Skip to contentSkip to conversation

Add QTI QFPROM-Efuse driver support

RFC In review Ravi Kumar Bokka 6 participants 18 messages Jun 11, 2020 – Jun 17, 2020

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

v2 cover Add QTI QFPROM-Efuse driver support

This patch series adds qfprom-efuse controller driver support.

This driver can access the raw qfprom regions for fuse blowing.

The current existed qfprom driver is only supports for cpufreq, thermal sensors drivers by read out calibration data, speed bins..etc which is stored by qfprom efuses.

Ravi Kumar Bokka (3):
dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse
drivers: nvmem: Add QTI qfprom-efuse support
arm64: dts: qcom: sc7180: Add qfprom-efuse

 .../devicetree/bindings/nvmem/qfprom.yaml          |  52 +++
 arch/arm64/boot/dts/qcom/sc7180-idp.dts            |   4 +
 arch/arm64/boot/dts/qcom/sc7180.dtsi               |  10 +-
 drivers/nvmem/Kconfig                              |   1 +
 drivers/nvmem/qfprom.c                             | 405 +++++++++++++++++++--
 5 files changed, 449 insertions(+), 23 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml
Signature · 1 line
Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.
#1 lore raw

RFC v2 1/3 dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

This patch adds dt-bindings document for qfprom-efuse controller.

---
 .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
 1 file changed, 52 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +52 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom
+
+  reg:
+    maxItems: 3
+
+required:
+   - compatible
+   - reg
+   - reg-names
+   - clocks
+   - clock-names
+
+examples:
+  - |
+    #include <dt-bindings/clock/qcom,gcc-sc7180.h>
+
+    qfprom@780000 {
+	compatible = "qcom,qfprom";
+	reg = <0 0x00780000 0 0x7a0>,
+	      <0 0x00782000 0 0x100>,
+	      <0 0x00784000 0 0x8ff>;
+	reg-names = "raw", "conf", "corrected";
+	clocks = <&gcc GCC_SEC_CTRL_CLK_SRC>;
+	clock-names = "secclk";
+
+	qusb2p_hstx_trim: hstx-trim-primary@25b {
+		reg = <0x25b 0x1>;
+		bits = <1 3>;
+	};
+    };
+
+    &qfprom {
+        vcc-supply = <&vreg_l11a_1p8>;
+    };
+
Signature · 1 line
Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.
#2 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

Quoted 1 line · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

there is an existing bindings doc (in .txt) which you should convert to yaml, and then add another patch to extend it.

Quoted 66 lines
---
  .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
  1 file changed, 52 insertions(+)

create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml

Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +52 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom
+
+  reg:
+    maxItems: 3
+
+required:
+   - compatible
+   - reg
+   - reg-names
+   - clocks
+   - clock-names
+
+examples:
+  - |
+    #include <dt-bindings/clock/qcom,gcc-sc7180.h>
+
+    qfprom@780000 {
+	compatible = "qcom,qfprom";
+	reg = <0 0x00780000 0 0x7a0>,
+	      <0 0x00782000 0 0x100>,
+	      <0 0x00784000 0 0x8ff>;
+	reg-names = "raw", "conf", "corrected";
+	clocks = <&gcc GCC_SEC_CTRL_CLK_SRC>;
+	clock-names = "secclk";
+
+	qusb2p_hstx_trim: hstx-trim-primary@25b {
+		reg = <0x25b 0x1>;
+		bits = <1 3>;
+	};
+    };
+
+    &qfprom {
+        vcc-supply = <&vreg_l11a_1p8>;
+    };
+
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation
maintainer
#3 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

In reply to Ravi Kumar Bokka
Quoted 8 lines · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

---
 .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
 1 file changed, 52 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml

My bot found errors running 'make dt_binding_check' on your patch:

Documentation/devicetree/bindings/nvmem/qfprom.yaml: while scanning a block scalar
in "<unicode string>", line 31, column 5
found a tab character where an indentation space is expected
in "<unicode string>", line 35, column 1
Documentation/devicetree/bindings/Makefile:12: recipe for target 'Documentation/devicetree/bindings/nvmem/qfprom.example.dts' failed make[1]: *** [Documentation/devicetree/bindings/nvmem/qfprom.example.dts] Error 1
make[1]: *** Waiting for unfinished jobs.... /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/nvmem/qfprom.yaml: ignoring, error parsing file
warning: no schema found in file: ./Documentation/devicetree/bindings/nvmem/qfprom.yaml /builds/robherring/linux-dt-review/Documentation/devicetree/bindings/nvmem/qfprom.yaml: ignoring, error parsing file
warning: no schema found in file: ./Documentation/devicetree/bindings/nvmem/qfprom.yaml
Makefile:1300: recipe for target 'dt_binding_check' failed
make: *** [dt_binding_check] Error 2

See https://patchwork.ozlabs.org/patch/1307399

If you already ran 'make dt_binding_check' and didn't see the above error(s), then make sure dt-schema is up to date:

pip3 install git+https://github.com/devicetree-org/dt-schema.git@master --upgrade

Please check and re-submit.

#4 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

In reply to Ravi Kumar Bokka

Hi,

Quoted 7 lines · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

---
 .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
 1 file changed, 52 insertions(+)

Overall comment: I reviewed your v1 series and so I'm obviously interested in your series. Please CC me on future versions.

I would also note that, since this is relevant to Qualcomm SoCs that you probably should be CCing "[email protected]" on your series.

Quoted 16 lines
 create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +8 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 10 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +10 -0
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom
- const: qcom,qfprom

NOTE: old SoCs won't have both of these and thus they will get flagged with "dtbs_check", but I believe that's fine (Rob can correct me if I'm wrong). The code should still work OK if the SoC isn't there but it would be good to fix old dts files to have the SoC specific string too.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 21

Quoted 3 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +3 -0
+
+  reg:
+    maxItems: 3

Please address feedback feedback on v1. If you disagree with my feedback it's OK to say so (I make no claims of being always right), but silently ignoring my feedback and sending the next version doesn't make me feel like it's a good use of my time to keep reviewing your series. Specifically I suggested that you actually add descriptions rather than just putting "maxItems: 3".

With all that has been discussed, I think the current best thing to put there is:

# If the QFPROM is read-only OS image then only the corrected region # needs to be provided. If the QFPROM is writable then all 3 regions # must be provided.
oneOf:
- items:
- description: The start of the corrected region.
- items:
- description: The start of the raw region.
- description: The start of the config region.
- description: The start of the corrected region.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 22

Quoted 1 line
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +1 -0
+

You missed a bunch of things that you should document:

# Clocks must be provided if QFPROM is writable from the OS image. clocks:
maxItems: 1
clock-names:
const: sec

# Supply reference must be provided if QFPROM is writable from the OS image. vcc-supply:
description: Our power supply.

# Needed if any child nodes are present.
"#address-cells":
const: 1
"#size-cells":
const: 1

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 26

Quoted 4 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+required:
+   - compatible
+   - reg
+   - reg-names

reg-names is discouraged. Please remove. I always point people here as a reference:

https://lore.kernel.org/r/CAL_Jsq+MMunmVWqeW9v2RyzsMKP+=kMzeTHNMG4JDHM7Fy0HBg@mail.gmail.com/

You can just figure out whether there are 3 register fields or 1 register field.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 28

Quoted 2 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +2 -0
+   - clocks
+   - clock-names

You can't retroactively make things required. In read-only mode I believe we don't require clocks/clock-names. Presumably the clock is only needed if we're writing?

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 35

Quoted 6 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +6 -0
+examples:
+  - |
+    #include <dt-bindings/clock/qcom,gcc-sc7180.h>
+
+    qfprom@780000 {
+       compatible = "qcom,qfprom";

As pointed out by Rob H in v1, you have a whole bunch of tabs in here that "dt_binding_check" yells about. Please fix and confirm that you ran "dt_binding_check" on your bindings and they passed with no errors. This is another case where someone took the time to respond to your v1 but you didn't address their comments in sending v2.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 39

Quoted 4 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+       reg = <0 0x00780000 0 0x7a0>,
+             <0 0x00782000 0 0x100>,
+             <0 0x00784000 0 0x8ff>;
+       reg-names = "raw", "conf", "corrected";

You are missing #address-cells and #size-cells, which are required because you have a sub-node.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 41

Quoted 2 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +2 -0
+       clocks = <&gcc GCC_SEC_CTRL_CLK_SRC>;
+       clock-names = "secclk";

As pointed out in v1, people don't like clock names to end in "clk". Just call this "sec".

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 51

Quoted 9 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +9 -0
+       qusb2p_hstx_trim: hstx-trim-primary@25b {
+               reg = <0x25b 0x1>;
+               bits = <1 3>;
+       };
+    };
+
+    &qfprom {
+        vcc-supply = <&vreg_l11a_1p8>;
+    };

Just fold the vcc-supply into the above node. Note that your current example refers to a phandle that doesn't exist.

Quoted 3 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +1 -1
+
--

Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.

For your edification, I have placed a patch that fixes all my own review feedback (and passes dt_binding_check) at:

https://crrev.com/c/2243853

Please either fold this into your next patch or provide comments about why you aren't taking one of my suggestions.

-Doug

#5 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

Hi,

Quoted 85 lines · Doug Anderson

Hi,

Quoted 7 lines · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

---
 .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
 1 file changed, 52 insertions(+)

Overall comment: I reviewed your v1 series and so I'm obviously interested in your series. Please CC me on future versions.

I would also note that, since this is relevant to Qualcomm SoCs that you probably should be CCing "[email protected]" on your series.

Quoted 16 lines
 create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +8 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 10 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +10 -0
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom
- const: qcom,qfprom

NOTE: old SoCs won't have both of these and thus they will get flagged with "dtbs_check", but I believe that's fine (Rob can correct me if I'm wrong). The code should still work OK if the SoC isn't there but it would be good to fix old dts files to have the SoC specific string too.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 21

Quoted 3 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +3 -0
+
+  reg:
+    maxItems: 3

Please address feedback feedback on v1. If you disagree with my feedback it's OK to say so (I make no claims of being always right), but silently ignoring my feedback and sending the next version doesn't make me feel like it's a good use of my time to keep reviewing your series. Specifically I suggested that you actually add descriptions rather than just putting "maxItems: 3".

With all that has been discussed, I think the current best thing to put there is:

Quoted 10 lines

# If the QFPROM is read-only OS image then only the corrected region # needs to be provided. If the QFPROM is writable then all 3 regions # must be provided.
oneOf:
- items:
- description: The start of the corrected region.
- items:
- description: The start of the raw region.
- description: The start of the config region.
- description: The start of the corrected region.

To address some of Srinivas's comments, I think you should actually add the 4th range in here too:

- description: The start of the security control region.

That allows you to read 0x6000 and get the major/minor/step properly.

I will also note that as I've started reviewing the driver code it'll probably make everyone's life easiest if you keep the "corrected" region first, even if it's not numerically first.

I've updated my FIXUP patch again. I might notice more comments as I review the driver more, but we'll see...

-Doug

#6 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

In reply to Doug Anderson
Quoted 50 lines · Doug Anderson

Hi,

Quoted 7 lines · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

---
  .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
  1 file changed, 52 insertions(+)

Overall comment: I reviewed your v1 series and so I'm obviously interested in your series. Please CC me on future versions.

I would also note that, since this is relevant to Qualcomm SoCs that you probably should be CCing "[email protected]" on your series.

Quoted 16 lines

create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml

Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +8 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 10 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +10 -0
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

Quoted 11 lines

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom

Above is not required for now in this patchset, as we can attach data at runtime specific to version of the qfprom.

This can be added when required!

Quoted 59 lines

- const: qcom,qfprom

NOTE: old SoCs won't have both of these and thus they will get flagged with "dtbs_check", but I believe that's fine (Rob can correct me if I'm wrong). The code should still work OK if the SoC isn't there but it would be good to fix old dts files to have the SoC specific string too.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 21

Quoted 3 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +3 -0
+
+  reg:
+    maxItems: 3

Please address feedback feedback on v1. If you disagree with my feedback it's OK to say so (I make no claims of being always right), but silently ignoring my feedback and sending the next version doesn't make me feel like it's a good use of my time to keep reviewing your series. Specifically I suggested that you actually add descriptions rather than just putting "maxItems: 3".

With all that has been discussed, I think the current best thing to put there is:

# If the QFPROM is read-only OS image then only the corrected region # needs to be provided. If the QFPROM is writable then all 3 regions # must be provided.
oneOf:
- items:
- description: The start of the corrected region.
- items:
- description: The start of the raw region.
- description: The start of the config region.
- description: The start of the corrected region.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 22

Quoted 1 line
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +1 -0
+

You missed a bunch of things that you should document:

# Clocks must be provided if QFPROM is writable from the OS image. clocks:
maxItems: 1
clock-names:
const: sec

# Supply reference must be provided if QFPROM is writable from the OS image. vcc-supply:
description: Our power supply.

   # Needed if any child nodes are present.
   "#address-cells":
     const: 1
   "#size-cells":
     const: 1

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 26

Quoted 4 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+required:
+   - compatible
+   - reg
+   - reg-names
Quoted 6 lines

reg-names is discouraged. Please remove. I always point people here as a reference:

https://lore.kernel.org/r/CAL_Jsq+MMunmVWqeW9v2RyzsMKP+=kMzeTHNMG4JDHM7Fy0HBg@mail.gmail.com/

You can just figure out whether there are 3 register fields or 1 register field.

Am not sure if I understand this correctly, reg-names are very useful in this particular case as we are dealing with multiple memory ranges with holes. I agree with not having this for cases where we have only one resource.

But having the ordering in DT without proper names associated with it seems fragile to me! And it makes very difficult to debug issues with wrong resource ordering in DT.

Rob, Is this the guidance for new bindings?

I have not seen any strong suggestion or guidance either in

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/resource-names.txt?h=v5.8-rc1
Or in ./drivers/of/address.c

Am I missing anything here?

#7 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

Quoted 134 lines · Srinivas Kandagatla
Quoted 50 lines · Doug Anderson

Hi,

Quoted 7 lines · Ravi Kumar Bokka

This patch adds dt-bindings document for qfprom-efuse controller.

---
  .../devicetree/bindings/nvmem/qfprom.yaml          | 52 ++++++++++++++++++++++
  1 file changed, 52 insertions(+)

Overall comment: I reviewed your v1 series and so I'm obviously interested in your series. Please CC me on future versions.

I would also note that, since this is relevant to Qualcomm SoCs that you probably should be CCing "[email protected]" on your series.

Quoted 16 lines

create mode 100644 Documentation/devicetree/bindings/nvmem/qfprom.yaml

Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +8 -0
new file mode 100644
index 0000000..7c8fc31
@@ -0,0 +1,52 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/nvmem/qfprom.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Qualcomm Technologies Inc, QFPROM Efuse bindings
+

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 10 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +10 -0
+maintainers:
+  - Ravi Kumar Bokka <[email protected]>
+
+allOf:
+  - $ref: "nvmem.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

Quoted 11 lines

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom

Above is not required for now in this patchset, as we can attach data at runtime specific to version of the qfprom.

This can be added when required!

Quoted 62 lines

- const: qcom,qfprom

NOTE: old SoCs won't have both of these and thus they will get flagged with "dtbs_check", but I believe that's fine (Rob can correct me if I'm wrong). The code should still work OK if the SoC isn't there but it would be good to fix old dts files to have the SoC specific string too.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 21

Quoted 3 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +3 -0
+
+  reg:
+    maxItems: 3

Please address feedback feedback on v1. If you disagree with my feedback it's OK to say so (I make no claims of being always right), but silently ignoring my feedback and sending the next version doesn't make me feel like it's a good use of my time to keep reviewing your series. Specifically I suggested that you actually add descriptions rather than just putting "maxItems: 3".

With all that has been discussed, I think the current best thing to put there is:

# If the QFPROM is read-only OS image then only the corrected region # needs to be provided. If the QFPROM is writable then all 3 regions # must be provided.
oneOf:
- items:
- description: The start of the corrected region.
- items:
- description: The start of the raw region.
- description: The start of the config region.
- description: The start of the corrected region.

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 22

Quoted 1 line
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +1 -0
+

You missed a bunch of things that you should document:

# Clocks must be provided if QFPROM is writable from the OS image. clocks:
maxItems: 1
clock-names:
const: sec

# Supply reference must be provided if QFPROM is writable from the OS image. vcc-supply:
description: Our power supply.

   # Needed if any child nodes are present.
   "#address-cells":
     const: 1
   "#size-cells":
     const: 1

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 26

Quoted 4 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+required:
+   - compatible
+   - reg
+   - reg-names

reg-names is discouraged. Please remove. I always point people here as a reference:

Quoted 8 lines
Quoted 3 lines

https://lore.kernel.org/r/CAL_Jsq+MMunmVWqeW9v2RyzsMKP+=kMzeTHNMG4JDHM7Fy0HBg@mail.gmail.com/

You can just figure out whether there are 3 register fields or 1 register field.

Am not sure if I understand this correctly, reg-names are very useful in this particular case as we are dealing with multiple memory ranges with holes. I agree with not having this for cases where we have only one resource.

1 or 3 doesn't sound like a case with holes.

Quoted 5 lines

But having the ordering in DT without proper names associated with it seems fragile to me! And it makes very difficult to debug issues with wrong resource ordering in DT.

Rob, Is this the guidance for new bindings?

This has *always* been the guidance since *-names were added.

Quoted 3 lines

I have not seen any strong suggestion or guidance either in

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/resource-names.txt?h=v5.8-rc1

The key word is 'supplemental'. Perhaps that could be clearer, but DT always required a defined order and the supplemental information doesn't throw that out.

Quoted 1 line

Or in ./drivers/of/address.c

How could it? Order is defined by the specific binding.

Quoted 2 lines

Am I missing anything here?

#8 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

In reply to Srinivas Kandagatla

Hi,

Quoted 23 lines · Srinivas Kandagatla
Quoted 10 lines

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 5 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

Quoted 11 lines

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom

Quoted 4 lines

Above is not required for now in this patchset, as we can attach data at runtime specific to version of the qfprom.

This can be added when required!

I'm OK with leaving off for this patchset. After we get everything landed for blowing fuses then I can send out a separate patch where we can debate the merits of adding the SoC-specific compatible strings. :-) Sound good?

-Doug

#9 lore raw

Re: [RFC v2 1/3] dt-bindings: nvmem: Add devicetree bindings for qfprom-efuse

Quoted 30 lines · Doug Anderson

Hi,

Quoted 26 lines · Srinivas Kandagatla
Quoted 10 lines

Documentation/devicetree/bindings/nvmem/qfprom.yaml line 18

Quoted 5 lines
Documentation/devicetree/bindings/nvmem/qfprom.yaml new file +4 -0
+properties:
+  compatible:
+    enum:
+      - qcom,qfprom

As per discussion in patch #1, I believe SoC compatible should be here too in case it is ever needed. This is standard practice for dts files for IP blocks embedded in an SoC. AKA, this should be:

Quoted 11 lines

items:
- enum:
- qcom,apq8064-qfprom
- qcom,apq8084-qfprom
- qcom,msm8974-qfprom
- qcom,msm8916-qfprom
- qcom,msm8996-qfprom
- qcom,msm8998-qfprom
- qcom,qcs404-qfprom
- qcom,sc7180-qfprom
- qcom,sdm845-qfprom

Above is not required for now in this patchset, as we can attach data at runtime specific to version of the qfprom.

Quoted 6 lines
Quoted 1 line

This can be added when required!

I'm OK with leaving off for this patchset. After we get everything landed for blowing fuses then I can send out a separate patch where we can debate the merits of adding the SoC-specific compatible strings. :-) Sound good?

Sounds good!

--srini

Quoted 3 lines

-Doug

#10 lore raw

RFC v2 2/3 drivers: nvmem: Add QTI qfprom-efuse support

In reply to Ravi Kumar Bokka

This patch adds support for QTI qfprom-efuse controller. This driver can access the raw qfprom regions for fuse blowing.

The current existed qfprom driver is only supports for cpufreq, thermal sensors drivers by read out calibration data, speed bins..etc which is stored by qfprom efuses.

---
 drivers/nvmem/Kconfig  |   1 +
 drivers/nvmem/qfprom.c | 405 ++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 385 insertions(+), 21 deletions(-)
drivers/nvmem/Kconfig +1 -0
index d7b7f6d..623d59e 100644
@@ -117,6 +117,7 @@ config QCOM_QFPROM
 	help
 	  Say y here to enable QFPROM support. The QFPROM provides access
 	  functions for QFPROM data to rest of the drivers via nvmem interface.
+	  And this driver provides access QTI qfprom efuse via nvmem interface.
 
 	  This driver can also be built as a module. If so, the module
 	  will be called nvmem_qfprom.
drivers/nvmem/qfprom.c +384 -21

448 lines not on this page · raw message

Signature · 1 line
Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.
#11 lore raw

Re: [RFC v2 2/3] drivers: nvmem: Add QTI qfprom-efuse support

Hi,

Quoted 10 lines · Ravi Kumar Bokka

This patch adds support for QTI qfprom-efuse controller. This driver can access the raw qfprom regions for fuse blowing.

The current existed qfprom driver is only supports for cpufreq, thermal sensors drivers by read out calibration data, speed bins..etc which is stored by qfprom efuses.

---

It makes all your code reviewers much happier (and much more likely to take the time to review your patches) if you include a changelog with what changed from one version to the next. If you would like some help maintaining such a thing, might I suggest "patman":

https://gitlab.denx.de/u-boot/u-boot/-/blob/master/tools/patman/README

...pay no mind that it's hosted in the U-Boot repo--it's really quite a great tool.

drivers/nvmem/Kconfig config QCOM_QFPROM · line 120

Quoted 13 lines
 drivers/nvmem/Kconfig  |   1 +
 drivers/nvmem/qfprom.c | 405 ++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 385 insertions(+), 21 deletions(-)
drivers/nvmem/Kconfig +1 -0
index d7b7f6d..623d59e 100644
@@ -117,6 +117,7 @@ config QCOM_QFPROM
        help
          Say y here to enable QFPROM support. The QFPROM provides access
          functions for QFPROM data to rest of the drivers via nvmem interface.
+         And this driver provides access QTI qfprom efuse via nvmem interface.

I'm not sure it was necessary to add that line, but I won't object if you/others really like it.

drivers/nvmem/qfprom.c

Quoted 6 lines
drivers/nvmem/Kconfig +0 -0
          This driver can also be built as a module. If so, the module
          will be called nvmem_qfprom.
drivers/nvmem/qfprom.c +0 -0
index 8a91717..312318c 100644

You've still mostly not addressed most of the review feedback I've now given you 3 times. Rather than repeating comments, I have simply provided a patch that makes the driver into a state that I'm happy with:

https://crrev.com/c/2244932

Rough summary:

* There should be no reason to provide "reset" values for things. For anything that you change for fuse blowing, just save and restore after.

* Use the major/minor version read from 0x6000 to pick the parameters to use. Please double-check that I got this right.

* Reading should still read "corrected", not "raw". Added a sysfs knob to allow you to read "raw", though.

* Simplified the SoC data structure.

* No need for quite so many levels of abstraction for setting clocks / regulator.

* Don't set regulator voltage. Rely on device tree to make sure it's right.

* Properly undo things in the case of failure.

* Don't just keep enabling the regulator over and over again.

* Enable / disable the clock each time; now we don't need a .remove function and yet we still don't leave the clock enabled/prepared.

* Polling every 100 us but timing out in 10 us didn't make sense. Swap those. Also no reason for 100 us to be SoC specific.

* No need for reg-names.

* We shouldn't be creating two separate nvmem devices.

In general I'm happy to post my series to the list myself to get review feedback. For now I'm expecting that you can squash my changes in and send the next version.

-Doug

#12 lore raw

Re: [RFC v2 2/3] drivers: nvmem: Add QTI qfprom-efuse support

Quoted 29 lines · Doug Anderson

Hi,

Quoted 10 lines · Ravi Kumar Bokka

This patch adds support for QTI qfprom-efuse controller. This driver can access the raw qfprom regions for fuse blowing.

The current existed qfprom driver is only supports for cpufreq, thermal sensors drivers by read out calibration data, speed bins..etc which is stored by qfprom efuses.

---

It makes all your code reviewers much happier (and much more likely to take the time to review your patches) if you include a changelog with what changed from one version to the next. If you would like some help maintaining such a thing, might I suggest "patman":

https://gitlab.denx.de/u-boot/u-boot/-/blob/master/tools/patman/README

...pay no mind that it's hosted in the U-Boot repo--it's really quite a great tool.

Quoted 4 lines
  drivers/nvmem/Kconfig  |   1 +
  drivers/nvmem/qfprom.c | 405 ++++++++++++++++++++++++++++++++++++++++++++++---
  2 files changed, 385 insertions(+), 21 deletions(-)
Quoted 13 lines

drivers/nvmem/Kconfig config QCOM_QFPROM · line 120

Quoted 9 lines
drivers/nvmem/Kconfig +1 -0
index d7b7f6d..623d59e 100644
@@ -117,6 +117,7 @@ config QCOM_QFPROM
         help
           Say y here to enable QFPROM support. The QFPROM provides access
           functions for QFPROM data to rest of the drivers via nvmem interface.
+         And this driver provides access QTI qfprom efuse via nvmem interface.

I'm not sure it was necessary to add that line, but I won't object if you/others really like it.

NAK from my side!

Quoted 21 lines

drivers/nvmem/qfprom.c

Quoted 7 lines
drivers/nvmem/Kconfig +0 -0
           This driver can also be built as a module. If so, the module
           will be called nvmem_qfprom.
drivers/nvmem/qfprom.c +0 -0
index 8a91717..312318c 100644

You've still mostly not addressed most of the review feedback I've now given you 3 times. Rather than repeating comments, I have simply provided a patch that makes the driver into a state that I'm happy with:

https://crrev.com/c/2244932

Rough summary:

* There should be no reason to provide "reset" values for things. For anything that you change for fuse blowing, just save and restore after.

Quoted 5 lines

* Use the major/minor version read from 0x6000 to pick the parameters to use. Please double-check that I got this right.

* Reading should still read "corrected", not "raw". Added a sysfs knob to allow you to read "raw", though.

We could create an additional nvmem read-only provider in future if required to read raw!.

Quoted 30 lines

* Simplified the SoC data structure.

* No need for quite so many levels of abstraction for setting clocks / regulator.

* Don't set regulator voltage. Rely on device tree to make sure it's right.

* Properly undo things in the case of failure.

* Don't just keep enabling the regulator over and over again.

* Enable / disable the clock each time; now we don't need a .remove function and yet we still don't leave the clock enabled/prepared.

* Polling every 100 us but timing out in 10 us didn't make sense. Swap those. Also no reason for 100 us to be SoC specific.

* No need for reg-names.

* We shouldn't be creating two separate nvmem devices.

In general I'm happy to post my series to the list myself to get review feedback. For now I'm expecting that you can squash my changes in and send the next version.

-Doug

#13 lore raw

Re: [RFC v2 2/3] drivers: nvmem: Add QTI qfprom-efuse support

Hi,

Quoted 6 lines · Srinivas Kandagatla
Quoted 3 lines

* Reading should still read "corrected", not "raw". Added a sysfs knob to allow you to read "raw", though.

We could create an additional nvmem read-only provider in future if required to read raw!.

Makes sense. For now module parameter (which can be tweaked from sysfs) seemed like at least an easy to get access to the raw values for testing. Other than for checking that the driver works OK, though, I'm under the impression that you should never read the raw values back.

-Doug

#14 lore raw

Re: [RFC v2 2/3] drivers: nvmem: Add QTI qfprom-efuse support

In reply to Doug Anderson

Hi,

Quoted 4 lines · Doug Anderson

In general I'm happy to post my series to the list myself to get review feedback. For now I'm expecting that you can squash my changes in and send the next version.

I talked with Ravi offline and he suggested that I send out the v3, so it's been posted. For the cover letter, see:

https://lore.kernel.org/r/[email protected]

-Doug

#15 lore raw

Re: [RFC v2 2/3] drivers: nvmem: Add QTI qfprom-efuse support

In reply to Ravi Kumar Bokka

I think Doug already covered most of the comments and his fixes seems be in right direction.

drivers/nvmem/qfprom.c static int qfprom_reg_read(void *context, · line 323

Quoted 4 lines · Ravi Kumar Bokka
drivers/nvmem/qfprom.c +4 -0
+	qfpraw = platform_get_resource_byname(pdev, IORESOURCE_MEM, "raw");
+
+	priv->qfpraw = devm_ioremap_resource(dev, qfpraw);
+	if (IS_ERR(priv->qfpraw)) {

General comment for up-streaming is that your patch should not break whats in mainline, Your patch is totally ignoring!! Please be mindful while doing changes to drivers which are used by other platforms.

--srini

Quoted 20 lines
drivers/nvmem/qfprom.c +20 -0
+		ret = PTR_ERR(priv->qfpraw);
+		goto err;
+	}
+
+	qfpconf = platform_get_resource_byname(pdev, IORESOURCE_MEM, "conf");
+
+	priv->qfpconf = devm_ioremap_resource(dev, qfpconf);
+	if (IS_ERR(priv->qfpconf)) {
+		ret = PTR_ERR(priv->qfpconf);
+		goto err;
+	}
+
+	qfpcorrected = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						    "corrected");
+
+	priv->qfpcorrected = devm_ioremap_resource(dev, qfpcorrected);
+	if (IS_ERR(priv->qfpcorrected)) {
+		ret = PTR_ERR(priv->qfpcorrected);
+		goto err;
+	}
#16 lore raw

RFC v2 3/3 arm64: dts: qcom: sc7180: Add qfprom-efuse

In reply to Ravi Kumar Bokka

This patch adds device tree node for qfprom-efuse controller.

---
 arch/arm64/boot/dts/qcom/sc7180-idp.dts |  4 ++++
 arch/arm64/boot/dts/qcom/sc7180.dtsi    | 10 ++++++++--
 2 files changed, 12 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sc7180-idp.dts +4 -0
index 4e9149d..2a9224e 100644
@@ -287,6 +287,10 @@
 	};
 };
 
+&qfprom {
+	vcc-supply = <&vreg_l11a_1p8>;
+};
+
 &qspi {
 	status = "okay";
 	pinctrl-names = "default";
arch/arm64/boot/dts/qcom/sc7180.dtsi +8 -2
index 31b9217..20f3480 100644
@@ -498,9 +498,15 @@
 			#power-domain-cells = <1>;
 		};
 
-		qfprom@784000 {
+		qfprom: qfprom@780000 {
 			compatible = "qcom,qfprom";
-			reg = <0 0x00784000 0 0x8ff>;
+			reg = <0 0x00780000 0 0x7a0>,
+			      <0 0x00782000 0 0x100>,
+			      <0 0x00784000 0 0x8ff>;
+			reg-names = "raw", "conf", "corrected";
+			
+			clocks = <&gcc GCC_SEC_CTRL_CLK_SRC>;
+			clock-names = "secclk";
 			#address-cells = <1>;
 			#size-cells = <1>;
 
Signature · 1 line
Qualcomm INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by the Linux Foundation.
#17 lore raw

Re: [RFC v2 3/3] arm64: dts: qcom: sc7180: Add qfprom-efuse

Hi,

Quoted 29 lines · Ravi Kumar Bokka

This patch adds device tree node for qfprom-efuse controller.

---
 arch/arm64/boot/dts/qcom/sc7180-idp.dts |  4 ++++
 arch/arm64/boot/dts/qcom/sc7180.dtsi    | 10 ++++++++--
 2 files changed, 12 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/qcom/sc7180-idp.dts +4 -0
index 4e9149d..2a9224e 100644
@@ -287,6 +287,10 @@
        };
 };
 
+&qfprom {
+       vcc-supply = <&vreg_l11a_1p8>;
+};
+
 &qspi {
        status = "okay";
        pinctrl-names = "default";
arch/arm64/boot/dts/qcom/sc7180.dtsi +0 -0
index 31b9217..20f3480 100644

arch/arm64/boot/dts/qcom/sc7180.dtsi line 501

Quoted 5 lines
arch/arm64/boot/dts/qcom/sc7180.dtsi +1 -1
@@ -498,9 +498,15 @@
                        #power-domain-cells = <1>;
                };
 
-               qfprom@784000 {
+               qfprom: qfprom@780000 {

The dt schema checker claims that your node should actually be called:

eeprom|efuse|nvram

So you probably want the above to actually be:

qfprom: efuse@780000

The label to the side doesn't matter and so it can stay qfprom--just the node name itself is what the checker cares about.

arch/arm64/boot/dts/qcom/sc7180.dtsi line 502

Quoted 1 line
arch/arm64/boot/dts/qcom/sc7180.dtsi +0 -0
                        compatible = "qcom,qfprom";

As per my response in the bindings, this should be:

"qcom,sc7180-qfprom", "qcom,qfprom"

...even if the driver only ever makes use of "qcom,qfprom" this future-proofs us a bit.

arch/arm64/boot/dts/qcom/sc7180.dtsi line 506

Quoted 5 lines
arch/arm64/boot/dts/qcom/sc7180.dtsi +4 -1
-                       reg = <0 0x00784000 0 0x8ff>;
+                       reg = <0 0x00780000 0 0x7a0>,
+                             <0 0x00782000 0 0x100>,
+                             <0 0x00784000 0 0x8ff>;
+                       reg-names = "raw", "conf", "corrected";

As per my response in the bindings, reg-names is discouraged so you should remove and make it so that the driver doesn't need.

arch/arm64/boot/dts/qcom/sc7180.dtsi line 507

Quoted 1 line
arch/arm64/boot/dts/qcom/sc7180.dtsi +1 -0
+

It's hard to tell in email, but checkpatch yells above the above line:

    ERROR: trailing whitespace
    #53: FILE: arch/arm64/boot/dts/qcom/sc7180.dtsi:512:
    +^I^I^I$

arch/arm64/boot/dts/qcom/sc7180.dtsi line 509

Quoted 2 lines
arch/arm64/boot/dts/qcom/sc7180.dtsi +2 -0
+                       clocks = <&gcc GCC_SEC_CTRL_CLK_SRC>;
+                       clock-names = "secclk";

As per the binding spec, clock name shouldn't end in "clk".

For your edification, I provided a patch to fix all my review feedback at:

https://crrev.com/c/2244933

Feel free to squash it into your next version.

-Doug