Chris Wilson [Wed, 23 Oct 2019 23:53:59 +0000 (00:53 +0100)]
drm/i915/selftests: Flush any i915_active callback work as well
Make trebly sure that all possible callbacks and their delayed brethren
are complete before asserting that the i915_active should be idle after
flushing all barriers.
We have the same CSB events being seen by process_csb() on two different
processors. One being issued by the reset in the test, the other by the
interrupt; this scenario is supposed to be prevented by flushing the
interrupt tasklet with tasklet_disable() before we enter the atomic
reset.
Zhenyu Wang [Wed, 23 Oct 2019 09:43:27 +0000 (17:43 +0800)]
drm/i915/gvt: fix dead locking in early workload shadow
As early workload scan and shadow happens in execlist mmio handler,
which has already taken vgpu_lock. So remove extra lock taking here.
Fixes: 952f89f098c7 ("drm/i915/gvt: Wean off struct_mutex") Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson [Wed, 23 Oct 2019 13:31:08 +0000 (14:31 +0100)]
drm/i915/gt: Replace hangcheck by heartbeats
Replace sampling the engine state every so often with a periodic
heartbeat request to measure the health of an engine. This is coupled
with the forced-preemption to allow long running requests to survive so
long as they do not block other users.
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: Jon Bloomfield <jon.bloomfield@intel.com> Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191023133108.21401-5-chris@chris-wilson.co.uk
Chris Wilson [Wed, 23 Oct 2019 13:31:07 +0000 (14:31 +0100)]
drm/i915/gem: Cancel contexts when hangchecking is disabled
Normally, we rely on our hangcheck to prevent persistent batches from
hogging the GPU. However, if the user disables hangcheck, this mechanism
breaks down. Despite our insistence that this is unsafe, the users are
equally insistent that they want to use endless batches and will disable
the hangcheck mechanism. We are looking at replacing hangcheck, in the
next patch, with a softer mechanism, that sends a pulse down the engine
to check if it is well. We can use the same preemptive pulse to flush an
active context off the GPU upon context close, preventing resources
being lost and unkillable requests remaining on the GPU after process
termination.
Testcase: igt/gem_ctx_exec/basic-nohangcheck Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Jon Bloomfield <jon.bloomfield@intel.com> Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191023133108.21401-4-chris@chris-wilson.co.uk
Chris Wilson [Wed, 23 Oct 2019 13:31:06 +0000 (14:31 +0100)]
drm/i915/execlists: Cancel banned contexts on schedule-out
On schedule-out (CS completion) of a banned context, scrub the context
image so that we do not replay the active payload. The intent is that we
skip banned payloads on request submission so that the timeline
advancement continues on in the background. However, if we are returning
to a preempted request, i915_request_skip() is ineffective and instead we
need to patch up the context image so that it continues from the start
of the next request.
v2: Fixup cancellation so that we only scrub the payload of the active
request and do not short-circuit the breadcrumbs (which might cause
other contexts to execute out of order).
v3: Grammar pass
Chris Wilson [Wed, 23 Oct 2019 13:31:05 +0000 (14:31 +0100)]
drm/i915/execlists: Force preemption
If the preempted context takes too long to relinquish control, e.g. it
is stuck inside a shader with arbitration disabled, evict that context
with an engine reset. This ensures that preemptions are reasonably
responsive, providing a tighter QoS for the more important context at
the cost of flagging unresponsive contexts more frequently (i.e. instead
of using an ~10s hangcheck, we now evict at ~100ms). The challenge of
lies in picking a timeout that can be reasonably serviced by HW for
typical workloads, balancing the existing clients against the needs for
responsiveness.
Note that coupled with timeslicing, this will lead to rapid GPU "hang"
detection with multiple active contexts vying for GPU time.
The forced preemption mechanism can be compiled out with
Chris Wilson [Wed, 23 Oct 2019 13:31:04 +0000 (14:31 +0100)]
drm/i915/gt: Try to more gracefully quiesce the system before resets
If we are doing a normal GPU reset triggered after detecting a long
period of stalled work, we can take our time and allow the engines to
quiesce. Since we've stopped submission to the engine, and if we wait
long enough an innocent context should complete, leaving the engine idle.
So by waiting a short amount of time, we should prevent clobbering other
users when resetting a stuck context.
Suggested-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Suggested-by: Jon Bloomfield <jon.bloomfield@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191023133108.21401-1-chris@chris-wilson.co.uk
GuC enable logging H2G action definition changed some time ago from 0xE000
to 0x40. All current GuC FW blobs use this definition, so fix the action
definition in driver to match.
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com> Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191022163754.23870-2-robert.m.fosha@intel.com
Robert M. Fosha [Tue, 22 Oct 2019 16:37:52 +0000 (09:37 -0700)]
drm/i915/guc: Enable guc logging on guc log relay write
Creating and opening the GuC log relay file enables and starts
the relay potentially before the caller is ready to consume logs.
Change the behavior so that relay starts only on an explicit call
to the write function (with a value of '1'). Other values flush
the log relay as before.
v2: Style changes and fix typos. Add guc_log_relay_stop()
function. (Daniele)
Cc: Matthew Brost <matthew.brost@intel.com> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191022163754.23870-1-robert.m.fosha@intel.com
Imre Deak [Tue, 22 Oct 2019 09:51:55 +0000 (12:51 +0300)]
drm/i915: Add new CNL PCH ID seen on a CML platform
Atm we don't detect a PCH with PCI ID 0xA3C1 which showed up now on a CML
platform. We don't have the official assignment of the PCH PCI IDs, but
this looks like a CNP which was already used on CML platforms. Let's add
the new ID->PCH type mapping accordingly.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=112051 Reported-and-tested-by: Cyrus <cyrus.lien@canonical.com> Cc: Cyrus <cyrus.lien@canonical.com> Signed-off-by: Imre Deak <imre.deak@intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191022095155.30991-1-imre.deak@intel.com
Jani Nikula [Tue, 22 Oct 2019 13:34:14 +0000 (16:34 +0300)]
drm/i915/dsc: move crtc state dp_dsc_cfg member under dsc as config
DSC isn't DP specific, so remove the dp_ prefix from the crtc state
member name. Also moving the member under the dsc sub-struct gives us
enough context to allow shortening the name to just config. No
functional changes.
Jani Nikula [Tue, 22 Oct 2019 13:34:13 +0000 (16:34 +0300)]
drm/i915/dsc: rename crtc state dsc_params member to dsc
Reduce verbosity in code by renaming dsc_params member of crtc state to
simply dsc. There is enough context for this to be clear. No functional
changes.
Chris Wilson [Tue, 22 Oct 2019 22:33:16 +0000 (23:33 +0100)]
drm/i915/selftests: Release ctx->engine_mutex after iteration
A lock once taken must be released again.
Fixes: c31c9e82ee8a ("drm/i915/selftests: Teach switch_to_context() to use the context") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191022223316.12662-1-chris@chris-wilson.co.uk
Chris Wilson [Tue, 22 Oct 2019 11:21:11 +0000 (12:21 +0100)]
drm/i915/selftests: Synchronize checking active status with retirement
If retirement is running on another thread, we may inspect the status of
the i915_active before its retirement callback is complete. As we expect
it to be running synchronously, we can wait for any callback to complete
by acquiring the i915_active.mutex.
Chris Wilson [Tue, 22 Oct 2019 13:10:16 +0000 (14:10 +0100)]
drm/i915/selftests: Move uncore fw selftests to operate on intel_gt
Forcewake is the speciality of the GT, so it is natural to run the
intel_uncore_forcewake tests over the GT. So pass intel_gt as the
parameter to our selftests.
Chris Wilson [Tue, 22 Oct 2019 14:19:35 +0000 (15:19 +0100)]
drm/i915: Teach record_defaults to operate on the intel_gt
Again we wish to operate on the engines, which are owned by the
intel_gt. As such it is easier, and much more consistent, to pass the
intel_gt parameter.
Chris Wilson [Tue, 22 Oct 2019 12:28:45 +0000 (13:28 +0100)]
drm/i915: Drop assertion that ce->pin_mutex guards state updates
The actual conditions are that we know the GPU is not accessing the
context, and we hold a pin on the context image to allow CPU access. We
used a fake lock on ce->pin_mutex so that we could try and use lockdep
to assert that access is serialised, but the various different
hardirq/softirq contexts where we need to *fake* holding the pin_mutex
are causing more trouble.
Still it would be nice if we did have a way to reassure ourselves that
the direct update to the context image is serialised with GPU execution.
In the meantime, stop lockdep complaining about false irq inversions.
Chris Wilson [Tue, 22 Oct 2019 10:17:04 +0000 (11:17 +0100)]
drm/i915/selftests: Make the mman object busy everywhere
Loop over all engines, issuing a request for the object on each in order
to make sure we leave no stone unturned when creating an active ref. The
purpose is to make sure that we can reap a zombie object (one that is
only alive due to an active reference on the GPU) no matter where that
active reference emanates from.
Chris Wilson [Mon, 21 Oct 2019 18:32:35 +0000 (19:32 +0100)]
drm/i915: Lift i915_vma_parked() onto the gt
Currently even though i915_vma_parked() operates on a per-gt struct, it
is called from a global pm notify. This oddity was only because the long
term plan is to decouple the vma cache from the pm notification, but
right now the oddity stands out like a sore thumb!
Chris Wilson [Mon, 21 Oct 2019 17:43:39 +0000 (18:43 +0100)]
drm/i915/gt: Introduce barrier pulses along engines
To flush idle barriers, and even inflight requests, we want to send a
preemptive 'pulse' along an engine. We use a no-op request along the
pinned kernel_context at high priority so that it should run or else
kick off the stuck requests. We can use this to ensure idle barriers are
immediately flushed, as part of a context cancellation mechanism, or as
part of a heartbeat mechanism to detect and reset a stuck GPU.
Ville Syrjälä [Thu, 18 Jul 2019 14:43:40 +0000 (17:43 +0300)]
drm/i915: Check some transcoder timing minimum limits
On ILK+ the documented min hdisplay is 64, min hblank is 32, and min
vblank is 5. On earlier platforms min hblank is also 32, and min
vblank is 3. Make sure the mode satisfies those limits.
There are further limits for HDMI and pfit use cases, but we'll check
for those in a more specific location.
Chris Wilson [Mon, 21 Oct 2019 08:02:11 +0000 (09:02 +0100)]
drm/i915: Don't set queue_priority_hint if we don't kick the submission
If we change the priority of the active context, then it has no impact
on the decision of whether to preempt the active context -- we don't
preempt the context with itself. In this situation, we elide the tasklet
rescheduling and should *not* be marking up the queue_priority_hint as
that may mask a later submission where we decide we don't have to kick
the tasklet as a higher priority submission is pending (spoiler alert,
it was not).
Andi Shyti [Sun, 20 Oct 2019 18:41:39 +0000 (19:41 +0100)]
drm/i915: Extract GT ring management
Although the ring management is much smaller compared to the other GT
power management functions, continue the theme of extracting it out of
the huge intel_pm.c for maintenance.
The current logic just reapplies the same configuration already stored
into stream->oa_config instead of the newly selected one.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> Fixes: 7831e9a965ea ("drm/i915/perf: Allow dynamic reconfiguration of the OA stream") Cc: Chris Wilson <chris@chris-wilson.co.uk> 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/20191019214647.27866-1-lionel.g.landwerlin@intel.com
Lucas De Marchi [Fri, 11 Oct 2019 01:09:03 +0000 (18:09 -0700)]
drm/i915: fix port checks for MST support on gen >= 11
Both Ice Lake and Elkhart Lake (gen 11) support MST on all external
connections except DDI A. Tiger Lake (gen 12) supports on all external
connections.
Move the check to happen inside intel_dp_mst_encoder_init() and add
specific platform checks.
v2: Replace != with == checks for ports on gen < 11 (Ville)
Manasi Navare [Fri, 18 Oct 2019 17:27:25 +0000 (10:27 -0700)]
drm/i915/display/icl: In port sync mode disable slaves first then master
In the transcoder port sync mode, the slave transcoders mask their vblanks
until master transcoder's vblank so while disabling them, make
sure slaves are disabled first and then the masters.
v5:
* Dont pass dev priv to get_slave_crtc (Ville)
v4:
* Obtain slave state from master (Maarten)
v3:
* Rebase
v2:
* Use the intel_old_crtc_state_disables() helper
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Manasi Navare <manasi.d.navare@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191018172725.1338-6-manasi.d.navare@intel.com
Manasi Navare [Fri, 18 Oct 2019 17:27:23 +0000 (10:27 -0700)]
drm/i915/display/icl: Enable master-slaves in trans port sync
As per the display enable sequence, we need to follow the enable sequence
for slaves first with DP_TP_CTL set to Idle and configure the transcoder
port sync register to select the corersponding master, then follow the
enable sequence for master leaving DP_TP_CTL to idle.
At this point the transcoder port sync mode is configured and enabled
and the Vblanks of both ports are synchronized so then set DP_TP_CTL
for the slave and master to Normal and do post crtc enable updates.
v11:
* Rebase (Manasi)
v10:
* in trans sync mode, dont stop link train for tgl (Manasi)
v9:
Remove update_scanline_offset to rebase on Maarten's patch (Manasi)
v8:
* Rebase on Maarten's patches (Manasi)
v7:
* Use ffs(slaves) to get slave crtc (Ville)
v6:
* Modeset implies active_changed, remove one condition (Maarten)
v5:
* Fix checkpatch warning (Manasi)
v4:
* Reuse skl_commit_modeset_enables() hook (Maarten)
* Obtain slave crtc and states from master (Maarten)
v3:
* Rebase on drm-tip (Manasi)
v2:
* Create a icl_update_crtcs hook (Maarten, Danvet)
* This sequence only for CRTCs in trans port sync mode (Maarten)
Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Signed-off-by: Manasi Navare <manasi.d.navare@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191018172725.1338-4-manasi.d.navare@intel.com
Manasi Navare [Fri, 18 Oct 2019 17:27:22 +0000 (10:27 -0700)]
drm/i915/display/icl: HW state readout for transcoder port sync config
After the state is committed, we readout the HW registers and compare
the HW state with the SW state that we just committed.
For Transcdoer port sync, we add master_transcoder and the
salves bitmask to the crtc_state, hence we need to read those during
the HW state readout to avoid pipe state mismatch.
v11:
* Move master trans init to get pipe_Config hooks (Ville)
v10:
* Initialize master_tarnscoder readout for all platforms (Ville)
v9:
* Initialize master_transcoder = INVALID at get config (Ville)
v8:
* Use master_select -1, address TRANS_EDP case (Ville)
* Rename master_transcoder to _readout (Lucas)
v7:
* NDont read HW state for DSI
v6:
* Go through both parts of HW readout (Maarten)
* Add a WARN if the same trans configured as
master and slave (Ville, Maarten)
v5:
* Add return INVALID in defaut case (Maarten)
v4:
* Get power domains in master loop for get_config (Ville)
v3:
* Add TRANSCODER_D (Maarten)
* v3 Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
v2:
* Add Transcoder_D and MISSING_CASE (Maarten)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Manasi Navare <manasi.d.navare@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191018172725.1338-3-manasi.d.navare@intel.com
Manasi Navare [Fri, 18 Oct 2019 17:27:21 +0000 (10:27 -0700)]
drm/i915/display/icl: Enable TRANSCODER PORT SYNC for tiled displays across separate ports
In case of tiled displays where different tiles are displayed across
different ports, we need to synchronize the transcoders involved.
This patch implements the transcoder port sync feature for
synchronizing one master transcoder with one or more slave
transcoders. This is only enbaled in slave transcoder
and the master transcoder is unaware that it is operating
in this mode.
This has been tested with tiled display connected to ICL.
v7:
* Rebase on Maarten's patches
v6:
* Use master_trans +1 and address missing trans_edp case (Ville)
v5:
* Add TRANSCODER_D case and MISSING_CASE (Maarten)
v4:
Rebase
v3:
* Check of DP_MST moved to atomic_check (Maarten)
v2:
* Do not use RMW, just write to the register in commit (Jani N)
Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Manasi Navare <manasi.d.navare@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191018172725.1338-2-manasi.d.navare@intel.com
Manasi Navare [Fri, 18 Oct 2019 17:27:20 +0000 (10:27 -0700)]
drm/i915/display/icl: Save Master transcoder in slave's crtc_state for Transcoder Port Sync
In case of tiled displays when the two tiles are sent across two CRTCs
over two separate DP SST connectors, we need a mechanism to synchronize
the two CRTCs and their corresponding transcoders.
So use the master-slave mode where there is one master corresponding
to last horizontal and vertical tile that needs to be genlocked with
all other slave tiles.
This patch identifies saves the master transcoder in all the slave
CRTC states. This is needed to select the master CRTC/transcoder
while configuring transcoder port sync for the corresponding slaves.
v6:
Rebase (manasi)
v5:
* Address Ville's comments
* Just pass crtc_state, no need to check GEN (Ville)
v4:
* Rebase
v3:
* Use master_tramscoder instead of master_crtc for valid
HW state readouts (Ville)
v2:
* Move this to intel_mode_set_pipe_config(Jani N, Ville)
* Use slave_bitmask to save associated slaves in master crtc state (Ville)
Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Signed-off-by: Manasi Navare <manasi.d.navare@intel.com> Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191018172725.1338-1-manasi.d.navare@intel.com
drm/i915: Restore full symmetry in i915_driver_modeset_probe/remove
Commit 2d6f6f359fd8 ("drm/i915: add i915_driver_modeset_remove()")
claimed removal of asymmetry in probe() and remove() calls, however, it
didn't take care of calling intel_irq_uninstall() on driver remove.
That doesn't hurt as long as we still call it from
intel_modeset_driver_remove() but in order to have full symmetry we
should call it again from i915_driver_modeset_remove().
Note that it's safe to call intel_irq_uninstall() twice thanks to
commit b318b82455bd ("drm/i915: Nuke drm_driver irq vfuncs"). We may
only want to mention the case we are adding in a related FIXME comment
provided by that commit. While being at it, update the name of
function mentioned as calling it out of sequence as that name has been
changed meanwhile by commit 78dae1ac35dd ("drm/i915: Propagate
"_remove" function name suffix down").
Suggested-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@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/6250061.7lZMOAyebC@jkrzyszt-desk.ger.corp.intel.com
Ville Syrjälä [Thu, 10 Oct 2019 14:51:27 +0000 (17:51 +0300)]
drm/i915: Make hdcp2_msg_timeout.timeout u16
All the timeout values fit in u16, so let's shrink the structure
a bit.
This ends up actually increasing the .text size a bit due to
some changes in instructions (constant imul+small jmps replaced
with mov+bigger jmpqs). Seems pretty arbitrary to me so I'll
just pretend I didn't see it.
text data bss dec hex filename
- 34521 360 0 34881 8841 intel_hdmi.o
+ 34537 360 0 34897 8851 intel_hdmi.o
Ville Syrjälä [Thu, 10 Oct 2019 14:51:26 +0000 (17:51 +0300)]
drm/i915: Remove hdcp2_hdmi_msg_timeout.timeout2
The only reason for the timeout2 value in the array is the
HDCP_2_2_AKE_SEND_HPRIME message. But that one still needs
special casing inside the loop, and so just ends up making
the code harder to read. Let's just remove this leaky
timeout2 abstraction and special case that one command
in a way that is easy to understand. We can then remove the
timeout2 member from struct entirely.
text data bss dec hex filename
- 34633 360 0 34993 88b1 intel_hdmi.o
+ 34521 360 0 34881 8841 intel_hdmi.o
Ville Syrjälä [Fri, 11 Oct 2019 20:09:43 +0000 (23:09 +0300)]
drm/i915: Nuke 'realloc_pipes'
The 'realloc_pipes' bitmask is pointless. It is either:
a) the set of pipes which are already part of the state,
in which case adding them again is entirely redundant
b) the set of all pipes which we then add to the state
Also the fact that 'realloc_pipes' uses the crtc indexes is
going to bite is at some point so best get rid of it quick.
Ville Syrjälä [Fri, 11 Oct 2019 20:09:42 +0000 (23:09 +0300)]
drm/i915: Nuke the useless changed param from skl_ddb_add_affected_pipes()
changed==true just means we have some crtcs in the state. All the
stuff following this only operates on crtcs in the state anyway so
there is no point in having this bool.
Ville Syrjälä [Thu, 18 Jul 2019 14:43:39 +0000 (17:43 +0300)]
drm/i915: Refuse modes with hdisplay==4096 on pre-HSW DP
The DP port/pipe goes wonky if we try to use timings with
hdisplay==4096 on pre-HSW platforms. The link fails to train
and the pipe may not signal vblank interrupts. On HDMI such at
mode works just fine (tested on ELK/SNB/CHV). So let's refuse
such modes on DP on older platforms.
Ville Syrjälä [Wed, 2 Oct 2019 16:25:00 +0000 (19:25 +0300)]
drm/i915: Polish possible_clones setup
Replace the hand rolled stuff with drm_encoder_mask() when populating
possible_clones, and rename the function to
intel_encoder_possible_clones() to make it clear what it's used for.
Ville Syrjälä [Tue, 15 Oct 2019 15:27:57 +0000 (18:27 +0300)]
drm/i915: Move the cursor rotation handling into intel_cursor_check_surface()
Unlike other planes the cursor currently handles 180 degree rotation
adjustment during the hardware programming phase. Let's move that
stuff into intel_cursor_check_surface() to match how we do things
with other plane types.
And while at we'll plop in the final src x/y coordinates (which will
actually always be zero) into the src rect and color_plane[0].x/y,
just for some extra consistency.
Matthew Auld [Fri, 18 Oct 2019 09:07:51 +0000 (10:07 +0100)]
drm/i915: treat stolen as a region
Convert stolen memory over to a region object. Still leaves open the
question with what to do with pre-allocated objects...
Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Abdiel Janulgue <abdiel.janulgue@linux.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/20191018090751.28295-3-matthew.auld@intel.com
Matthew Auld [Fri, 18 Oct 2019 09:07:50 +0000 (10:07 +0100)]
drm/i915: treat shmem as a region
Convert shmem to an intel_memory_region.
Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Abdiel Janulgue <abdiel.janulgue@linux.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/20191018090751.28295-2-matthew.auld@intel.com
Abdiel Janulgue [Fri, 18 Oct 2019 09:07:49 +0000 (10:07 +0100)]
drm/i915: enumerate and init each supported region
Nothing to enumerate yet...
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.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/20191018090751.28295-1-matthew.auld@intel.com
Chris Wilson [Fri, 18 Oct 2019 07:20:27 +0000 (08:20 +0100)]
drm/i915/execlists: Don't merely skip submission if maybe timeslicing
Normally, we try and skip submission if ELSP[1] is filled. However, we
may desire to enable timeslicing due to the queue priority, even if
ELSP[1] itself does not require timeslicing. That is the queue is equal
priority to ELSP[0] and higher priority then ELSP[1]. Previously, we
would wait until the context switch to preempt the current ELSP[1], but
with timeslicing, we want to preempt ELSP[0] and replace it with the
queue.
In writing the test case, it become quickly apparent that we were also
suppressing the tasklet during promotion and so failing to notice when
the queue started requiring timeslicing.
Tvrtko Ursulin [Thu, 17 Oct 2019 16:18:52 +0000 (17:18 +0100)]
drm/i915: Make for_each_engine_masked work on intel_gt
Medium term goal is to eliminate the i915->engine[] array and to get there
we have recently introduced equivalent array in intel_gt. Now we need to
migrate the code further towards this state.
This next step is to eliminate usage of i915->engines[] from the
for_each_engine_masked iterator.
For this to work we also need to use engine->id as index when populating
the gt->engine[] array and adjust the default engine set indexing to use
engine->legacy_idx instead of assuming gt->engines[] indexing.
v2:
* Populate gt->engine[] earlier.
* Check that we don't duplicate engine->legacy_idx
v3:
* Work around the initialization order issue between default_engines()
and intel_engines_driver_register() which sets engine->legacy_idx for
now. It will be fixed properly later.
drm/i915: Don't disable interrupts independently of the lock
The locks (active.lock and rq->lock) need to be taken with disabled
interrupts. This is done in i915_request_retire() by disabling the
interrupts independently of the locks itself.
While local_irq_disable()+spin_lock() equals spin_lock_irq() on vanilla
it does not on PREEMPT_RT.
Chris Wilson confirmed that local_irq_disable() was just introduced as
an optimisation to avoid enabling/disabling interrupts during
lock/unlock combo.
Enable/disable interrupts as part of the locking instruction.
Chris Wilson [Wed, 16 Oct 2019 12:52:36 +0000 (13:52 +0100)]
drm/i915/selftests: Teach requests to use all available engines
The request selftests straddle the boundary between checking the driver
and the hardware. They are subject to the quirks of the underlying HW,
but operate on top of the backend abstractions. The tests focus on the
scheduler elements and so should check for interactions of the scheduler
across all exposed engines.
Better explain the usage of the microcontroller and what i915 is
responsible of. While at it, fix the documentation for the auth
function, which doesn't do any pinning anymore.
v2: add a comment on HuC being optional and descrive how HuC accesses
memory (Martin)
v3: add extra newline for better text organization (Martin)
Add a short description of what we expect from GuC and some minor
improvements to existing documentation. Also remove a comment about a
difference between GuC and HuC that is not true anymore.
v2: add that the GuC is not mandatory (Martin)
v3: add extra newline for better text organization (Martin)
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Martin Peres <martin.peres@linux.intel.com> Acked-by: Anna Karas <anna.karas@intel.com> Reviewed-by: Martin Peres <martin.peres@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191014183602.3643-2-daniele.ceraolospurio@intel.com
Ville Syrjälä [Tue, 15 Oct 2019 19:05:38 +0000 (22:05 +0300)]
drm/i915: Fix MST oops due to MSA changes
The MSA MISC computation now depends on the connector state, and
we do it from the DDI .pre_enable() hook. All that is fine for
DP SST but with MST we don't actually pass the connector state
to the dig port's .pre_enable() hook which leads to an oops.
Need to think more how to solve this in a cleaner fashion, but
for now let's just add a NULL check to stop the oopsing.
Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com> Cc: Uma Shankar <uma.shankar@intel.com> Fixes: 0c06fa156006 ("drm/i915/dp: Add support of BT.2020 Colorimetry to DP MSA") Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191015190538.27539-1-ville.syrjala@linux.intel.com Reviewed-by: Uma Shankar <uma.shankar@intel.com> Reviewed-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Tvrtko Ursulin [Wed, 16 Oct 2019 09:38:02 +0000 (10:38 +0100)]
drm/i915/pmu: Support multiple GPUs
With discrete graphics system can have both integrated and discrete GPU
handled by i915.
Currently we use a fixed name ("i915") when registering as the uncore PMU
provider which stops working in this case.
To fix this we add the PCI device name string to non-integrated devices
handled by us. Integrated devices keep the legacy name preserving
backward compatibility.
v2:
* Detect IGP and keep legacy name. (Michal)
* Use PCI device name as suffix. (Michal, Chris)
v3:
* Constify the name. (Chris)
* Use pci_domain_nr. (Chris)
v4:
* Fix kfree_const usage. (Chris)
v5:
* kfree_const does not work for modules. (Chris)
* Changed is_igp helper to take i915.
Chris Wilson [Wed, 16 Oct 2019 14:32:34 +0000 (15:32 +0100)]
drm/i915: Move swizzle_bit under i915_ggtt
The HW performs swizzling as part of its fence tiling inside the Global
GTT. We already do the probing of the HW settings from the GGTT setup,
complete the picture by storing the information as part of the GGTT. The
primary benefit is the consistency of our probe routines do not break
the i915_ggtt encapsulation.
Chris Wilson [Wed, 16 Oct 2019 14:32:33 +0000 (15:32 +0100)]
drm/i915: Store i915_ggtt as the backpointer on fence registers
Now that i915_ggtt knows everything about its own paths to perform mmio,
we can use that as our primary backpointer for individual fence
registers. This reduces the amount of pointer dancing we have to perform
on the common paths, but more importantly finishes our fence register
encapsulation.
Chris Wilson [Wed, 16 Oct 2019 09:07:49 +0000 (10:07 +0100)]
drm/i915: Do initial mocs configuration directly
Now that we record the default "goldenstate" context, we do not need to
emit the mocs registers at the start of each context and can simply do
mmio before the first context and capture the registers as part of its
default image. As a consequence, this means that we repeat the mmio
after each engine reset, fixing up any platform and registers that were
zapped by the reset (for those platforms with global not context-saved
settings).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111723
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111645 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Prathap Kumar Valsan <prathap.kumar.valsan@intel.com> Reviewed-by: Prathap Kumar Valsan <prathap.kumar.valsan@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191016090749.7092-1-chris@chris-wilson.co.uk
Matt Roper [Tue, 15 Oct 2019 16:28:54 +0000 (09:28 -0700)]
drm/i915: Introduce Jasper Lake PCH
The Jasper Lake PCH follows ICP/TGP's south display behavior and is
identical to MCC graphics-wise except that it does not use the unusual
(port C -> TC1) pin mapping that MCC does.
Also, it turns out the extra PCH ID that we had previously thought was a
form of MCC is actually a second ID for JSP (i.e., port C uses the port
C pins instead of the TC1 pins).
v2:
- Also update the port masks (not just the pin table) in
mcc_hpd_irq_setup. (Vivek)
v3:
- Break jsp_hpd_irq_setup out into its own function for clarity.
(Vivek)
Ville Syrjälä [Mon, 7 Oct 2019 11:49:43 +0000 (14:49 +0300)]
drm/i915: Refactor timestamping constants update
Once we do the hw vs. uapi split we can no longer use
drm_atomic_helper_calc_timestamping_constants() as it'll
consult the uapi state instead of the hw state.
So let's just update the vblank timestamping constants whenever
we update the scanline offset. We use both to convert the hw
scanline count to something which matches the software timing
values.
First I thought to put these into intel_crtc_vblank_on() but
we may want to get the scanline counter value before that (eg.
from some early tracepoints), so let's stick to updating them
a bit earlier than intel_crtc_vblank_on().
Chris Wilson [Wed, 16 Oct 2019 10:08:51 +0000 (11:08 +0100)]
drm/i915/execlist: Trim immediate timeslice expiry
We perform timeslicing immediately upon receipt of a request that may be
put into the second ELSP slot. The idea behind this was that since we
didn't install the timer if the second ELSP slot was empty, we would not
have any idea of how long ELSP[0] had been running and so giving the
newcomer a chance on the GPU was fair. However, this causes us extra
busy work that we may be able to avoid if we wait a jiffie for the first
timeslice as normal.