PML4 has no special attributes, and there will always be a PML4.
So simply initialize it at creation, and destroy it at the end.

The code for 4lvl is able to call into the existing 3lvl page table code
to handle all of the lower levels.

v2: Return something at the end of gen8_alloc_va_range_4lvl to keep the
compiler happy. And define ret only in one place.
Updated gen8_ppgtt_unmap_pages and gen8_ppgtt_free to handle 4lvl.
v3: Use i915_dma_unmap_single instead of pci API. Fix a
couple of incorrect checks when unmapping pdp and pd pages (Akash).
v4: Call __pdp_fini also for 32b PPGTT. Clean up alloc_pdp param list.
v5: Prevent (harmless) out of range access in gen8_for_each_pml4e.
v6: Simplify alloc_vma_range_4lvl and gen8_ppgtt_init_common error
paths. (Akash)
v7: Rebase, s/gen8_ppgtt_free_*/gen8_ppgtt_cleanup_*/.
v8: Change location of pml4_init/fini. It will make next patches
cleaner.
v9: Rebase after Mika's ppgtt cleanup / scratch merge patch series, while
trying to reuse as much as possible for pdp alloc. pml4_init/fini
replaced by setup/cleanup_px macros.
v10: Rebase after Mika's merged ppgtt cleanup patch series.
v11: Rebase after final merged version of Mika's ppgtt/scratch
patches.
v12: Fix pdpe start value in trace (Akash)
v13: Define all 4lvl functions in this patch directly, instead of
previous patches, add i915_page_directory_pointer_entry_alloc here,
use test_bit to detect when pdp is already allocated (Akash).
v14: Move pdp allocation into a new gen8_ppgtt_alloc_page_dirpointers
funtion, as we do for pds and pts; move pd and pdp setup functions to
this patch (Akash).
v15: Added kfree(pdp) from previous patch to this (Akash).

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2+)
Reviewed-by: NAkash Goel <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Introduces the Page Map Level 4 (PML4), ie. the new top level structure
of the page tables.

To facilitate testing, 48b mode will be available on Broadwell and
GEN9+, when i915.enable_ppgtt = 3.

v2: Remove unnecessary CONFIG_X86_64 checks, ppgtt code is already
32/64-bit safe (Chris).
v3: Add goto free_scratch in temp 48-bit mode init code (Akash).
v4: kfree the pdp until the 4lvl alloc/free patch (Akash).
v5: Postpone 48-bit code in sanitize_enable_ppgtt (Akash).
v6: Keep _insert_pte_entries changes outside this patch (Akash).

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NMichel Thierry <michel.thierry@intel.com>
Reviewed-by: NAkash Goel <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The dynamic page allocation patch series added it for GEN6, this patch
adds them for GEN8.

v2: Consolidate pagetable/page_directory events
v3: Multiple rebases.
v4: Rebase after s/page_tables/page_table/.
v5: Rebase after Mika's ppgtt cleanup / scratch merge patch series.
v6: Rebase after gen8_map_pagetable_range removal.
v7: Use generic page name (px) in DECLARE_EVENT_CLASS (Akash)
v8: Defer define of i915_page_directory_pointer_entry_alloc (Akash)

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+)
Reviewed-by: NAkash Goel <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The insert_entries function was the function used to write PTEs. For the
PPGTT it was "hardcoded" to only understand two level page tables, which
was the case for GEN7. We can reuse this for 4 level page tables, and
remove the concept of insert_entries, which was never viable past 2
level page tables anyway, but it requires a bit of rework to make the
function a bit more generic.

v2: Rebase after Mika's ppgtt cleanup / scratch merge patch series.
v3: Rebase after final merged version of Mika's ppgtt/scratch patches.
v4: Check and warn for NULL value of pdp pointer (Akash).

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2)
Reviewed-by: NAkash Goel <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Up until now, ppgtt->pdp has always been the root of our page tables.
Legacy 32b addresses acted like it had 1 PDP with 4 PDPEs.

In preparation for 4 level page tables, we need to stop using ppgtt->pdp
directly unless we know it's what we want. The future structure will use
ppgtt->pml4 for the top level, and the pdp is just one of the entries
being pointed to by a pml4e. The temporal pdp local variable will be
removed once the rest of the 4-level code lands.

Also, start passing the vm pointer to the alloc functions, instead of
ppgtt.

v2: Updated after dynamic page allocation changes.
v3: Rebase after s/page_tables/page_table/.
v4: Rebase after changes in "Dynamic page table allocations" patch.
v5: Rebase after Mika's ppgtt cleanup / scratch merge patch series.
v6: Rebase after final merged version of Mika's ppgtt/scratch patches.
v7: Keep pagetable map in-line (and avoid unnecessary for_each_pde
loops), remove redundant ppgtt pointer in _alloc_pagetabs (Akash)
v8: Fix text indentation in _alloc_pagetabs/page_directories (Chris)
v9: Defer gen8_alloc_va_range_4lvl definition until 4lvl is implemented,
clean-up gen8_ppgtt_cleanup [pun intended] (Akash).
v10: Clean-up commit message (Akash).

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2+)
Reviewed-by: N"Akash Goel" <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This transitional patch doesn't do much for the existing code. However,
it should make upcoming patches to use the full 48b address space a bit
easier.

32-bit ppgtt uses just 4 PDPs, while 48-bit ppgtt will have up-to 512;
this patch prepares the existing functions to query the right number of pdps
at run-time. This also means that used_pdpes should also be allocated during
ppgtt_init, as the bitmap size will depend on the ppgtt address range
selected.

v2: Renamed  pdp_free to be similar to  pd/pt (unmap_and_free_pdp).
v3: To facilitate testing, 48b mode will be available on Broadwell and
GEN9+, when i915.enable_ppgtt = 3.
v4: Rebase after s/page_tables/page_table/, added extra information
about 4-level page table formats and use IS_ENABLED macro.
v5: Check CONFIG_X86_64 instead of CONFIG_64BIT.
v6: Rebase after Mika's ppgtt cleanup / scratch merge patch series, and
follow
his nomenclature in pdp functions (there is no alloc_pdp yet).
v7: Rebase after merged version of Mika's ppgtt cleanup patch series.
v8: Rebase after final merged version of Mika's ppgtt/scratch patches.
v9: Introduce PML4 (and 48-bit checks) until next patch (Akash).
v10: Also use test_bit to detect when pd/pt are already allocated (Akash)

Cc: Akash Goel <akash.goel@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2+)
Reviewed-by: NAkash Goel <akash.goel@intel.com>
[danvet: Amend commit message as suggested by Michel.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

gen8_clamp_pd clamps to the next page directory boundary, but the macro
gen8_for_each_pde already has a check to stop at the page directory
boundary.

Furthermore, i915_pte_count also restricts to the next page table
boundary.

v2: Rebase after Mika's ppgtt cleanup / scratch merge patch series.
Suggested-by: NAkash Goel <akash.goel@intel.com>
Signed-off-by: NMichel Thierry <michel.thierry@intel.com>
Reviewed-by: N"Akash Goel" <akash.goel@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

An OEM may request increased I_boost beyond the recommended values
by specifying an I_boost value to be applied to all swing entries for
a port. These override values are specified in VBT.

v2: rebase and remove unused iboost_bit variable

Issue: VIZ-5676
Signed-off-by: NAntti Koskipaa <antti.koskipaa@linux.intel.com>
Reviewed-by: NDavid Weinehall <david.weinehall@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Backmerge drm-intel-fixes because a bunch of atomic patch backporting
we had to do lead to horrible conflicts.

Conflicts:
	drivers/gpu/drm/drm_crtc.c
Just a bit of context conflict between -next and -fixes.
	drivers/gpu/drm/i915/intel_atomic.c
	drivers/gpu/drm/i915/intel_display.c
Atomic conflicts, always pick the code from -next.
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

There is currently conflicting documentation on which steppings the
workaround is needed, up to C vs. forever. However there is post-C
stepping hardware that doesn't report port presence on DDI A, leading to
black screen on eDP. Assume the strap isn't connected, and try to enable
DDI A on these machines. (We'll still check the VBT for the info in DDI
init.)

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Mika Westerberg <mika.westerberg@intel.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

DP spec requires the checksum of the last block read to be written
when replying to TEST_EDID_READ. This patch fixes the current code
to do the same.

v2: removed loop for jumping blocks and performed direct addition
as recommended by Daniel
Signed-off-by: NSivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Reviewed-by: NSonika Jindal <sonika.jindal@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Prevent leaking the if scoping by containing the WA_REG
macro inside its own scope.
Reported-by: NArun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NDave Gordon <david.s.gordon@intel.com>
[danvet: Appease checkpatch.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If we encounter an allocation failure during ppggt creation (trivial
even with 16Gib+ RAM!), we need to remove the dead context from the
fpriv->context_idr along with the references.

gem_exec_ctx: page allocation failure: order:0, mode:0x8004
CPU: 3 PID: 27272 Comm: gem_exec_ctx Tainted: G        W       4.2.0-rc5+ #37
 0000000000000000 ffff880086ff7a78 ffffffff816b947a ffff88041ed90038
 0000000000008004 ffff880086ff7b08 ffffffff8114b1a5 ffff880086ff7ac8
 ffffffff8108d848 0000000000000000 ffffffff81ce84b8 0000000000000000
Call Trace:
 [<ffffffff816b947a>] dump_stack+0x45/0x57
 [<ffffffff8114b1a5>] warn_alloc_failed+0xd5/0x120
 [<ffffffff8108d848>] ? __wake_up+0x48/0x60
 [<ffffffff8114e0ed>] __alloc_pages_nodemask+0x73d/0x8e0
 [<ffffffffc0472238>] ? i915_gem_execbuffer2+0x148/0x240 [i915]
 [<ffffffffc0474240>] __setup_page_dma+0x30/0x110 [i915]
 [<ffffffffc0477f61>] gen8_ppgtt_init+0x31/0x2f0 [i915]
 [<ffffffffc04785e0>] i915_ppgtt_init+0x30/0x80 [i915]
 [<ffffffffc0478928>] i915_ppgtt_create+0x48/0xc0 [i915]
 [<ffffffffc046c9c2>] i915_gem_create_context+0x1c2/0x390 [i915]
 [<ffffffffc046d9cb>] i915_gem_context_create_ioctl+0x5b/0xa0 [i915]

leading to an oops in i915_gem_context_close. Also note that this
benchmark should not be running out of memory in the first place...

Testcase: igt/benchmark/gem_exec_ctx -b create # ppgtt >= 2
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The IOMMU for Intel graphics has historically had many issues resulting
in random GPU hangs. Lets include its status when capturing the GPU hang
error state for post-mortem analysis.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If idle to active bit is set, the rest of the fields
in CSQ are not valid.

Bail out early if this is the case in order to prevent
rest of the loop inspecting stale values.

This was found by Bspec/code inspection. Doesn't seem to fix any of
the known issues.
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NArun Siluvery <arun.siluvery@linux.intel.com>
[danvet: Add note about how this was found.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There is no correspondent Aux channel for DDI-E.

So we need to rely on VBT to let us know witch one
is being used instead.

v2: Removing some trailing spaces and giving proper
credit to Xiong that added a nice way to avoid port
conflicts by setting supports_dp = 0 when using
equivalent aux for DDI-E.

Credits-to: Xiong Zhang <xiong.y.zhang@intel.com>
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: NXiong Zhang <xiong.y.zhang@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

DDI-E and DDI-A share 4 the same DDI-A lanes.
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: NXiong Zhang <xiong.y.zhang@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I started digging this when I noticed that the BDW code was just
reserving 1mb by coincidence since it was reading reserved fields.
Then I noticed we didn't have any values set for SNB and earlier, and
that the HSW sizes were wrong. After that, I noticed that the reserved
area has a specific start, and may not exactly end where the stolen
memory ends. I also noticed the base pointer can be zero. So I decided
to just write a single patch fixing everything instead of 20 patches
that would be much harder to review.

This patch may solve random stolen memory corruption/problems on
almost all platforms. Notice that since this is always dealing with
the top of the stolen memory, the problems are not so easy to
reproduce - especially since FBC is still disabled by default.

One of the major differences of this patch is that we now look at both
the size and base address. By only looking at the size we were
assuming that the reserved area was always at the very top of
stolen, which is not always true.

After we merge the patch series that allows user space to allocate
stolen memory we'll be able to write IGT tests that maybe catch the
bugs fixed by this patch.

v2:
  - s/BIOS reserved/stolen reserved/g (Chris)
  - Don't DRM_ERROR if we can't do anything about it (Chris)
  - Improve debug messages (Chris).
  - Use the gen7 version instead of gen6 on HSW. Tom found some
    documentation problems, so I think with gen7 we're on the safer
    side (Tom).
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This register needs to be updated with masked writes.

This was found by code inspection and comparison with Bspec and
doesn't seem to fix any known issue.
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NMichel Thierry <michel.thierry@intel.com>
[danvet: Add note about impact.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Add WaDisableSbeCacheDispatchPortSharing:skl

Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: NArun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If we encounter frequent problems with dp aux channel
communications, we end up spamming the dmesg with the
exact similar trace and status.

Inject a new backtrace only if we have new information
to share as otherwise we flush out all other important
stuff.
Signed-off-by: NMika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

-EDEADLK has special meaning in atomic, but get_fence may call
i915_find_fence_reg which can return -EDEADLK.

This has special meaning in the atomic world, so convert the error
to -EBUSY for this case.

Changes since v1:
- Add comment in the code.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The rest will be a noop anyway, since without modeset there will be
no updated dplls and no modeset state to update.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There are no more users, byebye!
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that everything's atomic, checking encoder->base.crtc is enough.
This function doesn't have the locks to dereference crtc->state, but
stealing an encoder bound to any crtc is probably enough reason to warn.

Changes since v1:
- Commit message.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

connectors_active will be removed, so just calculate this instead.

Changes since v1:
- Look for the right pointer in intel_sanitize_encoder.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is now done completely atomically.
Keep connectors_active for now, but make it mirror crtc_state->active.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Instead of allocating pipe_config on the stack use the old
crtc_state, it's only going to freed from this point on.

All crtc' are now only checked once during modeset,
because false positives can happen with encoders after
dpms changes and to limit the amount of errors for 1 failure.

Changes since v1:
- crtc_state -> old_crtc_state
- state -> old_state
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Connectors are updated atomically now, so the only interaction
with the encoder is through base.crtc.

If it's NULL the encoder's not part of any crtc, and if it's
not NULL then active should be equal to crtc_state->active.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is handled by the atomic core now, no need to check this for ourself.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Right now dpms callbacks can still fiddle with the connector state,
but it can only turn connectors off.

This is remediated by only checking crtc->state->active when the
connector is active, and ignore crtc->state->active when the
connector is off.

connectors_active is no longer checked, and will be removed later
in this series together with dpms.

Another check for !encoder->crtc is performed by check_encoder_state
too, so it can be removed.

Changes since v1:
- Add commit message.
- rename state to old_state.
- Move deletion of mst_port check to mst patch.
Changes since v2:
- Fix a null pointer dereference on MST now hw readout is fixed.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Fully remove the MST connector from the atomic state, and remove the
early returns in check_*_state for MST connectors.

With atomic the state can be made consistent all the time.

Thanks to Sivakumar Thulasimani for the idea of using
drm_atomic_helper_set_config.

Changes since v1:
- Remove the MST check in intel_connector_check_state too.
Changes since v2:
- Use drm_atomic_helper_set_config.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Reviewed-by: NSivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

First step in removing dpms and validating atomic state.

There can still be a mismatch in the connector state because the dpms
callbacks are still used, but this can not happen immediately after a modeset.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Set connectors_changed to force a modeset if the panel fitter's force
enabled on eDP.

Changes since v1:
- Use connectors_changed instead of active_changed because it's a
  routing update.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This patch is based on the upstream commit 5ac1c4bc and amended
for v4.2 to make sure it works as intended.

Repeated calls to begin_crtc_commit can cause warnings like this:
[  169.127746] BUG: sleeping function called from invalid context at kernel/locking/mutex.c:616
[  169.127835] in_atomic(): 0, irqs_disabled(): 1, pid: 1947, name: kms_flip
[  169.127840] 3 locks held by kms_flip/1947:
[  169.127843]  #0:  (&dev->mode_config.mutex){+.+.+.}, at: [<ffffffff814774bc>] __drm_modeset_lock_all+0x9c/0x130
[  169.127860]  #1:  (crtc_ww_class_acquire){+.+.+.}, at: [<ffffffff814774cd>] __drm_modeset_lock_all+0xad/0x130
[  169.127870]  #2:  (crtc_ww_class_mutex){+.+.+.}, at: [<ffffffff81477178>] drm_modeset_lock+0x38/0x110
[  169.127879] irq event stamp: 665690
[  169.127882] hardirqs last  enabled at (665689): [<ffffffff817ffdb5>] _raw_spin_unlock_irqrestore+0x55/0x70
[  169.127889] hardirqs last disabled at (665690): [<ffffffffc0197a23>] intel_pipe_update_start+0x113/0x5c0 [i915]
[  169.127936] softirqs last  enabled at (665470): [<ffffffff8108a766>] __do_softirq+0x236/0x650
[  169.127942] softirqs last disabled at (665465): [<ffffffff8108ae75>] irq_exit+0xc5/0xd0
[  169.127951] CPU: 1 PID: 1947 Comm: kms_flip Not tainted 4.1.0-rc4-patser+ #4039
[  169.127954] Hardware name: LENOVO 2349AV8/2349AV8, BIOS G1ETA5WW (2.65 ) 04/15/2014
[  169.127957]  ffff8800c49036f0 ffff8800cde5fa28 ffffffff817f6907 0000000080000001
[  169.127964]  0000000000000000 ffff8800cde5fa58 ffffffff810aebed 0000000000000046
[  169.127970]  ffffffff81c5d518 0000000000000268 0000000000000000 ffff8800cde5fa88
[  169.127981] Call Trace:
[  169.127992]  [<ffffffff817f6907>] dump_stack+0x4f/0x7b
[  169.128001]  [<ffffffff810aebed>] ___might_sleep+0x16d/0x270
[  169.128008]  [<ffffffff810aed38>] __might_sleep+0x48/0x90
[  169.128017]  [<ffffffff817fc359>] mutex_lock_nested+0x29/0x410
[  169.128073]  [<ffffffffc01635f0>] ? vgpu_write64+0x220/0x220 [i915]
[  169.128138]  [<ffffffffc017fddf>] ? ironlake_update_primary_plane+0x2ff/0x410 [i915]
[  169.128198]  [<ffffffffc0190e75>] intel_frontbuffer_flush+0x25/0x70 [i915]
[  169.128253]  [<ffffffffc01831ac>] intel_finish_crtc_commit+0x4c/0x180 [i915]
[  169.128279]  [<ffffffffc00784ac>] drm_atomic_helper_commit_planes+0x12c/0x240 [drm_kms_helper]
[  169.128338]  [<ffffffffc0184264>] __intel_set_mode+0x684/0x830 [i915]
[  169.128378]  [<ffffffffc018a84a>] intel_crtc_set_config+0x49a/0x620 [i915]
[  169.128385]  [<ffffffff817fdd39>] ? mutex_unlock+0x9/0x10
[  169.128391]  [<ffffffff81467b69>] drm_mode_set_config_internal+0x69/0x120
[  169.128398]  [<ffffffff8119b547>] ? might_fault+0x57/0xb0
[  169.128403]  [<ffffffff8146bf93>] drm_mode_setcrtc+0x253/0x620
[  169.128409]  [<ffffffff8145c600>] drm_ioctl+0x1a0/0x6a0
[  169.128415]  [<ffffffff810b3b41>] ? get_parent_ip+0x11/0x50
[  169.128424]  [<ffffffff811e9ab8>] do_vfs_ioctl+0x2f8/0x530
[  169.128429]  [<ffffffff810d0fcd>] ? trace_hardirqs_on+0xd/0x10
[  169.128435]  [<ffffffff812e7676>] ? selinux_file_ioctl+0x56/0x100
[  169.128439]  [<ffffffff811e9d71>] SyS_ioctl+0x81/0xa0
[  169.128445]  [<ffffffff81800697>] system_call_fastpath+0x12/0x6f

Solve it by using the newly introduced drm_atomic_helper_commit_planes_on_crtc.

The problem here was that the drm_atomic_helper_commit_planes() helper
we were using was basically designed to do

    begin_crtc_commit(crtc #1)
    begin_crtc_commit(crtc #2)
    ...
    commit all planes
    finish_crtc_commit(crtc #1)
    finish_crtc_commit(crtc #2)

The problem here is that since our hardware relies on vblank evasion,
our CRTC 'begin' function waits until we're out of the danger zone in
which register writes might wind up straddling the vblank, then disables
interrupts; our 'finish' function re-enables interrupts after the
registers have been written.  The expectation is that the operations between
'begin' and 'end' must be performed without sleeping (since interrupts
are disabled) and should happen as quickly as possible.  By clumping all
of the 'begin' calls together, we introducing a couple problems:
 * Subsequent 'begin' invocations might sleep (which is illegal)
 * The first 'begin' ensured that we were far enough from the vblank that
   we could write our registers safely and ensure they all fell within
   the same frame.  Adding extra delay waiting for subsequent CRTC's
   wasn't accounted for and could put us back into the 'danger zone' for
   CRTC #1.

This commit solves the problem by using a new helper that allows an
order of operations like:

   for each crtc {
        begin_crtc_commit(crtc)  // sleep (maybe), then disable interrupts
        commit planes for this specific CRTC
        end_crtc_commit(crtc)    // reenable interrupts
   }

so that sleeps will only be performed while interrupts are enabled and
we can be sure that registers for a CRTC will be written immediately
once we know we're in the safe zone.

The crtc->config->base.crtc update may seem unrelated, but the helper
will use it to obtain the crtc for the state. Without the update it
will dereference NULL and crash.

Changes since v1:
- Use Matt Roper's commit message.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NMatt Roper <matthew.d.roper@intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=90398Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

This should be much cleaner, with the same effects.

(cherry picked for v4.2 from commit fb9d6cf8)
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: NMatt Roper <matthew.d.roper@intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=90398Reviewed-by: NAnder Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

In

commit d328c9d7
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Apr 10 16:22:37 2015 +0200

    drm/i915: Select starting pipe bpp irrespective or the primary plane

we started to select the pipe bpp from sink capabilities and not from
the primary framebuffer - that one might change (and we don't want to
incur a modeset) and sprites might contain higher bpp content too.

We also selected dithering on a 8 bpc screen displaying a 24bpp rgb
primary, because pipe_bpp is 24 for such a typical 8 bpc sink, but since
the commit mentioned above, base_bpp is always the absolute maximum
supported by the hardware, e.g., 36 bpp on my Ironlake chip. Iow. the
only way to not get dithering would have been to connect a deep color 12
bpc display, so pipe_bpp == 36 == base_bpp.

Hence only enable dithering on 6bpc screens where we difinitely and
always want it.

Cc: Mario Kleiner <mario.kleiner.de@gmail.com>
Reported-by: NMario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Reviewed-and-tested-by: NMario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: NJani Nikula <jani.nikula@intel.com>

Pull input subsystem fixes from Dmitry Torokhov:
 "Just small ALPS and Elan touchpads, and other driver fixups"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: elantech - add special check for fw_version 0x470f01 touchpad
  Input: twl4030-vibra - fix ERROR: Bad of_node_put() warning
  Input: alps - only Dell laptops have separate button bits for v2 dualpoint sticks
  Input: axp20x-pek - add module alias
  Input: turbografx - fix potential out of bound access

Pull MIPS fixes from Ralf Baechle:
 "Another round of MIPS fixes for 4.2.  No area does particularly stand
  out but we have a two unpleasant ones:

   - Kernel ptes are marked with a global bit which allows the kernel to
     share kernel TLB entries between all processes.  For this to work
     both entries of an adjacent even/odd pte pair need to have the
     global bit set.  There has been a subtle race in setting the other
     entry's global bit since ~ 2000 but it take particularly
     pathological workloads that essentially do mostly vmalloc/vfree to
     trigger this.

     This pull request fixes the 64-bit case but leaves the case of 32
     bit CPUs with 64 bit ptes unsolved for now.  The unfixed cases
     affect hardware that is not available in the field yet.

   - Instruction emulation requires loading instructions from user space
     but the current fast but simplistic approach will fail on pages
     that are PROT_EXEC but !PROT_READ.  For this reason we temporarily
     do not permit this permission and will map pages with PROT_EXEC |
     PROT_READ.

  The remainder of this pull request is more or less across the field
  and the short log explains them well"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: Make set_pte() SMP safe.
  MIPS: Replace add and sub instructions in relocate_kernel.S with addiu
  MIPS: Flush RPS on kernel entry with EVA
  Revert "MIPS: BCM63xx: Provide a plat_post_dma_flush hook"
  MIPS: BMIPS: Delete unused Kconfig symbol
  MIPS: Export get_c0_perfcount_int()
  MIPS: show_stack: Fix stack trace with EVA
  MIPS: do_mcheck: Fix kernel code dump with EVA
  MIPS: SMP: Don't increment irq_count multiple times for call function IPIs
  MIPS: Partially disable RIXI support.
  MIPS: Handle page faults of executable but unreadable pages correctly.
  MIPS: Malta: Don't reinitialise RTC
  MIPS: unaligned: Fix build error on big endian R6 kernels
  MIPS: Fix sched_getaffinity with MT FPAFF enabled
  MIPS: Fix build with CONFIG_OF=y for non OF-enabled targets
  CPUFREQ: Loongson2: Fix broken build due to incorrect include.