mbox series

[v3,0/4] dts: qcom: arm64: sdm845: SLPI DSP enablement

Message ID 20230330165322.118279-1-me@dylanvanassche.be
Headers show
Series dts: qcom: arm64: sdm845: SLPI DSP enablement | expand

Message

Dylan Van Assche March 30, 2023, 4:53 p.m. UTC
* About *

The Qualcomm SDM845 SoC has a separate SLPI (Sensor Low Power Island)
DSP for sensors connected to the SoC which is responsible for exposing
sensors to userspace, power saving, and other features. 
While sensors are connected to GPIOs of the SoC, they cannot be used
because the hypervisor blocks direct access to the sensors, thus the 
DSP must be used to access any sensor on this SoC. The SLPI DSP uses a
GLink edge (dsps) to communicate with the host and has a FastRPC interface
to load files from the host filesystem such as sensor configuration files.
The FastRPC interface does not use regular FastRPC Compute Banks
but instead uses an allocated CMA region through which communication happens.

* Changes *

This patchseries completes support for the SLPI in the Qualcomm SDM845 SoC
by adding the SLPI to the SDM845 DTS and enable it for 2 hardware devices:
- Oneplus 6
- SHIFTPHONES SHIFT6mq

* Related patches *

1. Remoteproc changes (v2) to support the SLPI DSP in SDM845:
https://lore.kernel.org/linux-remoteproc/20230327183736.496170-1-me@dylanvanassche.be/
2. FastRPC changes (v2) to support the SLPI DSP in SDM845:
https://lore.kernel.org/linux-arm-msm/20230327184204.498032-1-me@dylanvanassche.be/

This serie does not depend on any serie, but all of them are necessary
to enable the feature in the end.

* Changelog *

Changes in v2:

- Removed double blank lines
- Enforce alphabetically order for 'slpi_pas'
- Reordered 'slpi_pas' properties
- Fixed FastRPC syntax
- Dropped qcom,assign-all-mem property as this is not necessary anymore

Changes in v3:

- Dropped FastRPC dt-bindings change, unnecessary to relax the bindings
- Add reg property to compute-cb for the SLPI on SDM845

Kind regards,
Dylan Van Assche

Dylan Van Assche (4):
  dts: arm64: qcom: sdm845: add SLPI remoteproc
  dts: arm64: qcom: sdm845: add SLPI FastRPC support
  dts: arm64: qcom: sdm845-oneplus: enable SLPI
  dts: arm64: qcom: sdm845-shift-axolotl: enable SLPI

 .../boot/dts/qcom/sdm845-oneplus-common.dtsi  |  5 ++
 .../boot/dts/qcom/sdm845-shift-axolotl.dts    |  5 ++
 arch/arm64/boot/dts/qcom/sdm845.dtsi          | 60 +++++++++++++++++++
 3 files changed, 70 insertions(+)

Comments

Dylan Van Assche March 31, 2023, 9:36 a.m. UTC | #1
Hi Konrad,

On Fri, 2023-03-31 at 04:03 +0200, Konrad Dybcio wrote:
> 
> 
> On 30.03.2023 18:53, Dylan Van Assche wrote:
> > Qualcomm SDM845 SoC features a SLPI DSP which uses FastRPC through
> > an allocated memory region to load files from the host filesystem
> > such as sensor configuration files.
> > 
> > Add a FastRPC node at /dev/fastrpc-sdsp and a DMA region, similar
> > to
> > downstream, to allow userspace to communicate with the SLPI via the
> > FastRPC interface for initializing the sensors on the SLPI.
> > 
> > Signed-off-by: Dylan Van Assche <me@dylanvanassche.be>
> > ---
> >  arch/arm64/boot/dts/qcom/sdm845.dtsi | 24 ++++++++++++++++++++++++
> >  1 file changed, 24 insertions(+)
> > 
> > diff --git a/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > index 3b547cb7aeb8..8ea4944f3ad6 100644
> > --- a/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > +++ b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > @@ -878,6 +878,14 @@ mdata_mem: mpss-metadata {
> >                         size = <0 0x4000>;
> >                         no-map;
> >                 };
> > +
> > +               fastrpc_mem: fastrpc {
> > +                       compatible = "shared-dma-pool";
> > +                       reusable;
> Please move it last to get a nice reverse-Christmas-tree layout.
> 

Will fix in v4.

> > +                       alloc-ranges = <0 0x00000000 0 0xffffffff>;
> Would there be any issues with it starting over (1<<31 - 1)?
> 

You mean a bigger region then, like the whole CMA region then? AFAIK,
the SLPI always use the same region expecting it to be in this range.
However, I cannot confirm more, as I have no insights in the firmware
running on there, this all comes from finding out what it exactly does
on downstream.

> > +                       alignment = <0 0x400000>;
> Please use 0x0 for the 0 here, as it's essentially reg.size with
> size-cells = 2

Will fix in v4.

> 
> > +                       size = <0 0x1000000>;
> Ditto

Will fix in v4.

> 
> > +               };
> >         };
> >  
> >         adsp_pas: remoteproc-adsp {
> > @@ -3344,6 +3352,22 @@ glink-edge {
> >                                 label = "dsps";
> >                                 qcom,remote-pid = <3>;
> >                                 mboxes = <&apss_shared 24>;
> > +
> > +                               fastrpc {
> > +                                       compatible =
> > "qcom,fastrpc";
> > +                                       qcom,glink-channels =
> > "fastrpcglink-apps-dsp";
> > +                                       label = "sdsp";
> > +                                       qcom,non-secure-domain;
> > +                                       qcom,vmids = <0x3 0xF 0x5
> > 0x6>;
> Please use the recently-introduced header and depend on (and
> make a patch atop)
> 
> https://lore.kernel.org/linux-arm-msm/8685b710-b74d-556a-815d-0ffef2b0eeff@linaro.org/T/#t
> 
> Konrad
> 
> > +                                       memory-region =
> > <&fastrpc_mem>;
> > +                                       #address-cells = <1>;
> > +                                       #size-cells = <0>;
> > +
> > +                                       compute-cb@0 {
> > +                                               compatible =
> > "qcom,fastrpc-compute-cb";
> > +                                               reg = <0>;
> > +                                       };
> > +                               };
> >                         };
> >                 };
> >  

Kind regards,
Dylan
Konrad Dybcio March 31, 2023, 12:20 p.m. UTC | #2
On 31.03.2023 11:36, Dylan Van Assche wrote:
> Hi Konrad,
> 
> On Fri, 2023-03-31 at 04:03 +0200, Konrad Dybcio wrote:
>>
>>
>> On 30.03.2023 18:53, Dylan Van Assche wrote:
>>> Qualcomm SDM845 SoC features a SLPI DSP which uses FastRPC through
>>> an allocated memory region to load files from the host filesystem
>>> such as sensor configuration files.
>>>
>>> Add a FastRPC node at /dev/fastrpc-sdsp and a DMA region, similar
>>> to
>>> downstream, to allow userspace to communicate with the SLPI via the
>>> FastRPC interface for initializing the sensors on the SLPI.
>>>
>>> Signed-off-by: Dylan Van Assche <me@dylanvanassche.be>
>>> ---
>>>  arch/arm64/boot/dts/qcom/sdm845.dtsi | 24 ++++++++++++++++++++++++
>>>  1 file changed, 24 insertions(+)
>>>
>>> diff --git a/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>> b/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>> index 3b547cb7aeb8..8ea4944f3ad6 100644
>>> --- a/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>> +++ b/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>> @@ -878,6 +878,14 @@ mdata_mem: mpss-metadata {
>>>                         size = <0 0x4000>;
>>>                         no-map;
>>>                 };
>>> +
>>> +               fastrpc_mem: fastrpc {
>>> +                       compatible = "shared-dma-pool";
>>> +                       reusable;
>> Please move it last to get a nice reverse-Christmas-tree layout.
>>
> 
> Will fix in v4.
> 
>>> +                       alloc-ranges = <0 0x00000000 0 0xffffffff>;
>> Would there be any issues with it starting over (1<<31 - 1)?
>>
> 
> You mean a bigger region then, like the whole CMA region then? AFAIK,
> the SLPI always use the same region expecting it to be in this range.
> However, I cannot confirm more, as I have no insights in the firmware
> running on there, this all comes from finding out what it exactly does
> on downstream.
I was asking about the <.. 0 0xfff.f> part specifically, as that means
it can't be allocated above 4 gigs. But I guess it's just how qcom
envisioned it.

Also, please use 0x0 in alloc-ranges as well, this is all addresses/
reg sizes.

Konrad
> 
>>> +                       alignment = <0 0x400000>;
>> Please use 0x0 for the 0 here, as it's essentially reg.size with
>> size-cells = 2
> 
> Will fix in v4.
> 
>>
>>> +                       size = <0 0x1000000>;
>> Ditto
> 
> Will fix in v4.
> 
>>
>>> +               };
>>>         };
>>>  
>>>         adsp_pas: remoteproc-adsp {
>>> @@ -3344,6 +3352,22 @@ glink-edge {
>>>                                 label = "dsps";
>>>                                 qcom,remote-pid = <3>;
>>>                                 mboxes = <&apss_shared 24>;
>>> +
>>> +                               fastrpc {
>>> +                                       compatible =
>>> "qcom,fastrpc";
>>> +                                       qcom,glink-channels =
>>> "fastrpcglink-apps-dsp";
>>> +                                       label = "sdsp";
>>> +                                       qcom,non-secure-domain;
>>> +                                       qcom,vmids = <0x3 0xF 0x5
>>> 0x6>;
>> Please use the recently-introduced header and depend on (and
>> make a patch atop)
>>
>> https://lore.kernel.org/linux-arm-msm/8685b710-b74d-556a-815d-0ffef2b0eeff@linaro.org/T/#t
>>
>> Konrad
>>
>>> +                                       memory-region =
>>> <&fastrpc_mem>;
>>> +                                       #address-cells = <1>;
>>> +                                       #size-cells = <0>;
>>> +
>>> +                                       compute-cb@0 {
>>> +                                               compatible =
>>> "qcom,fastrpc-compute-cb";
>>> +                                               reg = <0>;
>>> +                                       };
>>> +                               };
>>>                         };
>>>                 };
>>>  
> 
> Kind regards,
> Dylan
Dylan Van Assche March 31, 2023, 1:14 p.m. UTC | #3
On Fri, 2023-03-31 at 14:20 +0200, Konrad Dybcio wrote:
> 
> 
> On 31.03.2023 11:36, Dylan Van Assche wrote:
> > Hi Konrad,
> > 
> > On Fri, 2023-03-31 at 04:03 +0200, Konrad Dybcio wrote:
> > > 
> > > 
> > > On 30.03.2023 18:53, Dylan Van Assche wrote:
> > > > Qualcomm SDM845 SoC features a SLPI DSP which uses FastRPC
> > > > through
> > > > an allocated memory region to load files from the host
> > > > filesystem
> > > > such as sensor configuration files.
> > > > 
> > > > Add a FastRPC node at /dev/fastrpc-sdsp and a DMA region,
> > > > similar
> > > > to
> > > > downstream, to allow userspace to communicate with the SLPI via
> > > > the
> > > > FastRPC interface for initializing the sensors on the SLPI.
> > > > 
> > > > Signed-off-by: Dylan Van Assche <me@dylanvanassche.be>
> > > > ---
> > > >  arch/arm64/boot/dts/qcom/sdm845.dtsi | 24
> > > > ++++++++++++++++++++++++
> > > >  1 file changed, 24 insertions(+)
> > > > 
> > > > diff --git a/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > > b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > > index 3b547cb7aeb8..8ea4944f3ad6 100644
> > > > --- a/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > > +++ b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > > @@ -878,6 +878,14 @@ mdata_mem: mpss-metadata {
> > > >                         size = <0 0x4000>;
> > > >                         no-map;
> > > >                 };
> > > > +
> > > > +               fastrpc_mem: fastrpc {
> > > > +                       compatible = "shared-dma-pool";
> > > > +                       reusable;
> > > Please move it last to get a nice reverse-Christmas-tree layout.
> > > 
> > 
> > Will fix in v4.
> > 
> > > > +                       alloc-ranges = <0 0x00000000 0
> > > > 0xffffffff>;
> > > Would there be any issues with it starting over (1<<31 - 1)?
> > > 
> > 
> > You mean a bigger region then, like the whole CMA region then?
> > AFAIK,
> > the SLPI always use the same region expecting it to be in this
> > range.
> > However, I cannot confirm more, as I have no insights in the
> > firmware
> > running on there, this all comes from finding out what it exactly
> > does
> > on downstream.
> I was asking about the <.. 0 0xfff.f> part specifically, as that
> means
> it can't be allocated above 4 gigs. But I guess it's just how qcom
> envisioned it.
> 

I think it is limited by qcom, but I cannot be entirely sure :)
In any case, for v4, do I keep 0xffffffff?

> Also, please use 0x0 in alloc-ranges as well, this is all addresses/
> reg sizes.
> 
> Konrad

Yes, will use 0x0 instead of 0.

Dylan

> > 
> > > > +                       alignment = <0 0x400000>;
> > > Please use 0x0 for the 0 here, as it's essentially reg.size with
> > > size-cells = 2
> > 
> > Will fix in v4.
> > 
> > > 
> > > > +                       size = <0 0x1000000>;
> > > Ditto
> > 
> > Will fix in v4.
> > 
> > > 
> > > > +               };
> > > >         };
> > > >  
> > > >         adsp_pas: remoteproc-adsp {
> > > > @@ -3344,6 +3352,22 @@ glink-edge {
> > > >                                 label = "dsps";
> > > >                                 qcom,remote-pid = <3>;
> > > >                                 mboxes = <&apss_shared 24>;
> > > > +
> > > > +                               fastrpc {
> > > > +                                       compatible =
> > > > "qcom,fastrpc";
> > > > +                                       qcom,glink-channels =
> > > > "fastrpcglink-apps-dsp";
> > > > +                                       label = "sdsp";
> > > > +                                       qcom,non-secure-domain;
> > > > +                                       qcom,vmids = <0x3 0xF
> > > > 0x5
> > > > 0x6>;
> > > Please use the recently-introduced header and depend on (and
> > > make a patch atop)
> > > 
> > > https://lore.kernel.org/linux-arm-msm/8685b710-b74d-556a-815d-0ffef2b0eeff@linaro.org/T/#t
> > > 
> > > Konrad
> > > 
> > > > +                                       memory-region =
> > > > <&fastrpc_mem>;
> > > > +                                       #address-cells = <1>;
> > > > +                                       #size-cells = <0>;
> > > > +
> > > > +                                       compute-cb@0 {
> > > > +                                               compatible =
> > > > "qcom,fastrpc-compute-cb";
> > > > +                                               reg = <0>;
> > > > +                                       };
> > > > +                               };
> > > >                         };
> > > >                 };
> > > >  
> > 
> > Kind regards,
> > Dylan
Konrad Dybcio March 31, 2023, 4:58 p.m. UTC | #4
On 31.03.2023 15:14, Dylan Van Assche wrote:
> On Fri, 2023-03-31 at 14:20 +0200, Konrad Dybcio wrote:
>>
>>
>> On 31.03.2023 11:36, Dylan Van Assche wrote:
>>> Hi Konrad,
>>>
>>> On Fri, 2023-03-31 at 04:03 +0200, Konrad Dybcio wrote:
>>>>
>>>>
>>>> On 30.03.2023 18:53, Dylan Van Assche wrote:
>>>>> Qualcomm SDM845 SoC features a SLPI DSP which uses FastRPC
>>>>> through
>>>>> an allocated memory region to load files from the host
>>>>> filesystem
>>>>> such as sensor configuration files.
>>>>>
>>>>> Add a FastRPC node at /dev/fastrpc-sdsp and a DMA region,
>>>>> similar
>>>>> to
>>>>> downstream, to allow userspace to communicate with the SLPI via
>>>>> the
>>>>> FastRPC interface for initializing the sensors on the SLPI.
>>>>>
>>>>> Signed-off-by: Dylan Van Assche <me@dylanvanassche.be>
>>>>> ---
>>>>>  arch/arm64/boot/dts/qcom/sdm845.dtsi | 24
>>>>> ++++++++++++++++++++++++
>>>>>  1 file changed, 24 insertions(+)
>>>>>
>>>>> diff --git a/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>>>> b/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>>>> index 3b547cb7aeb8..8ea4944f3ad6 100644
>>>>> --- a/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>>>> +++ b/arch/arm64/boot/dts/qcom/sdm845.dtsi
>>>>> @@ -878,6 +878,14 @@ mdata_mem: mpss-metadata {
>>>>>                         size = <0 0x4000>;
>>>>>                         no-map;
>>>>>                 };
>>>>> +
>>>>> +               fastrpc_mem: fastrpc {
>>>>> +                       compatible = "shared-dma-pool";
>>>>> +                       reusable;
>>>> Please move it last to get a nice reverse-Christmas-tree layout.
>>>>
>>>
>>> Will fix in v4.
>>>
>>>>> +                       alloc-ranges = <0 0x00000000 0
>>>>> 0xffffffff>;
>>>> Would there be any issues with it starting over (1<<31 - 1)?
>>>>
>>>
>>> You mean a bigger region then, like the whole CMA region then?
>>> AFAIK,
>>> the SLPI always use the same region expecting it to be in this
>>> range.
>>> However, I cannot confirm more, as I have no insights in the
>>> firmware
>>> running on there, this all comes from finding out what it exactly
>>> does
>>> on downstream.
>> I was asking about the <.. 0 0xfff.f> part specifically, as that
>> means
>> it can't be allocated above 4 gigs. But I guess it's just how qcom
>> envisioned it.
>>
> 
> I think it is limited by qcom, but I cannot be entirely sure :)
> In any case, for v4, do I keep 0xffffffff?
Yep, let's not stray into the unknown..

Konrad
> 
>> Also, please use 0x0 in alloc-ranges as well, this is all addresses/
>> reg sizes.
>>
>> Konrad
> 
> Yes, will use 0x0 instead of 0.
> 
> Dylan
> 
>>>
>>>>> +                       alignment = <0 0x400000>;
>>>> Please use 0x0 for the 0 here, as it's essentially reg.size with
>>>> size-cells = 2
>>>
>>> Will fix in v4.
>>>
>>>>
>>>>> +                       size = <0 0x1000000>;
>>>> Ditto
>>>
>>> Will fix in v4.
>>>
>>>>
>>>>> +               };
>>>>>         };
>>>>>  
>>>>>         adsp_pas: remoteproc-adsp {
>>>>> @@ -3344,6 +3352,22 @@ glink-edge {
>>>>>                                 label = "dsps";
>>>>>                                 qcom,remote-pid = <3>;
>>>>>                                 mboxes = <&apss_shared 24>;
>>>>> +
>>>>> +                               fastrpc {
>>>>> +                                       compatible =
>>>>> "qcom,fastrpc";
>>>>> +                                       qcom,glink-channels =
>>>>> "fastrpcglink-apps-dsp";
>>>>> +                                       label = "sdsp";
>>>>> +                                       qcom,non-secure-domain;
>>>>> +                                       qcom,vmids = <0x3 0xF
>>>>> 0x5
>>>>> 0x6>;
>>>> Please use the recently-introduced header and depend on (and
>>>> make a patch atop)
>>>>
>>>> https://lore.kernel.org/linux-arm-msm/8685b710-b74d-556a-815d-0ffef2b0eeff@linaro.org/T/#t
>>>>
>>>> Konrad
>>>>
>>>>> +                                       memory-region =
>>>>> <&fastrpc_mem>;
>>>>> +                                       #address-cells = <1>;
>>>>> +                                       #size-cells = <0>;
>>>>> +
>>>>> +                                       compute-cb@0 {
>>>>> +                                               compatible =
>>>>> "qcom,fastrpc-compute-cb";
>>>>> +                                               reg = <0>;
>>>>> +                                       };
>>>>> +                               };
>>>>>                         };
>>>>>                 };
>>>>>  
>>>
>>> Kind regards,
>>> Dylan
>