Skip to contentSkip to conversation

drivers: tpm2: update reset gpio semantics

PATCH Superseded Jorge Ramirez-Ortiz 5 participants 8 messages May 26, 2021 – May 31, 2021

v1 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drivers: tpm2: update reset gpio semantics

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg			: SPI Chip select
 
 Optional properties:
-- gpio-reset		: Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios		: Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency	: See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +1 -1
index 4b33ac8fd3..94ac52d9ce 100644
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
 	if (CONFIG_IS_ENABLED(DM_GPIO)) {
 		struct gpio_desc reset_gpio;
 
-		ret = gpio_request_by_name(dev, "gpio-reset", 0,
+		ret = gpio_request_by_name(dev, "reset-gpios", 0,
 					   &reset_gpio, GPIOD_IS_OUT);
 		if (ret) {
 			log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",
Signature · 1 line
2.31.1
#1 lore raw
Quoted 27 lines · Jorge Ramirez-Ortiz

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg			: SPI Chip select
 
 Optional properties:
-- gpio-reset		: Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios		: Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency	: See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +0 -0
index 4b33ac8fd3..94ac52d9ce 100644

drivers/tpm/tpm2_tis_spi.c static int tpm_tis_spi_probe(struct udevice *dev) · line 595

Quoted 8 lines
drivers/tpm/tpm2_tis_spi.c +1 -1
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
 	if (CONFIG_IS_ENABLED(DM_GPIO)) {
 		struct gpio_desc reset_gpio;
 
-		ret = gpio_request_by_name(dev, "gpio-reset", 0,
+		ret = gpio_request_by_name(dev, "reset-gpios", 0,
 					   &reset_gpio, GPIOD_IS_OUT);
 		if (ret) {
 			log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",

I think you should deprecate gpio-reset but keep supporting that option with any warning and add code for reset-gpios.

Also would be good to add it as optional property to Linux kernel to keep it in sync.

Thanks,
Michal

#2 lore raw

Den tor. 27. maj 2021 kl. 09.15 skrev Michal Simek <[email protected]>:

Quoted 43 lines · Jorge Ramirez-Ortiz
Quoted 27 lines

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg                        : SPI Chip select
 
 Optional properties:
-- gpio-reset         : Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios                : Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency  : See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +0 -0
index 4b33ac8fd3..94ac52d9ce 100644

drivers/tpm/tpm2_tis_spi.c static int tpm_tis_spi_probe(struct udevice *dev) · line 595

Quoted 8 lines
drivers/tpm/tpm2_tis_spi.c +1 -1
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
      if (CONFIG_IS_ENABLED(DM_GPIO)) {
              struct gpio_desc reset_gpio;
 
-             ret = gpio_request_by_name(dev, "gpio-reset", 0,
+             ret = gpio_request_by_name(dev, "reset-gpios", 0,
                                         &reset_gpio, GPIOD_IS_OUT);
              if (ret) {
                      log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",

I think you should deprecate gpio-reset but keep supporting that option with any warning and add code for reset-gpios.

Quoted 2 lines

Also would be good to add it as optional property to Linux kernel to keep it in sync.

Hi

The reason the Linux kernel does not have a TPM reset signal, is that being able to reset the chip from software is a vulnerability. There was a discussion on it over on the Barebox mailing list a while ago.

TLDR: TPM reset needs to follow SOC reset.

/Bruno

#3 lore raw
Quoted 47 lines · Bruno Thomsen

Den tor. 27. maj 2021 kl. 09.15 skrev Michal Simek <[email protected]>:

Quoted 43 lines · Jorge Ramirez-Ortiz
Quoted 27 lines

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg                        : SPI Chip select
 
 Optional properties:
-- gpio-reset         : Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios                : Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency  : See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +0 -0
index 4b33ac8fd3..94ac52d9ce 100644

drivers/tpm/tpm2_tis_spi.c static int tpm_tis_spi_probe(struct udevice *dev) · line 595

Quoted 8 lines
drivers/tpm/tpm2_tis_spi.c +1 -1
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
      if (CONFIG_IS_ENABLED(DM_GPIO)) {
              struct gpio_desc reset_gpio;
 
-             ret = gpio_request_by_name(dev, "gpio-reset", 0,
+             ret = gpio_request_by_name(dev, "reset-gpios", 0,
                                         &reset_gpio, GPIOD_IS_OUT);
              if (ret) {
                      log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",

I think you should deprecate gpio-reset but keep supporting that option with any warning and add code for reset-gpios.

Quoted 2 lines

Also would be good to add it as optional property to Linux kernel to keep it in sync.

Quoted 8 lines

Hi

The reason the Linux kernel does not have a TPM reset signal, is that being able to reset the chip from software is a vulnerability. There was a discussion on it over on the Barebox mailing list a while ago.

TLDR: TPM reset needs to follow SOC reset.

I expect chip has the reset in both cases and it is just about who should be calling it. But we should be using the same DT for u-boot and Linux. It means it should be handled properly but described properly.

Thanks,
Michal

#4 lore raw
Quoted 57 lines · Michal Simek
Quoted 54 lines · Bruno Thomsen

Den tor. 27. maj 2021 kl. 09.15 skrev Michal Simek <[email protected]>:

Quoted 43 lines · Jorge Ramirez-Ortiz
Quoted 27 lines

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg                        : SPI Chip select
 
 Optional properties:
-- gpio-reset         : Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios                : Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency  : See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +0 -0
index 4b33ac8fd3..94ac52d9ce 100644

drivers/tpm/tpm2_tis_spi.c static int tpm_tis_spi_probe(struct udevice *dev) · line 595

Quoted 8 lines
drivers/tpm/tpm2_tis_spi.c +1 -1
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
      if (CONFIG_IS_ENABLED(DM_GPIO)) {
              struct gpio_desc reset_gpio;
 
-             ret = gpio_request_by_name(dev, "gpio-reset", 0,
+             ret = gpio_request_by_name(dev, "reset-gpios", 0,
                                         &reset_gpio, GPIOD_IS_OUT);
              if (ret) {
                      log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",

I think you should deprecate gpio-reset but keep supporting that option with any warning and add code for reset-gpios.

Quoted 2 lines

Also would be good to add it as optional property to Linux kernel to keep it in sync.

Hi

The reason the Linux kernel does not have a TPM reset signal, is that being able to reset the chip from software is a vulnerability. There was a discussion on it over on the Barebox mailing list a while ago.

Quoted 5 lines
Quoted 1 line

TLDR: TPM reset needs to follow SOC reset.

I expect chip has the reset in both cases and it is just about who should be calling it. But we should be using the same DT for u-boot and Linux. It means it should be handled properly but described properly.

right, I agree that it should be described properly (that was the patch intent).

but do we need to keep the legacy property?

Quoted 5 lines

Thanks,
Michal

#5 lore raw
Quoted 60 lines · Foundries
Quoted 59 lines · Michal Simek
Quoted 55 lines · Bruno Thomsen

Den tor. 27. maj 2021 kl. 09.15 skrev Michal Simek <[email protected]>:

Quoted 43 lines · Jorge Ramirez-Ortiz
Quoted 27 lines

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt +1 -1
index 3a2ee4bd17..bbcd12950f 100644
@@ -6,7 +6,7 @@ Required properties:
 - reg                        : SPI Chip select
 
 Optional properties:
-- gpio-reset         : Reset GPIO (if not connected to the SoC reset line)
+- reset-gpios                : Reset GPIO (if not connected to the SoC reset line)
 - spi-max-frequency  : See spi-bus.txt
 
 Example:
drivers/tpm/tpm2_tis_spi.c +0 -0
index 4b33ac8fd3..94ac52d9ce 100644

drivers/tpm/tpm2_tis_spi.c static int tpm_tis_spi_probe(struct udevice *dev) · line 595

Quoted 8 lines
drivers/tpm/tpm2_tis_spi.c +1 -1
@@ -589,7 +589,7 @@ static int tpm_tis_spi_probe(struct udevice *dev)
      if (CONFIG_IS_ENABLED(DM_GPIO)) {
              struct gpio_desc reset_gpio;
 
-             ret = gpio_request_by_name(dev, "gpio-reset", 0,
+             ret = gpio_request_by_name(dev, "reset-gpios", 0,
                                         &reset_gpio, GPIOD_IS_OUT);
              if (ret) {
                      log(LOGC_NONE, LOGL_NOTICE, "%s: missing reset GPIO\n",

I think you should deprecate gpio-reset but keep supporting that option with any warning and add code for reset-gpios.

Quoted 2 lines

Also would be good to add it as optional property to Linux kernel to keep it in sync.

Hi

The reason the Linux kernel does not have a TPM reset signal, is that being able to reset the chip from software is a vulnerability. There was a discussion on it over on the Barebox mailing list a while ago.

TLDR: TPM reset needs to follow SOC reset.

Quoted 8 lines
Quoted 3 lines

I expect chip has the reset in both cases and it is just about who should be calling it. But we should be using the same DT for u-boot and Linux. It means it should be handled properly but described properly.

right, I agree that it should be described properly (that was the patch intent).

but do we need to keep the legacy property?

I prefer all the time to have some time for transition. It means add support for new property. Keep there old one with message that this will be removed in near future. Not aware if there is any time defined. We normally keep it there for a year.

Thanks,
Michal

#7 lore raw
Quoted 11 lines · Simon Glass
Quoted 8 lines · Jorge Ramirez-Ortiz

Use the more generic reset-gpios propery name.

---
 doc/device-tree-bindings/tpm2/tis-tpm2-spi.txt | 2 +-
 drivers/tpm/tpm2_tis_spi.c                     | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

thanks.

should I add your reviewed-by by and resubmit the patch or do I also need to include the change proposed by Michal Simek (keeping the legacy property)?

TIA