Re: [PATCH v3 03/10] drm/msm/dsi: Fix DSI index detection when version clash occurs

[Marijn Suijten]

On 2023-03-14 12:59:40, Konrad Dybcio wrote:
> 
> 
> On 14.03.2023 00:51, Marijn Suijten wrote:
> > On 2023-03-07 14:01:41, Konrad Dybcio wrote:
> >> Currently, we allow for MAX_DSI entries in io_start to facilitate for
> >> MAX_DSI number of DSI hosts at different addresses. The configuration
> >> is matched against the DSI CTRL hardware revision read back from the
> >> component. We need a way to resolve situations where multiple SoCs
> >> with different register maps may use the same version of DSI CTRL. In
> >> preparation to do so, make msm_dsi_config a 2d array where each entry
> >> represents a set of configurations adequate for a given SoC.
> > 
> > Note that this code isn't fool-proof against different SoCs sharing the
> > same DSI host address but for different indices (for example, the
> > address at variant 0 DSI 0 could be the same as variant 1 DSI 1) and the
> > matching logic would wrongly return ID 0 instead of 1 for SoC variant 1,
> > because that's the first matching address it finds.
> I don't think we've had that happen yet, but if it ever does, that's out
> of scope of this patchset.

Sure, as long as we're at least aware of this.

> >> This is totally fine to do, as the only differentiating factors
> >> between same-version-different-SoCs configurations are the number of
> >> DSI hosts (1 or 2, at least as of today) and the set of base registers.
> >> The regulator setup is the same, because the DSI hardware is the same,
> >> regardless of the SoC it was implemented in.
> >>
> >> In addition to that, update the matching logic such that it will loop
> >> over VARIANTS_MAX variants, making sure they are all taken into account.
> > 
> > "in addition to that" makes it sound like you're doing a separate new
> > thing in this patch, when the match logic must in fact be updated to
> > make it compatible with the change described above (as in, it doesn't
> > compile if you don't account for the extra depth in the array).
> I really think you're nitpicking here..

It's not, this genuinely had me confused for a while.  Could have at
least been addressed as part of v4 that had to be sent regardless.

- Marijn