drm/i915: Extract i9xx_plane_ctl() and ironlake_plane_ctl()

Pull the code to calculate the pre-SKL primary plane control register
value into separate functions. Allows us to pre-compute it in the
future.

v2: Split the pre-ilk vs. ilk+ unification to a separate patch (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323192712.30682-2-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Use BIT() for computing the engine's flag

Since the engine's flag is just the bit of its id, use BIT().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324163540.31981-3-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Remove unused intel_flush_status_page()

intel_flush_status_page() is defunct since commit f8dd2934c4ec
("drm/i915: Remove BXT incoherent seqno write workaround"), time to
remove it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324163540.31981-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Fixup intel_write_status_page() for old CPUs without clflush

Not all of our target platforms have clflush. For those without, just
assume the status page is sufficiently coherent that we do not need our
paranoia.

Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: 14a6bbf9e535 ("drm/i915: Replace irq_seqno_barrier on hws write with a clflush")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324163540.31981-1-chris@chris-wilson.co.uk
Tested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Align "unfenced" tiled access on gen2, early gen3

Old devices have quite severe restrictions for using fences, and unlike
more recent device (anything from Pineview onwards) we need to enforce
those restrictions even for unfenced tiled access from the render
pipeline.

Fixes: 944397f04f24 ("drm/i915: Store required fence size/alignment for GGTT vma")
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.11-rc1+
Link: http://patchwork.freedesktop.org/patch/msgid/20170325113243.16438-1-chris@chris-wilson.co.uk
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Tested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Limit number of reads to stabilize rc6 counter reads

We have only 8bits of precise timestamps in which to complete our
upper/load reads, along with the switch between precision. This is not
always enough time to read the upper counter twice within the same time
slice, leading to hard lockups. Limit the number of times to prevent
an inifite loop (my fault for assuming we would have no trouble doing
the write + reads fast enough).

Fixes: 47c21d9a1a7b ("drm/i915: Extend vlv/chv residency resolution")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100377
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324165418.7455-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Check we have an wake device before flushing GTT writes

We can assume that if the device is asleep then all pending GTT writes
will have been posted, and so we can defer the flush from
i915_gem_object_flush_gtt_write_domain()

[ 1957.462568] WARNING: CPU: 0 PID: 6132 at drivers/gpu/drm/i915/intel_drv.h:1742 fwtable_read32+0x123/0x150 [i915]
[ 1957.462582] RPM wakelock ref not held during HW access
[ 1957.462583] Modules linked in: i915 intel_gtt drm_kms_helper prime_numbers
[ 1957.462607] CPU: 0 PID: 6132 Comm: gem_concurrent_ Tainted: G     U          4.11.0-rc1+ #464
[ 1957.462619] Hardware name:                  /        , BIOS PYBSWCEL.86A.0027.2015.0507.1758 05/07/2015
[ 1957.462630] Call Trace:
[ 1957.462646]  dump_stack+0x4d/0x6f
[ 1957.462657]  __warn+0xc1/0xe0
[ 1957.462667]  warn_slowpath_fmt+0x4a/0x50
[ 1957.462709]  fwtable_read32+0x123/0x150 [i915]
[ 1957.462750]  i915_gem_object_flush_gtt_write_domain+0x43/0x70 [i915]
[ 1957.462791]  i915_gem_object_set_to_cpu_domain+0x46/0xa0 [i915]
[ 1957.462831]  i915_gem_set_domain_ioctl+0x15d/0x220 [i915]
[ 1957.462843]  drm_ioctl+0x1d7/0x440
[ 1957.462885]  ? i915_gem_obj_prepare_shmem_write+0x1d0/0x1d0 [i915]
[ 1957.462896]  ? pick_next_task_fair+0x436/0x440
[ 1957.462906]  ? mntput+0x1f/0x30
[ 1957.462915]  do_vfs_ioctl+0x8f/0x5c0
[ 1957.462925]  ? __schedule+0x16f/0x5f0
[ 1957.462935]  ? ____fput+0x9/0x10
[ 1957.462943]  SyS_ioctl+0x3c/0x70
[ 1957.462952]  entry_SYSCALL_64_fastpath+0x17/0x98
[ 1957.462961] RIP: 0033:0x7fc542179ca7
[ 1957.462968] RSP: 002b:00007ffeef12ff98 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[ 1957.462982] RAX: ffffffffffffffda RBX: 00007ffeef1301d0 RCX: 00007fc542179ca7
[ 1957.462990] RDX: 00007ffeef12ffd0 RSI: 00000000400c645f RDI: 0000000000000003
[ 1957.462999] RBP: 0000000000000003 R08: 000055f433bc7c40 R09: 000000000000002c
[ 1957.463006] R10: 0000000000000073 R11: 0000000000000246 R12: 0000000000000018
[ 1957.463015] R13: 000055f432c89d20 R14: 000055f432c87690 R15: 0000000000000000

Fixes: 3b5724d702ef ("drm/i915: Wait for writes through the GTT to land before reading back")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323150053.28582-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/execlists: Trim irq handler

I noticed that gcc was spilling the CSB to the stack, so rearrange the
code to be more compact. Spilling in this function is slightly more
interesting due to the mmio reads acting as memory barriers and so
end up flushing the stack spills. Still miniscule to having to do at
least the pair of uncached reads :(

function                                     old     new   delta
intel_lrc_irq_handler                       1039     878    -161

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170325201053.21306-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/uc: Make intel_uc_prepare_fw() static

There is no need to expose this function as it is called from
one function only. Also move it up to avoid forward declaration.

v2: drop intel_ prefix (Oscar) and rename to fetch_uc_fw (Michal)

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170327094510.167400-1-michal.wajdeczko@intel.com
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Fix semaphore emission for BDW+ RCS ringbuffer emission

The required number of dwords for semaphore emission on BDW RCS is 8,
not 6 - leading to ring buffer corruption and immediate GPU hangs when
using ringbuffer submission.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324151724.32640-2-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

drm/i915: Disable MI_SET_CONTEXT psmi w/a for bdw

The current w/a for the gen7 psmi related hangs doesn't apply to bdw, so
disable it if using bdw ringbuffer submission.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170324151724.32640-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

drm/i915/guc: limit forcewake to blitter domain in guc_send

The forcewake_get call in the guc_send_mmio function was added to
avoid getting and releasing forcewake on each register access.
While this makes sense, all GuC registers are in the blitter range
so no need to wake all the wells.

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1490366919-34715-1-git-send-email-daniele.ceraolospurio@intel.com
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Refactor the retrieval of guc_process_desc

Move the common "client->vaddr + client->proc_desc_offset" to its own
function, __get_process_desc() to match the newly established pattern.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323230000.20786-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

drm/i915/execlists: Relax the locked clear_bit(IRQ_EXECLIST)

We only need to care about the ordering of the clearing of the bit with
the uncached CSB read in order to correctly detect a new interrupt
before the read completes. The uncached read itself acts as a full
memory barrier, so we do not need to enforce another in the form of a
locked clear_bit.

v2: Clarify why the split and unlocked test/clear is harmless.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323134803.10418-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Extract i845_cursor_ctl() and i9xx_cursor_ctl()

Pull the code to calculate the cursor control register value into
separate functions. Allows us to pre-compute them in the future.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-8-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Extract ilk_sprite_ctl()

Pull the code to calculate the ILK-SNB sprite control register value
into a separate function. Allows us to pre-compute it in the future.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-7-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Extract ivb_sprite_ctl()

Pull the code to calculate the IVB-BDW sprite control register value
into a separate function. Allows us to pre-compute it in the future.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-6-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Extract vlv_sprite_ctl()

Pull the code to calculate the VLV/CHV sprite control register value
into a separate function. Allows us to pre-compute it in the future.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-5-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Use skl_plane_ctl() for the SKL "sprite" planes

On SKL the planes are uniform so the "sprites" can use the
primary plane code perfectly fine. The only difference we
have is the color key handling, but since we never enable that
for the primary plane the same code works just fine.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-3-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Extract skl_plane_ctl()

Pull the code to calculate the SKL plane control register value into
a separate function. Allows us to pre-compute it in the future.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317211808.14693-2-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Move guc_interrupts_release next to guc_interrupts_capture

They go better together.

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Split out the mmio_white_list struct

We are going to need it for future platforms.

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Refactor the concept "GuC context descriptor" into "GuC stage descriptor"

A GuC context and a HW context are in no way related, so the name "GuC context descriptor"
is very unfortunate, because a new reader of the code gets overwhelmed very quickly with
a lot of things called "context" that refer to different things. We can improve legibility
a lot by simply renaming a few objects in the GuC code.

v2:
  - Rebased
  - s/ctx_desc_pool/stage_desc_pool
  - Move some explanations to the definition of the guc_stage_desc struct (Chris)

v3:
  - Calculate gemsize with less intermediate steps (Joonas)
  - Use BIT() macro (Joonas)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: A little bit more of doorbell sanitization

Some recent refactoring patches have left the doorbell creation outside
the GuC client allocation, which does not make a lot of sense (a client
without a doorbell is something useless). Move it back there, and
refactor the init_doorbell_hw consequently.

Thanks to this, we can do some other improvements, like hoisting the
check for GuC submission enabled out of the enable function.

v2: Rebased.

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Wait for doorbell to be inactive before deallocating

Doorbell release flow requires that we wait for GEN8_DRB_VALID bit to go
to zero after updating db_status before we call the GuC to release the
doorbell.

Kudos to Daniele for finding this out.

v2: WARN instead of DRM_ERROR (Joonas)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Improve the GuC documentation & comments about proxy submissions

While at it, fix a typo (s/ring_lcra/ring_lrca) and improve the naming of one
firware interface field (s/ring_tail/submit_element_info, since it can contain
more than just the ring tail).

No change in functionality.

v2:
  - Remove reference to "unique user" of the GuC (Daniele)
  - Keep mention to renaming from "GuC context" to "client" (Daniele)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Make intel_guc_send a function pointer

Prepare for an alternate GuC communication interface.

v2: Make a few functions static and name them correctly while we are at it (Oscar), but
leave an intel_guc_send_mmio interface for users that require old-style communication.

v3: Send intel_uc_init_early back to the top (Michal).

Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Break out the GuC log extras into their own "runtime" struct

When initializing the GuC log struct, there is an object we need to
allocate always, since the GuC needs its address at fw load time.
The rest is only needed during runtime, in the sense that we only
create if we actually enable GuC logging. Make that distinction
explicit by subdividing further the intel_guc_log struct.

v2: Call the new struct "runtime", instead of "extras" (Joonas)

v3: Check indent (Joonas)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: The Additional Data Struct (ADS) should get enabled together with GuC submission

It's mandatory and it gets created if and only if GuC submission is enabled, so that should be
the condition for informing the GuC about it.

Also s/guc_addon_create/guc_ads_create and s/guc_addon_destroy/guc_ads_destroy and, while
at it, add an explanation of what things go inside the ADS object.

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Add onion teardown to the GuC setup

Starting with intel_guc_loader, down to intel_guc_submission
and finally to intel_guc_log.

v2:
  - Null execbuf client outside guc_client_free (Daniele)
  - Assert if things try to get allocated twice (Daniele/Joonas)
  - Null guc->log.buf_addr when destroyed (Daniele)
  - Newline between returning success and error labels (Joonas)
  - Remove some unnecessary comments (Joonas)
  - Keep guc_log_create_extras naming convention (Joonas)
  - Helper function guc_log_has_extras (Joonas)
  - No need for separate relay_channel create/destroy. It's just another extra.
  - No need to nullify guc->log.flush_wq when destroyed (Joonas)
  - Hoist the check for has_extras out of guc_log_create_extras (Joonas)
  - Try to do i915_guc_log_register/unregister calls (kind of) symmetric (Daniele)
  - Make sure initel_guc_fini is not called before init is ever called (Daniele)

v3:
  - Remove unnecessary parenthesis (Joonas)
  - Check for logs enabled on debugfs registration
  - Rebase on top of Tvrtko's "Fix request re-submission after reset"

v4:
  - Rebased
  - Comment around enabling/disabling interrupts inside GuC logging (Joonas)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Keep the ctx_pool_vaddr mapped, for easy access

The GuC descriptor is big in size. If we use a local definition of
guc_desc we have a chance to overflow stack, so avoid it.

Also, Chris abhors scatterlists :)

v2: Rebased, helper function to retrieve the context descriptor,
s/ctx_pool_vma/ctx_pool/

v3: Zero out guc_context_desc before initialization

v4: Do not do arithmetic on void pointers (Daniele)

v5: Nicer than arithmetic on pointers (Chris, Joonas)

Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/guc: Sanitize GuC client initialization

Started adding proper teardown to guc_client_alloc, ended up removing
quite a few dead ends where errors communicating with the GuC were
silently ignored. There also seemed to be quite a few erronous
teardown actions performed in case of an error (ordering wrong).

v2:
  - Increase function symmetry/proximity (Michal/Daniele)
  - Fix __reserve_doorbell accounting for high priority (Daniele)
  - Call __update_doorbell_desc! (Daniele)
  - Isolate __guc_{,de}allocate_doorbell (Michal/Daniele)

v3:
  - "Select" a cacheline is a more accurate verb than "reserve" (Daniele).
  - We cannot update & create the doorbell without reserving it first, so
    move the whole doorbell creation for execbuf_client to the submission
    enable (Oscar).i
  - Add a fixme for ignoring possible doorbell destroy errors.

v4:
  - Remove comment about is_high_priority (Daniele)
  - Debug message typo (Daniele)
  - Reuse __get_doorbell in more places (Daniele)
  - Do not do arithmetic on void pointers (Daniele)
  - Add comment to __reset_doorbell (Daniele)

v5:
  - gccisms like arithmetic on void pointers are not frowned upon (Oscar)

Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>

drm/i915: Wait for all fences before installing an exclusive clflush fence

Ensure that before we overwrite the reservation_object with our
exclusive fence for the pending clflush operation, that we do wait upon
all the fences in the current reservation_object.

Fixes: 57822dc6b9cf ("drm/i915: Perform object clflushing asynchronously")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323085758.11695-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Drop uncore spinlock for reading debugfs forcewake counters

The set of available structs is not protected by the spinlock, and for
the single read we can use READ_ONCE instead.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-7-chris@chris-wilson.co.uk

drm/i915: All fw_domains share the same set/clear/reset values

Since we reuse the same values for each fw_domain, move them onto
uncore.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-6-chris@chris-wilson.co.uk

drm/i915: Remove posting-read for forcewake put

We can relax the requirement upon ourselves that the forcewake is
released immediately and just allow it to occur naturally following our
mmio request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-5-chris@chris-wilson.co.uk

drm/i915: Skip unused fw_domains

Use find-first-set bitop to quickly scan through the fw_domains mask and
skip iterating over unused domains.

v2: Move the WARN into the caller, to prevent compiler warnings in
normal builds.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-4-chris@chris-wilson.co.uk

drm/i915: Use correct fw_domains during reset

In the next patch we will begin to sanity check that we do not attempt
to obtain the forcewake on an unsupport domain. However, that is exactly
what we do during reset of the fw_domains - rectify it before it explodes.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-3-chris@chris-wilson.co.uk

drm/i915: Use correct fw_domains during initialisation

In the next patch we will begin to sanity check that we do not attempt
to obtain the forcewake on an unsupport domain. However, that is exactly
what we do during our actual initialisation of fw_domains - rectify it
before it explodes.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-2-chris@chris-wilson.co.uk

drm/i915: Eliminate per-fw_domain i915 backpointer

Pass along the drm_i915_private pointer from the caller, rather than
looking it up from each fw_domain during fw_domains_get/_put. This
allows us to then eliminate the backpointer, in exchange for a more
complicated unwrapping procedure in the rare
intel_uncore_fw_release_timer().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170323101944.21627-1-chris@chris-wilson.co.uk

drm/i915: Restore marking context objects as dirty on pinning

Commit e8a9c58fcd9a ("drm/i915: Unify active context tracking between
legacy/execlists/guc") converted the legacy intel_ringbuffer submission
to the same context pinning mechanism as execlists - that is to pin the
context until the subsequent request is retired. Previously it used the
vma retirement of the context object to keep itself pinned until the
next request (after i915_vma_move_to_active()). In the conversion, I
missed that the vma retirement was also responsible for marking the
object as dirty. Mark the context object as dirty when pinning
(equivalent to execlists) which ensures that if the context is swapped
out due to mempressure or suspend/hibernation, when it is loaded back in
it does so with the previous state (and not all zero).

Fixes: e8a9c58fcd9a ("drm/i915: Unify active context tracking between legacy/execlists/guc")
Reported-by: Dennis Gilmore <dennis@ausil.us>
Reported-by: Mathieu Marquer <mathieu.marquer@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99993
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100181
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.11-rc1
Link: http://patchwork.freedesktop.org/patch/msgid/20170322205930.12762-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Remove superfluous hw_flags from mi_set_context()

Why have both hw_flags and flags, when just one will do?

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170322210350.6208-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Actually pass the reclaim gfp_t along to shmemfs!

Words cannot describe the embarrassment at creating a new gfp_t relaim to
only prevent the oomkiller but allow direct|kswapd reclaim, and then not
use it in the shmem_read_mapping_page_gfp().

Fixes: 24f8e00a8a2e ("drm/i915: Prefer to report ENOMEM rather than incur the oom for gfx allocations")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20170322223447.7493-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Merge remote-tracking branch 'airlied/drm-next' into drm-intel-next-queued

Backmerge drm-next to get at the hdmi2.0 helper functions.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

BackMerge tag 'v4.11-rc3' into drm-next

Linux 4.11-rc3 as requested by Daniel

Merge tag 'drm-misc-next-2017-03-21' of git://anongit.freedesktop.org/git/drm-misc into drm-next

drm-misc for 4.12, 2nd attempt this week:

- topic branch from Jon Corbet for the new graph kerneldoc support
- lots of graphs for kms/atomic things using the above
- some vblank query tuning from Chris
- gem/cma_fops macros
- moar docs

Driver stuff:
- vc4 hdmi audio, yay (Eric)
- dw-hdmi polish from a bunch of people
- some rockchip dp updates that didn't make last week (Chris Zhong)
- misc bridge&driver updates

* tag 'drm-misc-next-2017-03-21' of git://anongit.freedesktop.org/git/drm-misc: (37 commits)
  drm/edid: detect SCDC support in HF-VSDB
  drm/edid: detect SCDC support in HF-VSDB
  drm/edid: check for HF-VSDB block
  drm: Add SCDC helpers
  drm: bridge: dw-hdmi: add HDMI vendor specific infoframe config
  drm/bridge: dw_hdmi: support i2c extended read mode
  drm/msm: add stubs for msm_{perf,rd}_debugfs_cleanup
  drm: bochs: Don't remove uninitialized fbdev framebuffer
  drm: vc4: remove redundant check of plane being non-null
  drm/vc4: use platform_register_drivers
  dma-fence: add dma_fence_match_context helper
  drm/vc4: Add HDMI audio support
  dt-bindings: Document the dmas and dma-names properties for VC4 HDMI
  drm/atmel-hlcdc: Fix suspend/resume implementation
  drm: Skip the waitqueue setup for vblank queries
  drm: Defer disabling the vblank IRQ until the next interrupt (for instant-off)
  drm/doc: atomic overview, with graph
  drm/doc: diagram for mode objects and properties
  drm/doc: Consistent kerneldoc include order
  drm/doc: Add KMS overview graphs
  ...

Merge tag 'drm-intel-next-2017-03-20' of git://anongit.freedesktop.org/git/drm-intel into drm-next

More in i915 for 4.12:

- designware i2c fixes from Hans de Goede, in a topic branch shared
  with other subsystems (maybe, they didn't confirm, but requested the
  pull)
- drop drm_panel usage from the intel dsi vbt panel (Jani)
- vblank evasion improvements and tracing (Maarten and Ville)
- clarify spinlock irq semantics again a bit (Tvrtko)
- new ->pwrite backend hook (right now just for shmem pageche writes),
  from Chris
- more planar/ccs work from Ville
- hotplug safe connector iterators everywhere
- userptr fixes (Chris)
- selftests for cache coloring eviction (Matthew Auld)
- extend debugfs drop_caches interface for shrinker testing (Chris)
- baytrail "the rps kills the machine" fix (Chris)
- use new atomic state iterators, a lot (Maarten)
- refactor guc/huc code some (Arkadiusz Hiler)
- tighten breadcrumbs rbtree a bit (Chris)
- improve wrap-around and time handling in rps residency counters
  (Mika)
- split reset-in-progress in two flags, backoff and handoff (Chris)
- other misc reset improvements from a few people
- bunch of vgpu interaction fixes with recent code changes
- misc stuff all over, as usual

* tag 'drm-intel-next-2017-03-20' of git://anongit.freedesktop.org/git/drm-intel: (144 commits)
  drm/i915: Update DRIVER_DATE to 20170320
  drm/i915: Initialise i915_gem_object_create_from_data() directly
  drm/i915: Correct error handling for i915_gem_object_create_from_data()
  drm/i915: i915_gem_object_create_from_data() doesn't require struct_mutex
  drm/i915: Retire an active batch pool object rather than allocate new
  drm/i915: Add i810/i815 pci-ids for completeness
  drm/i915: Skip execlists_dequeue() early if the list is empty
  drm/i915: Stop using obj->obj_exec_link outside of execbuf
  drm/i915: Squelch WARN for VLV_COUNTER_CONTROL
  drm/i915/glk: Enable pooled EUs for Geminilake
  drm/i915: Remove superfluous i915_add_request_no_flush() helper
  drm/i915/vgpu: Neuter forcewakes for VGPU more thoroughly
  drm/i915: Fix vGPU balloon for ggtt guard page
  drm/i915: Avoid use-after-free of ctx in request tracepoints
  drm/i915: Assert that the context pin_counts do not overflow
  drm/i915: Wait for reset to complete before returning from debugfs/i915_wedged
  drm/i915: Restore engine->submit_request before unwedging
  drm/i915: Move engine->submit_request selection to a vfunc
  drm/i915: Split I915_RESET_IN_PROGRESS into two flags
  drm/i915: make context status notifier head be per engine
  ...

drm/i915: Fix SKL cursor watermarks

Use intel_wm_plane_visible() to determine cursor visibility for SKL+
also. Previously SKL+ would check the actual visibility which now
conflicts with the assumptions in intel_legacy_cursor_update().

We also change SKL+ to compute the cursor watermarks based on the
unclipped cursor size, just as we do on all the other platforms.
Using the clipped size could now result in garbage results.

Testcase: igt/kms_chv_cursor_fail
Fixes: a5509abda48e ("drm/i915: Fix legacy cursor vs. watermarks for ILK-BDW")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100195
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170314151050.12194-2-ville.syrjala@linux.intel.com
Tested-by: Dorota Czaplejewicz <dorota.czaplejewicz@collabora.co.uk>
Tested-by: Jari Tahvanainen <jari.tahvanainen@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>

drm/i915: Extract intel_wm_plane_visible()

All platforms that lack double buffered watermarks will need to
handle the legacy cursor updates in the same way. So let's extract the
logic to determine the plane visibility into a small helper. For
simplicity we'll make the function DTRT for any plane, but only apply
the special sauce for cursor planes.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170314151050.12194-1-ville.syrjala@linux.intel.com
Tested-by: Dorota Czaplejewicz <dorota.czaplejewicz@collabora.co.uk>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Prefer to report ENOMEM rather than incur the oom for gfx allocations

Since gfx allocations tend to be large, unmovable and disposable, report
the allocation failure back to userspace as an ENOMEM rather than incur
the oomkiller. We have already tried to make room by purging our own
cached gfx objects, and the oomkiller doesn't attribute ownership of gfx
objects so will likely pick the wrong candidate. Instead, let userspace
see the ENOMEM.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170322110521.29930-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: simplify intel_ddi_pll_select()

Because {hsw,skl,bxt}_ddi_pll_select all pretty much do the same thing
in slightly different ways. Replace everything with a simple copy of
the function and inline it inside intle_ddi_pll_select().

v2: s/return pll/return pll != NULL/ (Ville)

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1490209125-20046-1-git-send-email-paulo.r.zanoni@intel.com

drm/i915: Implement cdclk restrictions based on Azalia BCLK

According to BSpec, "The CD clock frequency must be at least twice the
frequency of the Azalia BCLK." and BCLK is configured to 96 MHz by
default. This check is needed because BXT and GLK support cdclk
frequencies less than 192 MHz.

v2: Include other Gen9 platforms too for completeness.(Paulo)

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1489531556-2926-1-git-send-email-dhinakaran.pandiyan@intel.com

drm/i915/glk: Apply cdclk workaround for DP audio

Implement the DP-Audio cdclk restriction for GLK, similar to what is
implemented for BDW and other GEN9 platforms. The max. pixel clock
adjustment for GLK, however factors in the 2 pixels per clock output that
GLK generates.

Separating min. cdclk and max. pixel_rate would be nicer, but let's
defer that to future and fix the GLK bug for now.

Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1488931972-2865-1-git-send-email-dhinakaran.pandiyan@intel.com

drm/i915: Spinlocks in tasklets can use spin_(un)lock_irq

The tasklets callbacks are only called from tasklet context so
it is safe do to this.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170321105511.18269-1-tvrtko.ursulin@linux.intel.com

drm/i915: Remove intel_ring.last_retired_head

Storing the position of the breadcrumb of the last retired request as
a separate last_retired_head is superfluous as we always copy that into
head prior to recalculation of the intel_ring.space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170321102552.24357-1-chris@chris-wilson.co.uk

drm/i915/execlists: Split the atomic test_and_clear_bit for irq handler

Rather than impose the cost of a locked test before queuing a new
request, reduce it to a simple test_bit() with a following clear_bit()
prior to doing the CSB check. This ensure that if an interrupt does
occur whilst reading from the CSB, we still detect it (the interrupt
would trigger a rescheduling of the tasklet anyway).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170321113320.2603-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: split out check for noncontiguous pfn range

We get a warning with gcc-7 about a pointless comparison when
using a linear memmap:

drivers/gpu/drm/i915/selftests/scatterlist.c: In function 'alloc_table':
drivers/gpu/drm/i915/selftests/scatterlist.c:219:66: error: self-comparison always evaluates to false [-Werror=tautological-compare]

Splitting out the comparison into a separate function avoids the warning
and makes it slightly more obvious what happens.

Fixes: 935a2f776aa5 ("drm/i915: Add some selftests for sg_table manipulation")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: http://patchwork.freedesktop.org/patch/msgid/20170320094335.1266306-2-arnd@arndb.de
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: intel_engine_init_global_seqno() requires atomic kmap

As intel_engine_init_global_seqno() may be called by
nop_submit_request() from inside irq context, we have to use atomic
versions of kmap/kunmap. This is rare as this requires using gen8 legacy
ringbuffer submission.

Reported-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170320145609.4898-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Protect intel_engine_wakeup() for call from irq context

intel_engine_wakeup() is called by nop_request_submit() which is
installed to handle third party fences completed from within irq
context. As such, it needs the full irqsave/irqrestore and not the
partial spin_irq_lock handling.

[18942.714467] =================================
[18942.719076] [ INFO: inconsistent lock state ]
[18942.723522] 4.11.0-rc2-CI-CI_DRM_2368+ #1 Tainted: G     U  W
[18942.729970] ---------------------------------
[18942.734466] inconsistent {IN-SOFTIRQ-W} -> {SOFTIRQ-ON-W} usage.
[18942.740594] gem_eio/1275 [HC0[0]:SC0[0]:HE1:SE1] takes:
[18942.745932]  (&(&fence->lock)->rlock){+.?...}, at: [<ffffffff815ec100>] dma_fence_signal+0x100/0x
230
[18942.755331] {IN-SOFTIRQ-W} state was registered at:
[18942.760356]   __lock_acquire+0x5d0/0x1bb0
[18942.764444]   lock_acquire+0xc9/0x220
[18942.768196]   _raw_spin_lock_irqsave+0x41/0x60
[18942.772747]   dma_fence_signal+0x100/0x230
[18942.776927]   vgem_fence_timeout+0x9/0x10 [vgem]
[18942.781701]   call_timer_fn+0x92/0x380
[18942.785557]   expire_timers+0x150/0x1f0
[18942.789491]   run_timer_softirq+0x7c/0x160
[18942.793705]   __do_softirq+0x116/0x4c0
[18942.797560]   irq_exit+0xa9/0xc0
[18942.800873]   smp_apic_timer_interrupt+0x38/0x50
[18942.805611]   apic_timer_interrupt+0x90/0xa0
[18942.810008]   cpuidle_enter_state+0x135/0x380
[18942.814503]   cpuidle_enter+0x12/0x20
[18942.818250]   call_cpuidle+0x1e/0x40
[18942.821906]   do_idle+0x17e/0x1f0
[18942.825333]   cpu_startup_entry+0x18/0x20
[18942.829463]   rest_init+0x127/0x130
[18942.833025]   start_kernel+0x3f1/0x3fe
[18942.836908]   x86_64_start_reservations+0x2a/0x2c
[18942.841733]   x86_64_start_kernel+0x173/0x186
[18942.846234]   verify_cpu+0x0/0xfc
[18942.849604] irq event stamp: 30568
[18942.853140] hardirqs last  enabled at (30567): [<ffffffff8110b81f>] ktime_get+0xef/0x120
[18942.861468] hardirqs last disabled at (30568): [<ffffffff81876377>] _raw_spin_lock_irqsave+0x17/0
x60
[18942.870812] softirqs last  enabled at (30462): [<ffffffff81085cd9>] __do_softirq+0x1d9/0x4c0
[18942.879443] softirqs last disabled at (30439): [<ffffffff81086139>] irq_exit+0xa9/0xc0
[18942.887616]
[18942.887616] other info that might help us debug this:
[18942.894279]  Possible unsafe locking scenario:
[18942.894279]
[18942.900336]        CPU0
[18942.902851]        ----
[18942.905362]   lock(&(&fence->lock)->rlock);
[18942.909647]   <Interrupt>
[18942.912330]     lock(&(&fence->lock)->rlock);
[18942.916821]
[18942.916821]  *** DEADLOCK ***
[18942.916821]
[18942.922862] 1 lock held by gem_eio/1275:
[18942.926859]  #0:  (&(&fence->lock)->rlock){+.?...}, at: [<ffffffff815ec100>] dma_fence_signal+0x1
00/0x230
[18942.936651]
[18942.936651] stack backtrace:
[18942.941142] CPU: 3 PID: 1275 Comm: gem_eio Tainted: G     U  W       4.11.0-rc2-CI-CI_DRM_2368+ #
1
[18942.950367] Hardware name: Gigabyte Technology Co., Ltd. Z170X-UD5/Z170X-UD5-CF, BIOS F21 01/06/2
017
[18942.959756] Call Trace:
[18942.962244]  dump_stack+0x67/0x92
[18942.965626]  print_usage_bug.part.23+0x259/0x268
[18942.970362]  mark_lock+0x12c/0x6f0
[18942.973851]  ? check_usage_forwards+0x130/0x130
[18942.978487]  mark_held_locks+0x6f/0xa0
[18942.982329]  ? _raw_spin_unlock_irq+0x27/0x50
[18942.986797]  trace_hardirqs_on_caller+0x150/0x200
[18942.991599]  trace_hardirqs_on+0xd/0x10
[18942.995515]  _raw_spin_unlock_irq+0x27/0x50
[18942.999796]  intel_engine_wakeup+0x26/0x30 [i915]
[18943.004670]  intel_engine_init_global_seqno+0x131/0x1a0 [i915]
[18943.010745]  nop_submit_request+0x2e/0x40 [i915]
[18943.015476]  submit_notify+0x3f/0x5c [i915]
[18943.019763]  __i915_sw_fence_complete+0x176/0x220 [i915]
[18943.025234]  ? try_to_del_timer_sync+0x4d/0x60
[18943.029825]  i915_sw_fence_complete+0x25/0x40 [i915]
[18943.034887]  dma_i915_sw_fence_wake+0x26/0x60 [i915]
[18943.039959]  dma_fence_signal+0x146/0x230
[18943.044109]  vgem_fence_signal_ioctl+0x6c/0xc0 [vgem]
[18943.049275]  drm_ioctl+0x200/0x450
[18943.052758]  ? vgem_fence_attach_ioctl+0x270/0x270 [vgem]
[18943.058334]  do_vfs_ioctl+0x90/0x6e0
[18943.061991]  ? entry_SYSCALL_64_fastpath+0x5/0xb1
[18943.066843]  ? __this_cpu_preempt_check+0x13/0x20
[18943.071643]  ? trace_hardirqs_on_caller+0xe7/0x200
[18943.076532]  SyS_ioctl+0x3c/0x70
[18943.079842]  entry_SYSCALL_64_fastpath+0x1c/0xb1
[18943.084558] RIP: 0033:0x7f0dfcc14357
[18943.088240] RSP: 002b:00007ffeb4628da8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[18943.095996] RAX: ffffffffffffffda RBX: ffffffff8147eb93 RCX: 00007f0dfcc14357
[18943.103311] RDX: 00007ffeb4628de0 RSI: 0000000040086442 RDI: 0000000000000005
[18943.110574] RBP: ffffc9000176ff88 R08: 0000000000000004 R09: 0000000000000000
[18943.117845] R10: 0000000000000029 R11: 0000000000000246 R12: 0000000000000001
[18943.125168] R13: 0000000000000005 R14: 0000000040086442 R15: 0000000000000000
[18943.132520]  ? __this_cpu_preempt_check+0x13/0x20

Fixes: cdc3a4539034 ("drm/i915: No need to save/restore irq status in intel_engine_wakeup")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170320143133.1507-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/guc: Correct the request_in tracepoint position

It has to be called after the global seqno has been assigned.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 31de73501ac9 ("drm/i915/scheduler: emulate a scheduler for guc")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170320132556.29286-1-tvrtko.ursulin@linux.intel.com

drm/edid: detect SCDC support in HF-VSDB

This patch does following:
- Adds a new structure (drm_hdmi_info) in drm_display_info.
  This structure will be used to save and indicate if sink
  supports advanced HDMI 2.0 features
- Adds another structure drm_scdc within drm_hdmi_info, to
  reflect scdc support and capabilities in connected HDMI 2.0 sink.
- Checks the HF-VSDB block for presence of SCDC, and marks it
  in scdc structure
- If SCDC is present, checks if sink is capable of generating
  SCDC read request, and marks it in scdc structure.

V2: Addressed review comments
  Thierry:
  - Fix typos in commit message and make abbreviation consistent
    across the commit message.
  - Change structure object name from hdmi_info -> hdmi
  - Fix typos and abbreviations in description of structure drm_hdmi_info
    end the description with a full stop.
  - Create a structure drm_scdc, and keep all information related to SCDC
    register set (supported, read request supported) etc in it.

  Ville:
  - Change rr -> read_request
  - Call drm_detect_scrambling function drm_parse_hf_vsdb so that all
    of HF-VSDB parsing can be kept in same function, in incremental
    patches.

V3: Rebase.
V4: Rebase.
V5: Rebase.
V6: Addressed review comments from Ville
  - Add clock rate calculations for 1/10 and 1/40 ratios
  - Remove leftovers from old patchset
V7: Added R-B from Jose.
V8: Rebase.
V9: Rebase.
V10: Rebase.

Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1489404244-16608-5-git-send-email-shashank.sharma@intel.com

drm/edid: detect SCDC support in HF-VSDB

This patch does following:
- Adds a new structure (drm_hdmi_info) in drm_display_info.
  This structure will be used to save and indicate if sink
  supports advanced HDMI 2.0 features
- Adds another structure drm_scdc within drm_hdmi_info, to
  reflect scdc support and capabilities in connected HDMI 2.0 sink.
- Checks the HF-VSDB block for presence of SCDC, and marks it
  in scdc structure
- If SCDC is present, checks if sink is capable of generating
  SCDC read request, and marks it in scdc structure.

V2: Addressed review comments
 Thierry:
 - Fix typos in commit message and make abbreviation consistent
   across the commit message.
 - Change structure object name from hdmi_info -> hdmi
 - Fix typos and abbreviations in description of structure drm_hdmi_info
   end the description with a full stop.
 - Create a structure drm_scdc, and keep all information related to SCDC
   register set (supported, read request supported) etc in it.

Ville:
 - Change rr -> read_request
 - Call drm_detect_scrambling function drm_parse_hf_vsdb so that all
   of HF-VSDB parsing can be kept in same function, in incremental
   patches.

V3: Rebase.
V4: Rebase.
V5: Rebase.
V6: Rebase.
V7: Added R-B from Jose.
V8: Rebase.
V9: Rebase.
V10: Rebase.

Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1489404244-16608-4-git-send-email-shashank.sharma@intel.com

drm/edid: check for HF-VSDB block

This patch implements a small function that finds if a
given CEA db is hdmi-forum vendor specific data block
or not.

V2: Rebase.
V3: Added R-B from Jose.
V4: Rebase
V5: Rebase
V6: Rebase
V7: Rebase
V8: Rebase
V9: Rebase
V10: Rebase

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Reviewed-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1489404244-16608-3-git-send-email-shashank.sharma@intel.com

drm: Add SCDC helpers

SCDC is a mechanism defined in the HDMI 2.0 specification that allows
the source and sink devices to communicate.

This commit introduces helpers to access the SCDC and provides the
symbolic names for the various registers defined in the specification.

V2: Rebase.
V3: Added R-B from Jose.
V4: Rebase
V5: Addressed review comments from Ville
 - Handle the I2c return values in a better way (dp_dual_mode)
 - Make the macros for SCDC Major/Minor more readable, by adding
   a 'GET' in the macro names
V6: Rebase
V7: Rebase
V8: Rebase
V9: Rebase
V10: Rebase

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Reviewed-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1489404244-16608-2-git-send-email-shashank.sharma@intel.com

drm: bridge: dw-hdmi: add HDMI vendor specific infoframe config

Vendor specific infoframe is mandatory for 4K2K resolution.
Without this, the HDMI protocol compliance fails.

Signed-off-by: Nickey Yang <nickey.yang@rock-chips.com>
Reviewed-by: Jose Abreu <joabreu@synopsys.com>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1490081777-2232-1-git-send-email-nickey.yang@rock-chips.com

drm/bridge: dw_hdmi: support i2c extended read mode

"I2C Master Interface Extended Read Mode" implements a segment
pointer-based read operation using the Special Register configuration.

This patch fix https://patchwork.kernel.org/patch/7098101/ mentioned
"The current implementation does not support "I2C Master Interface
Extended Read Mode" to read data addressed by non-zero segment
pointer, this means that if EDID has more than 1 extension blocks,
EDID reading operation won't succeed"

With this patch, dw-hdmi can read EDID data with 1/2/4 blocks.

Signed-off-by: Nickey Yang <nickey.yang@rock-chips.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1489978651-16647-1-git-send-email-nickey.yang@rock-chips.com

drm/msm: add stubs for msm_{perf,rd}_debugfs_cleanup

We now call those two functions even when they are not defined
or declared anywhere because DEBUG_FS is disabled:

drivers/gpu/drm/msm/msm_drv.c: In function 'msm_drm_uninit':
drivers/gpu/drm/msm/msm_drv.c:244:2: error: implicit declaration of function 'msm_perf_debugfs_cleanup';did you mean 'msm_framebuffer_cleanup'? [-Werror=implicit-function-declaration]
drivers/gpu/drm/msm/msm_drv.c:245:2: error: implicit declaration of function 'msm_rd_debugfs_cleanup';did you mean 'msm_framebuffer_cleanup'? [-Werror=implicit-function-declaration]

This adds empty stub implementations for that case.

Fixes: 85eac4700ede ("drm/msm: Remove msm_debugfs_cleanup()")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20170320093936.1255573-1-arnd@arndb.de

drm/i915: Skip force-wake for uncached mmio flush of GGTT writes

The trick of using an uncached mmio read to ensure that the GGTT writes
are flushed does not require us to do the forcewake dance, so avoid it
in the hope of reducing the frequency that we do keep the device forced
awake.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170318104257.694-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Reset tasklet back to execlists after disabling guc

When switching back to execlists, we also now need to restore the
tasklet handler.

Reported-by: Oscar Mateo <oscar.mateo@intel.com>
Fixes: 31de73501ac9 ("drm/i915/scheduler: emulate a scheduler for guc")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170318102859.24101-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

drm: bochs: Don't remove uninitialized fbdev framebuffer

In the same spirit of the fix for QXL in commit 861078381ba5 ("drm: qxl:
Don't alloc fbdev if emulation is not supported"), prevent the Oops in
the unbind path of Bochs if fbdev emulation is disabled.

[  112.176009] Oops: 0002 [#1] SMP
[  112.176009] Modules linked in: bochs_drm
[  112.176009] CPU: 0 PID: 3002 Comm: bash Not tainted 4.11.0-rc1+ #111
[  112.176009] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.9.3-20161025_171302-gandalf 04/01/2014
[  112.176009] task: ffff8800743bbac0 task.stack: ffffc90000b5c000
[  112.176009] RIP: 0010:mutex_lock+0x18/0x30
[  112.176009] RSP: 0018:ffffc90000b5fc78 EFLAGS: 00010246
[  112.176009] RAX: 0000000000000000 RBX: 0000000000000260 RCX: 0000000000000000
[  112.176009] RDX: ffff8800743bbac0 RSI: ffff8800787176e0 RDI: 0000000000000260
[  112.176009] RBP: ffffc90000b5fc80 R08: ffffffff00000000 R09: 00000000ffffffff
[  112.176009] R10: ffff88007b463650 R11: 0000000000000000 R12: 0000000000000260
[  112.176009] R13: ffff8800787176e0 R14: ffffffffa0003068 R15: 0000000000000060
[  112.176009] FS:  00007f20564c7b40(0000) GS:ffff88007ce00000(0000) knlGS:0000000000000000
[  112.176009] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  112.176009] CR2: 0000000000000260 CR3: 000000006b89c000 CR4: 00000000000006f0
[  112.176009] Call Trace:
[  112.176009]  drm_mode_object_unregister+0x1e/0x50
[  112.176009]  drm_framebuffer_unregister_private+0x15/0x20
[  112.176009]  bochs_fbdev_fini+0x57/0x70 [bochs_drm]
[  112.176009]  bochs_unload+0x16/0x50 [bochs_drm]
[  112.176009]  drm_dev_unregister+0x37/0xd0
[  112.176009]  drm_put_dev+0x31/0x60
[  112.176009]  bochs_pci_remove+0x10/0x20 [bochs_drm]
[  112.176009]  pci_device_remove+0x34/0xb0
[  112.176009]  device_release_driver_internal+0x150/0x200
[  112.176009]  device_release_driver+0xd/0x10
[  112.176009]  unbind_store+0x108/0x150
[  112.176009]  drv_attr_store+0x20/0x30
[  112.176009]  sysfs_kf_write+0x32/0x40
[  112.176009]  kernfs_fop_write+0x10b/0x190
[  112.176009]  __vfs_write+0x23/0x120
[  112.176009]  ? security_file_permission+0x36/0xb0
[  112.176009]  ? rw_verify_area+0x49/0xb0
[  112.176009]  vfs_write+0xb0/0x190
[  112.176009]  SyS_write+0x41/0xa0
[  112.176009]  entry_SYSCALL_64_fastpath+0x1a/0xa9
[  112.176009] RIP: 0033:0x7f2055bd5620
[  112.176009] RSP: 002b:00007ffed2f487d8 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
[  112.176009] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f2055bd5620
[  112.176009] RDX: 000000000000000d RSI: 0000000000ee0008 RDI: 0000000000000001
[  112.176009] RBP: 0000000000000001 R08: 00007f2055e94760 R09: 00007f20564c7b40
[  112.176009] R10: 0000000000000073 R11: 0000000000000246 R12: 0000000000000000
[  112.176009] R13: 00007ffed2f48d70 R14: 0000000000000000 R15: 0000000000000000
[  112.176009] Code: 00 00 00 55 be 02 00 00 00 48 89 e5 e8 62 fb ff ff 5d c3 55 48 89 e5 53 48 89 fb e8 53 e9 ff ff 65 48 8b 14 25 40 c4 00 00 31 c0 <f0> 48 0f b1 13 48 85 c0 74 08 48 89 df e8c6 ff ff ff 5b 5d c3
[  112.176009] RIP: mutex_lock+0x18/0x30 RSP: ffffc90000b5fc78
[  112.176009] CR2: 0000000000000260
[  112.205622] ---[ end trace 76189cd7a9bdd155 ]---

Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170317181409.4183-1-krisman@collabora.co.uk
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

drm/i915: Update DRIVER_DATE to 20170320

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Merge tag 'imx-drm-next-2017-03-17' of git://git.pengutronix.de/git/pza/linux into drm-next

imx-drm PRE/PRG support, deferred plane disabling, separate alpha support

- Initial support for the Prefetch Resolve Engine/Gasket on i.MX6QP,
  improving linear scanout buffer memory bandwidth utilization. This
  will in the future grow reordering support and allow direct scanout
  of Vivante tiled renderbuffers from the GPU.
- Deferred plane disabling gets rid of some busy waiting in the atomic
  plane disable and crtc disable paths that lead to wait_for_vblank
  timeouts.
- Add support for RGBA formats with a separate alpha plane, that can
  reduce memory bandwidth utilization for mostly transparent overlay
  planes by skipping color reads for completely transparent regions.
- Allow moving an active overlay plane without enforcing a modeset.
- Add 8-bit and 16-bit bayer formats to ipu_cpmem_set_image.
- Set the base address in ipu_cpmem_set_image even for invalid formats
  to increase robustness against errors.
- Use drm_plane_helper_check_state in plane atomic_check.
- Some cleanup.

* tag 'imx-drm-next-2017-03-17' of git://git.pengutronix.de/git/pza/linux: (22 commits)
  drm/imx: Remove unneeded definition for structure imx_drm_component
  drm/imx: use PRG/PRE when possible
  drm/imx: enable/disable PRG on CRTC enable/disable
  gpu: ipu-v3: only set non-zero AXI ID for IC when PRG is absent
  gpu: ipu-v3: hook up PRG unit
  gpu: ipu-v3: document valid IPUv3 compatibles and extend for i.MX6 QuadPlus
  gpu: ipu-v3: add driver for Prefetch Resolve Gasket
  gpu: ipu-v3: add DT binding for the Prefetch Resolve Gasket
  gpu: ipu-v3: add driver for Prefetch Resolve Engine
  gpu: ipu-v3: add DT binding for the Prefetch Resolve Engine
  drm/imx: ipuv3-plane: add support for separate alpha planes
  drm/imx: extend drm_plane_state_to_eba for separate channel support
  gpu: ipu-v3: add support for separate alpha channels
  drm: add RGB formats with separate alpha plane
  drm/imx: add deferred plane disabling
  drm/imx: don't wait for vblank and stop calling cleanup_planes in commit_tail
  gpu: ipu-v3: add unsynchronised DP channel disabling
  gpu: ipu-v3: remove IRQ dance on DC channel disable
  gpu: ipu-cpmem: add bayer formats to ipu_cpmem_set_image
  gpu: ipu-cpmem: set image base address even for incorrect formats
  ...

Linux 4.11-rc3

mm/swap: don't BUG_ON() due to uninitialized swap slot cache

This BUG_ON() triggered for me once at shutdown, and I don't see a
reason for the check.  The code correctly checks whether the swap slot
cache is usable or not, so an uninitialized swap slot cache is not
actually problematic afaik.

I've temporarily just switched the BUG_ON() to a WARN_ON_ONCE(), since
I'm not sure why that seemingly pointless check was there.  I suspect
the real fix is to just remove it entirely, but for now we'll warn about
it but not bring the machine down.

Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'powerpc-4.11-5' of git://git./linux/kernel/git/powerpc/linux

Pull more powerpc fixes from Michael Ellerman:
 "A couple of minor powerpc fixes for 4.11:

   - wire up statx() syscall

   - don't print a warning on memory hotplug when HPT resizing isn't
     available

  Thanks to: David Gibson, Chandan Rajendra"

* tag 'powerpc-4.11-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/pseries: Don't give a warning when HPT resizing isn't available
  powerpc: Wire up statx() syscall

Merge branch 'parisc-4.11-2' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:

 - Mikulas Patocka added support for R_PARISC_SECREL32 relocations in
   modules with CONFIG_MODVERSIONS.

 - Dave Anglin optimized the cache flushing for vmap ranges.

 - Arvind Yadav provided a fix for a potential NULL pointer dereference
   in the parisc perf code (and some code cleanups).

 - I wired up the new statx system call, fixed some compiler warnings
   with the access_ok() macro and fixed shutdown code to really halt a
   system at shutdown instead of crashing & rebooting.

* 'parisc-4.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Fix system shutdown halt
  parisc: perf: Fix potential NULL pointer dereference
  parisc: Avoid compiler warnings with access_ok()
  parisc: Wire up statx system call
  parisc: Optimize flush_kernel_vmap_range and invalidate_kernel_vmap_range
  parisc: support R_PARISC_SECREL32 relocation in modules

Merge git://git./linux/kernel/git/nab/target-pending

Pull SCSI target fixes from Nicholas Bellinger:
 "The bulk of the changes are in qla2xxx target driver code to address
  various issues found during Cavium/QLogic's internal testing (stable
  CC's included), along with a few other stability and smaller
  miscellaneous improvements.

  There are also a couple of different patch sets from Mike Christie,
  which have been a result of his work to use target-core ALUA logic
  together with tcm-user backend driver.

  Finally, a patch to address some long standing issues with
  pass-through SCSI export of TYPE_TAPE + TYPE_MEDIUM_CHANGER devices,
  which will make folks using physical (or virtual) magnetic tape happy"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending: (28 commits)
  qla2xxx: Update driver version to 9.00.00.00-k
  qla2xxx: Fix delayed response to command for loop mode/direct connect.
  qla2xxx: Change scsi host lookup method.
  qla2xxx: Add DebugFS node to display Port Database
  qla2xxx: Use IOCB interface to submit non-critical MBX.
  qla2xxx: Add async new target notification
  qla2xxx: Export DIF stats via debugfs
  qla2xxx: Improve T10-DIF/PI handling in driver.
  qla2xxx: Allow relogin to proceed if remote login did not finish
  qla2xxx: Fix sess_lock & hardware_lock lock order problem.
  qla2xxx: Fix inadequate lock protection for ABTS.
  qla2xxx: Fix request queue corruption.
  qla2xxx: Fix memory leak for abts processing
  qla2xxx: Allow vref count to timeout on vport delete.
  tcmu: Convert cmd_time_out into backend device attribute
  tcmu: make cmd timeout configurable
  tcmu: add helper to check if dev was configured
  target: fix race during implicit transition work flushes
  target: allow userspace to set state to transitioning
  target: fix ALUA transition timeout handling
  ...

Merge branch 'libnvdimm-fixes' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull device-dax fixes from Dan Williams:
 "The device-dax driver was not being careful to handle falling back to
  smaller fault-granularity sizes.

  The driver already fails fault attempts that are smaller than the
  device's alignment, but it also needs to handle the cases where a
  larger page mapping could be established. For simplicity of the
  immediate fix the implementation just signals VM_FAULT_FALLBACK until
  fault-size == device-alignment.

  One fix is for -stable to address pmd-to-pte fallback from the
  original implementation, another fix is for the new (introduced in
  4.11-rc1) pud-to-pmd regression, and a typo fix comes along for the
  ride.

  These have received a build success notification from the kbuild
  robot"

* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  device-dax: fix debug output typo
  device-dax: fix pud fault fallback handling
  device-dax: fix pmd/pte fault fallback handling

qla2xxx: Update driver version to 9.00.00.00-k

Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
signed-off-by: Giridhar Malavali <giridhar.malavali@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Fix delayed response to command for loop mode/direct connect.

Current driver wait for FW to be in the ready state before
processing in-coming commands. For Arbitrated Loop or
Point-to- Point (not switch), FW Ready state can take a while.
FW will transition to ready state after all Nports have been
logged in. In the mean time, certain initiators have completed
the login and starts IO. Driver needs to start processing all
queues if FW is already started.

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Change scsi host lookup method.

For target mode, when new scsi command arrive, driver first performs
a look up of the SCSI Host. The current look up method is based on
the ALPA portion of the NPort ID. For Cisco switch, the ALPA can
not be used as the index. Instead, the new search method is based
on the full value of the Nport_ID via btree lib.

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Add DebugFS node to display Port Database

Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Use IOCB interface to submit non-critical MBX.

The Mailbox interface is currently over subscribed. We like
to reserve the Mailbox interface for the chip managment and
link initialization. Any non essential Mailbox command will
be routed through the IOCB interface. The IOCB interface is
able to absorb more commands.

Following commands are being routed through IOCB interface

- Get ID List (007Ch)
- Get Port DB (0064h)
- Get Link Priv Stats (006Dh)

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Add async new target notification

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Export DIF stats via debugfs

Signed-off-by: Anil Gurumurthy <anil.gurumurthy@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Improve T10-DIF/PI handling in driver.

Add routines to support T10 DIF tag.

Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Anil Gurumurthy <anil.gurumurthy@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Allow relogin to proceed if remote login did not finish

If the remote port have started the login process, then the
PLOGI and PRLI should be back to back. Driver will allow
the remote port to complete the process. For the case where
the remote port decide to back off from sending PRLI, this
local port sets an expiration timer for the PRLI. Once the
expiration time passes, the relogin retry logic is allowed
to go through and perform login with the remote port.

Signed-off-by: Quinn Tran <quinn.tran@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Fix sess_lock & hardware_lock lock order problem.

The main lock that needs to be held for CMD or TMR submission
to upper layer is the sess_lock. The sess_lock is used to
serialize cmd submission and session deletion. The addition
of hardware_lock being held is not necessary. This patch removes
hardware_lock dependency from CMD/TMR submission.

Use hardware_lock only for error response in this case.

Path1
       CPU0                    CPU1
       ----                    ----
  lock(&(&ha->tgt.sess_lock)->rlock);
                               lock(&(&ha->hardware_lock)->rlock);
                               lock(&(&ha->tgt.sess_lock)->rlock);
  lock(&(&ha->hardware_lock)->rlock);

Path2/deadlock
*** DEADLOCK ***
Call Trace:
dump_stack+0x85/0xc2
print_circular_bug+0x1e3/0x250
__lock_acquire+0x1425/0x1620
lock_acquire+0xbf/0x210
_raw_spin_lock_irqsave+0x53/0x70
qlt_sess_work_fn+0x21d/0x480 [qla2xxx]
process_one_work+0x1f4/0x6e0

Cc: <stable@vger.kernel.org>
Cc: Bart Van Assche <Bart.VanAssche@sandisk.com>
Reported-by: Bart Van Assche <Bart.VanAssche@sandisk.com>
Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Fix inadequate lock protection for ABTS.

Normally, ABTS is sent to Target Core as Task MGMT command.
In the case of error, qla2xxx needs to send response, hardware_lock
is required to prevent request queue corruption.

Cc: <stable@vger.kernel.org>
Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Fix request queue corruption.

When FW notify driver or driver detects low FW resource,
driver tries to send out Busy SCSI Status to tell Initiator
side to back off. During the send process, the lock was not held.

Cc: <stable@vger.kernel.org>
Signed-off-by: Quinn Tran <quinn.tran@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Fix memory leak for abts processing

Cc: <stable@vger.kernel.org>
Signed-off-by: Quinn Tran <quinn.tran@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

qla2xxx: Allow vref count to timeout on vport delete.

Cc: <stable@vger.kernel.org>
Signed-off-by: Joe Carnuccio <joe.carnuccio@cavium.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

tcmu: Convert cmd_time_out into backend device attribute

Instead of putting cmd_time_out under ../target/core/user_0/foo/control,
which has historically been used by parameters needed for initial
backend device configuration, go ahead and move cmd_time_out into
a backend device attribute.

In order to do this, tcmu_module_init() has been updated to create
a local struct configfs_attribute **tcmu_attrs, that is based upon
the existing passthrough_attrib_attrs along with the new cmd_time_out
attribute.  Once **tcm_attrs has been setup, go ahead and point
it at tcmu_ops->tb_dev_attrib_attrs so it's picked up by target-core.

Also following MNC's previous change, ->cmd_time_out is stored in
milliseconds but exposed via configfs in seconds.  Also, note this
patch restricts the modification of ->cmd_time_out to before +
after the TCMU device has been configured, but not while it has
active fabric exports.

Cc: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

tcmu: make cmd timeout configurable

A single daemon could implement multiple types of devices
using multuple types of real devices that may not support
restarting from crashes and/or handling tcmu timeouts. This
makes the cmd timeout configurable, so handlers that do not
support it can turn if off for now.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

tcmu: add helper to check if dev was configured

This adds a helper to check if the dev was configured. It
will be used in the next patch to prevent updates to some
config settings after the device has been setup.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

Merge tag 'openrisc-for-linus' of git://github.com/openrisc/linux

Pull OpenRISC fixes from Stafford Horne:
 "OpenRISC fixes for build issues that were exposed by kbuild robots
  after 4.11 merge. All from allmodconfig builds. This includes:

   - bug in the handling of 8-byte get_user() calls

   - module build failure due to multile missing symbol exports"

* tag 'openrisc-for-linus' of git://github.com/openrisc/linux:
  openrisc: Export symbols needed by modules
  openrisc: fix issue handling 8 byte get_user calls
  openrisc: xchg: fix `computed is not used` warning

target: fix race during implicit transition work flushes

This fixes the following races:

1. core_alua_do_transition_tg_pt could have read
tg_pt_gp_alua_access_state and gone into this if chunk:

if (!explicit &&
        atomic_read(&tg_pt_gp->tg_pt_gp_alua_access_state) ==
           ALUA_ACCESS_STATE_TRANSITION) {

and then core_alua_do_transition_tg_pt_work could update the
state. core_alua_do_transition_tg_pt would then only set
tg_pt_gp_alua_pending_state and the tg_pt_gp_alua_access_state would
not get updated with the second calls state.

2. core_alua_do_transition_tg_pt could be setting
tg_pt_gp_transition_complete while the tg_pt_gp_transition_work
is already completing. core_alua_do_transition_tg_pt then waits on the
completion that will never be called.

To handle these issues, we just call flush_work which will return when
core_alua_do_transition_tg_pt_work has completed so there is no need
to do the complete/wait. And, if core_alua_do_transition_tg_pt_work
was running, instead of trying to sneak in the state change, we just
schedule up another core_alua_do_transition_tg_pt_work call.

Note that this does not handle a possible race where there are multiple
threads call core_alua_do_transition_tg_pt at the same time. I think
we need a mutex in target_tg_pt_gp_alua_access_state_store.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

target: allow userspace to set state to transitioning

Userspace target_core_user handlers like tcmu-runner may want to set the
ALUA state to transitioning while it does implicit transitions. This
patch allows that state when set from configfs.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

target: fix ALUA transition timeout handling

The implicit transition time tells initiators the min time
to wait before timing out a transition. We currently schedule
the transition to occur in tg_pt_gp_implicit_trans_secs
seconds so there is no room for delays. If
core_alua_do_transition_tg_pt_work->core_alua_update_tpg_primary_metadata
needs to write out info to a remote file, then the initiator can
easily time out the operation.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>

target: Use system workqueue for ALUA transitions

If tcmu-runner is processing a STPG and needs to change the kernel's
ALUA state then we cannot use the same work queue for task management
requests and ALUA transitions, because we could deadlock. The problem
occurs when a STPG times out before tcmu-runner is able to
call into target_tg_pt_gp_alua_access_state_store->
core_alua_do_port_transition -> core_alua_do_transition_tg_pt ->
queue_work. In this case, the tmr is on the work queue waiting for
the STPG to complete, but the STPG transition is now queued behind
the waiting tmr.

Note:
This bug will also be fixed by this patch:
http://www.spinics.net/lists/target-devel/msg14560.html
which switches the tmr code to use the system workqueues.

For both, I am not sure if we need a dedicated workqueue since
it is not a performance path and I do not think we need WQ_MEM_RECLAIM
to make forward progress to free up memory like the block layer does.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>