2018 20:41:11 +0000 (13:41 -0700)]
scsi: lpfc: add Trunking support

Add trunking support to the driver. Trunking is found on more recent
asics. In general, trunking appears as a single "port" to the driver
and overall behavior doesn't differ. Link speed is reported as an
aggregate value, while link speed control is done on a per-physical
link basis with all links in the trunk symmetrical. Some commands
returning port information are updated to additionally provide
trunking information. And new ACQEs are generated to report physical
link events relative to the trunk.

This patch contains the following modifications:

- Added link speed settings of 128GB and 256GB.

- Added handling of trunk-related ACQEs, mainly logging and trapping
  of physical link statuses.

- Added additional bsg interface to query trunk state by applications.

- Augment link_state sysfs attribtute to display trunk link status

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Implement GID_PT on Nameserver query to support faster failover
James Smart [Tue, 23 Oct 2018 20:41:10 +0000 (13:41 -0700)]
scsi: lpfc: Implement GID_PT on Nameserver query to support faster failover

The switches seem to respond faster to GID_PT vs GID_FT NameServer
queries.  Add support for GID_PT to be used over GID_FT to enable
faster storage failover detection. Includes addition of new module
parameter to select between GID_PT and GID_FT (GID_FT is default).

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Correct loss of fc4 type on remote port address change
James Smart [Tue, 23 Oct 2018 20:41:09 +0000 (13:41 -0700)]
scsi: lpfc: Correct loss of fc4 type on remote port address change

An address change for a remote port cause PRLI for the wrong protocol
to be sent.  The node copy done in the discovery code skipped copying
the fc4 protocols supported as well.

Fix the copy logic for the address change.  Beefed up log messages in
this area as well.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Fix odd recovery in duplicate FLOGIs in point-to-point
James Smart [Tue, 23 Oct 2018 20:41:08 +0000 (13:41 -0700)]
scsi: lpfc: Fix odd recovery in duplicate FLOGIs in point-to-point

Testing a point-to-point topology and a case of re-FLOGI without
intervening link bouncing, showed an odd interaction with firmware and
a resulting scenario where the driver no longer probed after accepting
the new FLOGI.

Work around the firmware issue by issuing a link bounce if a FLOGI is
received after the link is already up and FLOGI's accepted.

While debugging the issue, realized that some debug traces should be
clarified to help in the future.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Correct LCB RJT handling
James Smart [Tue, 23 Oct 2018 20:41:07 +0000 (13:41 -0700)]
scsi: lpfc: Correct LCB RJT handling

When LCB's are rejected, if beaconing was already in progress, the
Reason Code Explanation was not being set. Should have been set to
command in progress.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: fcoe: Fix link down issue after 1000+ link bounces
James Smart [Tue, 23 Oct 2018 20:41:06 +0000 (13:41 -0700)]
scsi: lpfc: fcoe: Fix link down issue after 1000+ link bounces

On FCoE adapters, when running link bounce test in a loop, initiator
failed to login with switch switch and required driver reload to
recover. Switch reached a point where all subsequent FLOGIs would be
LS_RJT'd. Further testing showed the condition to be related to not
performing FCF discovery between FLOGI's.

Fix by monitoring FLOGI failures and once a repeated error is seen
repeat FCF discovery.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Correct errors accessing fw log
James Smart [Tue, 23 Oct 2018 20:41:05 +0000 (13:41 -0700)]
scsi: lpfc: Correct errors accessing fw log

This patch corrects two issues:

- An oops would occur if reading based on a non-zero offset.  Offset
  calculation was incorrect.

- Updates to ras config (logging level) were ignored if change was
  made while fw logging was enabled. Revise to dynamically update.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Reset link or adapter instead of doing infinite nameserver PLOGI retry
James Smart [Tue, 23 Oct 2018 20:41:04 +0000 (13:41 -0700)]
scsi: lpfc: Reset link or adapter instead of doing infinite nameserver PLOGI retry

Currently, PLOGI failures are infinitely delayed/retried.  There have
been some fabric situations where the PLOGI's were to the nameserver
and it stopped responding. The retries would never clear up.  A better
resolution in this situation is to retry a couple of times, then drop
the link and reinit. This brings back connectivity to the nameserver.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Fix LOGO/PLOGI handling when triggerd by ABTS Timeout event
James Smart [Tue, 23 Oct 2018 20:41:03 +0000 (13:41 -0700)]
scsi: lpfc: Fix LOGO/PLOGI handling when triggerd by ABTS Timeout event

After a LOGO in response to an ABTS timeout, a PLOGI wasn't issued to
re-establish the login.  An nlp_type check in the LOGO completion
handler failed to restart discovery for NVME targets.  Revised the
nlp_type check for NVME as well as SCSI.

While reviewing the LOGO handling a few other issues were seen and
were addressed:

- Better lock synchronization around ndlp data types

- When the ABTS times out, unregister the RPI before sending the LOGO
  so that all local exchange contexts are cleared and nothing received
  while awaiting LOGO/PLOGI handling will be accepted.

- LOGO handling optimized to:
   Wait only R_A_TOV for a response.
   It doesn't need to be retried on timeout. If there wasn't a
     response, a PLOGI will be sent, thus an implicit logout
     applies as well when the other port sees it.
   If there is a response, any kind of response is considered "good"
     and the XRI quarantined for a exchange qualifier window.

- PLOGI is issued as soon a LOGO state is resolved.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Fix lpfc_sli4_read_config return value check
James Smart [Tue, 23 Oct 2018 20:41:02 +0000 (13:41 -0700)]
scsi: lpfc: Fix lpfc_sli4_read_config return value check

An error is an error - but not to the existing return value check.

Revise check to handle any failure, not just EIO.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: lpfc: Correct speeds on SFP swap
James Smart [Tue, 23 Oct 2018 20:41:01 +0000 (13:41 -0700)]
scsi: lpfc: Correct speeds on SFP swap

Supported speeds is not updated when SFP is removed or replaced

Supported speed is obtained from lmt field in READ_CONFIG mailbox
response. Driver updates supported speeds only once from PCI probe
path. After that it is never updated. So, supported speeds remains the
same till reboot or driver reload.

When SFP is removed or inserted, driver gets SLI-Port Event ACQE.  If
SFP is removed, lmt wil have value 0. If a different SFP is inserted,
lmt will have value according to its supported speeds.  So, afterr
SLI-Port Event ACQE handling path, send READ_CONFIG mailbox and update
supported speeds. If READ_CONFIG fails, set supported speeds to
unknown and log.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Update driver version
Shivasharan S [Wed, 17 Oct 2018 06:37:57 +0000 (23:37 -0700)]
scsi: megaraid_sas: Update driver version

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: modify max supported lds related print
Shivasharan S [Wed, 17 Oct 2018 06:37:56 +0000 (23:37 -0700)]
scsi: megaraid_sas: modify max supported lds related print

The print related to number of VDs that a particular firmware supports is
misleading in some cases.  Even though supportMaxExtLDs is set, certain
firmware profiles only supports upto 64VDs.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: remove unused macro
Shivasharan S [Wed, 17 Oct 2018 06:37:55 +0000 (23:37 -0700)]
scsi: megaraid_sas: remove unused macro

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: increase timeout for IOC INIT to 180seconds
Shivasharan S [Wed, 17 Oct 2018 06:37:54 +0000 (23:37 -0700)]
scsi: megaraid_sas: increase timeout for IOC INIT to 180seconds

IOC INIT frame needs to be fired with a timeout of 180 seconds.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Remove double endian conversion
Shivasharan S [Wed, 17 Oct 2018 06:37:53 +0000 (23:37 -0700)]
scsi: megaraid_sas: Remove double endian conversion

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Re-use max_mfi_cmds to calculate queue sizes
Shivasharan S [Wed, 17 Oct 2018 06:37:52 +0000 (23:37 -0700)]
scsi: megaraid_sas: Re-use max_mfi_cmds to calculate queue sizes

In megasas_init_adapter_fusion(), max_mfi_cmds is being calcuated as
(MEGASAS_FUSION_INTERNAL_CMDS + MEGASAS_FUSION_IOCTL_CMDS).  max_mfi_cmds
can be used in megasas_configure_queue_sizes.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Rename scratch_pad registers
Shivasharan S [Wed, 17 Oct 2018 06:37:51 +0000 (23:37 -0700)]
scsi: megaraid_sas: Rename scratch_pad registers

Rename the scratch pad registers to match firmware headers.  No functional
change.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Remove spin lock for dpc operation
Shivasharan S [Wed, 17 Oct 2018 06:37:50 +0000 (23:37 -0700)]
scsi: megaraid_sas: Remove spin lock for dpc operation

Optimization: No need to hold hba_lock in dpc context for reading atomic
variable.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: optimize raid context access in IO path
Shivasharan S [Wed, 17 Oct 2018 06:37:49 +0000 (23:37 -0700)]
scsi: megaraid_sas: optimize raid context access in IO path

No functional change.  Use local variables when accessing raid context in
IO path.  Improves code readability.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Fail init if heartbeat timer fails
Shivasharan S [Wed, 17 Oct 2018 06:37:48 +0000 (23:37 -0700)]
scsi: megaraid_sas: Fail init if heartbeat timer fails

When driver fails to start the heartbeat timer, exit from FW init.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: For SRIOV, do not set STOP_ADP bit
Shivasharan S [Wed, 17 Oct 2018 06:37:47 +0000 (23:37 -0700)]
scsi: megaraid_sas: For SRIOV, do not set STOP_ADP bit

For SRIOV based adapters, driver should not set the STOP_ADP bit as part of
kill adapter.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Fix combined reply queue mode detection
Shivasharan S [Wed, 17 Oct 2018 06:37:46 +0000 (23:37 -0700)]
scsi: megaraid_sas: Fix combined reply queue mode detection

For Invader series, if FW supports more than 8 MSI-x vectors, driver needs
to enable combined reply queue mode. For Ventura series, driver enables
combined reply queue mode in case of more than 16 MSI-x vectors.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Fix module parameter description
Shivasharan S [Wed, 17 Oct 2018 06:37:45 +0000 (23:37 -0700)]
scsi: megaraid_sas: Fix module parameter description

Module parameter description for rdpq_enable incorrectly lists the default
as enabled.  Also, provide range of valid values for resetwaittime in the
description.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Fix goto labels in error handling
Shivasharan S [Wed, 17 Oct 2018 06:37:44 +0000 (23:37 -0700)]
scsi: megaraid_sas: Fix goto labels in error handling

During init, if pci_alloc_irq_vectors() fails, the driver has not yet setup
the IRQs. Fix the goto labels and error handling for this case.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Update copyright information
Shivasharan S [Wed, 17 Oct 2018 06:37:43 +0000 (23:37 -0700)]
scsi: megaraid_sas: Update copyright information

Change copyright to Broadcom Inc. Also update any references to Avago with
Broadcom. Update copyright duration wherever required.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Add check for reset adapter bit
Shivasharan S [Wed, 17 Oct 2018 06:37:42 +0000 (23:37 -0700)]
scsi: megaraid_sas: Add check for reset adapter bit

For SAS3 and later controllers, FW sets the reset adapter bit indicating
the driver to perform a controller reset.  Driver needs to check if this
bit is set before doing a reset.  This reduces the driver probe failure
time to 180seconds in case there is a faulty controller connected.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Fix msleep granularity
Shivasharan S [Wed, 17 Oct 2018 06:37:41 +0000 (23:37 -0700)]
scsi: megaraid_sas: Fix msleep granularity

In megasas_transition_to_ready() driver waits 180seconds for controller to
change FW state. Here we are calling msleep(1) in a loop for this.  As
explained in timers-howto.txt, msleep(1) will actually sleep longer than
1ms. If a faulty controller is connected, we will end up waiting for much
more than 180 seconds causing unnecessary delays during load.

Change the granularity of msleep() call from 1ms to 1000ms.

Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Add support for FW snap dump
Shivasharan S [Wed, 17 Oct 2018 06:37:40 +0000 (23:37 -0700)]
scsi: megaraid_sas: Add support for FW snap dump

Latest firmware adds a mechanism to save firmware logs just before
controller reset on pre-allocated internal controller DRAM. This feature is
called snapdump which will help debugging firmware issues.  This feature
requires extra time and firmware reports these values through new driver
interface. Before initiating an OCR, driver needs to inform FW to save a
snapdump and then wait for a specified time for the snapdump to complete.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: megaraid_sas: Add watchdog thread to detect Firmware fault
Shivasharan S [Wed, 17 Oct 2018 06:37:39 +0000 (23:37 -0700)]
scsi: megaraid_sas: Add watchdog thread to detect Firmware fault

Currently driver checks for Firmware state change from ISR context, and
only when there are interrupts tied with no I/O completions.  We have seen
multiple cases where doorbell interrupts sent by firmware to indicate FW
state change are not processed by driver and it takes long time for driver
to trigger OCR. And if there are no IOs running, since we only check the FW
state as part of ISR code, fault goes undetected by driver and OCR will not
be triggered.

This patch introduces a separate workqueue that runs every one second to
detect Firmware FAULT state and trigger reset immediately.  As an
additional gain, removing PCI reads from ISR to check FW state results in
improved performance as well.

Signed-off-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: mpt3sas: Bump driver version to 27.100.00.00
Suganath Prabu [Wed, 31 Oct 2018 13:23:39 +0000 (18:53 +0530)]
scsi: mpt3sas: Bump driver version to 27.100.00.00

Modify driver version to 27.100.00.00 (which is equivalent to PH8 OOB
driver)

Signed-off-by: Suganath Prabu <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: mpt3sas: Fix driver modifying persistent data in Manufacturing page11
Suganath Prabu [Wed, 31 Oct 2018 13:23:38 +0000 (18:53 +0530)]
scsi: mpt3sas: Fix driver modifying persistent data in Manufacturing page11

Currently driver is modifying both current & NVRAM/persistent data in
Manufacturing page11. Driver should change only current copy of
Manufacturing page11. It should not modify the persistent data.

So removed the section of code where driver is modifying the persistent
data of Manufacturing page11.

Signed-off-by: Suganath Prabu <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
6 years agoscsi: mpt3sas: Don't modify EEDPTagMode field setting on SAS3.5 HBA devices
Suganath Prabu [Wed, 31 Oct 2018 13:23:37 +0000 (18:53 +0530)]
scsi: mpt3sas: Don't modify EEDPTagMode field setting on SAS3.5 HBA devices

If EEDPTagMode field in manufacturing page11 is set then unset it. This is
needed to fix a hardware bug only in SAS3/SAS2 cards. So, skipping
EEDPTagMode changes in Manufacturing page11 for SAS 3.5 controllers.

Signed-off-by: Suganath Prabu <suganath-prabu.subramani@broadcom.com>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>