[PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
From: Thomas Abraham thomas...@samsung.com
Add a new optional boost-frequency binding for specifying the frequencies
usable in boost mode.
Cc: Rob Herring robh...@kernel.org
Cc: Pawel Moll pawel.m...@arm.com
Cc: Mark Rutland mark.rutl...@arm.com
Cc: Ian Campbell ijc+devicet...@hellion.org.uk
Cc: Kumar Gala ga...@codeaurora.org
Signed-off-by: Thomas Abraham thomas...@samsung.com
Acked-by: Viresh Kumar viresh.ku...@linaro.org
Acked-by: Nishanth Menon n...@ti.com
Acked-by: Lukasz Majewski l.majew...@samsung.com
---
.../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38
1 file changed, 38 insertions(+)
create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
new file mode 100644
index 000..63ed0fc
--- /dev/null
+++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
@@ -0,0 +1,38 @@
+* Device tree binding for CPU boost frequency (aka over-clocking)
+
+Certain CPU's can be operated in optional 'boost' mode (or sometimes referred
as
+overclocking) in which the CPU can operate at frequencies which are not
+specified by the manufacturer as CPU's operating frequency.
+
+Optional Properties:
+- boost-frequencies: list of frequencies in KHz to be used only in boost mode.
+ This list should be a subset of frequencies listed in operating-points
+ property. Refer to Documentation/devicetree/bindings/power/opp.txt for
+ details about operating-points property.
+
+Example:
+
+ cpus {
+ #address-cells = 1;
+ #size-cells = 0;
+ cpu@0 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 0;
+
+ operating-points =
+ 150 135
+ 140 1287500
+ 130 125
+ 120 1187500
+ 110 1137500
+ 100 1087500
+ ;
+ boost-frequencies = 150 140;
+ };
+ cpu@1 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 1;
+ };
+ };
--
1.7.9.5
--
To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
Hi,
Apologies for being somewhat late w.r.t. review on this.
On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote:
From: Thomas Abraham thomas...@samsung.com
Add a new optional boost-frequency binding for specifying the frequencies
usable in boost mode.
Cc: Rob Herring robh...@kernel.org
Cc: Pawel Moll pawel.m...@arm.com
Cc: Mark Rutland mark.rutl...@arm.com
Cc: Ian Campbell ijc+devicet...@hellion.org.uk
Cc: Kumar Gala ga...@codeaurora.org
Signed-off-by: Thomas Abraham thomas...@samsung.com
Acked-by: Viresh Kumar viresh.ku...@linaro.org
Acked-by: Nishanth Menon n...@ti.com
Acked-by: Lukasz Majewski l.majew...@samsung.com
---
.../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38
1 file changed, 38 insertions(+)
create mode 100644
Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
new file mode 100644
index 000..63ed0fc
--- /dev/null
+++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
@@ -0,0 +1,38 @@
+* Device tree binding for CPU boost frequency (aka over-clocking)
+
+Certain CPU's can be operated in optional 'boost' mode (or sometimes
referred as
Nit: CPUs (we're not greengrocers [1])
+overclocking) in which the CPU can operate at frequencies which are not
+specified by the manufacturer as CPU's operating frequency.
+
+Optional Properties:
+- boost-frequencies: list of frequencies in KHz to be used only in boost
mode.
+ This list should be a subset of frequencies listed in operating-points
+ property. Refer to Documentation/devicetree/bindings/power/opp.txt for
+ details about operating-points property.
What is 'boost-mode'?
What are the limitations on boost frequencies? When is a CPU expected to
go to these frequencies and for now long? When should I as a dt author
place elements in boost-frequencies?
Why are these in both operating-points and boost-frequencies? It'll be
really easy to accidentally forget to mark something as a
boost-frequency this way. Why not have a boost-points instead?
+
+Example:
+
+ cpus {
+ #address-cells = 1;
+ #size-cells = 0;
+ cpu@0 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 0;
+
+ operating-points =
+ 150 135
+ 140 1287500
+ 130 125
+ 120 1187500
+ 110 1137500
+ 100 1087500
+ ;
+ boost-frequencies = 150 140;
This is more of a general issue, but I hate the whole cpufreq-cpu0 way
of assuming that all CPUs mirror CPU0.
It would be nicer if either this were dropped in /cpus or repeated
per-cpu.
Cheers,
Mark.
+ };
+ cpu@1 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 1;
+ };
+ };
--
1.7.9.5
[1]
http://en.wikipedia.org/wiki/Apostrophe#Superfluous_apostrophes_.28.22greengrocers.27_apostrophes.22.29
--
To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
Hi Mark,
On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote:
Hi,
Apologies for being somewhat late w.r.t. review on this.
On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote:
From: Thomas Abraham thomas...@samsung.com
Add a new optional boost-frequency binding for specifying the frequencies
usable in boost mode.
Cc: Rob Herring robh...@kernel.org
Cc: Pawel Moll pawel.m...@arm.com
Cc: Mark Rutland mark.rutl...@arm.com
Cc: Ian Campbell ijc+devicet...@hellion.org.uk
Cc: Kumar Gala ga...@codeaurora.org
Signed-off-by: Thomas Abraham thomas...@samsung.com
Acked-by: Viresh Kumar viresh.ku...@linaro.org
Acked-by: Nishanth Menon n...@ti.com
Acked-by: Lukasz Majewski l.majew...@samsung.com
---
.../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38
1 file changed, 38 insertions(+)
create mode 100644
Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
new file mode 100644
index 000..63ed0fc
--- /dev/null
+++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
@@ -0,0 +1,38 @@
+* Device tree binding for CPU boost frequency (aka over-clocking)
+
+Certain CPU's can be operated in optional 'boost' mode (or sometimes
referred as
Nit: CPUs (we're not greengrocers [1])
+overclocking) in which the CPU can operate at frequencies which are not
+specified by the manufacturer as CPU's operating frequency.
+
+Optional Properties:
+- boost-frequencies: list of frequencies in KHz to be used only in boost
mode.
+ This list should be a subset of frequencies listed in operating-points
+ property. Refer to Documentation/devicetree/bindings/power/opp.txt for
+ details about operating-points property.
What is 'boost-mode'?
boost-mode activates additional one or more cpu clock speeds (which
are not specified as operating frequency of the cpu by the
manufacturer). The cpu is allowed to operate in these boost mode
speeds when the boost mode is active. The boost mode speeds are
usually undocumented. Some of the chip samples could be clocked in
boost mode speeds and only such samples can be safely operated in
boost mode.
The mechanism of entry into and exit out of boost mode is outside the
scope of this documentation.
What are the limitations on boost frequencies? When is a CPU expected to
go to these frequencies and for now long? When should I as a dt author
place elements in boost-frequencies?
I will add these details in the next revision of this patch.
Why are these in both operating-points and boost-frequencies? It'll be
really easy to accidentally forget to mark something as a
boost-frequency this way. Why not have a boost-points instead?
Does boost-points mean a set of clock speeds which are not listed as
part of operating-points property? If yes, that also is a possible
implementation (it was implemented in one of the earlier version of
this series). Could you confirm that you want the boost mode speeds to
be exclusive of the speeds listed in operating-points?
Thanks,
Thomas.
+
+Example:
+
+ cpus {
+ #address-cells = 1;
+ #size-cells = 0;
+ cpu@0 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 0;
+
+ operating-points =
+ 150 135
+ 140 1287500
+ 130 125
+ 120 1187500
+ 110 1137500
+ 100 1087500
+ ;
+ boost-frequencies = 150 140;
This is more of a general issue, but I hate the whole cpufreq-cpu0 way
of assuming that all CPUs mirror CPU0.
It would be nicer if either this were dropped in /cpus or repeated
per-cpu.
Cheers,
Mark.
+ };
+ cpu@1 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 1;
+ };
+ };
--
1.7.9.5
[1]
http://en.wikipedia.org/wiki/Apostrophe#Superfluous_apostrophes_.28.22greengrocers.27_apostrophes.22.29
--
To unsubscribe from this list: send the line unsubscribe devicetree in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On Fri, May 30, 2014 at 11:45 PM, Tomasz Figa tomasz.f...@gmail.com wrote:
On 30.05.2014 20:05, Thomas Abraham wrote:
Hi Mark,
On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote:
Hi,
Apologies for being somewhat late w.r.t. review on this.
On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote:
From: Thomas Abraham thomas...@samsung.com
Add a new optional boost-frequency binding for specifying the frequencies
usable in boost mode.
Cc: Rob Herring robh...@kernel.org
Cc: Pawel Moll pawel.m...@arm.com
Cc: Mark Rutland mark.rutl...@arm.com
Cc: Ian Campbell ijc+devicet...@hellion.org.uk
Cc: Kumar Gala ga...@codeaurora.org
Signed-off-by: Thomas Abraham thomas...@samsung.com
Acked-by: Viresh Kumar viresh.ku...@linaro.org
Acked-by: Nishanth Menon n...@ti.com
Acked-by: Lukasz Majewski l.majew...@samsung.com
---
.../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38
1 file changed, 38 insertions(+)
create mode 100644
Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
new file mode 100644
index 000..63ed0fc
--- /dev/null
+++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt
@@ -0,0 +1,38 @@
+* Device tree binding for CPU boost frequency (aka over-clocking)
+
+Certain CPU's can be operated in optional 'boost' mode (or sometimes
referred as
Nit: CPUs (we're not greengrocers [1])
+overclocking) in which the CPU can operate at frequencies which are not
+specified by the manufacturer as CPU's operating frequency.
+
+Optional Properties:
+- boost-frequencies: list of frequencies in KHz to be used only in boost
mode.
+ This list should be a subset of frequencies listed in operating-points
+ property. Refer to Documentation/devicetree/bindings/power/opp.txt for
+ details about operating-points property.
What is 'boost-mode'?
boost-mode activates additional one or more cpu clock speeds (which
are not specified as operating frequency of the cpu by the
manufacturer). The cpu is allowed to operate in these boost mode
speeds when the boost mode is active. The boost mode speeds are
usually undocumented. Some of the chip samples could be clocked in
boost mode speeds and only such samples can be safely operated in
boost mode.
IMHO the most important part that I believe should be stated in the
documentation is that CPU usually can operate in boost mode for limited
amount of time, which depends on thermal conditions, which makes the
boost operating points separate from normal ones, which can be used at
any time.
The factors that allow a CPU to operate in boost mode and the duration
the CPU can operate in boost mode clock speed is a CPU specific
description. This binding does not describe such CPU specific
behavior. This binding only states the additional clock speeds CPU is
allowed to operate in boost mode. The boost mode entry and exit
conditions are implementation specific and not part of the scope of
this binding and documentation.
The mechanism of entry into and exit out of boost mode is outside the
scope of this documentation.
What are the limitations on boost frequencies? When is a CPU expected to
go to these frequencies and for now long? When should I as a dt author
place elements in boost-frequencies?
I will add these details in the next revision of this patch.
Why are these in both operating-points and boost-frequencies? It'll be
really easy to accidentally forget to mark something as a
boost-frequency this way. Why not have a boost-points instead?
Does boost-points mean a set of clock speeds which are not listed as
part of operating-points property? If yes, that also is a possible
implementation (it was implemented in one of the earlier version of
this series). Could you confirm that you want the boost mode speeds to
be exclusive of the speeds listed in operating-points?
It seems reasonable to have boost operating points completely separate
from normal ones, so that a kernel without support for boost mode will
not use them. Also considering my comment above, logically boost
operating points are not considered normal operating points, due to
various constraints that need to be met to use them (i.e. mostly thermal
conditions).
Thanks,
Thomas.
+
+Example:
+
+ cpus {
+ #address-cells = 1;
+ #size-cells = 0;
+ cpu@0 {
+ device_type = cpu;
+ compatible = arm,cortex-a9;
+ reg = 0;
+
+ operating-points =
+ 150 135
+ 140 1287500
+ 130 125
+ 120 1187500
+ 110 1137500
+ 100
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 30.05.2014 20:33, Thomas Abraham wrote: On Fri, May 30, 2014 at 11:45 PM, Tomasz Figa tomasz.f...@gmail.com wrote: On 30.05.2014 20:05, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. IMHO the most important part that I believe should be stated in the documentation is that CPU usually can operate in boost mode for limited amount of time, which depends on thermal conditions, which makes the boost operating points separate from normal ones, which can be used at any time. The factors that allow a CPU to operate in boost mode and the duration the CPU can operate in boost mode clock speed is a CPU specific description. This binding does not describe such CPU specific behavior. This binding only states the additional clock speeds CPU is allowed to operate in boost mode. The boost mode entry and exit conditions are implementation specific and not part of the scope of this binding and documentation. The binding documentation describes properties related to boost mode and so it must explain what boost mode is, otherwise those properties will not make any sense. I'm not saying that specific constraints must be listed, but the text should mention that those operating points might need some. Best regards, Tomasz -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 30/05/14 19:15, Tomasz Figa wrote: On 30.05.2014 20:05, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. IMHO the most important part that I believe should be stated in the documentation is that CPU usually can operate in boost mode for limited amount of time, which depends on thermal conditions, which makes the boost operating points separate from normal ones, which can be used at any time. Yes exactly what I mentioned couple of times on previous version of this patch set[1][2] The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? I was told that index is not preferred based on the previous discussions when the OPP bindings were designed. In addition the OPP binding doesn't enforce any ordering. There are thermal bindings that assume otherwise and is broken. So boost-points is not feasible. Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? It seems reasonable to have boost operating points completely separate from normal ones, so that a kernel without support for boost mode will not use them. Also considering my comment above, logically boost operating points are not considered normal operating points, due to various constraints that need to be met to use them (i.e. mostly thermal conditions). IMO, at-least the existing OPP binding can't distinguish between under-, nominal- and over-drive OPP points. So my suggestion was to have a property that provides the beginning of these 3 points on the OPP curve. Regards, Sudeep [1] https://www.mail-archive.com/linux-samsung-soc@vger.kernel.org/msg26250.html [2] http://permalink.gmane.org/gmane.linux.kernel.samsung-soc/31552 -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 30.05.2014 20:38, Sudeep Holla wrote: On 30/05/14 19:15, Tomasz Figa wrote: On 30.05.2014 20:05, Thomas Abraham wrote: On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: [snip] Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? I was told that index is not preferred based on the previous discussions when the OPP bindings were designed. In addition the OPP binding doesn't enforce any ordering. There are thermal bindings that assume otherwise and is broken. So boost-points is not feasible. My understanding of Mark's comment was that the boost-points property would use the same format as operating-points and parsing code would just concatenate operating points with boost points after making the latter with necessary flag or whatever. Best regards, Tomasz -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 30/05/14 19:41, Tomasz Figa wrote: On 30.05.2014 20:38, Sudeep Holla wrote: On 30/05/14 19:15, Tomasz Figa wrote: On 30.05.2014 20:05, Thomas Abraham wrote: On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: [snip] Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? I was told that index is not preferred based on the previous discussions when the OPP bindings were designed. In addition the OPP binding doesn't enforce any ordering. There are thermal bindings that assume otherwise and is broken. So boost-points is not feasible. My understanding of Mark's comment was that the boost-points property would use the same format as operating-points and parsing code would just concatenate operating points with boost points after making the latter with necessary flag or whatever. Ah, I misunderstood that. That should be fine as it avoids duplication. Regards, Sudeep -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value in either table we might go into a boost mode when we didn't want to! Cheers, Mark. -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up device with dts typo. you could give a wrong voltage(extra 0?) and ensure you never use the chip ever again.. typos are dt bugs, we can do the best to write robust code to report them. [1] http://marc.info/?t=14005961858r=1w=2 -- Regards, Nishanth Menon -- To
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 30.05.2014 21:50, Nishanth Menon wrote: On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up device with dts typo. you could give a wrong voltage(extra 0?) and ensure you never use the chip ever again.. typos are dt bugs, we can do the best to write robust code to report them. Typos are not the
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 05/30/2014 03:02 PM, Tomasz Figa wrote: On 30.05.2014 21:50, Nishanth Menon wrote: On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up device with dts typo. you could give a wrong voltage(extra 0?) and ensure you never use the chip ever again.. typos are dt bugs, we can do the best to write robust code to
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 05/30/2014 03:19 PM, Tomasz Figa wrote: On 30.05.2014 22:13, Nishanth Menon wrote: On 05/30/2014 03:02 PM, Tomasz Figa wrote: On 30.05.2014 21:50, Nishanth Menon wrote: On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up device with dts typo. you could give a wrong voltage(extra 0?) and ensure
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On Fri, May 30, 2014 at 3:33 PM, Nishanth Menon n...@ti.com wrote: On 05/30/2014 03:19 PM, Tomasz Figa wrote: On 30.05.2014 22:13, Nishanth Menon wrote: On 05/30/2014 03:02 PM, Tomasz Figa wrote: On 30.05.2014 21:50, Nishanth Menon wrote: On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. On Fri, May 30, 2014 at 10:01:17AM +0100, Thomas Abraham wrote: From: Thomas Abraham thomas...@samsung.com Add a new optional boost-frequency binding for specifying the frequencies usable in boost mode. Cc: Rob Herring robh...@kernel.org Cc: Pawel Moll pawel.m...@arm.com Cc: Mark Rutland mark.rutl...@arm.com Cc: Ian Campbell ijc+devicet...@hellion.org.uk Cc: Kumar Gala ga...@codeaurora.org Signed-off-by: Thomas Abraham thomas...@samsung.com Acked-by: Viresh Kumar viresh.ku...@linaro.org Acked-by: Nishanth Menon n...@ti.com Acked-by: Lukasz Majewski l.majew...@samsung.com --- .../devicetree/bindings/cpufreq/cpufreq-boost.txt | 38 1 file changed, 38 insertions(+) create mode 100644 Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt new file mode 100644 index 000..63ed0fc --- /dev/null +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-boost.txt @@ -0,0 +1,38 @@ +* Device tree binding for CPU boost frequency (aka over-clocking) + +Certain CPU's can be operated in optional 'boost' mode (or sometimes referred as Nit: CPUs (we're not greengrocers [1]) +overclocking) in which the CPU can operate at frequencies which are not +specified by the manufacturer as CPU's operating frequency. + +Optional Properties: +- boost-frequencies: list of frequencies in KHz to be used only in boost mode. + This list should be a subset of frequencies listed in operating-points + property. Refer to Documentation/devicetree/bindings/power/opp.txt for + details about operating-points property. What is 'boost-mode'? boost-mode activates additional one or more cpu clock speeds (which are not specified as operating frequency of the cpu by the manufacturer). The cpu is allowed to operate in these boost mode speeds when the boost mode is active. The boost mode speeds are usually undocumented. Some of the chip samples could be clocked in boost mode speeds and only such samples can be safely operated in boost mode. The mechanism of entry into and exit out of boost mode is outside the scope of this documentation. What are the limitations on boost frequencies? When is a CPU expected to go to these frequencies and for now long? When should I as a dt author place elements in boost-frequencies? I will add these details in the next revision of this patch. Cheers. Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 05/30/2014 03:45 PM, Rob Herring wrote: On Fri, May 30, 2014 at 3:33 PM, Nishanth Menon n...@ti.com wrote: On 05/30/2014 03:19 PM, Tomasz Figa wrote: On 30.05.2014 22:13, Nishanth Menon wrote: On 05/30/2014 03:02 PM, Tomasz Figa wrote: On 30.05.2014 21:50, Nishanth Menon wrote: On 05/30/2014 01:55 PM, Mark Rutland wrote: On Fri, May 30, 2014 at 07:05:43PM +0100, Thomas Abraham wrote: Hi Mark, On Fri, May 30, 2014 at 6:38 PM, Mark Rutland mark.rutl...@arm.com wrote: Hi, Apologies for being somewhat late w.r.t. review on this. [...] Why are these in both operating-points and boost-frequencies? It'll be really easy to accidentally forget to mark something as a boost-frequency this way. Why not have a boost-points instead? Does boost-points mean a set of clock speeds which are not listed as part of operating-points property? If yes, that also is a possible implementation (it was implemented in one of the earlier version of this series). Could you confirm that you want the boost mode speeds to be exclusive of the speeds listed in operating-points? If these boost mode operating points are not generally advisable for use as the other operating-points are, then they should IMO been in an entirely separate property exclusive of (but in the same format as) the operating-points property, e.g. operating points = A B, C D; boost-points = E F; you are asking boost frequencies to remain for ever tied down to OPP style description. What we are trying to describe? What are my SoC's overclocked frequencies? That description can be used even in a system that does not use OPP style table (say ACPI based OPP tables or whatever acronymned table). Tying it down to operating points just because we have it today as a convenient description, is limiting. Further, if we decide to educate boost-frequencies to also indicate how long is it safe? That does indeed belong to boost-frequency description and not OPP description. Or if we decide to change operating-points description[1] in the future has an impact on boost-points description, when it should not have. Otherwise, without boost-mode support we have to parse the boost mode table to figure out which points to avoid. Or if someone typos a value That is OS usage of h/w description - yeah - in anycase, if OS has no ability to deal with boost-frequencies, it should skip it for sure. in either table we might go into a boost mode when we didn't want to! There are other ways to screw up device with dts typo. you could give a wrong voltage(extra 0?) and ensure you never use the chip ever again.. typos are dt bugs, we can do the best to write robust code to report them. Typos are not the primary thing to worry about here. Adding boost frequencies to the list of primary operating points is flawed, because an OS that has no idea of boost mode will use them as normal operating points and this is not desired. That means we have an implementation bug in OS since it does not consider the complete hardware description that device tree is providing. An analogy will be a regulator compatible match being used but regulator-min-microvolt and regulator-max-microvolt being ignored by OS. No. The operating points bindings were defined far before this boost-frequency and so there is no requirement to support the latter. So, we dont add any new bindings ever again? /me blinks. *IF* we add a new property in the future, do we expect the new description to be supported in older kernel(which could not have known about it)? How far are we taking this ABI thing? Documentation/devicetree/bindings/ABI.txt states: 3) Bindings can be augmented, but the driver shouldn't break when given the old binding. ie. add additional properties, but don't change the meaning of an existing property. For drivers, default to the original behaviour when a newly added property is missing. we are not changing the meaning of existing property, we are augumenting it. You are changing the meaning of entries in that they can have additional data which changes their properties. If we accept the DT changes (as DT maintainers) and reject the kernel changes (as kernel maintainers), you would be left with a broken system. That is true and I agree. In my opinion, *IF* we are concerned about polluting operating-points description, why dont we enforce that the boost operating points should NOT be defined in the current operating-points description - and - just follow what Rob suggested and iMx already does - add such operating points from platform code. I believe I also said on this and other attempts to bandage up the existing OPP binding, that we should not change/append it, but define a new binding for OPPs that addresses this and other issues. Otherwise, I'm going to just NAK every incomplete OPP binding bandaid. Are we open to creating a completely mutually-exclusive binding and maintain the legacy one as legacy
Re: [PATCH v6 2/2] Documentation: devicetree: Add boost-frequency binding to list boost mode frequency
On 05/30/2014 03:43 PM, Tomasz Figa wrote: [...] OK, so you add overclocked frequencies to operating-points property, boost-frequency property, build a dtb, use it with a kernel that doesn't support boost-frequency and nicely overheat (and likely destroy) your board. I don't think this makes too much sense, sorry. Yes, that is unfortunately true as well :( Sigh. -- Regards, Nishanth Menon -- To unsubscribe from this list: send the line unsubscribe linux-samsung-soc in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
