drm/i915: Kill intel_prepare_ddi()

Move the ddi buffer translation programming to occur from the encoder
.pre_enable() hook, for just the ddi port we are enabling. Previously
we used to reprogram the translations for all ddi ports during
init and during power well enabling.

v2: s/intel_prepare_ddi_buffers/intel_prepare_ddi_buffer/ (Daniel)
    Resolve conflicts due to  dev_priv->atomic_cdclk_freq

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Reject >9 ddi translation entried if port != A/E on SKL

Only DDI A and E support 10 translation entries in DP mode. For the
other ports the tenth entry is reserved for HDMI.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449597590-6971-9-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Pass around dev_priv for ddi buffer programming

Make the ddi buffer programming code a bit more neat by passing
around dev_priv instead of dev.

v2: Resolve conflicts due to KBL

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v1)

drm/i915: Eliminate duplicated skl_get_buf_trans_dp()

skl_get_buf_trans_edp() effectively contains another copy of
skl_get_buf_trans_dp(). Remove the duplication and just call
skl_get_buf_trans_dp() from  skl_get_buf_trans_edp().

v2: Resolve conflicts due to KBL

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v1)

drm/i915: Remove pointless 'ddi_translations' local variable

skl_get_buf_trans_*() don't need the 'ddi_translations' local variable
since all they with is assign and return. Just return the right thing
directly and get rid of the local variable.

v2: Resolve conflicts due to KBL

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v1)

drm/i915: Store max lane count in intel_digital_port

Rather than having open coded checks for the DDI A/E configuration,
just store the max supported lane count in intel_digital_port.

We had an open coded check for DDI A, but not for DDI E. So we may
have been vilating the DDI E max lane count.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Check max number of lanes when registering DDI ports

DDI A and E share some of the lanes, so check that we have enough
lanes for the purpose we need before registering the encoders.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449597590-6971-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Pass the correct encoder to intel_ddi_clk_select() with MST

We're supposed to pass the primary DP encoder to intel_ddi_clk_select(),
not the fake MST encoder. Do so.

There's no real bug here though, since intel_ddi_clk_select() only
checks if the encoder type is EDP (which it isn't for either the
primary DP encoder or the fake MST encoder), and it gets the DDI port
via intel_ddi_get_encoder_port() (which knows how to do the
fake->primary->port dance itself).

Fixes: e404ba8 ("drm/i915: Setup DDI clk for MST on SKL")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449597590-6971-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bdw+: Replace list_del+list_add_tail with list_move_tail

Same effect for slightly less source code and resulting binary.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1452521321-4032-2-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915/dsi: add debug printing of the new sequence block names

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452006497-28517-1-git-send-email-jani.nikula@intel.com

drm/i915/dsi: reduce tedious repetition

Make it a bit tidier and safer.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/cab24a84811ddbae72d8c3a5f59d29f57b1d3aad.1450702954.git.jani.nikula@intel.com

drm/i915/dsi: skip unknown elements for sequence block v3+

The sequence block has sizes of elements after the operation byte since
sequence block v3. Use it to skip elements we don't support yet.

v2: remove redundant exec_elem[operation_byte] check (Daniel)

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452006408-27688-1-git-send-email-jani.nikula@intel.com

drm/i915/bios: add support for MIPI sequence block v3

The changes since the sequence block v2 are:

* The whole MIPI bios data block has a separate 32-bit size field since
  v3, stored after the version. This facilitates big sequences.

* The size of the panel specific sequence blocks has grown to 32
  bits. This facilitates big sequences.

* The elements within sequences now have an 8-bit size field following
  the operation byte. This facilitates skipping unknown new operation
  bytes, i.e. forward compatibility.

v2 (of the patch): use DRM_ERROR for unknown operation byte
v3 (of the patch): even more bounds checking (Ville)

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452518102-3154-1-git-send-email-jani.nikula@intel.com

drm/i915/bios: add defines for v3 sequence block

New sequences, new operations within sequences.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/96335b9fb875f79882d694360bff060251bd2f17.1450702954.git.jani.nikula@intel.com

drm/i915: skip the i2c element in the generic VBT DSI driver

Don't choke on unknown elements when we do know how to skip them.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452518948-16469-1-git-send-email-jani.nikula@intel.com

drm/i915/bios: add sequences for MIPI sequence block v2

Properly parse the new sequences added in MIPI sequence block v2.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/cc1551bdfc4392d02413b78179f3a65c786c75ab.1450702954.git.jani.nikula@intel.com

drm/i915/bios: interpret the i2c element

Add parsing of the i2c element, defined in MIPI sequence block v2. Drop
the status operation byte while at it, that does not exist.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/d8a2998977feee2f5b5ad609aaca787adfb41479.1450702954.git.jani.nikula@intel.com

drm/i915: Arm the unclaimed mmio debugs on suspend path

If we go into suspend with unclaimed access detected,
it would be nice to catch that access on a next suspend path.
So instead of just notifying about it, arm the unclaimed
mmio checks on suspend side.

We want to keep the asymmetry on resume, as if it was
on resume path, it was not driver that is responsible so
no point in arming mmio debugs.

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>
Link: http://patchwork.freedesktop.org/patch/msgid/1452261080-6979-2-git-send-email-mika.kuoppala@intel.com

drm/i915: Enable mmio_debug for vlv/chv

With commit 8ac3e1bb76cc ("drm/i915: Add non claimed mmio checking
for vlv/chv") we now have chv/vlv support in place for detecting
unclaimed access. Also the perf hit of extra mmio read
is now only suffered if mmio_debug is set.

This allows us to stuff the macro for unclaimed reg
detection inside a generic gen6 register access, as now all
gens using these macros uses also unclaimed debugs, the one
exception being snb. We gain more clean and generic macros
and only downside is that snb will suffer one branch perf hit
without upside.

Note that the hsw write path debug register check now
happens before fifo check, but this should not make
any real difference.

As vlv/chv use the generic gen6 access macros, the consequence
is that they gain the mmio_debug feature.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
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>
Link: http://patchwork.freedesktop.org/patch/msgid/1452261080-6979-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Widen return value for reservation_object_wait_timeout_rcu to long.

This fixes a spurious warning from an integer overflow on 64-bits systems.
The function may return MAX_SCHEDULE_TIMEOUT which gets truncated to -1.

Explicitly handling this by casting to lret fixes it.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reported-and-tested-by: Joseph Yasi <joe.yasi@gmail.com>
Tested-by: Andreas Reis <andreas.reis@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: 3c28ff22f6e20c ("i915: wait for fence in prepare_plane_fb")
Link: http://patchwork.freedesktop.org/patch/msgid/5666EEC8.2000403@linux.intel.com

drm/i915: Update DRIVER_DATE to 20160111

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

drm/i915: Don't warn if the workaround list is empty part 2.

Extend the same reasoning as in the patch listed below.  It's not an
error for the workaround list to be empty if no workarounds are needed.

    commit 02235808b61cd9382d224b0df263193006dd9913
    Author: Francisco Jerez <currojerez@riseup.net>
    Date:   Wed Oct 7 14:44:01 2015 +0300
        drm/i915: Don't warn if the workaround list is empty.

Signed-off-by: Wayne Boyer <wayne.boyer@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452129330-3484-2-git-send-email-rodrigo.vivi@intel.com

drm/i915/kbl: Adding missing IS_KABYLAKE checks.

When adding IS_KABYLAKE definition I didn't included the
DC states related because I was planing to include them
with the patch that fixes DMC firmware loading, but I
forgot them.

Meanwhile this runtime pm code changed a lot for
Skylake.

Well, I didn't expect that this would crash the machine
and I just noticed now that Sarah warned me our driver
wasn't working. Thanks Sarah.

Michel had found the main error first and his
fix had better details on the history and got
merged already:

commit 16fbc291cb87c7defcd13ad715d3e4af0d523e43
Author: Michel Thierry <michel.thierry@intel.com>
Date:   Wed Jan 6 12:08:36 2016 +0000

    drm/i915/kbl: Enable PW1 and Misc I/O power wells

This one is a follow-up adding the other remaining
missing pieces.

v2: Rebased on top of Michel's patch as explained above.

Cc: Sarah Sharp <sarah.a.sharp@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452214179-22361-1-git-send-email-rodrigo.vivi@intel.com

drm/i915/dsi: remove unused dsi_rr_formula()

The dsi_rr_formula() function has been unused for almost two years,
since

commit 44d4c6eebb2ef04f698c292bb6eda5f2e80c663b
Author: Shobhit Kumar <shobhit.kumar@intel.com>
Date:   Tue Dec 10 12:14:56 2013 +0530

    drm/i915: Compute dsi_clk from pixel clock

citing the reason as pixel clock based calculation being recommended in
the MIPI host controller documentation. Remove the dead code, we can
always bring it back if it's needed.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452249940-2605-2-git-send-email-jani.nikula@intel.com

drm/i915/dsi: abstract get pclk platform differences

Hide away the platform differences in intel_dsi_get_pckl() within
intel_dsi_pll.c. No functional changes.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452249940-2605-1-git-send-email-jani.nikula@intel.com

drm/i915: Add non claimed mmio checking for vlv/chv

Imre mentioned that chv might also have capability to
track unclaimed mmio accesses. Ville added that
both chv and vlv has this capability and he had already
made this way back [1]. Mimic what Ville's patch does
but adapt on top of less frequent mmio accesses by
omitting checking always on reg writes.

This patch is untested as of now.

v2: overflow handling and posting omitted (Ville)

References: [1] http://lists.freedesktop.org/archives/intel-gfx/2013-May/027599.html
Cc: Imre Deak <imre.deak@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450201542-22918-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Streamline unclaimed reg debug trace

Remove char* assignments and add branching hint and
also constify the parameters.

This results in a 35 bytes shorter fast path, so author
boldly assumes it helps without doing in-depth assembly
analysis.

v2: use WARN's branching (Chris), commit name (Joonas)

Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450189512-30360-5-git-send-email-mika.kuoppala@intel.com

drm/i915: Do one shot unclaimed mmio detection less frequently

We have done unclaimed register access check in normal
(mmio_debug=0) mode once per write. This adds probability
of finding the exact sequence where we did the bad access, but
also adds burden to each write.

As we have mmio_debug available for more fine grained analysis,
give up accuracy of detecting correct spot at the first occurrence
by doing the one shot detection and arming of mmio_debug in hangcheck
and in modeset. This removes the write path performance burden.

v2: Remove gratuitous DRM_DEBUG and return value, comments (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450250808-14864-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Detect and clear unclaimed access on resume

If something, the usual suspect being bios, access hw
behind our back, don't let it slide into situation where
normal register access will detect this and spit out
a warn on into dmesg. On some bdw bioses this happens
during igt/bat run always and as there is not much we can
do about it, its better just to detect and flush this
explicitly on resume and only print a debug message.

v2: use DRM_DEBUG_DRIVER (Chris)
v3: s/access/mmio, s/prior/prior to, s/dev/dev_priv

Testcase: igt/pm_rpm/basic-rte
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <przanoni@gmail.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[Mika: fixed merge conflict]
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450189512-30360-3-git-send-email-mika.kuoppala@intel.com

drm/i915: Introduce intel_uncore_unclaimed_mmio

Currently interrupt code is the only place checking
for the unclaimed register access prior to actual register
macros using the same functionality. Rename the function
and make it return bool so that the possible error message
context is clear in the caller side. The motivation is to allow
usage of unclaimed detection on arbitrary places.

v2: rebase, s/access/mmio, s/dev/dev_priv

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450189512-30360-2-git-send-email-mika.kuoppala@intel.com

drm/i915: Consolidate unclaimed mmio detection

Access the unclaimed reg detection register through
one helper which also does cleanup. Note that we now access
the register only if the platform has the actual non claimed
access bit. This prevents reading the register with gens that
doesn't have the register or the unclaimed bit,
when debug_mmio > 0.

Note that we post after clearing the bit. This makes sure
that the next unclaimed write access would get detected
also if it happened right after clearing, and not fold
into the previous detection.

v2: s/unclaimed_reg_access/check_for_unclaimed_mmio (Chris)
    debug log on unclaimed detection on uncore init (Joonas)

v3: remove posting read (Ville)

Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450200287-24080-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Inspect subunit states on hangcheck

If head seems stuck and engine in question is rcs,
inspect subunit state transitions from undone to done,
before deciding that this really is a hang instead of limited
progress. Only account the transitions of subunits from
undone to done once, to prevent unstable subunit states
to keep us falsely active.

As this adds one extra steps to hangcheck heuristics,
before hang is declared, it adds 1500ms to to detect hang
for render ring to a total of 7500ms. We could sample
the subunit states on first head stuck condition but
decide not to do so only in order to mimic old behaviour. This
way the check order of promotion from seqno > atchd > instdone
is consistently done.

v2: Deal with unstable done states (Arun)
    Clear instdone progress on head and seqno movement (Chris)
    Report raw and accumulated instdone's in in debugfs (Chris)
    Return HANGCHECK_ACTIVE on undone->done

References: https://bugs.freedesktop.org/show_bug.cgi?id=93029
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1448985372-19535-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Cleaning up DDI translation tables

No functional changes.

That state the obvious and just duplicate the place we
need to change whenever the table is updated. So let's clean it.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452021535-22641-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Update Skylake DDI translation table for DP.

When reviewing DDI translation table I noticed few changes we
haven't incorporated yet and it is always good to follow latest
spec.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452009511-15064-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Update Skylake DDI translation table for HDMI.

When debuging an intermittent corrupted screen I suspected on DDI
translation table and checked we are out of date with the spec.

I'm not sure this will fix my bug yet, but it is always good to follow
the spec.

v2: Ville caught a switched i-boost value. Thanks!

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452021087-21673-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: intel_hpd_init(): Fix suspend/resume reprobing

This fixes reprobing of display connectors on resume.  After some
talking with danvet on IRC, I learned that calling
drm_helper_hpd_irq_event() does actually trigger a full reprobe of each
connector's status. It turns out this is the actual reason reprobing on
resume hasn't been working (this was observed on a T440s):

- We call hpd_init()
- We check each connector for a couple of things before marking
  connector->polled with DRM_CONNECTOR_POLL_HPD, one of which is an
  active encoder. Of course, a disconnected port won't have an
  active encoder, so we don't add the flag to any of the
  connectors.
- We call drm_helper_hpd_irq_event()
- drm_helper_irq_event() checks each connector for the
  DRM_CONNECTOR_POLL_HPD flag. The only one that has it is eDP-1,
  so we skip reprobing each connector except that one.

In addition, we also now avoid setting connector->polled to
DRM_CONNECTOR_POLL_HPD for MST connectors, since their reprobing is
handled by the mst helpers. This is probably what was originally
intended to happen here.

Changes since V1:
* Use the explanation of the issue as the commit message instead
* Change the title of the commit, since this does more then just stop a
  check for an encoder now
* Add "Fixes" line for the patch that introduced this regression
* Don't enable DRM_CONNECTOR_POLL_HPD for mst connectors

Changes since V2:
* Put patch changelog above Signed-off-by
* Follow Daniel Vetter's suggestion for making the code here a bit more
  legible

Fixes: 0e32b39ceed6 ("drm/i915: add DP 1.2 MST support (v0.7)")
Cc: stable@vger.kernel.org
Signed-off-by: Lyude <cpaul@redhat.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452181408-14777-1-git-send-email-cpaul@redhat.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Extract CSB status read

This is a useful thing to have around as a function because the mechanism may
change in the future.

There is a net increase in LOC here, and it will continue to be the case on GEN8
and GEN9 - but future GENs may have an alternate mechanism for doing this.

Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452018609-10142-4-git-send-email-benjamin.widawsky@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Change WARN to ERROR in CSB count

There is no point in emitting a WARN since the backtrace will always be the
same. Errors have actually become easier to spot given the large number of WARNs
which exist today in modesetting paths.

Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452018609-10142-3-git-send-email-benjamin.widawsky@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Cleanup some of the CSB handling

I think this patch is a worthwhile cleanup even if it might look only marginally
useful. It gets more useful in upcoming patches and for handling of future GEN
platforms.

The only non-mechanical part of this is the removal of the extra & operation on
the ring->next_context_status_buffer. This is safe because right above this, we
already did a modulus operation.

Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452018609-10142-2-git-send-email-benjamin.widawsky@intel.com
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/kbl: Enable PW1 and Misc I/O power wells

My kbl stopped working because of this.

Fixes regression from
commit 2f693e28b8df69f67beced5e18bb2b91c2bfcec2
Author: Damien Lespiau <damien.lespiau@intel.com>
Date:   Wed Nov 4 19:24:12 2015 +0200
    drm/i915: Make turning on/off PW1 and Misc I/O part of the init/fini
    sequences

Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452082116-16770-1-git-send-email-michel.thierry@intel.com
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove commit_plane function pointer.

With sprites, cursors and primary planes taking the atomic state
this is now unused. It's removed in a separate commit to allow
a revert.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-8-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Use plane state for primary plane updates.

Pass in the atomic states to allow for proper updates.
This removes uses of intel_crtc->config and direct access
to plane->state.

This breaks the last bit of kgdboc, but that appears to be dead code.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-7-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Use the plane state for cursor updates.

Cursor planes grab the state from plane->state instead of the state
that was passed. The only updates are atomic now, so use the plane_state
that's passed in.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-6-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Make disable_cursor_plane similar to commit_cursor_plane.

Update cursor_addr when disable_plane is called.
This is required to make commit_cursor_plane take a
crtc_state and a plane_state.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-5-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Remove some visibility checks from intel_crtc_update_cursor.

This is duplicated with intel_check_cursor_plane, and with all
non-atomic paths removed this should be dead code.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-4-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Do not use commit_plane for sprite planes.

Use update_plane and disable_plane directly.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-3-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Use passed plane state for sprite planes, v4.

Don't use plane->state directly, use the pointer from commit_plane.

Changes since v1:
- Fix uses of plane->state->rotation and color key to use the passed state too.
- Only pass crtc_state and plane_state to update_plane.
Changes since v2:
- Rebased.
Changes since v3:
- Small whitespace changes and only assign 1 variable per line.
- Constify plane_state and crtc_state. (vsyrjala)

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452164052-21752-2-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Add two-stage ILK-style watermark programming (v10)

In addition to calculating final watermarks, let's also pre-calculate a
set of intermediate watermark values at atomic check time.  These
intermediate watermarks are a combination of the watermarks for the old
state and the new state; they should satisfy the requirements of both
states which means they can be programmed immediately when we commit the
atomic state (without waiting for a vblank).  Once the vblank does
happen, we can then re-program watermarks to the more optimal final
value.

v2: Significant rebasing/rewriting.

v3:
 - Move 'need_postvbl_update' flag to CRTC state (Daniel)
 - Don't forget to check intermediate watermark values for validity
   (Maarten)
 - Don't due async watermark optimization; just do it at the end of the
   atomic transaction, after waiting for vblanks.  We do want it to be
   async eventually, but adding that now will cause more trouble for
   Maarten's in-progress work.  (Maarten)
 - Don't allocate space in crtc_state for intermediate watermarks on
   platforms that don't need it (gen9+).
 - Move WaCxSRDisabledForSpriteScaling:ivb into intel_begin_crtc_commit
   now that ilk_update_wm is gone.

v4:
 - Add a wm_mutex to cover updates to intel_crtc->active and the
   need_postvbl_update flag.  Since we don't have async yet it isn't
   terribly important yet, but might as well add it now.
 - Change interface to program watermarks.  Platforms will now expose
   .initial_watermarks() and .optimize_watermarks() functions to do
   watermark programming.  These should lock wm_mutex, copy the
   appropriate state values into intel_crtc->active, and then call
   the internal program watermarks function.

v5:
 - Skip intermediate watermark calculation/check during initial hardware
   readout since we don't trust the existing HW values (and don't have
   valid values of our own yet).
 - Don't try to call .optimize_watermarks() on platforms that don't have
   atomic watermarks yet.  (Maarten)

v6:
 - Rebase

v7:
 - Further rebase

v8:
 - A few minor indentation and line length fixes

v9:
 - Yet another rebase since Maarten's patches reworked a bunch of the
   code (wm_pre, wm_post, etc.) that this was previously based on.

v10:
 - Move wm_mutex to dev_priv to protect against racing commits against
   disjoint CRTC sets. (Maarten)
 - Drop unnecessary clearing of cstate->wm.need_postvbl_update (Maarten)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452108870-24204-1-git-send-email-matthew.d.roper@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: shut up gen8+ SDE irq dmesg noise, again

We still keep getting

[    4.249930] [drm:gen8_irq_handler [i915]] *ERROR* The master control interrupt lied (SDE)!

This reverts

commit 820da7ae46332fa709b171eb7ba57cbd023fa6df
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Nov 25 16:47:23 2015 +0200

    Revert "drm/i915: shut up gen8+ SDE irq dmesg noise"

which in itself is a revert, so this is just doing

commit 97e5ed1111dcc5300a0f59a55248cd243937a8ab
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Oct 23 10:56:12 2015 +0200

    drm/i915: shut up gen8+ SDE irq dmesg noise

all over again. I'll stop pretending I understand what's going on like I
did when I thought I'd fixed this for good in

commit 6a39d7c986be4fd18eb019e9cdbf774ec36c9f77
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Nov 25 16:47:22 2015 +0200

    drm/i915: fix the SDE irq dmesg warnings properly

Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Reference: http://mid.gmane.org/20151213124945.GA5715@nuc-i3427.alporthouse.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92084
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: 820da7ae4633 ("Revert "drm/i915: shut up gen8+ SDE irq dmesg noise"")
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452155350-14658-1-git-send-email-jani.nikula@intel.com

drm/i915: Allow fuzzy matching in intel_compare_link_m_n

This prevents a unnecessary modeset on a dell XPS 13 (2016).

N is always a power of 2, which means that for fuzzy matching we should
compare for inequality on the n values, then do fuzzy matching on the m
values.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Tested-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/568D0E93.304@linux.intel.com

drm/i915: Sanitize watermarks after hardware state readout (v4)

Although we can do a good job of reading out hardware state, the
graphics firmware may have programmed the watermarks in a creative way
that doesn't match how i915 would have chosen to program them.  We
shouldn't trust the firmware's watermark programming, but should rather
re-calculate how we think WM's should be programmed and then shove those
values into the hardware.

We can do this pretty easily by creating a dummy top-level state,
running it through the check process to calculate all the values, and
then just programming the watermarks for each CRTC.

v2:  Move watermark sanitization after our BIOS fb reconstruction; the
     watermark calculations that we do here need to look at pstate->fb,
     which isn't setup yet in intel_modeset_setup_hw_state(), even
     though we have an enabled & visible plane.

v3:
 - Don't move 'active = optimal' watermark assignment; we just undo
   that change in the next patch anyway.  (Ville)
 - Move atomic helper locking fix to separate patch.  (Maarten)

v4:
 - Grab connection_mutex before calling atomic helper to duplicate
   state.  The connector loop inside the helper will throw a WARN
   if we don't hold something to protect the connector list (and the
   helper itself doesn't try to lock the list).
 - Make failure to calculate watermarks for inherited state a WARN()
   since it probably indicates a serious problem in either our state
   readout code or our watermark code for this platform.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Add extra paranoia to ILK watermark calculations

Our low-level watermark calculation functions don't get called when the
CRTC is disabled or the relevant plane is invisible, so they should
never see a zero htotal or zero bpp.  However add some checks to ensure
this is true so that we don't wind up dividing by zero if we make a
mistake elsewhere in the driver (which the atomic watermark series has
revealed we might be).

References: http://lists.freedesktop.org/archives/intel-gfx/2015-October/077370.html
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449171462-30763-6-git-send-email-matthew.d.roper@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Convert hsw_compute_linetime_wm to use in-flight state

When watermark calculation was moved up to the atomic check phase, the
code was updated to calculate based on in-flight atomic state rather
than already-committed state.  However the hsw_compute_linetime_wm()
didn't get updated and continued to pull values out of the
currently-committed CRTC state.  On platforms that call this function
(HSW/BDW only), this will cause problems when we go to enable the CRTC
since we'll pull the current mode (off) rather than the mode we're
calculating for and wind up with a divide by zero error.

This was an oversight in commit:

        commit a28170f3389f4e42db95e595b0d86384a79de696
        Author: Matt Roper <matthew.d.roper@intel.com>
        Date:   Thu Sep 24 15:53:16 2015 -0700

            drm/i915: Calculate ILK-style watermarks during atomic check (v3)

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449171462-30763-5-git-send-email-matthew.d.roper@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Setup clipped src/dest coordinates during FB reconstruction (v2)

Plane state objects contain two copies of src/dest coordinates:  the
original (requested by userspace) coordinates in the base
drm_plane_state object, and a second, clipped copy (i.e., what we
actually want to program to the hardware) in intel_plane_state.  We've
only been setting up the former set of values during boot time FB
reconstruction, but we should really be initializing both.

Note that the code here probably still needs some more work since we
make a lot of assumptions about how the BIOS programmed the hardware
that may not always be true, especially on gen9+; e.g.,
 * Primary plane might not be positioned at 0,0
 * Primary plane could have been rotated by the BIOS
 * Primary plane might be scaled
 * The BIOS fb might be a single "extended mode" FB that spans
   multiple displays.
 * ...etc...

v2: Reword/expand commit message description of assumptions we make

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by(v1): Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1449171462-30763-4-git-send-email-matthew.d.roper@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Restore inhibiting the load of the default context

Following a GPU reset, we may leave the context in a poorly defined
state, and reloading from that context will leave the GPU flummoxed. For
secondary contexts, this will lead to that context being banned - but
currently it is also causing the default context to become banned,
leading to turmoil in the shared state.

This is a regression from

commit 6702cf16e0ba8b0129f5aa1b6609d4e9c70bc13b [v4.1]
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Mon Mar 16 16:00:58 2015 +0000

    drm/i915: Initialize all contexts

which quietly introduced the removal of the MI_RESTORE_INHIBIT on the
default context.

v2: Mark the global default context as uninitialized on GPU reset so
that the context-local workarounds are reloaded upon re-enabling.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1448630935-27377-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: stable@vger.kernel.org
[danvet: This seems to fix a gpu hand on after the first resume,
resulting in any future suspend operation failing with -EIO because
the gpu seems to be in a funky state. Somehow this patch fixes that.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Tune down rpm wakelock debug checks

They're causing massive amounts of dmesg noise and hence CI noise all
over the place. Enabling them for a bit was good enough to refresh our
task list of what's still needed to enable rpm by default.

To make sure we're not forgetting to make this noisy again add a FIXME
comment.

Fixes: da5827c36607 ("drm/i915: add assert_rpm_wakelock_held helper")
Cc: Imre Deak <imre.deak@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1452012847-4737-1-git-send-email-daniel.vetter@ffwll.ch
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Allow use of get_dma_address for stolen backed objects

i915_gem_object_get_dma_address function is used to retrieve the dma address
of a particular page so as to map it in a given GTT entry for CPU access.
This function would be used for stolen backed objects also for tasks like
pwrite,  clearing of the pages etc. So the obj->get_page.sg needs to be
initialized for the stolen objects also.

Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450765253-32104-2-git-send-email-ankitprasad.r.sharma@intel.com
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: add kerneldoc for intel_lr_context_size()

This function was recently renamed & exposed, so now it gets documented

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1451996493-16079-1-git-send-email-david.s.gordon@intel.com

drm/i915/dsi: be defensive about out of bounds operation byte

Untie the VBT based generic panel driver from the VBT parsing, so that
the two don't have to be updated in lockstep.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/01c71ac89a9db8bc7b8ae0fb05c50a5fae362dc4.1450702954.git.jani.nikula@intel.com

drm/i915/dsi: be defensive about out of bounds sequence id

Untie the VBT based generic panel driver from the VBT parsing, so that
the two don't have to be updated in lockstep.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/7a6e3e7c4404c0e4dbcf003acd8737a6ecbe218f.1450702954.git.jani.nikula@intel.com

drm/i915/bios: rewrite sequence block parsing

Make everything a bit more readable and clear.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/e8f2a62d78d90981a6b49fdf9ab3594f60a46033.1450702954.git.jani.nikula@intel.com

drm/i915/bios: abstract finding the panel sequence block

Make the whole thing easier to read. While at it, make the parsing more
robust, and ensure we don't read past buffer being parsed.

v2: improve commit message (Daniel)

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452001851-8967-1-git-send-email-jani.nikula@intel.com

drm/i915/bios: have get_blocksize() support MIPI sequence block v3+

Have get_blocksize() support the special case of MIPI sequence block v3+
which has a separate field for size. Provide and use abstractions for
getting the blocksize given a pointer to the block "envelope",
i.e. pointer to the block id, and given a pointer to the block payload
data.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/e935bd5e119a83dd91214c47e6cd4f6ce8b2a17e.1450702954.git.jani.nikula@intel.com

drm/i915/bios: split the MIPI DSI VBT block parsing to two

There's two blocks to parse, have one function per block. The existing
one cuts neatly into two.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/6c9598e2b4d07e8d264617cdfe8b6527a74261f7.1450702954.git.jani.nikula@intel.com

drm/i915/bios: fix header define name for intel_bios.h

Just for OCD.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/730e41760133dbaa1e3ab1b91631ada18676810c.1450702954.git.jani.nikula@intel.com

drm/i915/guc: Add GuC ADS - enabling ADS

Set ADS enabling flag during GuC init.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450468812-4882-6-git-send-email-yu.dai@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Add GuC ADS - MMIO reg state

GuC needs to know which registers and how they will be saved and
restored during event such as engine reset or power state changes.
For now only the base address of reg state is initialized. The
detail register table probably will be setup in future GuC TDR or
Preemption patch series.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450468812-4882-5-git-send-email-yu.dai@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Add GuC ADS - scheduler policies

GuC supports different scheduling policies for its four internal
queues. Currently these have been set to the same default values
as KMD_NORMAL queue.

Particularly POLICY_MAX_NUM_WI is set to 15 to match GuC internal
maximum submit queue numbers to avoid an out-of-space problem.
This value indicates max number of work items allowed to be queued
for one DPC process. A smaller value will let GuC schedule more
frequently while a larger number may increase chances to optimize
cmds (such as collapse cmds from same lrc) with risks that keeps
CS idle.

v1: tidy up code

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450468812-4882-4-git-send-email-yu.dai@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Add GuC ADS (Addition Data Structure) - allocation

The GuC firmware uses this for various purposes. The ADS itself is
a chunk of memory created by driver to share with GuC. Its members
are usually addresses telling where GuC to access them, including
things like scheduler policies, register list that will be saved
and restored during reset etc.

This is the first patch of a series to enable GuC ADS. For now, we
only create the ADS obj whilst keep it disabled.

v1: remove dead code checking return of kmap_atomic (Chris Wilson)
v2: use kmap instead of the atomic version of it.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450468812-4882-3-git-send-email-yu.dai@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Expose (intel)_lr_context_size()

The GuC code needs to know the size of a logical context, so we
expose get_lr_context_size(), renaming it intel_lr_context__size()
to fit the naming conventions for nonstatic functions.

For: VIZ-2021
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450468812-4882-2-git-send-email-yu.dai@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Move GuC wq_check_space to alloc_request_extras

Split GuC work queue space checking from submission and move it to
ring_alloc_request_extras. The reason is that failure in later
i915_add_request() won't be handled. In the case timeout happens,
driver can return early in order to handle the error.

v1: Move wq_reserve_space to ring_reserve_space
v2: Move wq_reserve_space to alloc_request_extras (Chris Wilson)
v3: The work queue head pointer is cached by driver now. So we can
    quickly return if space is available.
    s/reserve/check/g (Dave Gordon)
v4: Update cached wq head after ring doorbell; check wq space before
    ring doorbell in case unexpected error happens; call wq space
    check only when GuC submission is enabled. (Dave Gordon)

Signed-off-by: Alex Dai <yu.dai@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450295155-10050-1-git-send-email-yu.dai@intel.com
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Disable shrinker for non-swapped backed objects

If the system has no available swap pages, we cannot make forward
progress in the shrinker by releasing active pages, only by releasing
purgeable pages which are immediately reaped. Take total_swap_pages into
account when counting up available objects to be shrunk and subsequently
shrinking them. By doing so, we avoid unbinding objects that cannot be
shrunk and so wasting CPU cycles flushing those objects from the GPU to
the system and then immediately back again (as they will more than
likely be reused shortly after).

Based on a patch by Akash Goel.

v2: frontswap registers extra swap pages available for the system, so it
is already include in the count of available swap pages.

v3: Use get_nr_swap_pages() to query the currently available amount of
swap space. This should also stop us from shrinking the GPU buffers if
we ever run out of swap space. Though at that point, we would expect the
oom-notifier to be running and failing miserably...

Reported-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: linux-mm@kvack.org
Cc: Akash Goel <akash.goel@intel.com>
Cc: sourab.gupta@intel.com
Link: http://patchwork.freedesktop.org/patch/msgid/1449244734-25733-2-git-send-email-chris@chris-wilson.co.uk
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

mm: Export nr_swap_pages

Some modules, like i915.ko, use swappable objects and may try to swap
them out under memory pressure (via the shrinker). Before doing so, they
want to check using get_nr_swap_pages() to see if any swap space is
available as otherwise they will waste time purging the object from the
device without recovering any memory for the system. This requires the
nr_swap_pages counter to be exported to the modules.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: "Goel, Akash" <akash.goel@intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: linux-mm@kvack.org
Link: http://patchwork.freedesktop.org/patch/msgid/1449244734-25733-1-git-send-email-chris@chris-wilson.co.uk
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Avoid writing relocs with addresses in non-canonical form

According to PRM, some parts of HW require the addresses to be in
a canonical form, where bits [63:48] == [47]. Let's convert addresses to
canonical form prior to relocating and return converted offsets to
userspace. We also need to make sure that userspace is using addresses
in canonical form in case of softpin.

v2: Whitespace fixup, gen8_canonical_addr description (Chris, Ville)
v3: Rebase on top of softpin, fix a hole in relocate_entry,
    s/expect/require (Chris)
v4: Handle softpin in validate_exec_list (Chris)
v5: Convert back to canonical form at copy_to_user time (Chris)
v6: Don't use struct exec_object2 in place of exec_object
v7: Use sign_extend64 for converting to canonical form (Joonas),
    reject non-canonical and non-page-aligned offset for softpin (Chris)
v8: Convert back to non-canonical form in a function,
    split the test for EXEC_OBJECT_PINNED (Chris)
v9: s/canonial/canonical, drop accidental double newline (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1451409892-13708-1-git-send-email-michal.winiarski@intel.com
Testcase: igt/gem_bad_reloc/negative-reloc-blt
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92699
Cc: drm-intel-fixes@lists.freedesktop.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

i915: correctly handling failed allocation

Since devm_kzalloc can be failed, it needs to be checked
if not, NULL dereference could be happened.

Signed-off-by: Insu Yun <wuninsu@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1451491169-35068-1-git-send-email-wuninsu@gmail.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Fix whitespace (trivial)

Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1451427643-7266-1-git-send-email-benjamin.widawsky@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: increase the tries for HDMI hotplug live status checking

The total delay of HDMI hotplug detecting with 30ms is sometimes not
enoughtfor HDMI live status up with specific HDMI monitors in BSW platform.

After doing experiments for following monitors, it needs 80ms at least
for those worst cases.

Lenovo L246 1xwA (4 failed, necessary hot-plug delay: 58/40/60/40ms)
Philips HH2AP (9 failed, necessary hot-plug delay: 80/50/50/60/46/40/58/58/39ms)
BENQ ET-0035-N (6 failed, necessary hot-plug delay: 60/50/50/80/80/40ms)
DELL U2713HM (2 failed, necessary hot-plug delay: 58/59ms)
HP HP-LP2475w (5 failed, necessary hot-plug delay: 70/50/40/60/40ms)

It looks like 70-80 ms is BSW platform needs in some bad cases of the
monitors at this end (8 times delay at most). Keep less than 100ms for
HDCP pulse HPD low (with at least 100ms) to respond a plug out.

Reviewed-by: Cooper Chiou <cooper.chiou@intel.com>
Tested-by: Gary Wang <gary.c.wang@intel.com>
Cc: Gavin Hindman <gavin.hindman@intel.com>
Cc: Sonika Jindal <sonika.jindal@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Shobhit Kumar <shobhit.kumar@intel.com>
Signed-off-by: Gary Wang <gary.c.wang@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450858295-12804-1-git-send-email-gary.c.wang@intel.com
Tested-by: Shobhit Kumar <shobhit.kumar@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: 237ed86c693d ("drm/i915: Check live status before reading edid")
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Calculate visibility in check_plane correctly regardless of dpms.

When the crtc is configured but not active we currently clip to (0,0)x(0,0).
This results in differences in calculations depending on dpms setting.
When the crtc is enabled but not active run check_plane as if it were on,
but afterwards set plane_state->visible = false for the checks.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447945645-32005-13-git-send-email-maarten.lankhorst@linux.intel.com

drm/i915: Keep track of the cdclk as if all crtc's were active.

On skylake when calculating plane visibility with the crtc in
dpms off mode the real cdclk may be different from what it would be
if the crtc was active. This may result in a WARN_ON(cdclk < crtc_clock)
from skl_max_scale. The fix is to keep a atomic_cdclk that would be true
if all crtc's were active.

This is required to get the same calculations done correctly regardless
of dpms mode.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447945645-32005-12-git-send-email-maarten.lankhorst@linux.intel.com

drm/i915: Do not acquire crtc state to check clock during modeset, v4.

Parallel modesets are still not allowed, but this will allow updating
a different crtc during a modeset if the clock is not changed.

Additionally when all pipes are DPMS off the cdclk will be lowered
to the minimum allowed.

Changes since v1:
- Add dev_priv->active_crtcs for tracking which crtcs are active.
- Rename min_cdclk to min_pixclk and move to dev_priv.
- Add a active_crtcs mask which is updated atomically.
- Add intel_atomic_state->modeset which is set on modesets.
- Commit new pixclk/active_crtcs right after state swap.
Changes since v2:
- Make the changes related to max_pixel_rate calculations more readable.
Changes since v3:
- Add cherryview and missing WARN_ON to readout.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/skl: Do not allow scaling when crtc is disabled.

This fixes a warning when the crtc is turned off. In that case fb
will be NULL, and crtc_clock will be 0. Because the crtc is no longer
active this is not a bug, and shouldn't trigger the WARN_ON.

Also remove handling a null crtc_state, with all transitional helpers
gone this can no longer happen.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1448360945-5723-2-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/bios: add proper documentation for the Video BIOS Table (VBT)

Add an overview and documentation for the VBT/BDB header structures.

Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/3d826d4600688ca3518713776ab5bd8a8fc9f20f.1450702954.git.jani.nikula@intel.com

drm/i915: Unbreak check_digital_port_conflicts()

Atomic changes broke check_digital_port_conflicts(). It needs to look
at the global situation instead of just trying to find a conflict
within the current atomic state.

This bug made my HSW explode spectacularly after I had split the DDI
encoders into separate DP and HDMI encoders. With the fix, things
seem much more solid.

I hope holding the connection_mutex is enough protection that we can
actually walk the connectors even if they're not part of the current
atomic state...

v2: Regenerate the patch so that it actually applies (Jani)

Cc: stable@vger.kernel.org
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Fixes: 5448a00d3f06 ("drm/i915: Don't use staged config in check_digital_port_conflicts()")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1449764551-12466-1-git-send-email-ville.syrjala@linux.intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: Compile-time concatenate WARN_ON macro strings

Using __stringify(x) instead of #x adds support for macros as
a parameter and compile-time concatenation reduces the runtime
overhead.

Slightly increases the .text size but should not matter.

v2:
- Define I915_STATE_WARN_ON though I915_STATE_WARN
  (Bikeshed inspiration by Chris)

v3:
- More specific commit message

v4:
- Do not directly pass arbitary string as format, instead
  guard with "%s" (Dave)

Cc: Rob Clark <robdclark@gmail.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v3)
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450441647-23924-3-git-send-email-joonas.lahtinen@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Simplify _STATE_ debug macros

Take advantage of WARN return value to simplify the flow.

Cc: Rob Clark <robdclark@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450441647-23924-2-git-send-email-joonas.lahtinen@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove obsolete code from intelfb_alloc()

Clean up after 0c82312f3f15 ("drm/i915: Pin the ifbdev for the
info->system_base GGTT mmapping"):

At each of the remaining "goto out" in intelfb_alloc(), fb can only be
either an ERR_PTR or NULL, so the call to drm_framebuffer_unreference()
is now obsolete.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Link: http://patchwork.freedesktop.org/patch/msgid/56756c41.c306c20a.d0602.1830SMTPIN_ADDED_MISSING@mx.google.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Workaround CHV pipe C cursor fail

Turns out CHV pipe C was glued on somewhat poorly, and there's something
wrong with the cursor. If the cursor straddles the left screen edge,
and is then moved away from the edge or disabled, the pipe will often
underrun. If enough underruns are triggered quickly enough the pipe
will fall over and die (it just scans out a solid color and reports
a constant underrun). We need to turn the disp2d power well off and
on again to recover the pipe.

None of that is very nice for the user, so let's just refuse to place
the cursor in the compromised position. The ddx appears to fall back
to swcursor when the ioctl returns an error, so theoretically there's
no loss of functionality for the user (discounting swcursor bugs).
I suppose most cursors images actually have the hotspot not exactly
at 0,0 so under typical conditions the fallback will in fact kick in
as soon as the cursor touches the left edge of the screen.

Any atomic compositor should anyway be prepared to fall back to
GPU composition when things don't work out, so there should be no
problem with those.

Other things that I tried to solve this include flipping all
display related clock gating knobs I could find, increasing the
minimum gtt alignment all the way up to 512k. I also tried to see
if there are more specific screen coordinates that hit the bug, but
the findings were somewhat inconclusive. Sometimes the failures
happen almost across the whole left edge, sometimes more at the very
top and around the bottom half. I wasn't able to find any real pattern
to these variations, so it seems our only choice is to just refuse
to straddle the left screen edge at all.

Cc: stable@vger.kernel.org
Cc: Jason Plum <max@warheads.net>
Testcase: igt/kms_chv_cursor_fail
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92826
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450459479-16286-1-git-send-email-ville.syrjala@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Reorder i915_params struct.

Move all the bool variables to the end as per the comment.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450436898-20408-3-git-send-email-joonas.lahtinen@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Decouple struct i915_params i915 into i915_params.h

Otherwise usage in the i915 debug macros yields problems due to
i915_drv.h <-> i915_trace.h <-> intel_drv.h include loops.

v2:
- Document not-so-obvious need for linux/cache.h (Chris)

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450436898-20408-2-git-send-email-joonas.lahtinen@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Limit VF cache invalidate workaround usage to gen9

It is unclear if this is even required on BXT.

v2: Make sure to set the default value to false. Uncertain how my compiler
doesn't complain with v1.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450374597-7021-1-git-send-email-benjamin.widawsky@intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Correct max delay for HDMI hotplug live status checking

The total delay of HDMI hotplug detecting with 30ms have already
been split into a resolution of 3 retries of 10ms each, for the worst
cases. But it still suffered from only waiting 10ms at most in
intel_hdmi_detect(). This patch corrects it by reading hotplug status
with 4 times at most for 30ms delay.

v2:
- straight up to loop execution for more clear in code readability
- mdelay will replace with msleep by Daniel's new patch

drm/i915: mdelay(10) considered harmful

- suggest to re-evaluate try times for being compatible to old HDMI monitor

Reviewed-by: Cooper Chiou <cooper.chiou@intel.com>
Tested-by: Gary Wang <gary.c.wang@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Gavin Hindman <gavin.hindman@intel.com>
Cc: Sonika Jindal <sonika.jindal@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Gary Wang <gary.c.wang@intel.com>
[danvet: fixup conflict with s/mdelay/msleep/ patch.]
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Move Braswell stop_machine GGTT insertion workaround

There was a silent conflict between

commit 0a878716265e9af9f697264dc2e858fcc060d833
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Oct 15 14:23:01 2015 +0200

    drm/i915: restore ggtt double-bind avoidance

and

commit 5bab6f60cb4d1417ad7c599166bcfec87529c1a2
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Oct 23 18:43:32 2015 +0100

    drm/i915: Serialise updates to GGTT with access through GGTT on Braswell

thankfully caught by the extra WARN safegaurd in 0a878716. Since we now
override the GGTT insert_pages callback when installing the aliasing
ppgtt, we assert that the callback is the original ggtt routine.
However, on Braswell we now use a different insertion routine to
serialise access through the GGTT with updating the PTE and hence the
conflict. To avoid the conflict, move the custom insertion routine for
Braswell down a level.

Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447859979-20107-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Tested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Apply broader WaRsDisableCoarsePowerGating for guc also

commit 344df9809f45 ("drm/i915/skl: Disable coarse power gating up until F0")
failed to take into account that the same workaround is used in guc
when forcewake is sampled.

Wrap the condition check inside a macro and use it in both places
to fix the guc side scope.

Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450286318-6854-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Update DRIVER_DATE to 20151218

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

drm/i915/skl: Default to noncoherent access up to F0

The workarounds for disabling hdc invalidation and also forcing
context to be non coherent, are advised to be used up until rev D0.

However as it was found that rev F0, without the
WaForceEnableNonCoherent might system hang if the mesa
tried to use coherent mode.

As these two workarounds are about non coherent access, are
grouped in scope and they point the same HSD, increase the
scope of both to set default behaviour to non coherent access.

References: HSD: gen9lp/2131413
References: http://lists.freedesktop.org/archives/mesa-dev/2015-November/101515.html
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Cc: Francisco Jerez <currojerez@riseup.net>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450448093-22906-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Only spin whilst waiting on the current request

Limit busywaiting only to the request currently being processed by the
GPU. If the request is not currently being processed by the GPU, there
is a very low likelihood of it being completed within the 2 microsecond
spin timeout and so we will just be wasting CPU cycles.

v2: Check for logical inversion when rebasing - we were incorrectly
checking for this request being active, and instead busywaiting for
when the GPU was not yet processing the request of interest.

v3: Try another colour for the seqno names.
v4: Another colour for the function names.

v5: Remove the forced coherency when checking for the active request. On
reflection and plenty of recent experimentation, the issue is not a
cache coherency problem - but an irq/seqno ordering problem (timing issue).
Here, we do not need the w/a to force ordering of the read with an
interrupt.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Cc: "Rantala, Valtteri" <valtteri.rantala@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1449833608-22125-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Limit the busy wait on requests to 5us not 10ms!

When waiting for high frequency requests, the finite amount of time
required to set up the irq and wait upon it limits the response rate. By
busywaiting on the request completion for a short while we can service
the high frequency waits as quick as possible. However, if it is a slow
request, we want to sleep as quickly as possible. The tradeoff between
waiting and sleeping is roughly the time it takes to sleep on a request,
on the order of a microsecond. Based on measurements of synchronous
workloads from across big core and little atom, I have set the limit for
busywaiting as 10 microseconds. In most of the synchronous cases, we can
reduce the limit down to as little as 2 miscroseconds, but that leaves
quite a few test cases regressing by factors of 3 and more.

The code currently uses the jiffie clock, but that is far too coarse (on
the order of 10 milliseconds) and results in poor interactivity as the
CPU ends up being hogged by slow requests. To get microsecond resolution
we need to use a high resolution timer. The cheapest of which is polling
local_clock(), but that is only valid on the same CPU. If we switch CPUs
because the task was preempted, we can also use that as an indicator that
 the system is too busy to waste cycles on spinning and we should sleep
instead.

__i915_spin_request was introduced in
commit 2def4ad99befa25775dd2f714fdd4d92faec6e34 [v4.2]
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Apr 7 16:20:41 2015 +0100

     drm/i915: Optimistically spin for the request completion

v2: Drop full u64 for unsigned long - the timer is 32bit wraparound safe,
so we can use native register sizes on smaller architectures. Mention
the approximate microseconds units for elapsed time and add some extra
comments describing the reason for busywaiting.

v3: Raise the limit to 10us
v4: Now 5us.

Reported-by: Jens Axboe <axboe@kernel.dk>
Link: https://lkml.org/lkml/2015/11/12/621
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Cc: "Rantala, Valtteri" <valtteri.rantala@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1449833608-22125-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Break busywaiting for requests on pending signals

The busywait in __i915_spin_request() does not respect pending signals
and so may consume the entire timeslice for the task instead of
returning to userspace to handle the signal.

In the worst case this could cause a delay in signal processing of 20ms,
which would be a noticeable jitter in cursor tracking. If a higher
resolution signal was being used, for example to provide fairness of a
server timeslices between clients, we could expect to detect some
unfairness between clients (i.e. some windows not updating as fast as
others). This issue was noticed when inspecting a report of poor
interactivity resulting from excessively high __i915_spin_request usage.

Fixes regression from
commit 2def4ad99befa25775dd2f714fdd4d92faec6e34 [v4.2]
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Apr 7 16:20:41 2015 +0100

     drm/i915: Optimistically spin for the request completion

v2: Try to assess the impact of the bug

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc; "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Eero Tamminen <eero.t.tamminen@intel.com>
Cc: "Rantala, Valtteri" <valtteri.rantala@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1449833608-22125-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: don't enable autosuspend on platforms without RPM support

pm_runtime_{use,dont_use}_autosuspend() controls whether the device's
sysfs power/autosuspend_delay_ms file is writeable or returns -EIO on
access to user space. Since

commit 25b181b46e4340f69793a886f2cd46608487783b
Author: Imre Deak <imre.deak@intel.com>
Date:   Thu Dec 17 13:44:56 2015 +0200

  drm/i915: get a permanent RPM reference on platforms w/o RPM support

this sysfs file is writeable also on platforms without RPM support, but
userspace (at least IGT) depends on this file being unchangable to
determine whether the device supports runtime PM at all. So restore the
old behavior.

This gets rid of igt/pm_rpm failures on old platforms without RPM
support, where the test should be skipped.

Testcase: igt/pm_rpm/basic-rte
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450371873-878-1-git-send-email-imre.deak@intel.com

drm/i915/backlight: prefer dev_priv over dev pointer

Use dev_priv rather than dev pointer where applicable. Remove plenty of
unnecessary temp variables. No functional changes.

Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1450428695-28831-1-git-send-email-jani.nikula@intel.com