Skip to contentSkip to conversation

nvmem: core: Mark nWP GPIO as non-exclusive

PATCH In review Marek Vasut 2 participants 10 messages May 11 – Jul 29

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nvmem: core: Mark nWP GPIO as non-exclusive

Mark the nWP GPIO as non-exclusive, because there is hardware in the field which has multiple AT24 EEPROMs with nWP GPIO line tied to a single GPIO. This allows such hardware to be described in DT and the nWP GPIO operated via NVMEM force_ro sysfs attribute.

Note that GPIOD_FLAGS_BIT_NONEXCLUSIVE is a deprecated flag and should not be used, but thus far there seems to be no replacement.

---
Cc: Arnd Bergmann <[email protected]>
Cc: Bartosz Golaszewski <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Srinivas Kandagatla <[email protected]>
Cc: [email protected]
Cc: [email protected]

---
 drivers/nvmem/core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/nvmem/core.c +2 -1
index be28a366f6031..4fe2fad2f88b4 100644
@@ -928,7 +928,8 @@ struct nvmem_device *nvmem_register(const struct nvmem_config *config)
 
 	if (!config->ignore_wp)
 		nvmem->wp_gpio = gpiod_get_optional(config->dev, "wp",
-						    GPIOD_OUT_HIGH);
+						    GPIOD_OUT_HIGH |
+						    GPIOD_FLAGS_BIT_NONEXCLUSIVE);
 	if (IS_ERR(nvmem->wp_gpio)) {
 		rval = PTR_ERR(nvmem->wp_gpio);
 		nvmem->wp_gpio = NULL;
Signature · 1 line
2.53.0
#1 lore raw
Quoted 9 lines · Marek Vasut

Mark the nWP GPIO as non-exclusive, because there is hardware in the field which has multiple AT24 EEPROMs with nWP GPIO line tied to a single GPIO. This allows such hardware to be described in DT and the nWP GPIO operated via NVMEM force_ro sysfs attribute.

Note that GPIOD_FLAGS_BIT_NONEXCLUSIVE is a deprecated flag and should not be used, but thus far there seems to be no replacement.

There is a replacement now. It lives under drivers/gpio/gpiolib-shared.c and drivers/gpio/gpio-shared-proxy.c. Though the voting mechanism I implemented is hard-coded to drive the line high if there's at least one consumer sharing the line that votes high. I'm now thinking that this may be too rigid as yesterday, within two hours I had two submissions wanting to do it the other way around
- drive the line low if there's at least one consumer that wants it low.

I'll try to modify the gpio-shared-proxy driver to take into account the value set when the line is first requested and treat it as the "default" in the voting mechanism. IOW: when the first user requests it out-high, we'll drive it low on the first user that votes for low and drive it high again when the last user drivers it high. This should work for both the Qualcomm shared GPIOs for audio use-cases and the WP pin here as well as the SCL recovery shared GPIO we have in I2C.

With that said: I'd prefer you not set the GPIOD_FLAGS_BIT_NONEXCLUSIVE and see if you can make gpio-shared-proxy work for you instead. I'll submit something soon for you to test.

Bart

#2 lore raw
Quoted 20 lines · Bartosz Golaszewski
Quoted 9 lines · Marek Vasut

Mark the nWP GPIO as non-exclusive, because there is hardware in the field which has multiple AT24 EEPROMs with nWP GPIO line tied to a single GPIO. This allows such hardware to be described in DT and the nWP GPIO operated via NVMEM force_ro sysfs attribute.

Note that GPIOD_FLAGS_BIT_NONEXCLUSIVE is a deprecated flag and should not be used, but thus far there seems to be no replacement.

There is a replacement now. It lives under drivers/gpio/gpiolib-shared.c and drivers/gpio/gpio-shared-proxy.c. Though the voting mechanism I implemented is hard-coded to drive the line high if there's at least one consumer sharing the line that votes high. I'm now thinking that this may be too rigid as yesterday, within two hours I had two submissions wanting to do it the other way around
- drive the line low if there's at least one consumer that wants it low.

Quoted 8 lines

I'll try to modify the gpio-shared-proxy driver to take into account the value set when the line is first requested and treat it as the "default" in the voting mechanism. IOW: when the first user requests it out-high, we'll drive it low on the first user that votes for low and drive it high again when the last user drivers it high. This should work for both the Qualcomm shared GPIOs for audio use-cases and the WP pin here as well as the SCL recovery shared GPIO we have in I2C.

Shouldn't the polarity inversion be handled by DT GPIO_ACTIVE_* flags instead ?

#3 lore raw
In reply to Bartosz Golaszewski
Quoted 29 lines · Bartosz Golaszewski
Quoted 9 lines · Marek Vasut

Mark the nWP GPIO as non-exclusive, because there is hardware in the field which has multiple AT24 EEPROMs with nWP GPIO line tied to a single GPIO. This allows such hardware to be described in DT and the nWP GPIO operated via NVMEM force_ro sysfs attribute.

Note that GPIOD_FLAGS_BIT_NONEXCLUSIVE is a deprecated flag and should not be used, but thus far there seems to be no replacement.

There is a replacement now. It lives under drivers/gpio/gpiolib-shared.c and drivers/gpio/gpio-shared-proxy.c. Though the voting mechanism I implemented is hard-coded to drive the line high if there's at least one consumer sharing the line that votes high. I'm now thinking that this may be too rigid as yesterday, within two hours I had two submissions wanting to do it the other way around
- drive the line low if there's at least one consumer that wants it low.

I'll try to modify the gpio-shared-proxy driver to take into account the value set when the line is first requested and treat it as the "default" in the voting mechanism. IOW: when the first user requests it out-high, we'll drive it low on the first user that votes for low and drive it high again when the last user drivers it high. This should work for both the Qualcomm shared GPIOs for audio use-cases and the WP pin here as well as the SCL recovery shared GPIO we have in I2C.

Quoted 3 lines

With that said: I'd prefer you not set the GPIOD_FLAGS_BIT_NONEXCLUSIVE and see if you can make gpio-shared-proxy work for you instead. I'll submit something soon for you to test.

I'm sorry for my late reply, I was completely overwhelmed by emails.

I tried to look into the gpio-shared* , but I couldn't figure out how to describe it in DT. Do you have any examples?

Take for example this DT snippet:

gpio: gpio-controller { ... };

i2c {

  eeprom@0 {
    wp-gpios = <&gpio 0 0>;
    ...

};

  eeprom@1 {
    wp-gpios = <&gpio 0 0>;
    ...

};
};

I assume there has to be some new glue node between the eeproms and the GPIO controller somewhere ?

Thank you for your help !

Signature · 2 lines
Best regards,
Marek Vasut
#4 lore raw
Quoted 55 lines · Marek Vasut
Quoted 29 lines · Bartosz Golaszewski
Quoted 9 lines · Marek Vasut

Mark the nWP GPIO as non-exclusive, because there is hardware in the field which has multiple AT24 EEPROMs with nWP GPIO line tied to a single GPIO. This allows such hardware to be described in DT and the nWP GPIO operated via NVMEM force_ro sysfs attribute.

Note that GPIOD_FLAGS_BIT_NONEXCLUSIVE is a deprecated flag and should not be used, but thus far there seems to be no replacement.

There is a replacement now. It lives under drivers/gpio/gpiolib-shared.c and drivers/gpio/gpio-shared-proxy.c. Though the voting mechanism I implemented is hard-coded to drive the line high if there's at least one consumer sharing the line that votes high. I'm now thinking that this may be too rigid as yesterday, within two hours I had two submissions wanting to do it the other way around
- drive the line low if there's at least one consumer that wants it low.

I'll try to modify the gpio-shared-proxy driver to take into account the value set when the line is first requested and treat it as the "default" in the voting mechanism. IOW: when the first user requests it out-high, we'll drive it low on the first user that votes for low and drive it high again when the last user drivers it high. This should work for both the Qualcomm shared GPIOs for audio use-cases and the WP pin here as well as the SCL recovery shared GPIO we have in I2C.

Quoted 3 lines

With that said: I'd prefer you not set the GPIOD_FLAGS_BIT_NONEXCLUSIVE and see if you can make gpio-shared-proxy work for you instead. I'll submit something soon for you to test.

I'm sorry for my late reply, I was completely overwhelmed by emails.

I tried to look into the gpio-shared* , but I couldn't figure out how to describe it in DT. Do you have any examples?

Take for example this DT snippet:

gpio: gpio-controller { ... };

i2c {
   eeprom@0 {
     wp-gpios = <&gpio 0 0>;
     ...
   };

   eeprom@1 {
     wp-gpios = <&gpio 0 0>;
     ...
   };
};
Quoted 5 lines

I assume there has to be some new glue node between the eeproms and the GPIO controller somewhere ?

Thank you for your help !

No, this is literally how you do it. If you enable HAVE_SHARED_GPIOs and DEBUG_GPIO, you should see it being picked up in the kernel log.

Bartosz

#5 lore raw

On 7/24/26 9:30 AM, Bartosz Golaszewski wrote:

Hello Bartosz,

Quoted 19 lines
Quoted 19 lines

I tried to look into the gpio-shared* , but I couldn't figure out how to describe it in DT. Do you have any examples?

Take for example this DT snippet:

gpio: gpio-controller { ... };

i2c {
    eeprom@0 {
      wp-gpios = <&gpio 0 0>;
      ...
    };

    eeprom@1 {
      wp-gpios = <&gpio 0 0>;
      ...
    };
};
Quoted 8 lines
Quoted 5 lines

I assume there has to be some new glue node between the eeproms and the GPIO controller somewhere ?

Thank you for your help !

No, this is literally how you do it. If you enable HAVE_SHARED_GPIOs and DEBUG_GPIO, you should see it being picked up in the kernel log.

Ah, HAVE_SHARED_GPIOS is what I did not have enabled. Is there any reason why this is not enabled by default ?

Thank you !

#6 lore raw
Quoted 26 lines · Marek Vasut

On 7/24/26 9:30 AM, Bartosz Golaszewski wrote:

Hello Bartosz,

Quoted 22 lines
Quoted 22 lines

I tried to look into the gpio-shared* , but I couldn't figure out how to describe it in DT. Do you have any examples?

Take for example this DT snippet:

gpio: gpio-controller { ... };

i2c {
    eeprom@0 {
      wp-gpios = <&gpio 0 0>;
      ...
    };

    eeprom@1 {
      wp-gpios = <&gpio 0 0>;
      ...
    };
};

I assume there has to be some new glue node between the eeproms and the GPIO controller somewhere ?

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Thank you for your help !

No, this is literally how you do it. If you enable HAVE_SHARED_GPIOs and DEBUG_GPIO, you should see it being picked up in the kernel log.

Ah, HAVE_SHARED_GPIOS is what I did not have enabled. Is there any reason why this is not enabled by default ?

It will be. I'm doing it gradually. It's enabled for Qualcomm platforms by default and - by extension - in arm64 defconfig. I plan to enable it for other platforms and architectures too, I just don't want to rush it. Enabling it for arm64 defconfig had caused quite a lot of regressions a few months ago so I prefer to be careful.

What platform are you running this on?

Bart

#7 lore raw

On 7/27/26 3:10 PM, Bartosz Golaszewski wrote:

Hello Bartosz,

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

No, this is literally how you do it. If you enable HAVE_SHARED_GPIOs and DEBUG_GPIO, you should see it being picked up in the kernel log.

Ah, HAVE_SHARED_GPIOS is what I did not have enabled. Is there any reason why this is not enabled by default ?

Quoted 7 lines

It will be. I'm doing it gradually. It's enabled for Qualcomm platforms by default and - by extension - in arm64 defconfig. I plan to enable it for other platforms and architectures too, I just don't want to rush it. Enabling it for arm64 defconfig had caused quite a lot of regressions a few months ago so I prefer to be careful.

What platform are you running this on?

NXP i.MX8M Nano (arm64)

#8 lore raw
Quoted 10 lines · Marek Vasut

On 7/27/26 3:10 PM, Bartosz Golaszewski wrote:

Hello Bartosz,

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

No, this is literally how you do it. If you enable HAVE_SHARED_GPIOs and DEBUG_GPIO, you should see it being picked up in the kernel log.

Ah, HAVE_SHARED_GPIOS is what I did not have enabled. Is there any reason why this is not enabled by default ?

Quoted 9 lines
Quoted 7 lines

It will be. I'm doing it gradually. It's enabled for Qualcomm platforms by default and - by extension - in arm64 defconfig. I plan to enable it for other platforms and architectures too, I just don't want to rush it. Enabling it for arm64 defconfig had caused quite a lot of regressions a few months ago so I prefer to be careful.

What platform are you running this on?

NXP i.MX8M Nano (arm64)

Just consider sending the following once you can make this work for you?

arch/arm64/Kconfig.platforms +1 -0
index d2acfac730037..a02b37c752887 100644
@@ -281,6 +281,7 @@ config ARCH_MXC
 	select PM_GENERIC_DOMAINS
 	select SOC_BUS
 	select TIMER_IMX_SYS_CTR
+	select HAVE_SHARED_GPIOS
 	help
 	  This enables support for the ARMv8 based SoCs in the
 	  NXP i.MX family.

Thanks,
Bartosz

#9 lore raw

On 7/28/26 10:32 AM, Bartosz Golaszewski wrote:

Hello Bartosz,

Quoted 17 lines
Quoted 9 lines
Quoted 7 lines

It will be. I'm doing it gradually. It's enabled for Qualcomm platforms by default and - by extension - in arm64 defconfig. I plan to enable it for other platforms and architectures too, I just don't want to rush it. Enabling it for arm64 defconfig had caused quite a lot of regressions a few months ago so I prefer to be careful.

What platform are you running this on?

NXP i.MX8M Nano (arm64)

Just consider sending the following once you can make this work for you?

arch/arm64/Kconfig.platforms +0 -0
index d2acfac730037..a02b37c752887 100644

arch/arm64/Kconfig.platforms config ARCH_MXC · line 287

Quoted 7 lines
arch/arm64/Kconfig.platforms +1 -0
@@ -281,6 +281,7 @@ config ARCH_MXC
  	select PM_GENERIC_DOMAINS
  	select SOC_BUS
  	select TIMER_IMX_SYS_CTR
+	select HAVE_SHARED_GPIOS
  	help
  	  This enables support for the ARMv8 based SoCs in the
  	  NXP i.MX family.

Will do (sorted alphabetically).

Thank you !