[RFC PATCH v2 3/8] FWU: stm32mp1: Add helper functions for accessing FWU metadata

[Sughosh Ganu]

hi Masami,

On Fri, 24 Dec 2021 at 11:19, Masami Hiramatsu
<masami.hiramatsu at linaro.org> wrote:
>
> () should Hi Sughosh and Ilias,
>
> I would like to confirm that what the plat_fill_gpt_partition_guids()
> should return.
>
> 2021年12月19日(日) 16:07 Sughosh Ganu <sughosh.ganu at linaro.org>:
>
> > +static int plat_fill_gpt_partition_guids(struct blk_desc *desc,
> > +                                        efi_guid_t **part_guid_arr)
> > +{
> > +       int i, ret = 0;
> > +       u32 part;
> > +       struct dfu_entity *dfu;
> > +       struct disk_partition info;
> > +       efi_guid_t part_type_guid;
> > +       int alt_num = dfu_get_alt_number();
> > +
> > +       dfu_init_env_entities(NULL, NULL);
> > +
> > +       for (i = 0, part = 1; i < alt_num; i++) {
> > +               dfu = dfu_get_entity(i);
> > +
> > +               if (!dfu)
> > +                       continue;
> > +
> > +               /*
> > +                * Currently, Multi Bank update
> > +                * feature is being supported
> > +                * only on GPT partitioned
> > +                * MMC/SD devices.
> > +                */
> > +               if (dfu->dev_type != DFU_DEV_MMC)
> > +                       continue;
> > +
> > +               if (part_get_info(desc, part, &info)) {
> > +                       part++;
> > +                       continue;
> > +               }
> > +
> > +               uuid_str_to_bin(info.type_guid, part_type_guid.b,
> > +                               UUID_STR_FORMAT_GUID);
> > +               guidcpy((*part_guid_arr + i), &part_type_guid);
> > +               part++;
> > +       }
> > +
> > +       dfu_free_entities();
> > +
> > +       return ret;
> > +}
>
> So on this platform which uses GPT partition for each bank, it stores
> the GUID for each partition.
> I think this information will be shown in the ESRT, but not used for
> specifying capsule image_type_uuid.
>
> This is strange.
> Without this FWU multi-bank update, the array will only have the
> image_type UUID.
> Thus,  user can use ESRT to confirm what UUID should be specified for
> the capsule file. (it has image-type UUID)
> (of course that is fixed value on one platform anyway)
>
> And when FWU multi-bank update is enabled, that changes to the
> partition GUIDs. But those GUIDs are NOT used when making the capsule
> file nor doing the update process, because we should not specify the
> bank index directly, right?

The partition GUIDs are the same as ImageTypeId field in the
EFI_FIRMWARE_IMAGE_DESCRIPTOR and the UpdateImageTypeId field in the
EFI_FIRMWARE_MANAGEMENT_CAPSULE_IMAGE_HEADER. This value is the same
as the PartitionTypeGUID field in the GPT partition entry.

ImageTypeId  = UpdateImageTypeId = PartitionTypeGUID

This is the GUID value that is used when generating the capsule.

>
> We will set a unique image-type UUID on each image entry, and use it
> for making a capsule file,

Even with the FWU feature enabled, the PartitionTypeGUID/ImageTypeId
is used for making the capsule. What changes for the FWU feature is
that the value of UpdateImageIndex/ImageIndex that is passed to the
capsule generation tool becomes irrelevant. When the FWU feature is
not enabled, the ImageIndex value that is passed to the capsule
generation tool is important, since this is supposed to correspond to
the DFU alt number -- this ImageIndex value is then used to identify
the partition/address to which the capsule payload is to be written.
With the FWU feature enabled, this value is not obtained from the
capsule, since the partition/address to write the image to is
identified at runtime based on the value of the update bank.

> but the capsule file itself doesn't specify
> the banks. And acceptance/revert capsule file also doesn't specify it.
> Acceptance file will also uses image_type UUID, but not the partition
> UUID for the bank.

The image acceptance capsule will also contain the ImageTypeId. The
Accept bit will be set for the active bank image of the image type
specified in the acceptance capsule.

-sughosh

>
> Maybe I'm wrong, but I'm confusing. And this is important for the
> platform which doesn't use GPT.
>
> Thank you,
>
>
> --
> Masami Hiramatsu