From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tero Kristo Subject: Re: [PATCH 1/3] Documentation: dt: Add TI SCI clock driver Date: Mon, 21 Nov 2016 10:14:28 +0200 Message-ID: <0fe81866-8bfd-f3a7-d808-9cb23841f504@ti.com> References: <1477053961-27128-1-git-send-email-t-kristo@ti.com> <1477053961-27128-2-git-send-email-t-kristo@ti.com> <20161030204121.qvb5d33dh65awwzx@rob-hp-laptop> <41c58712-bc00-ed05-9d1d-42e31397a70c@ti.com> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8"; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: devicetree-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Rob Herring Cc: linux-clk , Michael Turquette , Stephen Boyd , Santosh Shilimkar , Nishanth Menon , "linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org" , "devicetree-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" List-Id: devicetree@vger.kernel.org On 18/11/16 19:20, Rob Herring wrote: > On Mon, Oct 31, 2016 at 7:50 AM, Tero Kristo wrote: >> On 30/10/16 22:41, Rob Herring wrote: >>> >>> On Fri, Oct 21, 2016 at 03:45:59PM +0300, Tero Kristo wrote: >>>> >>>> Add a clock implementation, TI SCI clock, that will hook to the common >>>> clock framework, and allow each clock to be controlled via TI SCI >>>> protocol. >>>> >>>> Signed-off-by: Tero Kristo >>>> --- >>>> .../devicetree/bindings/clock/ti,sci-clk.txt | 37 >>>> ++++++++++++++++++++++ >>>> MAINTAINERS | 1 + >>>> 2 files changed, 38 insertions(+) >>>> create mode 100644 >>>> Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> >>>> diff --git a/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> new file mode 100644 >>>> index 0000000..bfc3ca4 >>>> --- /dev/null >>>> +++ b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> @@ -0,0 +1,37 @@ >>>> +Texas Instruments TI-SCI Clocks >>>> +=============================== >>>> + >>>> +All clocks on Texas Instruments' SoCs that contain a System Controller, >>>> +are only controlled by this entity. Communication between a host >>>> processor >>>> +running an OS and the System Controller happens through a protocol known >>>> +as TI-SCI[1]. This clock implementation plugs into the common clock >>>> +framework and makes use of the TI-SCI protocol on clock API requests. >>>> + >>>> +[1] Documentation/devicetree/bindings/arm/keystone/ti,sci.txt >>>> + >>>> +Required properties: >>>> +------------------- >>>> +- compatible: Must be "ti,k2g-sci-clk" >>>> +- #clock-cells: Shall be 2. >>>> + In clock consumers, this cell represents the device ID and clock ID >>>> + exposed by the PM firmware. The assignments can be found in the header >>>> + files .h> (which covers the device IDs) and >>>> + .h> (which covers the clock IDs), where >>>> + is the SoC involved, for example 'k2g'. >>>> + >>>> +Examples: >>>> +-------- >>>> + >>>> +pmmc: pmmc { >>>> + compatible = "ti,k2g-sci"; >>>> + >>>> + k2g_clks: k2g_clks { >>> >>> >>> Use "clocks" for node name instead. >>> >>>> + compatible = "ti,k2g-sci-clk"; >>> >>> >>> I'm starting to think all these child nodes for SCI are pointless. Is >>> there any reason why the parent node can't be the clock provider (along >>> with all the other providers it acks as)? >> >> >> I believe the only reason to keep them separate is to have kernel side of >> things modular. If we have separate nodes, the drivers can be probed >> separately. >> >> If not, we need to build one huge blob with all the features in it, so the >> main driver can probe everything in one go, with annoying back-and-forth >> callbacks in place (assuming we still want to keep stuff somehow modular.) > > Since when is DT the only way to create a device? The main driver can > create devices for all the sub-functions like clocks. This is the same > as MFDs which have been done both ways. Yes obviously this can be done, my main point was that it will require building some sort of infra within the driver to handle this. With separate nodes, none of this is going to be needed. Also, we will lose any kind of configurability via DT if we don't have separate nodes; now we can select the available clocks / genpds via the compatible string of the clocks/genpd nodes themselves (this isn't clearly evident as of now as we only support a grand total of one device, which is k2g-evm.) Otherwise we need to probe against the main node and add a separate compatible string for every device, and carry this information to the sibling devices also somehow. It is just so much simpler if we can just keep separate nodes for them. Also, plenty of things are doing this kind of stuff already in DT/kernel, having a parent node in place and sub-functions added separately for ease of use, with apparently no visible point for having the nodes within the DT. -Tero -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Subject: Re: [PATCH 1/3] Documentation: dt: Add TI SCI clock driver To: Rob Herring References: <1477053961-27128-1-git-send-email-t-kristo@ti.com> <1477053961-27128-2-git-send-email-t-kristo@ti.com> <20161030204121.qvb5d33dh65awwzx@rob-hp-laptop> <41c58712-bc00-ed05-9d1d-42e31397a70c@ti.com> CC: linux-clk , Michael Turquette , Stephen Boyd , Santosh Shilimkar , Nishanth Menon , "linux-arm-kernel@lists.infradead.org" , "devicetree@vger.kernel.org" From: Tero Kristo Message-ID: <0fe81866-8bfd-f3a7-d808-9cb23841f504@ti.com> Date: Mon, 21 Nov 2016 10:14:28 +0200 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset="utf-8"; format=flowed List-ID: On 18/11/16 19:20, Rob Herring wrote: > On Mon, Oct 31, 2016 at 7:50 AM, Tero Kristo wrote: >> On 30/10/16 22:41, Rob Herring wrote: >>> >>> On Fri, Oct 21, 2016 at 03:45:59PM +0300, Tero Kristo wrote: >>>> >>>> Add a clock implementation, TI SCI clock, that will hook to the common >>>> clock framework, and allow each clock to be controlled via TI SCI >>>> protocol. >>>> >>>> Signed-off-by: Tero Kristo >>>> --- >>>> .../devicetree/bindings/clock/ti,sci-clk.txt | 37 >>>> ++++++++++++++++++++++ >>>> MAINTAINERS | 1 + >>>> 2 files changed, 38 insertions(+) >>>> create mode 100644 >>>> Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> >>>> diff --git a/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> new file mode 100644 >>>> index 0000000..bfc3ca4 >>>> --- /dev/null >>>> +++ b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> @@ -0,0 +1,37 @@ >>>> +Texas Instruments TI-SCI Clocks >>>> +=============================== >>>> + >>>> +All clocks on Texas Instruments' SoCs that contain a System Controller, >>>> +are only controlled by this entity. Communication between a host >>>> processor >>>> +running an OS and the System Controller happens through a protocol known >>>> +as TI-SCI[1]. This clock implementation plugs into the common clock >>>> +framework and makes use of the TI-SCI protocol on clock API requests. >>>> + >>>> +[1] Documentation/devicetree/bindings/arm/keystone/ti,sci.txt >>>> + >>>> +Required properties: >>>> +------------------- >>>> +- compatible: Must be "ti,k2g-sci-clk" >>>> +- #clock-cells: Shall be 2. >>>> + In clock consumers, this cell represents the device ID and clock ID >>>> + exposed by the PM firmware. The assignments can be found in the header >>>> + files .h> (which covers the device IDs) and >>>> + .h> (which covers the clock IDs), where >>>> + is the SoC involved, for example 'k2g'. >>>> + >>>> +Examples: >>>> +-------- >>>> + >>>> +pmmc: pmmc { >>>> + compatible = "ti,k2g-sci"; >>>> + >>>> + k2g_clks: k2g_clks { >>> >>> >>> Use "clocks" for node name instead. >>> >>>> + compatible = "ti,k2g-sci-clk"; >>> >>> >>> I'm starting to think all these child nodes for SCI are pointless. Is >>> there any reason why the parent node can't be the clock provider (along >>> with all the other providers it acks as)? >> >> >> I believe the only reason to keep them separate is to have kernel side of >> things modular. If we have separate nodes, the drivers can be probed >> separately. >> >> If not, we need to build one huge blob with all the features in it, so the >> main driver can probe everything in one go, with annoying back-and-forth >> callbacks in place (assuming we still want to keep stuff somehow modular.) > > Since when is DT the only way to create a device? The main driver can > create devices for all the sub-functions like clocks. This is the same > as MFDs which have been done both ways. Yes obviously this can be done, my main point was that it will require building some sort of infra within the driver to handle this. With separate nodes, none of this is going to be needed. Also, we will lose any kind of configurability via DT if we don't have separate nodes; now we can select the available clocks / genpds via the compatible string of the clocks/genpd nodes themselves (this isn't clearly evident as of now as we only support a grand total of one device, which is k2g-evm.) Otherwise we need to probe against the main node and add a separate compatible string for every device, and carry this information to the sibling devices also somehow. It is just so much simpler if we can just keep separate nodes for them. Also, plenty of things are doing this kind of stuff already in DT/kernel, having a parent node in place and sub-functions added separately for ease of use, with apparently no visible point for having the nodes within the DT. -Tero From mboxrd@z Thu Jan 1 00:00:00 1970 From: t-kristo@ti.com (Tero Kristo) Date: Mon, 21 Nov 2016 10:14:28 +0200 Subject: [PATCH 1/3] Documentation: dt: Add TI SCI clock driver In-Reply-To: References: <1477053961-27128-1-git-send-email-t-kristo@ti.com> <1477053961-27128-2-git-send-email-t-kristo@ti.com> <20161030204121.qvb5d33dh65awwzx@rob-hp-laptop> <41c58712-bc00-ed05-9d1d-42e31397a70c@ti.com> Message-ID: <0fe81866-8bfd-f3a7-d808-9cb23841f504@ti.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 18/11/16 19:20, Rob Herring wrote: > On Mon, Oct 31, 2016 at 7:50 AM, Tero Kristo wrote: >> On 30/10/16 22:41, Rob Herring wrote: >>> >>> On Fri, Oct 21, 2016 at 03:45:59PM +0300, Tero Kristo wrote: >>>> >>>> Add a clock implementation, TI SCI clock, that will hook to the common >>>> clock framework, and allow each clock to be controlled via TI SCI >>>> protocol. >>>> >>>> Signed-off-by: Tero Kristo >>>> --- >>>> .../devicetree/bindings/clock/ti,sci-clk.txt | 37 >>>> ++++++++++++++++++++++ >>>> MAINTAINERS | 1 + >>>> 2 files changed, 38 insertions(+) >>>> create mode 100644 >>>> Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> >>>> diff --git a/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> new file mode 100644 >>>> index 0000000..bfc3ca4 >>>> --- /dev/null >>>> +++ b/Documentation/devicetree/bindings/clock/ti,sci-clk.txt >>>> @@ -0,0 +1,37 @@ >>>> +Texas Instruments TI-SCI Clocks >>>> +=============================== >>>> + >>>> +All clocks on Texas Instruments' SoCs that contain a System Controller, >>>> +are only controlled by this entity. Communication between a host >>>> processor >>>> +running an OS and the System Controller happens through a protocol known >>>> +as TI-SCI[1]. This clock implementation plugs into the common clock >>>> +framework and makes use of the TI-SCI protocol on clock API requests. >>>> + >>>> +[1] Documentation/devicetree/bindings/arm/keystone/ti,sci.txt >>>> + >>>> +Required properties: >>>> +------------------- >>>> +- compatible: Must be "ti,k2g-sci-clk" >>>> +- #clock-cells: Shall be 2. >>>> + In clock consumers, this cell represents the device ID and clock ID >>>> + exposed by the PM firmware. The assignments can be found in the header >>>> + files .h> (which covers the device IDs) and >>>> + .h> (which covers the clock IDs), where >>>> + is the SoC involved, for example 'k2g'. >>>> + >>>> +Examples: >>>> +-------- >>>> + >>>> +pmmc: pmmc { >>>> + compatible = "ti,k2g-sci"; >>>> + >>>> + k2g_clks: k2g_clks { >>> >>> >>> Use "clocks" for node name instead. >>> >>>> + compatible = "ti,k2g-sci-clk"; >>> >>> >>> I'm starting to think all these child nodes for SCI are pointless. Is >>> there any reason why the parent node can't be the clock provider (along >>> with all the other providers it acks as)? >> >> >> I believe the only reason to keep them separate is to have kernel side of >> things modular. If we have separate nodes, the drivers can be probed >> separately. >> >> If not, we need to build one huge blob with all the features in it, so the >> main driver can probe everything in one go, with annoying back-and-forth >> callbacks in place (assuming we still want to keep stuff somehow modular.) > > Since when is DT the only way to create a device? The main driver can > create devices for all the sub-functions like clocks. This is the same > as MFDs which have been done both ways. Yes obviously this can be done, my main point was that it will require building some sort of infra within the driver to handle this. With separate nodes, none of this is going to be needed. Also, we will lose any kind of configurability via DT if we don't have separate nodes; now we can select the available clocks / genpds via the compatible string of the clocks/genpd nodes themselves (this isn't clearly evident as of now as we only support a grand total of one device, which is k2g-evm.) Otherwise we need to probe against the main node and add a separate compatible string for every device, and carry this information to the sibling devices also somehow. It is just so much simpler if we can just keep separate nodes for them. Also, plenty of things are doing this kind of stuff already in DT/kernel, having a parent node in place and sub-functions added separately for ease of use, with apparently no visible point for having the nodes within the DT. -Tero