[PATCH] tpm: tpm_tis: add settle delay after releasing locality

[PATCH] tpm: tpm_tis: add settle delay after releasing locality

[Daniel Golle]

tpm_tis_core_init() releases locality 0 then immediately reclaims it via
tpm_chip_start(); some TPMs (e.g. Nuvoton NPCT, TPM 2.0) need a few ms
before granting it again, so probe fails with -1. This back-to-back
release/request was added with the locality claim around TPM_INT_ENABLE.

Wait for the chip to settle after releasing the locality. A delay of
TPM_TIMEOUT (5 ms) in __tpm_tis_relinquish_locality() is reliable; values
below 3 ms are not.

Fixes: 0ef333f5ba7f ("tpm: add request_locality before write TPM_INT_ENABLE")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
---
 drivers/char/tpm/tpm_tis_core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
index 21d79ad3b164..6b90ff50c78d 100644
--- a/drivers/char/tpm/tpm_tis_core.c
+++ b/drivers/char/tpm/tpm_tis_core.c
@@ -171,6 +171,8 @@ static int __tpm_tis_relinquish_locality(struct tpm_tis_data *priv, int l)
 {
 	tpm_tis_write8(priv, TPM_ACCESS(l), TPM_ACCESS_ACTIVE_LOCALITY);
 
+	tpm_msleep(TPM_TIMEOUT);
+
 	return 0;
 }
 
-- 
2.54.0

Re: [PATCH] tpm: tpm_tis: add settle delay after releasing locality

[Jarkko Sakkinen]

On Mon, Jun 15, 2026 at 05:48:43AM +0100, Daniel Golle wrote:
> tpm_tis_core_init() releases locality 0 then immediately reclaims it via
> tpm_chip_start(); some TPMs (e.g. Nuvoton NPCT, TPM 2.0) need a few ms
> before granting it again, so probe fails with -1. This back-to-back
> release/request was added with the locality claim around TPM_INT_ENABLE.
> 
> Wait for the chip to settle after releasing the locality. A delay of
> TPM_TIMEOUT (5 ms) in __tpm_tis_relinquish_locality() is reliable; values
> below 3 ms are not.
> 
> Fixes: 0ef333f5ba7f ("tpm: add request_locality before write TPM_INT_ENABLE")
> Cc: stable@vger.kernel.org
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>
> ---
>  drivers/char/tpm/tpm_tis_core.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
> index 21d79ad3b164..6b90ff50c78d 100644
> --- a/drivers/char/tpm/tpm_tis_core.c
> +++ b/drivers/char/tpm/tpm_tis_core.c
> @@ -171,6 +171,8 @@ static int __tpm_tis_relinquish_locality(struct tpm_tis_data *priv, int l)
>  {
>  	tpm_tis_write8(priv, TPM_ACCESS(l), TPM_ACCESS_ACTIVE_LOCALITY);
>  
> +	tpm_msleep(TPM_TIMEOUT);
> +
>  	return 0;
>  }
>  
> -- 
> 2.54.0
> 

I think this is totally fine.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR, Jarkko

Re: [PATCH] tpm: tpm_tis: add settle delay after releasing locality

[Jarkko Sakkinen]

On Mon, Jun 15, 2026 at 05:48:43AM +0100, Daniel Golle wrote:
> tpm_tis_core_init() releases locality 0 then immediately reclaims it via
> tpm_chip_start(); some TPMs (e.g. Nuvoton NPCT, TPM 2.0) need a few ms
> before granting it again, so probe fails with -1. This back-to-back
> release/request was added with the locality claim around TPM_INT_ENABLE.
> 
> Wait for the chip to settle after releasing the locality. A delay of
> TPM_TIMEOUT (5 ms) in __tpm_tis_relinquish_locality() is reliable; values
> below 3 ms are not.
> 
> Fixes: 0ef333f5ba7f ("tpm: add request_locality before write TPM_INT_ENABLE")
> Cc: stable@vger.kernel.org
> Signed-off-by: Daniel Golle <daniel@makrotopia.org>
> ---
>  drivers/char/tpm/tpm_tis_core.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
> index 21d79ad3b164..6b90ff50c78d 100644
> --- a/drivers/char/tpm/tpm_tis_core.c
> +++ b/drivers/char/tpm/tpm_tis_core.c
> @@ -171,6 +171,8 @@ static int __tpm_tis_relinquish_locality(struct tpm_tis_data *priv, int l)
>  {
>  	tpm_tis_write8(priv, TPM_ACCESS(l), TPM_ACCESS_ACTIVE_LOCALITY);
>  
> +	tpm_msleep(TPM_TIMEOUT);
> +
>  	return 0;
>  }
>  
> -- 
> 2.54.0
> 

Can rebase to my for-next-tpm and resend v2?

BR, Jarkko

Re: [PATCH] tpm: tpm_tis: add settle delay after releasing locality

[Daniel Golle]

On Mon, Jun 15, 2026 at 03:17:43PM +0300, Jarkko Sakkinen wrote:
> On Mon, Jun 15, 2026 at 05:48:43AM +0100, Daniel Golle wrote:
> > tpm_tis_core_init() releases locality 0 then immediately reclaims it via
> > tpm_chip_start(); some TPMs (e.g. Nuvoton NPCT, TPM 2.0) need a few ms
> > before granting it again, so probe fails with -1. This back-to-back
> > release/request was added with the locality claim around TPM_INT_ENABLE.
> > 
> > Wait for the chip to settle after releasing the locality. A delay of
> > TPM_TIMEOUT (5 ms) in __tpm_tis_relinquish_locality() is reliable; values
> > below 3 ms are not.
> > 
> > Fixes: 0ef333f5ba7f ("tpm: add request_locality before write TPM_INT_ENABLE")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Daniel Golle <daniel@makrotopia.org>
> 
> Can rebase to my for-next-tpm and resend v2?

I figured the fix is already in place in your tree, added by
34bc0fabf166 ("tpm: tpm_tis: Add settle time for some TPMs")

device_id 0x00FE, vendor_id 0x1050 exactly matches my TPM as well,
so please drop my patch.

Sorry for the noise...