drm/i915: Drop pipe_enable checks in vblank funcs

With Ville's rework to use drm_crtc_vblank_on/off the core will take
care of rejecting drm_vblank_get calls when the pipe is off. Also the
core won't call the get_vblank_counter hooks in that case either. And
since we've dropped ums support recently we can now remove these
hacks, yay!

Noticed while trying to answer questions Laurent had about how the new
atomic helpers work.

Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/irq: Add drm_crtc_vblank_reset

At driver load we need to tell the vblank code about the state of the
pipes, so that the logic around reject vblank_get when the pipe is off
works correctly.

Thus far i915 used drm_vblank_off, but one of the side-effects of it
is that it also saves the vblank counter. And for that it calls down
into the ->get_vblank_counter hook. Which isn't really a good idea
when the pipe is off for a few reasons:
- With runtime pm the register might not respond.
- If the pipe is off some datastructures might not be around or
  unitialized.

The later is what blew up on gen3: We look at intel_crtc->config to
compute the vblank counter, and for a disabled pipe at boot-up that's
just not there. Thus far this was papered over by a check for
intel_crtc->active, but I want to get rid of that (since it's fairly
race, vblank hooks are called from all kinds of places).

So prep for that by adding a _reset functions which only does what we
really need to be done at driver load: Mark the vblank pipe as off,
but don't do any vblank counter saving or event flushing - neither of
that is required.

v2: Clarify the code flow slightly as suggested by Ville.

v3: Fix kerneldoc spelling, spotted by Laurent.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com> (v2)
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Update DRIVER_DATE to 20150214

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

drm/i915: Remove references to previously removed UMS config option

Commit 03dae59c72ffffd8 ("drm/i915: Ditch UMS config option") removed
CONFIG_DRM_I915_UMS from the Kconfig file, but i915_drv.c still
references this option in two #ifndef statements.

As an undefined config option will always be 'false', we can drop
the #ifndefs alltogether and adapt the printed error message.

This inconsistency was found with the undertaker tool.

Signed-off-by: Andreas Ruprecht <rupran@einserver.de>
[danvet: Undo logging change as requested by Jani.]
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Use a LRI for WaDisableDgMirrorFixInHalfSliceChicken5

I have no idea how that crept in, but we need to do the write from the
ring and this is a masked register. Two fixes in 1!

Cc: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Fix always true comparison in a revision id check

It's always a good idea to keep static analysis happy (also because it
prompts doing the check like I proposed :), this time smatch complains:

drivers/gpu/drm/i915/intel_ringbuffer.c:891 gen9_init_workarounds() warn:
  always true condition '((->dev->pdev->revision) >= (0)) => (0-255 >= 0)'

That's because revision is a u8. Tweak a bit the condition then.

Cc: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaEnableLbsSlaRetryTimerDecrement

This W/A is put in a gen9 specific function because it may well be
needed on other gen9 platforms.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add process identifier to requests

We use the pid of the process which opened our device when
we track which was the culprit of the gpu hang. But as that
file descriptor might get inherited, we might blame the
wrong process when we record the error state.

Track process identifiers in requests to always find
the correct offender.

v2: Track only user processes (Chris)

Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
[danvet: drop NULL check before put_pid as suggested by Chris.]
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaBarrierPerformanceFixDisable

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaCcsTlbPrefetchDisable:skl

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisableChickenBitTSGBarrierAckForFFSliceCS

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisableHDCInvalidation

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisableLSQCROPERFforOCL

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaDisablePartialResolveInVc

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Introduce a SKL specific init_workarounds()

This function will host SKL-only W/As.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Document that we implement WaRsClearFWBitsAtReset

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implement WaSetGAPSunitClckGateDisable

Let's also take the opportunity the remove the comment telling it's a
pre-prod W/A, it should be obvious from the stepping test.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Make the init clock gating function skylake specific

We'll gather cross-gen9 W/A in a separate function later.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Provide a gen9 specific init_render_ring()

WaDisableAsyncFlipPerfMode isn't listed for SKL and
INSTPM_FORCE_ORDERING is MBZ so let's make a gen9 specific render init
function.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Document the WM read latency W/A with its name

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Also detect eDRAM on SKL

Suggested-by: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Detect eDRAM with the enabled bit only

At the moment we compare the whole EDRAM_PRESENT/EDRAMCAP register value
to 1 while EDRAM_PRESENT is only bit 0 (the rest may be used to describe
eDRAM capabilities).

To be more future proof, only look at bit 0 to detect eDRAM presence.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove the IS_SNB_GT1 define

The last (only?) user of this was removed in:

  commit 2208d655a91f9879bd9a39ff9df05dd668b3512c
  Author: Daniel Vetter <daniel.vetter@ffwll.ch>
  Date:   Fri Nov 14 09:25:29 2014 +0100

    drm/i915: drop WaSetupGtModeTdRowDispatch:snb

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_unpin_fb_obj() static

This function is not used outside of intel_display.c since;

  commit cf4c7c12258ed9367f4fc45238f5f50d2db892c1
  Author: Matt Roper <matthew.d.roper@intel.com>
  Date:   Thu Dec 4 10:27:42 2014 -0800

      drm/i915: Make all plane disables use 'update_plane' (v5)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_dp_unpack_aux() static

This was introduced in:

  commit 0bc12bcb1b9686d7011f16410ba17ed0740167c3
  Author: Rodrigo Vivi <rodrigo.vivi@intel.com>
  Date:   Fri Nov 14 08:52:28 2014 -0800

      drm/i915: Introduce intel_psr.c

But the unpack function is unused at this date.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_dp_check_link_status() static

This function is only used in intel_dp.c since:

  commit 0e32b39ceed665bfa4a77a4bc307b6652b991632
  Author: Dave Airlie <airlied@redhat.com>
  Date:   Fri May 2 14:02:48 2014 +1000

      drm/i915: add DP 1.2 MST support (v0.7)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_logical_ring_advance_and_submit() static

This function is only used in intel_lrc.c, so restrict it to that file. The
function was moved around to avoid a forward declaration and group it with its
user.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_lr_context_render_state_init() static

This function is only used in intel_lrc.c, so restrict it to that file. The
function was moved around to avoid a forward declaration and group it with its
user.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove intel_dsi_cmd.h

This header has been unusued since:

  commit 063c86f60ad4064b2cf62041bee8c6389e180b76
  Author: Jani Nikula <jani.nikula@intel.com>
  Date:   Fri Jan 16 14:27:27 2015 +0200

      drm/i915/dsi: remove intel_dsi_cmd.c and the unused functions therein

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make intel_ring_setup_status_page() static

This function is only used in intel_ringbuffer.c, so restrict it to that
file. The function was moved around to avoid a forward declaration and
group it with its user.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Squash in fixup from Wu Fengguang.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Garbage collect orphaned prototypes

There have been quite a bit of development lately, leaving behing lonely
protypes. Time to bid them farewell.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Support alias ppgtt in VM if ppgtt is enabled

The current Intel GVT-g only supports alias ppgtt. And the
emulation is done in the host by first trapping PP_DIR_BASE
mmio accesses. Updating PP_DIR_BASE by using instructions such
as MI_LOAD_REGISTER_IMM are hard to detect and are not supported
in current code. Therefore this patch also adds a new callback
routine - vgpu_mm_switch() to set the PP_DIR_BASE by mmio writes.

v2:
take Chris' comments:
        - move the code into sanitize_enable_ppgtt()
v4:
take Tvrtko's comments:
        - fix the parenthesis alignment warning

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Create vGPU specific MMIO operations to reduce traps

In the virtualized environment, forcewake operations are not
necessary for the driver, because mmio accesses will be trapped
and emulated by the host side, and real forcewake operations are
also done in the host. New mmio access handlers are added to directly
call the __raw_i915_read/write, therefore will reduce many traps and
increase the overall performance for drivers running in the VM with
Intel GVT-g enhancement.

v2:
take Chris' comments:
        - register the mmio hooks in intel_uncore_init()
v3:
take Daniel's comments:
        - use macros to assign mmio write functions for vGPU
v4:
take Tvrtko's comments:
        - also use mmio hooks for read operations

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Kevin Tian <kevin.tian@intel.com>k
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Disable power management for i915 driver in VM

With Intel GVT-g, GPU power management is controlled by
host driver, so there is no need to provide virtualized
GPU PM support. In the future it might be useful to gather
VM input for freq boost, but now let's disable it simply.

v2:
take Chris' comments:
        - do not special case this to gen6+

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add the display switch logic for vGPU in i915 driver

Display switch logic is added to notify the host side that
current vGPU have a valid surface to show. It does so by
writing the display_ready field in PV INFO page, and then
will be handled in the host side. This is useful to avoid
trickiness when the VM's framebuffer is being accessed in
the middle of VM modesetting, e.g. compositing the framebuffer
in the host side.

v2:
        - move the notification code outside the 'else' in load sequence
        - remove the notification code in intel_crtc_set_config()

v4:
        - code rebase, no need to define another dev_priv
        - use #define instead of enum for display readiness

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Zhiyuan Lv <zhiyuan.lv@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Disable framebuffer compression for i915 driver in VM

Framebuffer compression is disabled when driver detects it's
running in a Intel GVT-g enlightened VM, because FBC is not
emulated and there is no stolen memory for a vGPU.

v2:
take Chris' comments:
        - move the code into intel_update_fbc()

v4:
take Tvrtko's comments:
        - rebase the code into intel_fbc_update()

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Zhiyuan Lv <zhiyuan.lv@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Partition the fence registers for vGPU in i915 driver

With Intel GVT-g, the fence registers are partitioned by multiple
vGPU instances in different VMs. Routine i915_gem_load() is modified
to reset the num_fence_regs, when the driver detects it's running in
a VM. Accesses to the fence registers from vGPU will be trapped and
remapped by the host side. And the allocated fence number is provided
in PV INFO page structure. By now, the value of fence number is fixed,
but in the future we can relax this limitation, to allocate the fence
registers dynamically from host side.

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Adds graphic address space ballooning logic

With Intel GVT-g, the global graphic memory space is partitioned by
multiple vGPU instances in different VMs. The ballooning code is called
in i915_gem_setup_global_gtt(), utilizing the drm mm allocator APIs to
mark the graphic address space which are partitioned out to other vGPUs
as reserved. With ballooning, host side does not need to translate a
grahpic address from guest view to host view. By now, current implementation
only support the static ballooning, but in the future, with more cooperation
from guest driver, the same interfaces can be extended to grow/shrink the
guest graphic memory dynamically.

v2:
take Chris and Daniel's comments:
- no guard page between different VMs
- use drm_mm_reserve_node() to do the reservation for ballooning,
instead of the previous drm_mm_insert_node_in_range_generic()

v3:
take Daniel's comments:
- move ballooning functions into i915_vgpu.c
- add kerneldoc to ballooning functions

v4:
take Tvrtko's comments:
- more accurate comments and commit message

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Introduce a PV INFO page structure for Intel GVT-g.

Introduce a PV INFO structure, to facilitate the Intel GVT-g
technology, which is a GPU virtualization solution with mediated
pass-through. This page contains the shared information between
i915 driver and the host emulator. For now, this structure utilizes
an area of 4K bytes on HSW GPU's unused MMIO space. Future hardware
will have the reserved window architecturally defined, and layout
of the page will be added in future BSpec.

The i915 driver load routine detects if it is running in a VM by
reading the contents of this PV INFO page. Thereafter a flag,
vgpu.active is set, and intel_vgpu_active() is used by checking
this flag to conclude if GPU is virtualized with Intel GVT-g. By
now, intel_vgpu_active() will return true, only when the driver
is running as a guest in the Intel GVT-g enhanced environment on
HSW platform.

v2:
take Chris' comments:
        - call the i915_check_vgpu() in intel_uncore_init()
        - sanitize i915_check_vgpu() by adding BUILD_BUG_ON() and debug info
take Daniel's comments:
        - put the definition of PV INFO into a new header - i915_vgt_if.h
other changes:
        - access mmio regs by readq/readw in i915_check_vgpu()

v3:
take Daniel's comments:
        - move the i915/vgt interfaces into a new i915_vgpu.c
        - update makefile
        - add kerneldoc to functions which are non-static
        - add a DOC: section describing some of the high-level design
        - update drm docbook
other changes:
        - rename i915_vgt_if.h to i915_vgpu.h

v4:
take Tvrtko's comments:
        - fix a typo in commit message
        - add debug message when vgt version mismatches
        - rename low_gmadr/high_gmadr to mappable/non-mappable in PV INFO
          structure

Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Introduce bit definitions of CTXT_SR_CTRL register.

This patch introduces 2 bit definitions of context save/restore
control register.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Suggested-by: Dave Gordon <david.s.gordon@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Announce support for framebuffer modifiers

Let the DRM core know we can handle it.

v2: Change to boolean true. (Daniel Vetter)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use fb modifiers in intel_pin_and_fence_fb_obj

And at the same time replace BUG() with a warning and handle it gracefuly.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use fb modifiers in intel_check_cursor_plane

Also drop the mutex since with universal planes there is always a
proper framebuffer around which wraps the underlying bo. Which means
tiling is locked down. This was different in the old code which
directly took gem handles. The looking though was always cargo-cult
since races where not prevented in any way.

v2: Unconditionally enforce untiled, because cursors are always
untiled. The check for physical or gtt cursor is irrelevant. Also
clarify the commit message a bit

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Use fb modifiers for sprites

While at it just outright remove the tiling check in
intel_check_sprite_plane because it's impossible: We only allow
untiled and X-tiled. This essentially reverts

commit 94c6419ed8f45e6682d766672b9ceb54867d3d2d
Author: Damien Lespiau <damien.lespiau@intel.com>
Date:   Mon Oct 29 15:14:51 2012 +0000

    drm/i915: Error out when trying to set a y-tiled as a sprite

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
[danvet: Drop the hunk in check_sprite, it's impossible.]
Cc: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: CS flips are not supported with execlists

And skl only works in execlist mode, not in legacy ring submission.

Therefore remove dead code.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use fb format modifiers in skylake_update_primary_plane

Just a little demo really. We probably need to introduce skl specific
functions for a lot of the format validation stuff, or at least
helpers. Specifically I think intel_framebuffer_init and
intel_fb_align_height must be adjusted to have an i915_ and a skl_
variant. And only shared code should be converted to fb modifiers,
platform code (like the plane config readout can keep on using old
tiling_mode defines to avoid some churn).

Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Switch intel_fb_align_height to fb format modifiers

With this we can treat the fb format modifier completely independently
from the fencing mode in obj->tiling_mode in the initial plane code.
Which means new tiling modes without any gtt fence are now fully
support in the core i915 driver code.

v2: Also add pixel_format while at it, we need this to compute the
height for the new tiling formats.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Set up fb format modifier for initial plane config

No functional changes yet since intel_framebuffer_init would have
fixed this up for us. But this is prep work to be able to handle new
tiling layouts in the initial plane config code.

Follow-up patches will start to make use of this and switch over to fb
modifiers where needed.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Show frame buffer modifier in debug info

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Add fb format modifier support

Currently we don't support anything but X tiled. And for an easier
transition it makes a lot of sense to just keep requiring that X tiled
is properly fenced.

Which means we need to do absolutely nothing in old code to support fb
modifiers, yay!

v2: Fix the Y tiling check, noticed by Tvrtko.

v3: Catch Y-tiled fb for legacy addfb again (Tvrtko) and explain why
we want X tiling to match in the comment.

Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Add tiled framebuffer modifiers

To be used from the new addfb2 extension.

v2:
- Drop Intel-specific untiled modfier.
- Move to drm_fourcc.h.
- Document layouts a bit and denote them as platform-specific and not
  useable for cross-driver sharing.
- Add Y-tiling for completeness.
- Drop special docstring markers to avoid confusing kerneldoc.

v3: Give Y-tiling a unique idea, noticed by Tvrtko.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> (v1)
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bdw: Implement WaForceContextSaveRestoreNonCoherent

v2: Reorder defines (Ben)
v3: More bikesheds, this time re-ordering comments! (Chris)

Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Resolve conflict.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Support not having an init clock gating function defined

When enabling new platforms, we may not have any W/A to apply,
especially that, now, a bunch of them have to be done from the ring.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: change dev_priv->fbc.plane to dev_priv->fbc.crtc

Since the mapping from CRTCs to planes is fixed, looking at the CRTC
is essentially the same as looking at the plane. Also, the next
patches wil start using the frontbuffer_bits macros, and they take the
pipe as the parameter instead of the plane, and this could differ on
gens 2 and 3.

Another nice thing is that we don't risk accidentally initializing
things to PLANE_A if we don't set the value before it is used for the
first time. But this shouldn't be a problem with the current code.

V2: Rebase.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> (v1)
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: don't keep reassigning FBC_UNSUPPORTED

This may save a few picoseconds on !HAS_FBC platforms. And it also
satisfies my OCD.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: don't try to find crtcs for FBC if it's disabled

.. because it would be a waste of time, so move the place where the
check is done. Also, with this we won't risk printing "more than one
pipe active, disabling compression" or "no output, disabling" when FBC
is actually disabled.

This patch also represents a small behavior difference when using
i915.powersave=0: it is now exactly the same as i915.enable_fbc=0 on
this part of the code.

V2: Rebase.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> (v1)
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: gen 9 h/w w/a (WaEnableForceRestoreInCtxtDescForVCS)

Add:
WaEnableForceRestoreInCtxtDescForVCS

v2: Add stepping check.

v3: Fixed stepping check direction. Cleaned up indentation.

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: gen 9 h/w w/a Fix stepping check

Fixed the stepping check on WaDisableDgMirrorFixInHalfSliceChicken5
to be for the correct SOC (Skylake)

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm: add support for tiled/compressed/etc modifier in addfb2

In DRM/KMS we are lacking a good way to deal with tiled/compressed
formats.  Especially in the case of dmabuf/prime buffer sharing, where
we cannot always rely on under-the-hood flags passed to driver specific
gem-create ioctl to pass around these extra flags.

The proposal is to add a per-plane format modifier.  This allows to, if
necessary, use different tiling patters for sub-sampled planes, etc.
The format modifiers are added at the end of the ioctl struct, so for
legacy userspace it will be zero padded.

v1: original
v1.5: increase modifier to 64b

v2: Incorporate review comments from the big thread, plus a few more.

- Add a getcap so that userspace doesn't have to jump through hoops.
- Allow modifiers only when a flag is set. That way drivers know when
  they're dealing with old userspace and need to fish out e.g. tiling
  from other information.
- After rolling out checks for ->modifier to all drivers I've decided
  that this is way too fragile and needs an explicit opt-in flag. So
  do that instead.
- Add a define (just for documentation really) for the "NONE"
  modifier. Imo we don't need to add mask #defines since drivers
  really should only do exact matches against values defined with
  fourcc_mod_code.
- Drop the Samsung tiling modifier on Rob's request since he's not yet
  sure whether that one is accurate.

v3:
- Also add a new ->modifier[] array to struct drm_framebuffer and fill
  it in drm_helper_mode_fill_fb_struct. Requested by Tvrkto Uruslin.
- Remove TODO in comment and add code comment that modifiers should be
  properly documented, requested by Rob.

Cc: Rob Clark <robdclark@gmail.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Daniel Stone <daniel@fooishbar.org>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Michel Dänzer <michel@daenzer.net>
Signed-off-by: Rob Clark <robdclark@gmail.com> (v1.5)
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915/gen8: Un-hardcode number of page directories

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Setup less PPGTT on failed page_directory

The current code will both potentially print a WARN, and setup part of
the PPGTT structure. Neither of these harm the current code, it is
simply for clarity, and to perhaps prevent later bugs, or weird
debug messages.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Rename to GEN8_LEGACY_PDPES

In gen8, 32b PPGTT has always had one "pdp" (it doesn't actually have
one, but it resembles having one). The #define was confusing as is, and
using "PDPE" is a much better description.

sed -i 's/GEN8_LEGACY_PDPS/GEN8_LEGACY_PDPES/' drivers/gpu/drm/i915/*.[ch]

It also matches the x86 pagetable terminology:
PTE  = Page Table Entry - pagetable level 1 page
PDE  = Page Directory Entry - pagetable level 2 page
PDPE = Page Directory Pointer Entry - pagetable level 3 page

And in the near future (for 48b addressing):
PML4E = Page Map Level 4 Entry

v2: Expanded information about Page Directory/Table nomenclature.

Cc: Daniel Vetter <daniel@ffwll.ch>
CC: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2)
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/trace: Fix offsets for 64b

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement WaForceEnableNonCoherent

v2: Don't add WaHdcDisableFetchWhenMasked. Add stepping check for
WaForceEnableNonCoherent

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement Wa4x4STCOptimizationDisable

Move Wa4x4STCOptimizationDisable to gen9_init_workarounds

v2: rebase

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement WaEnableYV12BugFixInHalfSliceChicken7

Move WaEnableYV12BugFixInHalfSliceChicken7 to gen9_init_workarounds

v2: Add stepping check.

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement WaDisableSDEUnitClockGating

v2: Add stepping check for WaDisableSDEUnitClockGating.

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Rebase.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: h/w w/a: syncing dependencies between camera and graphics

This one doesn't have one of these nice cryptic names unfortunately.

v2: Added missing register bitmap

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement WaDisableDgMirrorFixInHalfSliceChicken5

Move WaDisableDgMirrorFixInHalfSliceChicken5 to gen9_init_workarounds

v2: Added stepping check

v3: Removed unused register bitmap

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
[danvet: Bikesheds.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Implement WaDisablePartialInstShootdown

v2: Dont add WaDisableThreadStallDopClockGating as not SKL WA. (Found
by Damien Lespiau)

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Bikeshed commit message a bit as per Damien's suggestions.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: ring w/a gen 9 revision definitions

Add Skylake stepping Revision IDs definitions.

v1: Use existing revision id.

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Use magic __I915__ and bikeshed #defines as suggested by
Damien.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: ring w/a initialisation for gen 9

Add framework for gen 9 HW WAs

v1: Changed SOC specific WA function to gen 9 common function (Req: Damien Lespiau)

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make sure the primary plane is enabled before reading out the fb state

We don't want to end up in a state where we track that the pipe has its
primary plane enabled when primary plane registers are programmed with
values that look possible but the plane actually disabled.

Refuse to read out the fb state when the primary plane isn't enabled.

Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Suggested-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix atomic state when reusing the firmware fb

Right now, we get a warning when taking over the firmware fb:

  [drm:drm_atomic_plane_check] FB set but no CRTC

with the following backtrace:

  [<ffffffffa010339d>] drm_atomic_check_only+0x35d/0x510 [drm]
  [<ffffffffa0103567>] drm_atomic_commit+0x17/0x60 [drm]
  [<ffffffffa00a6ccd>] drm_atomic_helper_plane_set_property+0x8d/0xd0 [drm_kms_helper]
  [<ffffffffa00f1fed>] drm_mode_plane_set_obj_prop+0x2d/0x90 [drm]
  [<ffffffffa00a8a1b>] restore_fbdev_mode+0x6b/0xf0 [drm_kms_helper]
  [<ffffffffa00aa969>] drm_fb_helper_restore_fbdev_mode_unlocked+0x29/0x80 [drm_kms_helper]
  [<ffffffffa00aa9e2>] drm_fb_helper_set_par+0x22/0x50 [drm_kms_helper]
  [<ffffffffa050a71a>] intel_fbdev_set_par+0x1a/0x60 [i915]
  [<ffffffff813ad444>] fbcon_init+0x4f4/0x580

That's because we update the plane state with the fb from the firmware, but we
never associate the plane to that CRTC.

We don't quite have the full DRM take over from HW state just yet, so
fake enough of the plane atomic state to pass the checks.

v2: Fix the state on which we set the CRTC in the case we're sharing the
    initial fb with another pipe. (Matt)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Store the initial framebuffer in initial_plane_config

At the moment we use crtc->base.primary->fb to hold the initial
framebuffer allocation, disregarding if it's valid or not.

This lead to believe we were actually updating the fb at this point, but
it's not true and we haven't even called drm_framebuffer_init() on this
fb.

Instead, let's store the state in struct intel_initial_plane_config
until we know we can reuse that framebuffer.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Don't try to reference the fb in get_initial_plane_config()

Tvrtko noticed a new warning on boot:

  WARNING: CPU: 1 PID: 353 at include/linux/kref.h:47 drm_framebuffer_reference+0x6c/0x80 [drm]()
  Call Trace:
  [<ffffffff8161f10c>] dump_stack+0x4f/0x7b
  [<ffffffff81052caa>] warn_slowpath_common+0xaa/0xd0
  [<ffffffff81052d8a>] warn_slowpath_null+0x1a/0x20
  [<ffffffffa00d035c>] drm_framebuffer_reference+0x6c/0x80 [drm]
  [<ffffffffa01c0df7>] update_state_fb.isra.54+0x47/0x50 [i915]
  [<ffffffffa01ccd5c>] skylake_get_initial_plane_config+0x93c/0x950 [i915]
  [<ffffffffa01e8721>] intel_modeset_init+0x1551/0x17c0 [i915]
  [<ffffffffa02476e0>] i915_driver_load+0xed0/0x11e0 [i915]
  [<ffffffff81627aa1>] ? _raw_spin_unlock_irqrestore+0x51/0x70
  [<ffffffffa00ca8b7>] drm_dev_register+0x77/0x110 [drm]
  [<ffffffffa00cda3b>] drm_get_pci_dev+0x11b/0x1f0 [drm]
  [<ffffffff81098e3d>] ? trace_hardirqs_on+0xd/0x10
  [<ffffffff81627aa1>] ? _raw_spin_unlock_irqrestore+0x51/0x70
  [<ffffffffa0145276>] i915_pci_probe+0x56/0x60 [i915]
  [<ffffffff813ad59c>] pci_device_probe+0x7c/0x100
  [<ffffffff81466aad>] driver_probe_device+0x16d/0x380

We cannot take a reference at this point, not before
intel_framebuffer_init() and the underlying drm_framebuffer_init().

Introduced in:

  commit 706dc7b549175e47f23e913b7f1e52874a7d0f56
  Author: Matt Roper <matthew.d.roper@intel.com>
  Date:   Tue Feb 3 13:10:04 2015 -0800

      drm/i915: Ensure plane->state->fb stays in sync with plane->fb

v2: Don't move update_state_fb(). It was moved around because I
    originally put update_state_fb() in intel_alloc_plane_obj() before
    finding a better place. (Matt)

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reported-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Use an intermediate variable to avoid repeating ourselves

The code look slightly better this way and will ease the next commit,
changing where we take the fb pointer from.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Put update_state_fb() next to the fb update

update_state_fb() at the end of intel_find_plane_obj() is misleading as
it leads us to believe the update is done for all code path.

A successful call to intel_alloc_plane_obj() will return and
update_state_fb() is then only needed when we share a fb from another
CRTC. Put the update() function there then.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/fbc: fix the check for already reserved fbc size

The check for previously reserved stolen space size for FBC in
i915_gem_stolen_setup_compression() did not take the compression
threshold into account. Fix this by storing and comparing to
uncompressed size instead.

The bug has been introduced in

commit 5e59f7175f96550ede91f58d267d2b551cb6fbba
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Mon Jun 30 10:41:24 2014 -0700

    drm/i915: Try harder to get FBC

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88975
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Declare that GT3 has a second VCS

v2: leave intel_skylake_info alone (Rodrigo, Daniel)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Split the SKL PCI ids by GT

We need to have a separate GT3 struct intel_device_info to declare they
have a second VCS. Let's start by splitting the PCI ids per-GT.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Implementation of SKL display power well support

This patch implements core logic of SKL display power well.

v2: Addressed Imre's comments
- Added respective DDIs under power well #1 and #2
- Simplified repetitive code in power well programming

v3: Implemented Imre's comments
- Further simplified power well programming
- Made sure that PW 1 is enabled prior to PW 2

v4: Fix minor conflict with the the cherryview support (Damien)

v5: Add the PLL power domain to the always on power well (Damien)

v6: Disable BIOS power well (Imre)
    Use power well data for comparison (Imre)
    Put the PLL power domain into PW1 as its needed for CDCLK (Satheesh,
    Damien)

v7: Addressed Imre's comments
  - Lowered the time out to 1ms
  - Added parantheses in macro
  - Moved debug message and fixed wait_for interval

v8:
  - Add a WARN() when swiching on an unknown power well (Imre, done by Damien)
  - Whitespace fixes (spaces instead of tabs) (Damien)

v9: (Imre, done by Damien)
  - Merge the register definitions with this patch
  - Merge the MISC IO power well in this patch

v10: (Imre, done by Damien)

  - Define the Misc I/O power domains to be the power well 1 ones as Misc I/O
    needs to be enabled with PW1
  - Added Transcoder A and VGA domains to PW 2
  - Remove the MISC_IO power domains as well in the the always on
    domains definition
  - Move Misc I/O power well at the top of the power well list so it's turned
    on right after PW1.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Satheeshakrishna M <satheeshakrishna.m@intel.com> (v3,v6,v7)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Ensure plane->state->fb stays in sync with plane->fb

plane->state->fb and plane->fb should always reference the same FB so
that atomic and legacy codepaths have the same view of display state.
In commit

        commit db068420560511de80ac59222644f2bdf278c3d5
        Author: Matt Roper <matthew.d.roper@intel.com>
        Date:   Fri Jan 30 16:22:36 2015 -0800

            drm/i915: Keep plane->state updated on pageflip

we already fixed one case where these two pointers could get out of
sync.  However it turns out there are a few other places (mainly dealing
with initial FB setup at boot) that directly set plane->fb and neglect
to update plane->state->fb.  If we never do a successful update through
the atomic pipeline, the RmFB cleanup code will look at the
plane->state->fb pointer, which has never actually been set to a
legitimate value, and try to clean it up, leading to BUG's.

Add a quick helper function to synchronize plane->state->fb with
plane->fb (and update reference counts accordingly) and call it
everywhere the driver tries to manually set plane->fb outside of the
atomic pipeline.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88909
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Remove the check enforcing VCS2 to be gen8 only

We already track this in the intel_info struct.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
[danvet: Make the commit message a bit less terse.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Correct the variable holding the value for EOT to write

This isuue got introduced in -

commit 24ee0e64909bf7f1953d87d3e1e29d93eafcad73
Author: Gaurav K Singh <gaurav.k.singh@intel.com>
Date:   Fri Dec 5 14:24:21 2014 +0530

    drm/i915: Update the DSI enable path to support dual

Signed-off-by: Shobhit Kumar <shobhit.kumar@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Introduce intel_set_rps()

Replace the valleyview_set_rps() and gen6_set_rps() calls with
intel_set_rps() which itself does the IS_VALLEYVIEW() check. The
code becomes simpler since the callers don't have to do this check
themselves.

Most of the change was performe with the following semantic patch:
@@
expression E1, E2, E3;
@@
- if (IS_VALLEYVIEW(E1)) {
-  valleyview_set_rps(E2, E3);
- } else {
-  gen6_set_rps(E2, E3);
- }
+ intel_set_rps(E2, E3);

Adding intel_set_rps() and making valleyview_set_rps() and gen6_set_rps()
static was done manually. Also valleyview_set_rps() had to be moved a
bit avoid a forward declaration.

v2: Use a less greedy semantic patch

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Track old framebuffer instead of object

Daniel Vetter spotted a bug while reviewing some of my refactoring in this
are of the code. I'll quote:

"""
> @@ -9764,6 +9768,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
>   work->event = event;
>   work->crtc = crtc;
>   work->old_fb_obj = intel_fb_obj(old_fb);
> + work->old_tiling_mode = to_intel_framebuffer(old_fb)->tiling_mode;

Hm, that's actually an interesting bugfix - currently userspace could be
sneaky and destroy the old fb immediately after the flip completes and the
change the tiling of the underlying object before the unpin work had a
chance to run (needs some fudgin with rt prios to starve workers to make
this work though).

Imo the right fix is to hold a reference onto the fb and not the
underlying gem object. With that tiling is guaranteed not to change.
"""

This patch tries to implement the above proposed change.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Switch planes from transitional helpers to full atomic helpers

There are two sets of helper functions provided by the DRM core that can
implement the .update_plane() and .disable_plane() hooks in terms of a
driver's atomic entrypoints.  The transitional helpers (which we have
been using so far) create a plane state and then use the plane's atomic
entrypoints to perform the atomic begin/check/prepare/commit/finish
sequence on that single plane only.  The full atomic helpers create a
top-level atomic state (which is capable of holding multiple object
states for planes, crtc's, and/or connectors) and then passes the
top-level atomic state through the full "atomic modeset" pipeline.

Switching from the transitional to full helpers here shouldn't result in
any functional change, but will enable us to exercise/test more of the
internal atomic pipeline with the legacy API's used by existing
applications.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Keep plane->state updated on pageflip

Until all drivers have transitioned to atomic, the framebuffer
associated with a plane is tracked in both plane->fb (for legacy) and
plane->state->fb (for all the new atomic codeflow).  All of our modeset
and plane updates use drm_plane->update_plane(), so in theory plane->fb
and plane->state->fb should always stay in sync and point at the same
thing for i915.  However we forgot about the pageflip ioctl case, which
currently only updates plane->fb and leaves plane->state->fb at a stale
value.

Surprisingly, this doesn't cause any real problems at the moment since
internally we use the plane->fb pointer in most of the places that
matter, and on the next .update_plane() call, we use plane->fb to figure
out which framebuffer to cleanup.  However when we switch to the full
atomic helpers for update_plane()/disable_plane(), those helpers use
plane->state->fb to figure out which framebuffer to cleanup, so not
having updated the plane->state->fb pointer causes things to blow up
following a pageflip ioctl.

The fix here is to just make sure we update plane->state->fb at the same
time we update plane->fb in the pageflip ioctl.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove bogus locking check in the hangcheck code

You can _never_ assert that a lock is not held, except in some very
restricted corner cases where it's guranteed that your code is running
single-threade (e.g. driver load before you've published any pointers
leading to that lock).

In addition the early return breaks a bunch of testcases since with
highly concurrent hangcheck stress tests the reset fails to work and
the test doesn't recover and time out.

This regression has been introduced in

commit b8d24a06568368076ebd5a858a011699a97bfa42
Author: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Date:   Wed Jan 28 17:03:14 2015 +0200

    drm/i915: Remove nested work in gpu error handling

Aside: It is possible to check whether a given task doesn't hold a
lock, but only when lockdep is enabled, using the lockdep_assert_held
stuff.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88908
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Update DRIVER_DATE to 20150130

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

drm/i915: Use pipe_config's cpu_transcoder for reading encoder hw state

The get_config() functions for ddi and dp_mst, used to read the value
of cpu_transcoder from the crtc->config instead of the state passed as
an argument. On the hardware state readout path, that happens to work
since the proper value is written to it before encoder->get_config() is
called. However, in the check_crtc() path, the state will be read from
the cpu_transcoder in the software tracking, instead of the one just
read out from hw. Using the field in the supplied intel_crtc_state
should do the right thing in both cases.

v2: Fix intel_ddi_get_config() too. (Ander)

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix a use-after-free in intel_execlists_retire_requests

Remove request from list before unreferencing it, in case it's actually
the only reference. (Found by Tvrtko Ursulin)

This issue has been most likely introduced in

commit 6d3d8274bc45de4babb62d64562d92af984dd238
Author: Nick Hoath <nicholas.hoath@intel.com>
Date:   Thu Jan 15 13:10:39 2015 +0000

    drm/i915: Subsume intel_ctx_submit_request in to drm_i915_gem_request

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Split shared dpll setup out of __intel_set_mode()

This simplifies __intel_set_mode() a little.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Don't do posting reads on getting forcewake

The checking for ack and also any subsequent mmio access
will serialize with setting the forcewake bit. Drop the
posting read as superfluous.

Note that in the put side we still want to keep the posting read
as it will ensure that the hw sees our forcewake release in a
timely manner and doesn't keep the hw powered up.

Comment from Chris:

On Wed, Jan 28, 2015 at 05:54:14PM +0200, Mika Kuoppala wrote:
> Ville Syrjälä <ville.syrjala@linux.intel.com> writes:
> > IIRC the posting read from same cache line actually fixed real bugs. So
> > I'm a bit worried about dropping them. But I suppose it's possible only
> > the _put side was important for those bugs.
>
> I found these:
>
> commit 6af2d180f82151cf3d58952e35a4f96e45bc453a
> Author: Daniel Vetter <daniel.vetter@ffwll.ch>
> Date:   Thu Jul 26 16:24:50 2012 +0200
>
>     drm/i915: fix forcewake related hangs on snb
>
> commit 8dee3eea3ccd3b6c00a8d3a08dd715d6adf737dd
> Author: Ben Widawsky <ben@bwidawsk.net>
> Date:   Sat Sep 1 22:59:50 2012 -0700
>
>     drm/i915: Never read FORCEWAKE
>
> https://bugs.freedesktop.org/show_bug.cgi?id=51738
> https://bugs.freedesktop.org/show_bug.cgi?id=52424
>
> The snb here seems to survive gem_dummy_reloc_loop and
> gem_ring_sync_loop in here with the get side posting removed.

Note that we kept the once associated with #52424, but judging by my
comments in #51738 the posting read is just a band aid anyway as a full
mb() itself was not adequate.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: paste relevant review discussion in.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Do uncore early sanitize after domain init

intel_uncore_early_sanitize() will reset the forcewake registers. When
forcewake domains were introduced, the domain init was done after the
sanitization of the forcewake registers. And as the resetting of
registers use the domain accessors, we tried to reset the forcewake
registers with unitialized forcewake domains and failed.

Fix this by sanitizing after all the domains have been initialized. Do
per domain clearing of forcewake register on domain init so that
IVB can do early access to ECOBUS do determine the final configuration.

This regression was introduced in

commit 05a2fb157e44a53c79133805d30eaada43911941
Author: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Date:   Mon Jan 19 16:20:43 2015 +0200

    drm/i915: Consolidate forcewake code

v2: Carve out ellc detect, fw_domain_reset for ivb/ecobus (Chris)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88805
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reported-by: Olof Johansson <olof@lixom.net>
Tested-by: Darren Hart <dvhart@linux.intel.com> (v1)
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Handle CHV in vlv_set_rps_idle()

Move the CHV check into vlv_set_rps_idle() to simplify the caller a bit.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove nested work in gpu error handling

Now when we declare gpu errors only through our own dedicated
hangcheck workqueue there is no need to have a separate workqueue
for handling the resetting and waking up the clients as the deadlock
concerns are no more.

The only exception is i915_debugfs::i915_set_wedged, which triggers
error handling through process context. However as this is only used through
test harness it is responsibility for test harness not to introduce hangs
through both debug interface and through hangcheck mechanism at the same time.

Remove gpu_error.work and let the hangcheck work do the tasks it used to.

v2: Add a big warning sign into i915_debugfs::i915_set_wedged (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/documentation: Add intel_uncore.c to drm.tmpl

Include intel_uncore.c in template for it to include d
documentation for intel_uncore_forcewake_get and *_put.

Cc: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>