Quoting Ville Syrjälä (2018-09-20 20:41:30)
> On Thu, Sep 20, 2018 at 09:09:03AM +0100, Chris Wilson wrote:
> > Quoting Ville Syrjälä (2018-09-19 17:59:51)
> > > On Thu, Sep 13, 2018 at 11:01:40PM +0300, Ville Syrjala wrote:
> > > > From: Ville Syrjälä
> > > >
> > > > With gtt remapping in place
ALPM is a requirement and we don't need to keep it's cached, what
were done in commit 97c9de66ca80
("drm/i915/psr: Fix ALPM cap check for PSR2") but the alpm was not
removed from i915_psr.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/i915_drv.h | 1 -
1
ICL spec states that this bit is now reserved.
Spec: 7722
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/i915_reg.h | 4 ++--
drivers/gpu/drm/i915/intel_psr.c | 17 +++--
2 files changed, 13 insertions(+), 8 deletions(-)
diff --git
For PSR2 we don't have the option to keep main link enabled while
PSR2 is active, so don't configure sink DPCD with a wrong value.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/intel_psr.c | 9 +
1 file changed, 5 insertions(+), 4 deletions(-)
According to eDP spec, sink could required a granularity in the
start of x coordinate or in the width of the selective update region.
As it is not supported by hardware, lets not enable PSR2 in sinks
that requires it.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
This WA also works fine for PSR2, triggering a selective update when
possible.
Cc: Dhinakaran Pandiyan
Cc: Rodrigo Vivi
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/intel_psr.c | 24 ++--
1 file changed, 10 insertions(+), 14 deletions(-)
diff --git
We are already handling all PSR2 errors, so we can drop this TODO.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/intel_psr.c | 2 --
1 file changed, 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_psr.c b/drivers/gpu/drm/i915/intel_psr.c
index
Now both PSR and PSR2 have the same exit mask, so let's share then
instead of have the same code 2 times.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/intel_psr.c | 34
1 file changed, 13 insertions(+), 21 deletions(-)
Fix up the error unwind for logical_ring_init() failing by moving the
cleanup into the callers who own the various bits of state during
initialisation, so we don't forget to free the state allocated by the
caller.
Signed-off-by: Chris Wilson
Reviewed-by: Mika Kuoppala
---
Quoting Tvrtko Ursulin (2018-09-20 15:21:47)
>
> On 19/09/2018 20:55, Chris Wilson wrote:
> > Fix up the error unwind for logical_ring_init() failing by moving the
>
> Could you say in the commit what was broken?
We didn't cleanup all the state we allocated in the caller.
-Chris
From: Ville Syrjälä
Add the code to deal with the MPEG source infoframe.
Blindly typed from the spec, and totally untested.
Cc: Thierry Reding
Cc: Hans Verkuil
Cc: linux-me...@vger.kernel.org
Signed-off-by: Ville Syrjälä
---
drivers/video/hdmi.c | 229
From: Ville Syrjälä
Make life simpler by passing around intel_encoder instead of
drm_encoder.
@r1@
identifier F =~ "infoframe";
identifier I, M;
@@
F(
- struct drm_encoder *I
+ struct intel_encoder *I
, ...)
{
<...
(
- I->M
+ I->base.M
|
- I
+ >base
)
...>
}
@r2@
identifier F =~ "infoframe";
From: Ville Syrjälä
Store the infoframes in the crtc state and precompute them in
.compute_config(). While precomputing we'll also fill out the
inforames.enable bitmask appropriately.
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/intel_ddi.c | 1 +
drivers/gpu/drm/i915/intel_drv.h
From: Ville Syrjälä
Add code to read the infoframes from the video DIP and unpack them into
the crtc state.
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/intel_ddi.c | 17
drivers/gpu/drm/i915/intel_drv.h | 10 ++
drivers/gpu/drm/i915/intel_hdmi.c | 203
From: Ville Syrjälä
As with regular HDMI encoders, let's precompute the infoframes
(actually just AVI infoframe for the time being) with SDVO HDMI
encoders.
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/intel_sdvo.c | 58 +--
1 file changed, 43
From: Ville Syrjälä
Check the infoframes and infoframe enable state when comparing two
crtc states.
We'll use the infoframe logging functions from video/hdmi.c to
show the infoframes as part of the state dump.
TODO: Try to better integrate the infoframe dumps with
drm state dumps
v2:
From: Ville Syrjälä
Read the HDMI infoframes from the hbuf and unpack them into
the crtc state.
Well, actually just AVI infoframe for now but let's write the
infoframe readout code in a more generic fashion in case we
expand this later.
Signed-off-by: Ville Syrjälä
---
From: Ville Syrjälä
We have definitions and low level code for everything except the gamut
metadata HDMI packet. Add the missing bits.
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/i915_reg.h | 4 +++-
drivers/gpu/drm/i915/intel_hdmi.c | 12
2 files changed, 15
From: Ville Syrjälä
Dump out the infoframes in the normal crtc state dump.
TODO: Try to better integrate the infoframe dumps with
drm state dumps
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/intel_display.c | 26 ++
1 file changed, 26 insertions(+)
Quoting Ville Syrjala (2018-09-20 20:10:18)
> From: Ville Syrjälä
>
> Let's try to make sure the fb offset computations never hit
> an integer overflow by making sure the entire fb stays
> below 32bits. framebuffer_check() in the core already does
> the same check, but as it doesn't know about
On 2018-09-13 12:31 PM, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> Disable atomic on a per-device basis instead of for all devices.
> Made possible by the new device.driver_features thing.
>
> Cc: Alex Deucher
> Cc: "Christian König"
> Cc: "David (ChunMing) Zhou"
> Cc: Harry Wentland
>
== Series Details ==
Series: drm/i915: Infoframe precompute/check
URL : https://patchwork.freedesktop.org/series/49983/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
a017fc23cf36 video/hdmi: Constify 'buffer' to the unpack functions
0a03b408d82f video/hdmi: Pass buffer size to
From: Ville Syrjälä
Let's try to make sure the fb offset computations never hit
an integer overflow by making sure the entire fb stays
below 32bits. framebuffer_check() in the core already does
the same check, but as it doesn't know about tiling some things
can slip through. Repeat the check in
== Series Details ==
Series: drm/i915: Make sure fb gtt offsets stay within 32bits
URL : https://patchwork.freedesktop.org/series/49985/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4852 -> Patchwork_10243 =
== Summary - SUCCESS ==
No regressions found.
External
== Series Details ==
Series: drm/i915: Park the GPU on module load (rev2)
URL : https://patchwork.freedesktop.org/series/49693/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4852_full -> Patchwork_10241_full =
== Summary - SUCCESS ==
No regressions found.
== Known
On Thu, Sep 20, 2018 at 09:09:03AM +0100, Chris Wilson wrote:
> Quoting Ville Syrjälä (2018-09-19 17:59:51)
> > On Thu, Sep 13, 2018 at 11:01:40PM +0300, Ville Syrjala wrote:
> > > From: Ville Syrjälä
> > >
> > > With gtt remapping in place we can use arbitraily large framebuffers.
> > > Let's
eDP spec states 2 different bits to enable sink to trigger a
interruption when there is a CRC mismatch.
DP_PSR_CRC_VERIFICATION is for PSR only and
DP_PSR_IRQ_HPD_WITH_CRC_ERRORS is for PSR2 only.
Cc: Dhinakaran Pandiyan
Signed-off-by: José Roberto de Souza
---
drivers/gpu/drm/i915/intel_psr.c
== Series Details ==
Series: drm/i915/selftests: Live tests emit requests and so require rpm
URL : https://patchwork.freedesktop.org/series/49972/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4851_full -> Patchwork_10240_full =
== Summary - WARNING ==
Minor unknown
From: Ville Syrjälä
We want to start tracking which infoframes are enabled, so let's replace
the boolean flag with a bitmask.
We'll abstract the bitmask so that it's not platform dependent. That
will allow us to examine the bitmask later in platform independent code.
Signed-off-by: Ville
From: Ville Syrjälä
Let's make the infoframe pack functions usable with a const infoframe
structure. This allows us to precompute the infoframe earlier, and still
pack it later when we're no longer allowed to modify the structure.
So now we end up with a _check()+_pack_only() or _pack()
From: Ville Syrjälä
We'll be wanting to send more than just infoframes over HDMI. So add an
enum for other packet types.
TODO: Maybe just include the infoframe types in the packet type enum
and get rid of the infoframe type enum?
Cc: Thierry Reding
Cc: Hans Verkuil
Cc:
From: Ville Syrjälä
Store the mask of enabled infoframes in the crtc state. We'll start
with just the readout for HDMI encoder, and we'll expand this
to compute the bitmask in .compute_config() later. SDVO will also
follow later.
Signed-off-by: Ville Syrjälä
---
From: Ville Syrjälä
Add the code to deal with the NTSC VBI infoframe.
I decided against parsing the PES_data_field and just leave
it as an opaque blob, just dumping it out as hex in the log.
Blindly typed from the spec, and totally untested.
Cc: Thierry Reding
Cc: Hans Verkuil
Cc:
From: Ville Syrjälä
Replace the hand rolled memmove() with the real thing.
Signed-off-by: Ville Syrjälä
---
drivers/gpu/drm/i915/intel_hdmi.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_hdmi.c
b/drivers/gpu/drm/i915/intel_hdmi.c
index
From: Ville Syrjälä
The log functions don't modify the passed in infoframe so make it const.
Cc: Thierry Reding
Cc: Hans Verkuil
Cc: linux-me...@vger.kernel.org
Signed-off-by: Ville Syrjälä
---
drivers/video/hdmi.c | 22 +++---
include/linux/hdmi.h | 2 +-
2 files changed,
== Series Details ==
Series: drm/i915: Infoframe precompute/check
URL : https://patchwork.freedesktop.org/series/49983/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Commit: video/hdmi: Constify 'buffer' to the unpack functions
Okay!
Commit: video/hdmi: Pass buffer size to
Both the .enable_signaling and .release of the null syncobj fence
can be replaced by the default callbacks for a small reduction in code
size. In particular the default callback for .release was changed in
commit e28bd101ae1b ("drm: rename null fence to stub fence in syncobj v2")
which neglected
After schedule() returns 0, we must do one last check of COND to
determine the reason for the wakeup with 0 jiffies remaining before
reporting the timeout -- otherwise we may lose the signal due to
scheduler delays.
References: https://bugs.freedesktop.org/show_bug.cgi?id=106690
Signed-off-by:
== Series Details ==
Series: series starting with [1/2] drm: Use default dma_fence hooks where
possible for null syncobj
URL : https://patchwork.freedesktop.org/series/49988/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4853 -> Patchwork_10245 =
== Summary - SUCCESS ==
From: Ville Syrjälä
The unpack functions just read from the passed in buffer,
so make it const.
Cc: Thierry Reding
Cc: Hans Verkuil
Cc: linux-me...@vger.kernel.org
Signed-off-by: Ville Syrjälä
---
drivers/video/hdmi.c | 23 ---
include/linux/hdmi.h | 3 ++-
2 files
From: Ville Syrjälä
Series aimed at precomputing the HDMI infoframes, and we also get
better validation by reading them back out from the hardware and
comparing with the expected data.
Looks like I typed these up about a year ago. Might be time to
get them in before the anniversary ;)
Cc:
From: Ville Syrjälä
To make sure the infoframe unpack functions don't end up examining
stack garbage or oopsing, let's pass in the size of the buffer.
v2: Convert tda1997x.c as well (kbuild test robot)
Cc: Thierry Reding
Cc: Hans Verkuil
Cc: linux-me...@vger.kernel.org
Signed-off-by: Ville
== Series Details ==
Series: drm/i915: Infoframe precompute/check
URL : https://patchwork.freedesktop.org/series/49983/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4852 -> Patchwork_10242 =
== Summary - SUCCESS ==
No regressions found.
External URL:
== Series Details ==
Series: drm/i915/execlists: Onion unwind for logical_ring_init() failure
URL : https://patchwork.freedesktop.org/series/49986/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4853 -> Patchwork_10244 =
== Summary - SUCCESS ==
No regressions found.
== Series Details ==
Series: ICL interrupt handling improvements
URL : https://patchwork.freedesktop.org/series/49971/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4851_full -> Patchwork_10239_full =
== Summary - SUCCESS ==
No regressions found.
== Known issues ==
== Series Details ==
Series: drm/i915: Infoframe precompute/check
URL : https://patchwork.freedesktop.org/series/49983/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4852_full -> Patchwork_10242_full =
== Summary - SUCCESS ==
No regressions found.
== Known issues
On Thu, Sep 20, 2018 at 01:43:25PM -0700, José Roberto de Souza wrote:
> This WA also works fine for PSR2, triggering a selective update when
> possible.
Oh! really?! It didn't work when I chacked on my CNL,
but we probably had other bugs back there...
Thanks for finding this
Reviewed-by:
On 2018-09-19 07:08 PM, Lyude Paul wrote:
> Currently the way that we prevent userspace from performing new modesets
> on MST connectors that have just been destroyed is rather broken.
> There's nothing in the actual DRM DP MST topology helpers that checks
> whether or not a connector still
On 2018-09-18 07:06 PM, Lyude Paul wrote:
> Currently the way that we prevent userspace from performing new modesets
> on MST connectors that have just been destroyed is rather broken.
> There's nothing in the actual DRM DP MST topology helpers that checks
> whether or not a connector still
== Series Details ==
Series: series starting with [1/2] drm: Use default dma_fence hooks where
possible for null syncobj
URL : https://patchwork.freedesktop.org/series/49988/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4853_full -> Patchwork_10245_full =
== Summary -
On Thu, Sep 20, 2018 at 12:27:05PM +0200, Maarten Lankhorst wrote:
> While we may not update new_crtc_state, we may clear active_planes
> if the new cursor update state will disable the cursor, but we fail
> after. If this is immediately followed by a modeset disable, we may
> soon not disable the
On Thu, Sep 20, 2018 at 12:27:06PM +0200, Maarten Lankhorst wrote:
> We need to assume the plane has been visible before, even if no CRTC
> is assigned to the plane. This is because nv12 will enable a a extra
You may want to clarify that this is future, gen11-style NV12, not
current, gen9-style
== Series Details ==
Series: series starting with [1/8] drm/i915/psr: Share PSR and PSR2 exit mask
URL : https://patchwork.freedesktop.org/series/49993/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4854 -> Patchwork_10246 =
== Summary - SUCCESS ==
No regressions found.
== Series Details ==
Series: series starting with [1/8] drm/i915/psr: Share PSR and PSR2 exit mask
URL : https://patchwork.freedesktop.org/series/49993/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4854_full -> Patchwork_10246_full =
== Summary - WARNING ==
Minor
On Thu, Sep 20, 2018 at 12:27:11PM +0200, Maarten Lankhorst wrote:
> We can't make NV12 work any other way. The scaler doesn't handle odd
> coordinates well, and we will get visual corruption on the screen.
>
> Changes since v1:
> - Put the check in intel_plane_check_src_coordinates. (Ville)
>
>
On Thu, Sep 20, 2018 at 12:27:09PM +0200, Maarten Lankhorst wrote:
> On skylake we can switch to a high quality scaler mode when only 1 out
> of 2 scalers are used, but on GLK and later bit 28 has a different
> meaning. Don't set it, and make clear the distinction between
> SKL and later PS
On Thu, Sep 20, 2018 at 12:27:10PM +0200, Maarten Lankhorst wrote:
> This cleans the code up slightly, and will make other changes easier.
>
> Signed-off-by: Maarten Lankhorst
Reviewed-by: Matt Roper
> ---
> drivers/gpu/drm/i915/intel_sprite.c | 90 +
> 1 file
== Series Details ==
Series: series starting with [1/8] drm/i915/psr: Share PSR and PSR2 exit mask
URL : https://patchwork.freedesktop.org/series/49993/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Commit: drm/i915/psr: Share PSR and PSR2 exit mask
Okay!
Commit: drm/i915/psr: Do
On Thu, 2018-09-20 at 15:54 -0700, Rodrigo Vivi wrote:
> On Thu, Sep 20, 2018 at 01:43:25PM -0700, José Roberto de Souza
> wrote:
> > This WA also works fine for PSR2, triggering a selective update
> > when
> > possible.
>
> Oh! really?! It didn't work when I chacked on my CNL,
> but we probably
== Series Details ==
Series: drm/i915: Make sure fb gtt offsets stay within 32bits
URL : https://patchwork.freedesktop.org/series/49985/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4852_full -> Patchwork_10243_full =
== Summary - SUCCESS ==
No regressions found.
On 2018-09-18 07:06 PM, Lyude Paul wrote:
> Hook this into amdgpu's atomic check for their connectors so they never
> get modesets on no-longer-present MST connectors. We'll also expand on
> this later once we add DP MST fallback retraining support.
>
> As well, turns out that the only atomic DRM
== Series Details ==
Series: drm/i915/execlists: Onion unwind for logical_ring_init() failure
URL : https://patchwork.freedesktop.org/series/49986/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4853_full -> Patchwork_10244_full =
== Summary - SUCCESS ==
No regressions
On Wed, Sep 19, 2018 at 8:08 PM, Jani Nikula wrote:
> On Wed, 19 Sep 2018, Chris Chiu wrote:
>> I tried to add a slight delay in the hotplug work as follows
>>
>> --- a/drivers/gpu/drm/i915/intel_hotplug.c
>> +++ b/drivers/gpu/drm/i915/intel_hotplug.c
>> @@ -378,6 +378,8 @@ static void
Quoting Ville Syrjälä (2018-09-19 17:59:51)
> On Thu, Sep 13, 2018 at 11:01:40PM +0300, Ville Syrjala wrote:
> > From: Ville Syrjälä
> >
> > With gtt remapping in place we can use arbitraily large framebuffers.
> > Let's bump the limits as high as we can (32k-1). Going beyond that
> > would
Quoting Zhang, Ning A (2018-09-20 02:34:42)
> I don't understand why CI failure, could you help me to resolve it?
https://patchwork.freedesktop.org/patch/251045/
-Chris
___
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
thank you, by your patch I understand why CI fail for my patch.
BR.
Ning.
在 2018-09-20四的 08:44 +0100，Chris Wilson写道：
Quoting Zhang, Ning A (2018-09-20 02:34:42)
I don't understand why CI failure, could you help me to resolve it?
https://patchwork.freedesktop.org/patch/251045/
-Chris
Quoting Mika Kuoppala (2018-09-20 11:38:08)
> Chris Wilson writes:
> > + end_time = jiffies + i915_selftest.timeout_jiffies;
> > + for_each_prime_number_from(prime, 2, 8192) {
> > + times[1] = ktime_get_raw();
> > +
> > + for (n = 0;
We need to exercise the HW and submission paths for switching contexts
rapidly to check that features such as execlists' wa_tail are adequate.
Plus it's an interesting baseline latency metric.
v2: Check the initial request for allocation errors
v3: Use finite waits for more robust handling of
== Series Details ==
Series: drm/i915: Preparations for adding gen11 planar formats.
URL : https://patchwork.freedesktop.org/series/49956/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4849 -> Patchwork_10236 =
== Summary - SUCCESS ==
No regressions found.
External
== Series Details ==
Series: drm/i915/selftests: Basic stress test for rapid context switching (rev2)
URL : https://patchwork.freedesktop.org/series/48255/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
5a8436e6be6e drm/i915/selftests: Basic stress test for rapid context
== Series Details ==
Series: drm/i915/selftests: Basic stress test for rapid context switching (rev2)
URL : https://patchwork.freedesktop.org/series/48255/
State : success
== Summary ==
= CI Bug Log - changes from CI_DRM_4849 -> Patchwork_10237 =
== Summary - SUCCESS ==
No regressions
Quoting Michał Winiarski (2018-09-20 12:47:07)
> On Wed, Sep 19, 2018 at 09:54:32PM +0100, Chris Wilson wrote:
> > Stolen memory is lost across S4 (hibernate) or S3-RST as it is a portion
> > of ordinary volatile RAM. As we allocate our rings from stolen, this may
> > include the rings used for
Chris Wilson writes:
> We need to exercise the HW and submission paths for switching contexts
> rapidly to check that features such as execlists' wa_tail are adequate.
> Plus it's an interesting baseline latency metric.
>
> v2: Check the initial request for allocation errors
>
> Signed-off-by:
== Series Details ==
Series: drm/i915: Preparations for adding gen11 planar formats.
URL : https://patchwork.freedesktop.org/series/49956/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
a9c37aaf2d75 drm/i915: Clean up casts to crtc_state in
intel_atomic_commit_tail()
== Series Details ==
Series: drm/i915: Preparations for adding gen11 planar formats.
URL : https://patchwork.freedesktop.org/series/49956/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Commit: drm/i915: Clean up casts to crtc_state in intel_atomic_commit_tail()
Okay!
Commit:
== Series Details ==
Series: drm/i915/selftests: Basic stress test for rapid context switching (rev2)
URL : https://patchwork.freedesktop.org/series/48255/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Commit: drm/i915/selftests: Basic stress test for rapid context switching
On Wed, Sep 19, 2018 at 09:54:32PM +0100, Chris Wilson wrote:
> Stolen memory is lost across S4 (hibernate) or S3-RST as it is a portion
> of ordinary volatile RAM. As we allocate our rings from stolen, this may
> include the rings used for our preempt context and their breadcrumb
> instructions.
== Series Details ==
Series: idea for optimize i915 initial time with eDP
URL : https://patchwork.freedesktop.org/series/49952/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
32ea43a57a0b initial panel_power_off_time should be 0
-:38: WARNING:COMMIT_LOG_LONG_LINE: Possible
Op 20-09-18 om 02:12 schreef Matt Roper:
> On Wed, Sep 19, 2018 at 03:56:31PM +0200, Maarten Lankhorst wrote:
>> While we may not update new_crtc_state, we may clear active_planes
>> if the new cursor update state will disable the cursor, but we fail
>> after. If this is immediately followed by a
While we may not update new_crtc_state, we may clear active_planes
if the new cursor update state will disable the cursor, but we fail
after. If this is immediately followed by a modeset disable, we may
soon not disable the planes correctly when we start depending on
active_planes.
Changes since
On skylake we can switch to a high quality scaler mode when only 1 out
of 2 scalers are used, but on GLK and later bit 28 has a different
meaning. Don't set it, and make clear the distinction between
SKL and later PS values.
Signed-off-by: Maarten Lankhorst
---
drivers/gpu/drm/i915/i915_reg.h
Exercise the in-kernel load balancer checking that we can distribute
batches across the set of ctx->engines to avoid load.
Signed-off-by: Chris Wilson
---
tests/Makefile.am | 1 +
tests/Makefile.sources| 1 +
tests/gem_exec_balancer.c | 469 ++
Quoting Matthew Auld (2018-09-20 15:27:06)
> Which means we can now also put it to work in fake_get_huge_pages.
>
> Signed-off-by: Matthew Auld
> Cc: Tvrtko Ursulin
Can we take this into account for selftests/scatterlist.c?
Reviewed-by: Chris Wilson
-Chris
Add static inline helpers to master irq control. Related
comments become superfluous and can be removed.
Signed-off-by: Mika Kuoppala
---
drivers/gpu/drm/i915/i915_irq.c | 17 -
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_irq.c
If we ack the identities immediately after they have been
handled, it should unblock next interrupt accumulation
in the gathering register earlier. It also allows us to
remove time based polling of valid bit. If we don't get a valid
sample now, we will likely get a valid sample on next interrupt,
Don't keep master disabled while we handle the current
interrupts. This should help a little on latency of
generating the next interrupt.
Suggested-by: Chris Wilson
Cc: Chris Wilson
Signed-off-by: Mika Kuoppala
---
drivers/gpu/drm/i915/i915_irq.c | 3 +--
1 file changed, 1 insertion(+), 2
Hi,
This series brings gen11 interrupt handling closer to what
we have with previous gens. Namely the early releasing of
master intr to reduce interrupt latency.
Also cleanups in guc interrupts and identity selector
handling are included.
Code size shrinks a little:
add/remove: 0/0 grow/shrink:
Getting interrupt without any second level indications
is unlikely. So there is no real advantage to bailout early
as all the second level handlers can handle empty master
control status.
Signed-off-by: Mika Kuoppala
---
drivers/gpu/drm/i915/i915_irq.c | 6 +-
1 file changed, 1
Don't keep master disabled while handling display interrupts.
This should help a little with latency of generating the
next interrupt.
Cc: Ville Syrjälä
Cc: Chris Wilson
Signed-off-by: Mika Kuoppala
---
drivers/gpu/drm/i915/i915_irq.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
The returning of iir through function parameter is eyesore.
Make guc irq acking inline and return the iir directly, handling
the empty iir exception early. We can then omit passing the
master control to guc handler as the iir now contains everything
we need.
Cc: Chris Wilson
Cc: Dhinakaran
All other master control register bits, except the enable,
are read only and they are level indications of the second
level interrupt status. Only touch enable bit and rectify
the comment.
Cc: Chris Wilson
Cc: Dhinakaran Pandiyan
Signed-off-by: Mika Kuoppala
---
Move display interrupt handling outside of generic handler.
Cc: Chris Wilson
Cc: Ville Syrjälä
Signed-off-by: Mika Kuoppala
---
drivers/gpu/drm/i915/i915_irq.c | 36 -
1 file changed, 22 insertions(+), 14 deletions(-)
diff --git
Disable master interrupt first before reading. This
guarantees that the sample we act upon is a frozen sample
of level indications and no interrupt was missed between
reading and disabling, possibly then ending up triggering
another interrupt later.
Cc: Chris Wilson
Cc: Dhinakaran Pandiyan
Quoting Matthew Auld (2018-09-20 15:27:07)
> It really wants dev_priv anyway, also now matches i915_gem_init_stolen.
>
> Signed-off-by: Matthew Auld
> Cc: Joonas Lahtinen
> Cc: Tvrtko Ursulin
> Reviewed-by: Tvrtko Ursulin
Reviewed-by: Chris Wilson
As a bonus, start feeding in
Quoting Mika Kuoppala (2018-09-20 15:33:42)
> Disable master interrupt first before reading. This
> guarantees that the sample we act upon is a frozen sample
> of level indications and no interrupt was missed between
> reading and disabling, possibly then ending up triggering
> another interrupt
On Thu, Sep 20, 2018 at 03:27:06PM +0100, Matthew Auld wrote:
> Which means we can now also put it to work in fake_get_huge_pages.
Just another reminder to write self-contained commit msgs. Reading
this wihtout copy pasting the subject line in makes one wonder
what the "which" is here.
>
>
If we return a master control value on disable, it will act
as a posting read and further streamline our interrupt handler.
We can then safely use the inlined helpers on irq reset
and on postinstall. Posting read on postinstall is
superfluous as nothing beneath it is depedent.
Cc: Chris Wilson
Quoting Tvrtko Ursulin (2018-09-20 15:01:38)
>
> On 19/09/2018 20:55, Chris Wilson wrote:
> > As we emit requests or touch HW directly for some of the live tests, the
> > requirement is that we hold the rpm wakeref before doing so. We want a
> > mix of granularity since we will want to test
As we emit requests or touch HW directly for some of the live tests, the
requirement is that we hold the rpm wakeref before doing so. We want a
mix of granularity since we will want to test runtime suspend, so try to
mark up only the critical sections where we need rpm for the live test.
On 20/09/2018 15:27, Matthew Auld wrote:
Which means we can now also put it to work in fake_get_huge_pages.
Signed-off-by: Matthew Auld
Cc: Tvrtko Ursulin
---
drivers/gpu/drm/i915/i915_gem.c | 4 +++-
drivers/gpu/drm/i915/selftests/huge_pages.c | 2 ++
2 files changed, 5
1 - 100 of 151 matches
Mail list logo
