Commit 64d513ac ("scsi: use host wide tags by default") causes
the SCSI core to queue more commands then we can handle on devices with
multiple LUNs, limit the queue depth at the scsi-host level instead of
per slave to fix this.

BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1315013
Cc: stable@vger.kernel.org # 4.4.x and 4.5.x
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The uas driver can never queue more then MAX_CMNDS (- 1) tags and tags
are shared between luns, so there is no need to claim that we can_queue
some random large number.

Not claiming that we can_queue 65536 commands, fixes the uas driver
failing to initialize while allocating the tag map with a "Page allocation
failure (order 7)" error on systems which have been running for a while
and thus have fragmented memory.

Cc: stable@vger.kernel.org
Reported-and-tested-by: NYves-Alexis Perez <corsac@corsac.net>
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Some devices send response IUs when you'd expect a sense IU.
As a response to a wrong LUN that is within spec.
We cannot get away without handling for response IUs.
This version fixes the issues Hans raised.
Signed-off-by: NOliver Neukum <oneukum@suse.com>
Reviewed-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This function must be called with a spinlock held.
Memory can be allocated only with GFP_ATOMIC. Passing
a gfp_t argument is a waste.
Signed-off-by: NOliver Neukum <oneukum@suse.com>
Reviewed-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use this macro to make the driver more readable.
Signed-off-by: NOliver Neukum <oneukum@suse.com>
Reviewed-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The patch extends the family of SATA-to-USB JMicron adapters that need
FUA to be disabled and applies the same policy for uas driver.
See details in http://unix.stackexchange.com/questions/237204/Signed-off-by: NDmitry Katsubo <dmitry.katsubo@gmail.com>
Tested-by: NDmitry Katsubo <dmitry.katsubo@gmail.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch changes the !blk-mq path to the same defaults as the blk-mq
I/O path by always enabling block tagging, and always using host wide
tags.  We've had blk-mq available for a few releases so bugs with
this mode should have been ironed out, and this ensures we get better
coverage of over tagging setup over different configs.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NJens Axboe <axboe@kernel.dk>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Signed-off-by: NJames Bottomley <JBottomley@Odin.com>

-ESHUTDOWN means that the HC has been unplugged.
Reporting an error in that case makes no sense.
Signed-off-by: NOliver Neukum <oneukum@suse.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

'0' is now used as the default cmd_per_lun value,
so there's no need to explicitly set it to '1' in the
host template.
Signed-off-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJames Bottomley <JBottomley@Odin.com>

The usb-storage driver sets max_sectors = 240 in its scsi-host template,
for uas we do not want to do that for all devices, but testing has shown
that some devices need it.

This commit adds a US_FL_MAX_SECTORS_240 flag for such devices, and
implements support for it in uas.c, while at it it also adds support
for US_FL_MAX_SECTORS_64 to uas.c.

Cc: stable@vger.kernel.org # 3.16
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

uas_use_uas_driver may set some US_FL_foo flags during detection, currently
these are stored in a local variable and then throw away, but these may be
of interest to the caller, so add an extra parameter to (optionally) return
the detected flags, and use this in the uas driver.

Cc: stable@vger.kernel.org # 3.16
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Acked-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Drop the now unused reason argument from the ->change_queue_depth method.
Also add a return value to scsi_adjust_queue_depth, and rename it to
scsi_change_queue_depth now that it can be used as the default
->change_queue_depth implementation.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NHannes Reinecke <hare@suse.de>

Remove the tagged argument from scsi_adjust_queue_depth, and just let it
handle the queue depth.  For most drivers those two are fairly separate,
given that most modern drivers don't care about the SCSI "tagged" status
of a command at all, and many old drivers allow queuing of multiple
untagged commands in the driver.

Instead we start out with the ->simple_tags flag set before calling
->slave_configure, which is how all drivers actually looking at
->simple_tags except for one worke anyway.  The one other case looks
broken, but I've kept the behavior as-is for now.

Except for that we only change ->simple_tags from the ->change_queue_type,
and when rejecting a tag message in a single driver, so keeping this
churn out of scsi_adjust_queue_depth is a clear win.

Now that the usage of scsi_adjust_queue_depth is more obvious we can
also remove all the trivial instances in ->slave_alloc or ->slave_configure
that just set it to the cmd_per_lun default.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NHannes Reinecke <hare@suse.de>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>

Allow a driver to ask for block layer tags by setting .use_blk_tags in the
host template, in which case it will always see a valid value in
request->tag, similar to the behavior when using blk-mq.  This means even
SCSI "untagged" commands will now have a tag, which is especially useful
when using a host-wide tag map.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.de>

Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBart Van Assche <bvanassche@acm.org>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.de>

Currently scsi piggy backs on the block layer to define the concept
of a tagged command.  But we want to be able to have block-level host-wide
tags assigned even for untagged commands like the initial INQUIRY, so add
a new SCSI-level flag for commands that are tagged at the scsi level, so
that even commands without that set can have tags assigned to them.  Note
that this alredy is the case for the blk-mq code path, and this just lets
the old path catch up with it.

We also set this flag based upon sdev->simple_tags instead of the block
queue flag, so that it is entirely independent of the block layer tagging,
and thus always correct even if a driver doesn't use block level tagging
yet.

Also remove the old blk_rq_tagged; it was only used by SCSI drivers, and
removing it forces them to look for the proper replacement.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMike Christie <michaelc@cs.wisc.edu>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.de>

With uas over usb-3 the tags inside the uas iu-s must match the usb-3 stream
ids, and those go from 1 - qdepth.

Before blk-mq calling scsi_activate_tcq(sdev, qdepth) guaranteed that we would
only get cmnd->request->tag from 0 - (qdepth - 1), and we used those as
uas-tags / stream-ids.

With blk-mq however we are guaranteed to never get more then qdepth commands
queued at the same time, but the cmnd->request->tag values may be much larger,
which breaks uas.

This commit fixes this by generating uas tags in the 1 - qdepth range ourselves
instead of using cmnd->request->tag.

While touching all involved code anyways also rename the uas_cmd_info stream
field to uas_tag, because when using uas over usb-2 streams are not used.

Cc: Christoph Hellwig <hch@infradead.org>
Reported-by: NDouglas Gilbert <dgilbert@interlog.com>
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>

--
Changes in v2:
-Remove ".disable_blk_mq = true" from uas_host_template
Changes in v3:
-Rebased on top of Linus' current master branch
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The stream_id and pipe are already present in uas_cmd_info resp uas_dev_info,
so there is no need to pass a copy along.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The uas driver uses the block layer tag for USB3 stream IDs.  With
blk-mq we can get larger tag numbers that the queue depth, which breaks
this assumption.  A fix is under way for 3.18, but sits on top of
large changes so can't easily be backported.   Set the disable_blk_mq
path so that a uas device can't easily crash the system when using
blk-mq for SCSI.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NJames Bottomley <JBottomley@Parallels.com>

If something goes wrong in our communication with an uas device we may get
a response iu in reaction to a cmnd, rather then a status iu. In this case
propagate an error upwards, rather then logging a bogus iu message.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Instead of doing:

uas_log_cmd_state(cmnd, __func__)
scmd_printk(KERN_ERR, cmnd, "error doing foo %d\n", err)

On error, resulting in 2 log calls for a single error, make uas_log_cmd_state
take a status code, and change calls like the above to:

uas_log_cmd_state(cmnd, "error doing foo", err)

Also change various sanity checks (which should never trigger) from:
"scmd_printk(KERN_ERR, cmnd, "sanity foo failed\n")" to calling the new
uas_log_cmd_state(), so that when they do trigger we get more info.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We've removed all hack from the driver for pre-production hardware.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

I've access to a number of different uas devices now, and none of them use
old style sense urbs. The only case where these code-paths trigger is with
the asm1051 and there they do the wrong thing, as the asm1051 sends 8 bytes
status iu-s when it does not have any sense data, but uses new style
sense iu-s regardless, as can be seen for scsi cmnds where there is sense
data.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It was only used to sanity check against completing the same cmnd twice,
but that is the case we're likely operating on free-ed memory, and doing
sanity checks on free-ed memory is not really helpful.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use scsi_print_command to print commands during errors, rather then printing
the rather meaningless pointer to the command.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Check for both type of cancellation codes for sense and data urbs.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Limit the no-streams case to speeds less then USB_SPEED_SUPER.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The purpose of uas_pre_reset is to:

1) Stop any new commands from being submitted while an externally triggered
   usb-device-reset is running
2) Wait for any pending commands to finish before allowing the usb-device-reset
   to continue

The purpose of uas_suspend is to:
2) Wait for any pending commands to finish before suspending

This commit fixes races in both paths:

1) For 1) we use scsi_block_requests, but the scsi midlayer calls queuecommand
   without holding any locks, so a queuecommand may already past the midlayer
   scsi_block_requests checks when we call it, add a check to uas_queuecommand
   to fix this

2) For 2) we were waiting for all sense-urbs to complete, there are 2 problems
   with this approach:
a) data-urbs may complete after the sense urb, so we need to check for those
   too
b) if a sense-urb completes with a iu id of READ/WRITE_READY a command is not
   yet done. We submit a new sense-urb immediately in this case, but that
   submit may fail (in which case it will get retried by uas_do_work), if this
   happens the sense_urbs anchor may become empty while the cmnd is not yet
   done

Also unblock requests on timeout, to avoid things getting stuck in that case.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Not all urbs we've allocated are necessarily also submitted, non-submitted
urbs will not be free-ed by their completion handler. So we need to free
them manually.

There are 2 scenarios where this can happen:

1) We have failed to submit some urbs at abort / disconnect
2) When running over usb-2 we may have never tried to submit the data urbs
   when completing the scsi cmnd, because we never got a READ/WRITE_READY iu
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Do not keep references around to a cmnd which is under error handling.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It is not strictly necessary for the cmd urb to have a reference to the
cmnd, and without this reference it becomes easier to drop all references to
a cmnd on an abort.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We've the same info doubled in both the inflight list and the cmnd array,
drop the list.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The data urbs are all killed before calling zap_pending, and their completion
handler should have cleared their inflight flag.

Do not 0 the data inflight flags, and add a check for try_complete succeeding,
as it should always succeed when called from zap_pending.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Drop the whole dance with first moving cmnds to a dead-list. The resetting
flag ensures that no new cmds / urbs will be submitted, and that any urb
completions are short-circuited without trying to complete the scsi cmnd.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Now that we no longer drop our lock to unlink the data urbs, we can simply
free them on completion, making their handling consistent with the other urbs.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

There is no need for all the trickery with dropping the lock, we can
simply reference the urbs while we hold the lock to ensure the urbs don't
disappear beneath us, and do the actual unlink (+ unreference) after we've
dropped the lock.

This also fixes a race where we may loose of cmnd ownership to the scsi
midlayer without holding the lock due to the midlayer re-claiming ownership
through an abort (which will be handled by a future patch in this series).
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The status urb should not complete before the command has been submitted, nor
should we get a second status urb for the same tag after a IU_ID_STATUS.

Data urbs should not complete before the command has been submitted, but may
complete after the IU_ID_STATUS.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Using scsi_host_find_tag with tags returned by the device is unsafe for
multiple reasons:

1) It returns tags->rqs[tag], which may be non NULL even when the cmnd is
   not owned by us
2) It returns tags->rqs[tag], without holding any locks protecting it
3) It returns tags->rqs[tag], without doing any boundary checking

Instead keep our own list which maps tags -> inflight cmnds.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Factor out the mapping of scsi-tags -> uas-tags/stream-ids to a helper function
so that there is a single place where this "magic" happens.
Signed-off-by: NHans de Goede <hdegoede@redhat.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>