In some cases, like cursor updates, it is interesting to update the
plane in an asynchronous fashion to avoid big delays. The current queued
update could be still waiting for a fence to signal and thus block any
subsequent update until its scan out. In cases like this if we update the
cursor synchronously through the atomic API it will cause significant
delays that would even be noticed by the final user.
This patch creates a fast path to jump ahead the current queued state and
do single planes updates without going through all atomic steps in
drm_atomic_helper_commit(). We take this path for legacy cursor updates.
For now only single plane updates are supported, but we plan to support
multiple planes updates and async PageFlips through this interface as well
in the near future.
v6: - move check code to drm_atomic_helper.c (Daniel Vetter)
v5:
- improve comments (Eric Anholt)
v4:
- fix state->crtc NULL check (Archit Taneja)
v3:
- fix iteration on the wrong crtc state
- put back code to forbid updates if there is a queued update for
the same plane (Ville Syrjälä)
- move size checks back to drivers (Ville Syrjälä)
- move ASYNC_UPDATE flag addition to its own patch (Ville Syrjälä)
v2:
- allow updates even if there is a queued update for the same
plane.
- fixes on the documentation (Emil Velikov)
- unconditionally call ->atomic_async_update (Emil Velikov)
- check for ->atomic_async_update earlier (Daniel Vetter)
- make ->atomic_async_check() the last step (Daniel Vetter)
- add ASYNC_UPDATE flag (Eric Anholt)
- update state in core after ->atomic_async_update (Eric Anholt)
- update docs (Eric Anholt)
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Eric Anholt <eric@anholt.net>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Archit Taneja <architt@codeaurora.org> (v5)
Acked-by: Eric Anholt <eric@anholt.net> (v5)
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20170630180322.29007-2-gustavo@padovan.org
The PCI Power Management Spec, r1.2, sec 5.6.1, requires a 10 millisecond
delay when powering on a device, i.e., transitioning from state D3hot to
D0.
Apparently some devices require more time, and d1f9809ed1 ("drm/radeon:
add quirk for d3 delay during switcheroo poweron for apple macbooks") added
an additional delay for the Radeon device in a MacBook Pro. 4807c5a8a0
("drm/radeon: add a PX quirk list") made the affected device more explicit.
Add a generic PCI quirk to increase the d3_delay. This means we will use
the additional delay for *all* wakeups from D3, not just those initiated by
radeon_switcheroo_set_state().
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andreas Boll <andreas.boll.dev@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
CC: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Remove unnecessary save/restore of pdev->d3_delay.
The only assignments to pdev->d3_delay are in radeon_switcheroo_set_state()
and some quirks, none of which should be relevant in the
amdgpu_switcheroo_set_state() path.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andreas Boll <andreas.boll.dev@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
During the review of Coffee Lake workarounds Mika pointed out
that WaDisableKillLogic and GEN9_DISABLE_OCL_OOB_SUPPRESS_LOGIC
should be removed from CFL and with that I should carry the rv-b.
However when doing the v2 I removed another Workaround that should
remain because although not mentioned by spec the history of hangs
around it advocates on its favor.
On some follow-up patches I continued operating on the wrong
workardound, but Ville noticed that, so here is the fix for the
current CFL code that is upstream already.
Fixes: 46c26662d2 ("drm/i915/cfl: Introduce Coffee Lake workarounds.")
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
The debugfs interface has calls a function that was evidently
defined under the wrong name in some configurations:
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:64:12: error: 'amdgpu_debugfs_test_ib_ring_init' used but never defined [-Werror]
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:3803:12: error: 'amdgpu_debugfs_test_ib_init' defined but not used [-Werror=unused-function]
This fixes the function name.
Fixes: 4f0955fcc0 ("drm/amdgpu: export test ib debugfs interface")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
caused by not program dynamic_cu_mask_addr in the KIQ MQD.
v2: create struct vi_mqd_allocation in FB which will contain
1. PM4 MQD structure.
2. Write Pointer Poll Memory.
3. Read Pointer Report Memory
4. Dynamic CU Mask.
5. Dynamic RB Mask.
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When computing a hash for looking up relocation target handles in an
execbuf, we start with a large size for the hashtable and proceed to
halve it until the allocation succeeds. The final attempt is with an
order of 0 (i.e. a single element). This means that we then pass bits=0
to hash_32() which then computes "hash >> (32 - 0)" to lookup the single
element. Right shifting a value by the width of the operand is
undefined, so limit the smallest hash table we use to order 1.
v2: Keep the retry allocation flag for the final pass
Fixes: 4ff4b44cbb ("drm/i915: Store a direct lookup from object handle to vma")
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>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170629150425.27508-1-chris@chris-wilson.co.uk
The dpy_reg_mmio_read_x functions directly copy 4 bytes data to the
target address with considering the length. If may cause the target
memory corrupted if the requested length less than 4 bytes. Fix it
for safety even we already have some checking to avoid this happen.
And for convince, the 3 functions are merged.
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Pull drm fixes from Dave Airlie:
"This is the final set of fixes for -rc8, just a few i915 and one
vmwgfx ones.
I'm off on holidays for a week, so if anything shows up for fixes I've
asked Daniel or Sean Paul to herd it in the right direction"
[ The additional etnaviv fixes were already herded towards me as seen in
my previous pull - Linus ]
* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
drm/vmwgfx: Free hash table allocated by cmdbuf managed res mgr
drm/i915: Disable EXEC_OBJECT_ASYNC when doing relocations
drm/i915: Hold struct_mutex for per-file stats in debugfs/i915_gem_object
drm/i915: Retire the VMA's fence tracker before unbinding
Pull drm/etnaviv fixes from Lucas Stach:
"I realized I just missed the cut-off point for the final drm fixes
pull, but I have 2 more etnaviv fixes that need to go into 4.12, as
they fix fallout from the explicit sync work introduced in the last
merge window"
[ Pulling directly because Dave is on vacation. Noted by Daniel Vetter,
and acked by Dave Airlie - Linus ]
* 'etnaviv/fixes' of git://git.pengutronix.de/git/lst/linux:
drm/etnaviv: Fix implicit/explicit sync sense inversion
drm/etnaviv: fix submit flags getting overwritten by BO content
ALSA SoC needs to know connected DAI ID for probing. Using
the new audio-card-graph approach, ports/endpoints are used
to describe how the links are connected. Unfortunately, since
ports/endpoints are used as well for video linkages, there
are some issues mixing the port ids to the two (video and
audio) namespaces.
To solve this issue, this patch adds new .get_dai_id callback
on hdmi_codec_ops.
The will assume that HDMI audio out will be connected to
reg = <2>. This will then be remapped to the ALSA SoC side will
as DAI 0. Allowing the adv7511's hdmi audio support to be used
with the audio-card-graph.
Credit to Kuninori Morimoto who's patch to dw-hdmi-i2s-audio.c
was what this was mostly copy-pasted from.
Cc: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Cc: Archit Taneja <architt@codeaurora.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: David Airlie <airlied@linux.ie>
Cc: Lars-Peter Clausen <lars@metafoo.de>
Cc: Linux-ALSA <alsa-devel@alsa-project.org>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
ALSA SoC needs to know connected DAI ID for probing.
It is not a big problem if device/driver was only for sound,
but getting DAI ID will be difficult if device includes both
Video/Sound, like HDMI.
To solve this issue, this patch adds new .get_dai_id callback
on hdmi_codec_ops.
dw-hdmi-i2s will assume that HDMI sound will be connected
to reg = <2>. Then, ALSA SoC side will recognized it as DAI 0
ports {
#address-cells = <1>;
#size-cells = <0>;
port@0 {
reg = <0>;
/* HDMI Video IN */
};
port@1 {
reg = <1>;
/* HDMI OUT */
};
port@2 {
reg = <2>;
/* HDMI Sound IN */
};
};
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Acked-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Once a client has requested a waitboost, we keep that waitboost active
until all clients are no longer waiting. This is because we don't
distinguish which waiter deserves the boost. However, with the advent of
fence signaling, the signaler threads appear as waiters to the RPS
interrupt handler. So instead of using a single boolean to track when to
keep the waitboost active, use a counter of all outstanding waitboosted
requests.
At this point, I have removed all vestiges of the rate limiting on
clients. Whilst this means that compositors should remain more fluid,
it also means that boosts are more prevalent. See commit b29c19b645
("drm/i915: Boost RPS frequency for CPU stalls") for a longer discussion
on the pros and cons of both approaches.
A drawback of this implementation is that it requires constant request
submission to keep the waitboost trimmed (as it is now cancelled when the
request is completed). This will be fine for a busy system, but near
idle the boosts may be kept for longer than desired (effectively tens of
vblanks worstcase) and there is a reliance on rc6 instead.
v2: Remove defunct rps.client_lock
Reported-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170628123548.9236-1-chris@chris-wilson.co.uk
i915_gem_suspend() is called from all of our finalization paths
(suspend, hibernate, unload). i915_gem_drain_freed_objects() adds an
arbitrary delay as it uses an rcu_barrier() to ensure that there are no
more freed objects in flight, and this delay causes a large amount of
variability in suspend timings. For S3 suspend, we do not need to free
pages as doing so does not impact at all upon the system in its
suspended state, unlike S4 hibernation where we do want the hibernation
image to be as small as possible. Therefore we can forgo waiting inside
i915_gem_suspend(), so long as we ensure that we do cleanup before
unload (see i915_gem_load_cleanup()) and prefer to reap our objects
prior to hibernation (see i915_gem_freeze()).
Removing the rcu_barrier() from i915_gem_suspend() improves S3 latency
by about 30ms on Skylake (ymmv).
Reported-by: David Weinehall <david.weinehall@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170627173731.11566-1-chris@chris-wilson.co.uk
Tested-by: David Weinehall <david.weinehall@linux.intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
