Chris Wilson [Thu, 2 Aug 2018 10:06:28 +0000 (11:06 +0100)]
drm/i915: Drop stray clearing of rps->last_adj
We used to reset last_adj to 0 on crossing a power domain boundary, to
slow down our rate of change. However, commit 8e3ae8f3d4b3 ("drm/i915:
Interactive RPS mode") accidentally caused it to be reset on every
frequency update, nerfing the fast response granted by the slow start
algorithm.
Chris Wilson [Mon, 30 Jul 2018 16:43:25 +0000 (17:43 +0100)]
drm/i915/execlists: Terminate the context image with BB_END
In the aub trace utility, the context images are terminated with a
MI_BATCH_BUFFER_END; the simulator is reported as complaining otherwise.
Do the same for our protocontext image for completeness, and in passing
apply the magic bit for gen10 to mark the end of the context image.
The register for 0xe420 is unable to hold any value, including
this bit. The documentation is also mixed between having a
register bit for toggle and having a state command setup
for it. Apparently the register toggle is deprecated.
Remove the register toggle as evidence shows it's futile.
The thing remaining is an apology and humble request for
Mesa folks to resurrect their state setup for this as they
were on right track from start.
Chris Wilson [Wed, 1 Aug 2018 10:47:21 +0000 (11:47 +0100)]
drm/i95: Mark GGTT as incoherent for gen10+
The evidence suggests that we need to start treating writes via GGTT as
incoherent for gen10+, that is that they are internally buffered and not
immediately visible via a read along a different physical path.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107398
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107400
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107435 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180801104721.4030-1-chris@chris-wilson.co.uk
Chris Wilson [Tue, 31 Jul 2018 13:26:29 +0000 (14:26 +0100)]
drm/i915: Interactive RPS mode
RPS provides a feedback loop where we use the load during the previous
evaluation interval to decide whether to up or down clock the GPU
frequency. Our responsiveness is split into 3 regimes, a high and low
plateau with the intent to keep the gpu clocked high to cover occasional
stalls under high load, and low despite occasional glitches under steady
low load, and inbetween. However, we run into situations like kodi where
we want to stay at low power (video decoding is done efficiently
inside the fixed function HW and doesn't need high clocks even for high
bitrate streams), but just occasionally the pipeline is more complex
than a video decode and we need a smidgen of extra GPU power to present
on time. In the high power regime, we sample at sub frame intervals with
a bias to upclocking, and conversely at low power we sample over a few
frames worth to provide what we consider to be the right levels of
responsiveness respectively. At low power, we more or less expect to be
kicked out to high power at the start of a busy sequence by waitboosting.
Prior to commit 8ae02596dfe9 ("drm/i915: Avoid waitboosting on the active
request") whenever we missed the frame or stalled, we would immediate go
full throttle and upclock the GPU to max. But in commit 8ae02596dfe9, we
relaxed the waitboosting to only apply if the pipeline was deep to avoid
over-committing resources for a near miss. Sadly though, a near miss is
still a miss, and perceptible as jitter in the frame delivery.
To try and prevent the near miss before having to resort to boosting
after the fact, we use the pageflip queue as an indication that we are
in an "interactive" regime and so should sample the load more frequently
to provide power before the frame misses it vblank. This will make us
more favorable to providing a small power increase (one or two bins) as
required rather than going all the way to maximum and then having to
work back down again. (We still keep the waitboosting mechanism around
just in case a dramatic change in system load requires urgent uplocking,
faster than we can provide in a few evaluation intervals.)
v2: Reduce rps_set_interactive to a boolean parameter to avoid the
confusion of what if they wanted a new power mode after pinning to a
different mode (which to choose?)
v3: Only reprogram RPS while the GT is awake, it will be set when we
wake the GT, and while off warns about being used outside of rpm.
v4: Fix deferred application of interactive mode
v5: s/state/interactive/
v6: Group the mutex with its principle in a substruct
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107111 Fixes: 8ae02596dfe9 ("drm/i915: Avoid waitboosting on the active request") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180731132629.3381-1-chris@chris-wilson.co.uk
Matthew Auld [Mon, 30 Jul 2018 12:05:44 +0000 (13:05 +0100)]
drm/i915/gtt: remove px_page
Entries will either be pointing to scratch or real PD, making the
px_page(pd) check pointless. Also since there are no other users of
px_page, just remove it.
Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michel Thierry <michel.thierry@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180730120544.20784-1-matthew.auld@intel.com
Chris Wilson [Mon, 30 Jul 2018 07:53:51 +0000 (08:53 +0100)]
drm/i915/selftests: Replace opencoded clflush with drm_clflush_virt_range
We occasionally see that the clflush prior to a read of GPU data is
returning stale data, reminiscent of much earlier bugs fixed by adding a
second clflush for serialisation. As drm_clflush_virt_range() already
supplies the workaround, use it rather than open code the clflush
instruction.
Chris Wilson [Mon, 30 Jul 2018 07:53:50 +0000 (08:53 +0100)]
drm/i915: Kick waiters on resetting legacy rings
For reasons unknown, interrupts following a reset do not arrive, but
this can be papered over by kicking any waiter and peeking at the
breadcrumbs following the reset.
Chris Wilson [Thu, 26 Jul 2018 16:15:27 +0000 (17:15 +0100)]
drm/i915: Downgrade Gen9 Plane WM latency error
According to intel_read_wm_latency() it is perfectly legal for one WM
and all subsequent levels to be 0 (and the deeper powersaving states
disabled), so don't shout *ERROR*, over and over again.
Paulo Zanoni [Thu, 7 Jun 2018 23:07:00 +0000 (16:07 -0700)]
drm/i915: inline skl_copy_ddb_for_pipe() to its only caller
While things may have been different before, right now the function is
very simple and has a single caller. IMHO any possible benefits from
an abstraction here are gone and not worth the price of the current
indirection while reading the code.
Paulo Zanoni [Thu, 26 Jul 2018 00:12:29 +0000 (17:12 -0700)]
drm/i915/icl: don't set CNL_DDI_CLOCK_REG_ACCESS_ON anymore
The new recommendation from the spec is to simply not set this bit
anymore. Not setting the bit would prevent some hangs that our driver
manages to avoid since commit 49544916aa08 ("drm/i915: enable the
pipe/transcoder/planes later on HSW+"), and the theoretical downside
of not setting the bit doesn't seem realistic according to the HW
team. Let's follow their recommendation.
Jakub Bartmiński [Fri, 27 Jul 2018 14:11:47 +0000 (16:11 +0200)]
drm/i915: Add a fault injection point to WOPCM init
Add a fault injection point in the WOPCM initialization path.
v4:
Move the injection inside the WOPCM init function.
Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-5-jakub.bartminski@intel.com
Jakub Bartmiński [Fri, 27 Jul 2018 14:11:46 +0000 (16:11 +0200)]
drm/i915: Remove unnecessary ggtt_offset_bias from i915_gem_context
Since ggtt_offset_bias is now stored in ggtt.pin_bias, it is duplicated
inside i915_gem_context, and can instead be accessed directly from ggtt.
v3:
Added a helper function to retrieve the ggtt.pin_bias from the vma.
v4:
Moved the helper function to the previous patch in the series.
Dropped the bias from intel_ring_pin. This introduces a slight functional
change since we are always pinning the ring a bit higher if GuC is present
even though we don't really need to.
v8:
Fixed patch not applying on the most recent upstream.
Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-4-jakub.bartminski@intel.com
Jakub Bartmiński [Fri, 27 Jul 2018 14:11:45 +0000 (16:11 +0200)]
drm/i915/guc: Move the pin bias value from GuC to GGTT
Removing the pin bias from GuC allows us to not check for GuC every time
we pin a context, which fixes the assertion error on unresolved GuC
platform default in mock contexts selftest.
It also seems that we were using uninitialized WOPCM variables when
setting the GuC pin bias. The pin bias has to be set after the WOPCM,
but before the call to i915_gem_contexts_init where the first contexts
are pinned.
v2:
This also makes it so that there's no need to set GuC variables from
within the WOPCM init function or to move the WOPCM init, while keeping
the correct initialization order. Also for mock tests the pin bias is
left at 0 and we make sure that the pin bias with GuC will not be
smaller than without GuC.
v3:
Avoid unused i915 in intel_guc_ggtt_offset if debug is disabled.
v4:
Squash with WOPCM init reordering.
Moved the i915_ggtt_pin_bias helper to this patch, and made some
functions use it instead of directly dereferencing i915->ggtt.
v5:
Since we now don't use wopcm.guc.base for the pin bias there's no need to
validate it. It also has already been verified in WOPCM init.
v6:
Deleted the now unnecessarily introduced includes from previous versions.
Dropped naming changes from dev_priv to i915 for better patch readability.
v7:
Changed some comments to make more sense in the context they're in.
v8:
Moved and renamed the function which now returns the wopcm.guc.size to
intel_guc.c:intel_guc_reserved_gtt_size to avoid any possible confusion
with the pin_bias in ggtt, which should be used for pinning.
Fixed patch not applying or the most recent upstream.
Fixes: d0492e8fa930 ("drm/i915/uc: Fetch GuC/HuC firmwares from guc/huc specific init")
Testcase: igt/drv_selftest/mock_contexts #GuC Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-3-jakub.bartminski@intel.com
Jakub Bartmiński [Fri, 27 Jul 2018 14:11:44 +0000 (16:11 +0200)]
drm/i915/guc: Do not partition WOPCM if GuC is not used
There seems to be no reason for doing extra work on WOPCM partitioning
in the case GuC is not used, as the partitioning will not be used by the
intel_wopcm_init_hw function anyway.
Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-2-jakub.bartminski@intel.com
Jakub Bartmiński [Fri, 27 Jul 2018 14:11:43 +0000 (16:11 +0200)]
drm/i915/guc: Avoid wasting memory on incorrect GuC pin bias
It would appear that the calculated GuC pin bias was larger than it should
be, as the GuC address space does NOT contain the "HW contexts RSVD" part
of the WOPCM. Thus, the GuC pin bias is simply the GuC WOPCM size.
v5:
Clarify the diagram to better represent the GuC address space.
Since we now don't use guc.base for the pin bias there's no need to
validate it. It also has already been verified in WOPCM init.
Bspec: 1180
Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Michał Winiarski <michal.winiarski@intel.com> Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-1-jakub.bartminski@intel.com
As GEN8_LR_CONTEXT_ALIGN is I915_GTT_MIN_ALIGNMENT is it functionally
equivalent to 0, and we will not be able to reduce the min-alignment for
the GTT, so passing 0 is and will remain equivalent.
Chris Wilson [Fri, 27 Jul 2018 09:18:55 +0000 (10:18 +0100)]
drm/i915: Eliminate use of PAGE_SIZE as a virtual alignment
Using PAGE_SIZE for virtual offset alignment is superfluous as it is
equal to the minimum gtt alignment and so equivalent to 0. It is also
the wrong value to use as we stopped using physical page constructs for
the virtual GTT, i.e. it would be preferrable to use I915_GTT_PAGE_SIZE
and in these cases merely imply I915_GTT_MIN_ALIGNMENT.
Chris Wilson [Thu, 19 Jul 2018 19:47:46 +0000 (20:47 +0100)]
drm/i915/selftests: Exercise resetting in the middle of a wait-on-fence
On older HW, gen2/3, fence registers are used for detiling GPU commands
and as such changing those registers requires serialisation with the
requests on the GPU. Anything running on the GPU is subject to a hang,
and so we must be able to recover cleanly in the middle of a stuck wait
on a fence register.
We can simulate using the fence on the GPU simply by marking the fence
as active on the request for this vma, the interface being common to all
gen, thus broadening the test.
Chris Wilson [Thu, 19 Jul 2018 19:47:45 +0000 (20:47 +0100)]
drm/i915/selftests: Use a full emulation of a user ppgtt context
To test eviction from a ppgtt, we just want a ppgtt i.e. something other
than the Global GTT which is shared and used by the kernel for HW
features like fencing and scanout. However, we also need it to pass
!i915_is_ggtt() and the simplest way is to emulate a full user context
rather than the internal kernel context that is used for the GGTT.
Chris Wilson [Thu, 26 Jul 2018 08:50:33 +0000 (09:50 +0100)]
drm/i915: Don't disable the GPU for older gen on wedging
If we issue a device level GPU reset on the older gen, it will disable
key components of the GMCH and the display engine. The purpose of
wedging is to simply prevent further GEM usage without disabling KMS, so
we need to be careful when we do issue the reset on wedging.
Chris Wilson [Thu, 26 Jul 2018 08:50:32 +0000 (09:50 +0100)]
drm/i915: Restore sane defaults for KMS on GEM error load
If we fail during GEM initialisation, we scrub the HW state by
performing a device level GPU resuet. However, we want to leave the
system in a usable state (with functioning KMS but no GEM) so after
scrubbing the HW state, we need to restore some sane defaults and
re-enable the low-level common parts of the GPU (such as the GMCH).
Chris Wilson [Thu, 26 Jul 2018 10:47:59 +0000 (11:47 +0100)]
drm/i915: Avoid computing tile_row_size() for untiled objects
i915_gem_tile_height() asserts that the object is tiled, but inside the
error printer for the selftest we computed the row size regardless of
tiling, tripping over the assert.
drm/i915/mst: Continue state updates even if AUX writes fail.
We are too late in the enabling sequence to back out cleanly, not updating
state tracking variables, like intel_dp->active_mst_links in this
instance, results in incorrect behaviour further along.
The short pulse handler checks if channel equalization is okay and
goes onto retrain a link if there are active MST links. This retraining
path is not meant for new MST connections, but due to a bug elsewhere, if
active_mst_links is < 0 the boolean check for active_mst_links passes and
we proceed to retrain a new link. This results in a sequence of failed link
training attempts, most likely due to the hardware not setup for link
training at that point i.e., missing the DDI pre_enable sequence.
[ 80.301272] [drm:intel_dp_check_mst_status] channel EQ not ok, retraining
[ 80.301312] [drm:intel_ddi_prepare_link_retrain] *ERROR* Timeout waiting for DDI BUF C idle bit
The above error gives us a hint something went wrong before link
training started.
Check for a positive value of active_mst_links and throw in a warning for
invalid active_mst_links as debug aid.
Paulo Zanoni [Wed, 25 Jul 2018 00:28:13 +0000 (17:28 -0700)]
drm/i915/icl: toggle PHY clock gating around link training
The Gen11 TypeC PHY DDI Buffer chapter, PHY Clock Gating Programming
section says that PHY clock gating should be disabled before starting
voltage swing programming, then enabled after any link training is
complete.
Animesh Manna [Wed, 25 Jul 2018 00:28:11 +0000 (17:28 -0700)]
drm/i915/icl: Update FIA supported lane count for hpd.
In ICL, Flexible IO Adapter (FIA) muxes data and clocks of USB 3.1,
tbt and display controller. In DP alt mode FIA configure the
number of lanes and will be used apart from DPCD read to calculate max
available lanes for DP enablement.
Do like the other functions and check for the status bits. The "Hot
Plug Detection" page from our documentation says we can't just use the
ISR bits on the CPU side (North Display, which has the TC and TBT
modes), so use the correct register: DFLEXDPSP, TC Live State field.
v2: Rebase.
v3:
- Simplify true/false assignment (Rodrigo).
- Reorganize is_gen if ladder (Rodrigo).
- Don't use the ISR for TC/TBT CPU bits.
v4:
- Improve commit message wording (Lucas).
v5:
- COMMIT_LOG_LONG_LINE (Checkpatch).
Cc: Animesh Manna <animesh.manna@intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Lucas De Marchi <lucas.demarchi@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> (v3). Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180725195927.12059-1-paulo.r.zanoni@intel.com
Clarifies the clock recovery loop limit comment that 80
max_cr_tries for pre-DP1.4 devices was chosen as a very
tolerant upper bound.
Assumptions made:
- DP1.4 syncs should be smarter so they won't need more
than 10 tries
- pre-DP1.4 syncs should be compliant enough to not need
that many tries (80) but we should tolerate any that may
trigger this corner case
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Marc Herbert <marc.herbert@intel.com> Suggested-by: Marc Herbert <marc.herbert@intel.com> Signed-off-by: Nathan Ciobanu <nathan.d.ciobanu@linux.intel.com> Reviewed-by: Marc Herbert <marc.herbert@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/1532471612-30001-1-git-send-email-nathan.d.ciobanu@linux.intel.com
Manasi Navare [Thu, 28 Jun 2018 22:35:44 +0000 (15:35 -0700)]
drm/i915/icl: Implement voltage swing programming sequence for MG PHY DDI
This sequence is used to setup voltage swing before enabling MG PHY DDI
as well as for changing the voltage during DisplayPort Link training.
For ICL, there are two types of DDIs. This sequence needs to be used
for MG PHY DDI which is ports C-F.
v6 (From Manasi):
* Add programming for MG_CLKHUB and MG_TX_DCC as per the
spec updates
v5 (from Paulo):
* Checkpatch.
v4 (from Paulo):
* Fix bogus error message
* Fix copy+paste bugs (missing s/TX1/TX2/ after copy+paste)
* Use the new mask names
* Stay under 80 columns
* Add some blank lines
v3:
* Clear the regs before writing (Paulo)
v2:
* Rename to MG PHY in the function def (Jani Nikula)
* Rebase on top of new revision of other patches in series
drm/i915/icl: Add remaining registers and bitfields for MG PHY DDI
This patch adds the remaining register definitions and bit fields
required for MG PHy DDI buffer initializations and voltage
swing programming for MG PHy DDI ports.
While at it this patch also fixes the naming for previously defined
MG PHY registers in original commit id (0d9bd8f8b7b620 "drm/i915/icl:
Add register defs for voltage swing sequences for MG PHY DDI").
Since the MG PHY registers are first defined in ICL platform, there
is no need for _ICL prefix.
v4 (from Paulo): add two white spaces to CRI_CALCINIT too.
v3:
* Fix register names, add spaces for MASK defines, correct the order
of #defines (Paulo)
v2:
* Change the MG_TX_DRVCTL registers names to match the spec (Anusha)
Chris Wilson [Fri, 20 Jul 2018 11:11:02 +0000 (12:11 +0100)]
drm/i915: Show stack (by WARN) for hitting forcewake errors
On Sandybridge, we need a workaround to wait for the CPU thread to wake
up before we are sure that we have enabled the GT power well. However,
we do see the errors being reported and failed reads returning spurious
results. To try and capture more details as it fails, promote the error
into a WARN so we grab the stacktrace, and to try and reduce the
frequency of error increase the timeout from 500us to 5ms.
Chris Wilson [Sat, 21 Jul 2018 12:50:37 +0000 (13:50 +0100)]
drm/i915: Pull unpin map into vma release
A reasonably common operation is to pin the map of the vma alongside the
vma itself for the lifetime of the vma, and so release both pins at the
same time as destroying the vma. It is common enough to pull into the
release function, making that central function more attractive to a
couple of other callsites.
The continual ulterior motive is to sweep over errors on module load
aborting...
Changes the type and renames the max_vswing_tries variable
which was declared as an integer but used as a boolean
making it easy to be confused with a counter.
Changes in v2:
- updated the title and commit message
- left the loop exit point in place
v3: fix typo in title
v4: renamed max_vswing to max_vswing_reached (Ville)
drm/i915/dp: Limit link training clock recovery loop
Limit the link training clock recovery loop to 10 attempts at
LANEx_CR_DONE per DP 1.4 spec section 3.5.1.2.2 and 80 attempts for
pre-DP 1.4 (4 voltage levels x 4 preemphasis levels x
x 5 identical voltages tries). Some faulty USB-C MST hubs can
cause us to get stuck in this loop indefinitely requesting something
like:
voltage swing: 0, pre-emphasis level: 2
voltage swing: 1, pre-emphasis level: 2
voltage swing: 0, pre-emphasis level: 3
over and over so max_vswing would never be reached,
drm_dp_clock_recovery_ok() would never return true and voltage_tries
would always get reset to 1. The driver sends those values to the hub
but the hub keeps requesting new values every time.
Changes in v2:
- updated commit message (DK, Manasi)
- defined DP_DP14_MAX_CR_TRIES (Marc)
- made the loop iterate for max 10 times (Rodrigo, Marc)
Changes in v3:
- changed error message to use DP_DP14_MAX_CR_TRIES
Changes in v4:
- Updated the title to reflect the change
- Updated the commit message
- Added 80 attempts for pre-DP 1.4 devices
Changes in v5:
- Removed DP_DP14_MAX_CR_TRIES from drm
v6: Updated comment to match kernel style (Rodrigo)
Michał Winiarski [Thu, 12 Jul 2018 15:53:30 +0000 (17:53 +0200)]
drm/i915/kvmgt: Fix compilation error
gvt_pin_guest_page extracted some of the gvt_dma_map_page functionality:
commit 6a73088bb4bf ("drm/i915/kvmgt: Support setting dma map for huge pages")
And yet, part of it was reintroduced in:
commit 940bd0233601 ("drm/i915/kvmgt: Check the pfn got from vfio_pin_pages")
Causing kvmgt part to no longer build. Let's remove it.
Reported-by: Tomasz Lis <tomasz.lis@intel.com> Signed-off-by: Michał Winiarski <michal.winiarski@intel.com> Cc: Changbin Du <changbin.du@intel.com> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180712155330.32055-1-michal.winiarski@intel.com
First of all don't try to read dpcd if PSR is not even supported.
But also, if read failed return -EIO instead of reporting via a
backchannel.
v2: fix dev_priv: At this level m->private is the connector. (CI/DK)
don't convert dpcd read errors to EIO. (DK)
Fixes: fe0043a65b2d ("drm/i915/psr: Split sink status into a separate debugfs node") Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Cc: José Roberto de Souza <jose.souza@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180720003155.16290-1-rodrigo.vivi@intel.com
Chris Wilson [Fri, 20 Jul 2018 10:19:10 +0000 (11:19 +0100)]
drm/i915: Only force GGTT coherency w/a on required chipsets
Not all chipsets have an internal buffer delaying the visibility of
writes via the GGTT being visible by other physical paths, but we use a
very heavy workaround for all. We only need to apply that workarounds to
the chipsets we know suffer from the delay and the resulting coherency
issue.
Similarly, the same inconsistent coherency fouls up our ABI promise that
a write into a mmap_gtt is immediately visible to others. Since the HW
has made that a lie, let userspace know when that contract is broken.
(Not that userspace would want to use mmap_gtt on those chipsets for
other performance reasons...)
Testcase: igt/drv_selftest/live_coherency
Testcase: igt/gem_mmap_gtt/coherency
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100587 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Tomasz Lis <tomasz.lis@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180720101910.11153-1-chris@chris-wilson.co.uk
Chris Wilson [Fri, 20 Jul 2018 09:51:44 +0000 (10:51 +0100)]
drm/i915: Suppress assertion for i915_ggtt_disable_guc
Another step in the drv_module_reload fault-injection saga, is that we
try to disable the guc twice. Probably. It's a little unclear exactly
what is going on in the unload sequence that catches us out, so for the
time being suppress the assertion to get the test re-enabled.
Dave Airlie [Fri, 20 Jul 2018 04:30:18 +0000 (14:30 +1000)]
Merge branch 'drm-next-4.19' of git://people.freedesktop.org/~agd5f/linux into drm-next
More features for 4.19:
- Map processes to vmids for debugging GPUVM faults
- Raven gfxoff fixes
- Initial gfxoff support for vega12
- Use defines for interrupt sources rather than magic numbers
- DC aux fixes
- Finish DC logging TODO
- Add more DC debugfs interfaces for conformance testing
- Add CRC support for DCN
- Scheduler rework in preparation for load balancing
- Unify common smu9 code
- Clean up UVD instancing support
- ttm cleanups
- Misc fixes and cleanups
Dave Airlie [Fri, 20 Jul 2018 02:29:23 +0000 (12:29 +1000)]
Merge tag 'drm-intel-next-2018-07-19' of git://anongit.freedesktop.org/drm/drm-intel into drm-next
On GEM side:
- GuC related fixes (Chris, Michal)
- GTT read-only pages support (Jon, Chris)
- More selftests fixes (Chris)
- More GPU reset improvements (Chris)
- Flush caches after GGTT writes (Chris)
- Handle recursive shrinker for vma->last_active allocation (Chris)
- Other execlists fixes (Chris)
On Display side:
- GLK HDMI fix (Clint)
- Rework and cleanup around HPD pin (Ville)
- Preparation work for Display Stream Compression support coming on ICL (Anusha)
- Nuke LVDS lid notification (Ville)
- Assume eDP is always connected (Ville)
- Kill intel panel detection (Ville)
Signed-off-by: Dave Airlie <airlied@redhat.com>
# gpg: Signature made Fri 20 Jul 2018 01:51:45 AM AEST
# gpg: using RSA key FA625F640EEB13CA
# gpg: Good signature from "Rodrigo Vivi <rodrigo.vivi@intel.com>"
# gpg: aka "Rodrigo Vivi <rodrigo.vivi@gmail.com>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 6D20 7068 EEDD 6509 1C2C E2A3 FA62 5F64 0EEB 13CA
drm/i915: Fix assert_plane() warning on bootup with external display
On KBL, WHL RVPs, booting up with an external display connected, triggers
below warning, when the BiOS brings up the external display too.
This warning is not seen during hotplug.
The warning is seen when mode_setcrtc() is called for pipeB
during bootup and before we get a mode_setcrtc() for pipeA,
while doing update_crtcs() in intel_atomic_commit_tail().
Now since, plane1A is still active after commit, update_crtcs()
is done for pipeA and eventually update_plane() for plane1A.
intel_plane_state->ctl for plane1A is not updated since set_modecrtc() is
called for pipeB. So intel_plane_state->ctl for plane 1A will be 0x0.
So doing an update_plane() for plane1A, will result in clearing
PLANE_CTL_ENABLE bit, and hence the warning.
To fix this warning, force all active planes to recompute their states
in probe.
Changes in v8:
- Actually add Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Changes in v7:
- Move call to intel_initial_commit() after sanitize_watermarks()
Otherwise the plane update will still consult potentially bogus
watermarks we read out from the hardware. (Ville)
- Carry Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
from v6
Changes in v6:
- Handle EDEADLK for drm_atomic_get_crtc_state() and
drm_atomic_add_affected_planes()
- Remove optimization of calling intel_initial_commit()
only when there is more than one active pipe in probe.
- Avoid using intel_ types.
Changes in v5:
- Drop drm_modeset_lock_all_ctx() since locks will be taken later.
Changes in v4:
- Handle locking in intel_initial_commit()
- Move the for loop inside intel_initial_commit() so that
drm_atomic_commit() is called only once
- Call intel_initial_commit() only for more than one active crtc on boot.
- Save the return value of intel_initial_commit() and print a message in
case of an error
Changes in v3:
- Add comments
Changes in v2:
- Force all planes to recompute their states.(Ville Syrjälä)
- Update the commit message
Christian König [Wed, 18 Jul 2018 18:30:51 +0000 (20:30 +0200)]
drm/amdgpu: clean up UVD instance handling v2
The whole handle, filp and entity handling is superfluous here.
We should have reviewed that more thoughtfully. It looks like somebody
just made the code instance aware without knowing the background.
v2: fix one more missed case in amdgpu_uvd_suspend
Reviewed-by: Leo Liu <leo.liu@amd.com> Signed-off-by: Christian König <christian.koenig@amd.com> Acked-by: Chunming Zhou <david1.zhou@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Christian König [Wed, 18 Jul 2018 18:28:08 +0000 (20:28 +0200)]
drm/amdgpu: remove superflous UVD encode entity
Not sure what that was every used for, but now it is completely unused.
Reviewed-by: Leo Liu <leo.liu@amd.com> Signed-off-by: Christian König <christian.koenig@amd.com> Acked-by: Chunming Zhou <david1.zhou@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Michel Dänzer [Tue, 17 Jul 2018 10:37:45 +0000 (12:37 +0200)]
drm/amdgpu/display: Replace CONFIG_DRM_AMD_DC_DCN1_0 with CONFIG_X86
Allowing CONFIG_DRM_AMD_DC_DCN1_0 to be disabled on X86 was an
opportunity for display with Raven Ridge accidentally not working.
Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Michel Dänzer <michel.daenzer@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Chris Wilson [Tue, 17 Jul 2018 09:57:50 +0000 (10:57 +0100)]
drm/i915/gtt: Enable full-ppgtt by default everywhere
We should we have all the kinks worked out and full-ppgtt now works
reliably on gen7 (Ivybridge, Valleyview/Baytrail and Haswell). If we can
let userspace have full control over their own ppgtt, it makes softpinning
far more effective, in turn making GPU dispatch far more efficient by
virtue of better mm segregation. On the other hand, switching over to a
different GTT for every client does incur noticeable overhead, but only
for very lightweight tasks.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Matthew Auld <matthew.william.auld@gmail.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Jason Ekstrand <jason.ekstrand@intel.com> Cc: Kenneth Graunke <kenneth@whitecape.org> Acked-by: Kenneth Graunke <kenneth@whitecape.org> Link: https://patchwork.freedesktop.org/patch/msgid/20180717095751.1034-1-chris@chris-wilson.co.uk
Ville Syrjälä [Tue, 17 Jul 2018 17:42:15 +0000 (20:42 +0300)]
drm/i915: Assume eDP is always connected
We never registered any kind of lid notifier for eDP, so looking at the
lid status is pretty much bonkers. Let's just consider eDP always
connected instead.
Ville Syrjälä [Tue, 17 Jul 2018 17:42:14 +0000 (20:42 +0300)]
drm/i915: Nuke the LVDS lid notifier
We broke the LVDS notifier resume thing in (presumably) commit 8c19bbaadc91 ("drm/i915: Use atomic helpers for suspend, v2.") as
we no longer duplicate the current state in the LVDS notifier and
thus we never resume it properly either.
Instead of trying to fix it again let's just kill off the lid
notifier entirely. None of the machines tested thus far have
apparently needed it. Originally the lid notifier was added to
work around cases where the VBIOS was clobbering some of the
hardware state behind the driver's back, mostly on Thinkpads.
We now have a few report of Thinkpads working just fine without
the notifier. So maybe it was misdiagnosed originally, or
something else has changed (ACPI video stuff perhaps?).
If we do end up finding a machine where the VBIOS is still causing
problems I would suggest that we first try setting various bits in
the VBIOS scratch registers. There are several to choose from that
may instruct the VBIOS to steer clear.
With the notifier gone we'll also stop looking at the panel status
in ->detect().
Chris Wilson [Thu, 19 Jul 2018 07:50:29 +0000 (08:50 +0100)]
drm/i915/execlists: Move the assertion we have the rpm wakeref down
There's a race between idling the engine and finishing off the last
tasklet (as we may kick the tasklets after declaring an individual
engine idle). However, since we do not need to access the device until
we try to submit to the ELSP register (processing the CSB just requires
normal CPU access to the HWSP, and when idle we should not need to
submit!) we can defer the assertion unto that point. The assertion is
still useful as it does verify that we do hold the longterm GT wakeref
taken from request allocation until request completion.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107274 Fixes: 8c6668acdab0 ("drm/i915/execlists: Direct submission of new requests (avoid tasklet/ksoftirqd)") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180719075029.28643-1-chris@chris-wilson.co.uk
Chris Wilson [Thu, 19 Jul 2018 07:22:06 +0000 (08:22 +0100)]
drm/i915: Handle recursive shrinker for vma->last_active allocation
If we call into the shrinker for direct relcaim inside kmalloc, it will
retire the requests. If we retire the vma->last_active while processing a
new i915_vma_move_to_active() we can upset the delicate bookkeeping
required for the cache. After the possible invocation of the shrinker, we
need to double check the vma->last_active is still valid.
Fixes: 2985a48c2cac ("drm/i915: Track the last-active inside the i915_vma")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105600#c39 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180719072206.16015-1-chris@chris-wilson.co.uk
We make a decision at module load whether to use the GuC backend or not,
but lose that setup across set-wedge. Currently, the guc doesn't
override the engine->set_default_submission hook letting execlists sneak
back in temporarily on unwedging leading to an unbalanced park/unpark.
v2: Remove comment about switching back temporarily to execlists on
guc_submission_disable(). We currently only call disable on shutdown,
and plan to also call disable before suspend and reset, in which case we
will either restore guc submission or mark the driver as wedged, making
the reset back to execlists pointless.
v3: Move reset.prepare across
Fixes: 16be45f3e3fc ("drm/i915/execlists: Flush pending preemption events during reset")
Testcase: igt/drv_module_reload/basic-reload-inject
Testcase: igt/gem_eio Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Michał Winiarski <michal.winiarski@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180717202932.1423-1-chris@chris-wilson.co.uk
Ben Skeggs [Tue, 17 Jul 2018 23:33:39 +0000 (09:33 +1000)]
drm/nouveau/kms/nv50-: allocate push buffers in vidmem on pascal
Workaround for issues seen on systems with large amounts of RAM, caused
by display not supporting the same physical address limits as the other
parts of the GPU.
i915/dp/dsc: Add Rate Control Range Parameter Registers
RC model has these parameters that correspond with each of
15 ranges of RC buffer threshold value in the RC model.
The three elements are range_min_qp, range_max_qp and
range_bpg_offset.
Add the Rate Control range values for eDP/MIPI and DP case.
The actual values are calculated usung a helper function.
This patch adds the shifts to registers where the value will
be written during atomic commit.
v2:
- Use _MMIO_PIPE() instead of _MMIO(_PICK()) (Manasi)
- Combine shifts (Manasi)
Cc: Jose Souza <jose.souza@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Cc: Ville Syrjala <ville.syrjala@linux.intel.com> Cc: Manasi Navare <manasi.d.navare@intel.com> Signed-off-by: Anusha Srivatsa <anusha.srivatsa@intel.com> Reviewed-by: Manasi Navare <manasi.d.navare@intel.com> Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/1531861861-10950-4-git-send-email-anusha.srivatsa@intel.com
i915/dp/dsc: Add Rate Control Buffer Threshold Registers
Add register defines and shifts that control the RC buffer threshold
between encoder and decoder for eDP/MIPI and DP cases.
The actual values are calculated usung a helper function.
This patch adds the shifts to registers where the value will
be written during atomic commit.
v2:
- Use _MMIO_PIPE() instead of _MMIO_(_PICK()) (Manasi)
- Combine shifts (Manasi)
The Picture Parameter Set metadata for DSC has to be sent
to the panel through secondary data packets. Add the error
correction registers, data registers and control registers
for the same.
The control registers for transcoders A and B are already
defined and will be reused for Icelake purpose. This patch adds
Control register for EDP and transcoder C apart from adding the
PPS data and error registers.
v2: reuse MMIO_TRANS2 for _PPS_DATA and _PPS_ECC.
The _MMIO_TRANS2(pipe, reg) macro definition takes care of the eDp case
drm/amdgpu: simplify the bo reference on amdgpu_bo_update
BO ptr already be initialized at definition, we needn't use the complicated
reference.
v2: fix typo at subject line
Signed-off-by: Huang Rui <ray.huang@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com> Reviewed-by: Chunming Zhou <david1.zhou@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Christian König [Mon, 16 Jul 2018 11:47:34 +0000 (13:47 +0200)]
drm/amdgpu: allow for more flexible priority handling
Allow to call amdgpu_ring_priority_get() after pushing the ring to the
scheduler.
Signed-off-by: Christian König <christian.koenig@amd.com> Reviewed-by: Chunming Zhou <david1.zhou@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It was originally introduced following the VESA spec in order to validate PSR.
However we found so many issues around sink_crc that instead of helping PSR
development it only brought another layer of trouble to the table.
So, sink_crc has been a black whole for us in question of time, effort and hope.
First of the problems is that HW statement is clear: "Do not attempt to use
aux communication with PSR enabled". So the main reason behind sink_crc is
already compromised.
For a while we had hope on the aux-mutex could workaround this problem on SKL+
platforms, but that mutex was not reliable, not tested,
and we shouldn't use according to HW engineers.
Also, nor source, nor sink designed and implemented the sink_crc to be used like
we are trying to use here.
Well, the sink side of things is also apparently not prepared for this
case. Each panel that we tried seemed to have a different behavior with same
code and same source.
So, for all the time we lost on trying to ducktape all these different issues
I believe it is now time to move PSR to a more reliable validation.
Maybe not a perfect one as we dreamed for this sink_crc, but at least more
reliable.
Dave Airlie [Wed, 18 Jul 2018 19:46:24 +0000 (05:46 +1000)]
Merge tag 'drm-intel-next-2018-07-09' of git://anongit.freedesktop.org/drm/drm-intel into drm-next
Higlights here goes to many PSR fixes and improvements; to the Ice lake work with
power well support and begin of DSI support addition. Also there were many improvements
on execlists and interrupts for minimal latency on command submission; and many fixes
on selftests, mostly caught by our CI.
General driver:
- Clean-up on aux irq (Lucas)
- Mark expected switch fall-through for dealing with static analysis tools (Gustavo)
Gem:
- Different fixes for GuC (Chris, Anusha, Michal)
- Avoid self-relocation BIAS if no relocation (Chris)
- Improve debugging cases in on EINVAL return and vma allocation (Chris)
- Fixes and improvements on context destroying and freeing (Chris)
- Wait for engines to idle before retiring (Chris)
- Many improvements on execlists and interrupts for minimal latency on command submission (Chris)
- Many fixes in selftests, specially on cases highlighted on CI (Chris)
- Other fixes and improvements around GGTT (Chris)
- Prevent background reaping of active objects (Chris)
Display:
- Parallel modeset cleanup to fix driver reset (Chris)
- Get AUX power domain for DP main link (Imre)
- Clean-up on PSR unused func pointers (Rodrigo)
- Many PSR/PSR2 fixes and improvements (DK, Jose, Tarun)
- Add a PSR1 live status (Vathsala)
- Replace old drm_*_{un/reference} with put,get functions (Thomas)
- FBC fixes (Maarten)
- Abstract and document the usage of picking macros (Jani)
- Remove unnecessary check for unsupported modifiers for NV12. (DK)
- Interrupt fixes for display (Ville)
- Clean up on sdvo code (Ville)
- Clean up on current DSI code (Jani)
- Remove support for legacy debugfs crc interface (Maarten)
- Simplify get_encoder_power_domains (Imre)
Icelake:
- MG PLL fixes (Imre)
- Add hw workaround for alpha blending (Vandita)
- Add power well support (Imre)
- Add Interrupt Support (Anusha)
- Start to add support for DSI on Ice Lake (Madhav)
Signed-off-by: Dave Airlie <airlied@redhat.com>
# gpg: Signature made Tue 10 Jul 2018 08:41:37 AM AEST
# gpg: using RSA key FA625F640EEB13CA
# gpg: Good signature from "Rodrigo Vivi <rodrigo.vivi@intel.com>"
# gpg: aka "Rodrigo Vivi <rodrigo.vivi@gmail.com>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 6D20 7068 EEDD 6509 1C2C E2A3 FA62 5F64 0EEB 13CA Link: https://patchwork.freedesktop.org/patch/msgid/20180710234349.GA16562@intel.com
drm/sun4i: Substitute sun4i_backend_format_is_yuv() with format->is_yuv
drm_format_info table has a field 'is_yuv' to denote if the format
is yuv or not. The driver is expected to use this instead of
having a function for the same purpose.
drm/rockchip: Substitute is_yuv_support() with format->is_yuv
drm_format_info table has a field 'is_yuv' to denote if the format
is yuv or not. The driver is expected to use this instead of
having a function for the same purpose.
drm/i915: Substitute intel_format_is_yuv() with format->is_yuv
drm_format_info table has a field 'is_yuv' to denote if the format
is yuv or not. The driver is expected to use this instead of
having a function for the same purpose.
drm/fourcc: Add is_yuv field to drm_format_info to denote if the format is yuv
A lot of drivers duplicate the function to check if a format is yuv or not.
If we add a field (to denote whether the format is yuv or not) in the
drm_format_info table, all the drivers can use this field and it will
prevent duplication of similar logic.
Chris Wilson [Tue, 17 Jul 2018 08:41:21 +0000 (09:41 +0100)]
drm/i915: Always retire residual requests before suspend
If the driver is wedged, we skip idling the GPU. However, we may still
have a few requests still not retired following the wedging (since they
will be waiting for a background worker trying to acquire struct_mutex).
As we hold the struct_mutex, always do a quick request retirement in
order to flush the wedged path.
drm/pl111: Use 64-bit arithmetic instead of 32-bit
Add suffix ULL to constant 1000 in order to give the compiler complete
information about the proper arithmetic to use.
Notice that such constant is used in a context that expects an
expression of type u64 (64 bits, unsigned) and the following
expression is currently being evaluated using 32-bit arithmetic:
This patch unifies the naming of DRM functions for reference counting
of struct drm_device. The resulting code is more aligned with the rest
of the Linux kernel interfaces.
This patch unifies the naming of DRM functions for reference counting
of struct drm_device. The resulting code is more aligned with the rest
of the Linux kernel interfaces.
Chris Wilson [Tue, 17 Jul 2018 09:26:55 +0000 (10:26 +0100)]
drm/i915: Flush chipset caches after GGTT writes
Our I915g (early gen3, the oldest machine we have in the farm) is still
reporting occasional incoherency performing the following operations:
1) write through GGTT (indirect write into memory)
2) write through either CPU or WC (direct write into memory)
3) read from GGTT (indirect read)
Instead of reporting the value from (2), the read from GGTT reports the
earlier value written via the GGTT. We have made sure that the writes are
flushed from the CPU (commit dd7542cbf28c ("drm/i915/selftests: Provide
full mb() around clflush") and commit 94b56690278b ("drm/i915: Flush the
WCB following a WC write")), but still see the error, just less
frequently. The only remaining cache that might be affected here is a
chipset cache, so flush that as well.
drm/sun4i: sun8i: Avoid clearing blending order at each atomic commit
Blending order is set based on the z position of each DRM plane. The
blending order register is currently cleared at each atomic DRM commit,
with the intent that each committed plane will set the appropriate
bits (based on its z-pos) when enabling the plane.
However, it sometimes happens that a particular plane is left unchanged
by an atomic commit and thus will not be configured again. In that
scenario, blending order is cleared and only the bits relevant for the
planes affected by the commit are set. This leaves the planes that did
not change without their blending order set in the register, leading
to that plane not being displayed.
Instead of clearing the blending order register at every atomic commit,
this change moves the register's initial clear at bind time and only
clears the bits for a specific plane when disabling it or changing its
zpos.
This way, planes that are left untouched by a DRM atomic commit are
no longer disabled.
This patch unifies the naming of DRM functions for reference counting
of struct drm_device. The resulting code is more aligned with the rest
of the Linux kernel interfaces.
Chris Wilson [Tue, 17 Jul 2018 08:23:34 +0000 (09:23 +0100)]
drm/i915/selftests: Free the backing store between iterations
In the huge pages tests, we may have lots of objects being trapped on
the freelist as we hold the struct_mutex allowing the free worker no
opportunity to recover the backing store. We also have stricter
requirements and the desire for large contiguous pages, further
increasing the allocation pressure. To reduce the chance of running out
of memory, we could either drop the mutex and flush the free worker, or
we could release the backing store directly. We do the latter in this
patch for simplicity.
projects / kernel.git / log