diff mbox series

[v4,4/4] arm64/dts/qcom/sm8250: remove assigned-clock-rate property for mdp clk

Message ID 1646300401-9063-5-git-send-email-quic_vpolimer@quicinc.com
State Superseded
Headers show
Series Update mdp clk to max supported value to support higher refresh rates | expand

Commit Message

Vinod Polimera March 3, 2022, 9:40 a.m. UTC
Kernel clock driver assumes that initial rate is the
max rate for that clock and was not allowing it to scale
beyond the assigned clock value.

Drop the assigned clock rate property and vote on the mdp clock as per
calculated value during the usecase.

Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")
Signed-off-by: Vinod Polimera <quic_vpolimer@quicinc.com>
---
 arch/arm64/boot/dts/qcom/sm8250.dtsi | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

Comments

Dmitry Baryshkov March 3, 2022, 11:50 p.m. UTC | #1
On Thu, 3 Mar 2022 at 12:40, Vinod Polimera <quic_vpolimer@quicinc.com> wrote:
>
> Kernel clock driver assumes that initial rate is the
> max rate for that clock and was not allowing it to scale
> beyond the assigned clock value.
>
> Drop the assigned clock rate property and vote on the mdp clock as per
> calculated value during the usecase.
>
> Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")

Please remove the Fixes tags from all commits. Otherwise the patches
might be picked up into earlier kernels, which do not have a patch
adding a vote on the MDP clock.

> Signed-off-by: Vinod Polimera <quic_vpolimer@quicinc.com>
> ---
>  arch/arm64/boot/dts/qcom/sm8250.dtsi | 9 ++-------
>  1 file changed, 2 insertions(+), 7 deletions(-)
>
> diff --git a/arch/arm64/boot/dts/qcom/sm8250.dtsi b/arch/arm64/boot/dts/qcom/sm8250.dtsi
> index fdaf303..2105eb7 100644
> --- a/arch/arm64/boot/dts/qcom/sm8250.dtsi
> +++ b/arch/arm64/boot/dts/qcom/sm8250.dtsi
> @@ -3164,9 +3164,6 @@
>                                  <&dispcc DISP_CC_MDSS_MDP_CLK>;
>                         clock-names = "iface", "bus", "nrt_bus", "core";
>
> -                       assigned-clocks = <&dispcc DISP_CC_MDSS_MDP_CLK>;
> -                       assigned-clock-rates = <460000000>;
> -
>                         interrupts = <GIC_SPI 83 IRQ_TYPE_LEVEL_HIGH>;
>                         interrupt-controller;
>                         #interrupt-cells = <1>;
> @@ -3191,10 +3188,8 @@
>                                          <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
>                                 clock-names = "iface", "bus", "core", "vsync";
>
> -                               assigned-clocks = <&dispcc DISP_CC_MDSS_MDP_CLK>,
> -                                                 <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
> -                               assigned-clock-rates = <460000000>,
> -                                                      <19200000>;
> +                               assigned-clocks = <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
> +                               assigned-clock-rates = <19200000>;
>
>                                 operating-points-v2 = <&mdp_opp_table>;
>                                 power-domains = <&rpmhpd SM8250_MMCX>;
> --
> 2.7.4
>
Stephen Boyd March 3, 2022, 11:56 p.m. UTC | #2
Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> On Thu, 3 Mar 2022 at 12:40, Vinod Polimera <quic_vpolimer@quicinc.com> wrote:
> >
> > Kernel clock driver assumes that initial rate is the
> > max rate for that clock and was not allowing it to scale
> > beyond the assigned clock value.
> >
> > Drop the assigned clock rate property and vote on the mdp clock as per
> > calculated value during the usecase.
> >
> > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")
>
> Please remove the Fixes tags from all commits. Otherwise the patches
> might be picked up into earlier kernels, which do not have a patch
> adding a vote on the MDP clock.

What patch is that? The Fixes tag could point to that commit.
Dmitry Baryshkov March 4, 2022, 12:15 a.m. UTC | #3
On Fri, 4 Mar 2022 at 02:56, Stephen Boyd <swboyd@chromium.org> wrote:
>
> Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> > On Thu, 3 Mar 2022 at 12:40, Vinod Polimera <quic_vpolimer@quicinc.com> wrote:
> > >
> > > Kernel clock driver assumes that initial rate is the
> > > max rate for that clock and was not allowing it to scale
> > > beyond the assigned clock value.
> > >
> > > Drop the assigned clock rate property and vote on the mdp clock as per
> > > calculated value during the usecase.
> > >
> > > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")
> >
> > Please remove the Fixes tags from all commits. Otherwise the patches
> > might be picked up into earlier kernels, which do not have a patch
> > adding a vote on the MDP clock.
>
> What patch is that? The Fixes tag could point to that commit.

Please correct me if I'm wrong.
Currently the dtsi enforces bumping the MDP clock when the mdss device
is being probed and when the dpu device is being probed.
Later during the DPU lifetime the core_perf would change the clock's
rate as it sees fit according to the CRTC requirements.

However it would happen only when the during the
dpu_crtc_atomic_flush(), before we call this function, the MDP clock
is left in the undetermined state. The power rails controlled by the
opp table are left in the undetermined state.

I suppose that during the dpu_bind we should bump the clock to the max
possible freq and let dpu_core_perf handle it afterwards.


--
With best wishes
Dmitry
Doug Anderson March 4, 2022, 9:49 p.m. UTC | #4
Hi,

On Thu, Mar 3, 2022 at 4:16 PM Dmitry Baryshkov
<dmitry.baryshkov@linaro.org> wrote:
>
> On Fri, 4 Mar 2022 at 02:56, Stephen Boyd <swboyd@chromium.org> wrote:
> >
> > Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> > > On Thu, 3 Mar 2022 at 12:40, Vinod Polimera <quic_vpolimer@quicinc.com> wrote:
> > > >
> > > > Kernel clock driver assumes that initial rate is the
> > > > max rate for that clock and was not allowing it to scale
> > > > beyond the assigned clock value.
> > > >
> > > > Drop the assigned clock rate property and vote on the mdp clock as per
> > > > calculated value during the usecase.
> > > >
> > > > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")
> > >
> > > Please remove the Fixes tags from all commits. Otherwise the patches
> > > might be picked up into earlier kernels, which do not have a patch
> > > adding a vote on the MDP clock.
> >
> > What patch is that? The Fixes tag could point to that commit.
>
> Please correct me if I'm wrong.
> Currently the dtsi enforces bumping the MDP clock when the mdss device
> is being probed and when the dpu device is being probed.
> Later during the DPU lifetime the core_perf would change the clock's
> rate as it sees fit according to the CRTC requirements.

"Currently" means _before_ ${SUBJECT} patch lands, right? Since
${SUBJECT} patch is removing the bump to max.


> However it would happen only when the during the
> dpu_crtc_atomic_flush(), before we call this function, the MDP clock
> is left in the undetermined state. The power rails controlled by the
> opp table are left in the undetermined state.
>
> I suppose that during the dpu_bind we should bump the clock to the max
> possible freq and let dpu_core_perf handle it afterwards.

Definitely feels like seeing the clock to something predictable during
the initial probe makes sense. If it's just for the initial probe then
setting it to max (based on the opp table) seems fine. I think an
earlier version of this series set it to max every time we did runtime
resume. We'd have to have a good reason to do that.

-Doug
Dmitry Baryshkov March 4, 2022, 10:11 p.m. UTC | #5
On Sat, 5 Mar 2022 at 00:49, Doug Anderson <dianders@chromium.org> wrote:
> On Thu, Mar 3, 2022 at 4:16 PM Dmitry Baryshkov
> <dmitry.baryshkov@linaro.org> wrote:
> >
> > On Fri, 4 Mar 2022 at 02:56, Stephen Boyd <swboyd@chromium.org> wrote:
> > >
> > > Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> > > > On Thu, 3 Mar 2022 at 12:40, Vinod Polimera <quic_vpolimer@quicinc.com> wrote:
> > > > >
> > > > > Kernel clock driver assumes that initial rate is the
> > > > > max rate for that clock and was not allowing it to scale
> > > > > beyond the assigned clock value.
> > > > >
> > > > > Drop the assigned clock rate property and vote on the mdp clock as per
> > > > > calculated value during the usecase.
> > > > >
> > > > > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display system nodes")
> > > >
> > > > Please remove the Fixes tags from all commits. Otherwise the patches
> > > > might be picked up into earlier kernels, which do not have a patch
> > > > adding a vote on the MDP clock.
> > >
> > > What patch is that? The Fixes tag could point to that commit.
> >
> > Please correct me if I'm wrong.
> > Currently the dtsi enforces bumping the MDP clock when the mdss device
> > is being probed and when the dpu device is being probed.
> > Later during the DPU lifetime the core_perf would change the clock's
> > rate as it sees fit according to the CRTC requirements.
>
> "Currently" means _before_ ${SUBJECT} patch lands, right? Since
> ${SUBJECT} patch is removing the bump to max.

Yes. 'Before this patch'.

>
>
> > However it would happen only when the during the
> > dpu_crtc_atomic_flush(), before we call this function, the MDP clock
> > is left in the undetermined state. The power rails controlled by the
> > opp table are left in the undetermined state.
> >
> > I suppose that during the dpu_bind we should bump the clock to the max
> > possible freq and let dpu_core_perf handle it afterwards.
>
> Definitely feels like seeing the clock to something predictable during
> the initial probe makes sense. If it's just for the initial probe then
> setting it to max (based on the opp table) seems fine.

Vinod, could you please implement it?

> I think an
> earlier version of this series set it to max every time we did runtime
> resume. We'd have to have a good reason to do that.

Yes, this is correct. Based on the comments I had the impression that
there was a suggestion that the place for the calls was wrong. Most
probably I was instead projecting my own thoughts.
Vinod Polimera March 7, 2022, 4:05 p.m. UTC | #6
> WARNING: This email originated from outside of Qualcomm. Please be wary
> of any links or attachments, and do not enable macros.
> 
> On Sat, 5 Mar 2022 at 00:49, Doug Anderson <dianders@chromium.org>
> wrote:
> > On Thu, Mar 3, 2022 at 4:16 PM Dmitry Baryshkov
> > <dmitry.baryshkov@linaro.org> wrote:
> > >
> > > On Fri, 4 Mar 2022 at 02:56, Stephen Boyd <swboyd@chromium.org>
> wrote:
> > > >
> > > > Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> > > > > On Thu, 3 Mar 2022 at 12:40, Vinod Polimera
> <quic_vpolimer@quicinc.com> wrote:
> > > > > >
> > > > > > Kernel clock driver assumes that initial rate is the
> > > > > > max rate for that clock and was not allowing it to scale
> > > > > > beyond the assigned clock value.
> > > > > >
> > > > > > Drop the assigned clock rate property and vote on the mdp clock as
> per
> > > > > > calculated value during the usecase.
> > > > > >
> > > > > > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display
> system nodes")
> > > > >
> > > > > Please remove the Fixes tags from all commits. Otherwise the
> patches
> > > > > might be picked up into earlier kernels, which do not have a patch
> > > > > adding a vote on the MDP clock.
> > > >
> > > > What patch is that? The Fixes tag could point to that commit.
> > >
> > > Please correct me if I'm wrong.
> > > Currently the dtsi enforces bumping the MDP clock when the mdss
> device
> > > is being probed and when the dpu device is being probed.
> > > Later during the DPU lifetime the core_perf would change the clock's
> > > rate as it sees fit according to the CRTC requirements.
> >
> > "Currently" means _before_ ${SUBJECT} patch lands, right? Since
> > ${SUBJECT} patch is removing the bump to max.
> 
> Yes. 'Before this patch'.
> 
> >
> >
> > > However it would happen only when the during the
> > > dpu_crtc_atomic_flush(), before we call this function, the MDP clock
> > > is left in the undetermined state. The power rails controlled by the
> > > opp table are left in the undetermined state.
> > >
> > > I suppose that during the dpu_bind we should bump the clock to the max
> > > possible freq and let dpu_core_perf handle it afterwards.
> >
> > Definitely feels like seeing the clock to something predictable during
> > the initial probe makes sense. If it's just for the initial probe then
> > setting it to max (based on the opp table) seems fine.
> 
> Vinod, could you please implement it?
> 
> > I think an
> > earlier version of this series set it to max every time we did runtime
> > resume. We'd have to have a good reason to do that.
> 
> Yes, this is correct. Based on the comments I had the impression that
> there was a suggestion that the place for the calls was wrong. Most
> probably I was instead projecting my own thoughts.
> 
I had discussed internally with the team. Traditionally, mdp clk vote during
probe/bind is required when display is turned on in bootloader and persists
till first update in kernel. As in chromebook, timing engine will be turned 
off during depthcharge exit and as there is no display transition from 
bootloader to kernel, mdp clk can be voted based on the calculated value 
during framework update and does not required vote during probe/bind.

Thanks,
Vinod.
> --
> With best wishes
> Dmitry
Dmitry Baryshkov March 7, 2022, 4:22 p.m. UTC | #7
On Mon, 7 Mar 2022 at 19:05, Vinod Polimera <vpolimer@qti.qualcomm.com> wrote:
>
> > WARNING: This email originated from outside of Qualcomm. Please be wary
> > of any links or attachments, and do not enable macros.
> >
> > On Sat, 5 Mar 2022 at 00:49, Doug Anderson <dianders@chromium.org>
> > wrote:
> > > On Thu, Mar 3, 2022 at 4:16 PM Dmitry Baryshkov
> > > <dmitry.baryshkov@linaro.org> wrote:
> > > >
> > > > On Fri, 4 Mar 2022 at 02:56, Stephen Boyd <swboyd@chromium.org>
> > wrote:
> > > > >
> > > > > Quoting Dmitry Baryshkov (2022-03-03 15:50:50)
> > > > > > On Thu, 3 Mar 2022 at 12:40, Vinod Polimera
> > <quic_vpolimer@quicinc.com> wrote:
> > > > > > >
> > > > > > > Kernel clock driver assumes that initial rate is the
> > > > > > > max rate for that clock and was not allowing it to scale
> > > > > > > beyond the assigned clock value.
> > > > > > >
> > > > > > > Drop the assigned clock rate property and vote on the mdp clock as
> > per
> > > > > > > calculated value during the usecase.
> > > > > > >
> > > > > > > Fixes: 7c1dffd471("arm64: dts: qcom: sm8250.dtsi: add display
> > system nodes")
> > > > > >
> > > > > > Please remove the Fixes tags from all commits. Otherwise the
> > patches
> > > > > > might be picked up into earlier kernels, which do not have a patch
> > > > > > adding a vote on the MDP clock.
> > > > >
> > > > > What patch is that? The Fixes tag could point to that commit.
> > > >
> > > > Please correct me if I'm wrong.
> > > > Currently the dtsi enforces bumping the MDP clock when the mdss
> > device
> > > > is being probed and when the dpu device is being probed.
> > > > Later during the DPU lifetime the core_perf would change the clock's
> > > > rate as it sees fit according to the CRTC requirements.
> > >
> > > "Currently" means _before_ ${SUBJECT} patch lands, right? Since
> > > ${SUBJECT} patch is removing the bump to max.
> >
> > Yes. 'Before this patch'.
> >
> > >
> > >
> > > > However it would happen only when the during the
> > > > dpu_crtc_atomic_flush(), before we call this function, the MDP clock
> > > > is left in the undetermined state. The power rails controlled by the
> > > > opp table are left in the undetermined state.
> > > >
> > > > I suppose that during the dpu_bind we should bump the clock to the max
> > > > possible freq and let dpu_core_perf handle it afterwards.
> > >
> > > Definitely feels like seeing the clock to something predictable during
> > > the initial probe makes sense. If it's just for the initial probe then
> > > setting it to max (based on the opp table) seems fine.
> >
> > Vinod, could you please implement it?
> >
> > > I think an
> > > earlier version of this series set it to max every time we did runtime
> > > resume. We'd have to have a good reason to do that.
> >
> > Yes, this is correct. Based on the comments I had the impression that
> > there was a suggestion that the place for the calls was wrong. Most
> > probably I was instead projecting my own thoughts.
> >
> I had discussed internally with the team. Traditionally, mdp clk vote during
> probe/bind is required when display is turned on in bootloader and persists
> till first update in kernel.

Not each and every board has a display setup in the bootloader. For
example the RB5 I have here doesn't support setting up the display.
Not to mention that we should tell Linux, which vote is cast,
otherwise the .sync_state can turn respective votes off.

> As in chromebook, timing engine will be turned
> off during depthcharge exit and as there is no display transition from
> bootloader to kernel, mdp clk can be voted based on the calculated value
> during framework update and does not required vote during probe/bind.

Generally Linux should not depend on the bootloader setup. You can not
be sure. What if we kexec next kernel?
diff mbox series

Patch

diff --git a/arch/arm64/boot/dts/qcom/sm8250.dtsi b/arch/arm64/boot/dts/qcom/sm8250.dtsi
index fdaf303..2105eb7 100644
--- a/arch/arm64/boot/dts/qcom/sm8250.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8250.dtsi
@@ -3164,9 +3164,6 @@ 
 				 <&dispcc DISP_CC_MDSS_MDP_CLK>;
 			clock-names = "iface", "bus", "nrt_bus", "core";
 
-			assigned-clocks = <&dispcc DISP_CC_MDSS_MDP_CLK>;
-			assigned-clock-rates = <460000000>;
-
 			interrupts = <GIC_SPI 83 IRQ_TYPE_LEVEL_HIGH>;
 			interrupt-controller;
 			#interrupt-cells = <1>;
@@ -3191,10 +3188,8 @@ 
 					 <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
 				clock-names = "iface", "bus", "core", "vsync";
 
-				assigned-clocks = <&dispcc DISP_CC_MDSS_MDP_CLK>,
-						  <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
-				assigned-clock-rates = <460000000>,
-						       <19200000>;
+				assigned-clocks = <&dispcc DISP_CC_MDSS_VSYNC_CLK>;
+				assigned-clock-rates = <19200000>;
 
 				operating-points-v2 = <&mdp_opp_table>;
 				power-domains = <&rpmhpd SM8250_MMCX>;