[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[christopher.spinrath at rwth-aachen.de]

From: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>

The cm-fx6 module has an on-board AW-NH387 WiFi/BT module which is
based on Marvell's SD8787 chip and is connected to the usdhc1
controller.

Unfortunately, the chip gets unresponsive if the Bluetooth AMP
(Alternate Mac/Phy) function gets probed but the loaded firmware
doesn't support it. For instance, this is the case for the most
recent firmware in linux-firmware (Version 14.66.35.p52).

Thus, just add the required nodes but leave the usdhc1 node disabled
explicitly. Users who disabled the Bluetooth (AMP) support of their
OS can then conveniently enable WiFi (or even plain Bluetooth) support
with a simple device tree overlay/bootloader configuration.

Signed-off-by: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
---

Hi all,

some more information on the Bluetooth AMP dilemma: the vendor (CompuLab) just
remove Bluetooth AMP support from the kernel. However, this is not an option
here, since it would introduce a regression for its actual users.

Neither is degrading the firmware a good option: recent firmare (without AMP
support) provides WiFi that is more stable (in particular, for access points).

I wrote a RFC to linux-bluetooth [1] two weeks ago about this problem and the
only feedback I got was that Marvell has to provide information about how to
detect whether the firmware supports AMP properly -- but this is unlikely given
Marvell's information policy (in particular, since the sd8787 seems to be
unsupported as of now).

IMHO this patch is currently the best option, since it allows users in a
convenient way to enable WiFi support (e.g. no pinctrl handling in overlays),
even when they are using distribution kernels/device trees. Furthermore, the
hardware description is entirely accurate -- we have a driver/firmware problem
here.

Thanks,
Christopher

[1] https://www.spinics.net/lists/linux-bluetooth/msg70493.html

 arch/arm/boot/dts/imx6q-cm-fx6.dts | 41 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)

diff --git a/arch/arm/boot/dts/imx6q-cm-fx6.dts b/arch/arm/boot/dts/imx6q-cm-fx6.dts
index a8af382..fe6ab0a 100644
--- a/arch/arm/boot/dts/imx6q-cm-fx6.dts
+++ b/arch/arm/boot/dts/imx6q-cm-fx6.dts
@@ -64,6 +64,14 @@
 		};
 	};
 
+	awnh387_pwrseq: pwrseq {
+		pinctrl-names = "default";
+		pinctrl-0 = <&pinctrl_pwrseq>;
+		compatible = "mmc-pwrseq-sd8787";
+		powerdown-gpios = <&gpio7 12 GPIO_ACTIVE_HIGH>;
+		reset-gpios = <&gpio6 16 GPIO_ACTIVE_HIGH>;
+	};
+
 	reg_pcie_power_on_gpio: regulator-pcie-power-on-gpio {
 		compatible = "regulator-fixed";
 		regulator-name = "regulator-pcie-power-on-gpio";
@@ -304,6 +312,13 @@
 		>;
 	};
 
+	pinctrl_pwrseq: pwrseqgrp {
+		fsl,pins = <
+			MX6QDL_PAD_GPIO_17__GPIO7_IO12		0x1b0b0
+			MX6QDL_PAD_NANDF_CS3__GPIO6_IO16	0x1b0b0
+		>;
+	};
+
 	pinctrl_spdif: spdifgrp {
 		fsl,pins = <
 			MX6QDL_PAD_GPIO_16__SPDIF_IN  0x1b0b0
@@ -330,6 +345,17 @@
 			MX6QDL_PAD_EIM_D22__GPIO3_IO22	0x130b0
 		>;
 	};
+
+	pinctrl_usdhc1: usdhc1grp {
+		fsl,pins = <
+			MX6QDL_PAD_SD1_CMD__SD1_CMD	0x17071
+			MX6QDL_PAD_SD1_CLK__SD1_CLK	0x10071
+			MX6QDL_PAD_SD1_DAT0__SD1_DATA0	0x17071
+			MX6QDL_PAD_SD1_DAT1__SD1_DATA1	0x17071
+			MX6QDL_PAD_SD1_DAT2__SD1_DATA2	0x17071
+			MX6QDL_PAD_SD1_DAT3__SD1_DATA3	0x17071
+		>;
+	};
 };
 
 &pcie {
@@ -382,3 +408,18 @@
 	dr_mode = "otg";
 	status = "okay";
 };
+
+&usdhc1 {
+	pinctrl-names = "default";
+	pinctrl-0 = <&pinctrl_usdhc1>;
+	mmc-pwrseq = <&awnh387_pwrseq>;
+	non-removable;
+	/*
+	 * If the OS probes the Bluetooth AMP function advertised on this bus
+	 * but the firmware in place does not support it, the WiFi/BT module
+	 * gets unresponsive.
+	 * Users who configured their OS properly can enable this node to gain
+	 * WiFi and/or plain Bluetooth support.
+	 */
+	status = "disabled";
+};
-- 
2.10.2

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Fabio Estevam]

Hi Christopher,

On Tue, May 30, 2017 at 5:16 PM,  <christopher.spinrath@rwth-aachen.de> wrote:

> +       awnh387_pwrseq: pwrseq {
> +               pinctrl-names = "default";
> +               pinctrl-0 = <&pinctrl_pwrseq>;
> +               compatible = "mmc-pwrseq-sd8787";
> +               powerdown-gpios = <&gpio7 12 GPIO_ACTIVE_HIGH>;
> +               reset-gpios = <&gpio6 16 GPIO_ACTIVE_HIGH>;

At least in the AW-NH387 datasheet I found it says:

RESETn: Reset(active low at least 10ns)

Please double check the polarity in device tree.

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Christopher Spinrath]

Hi Fabio,

thanks for the fast reply.

On 05/30/2017 10:40 PM, Fabio Estevam wrote:
> Hi Christopher,
> 
> On Tue, May 30, 2017 at 5:16 PM,  <christopher.spinrath@rwth-aachen.de> wrote:
> 
>> +       awnh387_pwrseq: pwrseq {
>> +               pinctrl-names = "default";
>> +               pinctrl-0 = <&pinctrl_pwrseq>;
>> +               compatible = "mmc-pwrseq-sd8787";
>> +               powerdown-gpios = <&gpio7 12 GPIO_ACTIVE_HIGH>;
>> +               reset-gpios = <&gpio6 16 GPIO_ACTIVE_HIGH>;
> 
> At least in the AW-NH387 datasheet I found it says:
> 
> RESETn: Reset(active low at least 10ns)

The BSP device tree uses an active-low regulator called nreset, too, but 
strangly it has a startup delay of 10ms while the powerdown-regulator 
"starts up" immediately.

(btw: the "regulator setup" is not able to handle resets correctly.)

It seems to me that the pwrseq-sd8787 uses inverted logic: it first sets 
the -- in our case -- active-high gpio to high, waits, and then enables 
the powerdown-gpio.

This coincides with the gpio being called "reset" in the pwrseq binding, 
neither "nreset" nor "RESETn" (beware the missing *n*ot).

> Please double check the polarity in device tree.
> 

I just double checked: high/high for powerdown/reset gpios is the only 
of the four possible combinations that works.

So, all in all, I think the polarity is correct (or either the pwrseq 
driver or hardware invertes the polarity for some reason).

Cheers,
Christopher

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Shawn Guo]

On Tue, May 30, 2017 at 10:16:01PM +0200, christopher.spinrath at rwth-aachen.de wrote:
> From: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
> 
> The cm-fx6 module has an on-board AW-NH387 WiFi/BT module which is
> based on Marvell's SD8787 chip and is connected to the usdhc1
> controller.
> 
> Unfortunately, the chip gets unresponsive if the Bluetooth AMP
> (Alternate Mac/Phy) function gets probed but the loaded firmware
> doesn't support it. For instance, this is the case for the most
> recent firmware in linux-firmware (Version 14.66.35.p52).
> 
> Thus, just add the required nodes but leave the usdhc1 node disabled
> explicitly. Users who disabled the Bluetooth (AMP) support of their
> OS can then conveniently enable WiFi (or even plain Bluetooth) support
> with a simple device tree overlay/bootloader configuration.
> 
> Signed-off-by: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
> ---
> 
> Hi all,
> 
> some more information on the Bluetooth AMP dilemma: the vendor (CompuLab) just
> remove Bluetooth AMP support from the kernel. However, this is not an option
> here, since it would introduce a regression for its actual users.
> 
> Neither is degrading the firmware a good option: recent firmare (without AMP
> support) provides WiFi that is more stable (in particular, for access points).
> 
> I wrote a RFC to linux-bluetooth [1] two weeks ago about this problem and the
> only feedback I got was that Marvell has to provide information about how to
> detect whether the firmware supports AMP properly -- but this is unlikely given
> Marvell's information policy (in particular, since the sd8787 seems to be
> unsupported as of now).
> 
> IMHO this patch is currently the best option, since it allows users in a
> convenient way to enable WiFi support (e.g. no pinctrl handling in overlays),
> even when they are using distribution kernels/device trees. Furthermore, the
> hardware description is entirely accurate -- we have a driver/firmware problem
> here.

Hi Valentin, Igor,

Are you guys fine with the patch?

Shawn

> 
> Thanks,
> Christopher
> 
> [1] https://www.spinics.net/lists/linux-bluetooth/msg70493.html
> 
>  arch/arm/boot/dts/imx6q-cm-fx6.dts | 41 ++++++++++++++++++++++++++++++++++++++
>  1 file changed, 41 insertions(+)
> 
> diff --git a/arch/arm/boot/dts/imx6q-cm-fx6.dts b/arch/arm/boot/dts/imx6q-cm-fx6.dts
> index a8af382..fe6ab0a 100644
> --- a/arch/arm/boot/dts/imx6q-cm-fx6.dts
> +++ b/arch/arm/boot/dts/imx6q-cm-fx6.dts
> @@ -64,6 +64,14 @@
>  		};
>  	};
>  
> +	awnh387_pwrseq: pwrseq {
> +		pinctrl-names = "default";
> +		pinctrl-0 = <&pinctrl_pwrseq>;
> +		compatible = "mmc-pwrseq-sd8787";
> +		powerdown-gpios = <&gpio7 12 GPIO_ACTIVE_HIGH>;
> +		reset-gpios = <&gpio6 16 GPIO_ACTIVE_HIGH>;
> +	};
> +
>  	reg_pcie_power_on_gpio: regulator-pcie-power-on-gpio {
>  		compatible = "regulator-fixed";
>  		regulator-name = "regulator-pcie-power-on-gpio";
> @@ -304,6 +312,13 @@
>  		>;
>  	};
>  
> +	pinctrl_pwrseq: pwrseqgrp {
> +		fsl,pins = <
> +			MX6QDL_PAD_GPIO_17__GPIO7_IO12		0x1b0b0
> +			MX6QDL_PAD_NANDF_CS3__GPIO6_IO16	0x1b0b0
> +		>;
> +	};
> +
>  	pinctrl_spdif: spdifgrp {
>  		fsl,pins = <
>  			MX6QDL_PAD_GPIO_16__SPDIF_IN  0x1b0b0
> @@ -330,6 +345,17 @@
>  			MX6QDL_PAD_EIM_D22__GPIO3_IO22	0x130b0
>  		>;
>  	};
> +
> +	pinctrl_usdhc1: usdhc1grp {
> +		fsl,pins = <
> +			MX6QDL_PAD_SD1_CMD__SD1_CMD	0x17071
> +			MX6QDL_PAD_SD1_CLK__SD1_CLK	0x10071
> +			MX6QDL_PAD_SD1_DAT0__SD1_DATA0	0x17071
> +			MX6QDL_PAD_SD1_DAT1__SD1_DATA1	0x17071
> +			MX6QDL_PAD_SD1_DAT2__SD1_DATA2	0x17071
> +			MX6QDL_PAD_SD1_DAT3__SD1_DATA3	0x17071
> +		>;
> +	};
>  };
>  
>  &pcie {
> @@ -382,3 +408,18 @@
>  	dr_mode = "otg";
>  	status = "okay";
>  };
> +
> +&usdhc1 {
> +	pinctrl-names = "default";
> +	pinctrl-0 = <&pinctrl_usdhc1>;
> +	mmc-pwrseq = <&awnh387_pwrseq>;
> +	non-removable;
> +	/*
> +	 * If the OS probes the Bluetooth AMP function advertised on this bus
> +	 * but the firmware in place does not support it, the WiFi/BT module
> +	 * gets unresponsive.
> +	 * Users who configured their OS properly can enable this node to gain
> +	 * WiFi and/or plain Bluetooth support.
> +	 */
> +	status = "disabled";
> +};
> -- 
> 2.10.2
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Christopher Spinrath]

Hi Shawn,

On 06/05/2017 04:51 PM, Shawn Guo wrote:
> On Tue, May 30, 2017 at 10:16:01PM +0200, christopher.spinrath at rwth-aachen.de wrote:
>> From: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
>>
>> The cm-fx6 module has an on-board AW-NH387 WiFi/BT module which is
>> based on Marvell's SD8787 chip and is connected to the usdhc1
>> controller.
>>
>> Unfortunately, the chip gets unresponsive if the Bluetooth AMP
>> (Alternate Mac/Phy) function gets probed but the loaded firmware
>> doesn't support it. For instance, this is the case for the most
>> recent firmware in linux-firmware (Version 14.66.35.p52).
>>
>> Thus, just add the required nodes but leave the usdhc1 node disabled
>> explicitly. Users who disabled the Bluetooth (AMP) support of their
>> OS can then conveniently enable WiFi (or even plain Bluetooth) support
>> with a simple device tree overlay/bootloader configuration.
>>
>> Signed-off-by: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
>> ---
>>
>> Hi all,
>>
>> some more information on the Bluetooth AMP dilemma: the vendor (CompuLab) just
>> remove Bluetooth AMP support from the kernel. However, this is not an option
>> here, since it would introduce a regression for its actual users.
>>
>> Neither is degrading the firmware a good option: recent firmare (without AMP
>> support) provides WiFi that is more stable (in particular, for access points).
>>
>> I wrote a RFC to linux-bluetooth [1] two weeks ago about this problem and the
>> only feedback I got was that Marvell has to provide information about how to
>> detect whether the firmware supports AMP properly -- but this is unlikely given
>> Marvell's information policy (in particular, since the sd8787 seems to be
>> unsupported as of now).
>>
>> IMHO this patch is currently the best option, since it allows users in a
>> convenient way to enable WiFi support (e.g. no pinctrl handling in overlays),
>> even when they are using distribution kernels/device trees. Furthermore, the
>> hardware description is entirely accurate -- we have a driver/firmware problem
>> here.
> 
> Hi Valentin, Igor,
> 
> Are you guys fine with the patch?

Do you have a specific question here?

The devicetree additions are mostly identical to their counterparts in
the BSP devicetree. The only real exception being that the usdhc1
regulators have been replaced by the sd8787 power sequence, which allows
the mmc core to reset the wifi/bt chip (which is not possible with the
regulator setup). Note that the sd8787 power sequence binding/driver has
only recently been added to Linux (v4.11) and is not present in the BSP
kernel.

Furthermore, the Bluetooth AMP function is disabled in the BSP kernel
by CompuLab [1] for the same reason we do not enable the usdhc1 node in
this patch. (We cannot disable it in the driver because, as explained in
the commit message, this would yield a regression.)

Thanks,
Christopher

[1]
https://github.com/utilite-computer/linux-kernel/commit/d2ef249cc37a61ec88bc0da0ffcc75799e12f25f.patch

> Shawn
> 
>>
>> Thanks,
>> Christopher
>>
>> [1] https://www.spinics.net/lists/linux-bluetooth/msg70493.html
>>
>>  arch/arm/boot/dts/imx6q-cm-fx6.dts | 41 ++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 41 insertions(+)
>>
>> diff --git a/arch/arm/boot/dts/imx6q-cm-fx6.dts b/arch/arm/boot/dts/imx6q-cm-fx6.dts
>> index a8af382..fe6ab0a 100644
>> --- a/arch/arm/boot/dts/imx6q-cm-fx6.dts
>> +++ b/arch/arm/boot/dts/imx6q-cm-fx6.dts
>> @@ -64,6 +64,14 @@
>>  		};
>>  	};
>>  
>> +	awnh387_pwrseq: pwrseq {
>> +		pinctrl-names = "default";
>> +		pinctrl-0 = <&pinctrl_pwrseq>;
>> +		compatible = "mmc-pwrseq-sd8787";
>> +		powerdown-gpios = <&gpio7 12 GPIO_ACTIVE_HIGH>;
>> +		reset-gpios = <&gpio6 16 GPIO_ACTIVE_HIGH>;
>> +	};
>> +
>>  	reg_pcie_power_on_gpio: regulator-pcie-power-on-gpio {
>>  		compatible = "regulator-fixed";
>>  		regulator-name = "regulator-pcie-power-on-gpio";
>> @@ -304,6 +312,13 @@
>>  		>;
>>  	};
>>  
>> +	pinctrl_pwrseq: pwrseqgrp {
>> +		fsl,pins = <
>> +			MX6QDL_PAD_GPIO_17__GPIO7_IO12		0x1b0b0
>> +			MX6QDL_PAD_NANDF_CS3__GPIO6_IO16	0x1b0b0
>> +		>;
>> +	};
>> +
>>  	pinctrl_spdif: spdifgrp {
>>  		fsl,pins = <
>>  			MX6QDL_PAD_GPIO_16__SPDIF_IN  0x1b0b0
>> @@ -330,6 +345,17 @@
>>  			MX6QDL_PAD_EIM_D22__GPIO3_IO22	0x130b0
>>  		>;
>>  	};
>> +
>> +	pinctrl_usdhc1: usdhc1grp {
>> +		fsl,pins = <
>> +			MX6QDL_PAD_SD1_CMD__SD1_CMD	0x17071
>> +			MX6QDL_PAD_SD1_CLK__SD1_CLK	0x10071
>> +			MX6QDL_PAD_SD1_DAT0__SD1_DATA0	0x17071
>> +			MX6QDL_PAD_SD1_DAT1__SD1_DATA1	0x17071
>> +			MX6QDL_PAD_SD1_DAT2__SD1_DATA2	0x17071
>> +			MX6QDL_PAD_SD1_DAT3__SD1_DATA3	0x17071
>> +		>;
>> +	};
>>  };
>>  
>>  &pcie {
>> @@ -382,3 +408,18 @@
>>  	dr_mode = "otg";
>>  	status = "okay";
>>  };
>> +
>> +&usdhc1 {
>> +	pinctrl-names = "default";
>> +	pinctrl-0 = <&pinctrl_usdhc1>;
>> +	mmc-pwrseq = <&awnh387_pwrseq>;
>> +	non-removable;
>> +	/*
>> +	 * If the OS probes the Bluetooth AMP function advertised on this bus
>> +	 * but the firmware in place does not support it, the WiFi/BT module
>> +	 * gets unresponsive.
>> +	 * Users who configured their OS properly can enable this node to gain
>> +	 * WiFi and/or plain Bluetooth support.
>> +	 */
>> +	status = "disabled";
>> +};
>> -- 
>> 2.10.2
>>
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Shawn Guo]

On Thu, Jun 15, 2017 at 11:23:46PM +0200, Christopher Spinrath wrote:
> > Hi Valentin, Igor,
> > 
> > Are you guys fine with the patch?
> 
> Do you have a specific question here?

No, I don't.  I just wanted to hear the voice of Compulab people.  Okay,
we haven't waited for time enough, so I just applied patch.

> 
> The devicetree additions are mostly identical to their counterparts in
> the BSP devicetree. The only real exception being that the usdhc1
> regulators have been replaced by the sd8787 power sequence, which allows
> the mmc core to reset the wifi/bt chip (which is not possible with the
> regulator setup). Note that the sd8787 power sequence binding/driver has
> only recently been added to Linux (v4.11) and is not present in the BSP
> kernel.
> 
> Furthermore, the Bluetooth AMP function is disabled in the BSP kernel
> by CompuLab [1] for the same reason we do not enable the usdhc1 node in
> this patch. (We cannot disable it in the driver because, as explained in
> the commit message, this would yield a regression.)

Thanks for the comments.

Shawn

[PATCH] ARM: dts: imx6q-cm-fx6: add sdio wifi/bt nodes

[Shawn Guo]

On Sat, Jun 17, 2017 at 09:23:31PM +0800, Shawn Guo wrote:
> On Thu, Jun 15, 2017 at 11:23:46PM +0200, Christopher Spinrath wrote:
> > > Hi Valentin, Igor,
> > > 
> > > Are you guys fine with the patch?
> > 
> > Do you have a specific question here?
> 
> No, I don't.  I just wanted to hear the voice of Compulab people.  Okay,
> we haven't waited for time enough, so I just applied patch.

s/haven't/have