drm/i915: Introduce GEM proxy

GEM proxy is a kind of GEM, whose backing physical memory is pinned
and produced by guest VM and is used by host as read only. With GEM
proxy, host is able to access guest physical memory through GEM object
interface. As GEM proxy is such a special kind of GEM, a new flag
I915_GEM_OBJECT_IS_PROXY is introduced to ban host from changing the
backing storage of GEM proxy.

v3:
- update "Reviewed-by". (Joonas)

v2:
- return -ENXIO when pin and map pages of GEM proxy to kernel space.
  (Chris)

Here are the histories of this patch in "Dma-buf support for Gvt-g"
patch-set:

v14:
- return -ENXIO when gem proxy object is banned by ioctl.
  (Chris) (Daniel)

v13:
- add comments to GEM proxy. (Chris)
- don't ban GEM proxy in i915_gem_sw_finish_ioctl. (Chris)
- check GEM proxy bar after finishing i915_gem_object_wait. (Chris)
- remove GEM proxy bar in i915_gem_madvise_ioctl.

v6:
- add gem proxy barrier in the following ioctls. (Chris)
  i915_gem_set_caching_ioctl
  i915_gem_set_domain_ioctl
  i915_gem_sw_finish_ioctl
  i915_gem_set_tiling_ioctl
  i915_gem_madvise_ioctl

Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-2-git-send-email-tina.zhang@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-2-chris@chris-wilson.co.uk

drm/i915: Object w/o backing storage is banned by -ENXIO

-ENXIO should be returned when operations are banned from changing
backing storage of objects without backing storage.

v4:
- update "Reviewed-by". (Joonas)

v3:
- separate this patch from "Introduce GEM proxy" patch-set. (Joonas)

v2:
- update the patch description and subject to just mention objects w/o
  backing storage, instead of "GEM proxy". (Joonas)

Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-1-git-send-email-tina.zhang@intel.com
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-1-chris@chris-wilson.co.uk

drm/i915: Fix function name in comment

Commit 78597996370c (drm/i915/bxt: Fix PPS lost state after suspend
breaking eDP link training) renamed the function to
intel_power_sequencer_reset() but forgot to update comment.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114004638.5186-1-lucas.demarchi@intel.com

drm/i915: There is only one fault register from GEN8 onwards

Until Haswell/Baytrail, the hardware used to have a per engine fault
register (e.g. 0x4094 - render fault register, 0x4194 - media fault
register and so on). But since Broadwell, all these registers were
combined into a singe one and the engine id stored in bits 14:12.

Not only we should not been reading (and writing to) registers that do
not exist, in platforms with VCS2 (SKL), the address that would belong
this engine (0x4494, VCS2_HW = 4) is already assigned to other register.

v2: use less controversial function names (Chris).
v3: make non-exported functions static, remove now obsolete check for
engine presence before posting_read (Chris).

References: IHD-OS-BDW-Vol 2c-11.15, page 75.
References: IHD-OS-SKL-Vol 2c-05.16, page 350.
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171113173628.11689-1-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Clear per-engine fault register as early as possible

From gen6, the hardware tracks address lookup failures and we should
clear those registers upon startup to prevent false positives. However,
this was happening before we have the engines defined (intel_uncore_init())
and the for_each_engine loop was just a nop. The earliest we can call
this is inside intel_engines_init_mmio().

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171111004448.12360-1-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: expose command stream timestamp frequency to userspace

We use to have this fixed per generation, but starting with CNL userspace
cannot tell just off the PCI ID. Let's make this information available. This
is particularly useful for performance monitoring where much of the
normalization work is done using those timestamps (this include pipeline
statistics in both GL & Vulkan as well as OA reports).

v2: Use variables for 24MHz/19.2MHz values (Ewelina)
    Renamed function & coding style (Sagar)

v3: Fix frequency read on Broadwell (Sagar)
    Fix missing divide by 4 on <= gen4 (Sagar)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-7-lionel.g.landwerlin@intel.com

drm/i915/perf: enable perf support on CNL

This adds new registers to the whitelist to configs emitted from userspace.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-6-lionel.g.landwerlin@intel.com

drm/i915: fix register naming

This name was added with the whitelisting of registers for building up OA
configs. It is contained in a range gen8 whitelist :

   addr >= RPM_CONFIG0.reg && addr <= NOA_CONFIG(8).reg

Hence why the name isn't used anywhere.

v2: Fix register name again RPC->RCP (Matthew)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-5-lionel.g.landwerlin@intel.com

drm/i915/perf: refactor perf setup

Gen8/9 aren't very different and we can merge some of this code.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-4-lionel.g.landwerlin@intel.com

drm/i915/perf: add support for Coffeelake GT3

We can enable GT3 as well as GT2.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-3-lionel.g.landwerlin@intel.com

drm/i915/perf: complete whitelisting for OA programming on HSW

We were missing some registers and also can name one for which we only had
the offset.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-2-lionel.g.landwerlin@intel.com

drm/i915: Handle locking better in i915_sink_crc.

Lock the bare minimum, instead of the entire world, and
use interruptible locking because we can.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-6-maarten.lankhorst@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Handle adjust better in intel_pipe_config_compare

Some parameters use CHECK_BOOL, but should really use
CHECK_BOOL_INCOMPLETE. We cannot currently check whether
the inherited infoframes and audio are set up correctly.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-4-maarten.lankhorst@linux.intel.com
[mlankhorst: Add danvet's comment about why this is needed.]
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Check boolean options in intel_pipe_config_compare with its own macro

Add PIPE_CONF_CHECK_BOOL for boolean options, which are printed with yesno.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-3-maarten.lankhorst@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove bogus ips_enabled check.

The flag just tells us IPS can be enabled, if the primary plane
is not enabled it means IPS might not be. This never triggered
in CI because we don't have a haswell ULT there, but can be
reproduced easily with kms_atomic_transitions.plane-all-modeset-transition

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-2-maarten.lankhorst@linux.intel.com
[mlankhorst: Remove from haswell_get_pipe_config too. (danvet)]
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove Gen9 WAs with no effect

GEN8_CONFIG0 (0xD00) is a protected by a lock (bit 31) which is set by
the BIOS, so there is no way we can enable the three chicken bits
mandated by the WA (the BIOS should be doing it instead).

v2: Rebased
v3: Standalone patch

References: b033bb6d5d3a ("drm/i915/gen9: Enable must set chicken bits in config0 reg")
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1510185589-9100-2-git-send-email-oscar.mateo@intel.com
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Remove redundant intel_autoenable_gt_powersave()

Now that we always execute a context switch upon module load, there is
no need to queue a delayed task for doing so. The purpose of the delayed
task is to enable GT powersaving, for which we need the HW state to be
valid (i.e. having loaded a context and initialised basic state). We
used to defer this operation as historically it was slow (due to slow
register polling, fixed with commit 1758b90e38f5 ("drm/i915: Use a hybrid
scheme for fast register waits")) but now we have a requirement to save
the default HW state.

v2: Load the kernel context (to provide the power context) upon resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171112112738.1463-3-chris@chris-wilson.co.uk

drm/i915: Update watermark state correctly in sanitize_watermarks

We no longer use intel_crtc->wm.active for watermarks any more,
which was incorrect. But this uncovered a bug in sanitize_watermarks(),
which meant that we wrote the correct watermarks, but the next
update would still use the wrong hw watermarks for calculating.
This caused all further updates to fail with -EINVAL and the
log would reveal an error like the one below:

[   10.043902] [drm:ilk_validate_wm_level.part.8 [i915]] Sprite WM0 too large 56 (max 0)
[   10.043960] [drm:ilk_validate_pipe_wm [i915]] LP0 watermark invalid
[   10.044030] [drm:intel_crtc_atomic_check [i915]] No valid intermediate pipe watermarks are possible

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: b6b178a77210 ("drm/i915: Calculate ironlake intermediate watermarks correctly, v2.")
Cc: stable@vger.kernel.org #v4.8+
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-1-maarten.lankhorst@linux.intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/selftests: Yet another forgotten mock_i915->mm initialiser

Move all of the i915->mm initialisation to a private function that can
be reused by the mock i915 device to save forgetting any more steps.

For example,
<7>[ 1542.046332] [IGT] drv_selftest: starting subtest mock_objects
<4>[ 1542.123924] Setting dangerous option mock_selftests - tainting kernel
<6>[ 1542.167941] i915: Performing mock selftests with st_random_seed=0x246f5ab5 st_timeout=1000
<4>[ 1542.178012] INFO: trying to register non-static key.
<4>[ 1542.178027] the code is fine but needs lockdep annotation.
<4>[ 1542.178032] turning off the locking correctness validator.
<4>[ 1542.178041] CPU: 3 PID: 6008 Comm: kworker/3:7 Tainted: G     U          4.14.0-rc8-CI-CI_DRM_3332+ #1
<4>[ 1542.178049] Hardware name:                  /NUC6CAYB, BIOS AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017
<4>[ 1542.178144] Workqueue: events __i915_gem_free_work [i915]
<4>[ 1542.178152] Call Trace:
<4>[ 1542.178163]  dump_stack+0x68/0x9f
<4>[ 1542.178170]  register_lock_class+0x3fd/0x580
<4>[ 1542.178177]  ? unwind_next_frame+0x14/0x20
<4>[ 1542.178184]  ? __save_stack_trace+0x73/0xd0
<4>[ 1542.178191]  __lock_acquire+0xa4/0x1b00
<4>[ 1542.178254]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178261]  ? __lock_acquire+0x4ab/0x1b00
<4>[ 1542.178268]  lock_acquire+0xb0/0x200
<4>[ 1542.178273]  ? lock_acquire+0xb0/0x200
<4>[ 1542.178336]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178344]  _raw_spin_lock+0x32/0x50
<4>[ 1542.178405]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178468]  __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178476]  process_one_work+0x221/0x650
<4>[ 1542.178483]  worker_thread+0x4e/0x3c0
<4>[ 1542.178489]  kthread+0x114/0x150
<4>[ 1542.178494]  ? process_one_work+0x650/0x650
<4>[ 1542.178499]  ? kthread_create_on_node+0x40/0x40
<4>[ 1542.178506]  ret_from_fork+0x27/0x40

v2: Fish out i915->mm.object_stat_lock which was being inited over in
i915_drv.c (Matthew)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110232447.21618-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Kerneldoc typo s/rps/rps_client/

Update the kerneldoc parameter name to match the real parameter name.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109140644.10805-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Stop caching the "golden" renderstate

As we now record the default HW state and so only emit the "golden"
renderstate once to prepare the HW, there is no advantage in keeping the
renderstate batch around as it will never be used again.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-8-chris@chris-wilson.co.uk

drm/i915: Record the default hw state after reset upon load

Take a copy of the HW state after a reset upon module loading by
executing a context switch from a blank context to the kernel context,
thus saving the default hw state over the blank context image.
We can then use the default hw state to initialise any future context,
ensuring that each starts with the default view of hw state.

v2: Unmap our default state from the GTT after stealing it from the
context. This should stop us from accidentally overwriting it via the
GTT (and frees up some precious GTT space).

Testcase: igt/gem_ctx_isolation
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-7-chris@chris-wilson.co.uk

drm/i915: Mark the context state as dirty/written

In the next few patches, we will want to both copy out of the context
image and write a valid image into a new context. To be completely safe,
we should then couple in our domain tracking to ensure that we don't
have any issues with stale data remaining in unwanted cachelines.

Historically, we omitted the .write=true from the call to set-gtt-domain
in i915_switch_context() in order to avoid a stall between every request
as we would want to wait for the previous context write from the gpu.
Since then, we limit the set-gtt-domain to only occur when we first bind
the vma, so once in use we will never stall, and we are sure to flush
the context following a load from swap.

Equally we never applied the lessons learnt from ringbuffer submission
to execlists; so time to apply the flush of the lrc after load as well.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-6-chris@chris-wilson.co.uk

drm/i915: Inline intel_modeset_gem_init()

intel_modeset_gem_init() now only sets up the legacy overlay, so let's
remove the function and call the setup directly during driver load. This
should help us find a better point in the initialisation sequence for it
later.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-5-chris@chris-wilson.co.uk

drm/i915: Move intel_init_clock_gating() to i915_gem_init()

Despite its name intel_init_clock_gating applies both display clock gating
workarounds; GT mmio workarounds and the occasional GT power context
workaround. Worse, sometimes it includes a context register workaround
which we need to apply before we record the default HW state for all
contexts.

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>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-4-chris@chris-wilson.co.uk

drm/i915: Move GT powersaving init to i915_gem_init()

GT powersaving is tightly coupled to the request infrastructure. To
avoid complications with the order of initialisation in the next patch
(where we want to send requests to hw during GEM init) move the
powersaving initialisation into the purview of i915_gem_init().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-3-chris@chris-wilson.co.uk

drm/i915: Force the switch to the i915->kernel_context

In the next few patches, we will have a hard requirement that we emit a
context-switch to the perma-pinned i915->kernel_context (so that we can
save the HW state using that context-switch). As the first context
itself may be classed as a kernel context, we want to be explicit in our
comparison. For an extra-layer of finesse, we can check the last
unretired context on the engine; as well as the last retired context
when idle.

v2: verbose verbosity
v3: Always force the switch, even when the engine is idle, and update
the assert that this happens before suspend.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v1
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-2-chris@chris-wilson.co.uk

drm/i915: Define an engine class enum for the uABI

We want to be able to report back to userspace details about an engine's
class, and in return for userspace to be able to request actions
regarding certain classes of engines. To isolate the uABI from any
variations between hw generations, we define an abstract class for the
engines and internally map onto the hw.

v2: Remove MAX from the uABI; keep it internal if we need it, but don't
let userspace make the mistake of using it themselves.
v3: s/OTHER/INVALID/
  The use of OTHER is ill-defined, so remove it from the uABI as any
  future new type of engine can define a class to suit it. But keep a
  reserved value for an invalid class, so that we can always
  unambiguously express when something doesn't belong to the
  classification.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v2
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-1-chris@chris-wilson.co.uk

drm/i915/selftests: Initialise mock_i915->mm.obj_lock

lockdep spotted that the mock tests were using the i915->mm.obj_lock
without first initialiasing it:

>[ 1303.217043] [IGT] drv_selftest: starting subtest mock_objects
<4>[ 1303.240898] Setting dangerous option mock_selftests - tainting kernel
<6>[ 1303.253665] i915: Performing mock selftests with st_random_seed=0xd87ea6c6 st_timeout=1000
<4>[ 1303.254812] INFO: trying to register non-static key.
<4>[ 1303.254816] the code is fine but needs lockdep annotation.
<4>[ 1303.254818] turning off the locking correctness validator.
<4>[ 1303.254820] CPU: 4 PID: 13112 Comm: drv_selftest Tainted: G     U  W       4.14.0-rc8-CI-Patchwork_7058+ #1
<4>[ 1303.254823] Hardware name: MSI MS-7924/Z97M-G43(MS-7924), BIOS V1.12 02/15/2016
<4>[ 1303.254825] Call Trace:
<4>[ 1303.254829]  dump_stack+0x68/0x9f
<4>[ 1303.254832]  register_lock_class+0x3fd/0x580
<4>[ 1303.254835]  ? ___slab_alloc.constprop.29+0x157/0x3d0
<4>[ 1303.254837]  ? ___slab_alloc.constprop.29+0x157/0x3d0
<4>[ 1303.254840]  ? sg_kmalloc+0x1e/0x50
<4>[ 1303.254842]  ? debug_smp_processor_id+0x17/0x20
<4>[ 1303.254845]  __lock_acquire+0xa4/0x1b00
<4>[ 1303.254884]  ? __i915_gem_object_set_pages+0x116/0x1f0 [i915]
<4>[ 1303.254887]  ? __this_cpu_preempt_check+0x13/0x20
<4>[ 1303.254889]  ? sg_kmalloc+0x1e/0x50
<4>[ 1303.254891]  lock_acquire+0xb0/0x200
<4>[ 1303.254893]  ? lock_acquire+0xb0/0x200
<4>[ 1303.254917]  ? __i915_gem_object_set_pages+0x116/0x1f0 [i915]
<4>[ 1303.254920]  _raw_spin_lock+0x32/0x50
<4>[ 1303.254944]  ? __i915_gem_object_set_pages+0x116/0x1f0 [i915]
<4>[ 1303.254967]  __i915_gem_object_set_pages+0x116/0x1f0 [i915]
<4>[ 1303.254991]  i915_gem_object_get_pages_phys+0x286/0x2b0 [i915]
<4>[ 1303.255015]  ____i915_gem_object_get_pages+0x20/0x60 [i915]
<4>[ 1303.255039]  i915_gem_object_attach_phys+0x137/0x1a0 [i915]
<4>[ 1303.255063]  igt_phys_object+0x45/0x120 [i915]
<4>[ 1303.255094]  __i915_subtests+0x40/0xd0 [i915]
<4>[ 1303.255099]  ? work_on_cpu_safe+0x60/0x60
<4>[ 1303.255131]  i915_gem_object_mock_selftests+0x34/0x50 [i915]

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110151919.18451-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Restore the wait for idle engine after flushing interrupts

So it appears that commit 5427f207852d ("drm/i915: Bump wait-times for
the final CS interrupt before parking") was a little over optimistic in
its belief that it had successfully waited for all residual activity on
the engines before parking. Numerous sightings in CI since then of

<7>[   52.542886] [IGT] core_auth: executing
<3>[   52.561013] [drm:intel_engines_park [i915]] *ERROR* vcs0 is not idle before parking
<7>[   52.561215] intel_engines_park vcs0
<7>[   52.561229] intel_engines_park  current seqno 98, last 98, hangcheck 0 [-247449 ms], inflight 0
<7>[   52.561238] intel_engines_park  Reset count: 0
<7>[   52.561266] intel_engines_park  Requests:
<7>[   52.561363] intel_engines_park  RING_START: 0x00000000 [0x00000000]
<7>[   52.561377] intel_engines_park  RING_HEAD:  0x00000000 [0x00000000]
<7>[   52.561390] intel_engines_park  RING_TAIL:  0x00000000 [0x00000000]
<7>[   52.561406] intel_engines_park  RING_CTL:   0x00000000
<7>[   52.561422] intel_engines_park  RING_MODE:  0x00000200 [idle]
<7>[   52.561442] intel_engines_park  ACTHD:  0x00000000_00000000
<7>[   52.561459] intel_engines_park  BBADDR: 0x00000000_00000000
<7>[   52.561474] intel_engines_park  Execlist status: 0x00000301 00000000
<7>[   52.561489] intel_engines_park  Execlist CSB read 5 [5 cached], write 5 [5 from hws], interrupt posted? no
<7>[   52.561500] intel_engines_park  ELSP[0] idle
<7>[   52.561510] intel_engines_park  ELSP[1] idle
<7>[   52.561519] intel_engines_park  HW active? 0x0
<7>[   52.561608] intel_engines_park Idle? yes
<7>[   52.561617] intel_engines_park

on Braswell, which indicates that the engine just needs that little bit
longer after flushing the tasklet to settle. So give it a few more
milliseconds before declaring an err and applying the emergency brake.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103479
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110112550.28909-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915: Acquire PUNIT->PMIC bus for intel_uncore_forcewake_reset()

intel_uncore_forcewake_reset() does forcewake puts and gets as such
we need to make sure that no-one tries to access the PUNIT->PMIC bus
(on systems where this bus is shared) while it runs, otherwise bad
things happen.

Normally this is taken care of by the i915_pmic_bus_access_notifier()
which does an intel_uncore_forcewake_get(FORCEWAKE_ALL) when some other
driver tries to access the PMIC bus, so that later forcewake gets are
no-ops (for the duration of the bus access).

But intel_uncore_forcewake_reset gets called in 3 cases:
1) Before registering the pmic_bus_access_notifier
2) After unregistering the pmic_bus_access_notifier
3) To reset forcewake state on a GPU reset

In all 3 cases the i915_pmic_bus_access_notifier() protection is
insufficient.

This commit fixes this race by calling iosf_mbi_punit_acquire() before
calling intel_uncore_forcewake_reset(). In the case where it is called
directly after unregistering the pmic_bus_access_notifier, we need to
hold the punit-lock over both calls to avoid a race where
intel_uncore_fw_release_timer() may execute between the 2 calls.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171019111620.26761-3-hdegoede@redhat.com

x86/platform/intel/iosf_mbi: Add unlocked PMIC bus access notifier unregister

For race free unregistration drivers may need to acquire PMIC bus access
through iosf_mbi_punit_acquire() and then (un)register the notifier without
dropping the lock.

This commit adds an unlocked variant of
iosf_mbi_unregister_pmic_bus_access_notifier for this use case.

Acked-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171019111620.26761-2-hdegoede@redhat.com

drm/i915: Move irqs enabled assertion deeper for mock breadcrumbs

In order to allow the mock breadcrumbs tests to run without device irqs
being enabled, move the intel_irqs_enabled() assert deeper to just
before we commit to enabling the HW irq.

v2: Add a FIXME explaining that placing the assertion so deep is not
ideal, but a compromise for mock breadcrumbs.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107102003.1802-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/selftests: Reduce the volume of the timeout message

Originally it was anticipated that timeouts would be a rare event, and
so merit a warning that the test was incomplete. However, for igt we
keep the timeout low, and hitting the timeout is intentional. It no
longer necessitates a warning, but to be expected.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110101110.12042-1-chris@chris-wilson.co.uk
Reviwed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Update DRIVER_DATE to 20171109

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Update DRIVER_DATE to 20171109

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Mark up i915_vma_unbind() as a potential sleeper

Whenever we want to unbind a vma, we must wait on all GPU activity to
complete first. (This is what gives us the ability to do fine grained
eviction and purging by only having to wait on the VMA that we need to
unbind to proceed; though if pushed we can make it a rule that we are
only allowed to unbind already idle VMA and move the burden of the work
and organising the sleep onto the caller.) Currently, we might only
sleep if the vma is still active on the GPU, but in principle
i915_vma_unbind() always implies a sleep, so mark it up with a
might_sleep().

Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=103638
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109213450.13875-2-chris@chris-wilson.co.uk

drm/i915: Mark vm_free_page() as a potential sleeper agent

vm_free_page() may call down into set_pages_array_wb() (which itself
sleeps, on x86 at least) but only if on !llc and the caches overflow.
Since this is unlikely, we only rarely trigger the error in practice,
and so to make CI detection of this sleeping bug possible we want to
mark the common vm_free_page() as a potential sleep.

Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=103638
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109213450.13875-1-chris@chris-wilson.co.uk

drm/i915: Use trace_printk to provide a death rattle for GEM

Trying to enable printk debugging for GEM is fraught with the issue of
spam; interactions with HW are very frequent and often boring. However,
one instance where they are not so boring is just before a BUG; here
ftrace provides a facility to dump its ringbuffer on an oops. So for CI
let's enable trace_printk() to capture the last exchanges with HW as a
death rattle.

For example,
[   79.234110] ------------[ cut here ]------------
[   79.234137] kernel BUG at drivers/gpu/drm/i915/intel_lrc.c:907!
[   79.234145] invalid opcode: 0000 [#1] SMP
[   79.234153] Dumping ftrace buffer:
[   79.234158] ---------------------------------
...
[   79.314044] gem_conc-1059    1..s1 79203443us : intel_lrc_irq_handler: bcs0 out[0]: ctx=5.2, seqno=145
[   79.314089] gem_conc-1059    1..s. 79220800us : intel_lrc_irq_handler: bcs0 csb[1/1]: status=0x00000018:0x00000005
[   79.314133] gem_conc-1059    1..s. 79220803us : intel_lrc_irq_handler: bcs0 out[0]: ctx=5.1, seqno=145
[   79.314177] gem_conc-1062    2..s1 79230458us : intel_lrc_irq_handler: bcs0 in[0]:  ctx=8.1, seqno=146
[   79.314220] gem_conc-1062    2..s1 79230515us : intel_lrc_irq_handler: bcs0 in[0]:  ctx=8.2, seqno=147
[   79.314265] gem_conc-1059    1..s1 79230951us : intel_lrc_irq_handler: bcs0 csb[2/3]: status=0x00000012:0x00000008
[   79.314309] gem_conc-1059    1..s1 79230954us : intel_lrc_irq_handler: bcs0 out[0]: ctx=8.2, seqno=147
[   79.314353] gem_conc-1059    1..s1 79230954us : intel_lrc_irq_handler: bcs0 csb[3/3]: status=0x00008002:0x00000008
[   79.314396] gem_conc-1059    1..s1 79230955us : intel_lrc_irq_handler: bcs0 out[0]: ctx=8.1, seqno=147
[   79.314402] ---------------------------------

v2: Tweak the formatting to be more consistent between in/out.
v3: do {} while (0) stub macro protection

Suggested-by: Michał Winiarski <michal.winiarski@intel.com>
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: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109143019.16568-1-chris@chris-wilson.co.uk

drm/i915: Clean up DP code local variables and calling conventions

Eliminate a ton of pointless 'dev' variables in the DP code, and pass
around 'dev_priv' instead of 'dev'.

v2: Rebase

Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109152758.32257-1-ville.syrjala@linux.intel.com

drm/i915: Clean up PPS code calling conventions

No need to pass 'dev' or 'dev_priv' when the function already takes
'intel_dp'. Also let's prefer passing 'dev_priv' instead of 'dev'
when we have to pass one or the other.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-10-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Nuke intel_digital_port->port

Remove intel_digital_port->port and replace its users with
intel_encoder->port. intel_encoder->port is a superset of
intel_digital_port->port, and it works correctly even for
MST encoders.

v2: Eliminate a few dp_to_dig_port()->base.port cases too (DK)

Performed with cocci:
@@
@@
struct intel_digital_port {
       ...
-       enum port port;
       ...
}

@@
struct intel_digital_port *D;
expression E;
@@
- D->port = E;

@@
struct intel_digital_port *D;
@@
- D->port
+ D->base.port

@
expression E;
@@
(
- dp_to_dig_port(E)->port
+ dp_to_dig_port(E)->base.port
|
- enc_to_dig_port(E)->port
+ to_intel_encoder(E)->port
)

@@
expression E;
@@
- to_intel_encoder(&E->base)
+ E

@@
struct intel_digital_port *D;
identifier I, M;
@@
  I = &D->base
<...
(
- D->base.M
+ I->M
|
- &D->base
+ I
)
...>

@@
identifier D;
expression E;
identifier M;
@@
 D = enc_to_dig_port(&E->base)
<...
(
- D->base.M
+ E->M
|
- &D->base
+ E
)
...>

@@
identifier D, DP;
expression E;
identifier M;
@@
 DP = enc_to_intel_dp(&E->base)
<...
(
- dp_to_dig_port(DP)->base.M
+ E->M
|
- &dp_to_dig_port(DP)->base
+ E
)
...>

@@
expression E;
identifier M;
@@
(
- enc_to_dig_port(&E->base)->base.M
+ E->M
|
- enc_to_dig_port(&E->base)->base
+ E
|
- enc_to_mst(&E->base)->primary->base.port
+ E->port
)

@@
expression E;
identifier D;
@@
- struct intel_digital_port *D = E;
... when != D

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109152434.32074-1-ville.syrjala@linux.intel.com

drm/i915: Replace dig_port->port with encoder port for BXT DPLL selection

Replace dig_port->port with encoder->port in the BXT DPLL selection.
We can do this because both the master encoder and the fake MST encoders
have the same encoder->port value, whereas using dig_port->port only
worked for the master encoder since the fake encoders were't derived
from intel_digital_port. This eliminates the DP MST special case.

Do this by hand because spatch is having problems with the control
flow due to the dig_port assignment happening in two different
branches.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-8-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Eliminate crtc->config usage from CRT code

Replace crtc->config usage with the passed down crtc state.

Also take the opportunity for some s/pipe_config/crtc_state/ while at it.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-7-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Pass crtc state to DPIO PHY functions

Rather than digging through encoder->crtc and crtc->config in the
DPIO PHY functions, pass down the correct crtc state from the caller.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-6-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Eliminate some encoder->crtc usage from TV code

Extract the current crtc from the crtc state rather than via
the legacy encoder->crtc pointer whenever possible.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-5-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Eliminate some encoder->crtc usage from SDVO code

Extract the current crtc from the crtc state rather than via
the legacy encoder->crtc pointer whenever possible.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-4-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Eliminate some encoder->crtc usage from DSI code

Extract the current crtc from the crtc state rather than via
the legacy encoder->crtc pointer whenever possible.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-3-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Eliminate some encoder->crtc usage from DP code

Extract the current crtc from the crtc state rather than via
the legacy encoder->crtc pointer whenever possible.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-2-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU

When we close the VMA, we unbind it from the ppgtt and tear down the
page directory pointing at it. That may trigger us to return WC pages
back to the system, requiring conversion back to WB which itself may
sleep. That makes i915_vma_close() unsuitable for use inside the RCU
read lock, which we need to hold to iterate the radixtree.

The fix is quite simple, we can close all the VMA as we close the ppgtt,
we only need to do that instead of closing them during destruction of
the LUT.

v2: Order between closing the LUT and the ppgtt is important; we use the
vma inside the LUT as a means of retrieving the object, and so we must
clear the LUT before freeing the VMA when closing the ppgtt.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103638
Fixes: 547da76b5777 ("drm/i915: Hold rcu_read_lock when iterating over the radixtree (vma idr)")
Fixes: d1b48c1e7184 ("drm/i915: Replace execbuf vma ht with an idr")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109085540.32264-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915/guc: Simplify programming of GUC_SHIM_CONTROL

We can program GUC_SHIM_CONTROL register with all expected
bits without use of extra macro defined in fwif.h

v2: rebased without pre-prod code
v3: fixed typo

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-4-michal.wajdeczko@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Drop legacy workarounds from guc_prepare_xfer

We don't keep the workarounds for pre-production hardware
(see intel_detect_preproduction_hw) thus we can drop some
extra steps during firmware upload needed only for unsupported
platforms.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-3-michal.wajdeczko@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Wait for ucode DMA transfer completion

We silently assumed that DMA transfer will be completed
within assumed timeout and thus we were waiting only at
last step for GuC to become ready. Add intermediate wait
to catch unexpected delays in DMA transfer.

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-2-michal.wajdeczko@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Split GuC firmware xfer function into clear steps

Transfer of GuC firmware requires few steps that currently
are implemented in two large functions. Split this code into
smaller functions to make these steps small and clear.
Also be prepared for potential DMA xfer step failure.

v2: rename function steps (Sagar)

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-1-michal.wajdeczko@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Implement ReadHitWriteOnlyDisable.

The workaround for this is described as:

"if RenderSurfaceState.Num_Multisamples > 1, disable RCC clock gating if
RenderSurfaceState.Num_Multisamples == 1, set 0x7010[14] = 1"

Further documentation in the internal bug referenced by the bspec
suggest that any of the above suggestions should suffice to fix the
issue. We are going with disabling RCC clock gating.

Unfortunately, what we are doing doesn't match the name of the
workaround, but at least it matches its description.

This change improves CNL stability by avoiding some of the hangs seen in
the platform.

v2: Only disable RCC clock gating.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103183027.5051-1-rafael.antognolli@intel.com

drm/i915: Move init_clock_gating() back to where it was

Apparently setting up a bunch of GT registers before we've properly
initialized the rest of the GT hardware leads to these setting being
lost. So looks like I broke HSW with commit b7048ea12fbb ("drm/i915:
Do .init_clock_gating() earlier to avoid it clobbering watermarks")
by doing init_clock_gating() too early. This should actually affect
other platforms as well, but apparently not to such a great degree.

What I was ultimately after in that commit was to move the
ilk_init_lp_watermarks() call earlier. So let's undo the damage and
move init_clock_gating() back to where it was, and call
ilk_init_lp_watermarks() just before the watermark state readout.

This highlights how fragile and messed up our init order really is.
I wonder why we even initialize the display before gem. The opposite
order would make much more sense to me...

v2: Keep WaRsPkgCStateDisplayPMReq:hsw early as it really must
    be done before all planes might get disabled.

Cc: stable@vger.kernel.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mark Janes <mark.a.janes@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reported-by: Mark Janes <mark.a.janes@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103549
Fixes: b7048ea12fbb ("drm/i915: Do .init_clock_gating() earlier to avoid it clobbering watermarks")
References: https://lists.freedesktop.org/archives/intel-gfx/2017-November/145432.html
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171108133555.14091-1-ville.syrjala@linux.intel.com
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Prune the reservation shared fence array

The shared fence array is not autopruning and may continue to grow as an
object is shared between new timelines. Take the opportunity when we
think the object is idle (we have to confirm that any external fence is
also signaled) to decouple all the fences.

We apply a similar trick after waiting on an object, see commit
e54ca9774777 ("drm/i915: Remove completed fences after a wait")

v2: No longer need to handle the batch pool as a special case.
v3: Need to trylock from within i915_vma_retire as this may be called
form the shrinker - and we may later try to allocate underneath the
reservation lock, so a deadlock is possible.

References: https://bugs.freedesktop.org/show_bug.cgi?id=102936
Fixes: d07f0e59b2c7 ("drm/i915: Move GEM activity tracking into a common struct reservation_object")
Fixes: 80b204bce8f2 ("drm/i915: Enable multiple timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107220656.5020-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Idle the GPU before shinking everything

The handling of contexts are peculiar. Instead of tieing their vma to
activity, we pin the context. This means that we cannot simply unbind
the context object itself at will (which would normally cause us to wait
for the vma to be idle), but must manually idle the GPU and retire
requests first.

A consequence of this peculiarity is when doing a last desperate attempt
to recover memory. If the memory is tied up inside active context
objects, we will fail to recover any memory simply by trying to unbind
the objects without first doing a wait-for-idle.

A side-effect of removing the call to shrinker_lock_uninterruptible()
from i915_gem_shrinker_oom() was that we removed an unlocked
wait-for-idle, and so lost the "natural" shrinkage of context objects.
By replacing that with a locked wait from inside i915_gem_shrink(), we
not only replace it with the ability to recover all context objects, but
do so for all i915_gem_shrink_all() callers.

v2: Switching requires request allocation, which is not permitted from
inside the shrinker as it only uses ordinary allocations.

References: https://bugs.freedesktop.org/show_bug.cgi?id=102936
Fixes: f2123818ffad ("drm/i915: Move dev_priv->mm.[un]bound_list to its own lock")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171108094400.1386-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Include intel_engine_is_idle() status in engine pretty-printer

Upon parking, if we discover an active engine we dump its state. Follow
that state with an indication of whether the engine was idle.

References: https://bugs.freedesktop.org/show_bug.cgi?id=103479
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107152211.19930-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Read ilk FDI PLL frequency once during initialisation

During intel_atomic_check(), we do not take the intel_runtime_pm_get()
wakeref and so should do the atomic modeset precalculations without
referring to the HW. However, on Ironlake we see

<7>[   23.487557] [drm:intel_atomic_check [i915]] [CONNECTOR:47:VGA-1] checking for sink bpp constrains
<7>[   23.487615] [drm:intel_atomic_check [i915]] clamping display bpp (was 36) to default limit of 24
<4>[   23.487621] RPM wakelock ref not held during HW access
<4>[   23.487652] ------------[ cut here ]------------
<4>[   23.487697] WARNING: CPU: 0 PID: 1343 at drivers/gpu/drm/i915/intel_drv.h:1813 gen5_read32+0x183/0x200 [i915]
<4>[   23.487701] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic i915 intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul snd_hda_intel ghash_clmulni_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm lpc_ich e1000e mei_me ptp mei pps_core prime_numbers
<4>[   23.487784] CPU: 0 PID: 1343 Comm: debugfs_test Tainted: G        W       4.14.0-rc7-CI-Trybot_1378+ #1
<4>[   23.487788] Hardware name: Hewlett-Packard HP Compaq 8100 Elite SFF PC/304Ah, BIOS 786H1 v01.13 07/14/2011
<4>[   23.487793] task: ffff8801f90aa6c0 task.stack: ffffc900013ec000
<4>[   23.487838] RIP: 0010:gen5_read32+0x183/0x200 [i915]
<4>[   23.487842] RSP: 0018:ffffc900013efb58 EFLAGS: 00010292
<4>[   23.487849] RAX: 000000000000002a RBX: ffff880205c00000 RCX: 0000000000000006
<4>[   23.487854] RDX: 000000000000140a RSI: ffffffff81d0eb14 RDI: ffffffff81cc26f6
<4>[   23.487857] RBP: ffffc900013efb80 R08: ffff8801f90aaff8 R09: 0000000000000000
<4>[   23.487861] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
<4>[   23.487865] R13: 0000000000046000 R14: ffff88020ffaba78 R15: ffff88020b109bf8
<4>[   23.487870] FS:  00007f53b5e40a40(0000) GS:ffff88021bc00000(0000) knlGS:0000000000000000
<4>[   23.487874] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
<4>[   23.487878] CR2: 000055e41900c0e8 CR3: 00000001fa0d6005 CR4: 00000000000206f0
<4>[   23.487882] Call Trace:
<4>[   23.487931]  intel_atomic_check+0x745/0x1290 [i915]
<4>[   23.487948]  drm_atomic_check_only+0x459/0x560
<4>[   23.487956]  ? drm_atomic_set_crtc_for_connector+0xc9/0x100
<4>[   23.488025]  drm_atomic_commit+0x18/0x50
<4>[   23.488035]  restore_fbdev_mode_atomic+0x190/0x1f0
<4>[   23.488059]  restore_fbdev_mode+0x32/0x120
<4>[   23.488072]  drm_fb_helper_restore_fbdev_mode_unlocked+0x50/0xa0
<4>[   23.488139]  intel_fbdev_restore_mode+0x34/0x90 [i915]
<4>[   23.488194]  i915_driver_lastclose+0xe/0x10 [i915]
<4>[   23.488208]  drm_lastclose+0x39/0xf0
<4>[   23.488219]  drm_release+0x30c/0x3c0
<4>[   23.488236]  __fput+0xb9/0x200
<4>[   23.488252]  ____fput+0xe/0x10
<4>[   23.488264]  task_work_run+0x89/0xc0
<4>[   23.488278]  exit_to_usermode_loop+0x83/0x90
<4>[   23.488290]  syscall_return_slowpath+0xd0/0x110
<4>[   23.488304]  entry_SYSCALL_64_fastpath+0xaf/0xb1
<4>[   23.488312] RIP: 0033:0x7f53b4317560
<4>[   23.488320] RSP: 002b:00007ffca7e70748 EFLAGS: 00000246 ORIG_RAX: 0000000000000003
<4>[   23.488333] RAX: 0000000000000000 RBX: 0000000000000001 RCX: 00007f53b4317560
<4>[   23.488340] RDX: 0000000000000005 RSI: 00007ffca7e70640 RDI: 0000000000000004
<4>[   23.488347] RBP: 000055e417783900 R08: 000055e418f9e290 R09: 0000000000000000
<4>[   23.488356] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000001
<4>[   23.488363] R13: 00007f53b4302c40 R14: 0000000000000000 R15: 0000000000000000
<4>[   23.488384] Code: b5 f2 f2 e0 0f ff e9 c5 fe ff ff 80 3d 0e 4b 10 00 00 0f 85 c6 fe ff ff 48 c7 c7 30 73 29 a0 c6 05 fa 4a 10 00 01 e8 8e f2 f2 e0 <0f> ff e9 ac fe ff ff e8 51 9d f3 e0 85 c0 0f 85 01 ff ff ff 48
<4>[   23.488780] ---[ end trace 6bc72ce7f1596190 ]---
<7>[   23.488844] [drm:intel_atomic_check [i915]] checking fdi config on pipe A, lanes 1
<7>[   23.488911] [drm:intel_atomic_check [i915]] hw max bpp: 36, pipe bpp: 24, dithering: 0

due to intel_fdi_link_freq() poking at FDI_PLL_BIOS_0. Avoid this by
recording the fdi pll frequency during device initiailisation.

v2: Also extract the static FDI PLL frequencies for Sandybridge and
Ivybridge.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107214713.18704-1-chris@chris-wilson.co.uk
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/selftests: Take rpm wakeref around partial tiling tests

Since the partial tiling tests are poking into the GGTT to watch the
fence registers in operation, it itself needs the device rpm wakeref in
order for the GGTT to remain accessible.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107115653.10716-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915/selftests: Take rpm wakeref around GGTT lowlevel tests

The vma routines are responsible for acquiring the device rpm wakeref
before they poke the HW. However, some of the selftests bypass the
higher level vma routines in order to poke directly at the lowlevel GGTT
functions; these are then responsible for managing rpm themselves.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107114051.10583-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915/selftests: Skip mixed page exhaustion if only small pages available

If we only have 4k pages, we can't mix together different combinations
of hugepages to see if the world burns. However, as the loops did
nothing, we never set err to 0 and reported ENODEV aborting the test.
Teach the test to skip instead.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107110559.6098-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>

drm/i915: Prevent unbounded wm results in g4x_compute_wm()

Smatch warns of

drivers/gpu/drm/i915/intel_pm.c:1161 g4x_compute_wm() warn: signedness bug returning '(-33554430)'

which is a result of it believing that wm may be INT_MAX following
g4x_tlb_miss_wa(). Just declaring g4x_tlb_miss_wa() as returning an
unsigned integer is not sufficient, we need to tell smatch that wm itself
is unsigned for it to not worry. So mark up the locals we expect to be
non-negative, and so silence smatch.

v2: Mark up vlv_compute_wm_level() as unsigned similarly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20171107140338.13748-1-chris@chris-wilson.co.uk

drm/i915: Simplify onion for bxt_ddi_phy_init()

Older compilers (gcc-4.9) are not as able to track uninitialised
variables as well as more recent compilers. In particular,

drivers/gpu/drm/i915/intel_dpio_phy.c: In function ‘bxt_ddi_phy_init’:
drivers/gpu/drm/i915/intel_dpio_phy.c:482:25: warning: ‘was_enabled’ may be used uninitialized in this function [-Wmaybe-uninitialized]

In this case, we can rearrange code slightly to make the control flow
clearer to the reader, as well as the compiler. That is we only call
uninit using the same predicate as calling init

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107135324.28300-1-chris@chris-wilson.co.uk
Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>

drm/i915: Silence compiler for csr_load_work_fn()

gcc-4.7 is not very smart and can not tell that "si" is guarded by size
being 0. So it complains,

drivers/gpu/drm/i915/intel_csr.c: In function ‘csr_load_work_fn’:
drivers/gpu/drm/i915/intel_csr.c:204:3: warning: ‘si’ may be used uninitialized in this function [-Wmaybe-uninitialized]
drivers/gpu/drm/i915/intel_csr.c:190:30: note: ‘si’ was declared in

Give in and mark si as NULL.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107145334.27154-1-chris@chris-wilson.co.uk
Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/i915: Silence smatch for cmdparser

drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue
drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue

If we move the shift into each case not only do we kill the warning from
smatch, but we shrink the code slightly:

   text    data     bss     dec     hex filename
1267906   20587    3168 1291661  13b58d before
1267890   20587    3168 1291645  13b57d after

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107154055.19460-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>

drm/i915: Deconstruct struct sgt_dma initialiser

gcc-4.4 complains about:

struct sgt_dma iter = {
.sg = vma->pages->sgl,
.dma = sg_dma_address(iter.sg),
.max = iter.dma + iter.sg->length,
};

drivers/gpu/drm/i915/i915_gem_gtt.c: In function ‘gen8_ppgtt_insert_4lvl’:
drivers/gpu/drm/i915/i915_gem_gtt.c:938: error: ‘iter.sg’ is used uninitialized in this function
drivers/gpu/drm/i915/i915_gem_gtt.c:939: error: ‘iter.dma’ is used uninitialized in this function

and worse generates invalid code that triggers a GPF:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915]
PGD 0
Oops: 0000 [#1] SMP
Modules linked in: snd_aloop nf_conntrack_ipv6 nf_defrag_ipv6 nf_log_ipv6 ip6table_filter ip6_tables ctr ccm xt_state nf_log_ipv4
nf_log_common xt_LOG xt_limit xt_recent xt_owner xt_addrtype iptable_filter ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat
nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack libcrc32c ip_tables dm_mod vhost_net macvtap macvlan vhost tun kvm_intel kvm
irqbypass uas usb_storage hid_multitouch btusb btrtl uvcvideo videobuf2_v4l2 videobuf2_core videodev media videobuf2_vmalloc videobuf2_memops
sg ppdev dell_wmi sparse_keymap mei_wdt sd_mod iTCO_wdt iTCO_vendor_support rtsx_pci_ms memstick rtsx_pci_sdmmc mmc_core dell_smm_hwmon hwmon
dell_laptop dell_smbios dcdbas joydev input_leds hci_uart btintel btqca btbcm bluetooth parport_pc parport i2c_hid
  intel_lpss_acpi intel_lpss pcspkr wmi int3400_thermal acpi_thermal_rel dell_rbtn mei_me mei snd_hda_codec_hdmi snd_hda_codec_realtek
snd_hda_codec_generic ahci libahci acpi_pad xhci_pci xhci_hcd snd_hda_intel snd_hda_codec snd_hda_core snd_hwdep snd_seq snd_seq_device
snd_pcm snd_timer snd soundcore int3403_thermal arc4 e1000e ptp pps_core i2c_i801 iwlmvm mac80211 rtsx_pci iwlwifi cfg80211 rfkill
intel_pch_thermal processor_thermal_device int340x_thermal_zone intel_soc_dts_iosf i915 video fjes
CPU: 2 PID: 2408 Comm: X Not tainted 4.10.0-rc5+ #1
Hardware name: Dell Inc. Latitude E7470/0T6HHJ, BIOS 1.11.3 11/09/2016
task: ffff880219fe4740 task.stack: ffffc90005f98000
RIP: 0010:gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915]
RSP: 0018:ffffc90005f9b8c8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8802167d8000 RCX: 0000000000000001
RDX: 00000000ffff7000 RSI: ffff880219f94140 RDI: ffff880228444000
RBP: ffffc90005f9b948 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000080
R13: 0000000000000001 R14: ffffc90005f9bcd7 R15: ffff88020c9a83c0
FS:  00007fb53e1ee920(0000) GS:ffff88024dd00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000010 CR3: 000000022ef95000 CR4: 00000000003406e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
  ppgtt_bind_vma+0x40/0x50 [i915]
  i915_vma_bind+0xcb/0x1c0 [i915]
  __i915_vma_do_pin+0x6e/0xd0 [i915]
  i915_gem_execbuffer_reserve_vma+0x162/0x1d0 [i915]
  i915_gem_execbuffer_reserve+0x4fc/0x510 [i915]
  ? __kmalloc+0x134/0x250
  ? i915_gem_wait_for_error+0x25/0x100 [i915]
  ? i915_gem_wait_for_error+0x25/0x100 [i915]
  i915_gem_do_execbuffer+0x2df/0xa00 [i915]
  ? drm_malloc_gfp.clone.0+0x42/0x80 [i915]
  ? path_put+0x22/0x30
  ? __check_object_size+0x62/0x1f0
  ? terminate_walk+0x44/0x90
  i915_gem_execbuffer2+0x95/0x1e0 [i915]
  drm_ioctl+0x243/0x490
  ? handle_pte_fault+0x1d7/0x220
  ? i915_gem_do_execbuffer+0xa00/0xa00 [i915]
  ? handle_mm_fault+0x10d/0x2a0
  vfs_ioctl+0x18/0x30
  do_vfs_ioctl+0x14b/0x3f0
  SyS_ioctl+0x92/0xa0
  entry_SYSCALL_64_fastpath+0x1a/0xa9
RIP: 0033:0x7fb53b4fcb77
RSP: 002b:00007ffe0c572898 EFLAGS: 00003246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00007fb53e17c038 RCX: 00007fb53b4fcb77
RDX: 00007ffe0c572900 RSI: 0000000040406469 RDI: 000000000000000b
RBP: 00007fb5376d67e0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000028 R11: 0000000000003246 R12: 0000000000000000
R13: 0000000000000000 R14: 000055eecb314d00 R15: 000055eecb315460
Code: 0f 84 5d ff ff ff eb a2 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 53 48 83 ec 58 0f 1f 44 00 00 31 c0 89 4d b0 <4c>
8b 60 10 44 8b 70 0c 48 89 d0 4c 8b 2e 48 c1 e8 27 25 ff 01
RIP: gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915] RSP: ffffc90005f9b8c8
CR2: 0000000000000010

Recent gccs, such as 4.9, 6.3 or 7.2, do not generate the warning nor do
they explode on use. If we manually create the struct using locals from
the stack, this should eliminate this issue, and does not alter code
generation with gcc-7.2.

Fixes: 894ccebee2b0 ("drm/i915: Micro-optimise gen8_ppgtt_insert_entries()")
Reported-by: Kelly French <kfrench@federalhill.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Kelly French <kfrench@federalhill.net>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171106211128.12538-1-chris@chris-wilson.co.uk
Tested-by: Kelly French <kfrench@federalhill.net>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Handle error-state modparams in dedicated functions

Capturing and cleanup and modparams in error state requires
some macro tricks. Move that code into separated functions
for easier maintenance.

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-3-michal.wajdeczko@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Make GuC log part of the uC error state

We keep details of GuC and HuC in separate error state struct.
Make GuC log part of it to group all related data together.
Since we are printing uC details at the end, with this change
GuC log will be moved there too.

v2: comment on new placement of the log (Chris)

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-2-michal.wajdeczko@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Add Guc/HuC firmware details to error state

Include GuC and HuC firmware details in captured error state
to provide additional debug information. To reuse existing
uc firmware pretty printer, introduce new drm-printer variant
that works with our i915_error_state_buf output. Also update
uc firmware pretty printer to accept const input.

v2: don't rely on current caps (Chris)
    dump correct fw info (Michal)
v3: simplify capture of custom paths (Chris)
v4: improve 'why' comment (Joonas)
    trim output if no fw path (Michal)
    group code around uc error state (Michal)
v5: use error in cleanup_uc (Michal)

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-1-michal.wajdeczko@intel.com
[ickle: allow printing uc_fw after allocation failure]
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Assert ctch->vma is allocated

Silence smatch by demonstrating that ctch->vma is allocated
following a successful chch_init()

drivers/gpu/drm/i915/intel_guc_ct.c:204 ctch_open() error:
 we previously assumed 'ctch->vma' could be null (see line 197)

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171106135154.52520-1-michal.wajdeczko@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Assert guc->stage_desc_pool is allocated

Silence smatch by demonstrating that guc->stage_desc_pool is allocated
following a successful guc_stage_desc_pool_create(),

drivers/gpu/drm/i915/i915_guc_submission.c:1293 i915_guc_submission_init() error: we previously assumed 'guc->stage_desc_pool' could be null (see line 1261)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171106114833.31199-1-chris@chris-wilson.co.uk
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/i915: Lock llist_del_first() vs llist_del_all()

An oversight in commit 87701b4b5593 ("drm/i915: Only free the oldest
stale object before a fresh allocation") was that not only do we have to
serialise concurrent users of llist_del_first(), but we also have to
lock llist_del_first() vs llist_del_all().

From llist.h,

 * This can be summarized as follows:
 *
 *           |   add    | del_first |  del_all
 * add       |    -     |     -     |     -
 * del_first |          |     L     |     L
 * del_all   |          |           |     -
 *
 * Where, a particular row's operation can happen concurrently with a column's
 * operation, with "-" being no lock needed, while "L" being lock is needed.

This should hopefully explain:

<4>[   89.287106] general protection fault: 0000 [#1] PREEMPT SMP
<4>[   89.287126] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic x86_pkg_temp_thermal intel_powerclamp coretemp i915 crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core r8169 mii mei_me mei snd_pcm prime_numbers i2c_hid pinctrl_geminilake pinctrl_intel
<4>[   89.287226] CPU: 2 PID: 23 Comm: ksoftirqd/2 Tainted: G     U          4.14.0-rc8-CI-CI_DRM_3315+ #1
<4>[   89.287247] Hardware name: Intel Corp. Geminilake/GLK RVP2 LP4SD (07), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017
<4>[   89.287270] task: ffff88017ab34ec0 task.stack: ffffc90000128000
<4>[   89.287290] RIP: 0010:llist_add_batch+0x4/0x20
<4>[   89.287301] RSP: 0018:ffffc9000012bdb8 EFLAGS: 00010296
<4>[   89.287314] RAX: ffffffff811017ad RBX: 6e468801a1560000 RCX: ef3e53fceecdeb81
<4>[   89.287330] RDX: 6e468801a1566130 RSI: ffff880103d73d98 RDI: ffff880103d73d98
<4>[   89.287346] RBP: ffffc9000012bdb8 R08: ffff88017ab35780 R09: 0000000000000000
<4>[   89.287361] R10: ffffc9000012bd68 R11: 00000000abb18c3d R12: ffffffffa01369e0
<4>[   89.287377] R13: ffff88017fd1b8f8 R14: ffff88017ab34ec0 R15: 000000000000000a
<4>[   89.287393] FS:  0000000000000000(0000) GS:ffff88017fd00000(0000) knlGS:0000000000000000
<4>[   89.287411] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
<4>[   89.287424] CR2: 00007ff0c0755018 CR3: 000000016df9b000 CR4: 00000000003406e0
<4>[   89.287440] Call Trace:
<4>[   89.287511]  __i915_gem_free_object_rcu+0x20/0x40 [i915]
<4>[   89.287527]  rcu_process_callbacks+0x27a/0x730
<4>[   89.287544]  __do_softirq+0xc0/0x4ae
<4>[   89.287559]  ? smpboot_thread_fn+0x2d/0x280
<4>[   89.287571]  run_ksoftirqd+0x1f/0x70
<4>[   89.287582]  smpboot_thread_fn+0x18a/0x280
<4>[   89.287595]  kthread+0x114/0x150
<4>[   89.287605]  ? sort_range+0x30/0x30
<4>[   89.287615]  ? kthread_create_on_node+0x40/0x40
<4>[   89.287628]  ret_from_fork+0x27/0x40
<4>[   89.287641] Code: 0d 48 83 ea 01 4c 89 c1 48 83 fa ff 74 12 48 23 0c d7 74 ed 48 c1 e2 06 48 0f bd c9 48 8d 04 0a 5d c3 90 90 90 90 90 55 48 89 e5 <48> 8b 0a 48 89 0e 48 89 c8 f0 48 0f b1 3a 48 39 c1 75 ed 48 85
<1>[   89.287774] RIP: llist_add_batch+0x4/0x20 RSP: ffffc9000012bdb8
<4>[   89.287826] ---[ end trace e775d15174d8ae02 ]---

(Lockless lists are only easy (and lockless) when only using
llist_add/llist_del_all!)

Fixes: 87701b4b5593 ("drm/i915: Only free the oldest stale object before
a fresh allocation")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171106111508.11941-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915/selftests: Hide dangerous tests

Some tests are designed to exercise the limits of the HW and may trigger
unintended side-effects making the machine unusable. This should not be
executed by default, but are still useful for early platform validation.

References: https://bugs.freedesktop.org/show_bug.cgi?id=103453
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171025153207.9589-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Assert vma->flags are updated correctly during binding

As we bind, and unbind on error, we want to be sure that the vma->flags
are updated to reflect the binding state so that on the next invocation
all is well.

v2: Take two.
v3: Take three; vma-misplaced is checking map-and-fenceable so keep it
last!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171105124550.32715-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>

drm/i915: Set up mocs tables before restarting the engines

After a reset, we may immediately begin executing requests on restarting
the engines. Ergo this has to be last step with all re-initialisation
completed beforehand. The mocs setup was after we started executing the
requests; do it earlier!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171102131430.22328-1-chris@chris-wilson.co.uk

drm/i915: Warn in debug builds of incorrect usages of ptr_pack_bits

GEM_BUG_ON if the packed bits do not fit into the specified width.

v2: Avoid using the macro argument twice.
v3: Drop unnecessary braces. (Joonas)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103090538.14474-1-tvrtko.ursulin@linux.intel.com

drm/i915: Reject unknown syncobj flags

We have to reject unknown flags for uAPI considerations, and also
because the curent implementation limits their i915 storage space
to two bits.

v2: (Chris Wilson)
 * Fix fail in ABI check.
 * Added unknown flags and BUILD_BUG_ON.

v3:
 * Use ARCH_KMALLOC_MINALIGN instead of alignof. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: cf6e7bac6357 ("drm/i915: Add support for drm syncobjs")
Cc: Jason Ekstrand <jason@jlekstrand.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: David Airlie <airlied@linux.ie>
Cc: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031102326.9738-1-tvrtko.ursulin@linux.intel.com

drm/i915: ensure oa config uuid is null terminated

Because dev_priv is 0-ed it's not currently an issue, but since we
have dev_priv->perf.oa.test_config.uuid size at uuid + 1, we could
just copy the null character.

v2: Use strlcpy instead of strncpy (Chris)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171102121827.436-1-lionel.g.landwerlin@intel.com

drm/i915: Flush the irq and tasklets before asserting engine is idle

Before we assert that the engine is idle, make sure we flush any
residual tasklet. After that point, if the engine is not idle, more work
may be queued despite us trying to park the engine and go to sleep.

References: https://bugs.freedesktop.org/show_bug.cgi?id=103479
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171101202149.32493-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915: Use fallback forcewake if primary ack missing

There is a possibility on gen9 hardware to miss the forcewake ack
message. The recommended workaround is to use another free
bit and toggle it until original bit is successfully acknowledged.

Some future gen9 revs might or might not fix the underlying issue but
using fallback forcewake bit dance can be considered as harmless:
without the ack timeout we never reach the fallback bit forcewake.
Thus as of now we adopt a blanket approach for all gen9 and leave
the bypassing the fallback bit approach for future patches if
corresponding hw revisions do appear.

Commit 83e3337204b2 ("drm/i915: Increase maximum polling time to 50ms
for forcewake request/clear ack") did increase the forcewake timeout.
If the issue was a delayed ack, future work could include finding
a suitable timeout value both for primary ack and reserve toggle
to reduce the worst case latency.

v2: use bit 15, naming, comment (Chris), only wait fallback ack
v3: fix return on fallback, backoff after fallback write (Chris)
v4: udelay on first pass, grammar (Chris)
v4: s/reserve/fallback

References: HSDES #1604254524
References: https://bugs.freedesktop.org/show_bug.cgi?id=102051
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171102094836.2506-1-mika.kuoppala@linux.intel.com

drm/i915/guc: Add support for reset engine using GuC commands

This patch adds per engine reset and recovery (TDR) support when GuC is
used to submit workloads to GPU.

In the case of i915 directly submission to ELSP, driver manages hang
detection, recovery and resubmission. With GuC submission these tasks
are shared between driver and GuC. i915 is still responsible for detecting
a hang, and when it does it only requests GuC to reset that Engine. GuC
internally manages acquiring forcewake and idling the engine before
resetting it.

Once the reset is successful, i915 takes over again and handles the
resubmission. The scheduler in i915 knows which requests are pending so
after resetting a engine, pending workloads/requests are resubmitted
again.

v2: s/i915_guc_request_engine_reset/i915_guc_reset_engine/ to match the
non-guc function names.

v3: Removed debug message about engine restarting from which request,
since the new baseline do it regardless of submission mode. (Chris)

v4: Rebase.

v5: Do not pass unnecessary reporting flags to the fw (Jeff);
tasklet_schedule(&execlists->irq_tasklet) handles the resubmit; rebase.

v6: Rename the existing reset engine function and share a similar
interface between guc and non-guc paths (Chris).

Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031225309.10888-1-michel.thierry@intel.com
Reviewed-by: Jeff McGee <jeff.mcgee@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Rename the function that resets the GuC

intel_guc_reset sounds more like the microcontroller is the one performing
a reset, while in this case is the opposite. intel_reset_guc not only
makes it clearer, it follows the other intel_reset functions available.

v2: Print error message in English (Tvrtko).

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171030185616.32836-2-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/guc: Clear terminated attribute bit on GuC preemption context

If GuC firmware performs an engine reset while that engine had a
preemption pending, it will set the terminated attribute bit on our
preemption stage descriptor. GuC firmware retains all pending work
items for a high-priority GuC client, unlike the normal-priority GuC
client where work items are dropped. It wants to make sure the preempt-
to-idle work doesn't run when scheduling resumes, and uses this bit to
inform its scheduler and presumably us as well. Our job is to clear it
for the next preemption after reset, otherwise that and future
preemptions will never complete. We'll just clear it every time.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171101221630.25086-1-jeff.mcgee@intel.com
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/cnl: Symmetric scalers for each pipe

For Cannonlake the number of scalers for each pipe is 2. Let's increase
the number of scalers for pipe C.

v2: Use INTEL_GEN() instead of IS_CANNONLAKE()

Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1509530930-24960-1-git-send-email-mika.kahola@intel.com

drm/i915: Give more details for the active-when-parking warning for the engines

If the we think the engine is still active when we attempt to park it,
we want more details -- so dump the engine state.

References: https://bugs.freedesktop.org/show_bug.cgi?id=103479
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027110617.31745-4-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915: Move parking-while-active warning to intel_engines_park()

We will want to break this down to give detailed per-engine warnings as
to why we still think we are active as we attempt to park the engines.
For the first step, just move the warning verbatim from the idle-worker
to intel_engines_park().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027110617.31745-3-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915: Check that the breadcrumb wasn't disarmed automatically before parking

We will disarm the breadcrumb interrupt if we see a user interrupt
whilst no one is waiting. This may race with the call to
intel_engine_disarm_breadcrumbs() triggering an assert that we aren't
trying to do the same job twice. Prevent this by checking that the irq
is still armed after flushing the interrupt (for the irq spinlock).

Fixes: bcbd5c33a342 ("drm/i915/guc: Always enable the breadcrumbs irq")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031122235.1395-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Check incoming alignment for unfenced buffers (on i915gm)

In case the object has changed tiling between calls to execbuf, we need
to check if the existing offset inside the GTT matches the new tiling
constraint. We even need to do this for "unfenced" tiled objects, where
the 3D commands use an implied fence and so the object still needs to
match the physical fence restrictions on alignment (only required for
gen2 and early gen3).

In commit 2889caa92321 ("drm/i915: Eliminate lots of iterations over
the execobjects array"), the idea was to remove the second guessing and
only set the NEEDS_MAP flag when required. However, the entire check
for an unusable offset for fencing was removed and not just the
secondary check. I.e.

/* avoid costly ping-pong once a batch bo ended up non-mappable */
        if (entry->flags & __EXEC_OBJECT_NEEDS_MAP &&
            !i915_vma_is_map_and_fenceable(vma))
                return !only_mappable_for_reloc(entry->flags);

was entirely removed as the ping-pong between execbuf passes was fixed,
but its primary purpose in forcing unaligned unfenced access to be
rebound was forgotten.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103502
Fixes: 2889caa92321 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031103607.17836-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/cnl: Remove unnecessary check in cnl_setup_private_ppat

There is no need check if PPGTT is disabled because that not possible
in CNL. Execlists and GuC submission modes rely on at least aliasing
PPGTT and even intel_sanitize_enable_ppgtt says: "We don't allow disabling
PPGTT for gen9+ as it's a requirement for execlists, the sole mechanism
available to submit work."

Suggested-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027223207.7869-1-michel.thierry@intel.com

drm/i915: Remove most encoder->type uses from the audio code

encoder->type isn't genreally safe around DDI ports, so let's
replace some uses in the audio code with the crtc state's
output_types instead.

Actually in these cases encoder->type would work since the DP
SST case is only relevant for VLV/CHV and encoder->type==DP
is a thing on those platforms. The DP MST cases would work
as well since MST encoder->type==DP_MST always. But I think
it's best to try and minimize the encoder->type use in general
to avoid showing a bad example to people.

Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171030184654.17429-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Pass around crtc and connector states for audio

Explicitly pass the crtc and connector states into the audio
code enable/disable hooks, and plumb them all the way down.

This gets rid of almost all crtc->config and encoder->crtc
uses. The one place where we still use them is
i915_audio_component_sync_audio_rate() since that gets called from
the audio driver and we don't have explicit states around then.

Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171030184654.17429-1-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Replace "cc-option -Wno-foo" with "cc-disable-warning foo"

To quote kbuild/makefiles.txt:

    cc-disable-warning checks if gcc supports a given warning and returns
    the commandline switch to disable it. This special function is needed,
    because gcc 4.4 and later accept any unknown -Wno-* option and only
    warn about it if there is another warning in the source file.

This is exactly what we were trying to achieve with cc-option -Wno-foo and
failed miserably.

Reported-by: kbuild-all@01.org
Fixes: 39bf4de89ff7 ("drm/i915: Add -Wall -Wextra to our build, set warnings to full")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171030172927.18158-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Use intel_ddi_get_config() for MST

Eliminate the partially duplicated DDI readout code from MST, and
instead just call intel_ddi_get_config(). As a nice bonus we get
more cross checking as intel_ddi_get_config() will populate
output_types based on the actual mode of the DDI port.

Additonally intel_ddi_get_config() must be changed to get the crtc
from the passed in crtc state rather than from the encoder->crtc link.
encoder->crtc really shouldn't be used anyway.

v2: Rebased on BXT MST latency_optim fix
    Make intel_ddi_clock_get() static

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-7-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Pass a crtc state to ddi post_disable from MST code

Pass an old crtc state to intel_ddi_post_disable() from the MST code.

Note that this crtc state won't necessaitly match the one that was
passed to intel_ddi_pre_enable() if the first stream to be enabled isn't
the last stream to be disabled. But this is fine since the states should
be identical in every important way. This does mean people frobbing
the DDI pre_enable/post_disable hooks have to pay attention in what
parts of the state they consult.

The alternative would be to inline the relevant code into the MST code.
That is actually what we used to do for pre_enable before
commit e081c8463ac9 ("drm/i915: Remove duplicate DDI enabling logic
from MST path"). For post_disable we've always called the DDI hook.

v2: Pimp up the comments explaining the MST issues

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-6-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Eliminate pll->state usage from bxt_calc_pll_link()

We should be using the DPLL hw state we got from the current crtc state
to determine the corresponding port clock frequency rather than getting
it via the current state programmed into the DPLL.

v2: Rebase due to intel_dpll_id changes

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-5-ville.syrjala@linux.intel.com

drm/i915: Nuke intel_ddi_get_encoder_port()

encoder->port works for FDI, and it also works for MST (regardless of
whether we're dealing with the "fake" MST encoder, or mst->primary).
So let's eliminate intel_ddi_get_encoder_port().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-4-ville.syrjala@linux.intel.com

drm/i915: Stop frobbing with DDI encoder->type

Currently the DDI encoder->type will change at runtime depending on
what kind of hotplugs we've processed. That's quite bad since we can't
really trust that that current value of encoder->type actually matches
the type of signal we're trying to drive through it.

Let's eliminate that problem by declaring that non-eDP DDI port will
always have the encoder type as INTEL_OUTPUT_DDI. This means the code
can no longer try to distinguish DP vs. HDMI based on encoder->type.
We'll leave eDP as INTEL_OUTPUT_EDP, since it'll never change and
there's a bunch of code that relies on that value to identify eDP
encoders.

We'll introduce a new encoder .compute_output_type() hook. This allows
us to compute the full output_types before any encoder .compute_config()
hooks get called, thus those hooks can rely on output_types being
correct, which is useful for cloning on oldr platforms. For now we'll
just look at the connector type and pick the correct mode based on that.
In the future the new hook could be used to implement dynamic switching
between LS and PCON modes for LSPCON.

v2: Fix BXT/GLK PPS explosion with DSI/MST encoders
v3: Avoid the PPS warn on pure HDMI/DVI DDI encoders by checking dp.output_reg
v4: Rebase
v5: Populate output_types in .get_config() rather than in the caller
v5: Split out populating output_types in .get_config() (Maarten)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-3-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Populate output_types from .get_config()

Rather than having the caller of .get_config() set output_types based on
encoder->type, let's just have .get_config() itself populate
output_types. This way we are isolated from encoder->type, which won't
be useable for this purpose anyway soon (at least for DDI encoders).

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-2-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>