Hi Simon,
On 13/08/2015 03:30, Simon Glass wrote:
Hi Christophe,
On 11 August 2015 at 15:47, christophe.ricard christophe.ricard@gmail.com wrote:
Hi Simon,
Locality concept are valid almost on any chip assuming if no locality are supported the default one is locality 0. I would leave this change open for discussion.
However, as per patch 06 & 07, i would keep req_complete_mask, req_complete_val, req_canceled, timeout_a, timeout_b, timeout_c, timeout_d in tpm_vendor_specific structure as this is chip specific.
I really think tpm_vendor_specific is usefull for managing different kind of TPM "the same way"/following standards.
That code belongs in the uclass I think. If there really are generic settings that are needed for all TPMs then it should sit there. We don't want to have an additional layer of stuff that doesn't relate to driver model.
After reviewing your previous comments, i think we can drop this tpm_vendor_specific structure to simplify the code a bit. However, the work we are doing may stick only to TPM1.2. I think it will be fine as we have only drivers for those kind of TPMs. I believe a new uclass may be necessary when going to provide support TPM 2.0.
In short, may be we can anticipate that and make it explicit in the uclass name ? (UCLASS_TPM12 ?)
[...]
Best Regards Christophe