scsi: cxlflash: Combine the send queue locks

Currently there are separate spin locks for the two supported I/O queueing
models. This makes it difficult to serialize with paths outside the enqueue
path.

As a design simplification and to support serialization with enqueue
operations, move to only a single lock that is used for enqueueing
regardless of the queueing model.

Signed-off-by: Uma Krishnan <ukrishn@linux.vnet.ibm.com>
Acked-by: Matthew R. Ochs <mrochs@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: sgiwd93: switch to dma_alloc_attrs

Use dma_alloc_attrs directly instead of the dma_alloc_noncoherent
wrapper.

[mkp: fixed driver name]

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: 53c700: switch to dma_alloc_attrs

Use dma_alloc_attrs directly instead of the dma_alloc_noncoherent
wrapper.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qedi: Remove comparison of u16 idx with zero.

Variable idx is defined as u16 thus statement (idx < 0) is always false
and should be removed.

Signed-off-by: Christos Gkekas <chris.gekas@gmail.com>
Acked-by: Manish Rangankar <Manish.Rangankar@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hpsa: fix spelling mistake: "encrypytion" -> "encryption"

Trivial fix to spelling mistake in dev_info message

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: redefine hisi_sas_phy.phy_type as u32

Element phy_type is a bitmask and it only ever has 2 bits possibly set,
and it is overkill to define as a u64, so redefine as a u32.

This change resolves static code check complaint that "phy->phy_type &=
~PORT_TYPE_SAS;" would unintentionally clear the high 32 bits as well.

Structure hisi_sas_phy is also reordered to ensure packing efficiency.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Remove the definition of VLC_SA_RECEIVE_CREDENTIAL

The symbolic name VLC_SA_RECEIVE_CREDENTIAL is not used anywhere in the
kernel. Additionally, since SPC 5 the RECEIVE CREDENTIAL command is
obsolete. The VLC_SA_RECEIVE_CREDENTIAL definition is misleading since
it occurs outside the list of other variable length CDB service action
codes (READ_32, WRITE_32, ...). Hence remove this definition.

References: commit e9ccc998b70f ("[SCSI] Add missing SPC-4 CDB and
MAINTENANCE_[IN,OUT] service action definitions")
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: aacraid: Don't copy uninitialized stack memory to userspace

Both aac_send_raw_srb() and aac_get_hba_info() may copy stack allocated
structs to userspace without initializing all members of these
structs. Clear out this memory to prevent information leaks.

Fixes: 423400e64d377 ("scsi: aacraid: Include HBA direct interface")
Fixes: c799d519bf088 ("scsi: aacraid: Retrieve HBA host information ioctl")
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: bnx2i: missing error code in bnx2i_ep_connect()

If bnx2i_map_ep_dbell_regs() then we accidentally return NULL instead of
an error pointer.  It results in a NULL dereference in
iscsi_if_ep_connect().

Fixes: cf4e6363859d ("[SCSI] bnx2i: Add bnx2i iSCSI driver.")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hptiop: make function hptiop_iop_request_callback_itl static

The function hptiop_iop_request_callback_itl does not need to be in
global scope, so make it static.

Cleans up sparse warning:
"symbol 'hptiop_iop_request_callback_itl' was not declared. Should it
 be static?"

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: cxgb4i: assign rxqs in round robin mode

Assign rxq to TCP connections in round robin mode to use all available
rxqs.

Signed-off-by: Varun Prakash <varun@chelsio.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fix typos and grammar in comments of scsi_transport_fc.c

Signed-off-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com>
Reviewed-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: fix refcount error on node list

A change in remote port removal introduced a spurious put which can
cause a premature structure teardown. The affects were most notable when
the driver attempted to unload as a null pointer would be hit.

Fix by removing the unnecessary put.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix nvme io stoppage after link bounce

On link down, transport is calling driver to abort outstanding ios.
Driver erroneously rejects the abort if the port indicates it isn't
logged in - which will be the case after the link down. Thus, the io
can't clean up. This prevents reconnection at the transport level.

Fix by allowing abort to proceed.

Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: virtio_scsi: let host do exception handling

virtio_scsi tries to do exception handling after the default 30 seconds
timeout expires.  However, it's better to let the host control the
timeout, otherwise with a heavy I/O load it is likely that an abort will
also timeout.  This leads to fatal errors like filesystems going
offline.

Disable the 'sd' timeout and allow the host to do exception handling,
following the precedent of the storvsc driver.

Hannes has a proposal to introduce timeouts in virtio, but this provides
an immediate solution for stable kernels too.

[mkp: fixed typo]

Reported-by: Douglas Miller <dougmill@linux.vnet.ibm.com>
Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: Hannes Reinecke <hare@suse.de>
Cc: linux-scsi@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qla2xxx: don't include <generated/utsrelease.h>

There's no need to use the static UTS_RELEASE string, since
utsname()->release contains the same.

This avoids rebuilding this file for every change of the release string.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: fc: drop residual tsk_mgmt_response and it_nexus_response

After commit 556e26a70b64 ("scsi: remove tsk_mgmt_response and
it_nexus_response transport methods"), the target driver support was
removed totally. Drop the residual.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: csiostor: update module version

Add -ko to the module version similar to module version of other Chelsio
drivers.

Signed-off-by: Varun Prakash <varun@chelsio.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: sd: Use sysfs_match_string()

Avoid unnecessary snprintf() when formatting variables for display in
sysfs and switch to sysfs_match_string() for validating user input.

Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: scsi_dh_alua: remove synchronous STPG support

Since 9c58b395 ("scsi: scsi_devinfo: remove synchronous ALUA for NETAPP
devices") this code is unused.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: remove various unused blist flags

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: storvsc: use default I/O timeout handler for FC devices

FC disks issue I/O directly to the host storage port driver, this is
diffirent to VHD disks where I/O is virtualized and timeout is handled
by the host VSP (Virtualization Service Provider).

FC disks are usually setup in a multipath system, and they don't want to
reset timer on I/O timeout. Timeout is detected by multipath as a good
time to failover and recover.

Signed-off-by: Long Li <longli@microsoft.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: aacraid: fix leak of data from stack back to userspace

The fields sense_data_size and sense_data are unitialized garbage from
the stack and are being copied back to userspace.  Fix this leak of
stack information by ensuring they are zero'd.

Detected by CoverityScan, CID#1435473 ("Uninitialized scalar variable")

Fixes: 423400e64d377 ("scsi: aacraid: Include HBA direct interface")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Dave Carroll <david.carroll@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: update to revision to 11.4.0.1

Set lpfc driver revision to 11.4.0.1

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Driver responds LS_RJT to Beacon Off ELS - Linux

Beacon OFF from switch is rejected by driver.

Driver fails Beacon OFF if frequency is set to 0. As per fc-ls spec,
status, capability, frequency and duration fields are only applicable
for Beacon ON.

Remove frequency and type checks. Reject Beacon ON if duration is non
zero.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix crash in lpfc_sli_ringtxcmpl_put when nvmet gets an abort request.

When running nvme detach-ns /dev/nvme0n1 -n 1 command, the nvmet lpfc
driver crashes with this stack dump:

kernel BUG at /root/NVME/lpfc_8.4/lpfc_sli.c:1393!
invalid opcode: 0000 [#1] SMP
Workqueue: nvmet-fc-cpu0 nvmet_fc_do_work_on_cpu [nvmet_fc]
 lpfc_sli4_issue_wqe+0x357/0x440 [lpfc]
 lpfc_nvmet_xmt_fcp_abort+0x36b/0x5c0 [lpfc]
 nvmet_fc_abort_op+0x30/0x50 [nvmet_fc]
 nvmet_fc_do_work_on_cpu+0xd9/0x130 [nvmet_fc]
 process_one_work+0x14e/0x410
 worker_thread+0x116/0x490
 kthread+0xc7/0xe0
 ret_from_fork+0x3f/0x70

Crash is due to an uninitialized iocbq->vport pointer.

Explicitly set the iocbq->vport field to phba->pport in
lpfc_nvmet_sol_fcp_issue_abort as it does all abort iocbq initialization
in the routine.  Using phba->pport is ok because target does not support
NPIV instances.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix crash doing IO with resets

During every reset, IOCBs are allocated. So, at one point, number of
allocated IOCBs reaches maximum limit and lpfc_sli_next_iotag fails.

Allocate IOCBs only during initialization. Reuse them after every reset
instead of allocating new set of IOCBs.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix crash after firmware flash when IO is running.

OS crashes after the completion of firmware download.

Failure in posting SCSI SGL buffers because number of SGL buffers is
less than total count. Some of the pending IOs are not completed by
driver. SGL buffers for these IOs are not added back to the list.
Pending IOs are not completed because lpfc_wq_list list is initialized
before completion of pending IOs.

Postpone lpfc_wq_list reinitialization by moving
lpfc_sli4_queue_destroy() after lpfc_hba_down_post().

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix SLI3 drivers attempting NVME ELS commands.

In a server with an 8G adapter and a 32G adapter, running NVME and FCP,
the server would crash with the following stack.

RIP: 0010: ... lpfc_nvme_register_port+0x38/0x420 [lpfc]
 lpfc_nlp_state_cleanup+0x154/0x4f0 [lpfc]
 lpfc_nlp_set_state+0x9d/0x1a0 [lpfc]
 lpfc_cmpl_prli_prli_issue+0x35f/0x440 [lpfc]
 lpfc_disc_state_machine+0x78/0x1c0 [lpfc]
 lpfc_cmpl_els_prli+0x17c/0x1f0 [lpfc]
 lpfc_sli_sp_handle_rspiocb+0x39b/0x6b0 [lpfc]
 lpfc_sli_handle_slow_ring_event_s3+0x134/0x2d0 [lpfc]
 lpfc_work_done+0x8ac/0x13b0 [lpfc]
 lpfc_do_work+0xf1/0x1b0 [lpfc]

Crash, on the 8G adapter, is due to a vport which does not have a nvme
local port structure. It's not supposed to have one. NVME is not
supported on the 8G adapter, so the NVME PRLI, which started this flow
shouldn't have been sent in the first place.

Correct discovery engine to recognize when on an SLI3 rport, which
doesn't support SLI3, if the rport supports only NVME, don't send a NVME
PRLI. Instead, as no FC4 will be used, a LOGO is sent.  If rport is FCP
and NVME, only execute the SCSI PRLI.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Break up IO ctx list into a separate get and put list

Since unsol rcv ISR and command cmpl ISR both access/lock this list,
separate get/put lists will reduce contention.

Replaced
struct list_head lpfc_nvmet_ctx_list;
with
struct list_head lpfc_nvmet_ctx_get_list;
struct list_head lpfc_nvmet_ctx_put_list;
and all correpsonding locks and counters.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Reduce time spent in IRQ for received NVME commands

Removed unnecessary bzero of context area. Due to size of sg list, added
a substantial delay and played havoc on cpu caches.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Vport creation is failing with "Link Down" error

Vport creation fails for SLI-3 adapters.

Mailbox submission fails because mailbox interrupt is disabled. Mailbox
interrupt is disabled during port reset.

Do reset only for physical port.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix nvme_info sysfs output to be consistent

First line of nvme_info output is not consistent. There is an Extra
colon in the format.

First line of output will contain one of the following strings:
NVME Initiator Enabled
NVME Target Enabled
NVME Disabled

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix system panic when express lane enabled.

There is a null pointer dereference that can happen in the FOF interrupt
handler.

The driver was not setting up cq->assoc_qp_for sli4_hba->oas_cq.

Initialize cq->assoc_qp before accessing it.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: modify internal abort dev flow for v3 hw

There is a change for abort dev for v3 hw: add registers to configure
unaborted iptt for a device, and then inform this to logic.

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>

scsi: hisi_sas: add v3 code to fill some more hw function pointers

Add code to fill the interface of phy_hard_reset, phy_get_max_linkrate,
and phy enable/disable.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add get_wideport_bitmap_v3_hw()

Add code for interface get_wideport_bitmap.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 code to send internal abort command

Add code to prepare internal abort command.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 code for itct setup and free

Add code to itct setup and free for v3 hw.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 code to send ATA frame

Add code to prepare ATA frame for v3 hw

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 code to send SMP frame

Add code to prepare SMP frame.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 code to send SSP frame

Add code to prepare SSP frame and deliver it to hardware.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 cq interrupt handler

Add v3 cq interrupt handler slot_complete_v3_hw().

Note: The slot error handling needs to be further refined in the future
to examine all fields in the error record, and handle appropriately,
instead of current solution - just report SAS_OPEN_REJECT.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add phy up/down/bcast and channel ISR

Add code to initialise interrupts and add some interrupt handlers.

Also add function hisi_sas_v3_destroy_irqs() to clean-up irqs upon
module unloading.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 hw PHY init

Add code to configure PHYs for v3 hw.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add v3 hw init

Add code to initialise v3 hardware.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add initialisation for v3 pci-based controller

Add the code to initialise the controller which is based on pci device
in hisi_sas_v3_hw.c

The core controller routines are still in hisi_sas_main.c; some common
initialisation functions are also exported from hisi_sas_main.c

For pci-based controller, the device properties, like phy count and sas
address are read from the firmware, same as platform device-based
controller.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add skeleton v3 hw driver

Add skeleton driver for v3 hw in hisi_sas_v3_hw.c

File hisi_sas_v3_hw.c will serve 2 purposes:
- probing and initialisation of the controller based on pci device
- hw layer for v3-based controllers

The controller design is quite similar to v2 hw in hip07.

However key differences include:
-All v2 hw bugs are fixed (hopefully), so workarounds are not required
-support for device deregistration
-some interrupt modifications
-configurable max device support

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: create hisi_sas_get_fw_info()

Move the functionality to retrieve the fw info into a dedicated device
type-agnostic function, hisi_sas_get_fw_info().

The reasoning is that this function will be required for future
pci-based platforms.

Also add some debug logs for failure.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: add pci_dev in hisi_hba struct

Since hip08 SAS controller is based on pci device, add hisi_hba.pci_dev
for hip08 (will be v3), and also rename hisi_hba.pdev to .platform_dev
for clarity.

In addition, for common code which wants to reference the controller
device struct, add hisi_hba.dev, and change the common code to use it.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: relocate get_ncq_tag_v2_hw()

Relocate get_ncq_tag_v2_hw() to a common location, as future hw versions
will require it.  Also rename with "hisi_sas_" prefix for consistency.

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>

scsi: hisi_sas: relocate sata_done_v2_hw()

Relocate get_ata_protocol() to a common location, as future hw versions
will require it.  Also rename with "hisi_sas_" prefix for consistency.

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>

scsi: hisi_sas: relocate get_ata_protocol()

Relocate get_ata_protocol() to a common location, as future hw versions
will require it.  Also rename with "hisi_sas_" prefix for consistency.

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>

scsi: hisi_sas: optimise the usage of hisi_hba.lock

Currently hisi_hba.lock is locked to deliver and receive a command
to/from any hw queue. This causes much contention at high data-rates.

To boost performance, lock on a per queue basis for sending and
receiving commands to/from hw.

Certain critical regions still need to be locked in the delivery and
completion stages with hisi_hba.lock.

New element hisi_sas_device.dq is added to store the delivery queue for
a device, so it does not need to be needlessly re-calculated for every
task.

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>

scsi: hisi_sas: define hisi_sas_device.device_id as int

Currently hisi_sas_device.device_id is a u64. This can create a problem
in selecting the queue for a device, in that this code does a 64b
division on device id. For some 32b systems, 64b division is slow and
the lib reference must be explicitly included.

The device id does not need to be 64b in size, so, as a solution, just
make as an int.

Also, struct hisi_sas_device elements are re-ordered to improve packing
efficiency.

Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: hisi_sas: fix timeout check in hisi_sas_internal_task_abort()

We need to check for timeout before task status, or the task will be
mistook as completed internal abort command.  Also add protection for
sas_task.task_state_flags in hisi_sas_tmf_timedout().

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>

scsi: lpfc: update to revision to 11.4.0.0

Set lpfc driver revision to 11.4.0.0

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Add auto EQ delay logic

Administrator intervention is currently required to get good numbers
when switching from running latency tests to IOPS tests.

The configured interrupt coalescing values will greatly effect the
results of these tests.  Currently, the driver has a single coalescing
value set by values of the module attribute.  This patch changes the
driver to support auto-configuration of the coalescing value based on
the total number of outstanding IOs and average number of CQEs processed
per interrupt for an EQ.  Values are checked every 5 seconds.

The driver defaults to the automatic selection. Automatic selection can
be disabled by the new lpfc_auto_imax module_parameter.

Older hardware can only change interrupt coalescing by mailbox
command. Newer hardware supports change via a register. The patch
support both.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix defects reported by Coverity Scan

Addressed the following reported defects:

** CID 1411552:  Control flow issues  (MISSING_BREAK)
/drivers/scsi/lpfc/lpfc_sli.c: 13259 in lpfc_sli4_nvmet_handle_rcqe()

** CID 1411553:  Memory - illegal accesses  (OVERRUN)
/drivers/scsi/lpfc/lpfc_sli.c: 16218 in lpfc_fc_frame_check()

** CID 1411553:  Memory - illegal accesses  (OVERRUN)
   Overrunning array "lpfc_rctl_names" of 202 8-byte elements at element
   index 244 (byte offset 1952) using index "fc_hdr->fh_r_ctl" (which
   evaluates to 244).

** CID 1411554:  Null pointer dereferences  (REVERSE_INULL)
/drivers/scsi/lpfc/lpfc_nvmet.c: 2131 in lpfc_nvmet_unsol_fcp_abort_cmp()

** CID 1411555:  Memory - illegal accesses  (UNINIT)
/drivers/scsi/lpfc/lpfc_nvmet.c: 180 in lpfc_nvmet_ctxbuf_post()

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix vports not logging into target

vports cannot login to target.

For vports, lpfc_nodelist is allocated for targets only on completion of
GFF_ID command. Driver checks if lpfc_nodelist exists for target before
sending GFF_ID. So, GFF_ID and PLOGI are not sent.

As mentioned by the comment in lpfc_prep_node_fc4type() routine, do not
send GFF_ID only if this NPortID is previously identified as FCP
target. Send GFF_ID if it is a newly identified remote port from GID_FT
response.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix PRLI retry handling when target rejects it.

The nvmet driver was rejecting the initiator's PRLI because its reg_rpi
for the PLOGI was still outstanding.  The initiator would resend the
PRLI without delay and get the same answer.  The PRLI retries would
exhaust causing the nvme initiator to set the nvmet ndlp to UNMAPPED.

The driver's lpfc_els_retry handler did not have a policy for an LS_RJT
with explanation CMD_IN_PROGRESS for PRLI or NVME_PRLI.  This caused the
delay to remain at 0 but retry set 1.

Fix: When the ELS response is LS_RJT, TPC and the command was PRLI or
NVME_PRLI, just set the delay to 1000 mS to get a 1 second delay on the
PRLI retry.  This was enough to allow the REG_RPI to complete at the
target.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Null pointer dereference when log_verbose is set to 0xffffffff

Kernel panic when log_verbose is set to 0xffffffff

phba->pport is dereferenced before it is initialized

Fix: Do not dereference phba->pport if it is NULL

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix System panic after loading the driver

System panic with general protection fault during driver load

The driver uses a static array sli4_hba.handler_name to store the irq
handler names. If the io_channel_irqs exceeds the pre-allocated size
(32+1), then the driver will overwrite other fields of sli4_hba.

Fix: Dynamically allocate handler_name.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix crash on powering off BFS VM with passthrough device

Null pointer dereference when BFS VM is powered off

The driver incorrectly uses sli3_ring on SLI-4 adapters

Use the correct ring structure based on sli_rev

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Tested-by: Raphael Silva <raphasil@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix return value of board_mode store routine in case of online failure

On hbacmd reset failure, observing wrong string "nline" in kernel log.

On failure, non negative value (1) is returned from sysfs store
routine. It is interpreted as count by kernel and store routine is
called again with the remaining characters as input.

Fix: Return negative error code (-EIO) in case of failure.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix counters so outstandng NVME IO count is accurate

NVME FC counters don't reflect actual results

Since counters are not atomic, or protected by a lock, the values often
get screwed up.

Make them atomic, like NVMET.  Fix up sysfs and debugfs display
accordingly Added Outstanding IOs to stats display

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix Port going offline after multiple resets.

Observing lpfc port down after issuing hbacmd reset command

Failure in posting SGL buffers. If there is only one SGL buffer and rrq
is valid for its XRI, we are rightly returning NULL but not adding the
buffer back to the SGL list. So, number of buffers become less than
total count and repost fails during reset.

Add SGL buffer back to list before returning NULL.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix nvmet node ref count handling

When unloading the driver, the NVMET driver would wait the full 30
seconds for its UNMAPPED initiator node to get removed before continuing
with the unload process.  NVMEI worked correctly.

For each rport put into UNMAPPED or MAPPED state by NVMET, the driver
puts a reference on the NDLP.  The difference is that NVMEI has a
unregister call for its rports and the extra reference is removed in the
unregister process.  For NVMET, the driver has to remove the reference
explicitly when dropping out of UNMAPPED or MAPPED because there is no
unregister call.

Add a call to lpfc_nlp_put on the ndlp when NVMET and the old state was
UNMAPPED or MAPPED.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix Lun Priority level shown as NA

Lun Priority level shown as NA

Remote port is not getting registered for nameserver and fdmi.  Due to
which dfc SendCTPassThru cmd is failing.

Made changes to register the remote port for both.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Add changes to assist in NVMET debugging

Inconsistent error messages and context state checks

Context state sanity checks were not accurate or inconsistent in the
code paths.

Separated LS context states from FCP.
Added and modified context state sanity checks.
Use context state to determine if a sol or unsol ABORT is needed.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix nvme port role handling in sysfs and debugfs handlers.

While debugging Devloss and recovery, debugfs and sysfs were found to
not show the NVME port roles consistently.

The port role FC_PORT_ROLE_NVME_DISCOVERY was added with the devloss
bringup and the other issues were just oversight.

Add NVME Target and DISCSRVC to debugfs nodeinfo and sysfs nvme info
handlers. The full port role was added to the NVME data only not the
generic nodelist.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Fix transition nvme-i rport handling to nport only.

As the devloss API was implemented in the nvmei driver, an evaluation of
the nvme transport and the lpfc driver showed dual management of the
rports.  This creates a bug possibility when the thread count and SAN
size increases.

The nvmei driver code was based on a very early transport and was not
revisited until the devloss API was introduced.

Remove the listhead in the driver's rport data structure and the
listhead in the driver's lport data structure.  Remove all rport_list
traversal.  Convert the driver to use the nrport (nvme rport) pointer
that is now NULL or nonNULL depending on a devloss action.  Convert
debugfs and nvme_info in sysfs to use the fc_nodes list in the vport.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: Add nvme initiator devloss support

Add nvme initiator devloss support

The existing implementation was based on no devloss behavior in the
transport (e.g. immediate teardown) so code didn't properly handle
delayed nvme rport device unregister calls.  In addition, the driver was
not correctly cycling the rport port role for each
register-unregister-reregister process.

This patch does the following:

Rework the code to properly handle rport device unregister calls and
potential re-allocation of the remoteport structure if the port comes
back in under dev_loss_tmo.

Correct code that was incorrectly cycling the rport port role for each
register-unregister-reregister process.

Prep the code to enable calling the nvme_fc transport api to dynamically
update dev_loss_tmo when the scsi sysfs interface changes it.

Memset the rpinfo structure in the registration call to enforce "accept
nvme transport defaults" in the registration call.  Driver parameters do
influence the dev_loss_tmo transport setting dynamically.

Simplifies the register function: the driver was incorrectly searching
its local rport list to determine resume or new semantics, which is not
valid as the transport already handles this.  The rport was resumed if
the rport handed back matches the ndlp->nrport pointer.  Otherwise,
devloss fired and the ndlp's nrport is NULL.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qedf: Merge a few quoted strings split across lines

Merge some quoted strings to improve readability and to save some lines
of code.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qedf: Use 'dma_zalloc_coherent' to reduce code verbosity.

Replace some 'dma_alloc_coherent+memset' by some quivalent
'dma_zalloc_coherent' in order to reduce code verbosity

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qedf: Fix a return value in case of error in 'qedf_alloc_global_queues'

We should return -ENOMEM in case of memory allocation error, as done
elsewhere in this function.

[mkp: fixed typo]

Fixes: 61d8658b4a435 ("scsi: qedf: Add QLogic FastLinQ offload FCoE driver framework.")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: esas2r: Replace semaphore fs_api_semaphore with mutex

The semaphore 'fs_api_semaphore' is used as a simple mutex, so it should
be written as one. Semaphores are going away in the future.

Signed-off-by: Binoy Jayan <binoy.jayan@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: esas2r: Replace semaphore fm_api_semaphore with mutex

The semaphore 'fm_api_semaphore' is used as a simple mutex, so it should
be written as one. Semaphores are going away in the future.

Signed-off-by: Binoy Jayan <binoy.jayan@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qla2xxx: Fix compile warning

Fixes following 0-day kernel build warnings:

drivers/scsi/qla2xxx/qla_init.c:6407:50: warning: format '%lx' expects
argument of type 'long unsigned int', but argument 5 has type 'unsigned
int' [-Wformat=]
drivers/scsi/qla2xxx/qla_init.c:6709:50: warning: format '%lx'
expects argument of type 'long unsigned int', but argument 5 has
type 'unsigned int' [-Wformat=]

Fixes: b95b9452aacf ("scsi: qla2xxx: Fix crash due to mismatch mumber of Q-pair creation for Multi queue")
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qla2xxx: remove redundant null check on tgt

An earlier commit ed7fb808477b846bb2 ("scsi: qla2xxx: Remove redundant
wait when target is stopped.") removed a null check on ha->tgt.tgt_ops
and replaced it with a new check that null checked tgt, thus making the
subsequent null check on tgt totally redundant. Remove it.

Detected by CoverityScan, CID#1440452 ("Logically Dead Code")

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufshcd-pci: Add Intel CNL support

Add PCI id and variant ops for Intel CNL UFS host controller.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufshcd-pci: Fix PM config

Put PM functions under correct config options and use standard PM macros
to set callbacks.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: lpfc: make a couple of functions static

functions lpfc_nvmet_cleanup_io_context and lpfc_nvmet_setup_io_context
can be made static as they do not need to be in global scope.

Cleans up sparse warnings:
  "warning: symbol 'lpfc_nvmet_cleanup_io_context' was not declared.
   Should it be static?"
  "warning: symbol 'lpfc_nvmet_setup_io_context' was not declared.
   Should it be static?"

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: xen-scsifront: Remove code that zeroes driver-private command data

Since the SCSI core zeroes driver-private command data, remove
that code from the xen-scsifront driver.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: xen-devel@lists.xenproject.org
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: virtio_scsi: Remove code that zeroes driver-private command data

Since the SCSI core zeroes driver-private command data, remove
that code from the virtio driver.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: snic: Remove code that zeroes driver-private command data

Since the SCSI core zeroes driver-private command data, remove that code
from the snic driver.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Narsimhulu Musini <nmusini@cisco.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Sesidhar Baddela <sebaddel@cisco.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Make scsi_mq_prep_fn() call scsi_init_command()

This patch reduces code duplication. There are two functional changes in
this patch:

- It causes scsi_mq_prep_fn() to clear driver-private command data, just
  like the already upstream commit 1bad6c4a57ef ("scsi: zero per-cmd
  private driver data for each MQ I/O").

- The initialization of .prot_sdb is moved from scsi_mq_prep_fn() into
  scsi_init_request().

[mkp: applied by hand]

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Introduce scsi_mq_sgl_size()

This patch does not change any functionality but makes the next patch
easier to read.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Only add commands to the device command list if required by the LLD

Just like for the scsi-mq code path, in the single queue SCSI code path
only add commands to the per-device command list if required by the SCSI
LLD. This patch will make it easier to merge the single-queue and
multiqueue command initialization code.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Make __scsi_remove_device go straight from BLOCKED to DEL

If a device is blocked, make __scsi_remove_device() cause it to
transition to the DEL state. This means that all the commands issued in
.shutdown() will error in the mid-layer, thus making the removal proceed
without being stopped.

This patch is a slightly modified version of a patch from James
Bottomley. This patch avoids that the following lockup occurs:

Call Trace:
 schedule+0x35/0x80
 schedule_timeout+0x237/0x2d0
 io_schedule_timeout+0xa6/0x110
 wait_for_completion_io+0xa3/0x110
 blk_execute_rq+0xdf/0x120
 scsi_execute+0xce/0x150 [scsi_mod]
 scsi_execute_req_flags+0x8f/0xf0 [scsi_mod]
 sd_sync_cache+0xa9/0x190 [sd_mod]
 sd_shutdown+0x6a/0x100 [sd_mod]
 sd_remove+0x64/0xc0 [sd_mod]
 __device_release_driver+0x8d/0x120
 device_release_driver+0x1e/0x30
 bus_remove_device+0xf9/0x170
 device_del+0x127/0x240
 __scsi_remove_device+0xc1/0xd0 [scsi_mod]
 scsi_forget_host+0x57/0x60 [scsi_mod]
 scsi_remove_host+0x72/0x110 [scsi_mod]
 srp_remove_work+0x8b/0x200 [ib_srp]

Reported-by: Israel Rukshin <israelr@mellanox.com>
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Israel Rukshin <israelr@mellanox.com>
Cc: Max Gurtovoy <maxg@mellanox.com>
Cc: Benjamin Block <bblock@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Introduce scsi_start_queue()

This patch does not change any functionality.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Israel Rukshin <israelr@mellanox.com>
Cc: Max Gurtovoy <maxg@mellanox.com>
Cc: Benjamin Block <bblock@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Protect SCSI device state changes with a mutex

Serializing SCSI device state changes avoids that two state changes can
occur concurrently, e.g. the state changes in scsi_target_block() and
__scsi_remove_device(). This serialization is essential to make patch
"Make __scsi_remove_device go straight from BLOCKED to DEL" work
reliably.

Enable this mechanism for all scsi_target_*block() callers but not for
the scsi_internal_device_unblock() calls from the mpt3sas driver because
that driver can call scsi_internal_device_unblock() from atomic context.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Create two versions of scsi_internal_device_unblock()

This will make it easier to serialize SCSI device state changes through
a mutex.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Sreekanth Reddy <sreekanth.reddy@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Split scsi_internal_device_block()

Instead of passing a "wait" argument to scsi_internal_device_block(),
split this function into a function that waits and a function that
doesn't wait. This will make it easier to serialize SCSI device state
changes through a mutex.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Sreekanth Reddy <sreekanth.reddy@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: Avoid that scsi_exit_rq() triggers a use-after-free

Dereferencing shost from scsi_exit_rq() is not safe because the SCSI
host may already have been freed when scsi_exit_rq() is called.
Increasing the shost reference count in scsi_init_rq() and dropping that
reference in scsi_exit_rq() is nontrivial since scsi_host_dev_release()
may sleep and since scsi_exit_rq() may be called from interrupt
context. Since scsi_exit_rq() only needs a single bit from shost, copy
that bit into struct scsi_cmnd.

Reported-by: Scott Bauer <scott.bauer@intel.com>
Fixes: e9c787e65c0c ("scsi: allocate scsi_cmnd structures as part of struct request")
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Scott Bauer <scott.bauer@intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: megaraid: Fix a sleep-in-atomic bug

The driver may sleep under a spin lock, and the function call path is:
mraid_mm_attach_buf (acquire the lock by spin_lock_irqsave)
  pci_pool_alloc(GFP_KERNEL) --> may sleep

To fix it, the "GFP_KERNEL" is replaced with "GFP_ATOMIC".

[mkp: fixed whitespace]

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Acked-by: Sumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: ufs: Tidy clocks list head usage

Move the initialization of clocks list head to ufshcd_alloc_host() so
that every driver doesn't have to do it. Remove checks for the list head
being NULL because that is not possible.

Signed-off-by: Szymon Mielczarek <szymonx.mielczarek@intel.com>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: atari_scsi: Use m68k_realnum_memory for FastRAM test

m68k_num_memory is unsuitable to test for the presence of FastRAM on
CT60 if the kernel is located in FastRAM: in arch/m68k/mm/motorola.c the
ST-RAM chunk is skipped and m68k_num_memory is decremented in this
case. m68k_realnum_memory still contains the actual number of RAM chunks
so use that.

Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Tested-by: Christian T. Steigies <cts@debian.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qla2xxx: remove writeq/readq function definitions

Instead of rewriting write/readq, use linux/io-64-nonatomic-lo-hi.h
which already have them.

Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Reviewed-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

scsi: qedf: Check if sense buffer has been allocated during completion

sc_cmd->sense_buffer is not guaranteed to be allocated so we need to
sc_cmd->check
if the pointer is NULL before trying to copy anything into it.

Fixes the crash:

[  143.793176] [0000:00:00.0]:[qedf_eh_device_reset:626]: LUN RESET Issued...
[  143.802996] BUG: unable to handle kernel NULL pointer dereference at           (null)
[  143.803063] IP: qedf_parse_fcp_rsp+0xe2/0x290 [qedf]
[  143.803077] PGD 0
[  143.803078] P4D 0

[  143.803103] Oops: 0002 [#1] SMP
[  143.803115] Modules linked in: msr(E) ebtable_filter(E) ebtables(E) ip6table_filter(E) ip6_tables(E) iptable_filter(E) ip_tables(E) x_tables(E) raw(E) scsi_transport_iscsi(E) br_netfilter(E) bridge(E) iscsi_ibft(E) iscsi_boot_sysfs(E) intel_rapl(E) sb_edac(E) x86_pkg_temp_thermal(E) intel_powerclamp(E) coretemp(E) kvm_intel(E) kvm(E) irqbypass(E) crct10dif_pclmul(E) crc32_pclmul(E) xfs(E) ghash_clmulni_intel(E) pcbc(E) aesni_intel(E) aes_x86_64(E) crypto_simd(E) ipmi_ssif(E) glue_helper(E) iTCO_wdt(E) iTCO_vendor_support(E) lpc_ich(E) ipmi_si(E) pcspkr(E) hpilo(E) ioatdma(E) cryptd(E) ipmi_devintf(E) hpwdt(E) mfd_core(E) shpchp(E) dca(E) thermal(E) pcc_cpufreq(E) ipmi_msghandler(E) acpi_cpufreq(E) af_packet(E) btrfs(E) xor(E) raid6_pq(E) sr_mod(E) cdrom(E) ata_generic(E) sd_mod(E) 8021q(E) garp(E)
[  143.803302]  stp(E) llc(E) mrp(E) bnx2fc(E) cnic(E) uio(E) mgag200(E) ata_piix(E) i2c_algo_bit(E) drm_kms_helper(E) syscopyarea(E) sysfillrect(E) sysimgblt(E) ahci(E) fb_sys_fops(E) bnx2x(E) qedf(E) serio_raw(E) libahci(E) ttm(E) uhci_hcd(E) ehci_pci(E) qed(E) mdio(E) libcrc32c(E) ehci_hcd(E) crc32c_intel(E) drm(E) libata(E) usbcore(E) tg3(E) ptp(E) hpsa(E) pps_core(E) scsi_transport_sas(E) libphy(E) wmi(E) button(E) fcoe(E) libfcoe(E) libfc(E) scsi_transport_fc(E) sg(E) dm_multipath(E) dm_mod(E) scsi_dh_rdac(E) scsi_dh_emc(E) scsi_dh_alua(E) scsi_mod(E) autofs4(E)
[  143.803438] CPU: 31 PID: 494 Comm: kworker/31:2 Tainted: G            E   4.12.0-rc1-69-default+ #1
[  143.803461] Hardware name: HP ProLiant DL380p Gen8, BIOS P70 08/20/2012
[  143.803480] Workqueue: qedf_io_wq qedf_fp_io_handler [qedf]
[  143.803496] task: ffff8804181a0000 task.stack: ffffc90003b64000
[  143.803514] RIP: 0010:qedf_parse_fcp_rsp+0xe2/0x290 [qedf]
[  143.803529] RSP: 0018:ffffc90003b67dc8 EFLAGS: 00010246
[  143.803544] RAX: 0000000000000000 RBX: ffff880401abdd48 RCX: 000000000000000c
[  143.803563] RDX: 0000000000000060 RSI: ffffffffa039c740 RDI: 0000000000000000
[  143.803581] RBP: ffffc90003b67df0 R08: ffffffffa039dba8 R09: 0000000000000000
[  143.803600] R10: 0000000000000000 R11: 0000000000000018 R12: 0000000000000000
[  143.803619] R13: ffff88040ac80bc8 R14: 0000000000000008 R15: ffff880407c14008
[  143.803638] FS:  0000000000000000(0000) GS:ffff88043f7c0000(0000) knlGS:0000000000000000
[  143.804360] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  143.805065] CR2: 0000000000000000 CR3: 0000000001c09000 CR4: 00000000000406e0
[  143.805753] Call Trace:
[  143.806436]  qedf_process_tmf_compl+0x19/0x30 [qedf]
[  143.807124]  qedf_process_cqe+0x265/0x280 [qedf]
[  143.807800]  qedf_fp_io_handler+0x26/0x60 [qedf]
[  143.808469]  process_one_work+0x138/0x370
[  143.809133]  worker_thread+0x4d/0x3b0
[  143.809797]  kthread+0x109/0x140
[  143.810451]  ? rescuer_thread+0x320/0x320
[  143.811100]  ? kthread_park+0x60/0x60
[  143.811743]  ret_from_fork+0x2c/0x40

Signed-off-by: Chad Dupuis <chad.dupuis@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>