diff mbox series

[v2,1/2] dt-bindings: arm,scmi: Do not use clocks for SCMI performance domains

Message ID 20201021183104.27949-1-sudeep.holla@arm.com
State New
Headers show
Series [v2,1/2] dt-bindings: arm,scmi: Do not use clocks for SCMI performance domains | expand

Commit Message

Sudeep Holla Oct. 21, 2020, 6:31 p.m. UTC 
Commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return
-EPROBE_DEFER") handles -EPROBE_DEFER for the clock/interconnects within
_allocate_opp_table() which is called from dev_pm_opp_add and it
now propagates the error back to the caller.

SCMI performance domain re-used clock bindings to keep it simple. However
with the above mentioned change, if clock property is present in a device
node, opps can't be added until clk_get succeeds. So in order to fix the
issue, we can register dummy clocks which is completely ugly.

Since there are no upstream users for the SCMI performance domain clock
bindings, let us introduce separate performance domain bindings for the
same.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

---
 .../devicetree/bindings/arm/arm,scmi.txt      | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

v1[1]->v2:
	- Changed the generic #perf-domain-cells to more SCMI specific
	  property #arm,scmi-perf-domain-cells

Hi Rob/Viresh,

This is actually a fix for the regression I reported here[2].
I am not adding fixes tag as I am targeting in the same release and
also because it is not directly related.

Regards,
Sudeep

[1] https://lore.kernel.org/r/20201020203710.10100-1-sudeep.holla@arm.com
[2] https://lore.kernel.org/r/20201015180555.gacdzkofpibkdn2e@bogus

P.S.:/me records that this binding needs to be moved to yaml in v5.11
--
2.17.1

Comments

Sudeep Holla Oct. 22, 2020, 6:22 p.m. UTC | #1 
On Wed, Oct 21, 2020 at 07:31:03PM +0100, Sudeep Holla wrote:
> Commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return

> -EPROBE_DEFER") handles -EPROBE_DEFER for the clock/interconnects within

> _allocate_opp_table() which is called from dev_pm_opp_add and it

> now propagates the error back to the caller.

> 

> SCMI performance domain re-used clock bindings to keep it simple. However

> with the above mentioned change, if clock property is present in a device

> node, opps can't be added until clk_get succeeds. So in order to fix the

> issue, we can register dummy clocks which is completely ugly.

> 

> Since there are no upstream users for the SCMI performance domain clock

> bindings, let us introduce separate performance domain bindings for the

> same.

> 

> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> ---

>  .../devicetree/bindings/arm/arm,scmi.txt      | 19 ++++++++++++++++---

>  1 file changed, 16 insertions(+), 3 deletions(-)

> 

> v1[1]->v2:

> 	- Changed the generic #perf-domain-cells to more SCMI specific

> 	  property #arm,scmi-perf-domain-cells

>


Is more specific #arm,scmi-perf-domain-cells acceptable ?
Sorry for the rush, but this fixes SCMI cpufreq which is broken after
commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return
-EPROBE_DEFER")

-- 
Regards,
Sudeep
Rob Herring Oct. 23, 2020, 1:24 p.m. UTC | #2 
On Thu, Oct 22, 2020 at 1:22 PM Sudeep Holla <sudeep.holla@arm.com> wrote:
>

> On Wed, Oct 21, 2020 at 07:31:03PM +0100, Sudeep Holla wrote:

> > Commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return

> > -EPROBE_DEFER") handles -EPROBE_DEFER for the clock/interconnects within

> > _allocate_opp_table() which is called from dev_pm_opp_add and it

> > now propagates the error back to the caller.

> >

> > SCMI performance domain re-used clock bindings to keep it simple. However

> > with the above mentioned change, if clock property is present in a device

> > node, opps can't be added until clk_get succeeds. So in order to fix the

> > issue, we can register dummy clocks which is completely ugly.

> >

> > Since there are no upstream users for the SCMI performance domain clock

> > bindings, let us introduce separate performance domain bindings for the

> > same.

> >

> > Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> > ---

> >  .../devicetree/bindings/arm/arm,scmi.txt      | 19 ++++++++++++++++---

> >  1 file changed, 16 insertions(+), 3 deletions(-)

> >

> > v1[1]->v2:

> >       - Changed the generic #perf-domain-cells to more SCMI specific

> >         property #arm,scmi-perf-domain-cells

> >

>

> Is more specific #arm,scmi-perf-domain-cells acceptable ?

> Sorry for the rush, but this fixes SCMI cpufreq which is broken after

> commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return

> -EPROBE_DEFER")


If you are in a rush, you'd better go the dummy clock route. We should
get this binding right and I think that means something common, not
SCMI specific.

Rob
Sudeep Holla Oct. 23, 2020, 1:59 p.m. UTC | #3 
On Fri, Oct 23, 2020 at 08:24:06AM -0500, Rob Herring wrote:
> On Thu, Oct 22, 2020 at 1:22 PM Sudeep Holla <sudeep.holla@arm.com> wrote:

> >

> > On Wed, Oct 21, 2020 at 07:31:03PM +0100, Sudeep Holla wrote:

> > > Commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return

> > > -EPROBE_DEFER") handles -EPROBE_DEFER for the clock/interconnects within

> > > _allocate_opp_table() which is called from dev_pm_opp_add and it

> > > now propagates the error back to the caller.

> > >

> > > SCMI performance domain re-used clock bindings to keep it simple. However

> > > with the above mentioned change, if clock property is present in a device

> > > node, opps can't be added until clk_get succeeds. So in order to fix the

> > > issue, we can register dummy clocks which is completely ugly.

> > >

> > > Since there are no upstream users for the SCMI performance domain clock

> > > bindings, let us introduce separate performance domain bindings for the

> > > same.

> > >

> > > Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>

> > > ---

> > >  .../devicetree/bindings/arm/arm,scmi.txt      | 19 ++++++++++++++++---

> > >  1 file changed, 16 insertions(+), 3 deletions(-)

> > >

> > > v1[1]->v2:

> > >       - Changed the generic #perf-domain-cells to more SCMI specific

> > >         property #arm,scmi-perf-domain-cells

> > >

> >

> > Is more specific #arm,scmi-perf-domain-cells acceptable ?

> > Sorry for the rush, but this fixes SCMI cpufreq which is broken after

> > commit dd461cd9183f ("opp: Allow dev_pm_opp_get_opp_table() to return

> > -EPROBE_DEFER")

> 

> If you are in a rush, you'd better go the dummy clock route. We should

> get this binding right and I think that means something common, not

> SCMI specific.

> 


Ah OK, I assumed you wanted to make it SCMI specific. It makes sense to
have something generic like clocks for OPP domains.

There was discussion on the other thread to use empty OPP list for
firmware discoverable OPP list. Random thought here is to add domain ID
to OPP binding and use that ?

--
Regards,
Sudeep
diff mbox series

Patch

diff --git a/Documentation/devicetree/bindings/arm/arm,scmi.txt b/Documentation/devicetree/bindings/arm/arm,scmi.txt
index 55deb68230eb..7af1be54f6c7 100644
--- a/Documentation/devicetree/bindings/arm/arm,scmi.txt
+++ b/Documentation/devicetree/bindings/arm/arm,scmi.txt
@@ -44,7 +44,7 @@  as described in the following sections. If the platform supports dedicated
 mboxes, mbox-names and shmem shall be present in the sub-node corresponding
 to that protocol.

-Clock/Performance bindings for the clocks/OPPs based on SCMI Message Protocol
+Clock bindings for the clocks based on SCMI Message Protocol
 ------------------------------------------------------------

 This binding uses the common clock binding[1].
@@ -52,6 +52,19 @@  This binding uses the common clock binding[1].
 Required properties:
 - #clock-cells : Should be 1. Contains the Clock ID value used by SCMI commands.

+Performance bindings for the OPPs based on SCMI Message Protocol
+------------------------------------------------------------
+
+Required properties:
+- #arm,scmi-perf-domain-cells: Should be 1. Contains the performance domain ID
+  value used by SCMI commands.
+
+* Property arm,scmi-perf-domain
+
+Devices supporting SCMI performance domain must set their "arm,scmi-perf-domain"
+property with phandle to a SCMI performance domain controller followed by the
+performance domain.
+
 Power domain bindings for the power domains based on SCMI Message Protocol
 ------------------------------------------------------------

@@ -152,7 +165,7 @@  firmware {

 		scmi_dvfs: protocol@13 {
 			reg = <0x13>;
-			#clock-cells = <1>;
+			#arm,scmi-perf-domain-cells = <1>;
 		};

 		scmi_clk: protocol@14 {
@@ -175,7 +188,7 @@  firmware {
 cpu@0 {
 	...
 	reg = <0 0>;
-	clocks = <&scmi_dvfs 0>;
+	arm,scmi-perf-domain = <&scmi_dvfs 0>;
 };

 hdlcd@7ff60000 {