While the 'state' attribute can (and will) change occasionally,
calling 'poll()' or 'select()' on it fails as sysfs is never
notified that the state has changed.
With this patch calling 'poll()' or 'select()' will work
properly.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Rework scsi_internal_device_unblock_nowait() into using a switch
statement. No functional changes.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

pci_device_id are not supposed to change at runtime. All functions
working with pci_device_id provided by <linux/pci.h> work with const
pci_device_id. So mark the non-const structs as const.
Signed-off-by: NArvind Yadav <arvind.yadav.cs@gmail.com>
Acked-by: NBradley Grove <bgrove@attotech.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Since switching to blk-mq as the default in commit 5c279bd9
("scsi: default to scsi-mq"), virtio-scsi LUNs consume about 10x as
much kernel memory.

qemu currently allocates a fixed 128 entry virtqueue.  can_queue
currently is set to 1024.  But with indirect descriptors, each command
in the queue takes 1 virtqueue entry, so the number of commands which
can be queued is equal to the length of the virtqueue.

Note I intend to send a patch to qemu to allow the virtqueue size to be
configured from the qemu command line.
Signed-off-by: NRichard W.M. Jones <rjones@redhat.com>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If using indirect descriptors, you can make the total_sg as large as you
want.  If not, BUG is too serious because the function later returns
-ENOSPC.
Signed-off-by: NRichard W.M. Jones <rjones@redhat.com>
Reviewed-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

[qed_sp_iscsi_func_start:189(host_7-0)]Cannot satisfy CQ amount. Queues
requested 8, CQs available 4. Aborting function start

Above condition will resolve as management firmware is capable of
telling us the number of CQs available for a given PF, qed will
communicate the same number to qedi, So that qedi will know how much CQs
are allowed.
Signed-off-by: NManish Rangankar <manish.rangankar@cavium.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The driver version is not updated with changes to the driver, so it has
no value, so just get rid of it.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Instances of kfree(shost) should be replaced with scsi_host_put().

In addition, a missing scsi_host_put() is added for error path in
hisi_sas_shost_alloc_pci() and v3 driver removal.

Signed-off-by: Pan Bian <bianpan2016@163.com> # For main.c changes
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Variable res only holds value 0, so remove it.

This cleans up a coccicheck warning.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add function to set linkrate for v3 hw.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch updates some register setting according to recommendation
from HW designer and experiment.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Use ACPI "_RST" method to reset the controller, since FLR is not
supported.

Function hisi_sas_stop_phys() is introduced to remove some code
duplication.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch adds calls to kill CQ takslets v3 hw during probe failure.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The channel interrupt is to process all the interrupts except PHY
UP/DOWN and broadcast interrupt. So we need to clear all the interrupts
except those 3 interrupts after processing channel interrupts.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Modify STP link timer from 10ms to 500ms. Also add the register address.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

For v3 hw, internal abort function required status and command buffer to
be set, so add necessary code for this.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add two ATA commands, ATA_CMD_ZAC_MGMT_IN and ATA_CMD_ZAC_MGMT_OUT in
hisi_sas_get_ata_protocol(), to support SATA SMR disk.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch is a fix related to freeing a device in v2 hw driver.

Before, we polled to ITCT CLR interrupt to check if a device is free.

This was error prone, as if the interrupt doesn't occur in 10us, we miss
processing it.

To avoid this situation, service this interrupt and sync the event with
a completion.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch adds support to clean-up allocated IRQs and kill tasklets
when probe fails and for driver removal.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch removes some repeated configurations:

(1) The device id of the device is already set in the alloc function, so
    we don't need to modify in free device function.

(2) Field dev_type and dev_status are configured in hisi_sas_dev_gone(),
    so there is no need for repeated config in free_device_v3_hw.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The code to print ECC errors in v2 hw driver is very repetitive.  This
patch condensed the code by looping an array of errors.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NShiju Jose <shiju.jose@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add DFX feature for v2 hw. We are adding support for
the following errors:
- loss_of_dword_sync_count
- invalid_dword_count
- phy_reset_problem_count
- running_disparity_error_count
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The value dw0 is the residual bytes when UNDERFLOW error happens, but we
filled the residual with the value of dw3 before. So change the residual
from dw3 to dw0.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When some interrupts happen together, we need to process every interrupt
one-by-one, and should not return immediately when one interrupt process
is finished being processed.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch provides fixes for the following issues:

1. Fix issue of controller reset required to send commands. For reset
   process, it may be required to send commands to the controller, but
   not during soft reset.  So add HISI_SAS_NOT_ACCEPT_CMD_BIT to prevent
   executing a task during this period.

2. Send a broadcast event in rescan topology to detect any topology
   changes during reset.

3. Previously it was not ensured that libsas has processed the PHY up
   and down events after reset. Potentially this could cause an issue
   that we still process the PHY event after reset. So resolve this by
   flushing shot workqueue in LLDD reset.

4. Port ID requires refresh after reset. The port ID generated after
   reset is not guaranteed to be the same as before reset, so it needs
   to be refreshed for each device's ITCT.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Update the driver’s PCI IDs to match the latest Microsemi controllers
Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add PQI reset to driver shutdown callback to work around controller bug.

During an 1.) OS shutdown or 2.) kexec outside of a kdump, the Linux
kernel will clear BME on our controller.

If BME is cleared during a controller/host PCIe transfer, the controller
will lock up.

So we perform a PQI reset in the driver's shutdown callback function to
eliminate the possibility of a controller/host PCIe transfer being
active when the kernel clears BME immediately after calling the driver's
shutdown callback.
Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

 - make pass-thru requests bi-directional
Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

 - distinguish between shutdown and non-shutdown.
Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Reviewed-by: NScott Benesh <scott.benesh@microsemi.com>
Signed-off-by: NKevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: NDon Brace <don.brace@microsemi.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The null check on pHba->channel[chan].device is redundant because
device is an array and hence can never be null. Remove the check.

Detected by CoverityScan, CID#115362 ("Array compared against 0")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The return value of dma_map_single() should be checked by
dma_mapping_error(). However, in function qla26xx_dport_diagnostics(), its
return value is checked against NULL, which could result in failures.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Acked-by: NHimanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The return value of scsi_host_alloc() should be released by
scsi_host_put(). However, in function mvs_pci_init(), kfree()
is used. This patch replaces kfree() with scsi_host_put() to avoid
possible memory leaks.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Reviewed-by: NJack Wang <jinpu.wang@profitbricks.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

In function pm8001_pci_probe(), on errors that the control flow jumps to
label err_out_ha_free, function pm8001_free() is called. In pm8001_free(),
scsi_host_put() is called to release shost, which keeps the return value
of scsi_host_alloc(). After pm8001_free() returns, kfree() is called to
free shost again, resulting in a double free bug. This patch removes
scsi_host_put() from pm8001_free() and explicitly calls scsi_host_put()
to release Scsi_Host in need.
Signed-off-by: NPan Bian <bianpan2016@163.com>
Reviewed-by: NJack Wang <jinpu.wang@profitbricks.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

fix allocate instance->pd_info twice which was introduced by 96188a89.
Signed-off-by: Nweiping zhang <zhangweiping@didichuxing.com>
Acked-by: NSumit Saxena <sumit.saxena@broadcom.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

After sending a message, always clear esp->msg_out_len. Otherwise,
eh_abort_handler may subsequently fail to send an ABORT TASK SET
message.
Tested-by: NStan Johnson <userm57@yahoo.com>
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If an LLD aborts a task set, it should complete the affected commands
with the appropriate result code. In a couple of cases esp_scsi doesn't
do so.

When the initiator receives an unhandled message, just respond by sending
a MESSAGE REJECT instead of ABORT TASK SET, and thus avoid the issue.

OTOH, a MESSAGE REJECT sent by a target can be taken as an indication
that the initiator messed up somehow. It isn't always possible to abort
correctly, so just fall back on a SCSI bus reset, which will complete the
affected commands with the appropriate result code.

For example, certain Apple (Sony) CD-ROM drives, when the non-existent
LUN 1 is scanned, can't handle the INQUIRY command. The problem is not
detected until the initiator gets a MESSAGE REJECT. Whenever esp_scsi
sees that message, it raises ATN and sends ABORT TASK SET -- but
neglects to complete the failed scmd.

The target then goes into DATA OUT phase (probably bogus), while the ESP
device goes into disconnected mode (surprising, given the bus phase).
The next Transfer Information command from esp_scsi then causes
an Invalid Command interrupt because that command is not valid when in
disconnected mode:

mac_esp: using PDMA for controller 0
mac_esp mac_esp.0: esp0: regs[50f10000:(null)] irq[19]
mac_esp mac_esp.0: esp0: is a ESP236, 16 MHz (ccf=4), SCSI ID 7
scsi host0: esp
scsi 0:0:0:0: Direct-Access     SEAGATE  ST318416N        0010 PQ: 0 ANSI: 3
scsi target0:0:0: Beginning Domain Validation
scsi target0:0:0: asynchronous
scsi target0:0:0: Domain Validation skipping write tests
scsi target0:0:0: Ending Domain Validation
scsi 0:0:3:0: CD-ROM            SONY     CD-ROM CDU-8003A 1.9a PQ: 0 ANSI: 2 CCS
scsi target0:0:3: Beginning Domain Validation
scsi target0:0:3: FAST-5 SCSI 2.0 MB/s ST (500 ns, offset 15)
scsi target0:0:3: Domain Validation skipping write tests
scsi target0:0:3: Ending Domain Validation
scsi host0: unexpected IREG 40
scsi host0: Dumping command log
scsi host0: ent[2] CMD val[c2] sreg[90] seqreg[cc] sreg2[00] ireg[20] ss[01] event[0c]
scsi host0: ent[3] CMD val[00] sreg[91] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c]
scsi host0: ent[4] EVENT val[0d] sreg[91] seqreg[04] sreg2[00] ireg[18] ss[00] event[0c]
scsi host0: ent[5] EVENT val[03] sreg[91] seqreg[04] sreg2[00] ireg[18] ss[00] event[0d]
scsi host0: ent[6] CMD val[90] sreg[91] seqreg[04] sreg2[00] ireg[18] ss[00] event[03]
scsi host0: ent[7] EVENT val[05] sreg[91] seqreg[04] sreg2[00] ireg[18] ss[00] event[03]
scsi host0: ent[8] EVENT val[0d] sreg[93] seqreg[cc] sreg2[00] ireg[10] ss[00] event[05]
scsi host0: ent[9] CMD val[01] sreg[93] seqreg[cc] sreg2[00] ireg[10] ss[00] event[0d]
scsi host0: ent[10] CMD val[11] sreg[93] seqreg[cc] sreg2[00] ireg[10] ss[00] event[0d]
scsi host0: ent[11] EVENT val[0b] sreg[93] seqreg[cc] sreg2[00] ireg[10] ss[00] event[0d]
scsi host0: ent[12] CMD val[12] sreg[97] seqreg[cc] sreg2[00] ireg[08] ss[00] event[0b]
scsi host0: ent[13] EVENT val[0c] sreg[97] seqreg[cc] sreg2[00] ireg[08] ss[00] event[0b]
scsi host0: ent[14] CMD val[44] sreg[90] seqreg[cc] sreg2[00] ireg[20] ss[00] event[0c]
scsi host0: ent[15] CMD val[01] sreg[90] seqreg[cc] sreg2[00] ireg[20] ss[01] event[0c]
scsi host0: ent[16] CMD val[c2] sreg[90] seqreg[cc] sreg2[00] ireg[20] ss[01] event[0c]
scsi host0: ent[17] CMD val[00] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[0c]
scsi host0: ent[18] EVENT val[0d] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[0c]
scsi host0: ent[19] EVENT val[06] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[0d]
scsi host0: ent[20] CMD val[01] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[06]
scsi host0: ent[21] CMD val[10] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[06]
scsi host0: ent[22] CMD val[1a] sreg[87] seqreg[ca] sreg2[00] ireg[08] ss[00] event[06]
scsi host0: ent[23] CMD val[12] sreg[87] seqreg[ca] sreg2[00] ireg[08] ss[00] event[06]
scsi host0: ent[24] EVENT val[0d] sreg[87] seqreg[ca] sreg2[00] ireg[08] ss[00] event[06]
scsi host0: ent[25] EVENT val[09] sreg[86] seqreg[ca] sreg2[00] ireg[10] ss[00] event[0d]
scsi host0: ent[26] CMD val[01] sreg[86] seqreg[ca] sreg2[00] ireg[10] ss[00] event[09]
scsi host0: ent[27] CMD val[10] sreg[86] seqreg[ca] sreg2[00] ireg[10] ss[00] event[09]
scsi host0: ent[28] EVENT val[0a] sreg[86] seqreg[ca] sreg2[00] ireg[10] ss[00] event[09]
scsi host0: ent[29] EVENT val[0d] sreg[80] seqreg[ca] sreg2[00] ireg[20] ss[00] event[0a]
scsi host0: ent[30] EVENT val[04] sreg[80] seqreg[ca] sreg2[00] ireg[20] ss[00] event[0d]
scsi host0: ent[31] CMD val[01] sreg[80] seqreg[ca] sreg2[00] ireg[20] ss[00] event[04]
scsi host0: ent[0] CMD val[90] sreg[80] seqreg[ca] sreg2[00] ireg[20] ss[00] event[04]
scsi host0: ent[1] EVENT val[05] sreg[80] seqreg[ca] sreg2[00] ireg[20] ss[00] event[04]
scsi target0:0:3: FAST-5 SCSI 2.0 MB/s ST (500 ns, offset 15)
scsi target0:0:0: asynchronous
sr 0:0:3:0: [sr0] scsi-1 drive
cdrom: Uniform CD-ROM driver Revision: 3.20
sd 0:0:0:0: Attached scsi generic sg0 type 0
sr 0:0:3:0: Attached scsi generic sg1 type 5

This patch resolves this issue because the bus reset causes the INQUIRY
command to fail earlier, and return the appropriate result code.
Tested-by: NStan Johnson <userm57@yahoo.com>
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch improves readability. There are no functional changes.

Since this touches on a questionable ESP_INTR_DC conditional, add some
commentary to help others who may (as I did) find themselves chasing an
"Invalid Command" error after the device flags this condition.

This cleanup also eliminates a warning from "make W=1":
drivers/scsi/esp_scsi.c: In function 'esp_finish_select':
drivers/scsi/esp_scsi.c:1233:5: warning: variable 'orig_select_state' set but not used [-Wunused-but-set-variable]
  u8 orig_select_state;
Tested-by: NStan Johnson <userm57@yahoo.com>
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>