2024-04-09 01:59:10

by Pavan Holla

[permalink] [raw]
Subject: [PATCH v3] usb: typec: ucsi: Wait 20ms before reading CCI after a reset

The PPM might take time to process a reset. Allow 20ms for the reset to
be processed before reading the CCI.

This should not slow down existing implementations because they would
not set any bits in the CCI after a reset, and would take a 20ms delay
to read the CCI anyway. This change makes the delay explicit, and
reduces a CCI read. Based on the spec, the PPM has 10ms to set busy, so,
20ms seems like a reasonable delay before we read the CCI.

Signed-off-by: Pavan Holla <[email protected]>
Reviewed-by: Prashant Malani <[email protected]>
---
The PPM might take time to process a reset and set the CCI. Give the PPM
20ms to update it's CCI.

Commit message now summarizes discussion in v1.
---
Changes in v3:
- Commit message updated.
- Link to v2: https://lore.kernel.org/r/[email protected]

Changes in v2:
- Commit message updated.
- Link to v1: https://lore.kernel.org/r/[email protected]
---
drivers/usb/typec/ucsi/ucsi.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c
index cf52cb34d285..c0706c40fa64 100644
--- a/drivers/usb/typec/ucsi/ucsi.c
+++ b/drivers/usb/typec/ucsi/ucsi.c
@@ -1280,6 +1280,9 @@ static int ucsi_reset_ppm(struct ucsi *ucsi)
goto out;
}

+ /* Give the PPM time to process a reset before reading CCI */
+ msleep(20);
+
ret = ucsi->ops->read(ucsi, UCSI_CCI, &cci, sizeof(cci));
if (ret)
goto out;
@@ -1293,7 +1296,6 @@ static int ucsi_reset_ppm(struct ucsi *ucsi)
goto out;
}

- msleep(20);
} while (!(cci & UCSI_CCI_RESET_COMPLETE));

out:

---
base-commit: 4cece764965020c22cff7665b18a012006359095
change-id: 20240325-ucsi-reset-delay-bdf6712455fd

Best regards,
--
Pavan Holla <[email protected]>



2024-04-09 10:38:07

by Heikki Krogerus

[permalink] [raw]
Subject: Re: [PATCH v3] usb: typec: ucsi: Wait 20ms before reading CCI after a reset

On Tue, Apr 09, 2024 at 01:58:57AM +0000, Pavan Holla wrote:
> The PPM might take time to process a reset. Allow 20ms for the reset to
> be processed before reading the CCI.
>
> This should not slow down existing implementations because they would
> not set any bits in the CCI after a reset, and would take a 20ms delay
> to read the CCI anyway. This change makes the delay explicit, and
> reduces a CCI read. Based on the spec, the PPM has 10ms to set busy, so,
> 20ms seems like a reasonable delay before we read the CCI.
>
> Signed-off-by: Pavan Holla <[email protected]>
> Reviewed-by: Prashant Malani <[email protected]>

Reviewed-by: Heikki Krogerus <[email protected]>

> ---
> The PPM might take time to process a reset and set the CCI. Give the PPM
> 20ms to update it's CCI.
>
> Commit message now summarizes discussion in v1.
> ---
> Changes in v3:
> - Commit message updated.
> - Link to v2: https://lore.kernel.org/r/[email protected]
>
> Changes in v2:
> - Commit message updated.
> - Link to v1: https://lore.kernel.org/r/[email protected]
> ---
> drivers/usb/typec/ucsi/ucsi.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/typec/ucsi/ucsi.c b/drivers/usb/typec/ucsi/ucsi.c
> index cf52cb34d285..c0706c40fa64 100644
> --- a/drivers/usb/typec/ucsi/ucsi.c
> +++ b/drivers/usb/typec/ucsi/ucsi.c
> @@ -1280,6 +1280,9 @@ static int ucsi_reset_ppm(struct ucsi *ucsi)
> goto out;
> }
>
> + /* Give the PPM time to process a reset before reading CCI */
> + msleep(20);
> +
> ret = ucsi->ops->read(ucsi, UCSI_CCI, &cci, sizeof(cci));
> if (ret)
> goto out;
> @@ -1293,7 +1296,6 @@ static int ucsi_reset_ppm(struct ucsi *ucsi)
> goto out;
> }
>
> - msleep(20);
> } while (!(cci & UCSI_CCI_RESET_COMPLETE));
>
> out:
>
> ---
> base-commit: 4cece764965020c22cff7665b18a012006359095
> change-id: 20240325-ucsi-reset-delay-bdf6712455fd
>
> Best regards,
> --
> Pavan Holla <[email protected]>

--
heikki