MCP2518FD : Failed to detect MCP2518FD (osc=0x00000000)
Hi Folks,
I am working on an IMX7D board. Some time ago I have back-ported the MCP2518FD driver to the official 5.4.70 NXP kernel and it works fine.
I am struggleing to get the mcp driver working on the latest nxp kernel 5.10.72.
Same hardware.
I did not find any unusal in the logs. pinctrl did not complain, spi seems to initialize normal.
The old kernel with the new (almost old one) device tree works. The new kernel did not work with the old or new device tree.
At the moment I am stuck. Could be anything. Must not be the driver itself.
BR
Niels
mcp2518fd_clk: mcp2818fd_clk {
compatible = "fixed-clock";
#clock-cells = <0>;
clock-frequency = <40000000>;
};&ecspi2 {
pinctrl-names = "default";
pinctrl-0 = <&pinctrl_ecspi2>;
status = "okay";
fsl,spi-num-chipselects = <2>;
cs-gpios = <&gpio5 1 0>, <&gpio1 3 0>;
mcp2518fd@0 {
compatible = "microchip,mcp2518fd";
reg = <0>;
interrupts-extended = <&gpio3 0 IRQ_TYPE_LEVEL_LOW>;
spi-max-frequency = <20000000>;
clocks = <&mcp2518fd_clk>;
};
mcp2518fd@1 {
compatible = "microchip,mcp2518fd";
reg = <1>;
interrupts-extended = <&gpio4 23 IRQ_TYPE_LEVEL_LOW>;
spi-max-frequency = <20000000>;
clocks = <&mcp2518fd_clk>;
};
};Marc Kleine-Budde
maintainerQuoted 8 lines · On 31.01.2022 09:33:04, [email protected] wrote:
I am working on an IMX7D board. Some time ago I have back-ported the MCP2518FD driver to the official 5.4.70 NXP kernel and it works fine.
I am struggleing to get the mcp driver working on the latest nxp kernel 5.10.72.
Same hardware.
Quoted 8 lines
I did not find any unusal in the logs. pinctrl did not complain, spi seems to initialize normal.
The old kernel with the new (almost old one) device tree works. The new kernel did not work with the old or new device tree.
At the moment I am stuck. Could be anything. Must not be the driver itself.
Is SPI DMA enabled? Try disabling it.
regards,
Marc
Signature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Hi Marc,
I have disabled DMAs in imx7s.dtsi for all ecspis and recompile the device tree. Still get the same failure.
Thanks
Niels
ecspi2: spi@30830000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,imx7d-ecspi", "fsl,imx51-ecspi";
reg = <0x30830000 0x10000>;
interrupts = <GIC_SPI 32 IRQ_TYPE_LEVEL_HIGH>;
clocks = <&clks IMX7D_ECSPI2_ROOT_CLK>,
<&clks IMX7D_ECSPI2_ROOT_CLK>;
clock-names = "ipg", "per";
/*
dmas = <&sdma 2 7 1>, <&sdma 3 7 2>;
dma-names = "rx", "tx";
*/
status = "disabled";
};Marc Kleine-Budde schrieb am 31.01.2022 08:42 (GMT +00:00):
Quoted 29 lines · On 31.01.2022 09:33:04, [email protected] wrote:
Quoted 8 lines
I am working on an IMX7D board. Some time ago I have back-ported the MCP2518FD driver to the official 5.4.70 NXP kernel and it works fine.
I am struggleing to get the mcp driver working on the latest nxp kernel 5.10.72.
Same hardware.
Quoted 8 lines
I did not find any unusal in the logs. pinctrl did not complain, spi seems to initialize normal.
The old kernel with the new (almost old one) device tree works. The new kernel did not work with the old or new device tree.
At the moment I am stuck. Could be anything. Must not be the driver itself.
Is SPI DMA enabled? Try disabling it.
regards,
MarcSignature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Marc Kleine-Budde
maintainerQuoted 2 lines · On 31.01.2022 10:35:54, [email protected] wrote:
I have disabled DMAs in
imx7s.dtsifor all ecspis and recompile the device tree. Still get the same failure.
You can check if that really works by looking at the used DT:
$ dtc -I fs /proc/device-tree
Next step would be to get a scope and attach it to the MISO/MOSI/CLk/CS pins.
Marc
Signature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Hi,
DMA is disabled. CLK and MOSI are looking fine.
Very interessing, CSs are not asserted when loading driver (on old kernel it does), but I can measure "high" after HW initialisation.
Since I did not change the IOs between the old and new device tree assuming the "high" on CS also showing correct PAD / GPIO configuration there seems a problem asserting the CS.
cat /sys/kernel/debug/gpio looks good (gpio-3 and gpio-129)
gpiochip0: GPIOs 0-31, parent: platform/30200000.gpio, 30200000.gpio:
gpio-3 ( |spi1 CS1 ) out hi ACTIVE LOW gpio-5 ( |regulator-vbus ) out lo
gpiochip1: GPIOs 64-95, parent: platform/30220000.gpio, 30220000.gpio:
gpio-67 ( |spi3 CS0 ) out hi ACTIVE LOW
gpiochip2: GPIOs 96-127, parent: platform/30230000.gpio, 30230000.gpio:
gpiochip3: GPIOs 128-159, parent: platform/30240000.gpio, 30240000.gpio:
gpio-128 ( |cd ) in hi IRQ ACTIVE LOW gpio-129 ( |spi1 CS0 ) out hi ACTIVE LOW
BR
Niels
Marc Kleine-Budde schrieb am 31.01.2022 09:58 (GMT +00:00):
Quoted 19 lines · On 31.01.2022 10:35:54, [email protected] wrote:
Quoted 2 lines
I have disabled DMAs in
imx7s.dtsifor all ecspis and recompile the device tree. Still get the same failure.You can check if that really works by looking at the used DT:
$ dtc -I fs /proc/device-treeNext step would be to get a scope and attach it to the MISO/MOSI/CLk/CS pins.
Marc
Signature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Marc Kleine-Budde
maintainerQuoted 4 lines · On 31.01.2022 13:44:28, [email protected] wrote:
DMA is disabled. CLK and MOSI are looking fine.
Very interessing, CSs are not asserted when loading driver (on old kernel it does), but I can measure "high" after HW initialisation.
Asserted means low?
Quoted 3 lines
Since I did not change the IOs between the old and new device tree assuming the "high" on CS also showing correct PAD / GPIO configuration there seems a problem asserting the CS.
| &ecspi2 {
| pinctrl-names = "default";
| pinctrl-0 = <&pinctrl_ecspi2>;
| status = "okay";
|
| fsl,spi-num-chipselects = <2>; V
| cs-gpios = <&gpio5 1 0>, <&gpio1 3 0>;
^ ^
Have a look at include/dt-bindings/gpio/gpio.h:
/* Bit 0 express polarity */
#define GPIO_ACTIVE_HIGH 0
#define GPIO_ACTIVE_LOW 1
I'm using GPIO_ACTIVE_LOW in my DT on imx6dl.
|
|
| mcp2518fd@0 {
| compatible = "microchip,mcp2518fd";
| reg = <0>;
| interrupts-extended = <&gpio3 0 IRQ_TYPE_LEVEL_LOW>;
| spi-max-frequency = <20000000>;
| clocks = <&mcp2518fd_clk>;
| };
|
| mcp2518fd@1 {
| compatible = "microchip,mcp2518fd";
| reg = <1>;
| interrupts-extended = <&gpio4 23 IRQ_TYPE_LEVEL_LOW>;
| spi-max-frequency = <20000000>;
| clocks = <&mcp2518fd_clk>;
| };
| }; Quoted 1 line
cat
/sys/kernel/debug/gpiolooks good (gpio-3 and gpio-129)
This doesn't look consistent with your DT, above. I'm always confused with numbering starting with 0 and 1 in the DT, but I think:
gpiochip0: gpio-3 == &gpio1 3
gpiochip3: gpio-129 == &gpio4 1 (not &gpio5 1 as in you DT)
Quoted 4 lines
gpiochip0: GPIOs 0-31, parent: platform/30200000.gpio, 30200000.gpio:
gpio-3 ( |spi1 CS1 ) out hi ACTIVE LOW gpio-5 ( |regulator-vbus ) out lo
Quoted 8 lines
gpiochip1: GPIOs 64-95, parent: platform/30220000.gpio, 30220000.gpio:
gpio-67 ( |spi3 CS0 ) out hi ACTIVE LOWgpiochip2: GPIOs 96-127, parent: platform/30230000.gpio, 30230000.gpio:
gpiochip3: GPIOs 128-159, parent: platform/30240000.gpio, 30240000.gpio:
gpio-128 ( |cd ) in hi IRQ ACTIVE LOW gpio-129 ( |spi1 CS0 ) out hi ACTIVE LOW
regards,
Marc
Signature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Hi Marc,
CS = High (de-asserted, slave not selected)
CS = low (asserted, slave selected)
I also did not like the 0/1 starting point, but it is defined here:
aliases {
gpio0 = &gpio1;
gpio1 = &gpio2;
gpio2 = &gpio3;
...&gpio1 3 = gpiochip0 = 0*32 + 3 = gpio3
&gpio5 1 = gpiochip4 = 4*32 + 1 = gpio129
Interresting with ACTIVE HIGH / LOW. cat debug/gpio tells me "ACTIVE LOW" but anyway I did not see CS toggle at all. Also tested "1" but nothing changed.
By the way, the old kernel is working with the new device tree.
I did a make clean and make sure that I really use latest build artefacts. Everything is okay. Still failure.
Maybe there are problems when using "make imx_v7_defconfig" ... possible conflicts between drivers because there are much changes between 5.4 and 5.10 whats enabled and/or new by default ? Try to go with a very small kernel.
BR
Niels
Marc Kleine-Budde schrieb am 31.01.2022 13:09 (GMT +00:00):
Quoted 84 lines · On 31.01.2022 13:44:28, [email protected] wrote:
Quoted 4 lines
DMA is disabled. CLK and MOSI are looking fine.
Very interessing, CSs are not asserted when loading driver (on old kernel it does), but I can measure "high" after HW initialisation.
Asserted means low?
Quoted 3 lines
Since I did not change the IOs between the old and new device tree assuming the "high" on CS also showing correct PAD / GPIO configuration there seems a problem asserting the CS.
| &ecspi2 { | pinctrl-names = "default"; | pinctrl-0 = <&pinctrl_ecspi2>; | status = "okay"; | | fsl,spi-num-chipselects = <2>;V
| cs-gpios = <&gpio5 1 0>, <&gpio1 3 0>;
^ ^
Have a look atinclude/dt-bindings/gpio/gpio.h:/* Bit 0 express polarity */
#defineGPIO_ACTIVE_HIGH0
#defineGPIO_ACTIVE_LOW1I'm using
GPIO_ACTIVE_LOWin my DT on imx6dl.| | | mcp2518fd@0 { | compatible = "microchip,mcp2518fd"; | reg = <0>; | interrupts-extended = <&gpio3 0 IRQ_TYPE_LEVEL_LOW>; | spi-max-frequency = <20000000>; | clocks = <&mcp2518fd_clk>; | };| | mcp2518fd@1 { | compatible = "microchip,mcp2518fd"; | reg = <1>; | interrupts-extended = <&gpio4 23 IRQ_TYPE_LEVEL_LOW>; | spi-max-frequency = <20000000>; | clocks = <&mcp2518fd_clk>; | };| };
Quoted 1 line
cat
/sys/kernel/debug/gpiolooks good (gpio-3 and gpio-129)This doesn't look consistent with your DT, above. I'm always confused with numbering starting with 0 and 1 in the DT, but I think:
gpiochip0: gpio-3 == &gpio1 3
gpiochip3: gpio-129 == &gpio4 1 (not &gpio5 1 as in you DT)Quoted 4 lines
gpiochip0: GPIOs 0-31, parent: platform/30200000.gpio, 30200000.gpio:
gpio-3 ( |spi1 CS1 ) out hi ACTIVE LOW gpio-5 ( |regulator-vbus ) out loQuoted 8 lines
gpiochip1: GPIOs 64-95, parent: platform/30220000.gpio, 30220000.gpio:
gpio-67 ( |spi3 CS0 ) out hi ACTIVE LOWgpiochip2: GPIOs 96-127, parent: platform/30230000.gpio, 30230000.gpio:
gpiochip3: GPIOs 128-159, parent: platform/30240000.gpio, 30240000.gpio:
gpio-128 ( |cd ) in hi IRQ ACTIVE LOW gpio-129 ( |spi1 CS0 ) out hi ACTIVE LOWregards,
MarcSignature · 4 lines
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Marc Kleine-Budde
maintainerQuoted 16 lines · On 31.01.2022 15:15:03, [email protected] wrote:
Hi Marc,
CS = High (de-asserted, slave not selected)
CS = low (asserted, slave selected)I also did not like the 0/1 starting point, but it is defined here:
aliases { gpio0 = &gpio1; gpio1 = &gpio2; gpio2 = &gpio3; ...
Quoted 2 lines
&gpio1 3 = gpiochip0 = 0*32 + 3 = gpio3
&gpio5 1 = gpiochip4 = 4*32 + 1 = gpio129
In your output of /sys/kernel/debug/gpio, gpiochip3 starts at 128:
Quoted 3 lines
Quoted 3 lines
gpiochip3: GPIOs 128-159, parent: platform/30240000.gpio, 30240000.gpio:
gpio-128 ( |cd ) in hi IRQ ACTIVE LOW gpio-129 ( |spi1 CS0 ) out hi ACTIVE LOW
Here it looks like this:
gpiochip4: GPIOs 128-159, parent: platform/20ac000.gpio, 20ac000.gpio:
gpio-130 (EIM_A25 |user1 ) out lo ACTIVE LOW gpio-133 (GPIO5_05 ) gpio-134 (GPIO5_06 |reset ) out lo gpio-135 (GPIO5_07 ) gpio-136 (GPIO5_08 ) gpio-137 (CSPI2_CS1 |spi1 CS1 ) out hi ACTIVE LOW gpio-138 (CSPI2_MOSI ) gpio-139 (CSPI2_MISO ) gpio-140 (CSPI2_CS0 |spi1 CS0 ) out hi ACTIVE LOW gpio-141 (CSPI2_CLK )
Quoted 6 lines
Interresting with ACTIVE HIGH / LOW. cat debug/gpio tells me "ACTIVE LOW" but anyway I did not see CS toggle at all. Also tested "1" but nothing changed.
By the way, the old kernel is working with the new device tree.
Quoted 7 lines
I did a make clean and make sure that I really use latest build artefacts. Everything is okay. Still failure.
Maybe there are problems when using "make
imx_v7_defconfig" ... possible conflicts between drivers because there are much changes between 5.4 and 5.10 whats enabled and/or new by default ? Try to go with a very small kernel.
There shouldn't be any conflicts between drivers.
regards,
Marc
--
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
Hi Folks,
Embarrassing ...
There was a small build of 5. There was an hw change afterwarts regardging CS. Reworked 4/5. I got the one without rework but special "local" working test software. I wonder why switching to master is not working ...
Sorry for your time lapse.
BR
Niels
Marc Kleine-Budde schrieb am 31.01.2022 15:17 (GMT +00:00):
Quoted 65 lines · On 31.01.2022 15:15:03, [email protected] wrote:
Quoted 16 lines
Hi Marc,
CS = High (de-asserted, slave not selected)
CS = low (asserted, slave selected)I also did not like the 0/1 starting point, but it is defined here:
aliases { gpio0 = &gpio1; gpio1 = &gpio2; gpio2 = &gpio3; ...Quoted 2 lines
&gpio1 3 = gpiochip0 = 0*32 + 3 = gpio3
&gpio5 1 = gpiochip4 = 4*32 + 1 = gpio129In your output of
/sys/kernel/debug/gpio, gpiochip3 starts at 128:Quoted 3 lines
Quoted 3 lines
gpiochip3: GPIOs 128-159, parent: platform/30240000.gpio, 30240000.gpio:
gpio-128 ( |cd ) in hi IRQ ACTIVE LOW gpio-129 ( |spi1 CS0 ) out hi ACTIVE LOWHere it looks like this:
gpiochip4: GPIOs 128-159, parent: platform/20ac000.gpio, 20ac000.gpio:
gpio-130 (EIM_A25 |user1 ) out lo ACTIVE LOW gpio-133 (GPIO5_05 ) gpio-134 (GPIO5_06 |reset ) out lo gpio-135 (GPIO5_07 ) gpio-136 (GPIO5_08 ) gpio-137 (CSPI2_CS1 |spi1 CS1 ) out hi ACTIVE LOW gpio-138 (CSPI2_MOSI ) gpio-139 (CSPI2_MISO ) gpio-140 (CSPI2_CS0 |spi1 CS0 ) out hi ACTIVE LOW gpio-141 (CSPI2_CLK )Quoted 6 lines
Interresting with ACTIVE HIGH / LOW. cat debug/gpio tells me "ACTIVE LOW" but anyway I did not see CS toggle at all. Also tested "1" but nothing changed.
By the way, the old kernel is working with the new device tree.
Quoted 7 lines
I did a make clean and make sure that I really use latest build artefacts. Everything is okay. Still failure.
Maybe there are problems when using "make
imx_v7_defconfig" ... possible conflicts between drivers because there are much changes between 5.4 and 5.10 whats enabled and/or new by default ? Try to go with a very small kernel.There shouldn't be any conflicts between drivers.
regards,
Marc--
Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |