[PATCH net 0/3] broadcom: report the supported flags for ancillary features

[PATCH net 0/3] broadcom: report the supported flags for ancillary features

[Jacob Keller]

James Clark reported off list that the broadcom PHY PTP driver was
incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
the conversion to the .supported_*_flags fields. This series fixes the
driver to correctly report its flags through the .supported_perout_flags
and .supported_extts_flags fields. It also contains an update to comment
the behavior of the PTP_STRICT_FLAGS being always enabled for
PTP_EXTTS_REQUEST2.

I plan to follow up this series with some improvements to the PTP
documentation better explaining each flag and the expectation of the driver
APIs.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
---
Jacob Keller (3):
      broadcom: fix support for PTP_PEROUT_DUTY_CYCLE
      broadcom: fix support for PTP_EXTTS_REQUEST2 ioctl
      ptp: document behavior of PTP_STRICT_FLAGS

 include/uapi/linux/ptp_clock.h | 3 +++
 drivers/net/phy/bcm-phy-ptp.c  | 6 ++----
 2 files changed, 5 insertions(+), 4 deletions(-)
---
base-commit: cbf658dd09419f1ef9de11b9604e950bdd5c170b
change-id: 20250918-jk-fix-bcm-phy-supported-flags-0796dddf7954

Best regards,
--  
Jacob Keller <jacob.e.keller@intel.com>

[PATCH net 1/3] broadcom: fix support for PTP_PEROUT_DUTY_CYCLE

[Jacob Keller]

The bcm_ptp_perout_locked() function has support for handling
PTP_PEROUT_DUTY_CYCLE, but its not listed in the supported_perout_flags.
Attempts to use the duty cycle support will be rejected since commit
d9f3e9ecc456 ("net: ptp: introduce .supported_perout_flags to
ptp_clock_info"), as this flag accidentally missed while doing the
conversion.

Drop the unnecessary supported flags check from the bcm_ptp_perout_locked()
function and correctly set the supported_perout_flags. This fixes use of
the PTP_PEROUT_DUTY_CYCLE support for the broadcom driver.

Reported-by: James Clark <jjc@jclark.com>
Fixes: d9f3e9ecc456 ("net: ptp: introduce .supported_perout_flags to ptp_clock_info")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/phy/bcm-phy-ptp.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/net/phy/bcm-phy-ptp.c b/drivers/net/phy/bcm-phy-ptp.c
index eba8b5fb1365..1cf695ac73cc 100644
--- a/drivers/net/phy/bcm-phy-ptp.c
+++ b/drivers/net/phy/bcm-phy-ptp.c
@@ -597,10 +597,6 @@ static int bcm_ptp_perout_locked(struct bcm_ptp_private *priv,
 
 	period = BCM_MAX_PERIOD_8NS;	/* write nonzero value */
 
-	/* Reject unsupported flags */
-	if (req->flags & ~PTP_PEROUT_DUTY_CYCLE)
-		return -EOPNOTSUPP;
-
 	if (req->flags & PTP_PEROUT_DUTY_CYCLE)
 		pulse = ktime_to_ns(ktime_set(req->on.sec, req->on.nsec));
 	else
@@ -741,6 +737,7 @@ static const struct ptp_clock_info bcm_ptp_clock_info = {
 	.n_pins		= 1,
 	.n_per_out	= 1,
 	.n_ext_ts	= 1,
+	.supported_perout_flags = PTP_PEROUT_DUTY_CYCLE,
 };
 
 static void bcm_ptp_txtstamp(struct mii_timestamper *mii_ts,

-- 
2.51.0.rc1.197.g6d975e95c9d7

[PATCH net 2/3] broadcom: fix support for PTP_EXTTS_REQUEST2 ioctl

[Jacob Keller]

Commit 7c571ac57d9d ("net: ptp: introduce .supported_extts_flags to
ptp_clock_info") modified the PTP core kernel logic to validate the
supported flags for the PTP_EXTTS_REQUEST ioctls, rather than relying on
each individual driver correctly checking its flags.

The bcm_ptp_enable() function implements support for PTP_CLK_REQ_EXTTS, but
does not check the flags, and does not forward the request structure into
bcm_ptp_extts_locked().

When originally converting the bcm-phy-ptp.c code, it was unclear what
edges the hardware actually timestamped. Thus, no flags were initialized in
the .supported_extts_flags field. This results in the kernel automatically
rejecting all userspace requests for the PTP_EXTTS_REQUEST2 ioctl.

This occurs because the PTP_STRICT_FLAGS is always assumed when operating
under PTP_EXTTS_REQUEST2. This has been the case since the flags
introduction by commit 6138e687c7b6 ("ptp: Introduce strict checking of
external time stamp options.").

The bcm-phy-ptp.c logic never properly supported strict flag validation,
as it previously ignored all flags including both PTP_STRICT_FLAGS and the
PTP_FALLING_EDGE and PTP_RISING_EDGE flags.

Reports from users in the field prove that the hardware timestamps the
rising edge. Encode this in the .supported_extts_flags field. This
re-enables support for the PTP_EXTTS_REQUEST2 ioctl.

Reported-by: James Clark <jjc@jclark.com>
Fixes: 7c571ac57d9d ("net: ptp: introduce .supported_extts_flags to ptp_clock_info")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/phy/bcm-phy-ptp.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/phy/bcm-phy-ptp.c b/drivers/net/phy/bcm-phy-ptp.c
index 1cf695ac73cc..d3501f8487d9 100644
--- a/drivers/net/phy/bcm-phy-ptp.c
+++ b/drivers/net/phy/bcm-phy-ptp.c
@@ -738,6 +738,7 @@ static const struct ptp_clock_info bcm_ptp_clock_info = {
 	.n_per_out	= 1,
 	.n_ext_ts	= 1,
 	.supported_perout_flags = PTP_PEROUT_DUTY_CYCLE,
+	.supported_extts_flags = PTP_STRICT_FLAGS | PTP_RISING_EDGE,
 };
 
 static void bcm_ptp_txtstamp(struct mii_timestamper *mii_ts,

-- 
2.51.0.rc1.197.g6d975e95c9d7

[PATCH net 3/3] ptp: document behavior of PTP_STRICT_FLAGS

[Jacob Keller]

Commit 6138e687c7b6 ("ptp: Introduce strict checking of external time stamp
options.") added the PTP_STRICT_FLAGS to the set of flags supported for the
external timestamp request ioctl.

It is only supported by PTP_EXTTS_REQUEST2, as it was introduced the
introduction of the new ioctls. Further, the kernel has always set this
flag for PTP_EXTTS_REQUEST2 regardless of whether or not the user requested
the behavior.

This effectively means that the flag is not useful for userspace. If the
user issues a PTP_EXTTS_REQUEST ioctl, the flag is ignored due to not being
supported on the old ioctl. If the user issues a PTP_EXTTS_REQUEST2 ioctl,
the flag will be set by the kernel regardless of whether the user set the
flag in their structure.

Add a comment documenting this behavior in the uAPI header file.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/uapi/linux/ptp_clock.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/uapi/linux/ptp_clock.h b/include/uapi/linux/ptp_clock.h
index 18eefa6d93d6..2c3346e91dbe 100644
--- a/include/uapi/linux/ptp_clock.h
+++ b/include/uapi/linux/ptp_clock.h
@@ -37,6 +37,9 @@
 
 /*
  * flag fields valid for the new PTP_EXTTS_REQUEST2 ioctl.
+ *
+ * Note: PTP_STRICT_FLAGS is always enabled by the kernel for
+ * PTP_EXTTS_REQUEST2 regardless of whether it is set by userspace.
  */
 #define PTP_EXTTS_VALID_FLAGS	(PTP_ENABLE_FEATURE |	\
 				 PTP_RISING_EDGE |	\

-- 
2.51.0.rc1.197.g6d975e95c9d7

Re: [PATCH net 0/3] broadcom: report the supported flags for ancillary features

[Richard Cochran]

On Thu, Sep 18, 2025 at 05:33:15PM -0700, Jacob Keller wrote:
> James Clark reported off list that the broadcom PHY PTP driver was
> incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
> the conversion to the .supported_*_flags fields. This series fixes the
> driver to correctly report its flags through the .supported_perout_flags
> and .supported_extts_flags fields. It also contains an update to comment
> the behavior of the PTP_STRICT_FLAGS being always enabled for
> PTP_EXTTS_REQUEST2.
> 
> I plan to follow up this series with some improvements to the PTP
> documentation better explaining each flag and the expectation of the driver
> APIs.
> 
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>

For series:

Acked-by: Richard Cochran <richardcochran@gmail.com>

Re: [PATCH net 0/3] broadcom: report the supported flags for ancillary features

[Kory Maincent]

On Thu, 18 Sep 2025 17:33:15 -0700
Jacob Keller <jacob.e.keller@intel.com> wrote:

> James Clark reported off list that the broadcom PHY PTP driver was
> incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
> the conversion to the .supported_*_flags fields. This series fixes the
> driver to correctly report its flags through the .supported_perout_flags
> and .supported_extts_flags fields. It also contains an update to comment
> the behavior of the PTP_STRICT_FLAGS being always enabled for
> PTP_EXTTS_REQUEST2.
> 
> I plan to follow up this series with some improvements to the PTP
> documentation better explaining each flag and the expectation of the driver
> APIs.
> 
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>

Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>

Thank you!
-- 
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

Re: [PATCH net 1/3] broadcom: fix support for PTP_PEROUT_DUTY_CYCLE

[Vadim Fedorenko]

On 19/09/2025 01:33, Jacob Keller wrote:
> The bcm_ptp_perout_locked() function has support for handling
> PTP_PEROUT_DUTY_CYCLE, but its not listed in the supported_perout_flags.
> Attempts to use the duty cycle support will be rejected since commit
> d9f3e9ecc456 ("net: ptp: introduce .supported_perout_flags to
> ptp_clock_info"), as this flag accidentally missed while doing the
> conversion.
> 
> Drop the unnecessary supported flags check from the bcm_ptp_perout_locked()
> function and correctly set the supported_perout_flags. This fixes use of
> the PTP_PEROUT_DUTY_CYCLE support for the broadcom driver.
> 
> Reported-by: James Clark <jjc@jclark.com>
> Fixes: d9f3e9ecc456 ("net: ptp: introduce .supported_perout_flags to ptp_clock_info")
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
> ---
>   drivers/net/phy/bcm-phy-ptp.c | 5 +----
>   1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/drivers/net/phy/bcm-phy-ptp.c b/drivers/net/phy/bcm-phy-ptp.c
> index eba8b5fb1365..1cf695ac73cc 100644
> --- a/drivers/net/phy/bcm-phy-ptp.c
> +++ b/drivers/net/phy/bcm-phy-ptp.c
> @@ -597,10 +597,6 @@ static int bcm_ptp_perout_locked(struct bcm_ptp_private *priv,
>   
>   	period = BCM_MAX_PERIOD_8NS;	/* write nonzero value */
>   
> -	/* Reject unsupported flags */
> -	if (req->flags & ~PTP_PEROUT_DUTY_CYCLE)
> -		return -EOPNOTSUPP;
> -
>   	if (req->flags & PTP_PEROUT_DUTY_CYCLE)
>   		pulse = ktime_to_ns(ktime_set(req->on.sec, req->on.nsec));
>   	else
> @@ -741,6 +737,7 @@ static const struct ptp_clock_info bcm_ptp_clock_info = {
>   	.n_pins		= 1,
>   	.n_per_out	= 1,
>   	.n_ext_ts	= 1,
> +	.supported_perout_flags = PTP_PEROUT_DUTY_CYCLE,
>   };
>   
>   static void bcm_ptp_txtstamp(struct mii_timestamper *mii_ts,
> 

Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>

Re: [PATCH net 2/3] broadcom: fix support for PTP_EXTTS_REQUEST2 ioctl

[Vadim Fedorenko]

On 19/09/2025 01:33, Jacob Keller wrote:
> Commit 7c571ac57d9d ("net: ptp: introduce .supported_extts_flags to
> ptp_clock_info") modified the PTP core kernel logic to validate the
> supported flags for the PTP_EXTTS_REQUEST ioctls, rather than relying on
> each individual driver correctly checking its flags.
> 
> The bcm_ptp_enable() function implements support for PTP_CLK_REQ_EXTTS, but
> does not check the flags, and does not forward the request structure into
> bcm_ptp_extts_locked().
> 
> When originally converting the bcm-phy-ptp.c code, it was unclear what
> edges the hardware actually timestamped. Thus, no flags were initialized in
> the .supported_extts_flags field. This results in the kernel automatically
> rejecting all userspace requests for the PTP_EXTTS_REQUEST2 ioctl.
> 
> This occurs because the PTP_STRICT_FLAGS is always assumed when operating
> under PTP_EXTTS_REQUEST2. This has been the case since the flags
> introduction by commit 6138e687c7b6 ("ptp: Introduce strict checking of
> external time stamp options.").
> 
> The bcm-phy-ptp.c logic never properly supported strict flag validation,
> as it previously ignored all flags including both PTP_STRICT_FLAGS and the
> PTP_FALLING_EDGE and PTP_RISING_EDGE flags.
> 
> Reports from users in the field prove that the hardware timestamps the
> rising edge. Encode this in the .supported_extts_flags field. This
> re-enables support for the PTP_EXTTS_REQUEST2 ioctl.
> 
> Reported-by: James Clark <jjc@jclark.com>
> Fixes: 7c571ac57d9d ("net: ptp: introduce .supported_extts_flags to ptp_clock_info")
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
> ---
>   drivers/net/phy/bcm-phy-ptp.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/phy/bcm-phy-ptp.c b/drivers/net/phy/bcm-phy-ptp.c
> index 1cf695ac73cc..d3501f8487d9 100644
> --- a/drivers/net/phy/bcm-phy-ptp.c
> +++ b/drivers/net/phy/bcm-phy-ptp.c
> @@ -738,6 +738,7 @@ static const struct ptp_clock_info bcm_ptp_clock_info = {
>   	.n_per_out	= 1,
>   	.n_ext_ts	= 1,
>   	.supported_perout_flags = PTP_PEROUT_DUTY_CYCLE,
> +	.supported_extts_flags = PTP_STRICT_FLAGS | PTP_RISING_EDGE,
>   };
>   
>   static void bcm_ptp_txtstamp(struct mii_timestamper *mii_ts,
> 

Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>

Re: [PATCH net 3/3] ptp: document behavior of PTP_STRICT_FLAGS

[Vadim Fedorenko]

On 19/09/2025 01:33, Jacob Keller wrote:
> Commit 6138e687c7b6 ("ptp: Introduce strict checking of external time stamp
> options.") added the PTP_STRICT_FLAGS to the set of flags supported for the
> external timestamp request ioctl.
> 
> It is only supported by PTP_EXTTS_REQUEST2, as it was introduced the
> introduction of the new ioctls. Further, the kernel has always set this
> flag for PTP_EXTTS_REQUEST2 regardless of whether or not the user requested
> the behavior.
> 
> This effectively means that the flag is not useful for userspace. If the
> user issues a PTP_EXTTS_REQUEST ioctl, the flag is ignored due to not being
> supported on the old ioctl. If the user issues a PTP_EXTTS_REQUEST2 ioctl,
> the flag will be set by the kernel regardless of whether the user set the
> flag in their structure.
> 
> Add a comment documenting this behavior in the uAPI header file.
> 
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
> ---
>   include/uapi/linux/ptp_clock.h | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/include/uapi/linux/ptp_clock.h b/include/uapi/linux/ptp_clock.h
> index 18eefa6d93d6..2c3346e91dbe 100644
> --- a/include/uapi/linux/ptp_clock.h
> +++ b/include/uapi/linux/ptp_clock.h
> @@ -37,6 +37,9 @@
>   
>   /*
>    * flag fields valid for the new PTP_EXTTS_REQUEST2 ioctl.
> + *
> + * Note: PTP_STRICT_FLAGS is always enabled by the kernel for
> + * PTP_EXTTS_REQUEST2 regardless of whether it is set by userspace.
>    */
>   #define PTP_EXTTS_VALID_FLAGS	(PTP_ENABLE_FEATURE |	\
>   				 PTP_RISING_EDGE |	\
> 

Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>

Re: [PATCH net 0/3] broadcom: report the supported flags for ancillary features

[Jakub Kicinski]

On Thu, 18 Sep 2025 17:33:15 -0700 Jacob Keller wrote:
> James Clark reported off list that the broadcom PHY PTP driver was
> incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
> the conversion to the .supported_*_flags fields. This series fixes the
> driver to correctly report its flags through the .supported_perout_flags
> and .supported_extts_flags fields. It also contains an update to comment
> the behavior of the PTP_STRICT_FLAGS being always enabled for
> PTP_EXTTS_REQUEST2.

James, would you be willing to test and send an official Tested-by tag?

Re: [PATCH net 0/3] broadcom: report the supported flags for ancillary features

[James Clark]

On Sat, Sep 20, 2025 at 7:04 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Thu, 18 Sep 2025 17:33:15 -0700 Jacob Keller wrote:
> > James Clark reported off list that the broadcom PHY PTP driver was
> > incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
> > the conversion to the .supported_*_flags fields. This series fixes the
> > driver to correctly report its flags through the .supported_perout_flags
> > and .supported_extts_flags fields. It also contains an update to comment
> > the behavior of the PTP_STRICT_FLAGS being always enabled for
> > PTP_EXTTS_REQUEST2.
>
> James, would you be willing to test and send an official Tested-by tag?

I have tested this on a Raspberry Pi CM4: PTP_EXTTS_REQUEST2 and
PTP_PEROUT_REQUEST2 both work. Thanks!

Going forward, I have got myself setup to be able to build and install
a vanilla kernel on a CM4, so if there is stuff that needs testing on
bcm-phy-ptp, just Cc me.

Tested-by: James Clark <jjc@jclark.com>

Re: [PATCH net 0/3] broadcom: report the supported flags for ancillary features

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Thu, 18 Sep 2025 17:33:15 -0700 you wrote:
> James Clark reported off list that the broadcom PHY PTP driver was
> incorrectly handling PTP_EXTTS_REQUEST and PTP_PEROUT_REQUEST ioctls since
> the conversion to the .supported_*_flags fields. This series fixes the
> driver to correctly report its flags through the .supported_perout_flags
> and .supported_extts_flags fields. It also contains an update to comment
> the behavior of the PTP_STRICT_FLAGS being always enabled for
> PTP_EXTTS_REQUEST2.
> 
> [...]

Here is the summary with links:
  - [net,1/3] broadcom: fix support for PTP_PEROUT_DUTY_CYCLE
    https://git.kernel.org/netdev/net/c/6e6c88d85623
  - [net,2/3] broadcom: fix support for PTP_EXTTS_REQUEST2 ioctl
    https://git.kernel.org/netdev/net/c/3200fdd4021d
  - [net,3/3] ptp: document behavior of PTP_STRICT_FLAGS
    https://git.kernel.org/netdev/net/c/cd875625b475

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html