Swati Sharma [Tue, 17 Mar 2020 13:57:36 +0000 (19:27 +0530)]
drm/i915/color: Extract icl_read_luts()
For icl+, have hw read out to create hw blob of gamma
lut values. icl+ platforms supports multi segmented gamma
mode by default, add hw lut creation for this mode.
This will be used to validate gamma programming using dsb
(display state buffer) which is a tgl specific feature.
v2: -readout code for multisegmented gamma has to come
up with some intermediate entries that aren't preserved
in hardware (Jani N)
-linear interpolation (Ville)
-moved common code to check gamma_enable to specific funcs,
since icl doesn't support that
v3: -use u16 instead of __u16 [Jani N]
-used single lut [Jani N]
-improved and more readable for loops [Jani N]
-read values directly to actual locations and then fill gaps [Jani N]
-moved cleaning to patch 1 [Jani N]
-renamed icl_read_lut_multi_seg() to icl_read_lut_multi_segment to
make it similar to icl_load_luts()
-renamed icl_compute_interpolated_gamma_blob() to
icl_compute_interpolated_gamma_lut_values() more sensible, I guess
v4: -removed interpolated func for creating gamma lut values
-removed readouts of fine and coarse segments, failure to read PAL_PREC_DATA
correctly
v5: -added gamma_enable check inside read_luts()
v6: -renamed intel_color_lut_entry_equal() to intel_color_lut_entries_equal() [Ville]
-changed if-else to switch [Ville]
-removed intel_color_lut_entry_multi_equal() [Ville]
v7: -checkpatch warnings
v8: -rebased
v9: -rebased, aligned with Ville's style of gamma cleanup
Anshuman Gupta [Wed, 18 Mar 2020 09:44:48 +0000 (15:14 +0530)]
drm/i915: Enable non-contiguous pipe fusing
As we have already enabled supported infrastructure for
non-contiguous pipe fusing in driver, we don't require non-contiguous
pipe_mask check anymore.
It is an unlike condition, it make sense to remove this condition.
changes since RFC:
- using intel_pipe_mask_is_valid() function to check integrity of
pipe_mask. [Ville]
v2:
- simplify condition in intel_pipe_mask_is_valid(). [Ville]
v3:
- removed non-contiguous pipe fusing check. [Lucas]
drm/i915/perf: Invalidate OA TLB on when closing perf stream
On running several back to back perf capture sessions involving closing
and opening the perf stream, invalid OA reports are seen in the
beginning of the OA buffer in some sessions. Fix this by invalidating OA
TLB when the perf stream is closed or disabled on gen12.
Chris Wilson [Tue, 17 Mar 2020 16:32:08 +0000 (16:32 +0000)]
drm/i915/gt: Always reschedule the new heartbeat
In order to better respond to new heartbeat intervals given via sysfs,
always reprogramme an active heartbeat upon change (i.e. use
mod_delayed_work to reschedule rather than queue_delayed_work which
ignores an already active work.)
On Gen11 powergating half the execution units is a functional
requirement when using the VME samplers. Not fullfilling this
requirement can lead to hangs.
This unfortunately plays fairly poorly with the NOA requirements. NOA
requires a stable power configuration to maintain its configuration.
As a result using OA (and NOA feeding into it) so far has required us
to use a power configuration that can work for all contexts. The only
power configuration fullfilling this is powergating half the execution
units.
This makes performance analysis for 3D workloads somewhat pointless.
Failing to find a solution that would work for everybody, this change
introduces a new i915-perf stream open parameter that punts the
decision off to userspace. If this parameter is omitted, the existing
Gen11 behavior remains (half EU array powergating).
This change takes the initiative to move all perf related sseu
configuration into i915_perf.c
v2: Make parameter priviliged if different from default
v3: Fix context modifying its sseu config while i915-perf is enabled
v4: Always consider global sseu a privileged operation (Tvrtko)
Override req_sseu point in intel_sseu_make_rpcs() (Tvrtko)
Remove unrelated changes (Tvrtko)
v5: Some typos (Tvrtko)
Process sseu param in read_properties_unlocked() (Tvrtko)
v6: Actually commit the bits from v5...
Fixup some checkpath warnings
Back when i915-perf was introduced (4.13), there was no way to
dynamically add new OA configurations to i915. Only the generated
configs baked in at build time were allowed.
It quickly became obvious that we would need to allow applications
to upload their own configurations, for instance to be able to test
new ones, and so by the next stable version (4.14) we added uAPIs
to allow uploading new configurations.
When adding that capability, we took the opportunity to remove most
HW configurations except the TestOa one which is a configuration
IGT would rely on to verify that the HW is outputting correct
values. At the time it made sense to have that confiuration in at
the same time a given HW platform added to the i915-perf driver.
Now that IGT has become the reference point for HW configurations (see
commit 53f8f541ca ("lib: Add i915_perf library"), previously this was
located in the GPUTop repository), the need for having those
configurations in i915-perf is gone.
On the Mesa side, we haven't relied on this test configuration for a
while. The MDAPI library always required 4.14 feature level and always
loaded its configuration into i915.
I'm sure nobody will miss this generated stuff in i915 :)
v2: Fix selftests by creating an empty config
v3: Fix unlocking on allocation error (Dan Carpenter)
Chris Wilson [Mon, 16 Mar 2020 11:38:45 +0000 (11:38 +0000)]
drm/i915: Remove manual save/resume of fence register state
Since we always reload the fence register state on runtime resume,
having it explicitly in the S0ix resume code is redundant. Indeed, it
is not even being used!
We have a few TGL machines in our CI and it is mostly green with
failures in tests that will not impact future Linux installations.
Also there is no warnings, errors, flickering or any visual defects
while doing ordinary tasks like browsing and editing documents in a
dual monitor setup.
As a reminder i915.require_force_probe was created to protect
future Linux installation's iso images that might contain a
kernel from the enabling time of the new platform. Without this
protection most of linux installation was recommending
nomodeset option during installation that was getting stick
there after installation.
Reference: https://intel-gfx-ci.01.org/tree/drm-tip/fi-tgl-u.html
Reference: https://intel-gfx-ci.01.org/tree/drm-tip/shard-tglb.html Cc: James Ausmus <james.ausmus@intel.com> Cc: Jani Saarinen <jani.saarinen@intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: José Roberto de Souza <jose.souza@intel.com> Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200218230822.66801-1-jose.souza@intel.com
Matt Roper [Wed, 11 Mar 2020 16:23:00 +0000 (09:23 -0700)]
drm/i915: Add Wa_1605460711 / Wa_1408767742 to ICL and EHL
This workaround appears under two different numbers (and with somewhat
confused stepping applicability on ICL). Ultimately it appears we
should just implement this for all stepping of ICL and EHL.
Note that this is identical to Wa_1407928979:tgl that already exists in
our driver too...yet another number referencing the same actual
workaround.
Matt Roper [Wed, 11 Mar 2020 16:22:59 +0000 (09:22 -0700)]
drm/i915: Apply Wa_1406680159:icl,ehl as an engine workaround
The register this workaround updates is a render engine register in the
MCR range, so we should initialize this in rcs_engine_wa_init() rather
than gt_wa_init().
Matt Roper [Wed, 11 Mar 2020 16:22:57 +0000 (09:22 -0700)]
drm/i915: Add Wa_1604278689:icl,ehl
The bspec description for this workaround tells us to program
0xFFFF_FFFF into both FBC_RT_BASE_ADDR_REGISTER_* registers, but we've
previously found that this leads to failures in CI. Our suspicion is
that the failures are caused by this valid turning on the "address valid
bit" even though we're intentionally supplying an invalid address.
Experimentation has shown that setting all bits _except_ for the
RT_VALID bit seems to avoid these failures.
v2:
- Mask off the RT_VALID bit. Experimentation with CI trybot indicates
that this is necessary to avoid reset failures on BCS.
v3:
- Program RT_BASE before RT_BASE_UPPER so that the valid bit is turned
off by the first write. (Chris)
Matt Roper [Wed, 11 Mar 2020 16:22:56 +0000 (09:22 -0700)]
drm/i915: Add Wa_1209644611:icl,ehl
On gen11 the XY_FAST_COPY_BLT command has some size restrictions on its
usage. Although this instruction is mainly used by userspace, i915 also
uses it to copy object contents during some selftests, so let's ensure
the restrictions are followed.
Chris Wilson [Thu, 12 Mar 2020 11:53:07 +0000 (11:53 +0000)]
drm/i915/gt: Wait for RCUs frees before asserting idle on unload
During driver unload, we have many asserts that we have released our
bookkeeping structs and are idle. In some cases, these struct are
protected by RCU and we do not release them until after an RCU grace
period.
Reported-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Fixes: 4f15002c8cb0 ("drm/i915/gem: Consolidate ctx->engines[] release") 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> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200312115307.16460-1-chris@chris-wilson.co.uk
Chris Wilson [Wed, 11 Mar 2020 16:03:10 +0000 (16:03 +0000)]
drm/i915/gem: Drop relocation slowpath
Since the relocations are no longer performed under a global
struct_mutex, or any other lock, that is also held by pagefault handlers,
we can relax and allow our fast path to take a fault. As we no longer
need to abort the fast path for lock avoidance, we no longer need the
slow path handling at all.
Tvrtko Ursulin [Thu, 12 Mar 2020 11:57:48 +0000 (11:57 +0000)]
drm/i915/gen12: Disable preemption timeout
Allow super long OpenCL workloads which cannot be preempted within
the default timeout to run out of the box.
v2:
* Make it stick out more and apply only to RCS. (Chris)
v3:
* Mention platform override in kconfig. (Joonas)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Michal Mrozek <michal.mrozek@intel.com> Cc: <stable@vger.kernel.org> # v5.6+ Acked-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Michal Mrozek <Michal.mrozek@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200312115748.29970-1-tvrtko.ursulin@linux.intel.com
Chris Wilson [Wed, 11 Mar 2020 22:17:39 +0000 (22:17 +0000)]
drm/i915/gem: Take a copy of the engines for context_barrier_task
When applying the context-barrier, we only care about the current
engines, as the next set of engines will be naturally after the barrier.
So we can skip holding the ctx->engines_mutex while constructing the
request by taking a sneaky reference to the i915_gem_engines instead.
Chris Wilson [Wed, 11 Mar 2020 22:17:38 +0000 (22:17 +0000)]
drm/i915/gem: Mark up sw-fence notify function
The sw-fence notify function requires to be at least 4-byte aligned so
that we can use the low bits in the function pointer for internal fence
flags. Make it so.
References: https://gitlab.freedesktop.org/drm/intel/issues/1433 Fixes: cffd25855556 ("drm/i915/gem: Don't leak non-persistent requests on changing engines") 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/20200311221739.30375-1-chris@chris-wilson.co.uk
Kai Vehmanen [Tue, 10 Mar 2020 16:23:38 +0000 (18:23 +0200)]
drm/i915: Add missing HDMI audio pixel clocks for gen12
Gen12 hardware supports HDMI audio pixel clocks of 296.7/297Mhz
and 593.4/594Mhz. Add the missing rates and add logic to ignore
them if running on older hardware.
Chris Wilson [Wed, 11 Mar 2020 09:26:22 +0000 (09:26 +0000)]
drm/i915/gt: Pull checking rps->pm_events under the irq_lock
Avoid angering kcsan by serialising the read of the pm_events with the
write in rps_disable_interrupts.
[ 6268.713419] BUG: KCSAN: data-race in intel_rps_park [i915] / rps_work [i915]
[ 6268.713437]
[ 6268.713449] write to 0xffff8881eda8efac of 4 bytes by task 1127 on cpu 3:
[ 6268.713680] intel_rps_park+0x136/0x260 [i915]
[ 6268.713905] __gt_park+0x61/0xa0 [i915]
[ 6268.714128] ____intel_wakeref_put_last+0x42/0x90 [i915]
[ 6268.714352] __intel_wakeref_put_work+0xd3/0xf0 [i915]
[ 6268.714369] process_one_work+0x3b1/0x690
[ 6268.714384] worker_thread+0x80/0x670
[ 6268.714398] kthread+0x19a/0x1e0
[ 6268.714412] ret_from_fork+0x1f/0x30
[ 6268.714423]
[ 6268.714435] read to 0xffff8881eda8efac of 4 bytes by task 950 on cpu 2:
[ 6268.714664] rps_work+0xc2/0x680 [i915]
[ 6268.714679] process_one_work+0x3b1/0x690
[ 6268.714693] worker_thread+0x80/0x670
[ 6268.714707] kthread+0x19a/0x1e0
[ 6268.714720] ret_from_fork+0x1f/0x30
v2: Mark all reads and writes of rpm->pm_events.
The flow of enabling/disabling rps is stronly ordered, so the writes and
interrupt generation are also strongly ordered -- just this may not be
visible to the compiler, so provide annotations.
Takashi Iwai [Wed, 11 Mar 2020 07:32:56 +0000 (08:32 +0100)]
drm/i915/gt: Use scnprintf() for avoiding potential buffer overflow
Since snprintf() returns the would-be-output size instead of the
actual output size, the succeeding calls may go beyond the given
buffer limit. Fix it by replacing with scnprintf().
Wambui Karuga [Tue, 10 Mar 2020 08:52:49 +0000 (10:52 +0200)]
drm/i915/overlay: convert to drm_device based logging.
Convert various instances of the printk based drm logging macros to the
struct drm_device based logging macros in i915/display/intel_overlay.c.
This transformation was achieved using the following coccinelle script:
@@
identifier fn, T;
@@
Wambui Karuga [Tue, 10 Mar 2020 08:52:48 +0000 (10:52 +0200)]
drm/i915/lvds: convert to drm_device based logging macros.
Converts various instances of the printk based drm logging macros to the
struct drm_device based logging macros in i915/display/intel_lvds.c.
This transformation was done by the following coccinelle script that
matches based on the existence of a drm_i915_private device:
@@
identifier fn, T;
@@
Wambui Karuga [Tue, 10 Mar 2020 08:52:46 +0000 (10:52 +0200)]
drm/i915/hotplug: convert to drm_device based logging.
Converts various instances of the printk based drm logging macros to the
struct drm_device based logging macros in i915/display/intel_hotplug.c.
In some cases, this involves extracting the drm_i915_private pointer from
the drm_device struct to be used in the logging macros.
Wambui Karuga [Tue, 10 Mar 2020 08:52:44 +0000 (10:52 +0200)]
drm/i915/gmbus: convert to drm_device based logging,
Conversion instances of printk based drm logging macros to use the
struct drm_device based logging macros in i915/display/intel_gmbus.c.
This was done using the following coccinelle semantic patch that
transforms based on the existence of an existing drm_i915_private
device:
@@
identifier fn, T;
@@
Wambui Karuga [Tue, 10 Mar 2020 08:52:43 +0000 (10:52 +0200)]
drm/i915/fifo_underrun: convert to drm_device based logging.
Convert various instances of the printk based drm logging macros to the
struct drm_device based logging macros in
i915/display/intel_fifo_underrun.c.
This was done using the following coccinelle script:
@@
identifier fn, T;
@@
Wambui Karuga [Mon, 9 Mar 2020 15:12:41 +0000 (17:12 +0200)]
drm/i915/dsb: convert to drm_device based logging macros.
This converts uses of the printk based drm logging macros to the struct
drm_device logging macros in i915/display/intel_dsb.c. This was done
using the following coccinelle script:
@@
identifier fn, T;
@@
Tvrtko Ursulin [Tue, 10 Mar 2020 16:47:33 +0000 (16:47 +0000)]
drm/i915: Remove debugfs i915_drpc_info and i915_forcewake_domains
The two files have been duplicated under the gt/ subdir and since there
are not apparent users looking for them at the old location lets simply
remove them and duplicated code.
Chris Wilson [Tue, 10 Mar 2020 14:24:03 +0000 (14:24 +0000)]
drm/i915: Mark up racy read of active rq->engine
As a virtual engine may change the rq->engine to point to the active
request in flight, we need to warn the compiler that an active request's
engine is volatile.
Chris Wilson [Tue, 10 Mar 2020 14:13:18 +0000 (14:13 +0000)]
drm/i915/gt: Mark up racy reads for intel_context.inflight
When being used across multiple real engines inside a virtual engine,
the intel_context.inflight is updated atomically, and so we must
annotate the racy read from outside the owning context.
[11142.482846] BUG: KCSAN: data-race in __execlists_submission_tasklet [i915] / __execlists_submission_tasklet [i915]
[11142.482867]
[11142.482878] write (marked) to 0xffff8881f257b5e0 of 8 bytes by interrupt on cpu 2:
[11142.483107] __execlists_submission_tasklet+0x1d33/0x2120 [i915]
[11142.483336] execlists_submission_tasklet+0xd3/0x170 [i915]
[11142.483355] tasklet_action_common.isra.0+0x42/0xa0
[11142.483371] __do_softirq+0xd7/0x2cd
[11142.483384] irq_exit+0xbe/0xe0
[11142.483401] do_IRQ+0x51/0x100
[11142.483424] ret_from_intr+0x0/0x1c
[11142.483446] do_idle+0x133/0x1f0
[11142.483465] cpu_startup_entry+0x14/0x16
[11142.483483] start_secondary+0x120/0x180
[11142.483498] secondary_startup_64+0xa4/0xb0
[11142.483512]
[11142.483528] read to 0xffff8881f257b5e0 of 8 bytes by interrupt on cpu 1:
[11142.483755] __execlists_submission_tasklet+0x14e/0x2120 [i915]
[11142.483981] execlists_submission_tasklet+0xd3/0x170 [i915]
[11142.483999] tasklet_action_common.isra.0+0x42/0xa0
[11142.484014] __do_softirq+0xd7/0x2cd
[11142.484028] do_softirq_own_stack+0x2a/0x40
[11142.484046] do_softirq.part.0+0x26/0x30
[11142.484071] __local_bh_enable_ip+0x46/0x50
[11142.484299] i915_gem_do_execbuffer+0x39c1/0x4e50 [i915]
[11142.484528] i915_gem_execbuffer2_ioctl+0x2c3/0x580 [i915]
[11142.484546] drm_ioctl_kernel+0xe4/0x120
[11142.484559] drm_ioctl+0x297/0x4c7
[11142.484572] ksys_ioctl+0x89/0xb0
[11142.484586] __x64_sys_ioctl+0x42/0x60
[11142.484610] do_syscall_64+0x6e/0x2c0
[11142.484627] entry_SYSCALL_64_after_hwframe+0x44/0xa9
Chris Wilson [Tue, 10 Mar 2020 11:59:47 +0000 (11:59 +0000)]
drm/i915: Tweak scheduler's kick_submission()
Skip useless priority bumping on adding a new dependency by making sure
that we do update the priority if we would have rescheduled the active
cotnext.
Chris Wilson [Tue, 10 Mar 2020 10:17:20 +0000 (10:17 +0000)]
drm/i915: Defer semaphore priority bumping to a workqueue
Since the semaphore fence may be signaled from inside an interrupt
handler from inside a request holding its request->lock, we cannot then
enter into the engine->active.lock for processing the semaphore priority
bump as we may traverse our call tree and end up on another held
request.
drm/i915/display: Deactive FBC in fastsets when disabled by parameter
Most of the kms_frontbuffer_tracking tests disables the feature being
tested, draw, get the CRC then enable the feature, draw again, get the
CRC and check if it matches.
Some times it is able to do that with a fastset, so
intel_pre_plane_update() is executed but intel_fbc_can_flip_nuke() was
not checking if FBC is now enabled in this CRTC leaving FBC active and
causing the warning bellow in __intel_fbc_disable()
i915 can enable aux device nodes for DP MST by calling
drm_dp_mst_connector_late_register()/
drm_dp_mst_connector_early_unregister(),
so let's hook that up.
Changes since v1:
* Call intel_connector_register/unregister() from
intel_dp_mst_connector_late_register/unregister() so we don't lose
error injection - Ville Syrjälä
Changes since v2:
* Don't forget to clean up if intel_connector_register() fails - Ville
Chris Wilson [Fri, 6 Mar 2020 13:38:38 +0000 (13:38 +0000)]
drm/i915: Improve the start alignment of bonded pairs
Always wait on the start of the signaler request to reduce the problem
of dequeueing the bonded pair too early -- we want both payloads to
start at the same time, with no latency, and yet still allow others to
make full use of the slack in the system. This reduce the amount of time
we spend waiting on the semaphore used to synchronise the start of the
bonded payload.
Ville Syrjälä [Mon, 2 Mar 2020 17:44:42 +0000 (19:44 +0200)]
drm/i915: Lock gmbus/aux mutexes while changing cdclk
gmbus/aux may be clocked by cdclk, thus we should make sure no
transfers are ongoing while the cdclk frequency is being changed.
We do that by simply grabbing all the gmbus/aux mutexes. No one
else should be holding any more than one of those at a time so
the lock ordering here shouldn't matter.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:13 +0000 (19:33 +0200)]
drm/i915: Pass the crtc to the low level read_lut() funcs
The low level read_lut() functions don't need the entire crtc state
as they know exactly what they're reading. Just need to pass in the
crtc to get at the pipe. This now neatly mirrors the load_lut()
direction.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:12 +0000 (19:33 +0200)]
drm/i915: Fix readout of PIPEGCMAX
PIPEGCMAX is a 11.6 (or 1.16 if you will) value. Ie. it can
represent a value of 1.0 when the maximum we can store in the
software LUT is 0.ffff. Clamp the value so that it gets
saturated to the max the uapi supports.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:11 +0000 (19:33 +0200)]
drm/i915: Refactor LUT read functions
Extract all the 'hw value -> LUT entry' stuff into small helpers
to make the main 'read out the entire LUT' loop less bogged down
by such mundane details.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:10 +0000 (19:33 +0200)]
drm/i915: Clean up integer types in color code
A variable called 'i' having an unsigned type is just looking for
trouble, and using a sized type generally makes no sense either.
Change all of them to just plain old int. And do the same for some
'lut_size' variables which generally provide the loop end codition
for 'i'.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:09 +0000 (19:33 +0200)]
drm/i915: s/chv_read_cgm_lut/chv_read_cgm_gamma/
chv_read_cgm_lut() specifically reads the CGM _gamma_ LUT so
let's rename it to reflect that fact. This also mirrors
the other direction's chv_load_cgm_gamma().
Ville Syrjälä [Tue, 3 Mar 2020 17:33:07 +0000 (19:33 +0200)]
drm/i915: Split i9xx_read_lut_8() to gmch vs. ilk variants
To mirror the load_luts path let's clone an ilk+ version
from i9xx_read_lut_8(). I guess the extra branch isn't a huge
issue but feels better to make a clean split.
Ville Syrjälä [Tue, 3 Mar 2020 17:33:05 +0000 (19:33 +0200)]
drm/i915: Polish CHV CGM CSC loading
Only load the CGM CSC based on the cgm_mode bit like we
do with the gamma/degamma LUTs. And make the function
naming and arguments consistent as well.
TODO: the code to convert the coefficients look totally
bogus. IIRC CHV uses two's complement format but the code
certainly doesn't generate that, so probably negative
coefficients are totally busted.
Chris Wilson [Mon, 9 Mar 2020 11:09:30 +0000 (11:09 +0000)]
drm/i915: Mark up unlocked update of i915_request.hwsp_seqno
During i915_request_retire() we decouple the i915_request.hwsp_seqno
from the intel_timeline so that it may be freed before the request is
released. However, we need to warn the compiler that the pointer may
update under its nose.
[ 171.438899] BUG: KCSAN: data-race in i915_request_await_dma_fence [i915] / i915_request_retire [i915]
[ 171.438920]
[ 171.438932] write to 0xffff8881e7e28ce0 of 8 bytes by task 148 on cpu 2:
[ 171.439174] i915_request_retire+0x1ea/0x660 [i915]
[ 171.439408] retire_requests+0x7a/0xd0 [i915]
[ 171.439640] engine_retire+0xa1/0xe0 [i915]
[ 171.439657] process_one_work+0x3b1/0x690
[ 171.439671] worker_thread+0x80/0x670
[ 171.439685] kthread+0x19a/0x1e0
[ 171.439701] ret_from_fork+0x1f/0x30
[ 171.439721]
[ 171.439739] read to 0xffff8881e7e28ce0 of 8 bytes by task 696 on cpu 1:
[ 171.439990] i915_request_await_dma_fence+0x162/0x520 [i915]
[ 171.440230] i915_request_await_object+0x2fe/0x470 [i915]
[ 171.440467] i915_gem_do_execbuffer+0x45dc/0x4c20 [i915]
[ 171.440704] i915_gem_execbuffer2_ioctl+0x2c3/0x580 [i915]
[ 171.440722] drm_ioctl_kernel+0xe4/0x120
[ 171.440736] drm_ioctl+0x297/0x4c7
[ 171.440750] ksys_ioctl+0x89/0xb0
[ 171.440766] __x64_sys_ioctl+0x42/0x60
[ 171.440788] do_syscall_64+0x6e/0x2c0
[ 171.440802] entry_SYSCALL_64_after_hwframe+0x44/0xa9
Matt Roper [Fri, 6 Mar 2020 17:11:39 +0000 (09:11 -0800)]
drm/i915/tgl: Don't treat unslice registers as masked
The UNSLICE_UNIT_LEVEL_CLKGATE and UNSLICE_UNIT_LEVEL_CLKGATE2 registers
that we update in a few engine workarounds are not masked registers
(i.e., we don't have to write a mask bit in the top 16 bits when
updating one of the lower 16 bits). As such, these workarounds should
be applied via wa_write_or() rather than wa_masked_en()
v2:
- Rebase
Reported-by: Nick Desaulniers <ndesaulniers@google.com> Reported-by: kernelci.org bot <bot@kernelci.org>
References: https://github.com/ClangBuiltLinux/linux/issues/918 Fixes: f85de99d49d2 ("drm/i915/tgl: Move and restrict Wa_1408615072") Fixes: e3747f971acc ("drm/i915/gen11: Moving WAs to rcs_engine_wa_init()") Cc: José Roberto de Souza <jose.souza@intel.com> Signed-off-by: Matt Roper <matthew.d.roper@intel.com> Tested-by: Nick Desaulniers <ndesaulniers@google.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200306171139.1414649-1-matthew.d.roper@intel.com
Imre Deak [Wed, 4 Mar 2020 15:09:18 +0000 (17:09 +0200)]
drm/i915: Fix documentation for intel_dpll_get_freq()
Fix the following kerneldoc warning and while at it also the doc for the
corresponding vfunc hook.
$ make htmldocs 2>&1 > /dev/null | grep i915
./drivers/gpu/drm/i915/display/intel_dpll_mgr.h:285: warning: Function parameter or member 'get_freq' not described in 'intel_shared_dpll_funcs'
Chris Wilson [Sat, 7 Mar 2020 12:24:25 +0000 (12:24 +0000)]
drm/i915/gt: Wait for the wa batch to be pinned
Be sure to wait for the vma to be in place before we tell the GPU to
execute from the wa batch. Since initialisation is mostly synchronous
(or rather at some point during start up we will need to sync anyway),
we can affort to do an explicit i915_vma_sync() during wa batch
construction rather than check for a required await on every context
switch. (We don't expect to change the wa bb at run time so paying the
cost once up front seems preferrable.)
Chris Wilson [Fri, 6 Mar 2020 15:46:47 +0000 (15:46 +0000)]
drm/i915/gt: Close race between cacheline_retire and free
If the cacheline may still be busy, atomically mark it for future
release, and only if we can determine that it will never be used again,
immediately free it.
Closes: https://gitlab.freedesktop.org/drm/intel/issues/1392 Fixes: 6f53fdf7650e ("drm/i915: Keep timeline HWSP allocated until idle across the system") 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: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: <stable@vger.kernel.org> # v5.2+ Link: https://patchwork.freedesktop.org/patch/msgid/20200306154647.3528345-1-chris@chris-wilson.co.uk
Chris Wilson [Fri, 6 Mar 2020 11:30:10 +0000 (11:30 +0000)]
drm/i915/execlists: Enable timeslice on partial virtual engine dequeue
If we stop filling the ELSP due to an incompatible virtual engine
request, check if we should enable the timeslice on behalf of the queue.
This fixes the case where we are inspecting the last->next element when
we know that the last element is the last request in the execution queue,
and so decided we did not need to enable timeslicing despite the intent
to do so!
Fixes: b0ea7ccafad8 ("drm/i915/execlists: Minimalistic timeslicing") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: <stable@vger.kernel.org> # v5.4+ Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200306113012.3184606-1-chris@chris-wilson.co.uk
Chris Wilson [Fri, 6 Mar 2020 13:38:36 +0000 (13:38 +0000)]
drm/i915/selftests: Apply a heavy handed flush to i915_active
Due to the ordering of cmpxchg()/dma_fence_signal() inside node_retire(),
we must also use the xchg() as our primary memory barrier to flush the
outstanding callbacks after expected completion of the i915_active.
Vivek Kasireddy [Wed, 4 Mar 2020 23:42:40 +0000 (15:42 -0800)]
drm/i915/hotplug: Use phy to get the hpd_pin instead of the port (v5)
On some platforms such as Elkhart Lake, although we may use DDI D
to drive a connector, we have to use PHY A (Combo Phy PORT A) to
detect the hotplug interrupts as per the spec because there is no
one-to-one mapping between DDIs and PHYs. Therefore, use the
function intel_port_to_phy() which contains the logic for such
mapping(s) to find the correct hpd_pin.
This change should not affect other platforms as there is always
a one-to-one mapping between DDIs and PHYs.
v2:
- Convert the case statements to use PHYs instead of PORTs (Jani)
v3:
- Refactor the function to reduce the number of return statements by
lumping all the case statements together except PHY_F which needs
special handling (Jose)
v4:
- Add a comment describing how the HPD pin value associated with any
port can be retrieved using port or phy enum value. (Jani)
v5:
- Use case ranges instead of individual labels and also normalize the
return statement by adding -PHY_A to the expression (Ville)
Cc: Jani Nikula <jani.nikula@intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: José Roberto de Souza <jose.souza@intel.com> Cc: Ville Syrjala <ville.syrjala@linux.intel.com> Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Signed-off-by: José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200304234240.12062-1-vivek.kasireddy@intel.com
Matthew Auld [Thu, 5 Mar 2020 20:47:11 +0000 (20:47 +0000)]
drm/i915/selftests: try to rein in alloc_smoke
Depending on RNG we might try to fill an 8G region for every possible
order, using the smallest possible chunk size of 4K, which seems to be
very slow. Try to remedy the situation by adding an overall timeout for
the test, while also selecting each order level in a random fashion.
Check the edge case where batch_start_offset sits exactly on the batch
size.
v2: add new range_overflows variant to capture the special case where
the size is permitted to be zero, like with batch_len.
v3: other way around. the common case is the exclusive one which should
just be >=, with that we then just need to convert the three odd ball
cases that don't apply to use the new inclusive _end version.
Testcase: igt/gem_exec_params/invalid-batch-start-offset Fixes: db38e7eb782e ("drm/i915/cmdparser: Use cached vmappings") Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20200306094735.258285-1-matthew.auld@intel.com
Chris Wilson [Fri, 6 Mar 2020 07:16:14 +0000 (07:16 +0000)]
drm/i915/gem: Limit struct_mutex to eb_reserve
We only need to serialise the multiple pinning during the eb_reserve
phase. Ideally this would be using the vm->mutex as an outer lock, or
using a composite global mutex (ww_mutex), but at the moment we are
using struct_mutex for the group.
Closes: https://gitlab.freedesktop.org/drm/intel/issues/1381 Fixes: 9b7d4b20d38a ("drm/i915/gem: Only call eb_lookup_vma once during execbuf ioctl") 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/20200306071614.2846708-3-chris@chris-wilson.co.uk
Chris Wilson [Fri, 6 Mar 2020 07:16:13 +0000 (07:16 +0000)]
drm/i915: Always propagate the invocation to i915_schedule
We only call i915_schedule() when we know we have changed the priority
on a request and so require to propagate any change in priority to its
signalers (for PI). By unconditionally checking all of our signalers, we
avoid skipping changes made prior to construction of the request (as the
request may be waited upon before submission when used in parallel).
On gen7 and gen7.5 devices, there could be leftover data residuals in
EU/L3 from the retiring context. This patch introduces workaround to clear
that residual contexts, by submitting a batch buffer with dedicated HW
context to the GPU with ring allocation for each context switching.
This security mitigation changes does not triggers any performance
regression. Performance is on par with current drm-tips.
v2: Add igt generated header file for CB kernel assembled with Mesa tool
and addressed use of Kernel macro for ptr_align comment.
v3: Resolve Sparse warnings with newly generated, and imported CB
kernel.
v4: Include new igt generated CB kernel for gen7 and gen7.5. Also
add code formatting and compiler warnings changes (Chris Wilson)
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Prathap Kumar Valsan <prathap.kumar.valsan@intel.com> Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com> Cc: Chris Wilson <chris.p.wilson@intel.com> Cc: Balestrieri Francesco <francesco.balestrieri@intel.com> Cc: Bloomfield Jon <jon.bloomfield@intel.com> Cc: Dutt Sudeep <sudeep.dutt@intel.com> Acked-by: Chris Wilson <chris@chris-wilso.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20200306000957.2836150-2-chris@chris-wilson.co.uk
Mika Kuoppala [Fri, 6 Mar 2020 00:09:56 +0000 (00:09 +0000)]
drm/i915: Add mechanism to submit a context WA on ring submission
This patch adds framework to submit an arbitrary batchbuffer on each
context switch to clear residual state for render engine on Gen7/7.5
devices.
The idea of always emitting the context and vm setup around each request
is primary to make reset recovery easy, and not require rewriting the
ringbuffer. As each request would set up its own context, leaving it to
the HW to notice and elide no-op context switches, we could restart the
ring at any point, and reorder the requests freely.
However, to avoid emitting clear_residuals() between consecutive requests
in the ringbuffer of the same context, we do want to track the current
context in the ring. In doing so, we need to be careful to only record a
context switch when we are sure the next request will be emitted.
This security mitigation change does not trigger any performance
regression. Performance is on par with current mainline/drm-tip.
v2: Update vm_alias params to point to correct address space "vm" due to
changes made in the patch "ad49bd9ca9e9deccd"
v3-v4: none
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Prathap Kumar Valsan <prathap.kumar.valsan@intel.com> Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Balestrieri Francesco <francesco.balestrieri@intel.com> Cc: Bloomfield Jon <jon.bloomfield@intel.com> Cc: Dutt Sudeep <sudeep.dutt@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/20200306000957.2836150-1-chris@chris-wilson.co.uk
Chris Wilson [Wed, 4 Mar 2020 03:23:06 +0000 (11:23 +0800)]
drm/i915/gvt: Wean gvt off dev_priv->engine[]
Stop trying to escape out of the gvt layer to find the engine that we
initially setup for use with gvt. Record the engines during initialisation
and use them henceforth.
Chris Wilson [Wed, 4 Mar 2020 03:23:05 +0000 (11:23 +0800)]
drm/i915/gvt: cleanup debugfs scan_nonprivbb
Remove extra chatty message for debugfs scan_nonprivbb which is used
to enable scan for non privileged batch on specific engine. Just write
target i915 engine mask instead.
Tina Zhang [Thu, 27 Feb 2020 01:00:41 +0000 (09:00 +0800)]
drm/i915/gvt: Fix dma-buf display blur issue on CFL
Commit 081157f765adb ("drm/i915/gvt: Enable gfx virtualiztion for CFL")
added the support on CFL. The vgpu emulation hotplug support on CFL was
supposed to be included in that patch. Without the vgpu emulation
hotplug support, the dma-buf based display gives us a blur face.
So fix this issue by adding the vgpu emulation hotplug support on CFL.
Chris Wilson [Thu, 5 Mar 2020 13:48:22 +0000 (13:48 +0000)]
drm/i915: Return early for await_start on same timeline
Requests within a timeline are ordered by that timeline, so awaiting for
the start of a request within the timeline is a no-op. This used to work
by falling out of the mutex_trylock() as the signaler and waiter had the
same timeline and not returning an error.
projects / kernel.git / log