Re: [PATCH] mmc: part_switch: fixes switch on gp3 partition

["Jorge Ramirez-Ortiz, Foundries" <jorge@foundries.io>]

On 06/03/24 14:18:49, Linus Walleij wrote:
> On Wed, Mar 6, 2024 at 10:05 AM Jorge Ramirez-Ortiz, Foundries
> <jorge@foundries.io> wrote:
>
> > That looked strange as there should be support for 4 GP but this code
> > kind of convinced me of the opposite.
> >
> >         if (idata->rpmb) {
> >                 /* Support multiple RPMB partitions */
> >                 target_part = idata->rpmb->part_index;
> >                 target_part |= EXT_CSD_PART_CONFIG_ACC_RPMB;
> >         }
> >
> > So if we apply the fix that you propose, how are multiple RPMB
> > partitions (ie, 4) going to be identified as RPMB? Unless there can't be
> > more than 3?
>
> As far as I can tell there can only be one RPMB partition per device.

that matches everything I have seen in the field too (and we have been
supporting RPMB on many designs lately (# > 30).

>
> The v5.1A spec says (section 6.2.1):
>
> "Two Boot Area Partitions, (...)"
> "One RPMB Partition accessed through a trusted mechanism, (...)"
> "Four General Purpose Area Partitions (...)"
>
> implying there can be only one RPMB.
>
> Also I have never seen more than one in practice.

+1

so I think it is safe to conclude that my commit did indeed cause these
regressions as it ignored the support for multiple GP. Sorry about it!.

I still cant grasp how "target_part = idata->rpmb->part_index" is
helping in the design.

What happens when:
1) EXT_CSD_PART_CONFIG_ACC_MASK > part_index > EXT_CSD_PART_CONFIG_ACC_RPMB
target_part now could be indicating a GP instead of an RPMB leading to failures.

2) part_index <= EXT_CSD_PART_CONFIG_ACC_RPMB
loses the part_index value .

So part_index should be larger than EXT_CSD_PART_CONFIG_ACC_MASK even
though the comment indicates it starts at 0?

/**
 * struct mmc_rpmb_data - special RPMB device type for these areas
 * @dev: the device for the RPMB area
 * @chrdev: character device for the RPMB area
 * @id: unique device ID number
 * @part_index: partition index (0 on first)    <---------------------
 * @md: parent MMC block device
 * @node: list item, so we can put this device on a list
 */
struct mmc_rpmb_data {
	struct device dev;
	struct cdev chrdev;
	int id;

is it just possible that "target_part = idata->rpmb->part_index" just
needs to be shifted to avoid issues?

I think the fix to the regression I introduced could perhaps address
this as well.