Skip to contentSkip to conversation

Fix emmc hs400 mode on mt7986

PATCH Superseded Frank Wunderlich 3 participants 7 messages Jun 25, 2023 – Jun 27, 2023

Subsystems (5)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover Fix emmc hs400 mode on mt7986

From: Frank Wunderlich <[email protected]>

Eric boots linux directly from ATF without using uboot. In this corner case clocks for emmc hs400 mode are not correctly setup.

Fix this by adding assigned-clocks and assigned.clock-parents to mmc node.

Eric Woudstra (1):
arm64: dts: mt7986: fix emmc hs400 mode without uboot initialization

Frank Wunderlich (1):
dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for

    mt7986
 Documentation/devicetree/bindings/mmc/mtk-sd.yaml | 4 ++++
 arch/arm64/boot/dts/mediatek/mt7986a.dtsi         | 4 ++++
 2 files changed, 8 insertions(+)
Signature · 1 line
2.34.1
#1 lore raw

PATCH 1/2 dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

From: Frank Wunderlich <[email protected]>

MT7986 has 2 clock-parents so update the binding for it.

---
 Documentation/devicetree/bindings/mmc/mtk-sd.yaml | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/devicetree/bindings/mmc/mtk-sd.yaml +4 -0
index 46eefdd19a2c..485516dcc759 100644
@@ -247,6 +247,10 @@ allOf:
               - mediatek,mt7986-mmc
     then:
       properties:
+        assigned-clocks:
+          maxItems: 2
+        assigned-clock-parents:
+          maxItems: 2
         clocks:
           minItems: 3
           items:
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 3 lines · Frank Wunderlich

From: Frank Wunderlich <[email protected]>

MT7986 has 2 clock-parents so update the binding for it.

You didn't test it, I think. If you do, then you will see errors from other trees.

Anyway, I don't understand why defining it in the first place. Just drop the assigned-clock* from the binding.

Best regards,
Krzysztof

#3 lore raw

Aw: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 11 lines

Gesendet: Dienstag, 27. Juni 2023 um 12:44 Uhr
Von: "Krzysztof Kozlowski" <[email protected]>
Betreff: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 3 lines · Frank Wunderlich

From: Frank Wunderlich <[email protected]>

MT7986 has 2 clock-parents so update the binding for it.

You didn't test it, I think. If you do, then you will see errors from other trees.

Hi,

i tested it of course...which errors do you see?

this is basicly how i tested it (in case anything has changed):

logfile=dtbs_arm64.log
exec 3> >(tee $logfile)
ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make DT_CHECKER_FLAGS=-m dt_binding_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 binding check failed!";cat $logfile;exit 1;fi ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make defconfig #dtbs_check need kernel-config ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j8 DT_CHECKER_FLAGS=-m dtbs_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 dtbs_check failed!";cat $logfile;exit 1;fi

and looked into the resulting logfile for keywords like mmc like mtk-sd

i tried running dtbs_check with passing the yaml-file, but of course all compatibles not matching this file were reported.

ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j6 DT_CHECKER_FLAGS=-m dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/mmc/mtk-sd.yaml

but this spits out many errors "failed to match any schema with compatible" because i defined only the changed one...

maybe there is another way to check only one yaml file against all dtbs without these unrelated errors.

pipeline in dt-bindings-patchwork is clean too
https://patchwork.ozlabs.org/project/devicetree-bindings/patch/[email protected]/

Quoted 2 lines

Anyway, I don't understand why defining it in the first place. Just drop the assigned-clock* from the binding.

as it was defined (not looked where it was used) i only used the soc-specific branch to update the MaxItems...just to not break anything. After that the message i got before was fixed

arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clocks: [[4, 35], [4, 34]] is too long arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clock-parents: [[5, 6], [4, 18]] is too long

but if the right way is to drop the MaxItems from generic (or the property itself - where is it taken from then?). The only include i see is Documentation/devicetree/bindings/mmc/mmc-controller.yaml and there the assigned-clock* is not defined. And the

unevaluatedProperties: false

tells me that removing the generic one will result in unevaluated propety error.

regards Frank

Quoted 4 lines

Best regards,
Krzysztof

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#4 lore raw

Re: Aw: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 7 lines · Frank Wunderlich
Quoted 7 lines

Gesendet: Dienstag, 27. Juni 2023 um 12:44 Uhr
Von: "Krzysztof Kozlowski" <[email protected]>
Betreff: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 2 lines · Frank Wunderlich

From: Frank Wunderlich <[email protected]>

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

MT7986 has 2 clock-parents so update the binding for it.

You didn't test it, I think. If you do, then you will see errors from other trees.

Hi,

i tested it of course...which errors do you see?

The top-level said it can be maximum 1, so raising it in allOf:if:then:
should not be enough

Quoted 10 lines

11
this is basicly how i tested it (in case anything has changed):

logfile=dtbs_arm64.log
exec 3> >(tee $logfile)
ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make DT_CHECKER_FLAGS=-m dt_binding_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 binding check failed!";cat $logfile;exit 1;fi ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make defconfig #dtbs_check need kernel-config ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j8 DT_CHECKER_FLAGS=-m dtbs_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 dtbs_check failed!";cat $logfile;exit 1;fi

A bit over-complicated... why not running dtbs_check against the schema you changed?

Quoted 5 lines

and looked into the resulting logfile for keywords like mmc like mtk-sd

i tried running dtbs_check with passing the yaml-file, but of course all compatibles not matching this file were reported.

Quoted 8 lines

ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j6 DT_CHECKER_FLAGS=-m dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/mmc/mtk-sd.yaml

but this spits out many errors "failed to match any schema with compatible" because i defined only the changed one...

maybe there is another way to check only one yaml file against all dtbs without these unrelated errors.

pipeline in dt-bindings-patchwork is clean too
https://patchwork.ozlabs.org/project/devicetree-bindings/patch/[email protected]/

Maybe that binding just fails to apply to DTS because of missing or not correct compatibles.

Quoted 11 lines
Quoted 3 lines

Anyway, I don't understand why defining it in the first place. Just drop the assigned-clock* from the binding.

as it was defined (not looked where it was used) i only used the soc-specific branch to update the MaxItems...just to not break anything. After that the message i got before was fixed

arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clocks: [[4, 35], [4, 34]] is too long arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clock-parents: [[5, 6], [4, 18]] is too long

but if the right way is to drop the MaxItems from generic (or the property itself - where is it taken from then?). The only include i see is Documentation/devicetree/bindings/mmc/mmc-controller.yaml and there the assigned-clock* is not defined. And the

The way is to entirely drop assigned-clocks. I don't think there are much benefits of having them in the bindings. Maybe if specific rates are required, then yes - device cannot work with other rates and you can verify it with dtbs_check. But otherwise it is like adding values of 'reg' or 'interrupts'. Plus some board might require third item and then what?

Best regards,
Krzysztof

#5 lore raw

Aw: Re: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Hi

Quoted 15 lines

Gesendet: Dienstag, 27. Juni 2023 um 15:37 Uhr
Von: "Krzysztof Kozlowski" <[email protected]>

Quoted 12 lines · Frank Wunderlich
Quoted 11 lines

Gesendet: Dienstag, 27. Juni 2023 um 12:44 Uhr
Von: "Krzysztof Kozlowski" <[email protected]>
Betreff: Re: [PATCH v1 1/2] dt-bindings: mmc: mtk-sd: update assigned-clocks/clock-parents for mt7986

Quoted 3 lines · Frank Wunderlich

From: Frank Wunderlich <[email protected]>

MT7986 has 2 clock-parents so update the binding for it.

You didn't test it, I think. If you do, then you will see errors from other trees.

Quoted 6 lines
Quoted 3 lines

Hi,

i tested it of course...which errors do you see?

The top-level said it can be maximum 1, so raising it in allOf:if:then:
should not be enough

but the previous error with too long list wasn't there after the change...so it seems to work...if it right, i don't know, but other properties are overriddedn the same way.

Quoted 11 lines
Quoted 10 lines

11
this is basicly how i tested it (in case anything has changed):

logfile=dtbs_arm64.log
exec 3> >(tee $logfile)
ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make DT_CHECKER_FLAGS=-m dt_binding_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 binding check failed!";cat $logfile;exit 1;fi ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make defconfig #dtbs_check need kernel-config ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j8 DT_CHECKER_FLAGS=-m dtbs_check 2>&3 if [[ $? -ne 0 ]];then echo "arm64 dtbs_check failed!";cat $logfile;exit 1;fi

Quoted 2 lines

A bit over-complicated... why not running dtbs_check against the schema you changed?

it is a script i only start ;) and having a logfile (with all messages - stdin+stdout) makes it easier to find something. i had always the problem, that second run does not show the errors again without changing the yaml or dts.

as you see below i tried the run against the yaml too, but there are many many unrelated compatibles which seem to be not documented or still in txt bindings.

Quoted 8 lines
Quoted 8 lines

and looked into the resulting logfile for keywords like mmc like mtk-sd

i tried running dtbs_check with passing the yaml-file, but of course all compatibles not matching this file were reported.

ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j6 DT_CHECKER_FLAGS=-m dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/mmc/mtk-sd.yaml

but this spits out many errors "failed to match any schema with compatible" because i defined only the changed one...

Quoted 7 lines
Quoted 4 lines

maybe there is another way to check only one yaml file against all dtbs without these unrelated errors.

pipeline in dt-bindings-patchwork is clean too
https://patchwork.ozlabs.org/project/devicetree-bindings/patch/[email protected]/

Maybe that binding just fails to apply to DTS because of missing or not correct compatibles.

the messages are unrelated to current yaml file i passed with the DT_SCHEMA_FILES parameter, e.g. the last messages:

From schema: Documentation/devicetree/bindings/display/mediatek/mediatek,dsi.yaml
arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@15020000: failed to match any schema with compatible: ['mediatek,mt8183-imgsys', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@16000000: failed to match any schema with compatible: ['mediatek,mt8183-vdecsys', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@17000000: failed to match any schema with compatible: ['mediatek,mt8183-vencsys', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@19000000: failed to match any schema with compatible: ['mediatek,mt8183-ipu_conn', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@19010000: failed to match any schema with compatible: ['mediatek,mt8183-ipu_adl', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@19180000: failed to match any schema with compatible: ['mediatek,mt8183-ipu_core0', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@19280000: failed to match any schema with compatible: ['mediatek,mt8183-ipu_core1', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /soc/syscon@1a000000: failed to match any schema with compatible: ['mediatek,mt8183-camsys', 'syscon'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb:0:0: /mt8183-sound: failed to match any schema with compatible: ['mediatek,mt8183_mt6358_ts3a227_max98357'] arch/arm64/boot/dts/mediatek/mt8183-kukui-jacuzzi-burnet.dtb: bt-sco: '#sound-dai-cells' is a required property

Quoted 11 lines
Quoted 10 lines
Quoted 2 lines

Anyway, I don't understand why defining it in the first place. Just drop the assigned-clock* from the binding.

as it was defined (not looked where it was used) i only used the soc-specific branch to update the MaxItems...just to not break anything. After that the message i got before was fixed

arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clocks: [[4, 35], [4, 34]] is too long arch/arm64/boot/dts/mediatek/mt7986a-bananapi-bpi-r3.dtb: mmc@11230000: assigned-clock-parents: [[5, 6], [4, 18]] is too long

but if the right way is to drop the MaxItems from generic (or the property itself - where is it taken from then?). The only include i see is Documentation/devicetree/bindings/mmc/mmc-controller.yaml and there the assigned-clock* is not defined. And the

Quoted 6 lines

The way is to entirely drop assigned-clocks. I don't think there are much benefits of having them in the bindings. Maybe if specific rates are required, then yes - device cannot work with other rates and you can verify it with dtbs_check. But otherwise it is like adding values of 'reg' or 'interrupts'. Plus some board might require third item and then what?

mhm, disabled all assigned-clock* properties and do not get any messages about this (don't know why because of "unevaluatedProperties: false", but looks like it works)

there are only some syscon-nodes in mt8173-*.dtb where these properties used are reported

btw. need binding also be backported to stable? so do i need to add the tag there?

regards Frank

#6 lore raw

PATCH 2/2 arm64: dts: mt7986: fix emmc hs400 mode without uboot initialization

In reply to Frank Wunderlich

From: Eric Woudstra <[email protected]>

Eric reports errors on emmc with hs400 mode when booting linux on bpi-r3 without uboot [1]. Booting with uboot does not show this because clocks seem to be initialized by uboot.

Fix this by adding assigned-clocks and assigned-clock-parents like it's done in uboot [2].

[1] https://forum.banana-pi.org/t/bpi-r3-kernel-fails-setting-emmc-clock-to-416m-depends-on-u-boot/15170
[2] https://github.com/u-boot/u-boot/blob/master/arch/arm/dts/mt7986.dtsi#L287

Fixes 513b49d19b34 ("arm64: dts: mt7986: add mmc related device nodes")
---
 arch/arm64/boot/dts/mediatek/mt7986a.dtsi | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/boot/dts/mediatek/mt7986a.dtsi +4 -0
index 68539ea788df..207510abda89 100644
@@ -374,6 +374,10 @@ mmc0: mmc@11230000 {
 			reg = <0 0x11230000 0 0x1000>,
 			      <0 0x11c20000 0 0x1000>;
 			interrupts = <GIC_SPI 143 IRQ_TYPE_LEVEL_HIGH>;
+			assigned-clocks = <&topckgen CLK_TOP_EMMC_416M_SEL>,
+					  <&topckgen CLK_TOP_EMMC_250M_SEL>;
+			assigned-clock-parents = <&apmixedsys CLK_APMIXED_MPLL>,
+						 <&topckgen CLK_TOP_NET1PLL_D5_D2>;
 			clocks = <&topckgen CLK_TOP_EMMC_416M_SEL>,
 				 <&infracfg CLK_INFRA_MSDC_HCK_CK>,
 				 <&infracfg CLK_INFRA_MSDC_CK>,
Signature · 1 line
2.34.1