Hi Aidan, On 2026-05-13T00:26:04, Aidan Garske <[email protected]> wrote: > tpm: export tpm_show_device, tpm_set_device, and get_tpm > > Remove static scope from tpm_show_device(), tpm_set_device(), and > get_tpm() in cmd/tpm-common.c so they can be called from other > translation units. Add corresponding declarations to > include/tpm-common.h. > > wolfTPM's command backend needs these functions for device enumeration > and selection when operating through U-Boot's driver model. > > Signed-off-by: Aidan Garske <[email protected]> > > cmd/tpm-common.c | 4 ++-- > include/tpm-common.h | 22 ++++++++++++++++++++++ > 2 files changed, 24 insertions(+), 2 deletions(-)
> diff --git a/cmd/tpm-common.c b/cmd/tpm-common.c > @@ -234,7 +234,7 @@ int type_string_write_vars(const char *type_str, u8 *data, > return 0; > } > > -static int tpm_show_device(void) > +int tpm_show_device(void) These three helpers live in cmd/ and operate on a static tpm_dev private to the command layer. Promoting them so a backend in lib/ or drivers/ can call into cmd/ is the wrong direction. If the wolfTPM backend genuinely needs shared 'currently selected TPM' state, please move that state and these helpers into the TPM uclass and have cmd/tpm-common.c call into them, not the other way round. > diff --git a/include/tpm-common.h b/include/tpm-common.h > @@ -337,4 +337,26 @@ enum tpm_version tpm_get_version(struct udevice *dev); > +/** > + * tpm_show_device() - Show all TPM devices > + * > + * Return: 0 on success, -ve on failure > + */ > +int tpm_show_device(void); > + > +/** > + * tpm_set_device() - Set the TPM device to use > + * > + * @num: The number of the TPM device to use > + * Return: 0 on success, -ve on failure > + */ > +int tpm_set_device(unsigned long num); > + > +/** > + * get_tpm() - Get the TPM device > + * > + * Return: 0 on success, -ve on failure > + */ > +int get_tpm(struct udevice **devp); The kerneldoc is wrong. tpm_set_device() and get_tpm() return CMD_RET_FAILURE (1, positive), not -ve. tpm_show_device() ignores per-device errors and unconditionally returns 0. Either document accurately, or make tpm_show_device() propagate the failure from tpm_get_desc(). If these become public API, please also fix the return convention to use proper negative errnos (CMD_RET_* is a command-layer thing) and rename get_tpm() - tpm_get_device() or similar. get_tpm() is too generic for a global symbol. Also, @devp is undocumented. Please add an entry for it, and note that the function selects device 0 the first time it is called. Regards, Simon
