It might be useful to mention the following, based on discussions here:
- Describe that Adreno probe fails in a nondescriptive way, making it hard to have seen that it was as simple as SMMU being disabled
- With that it's useful to have the dmesg failure in your commit, but **without** any custom logging to drive the point home
- Describe (what was already suggested in review) that the convention appears to be that other SoCs don't disable nodes if they're always available and don't require any board-specific configuration (that might cause them to fail to probe otherwise)
- And perhaps that you based your board-specific Adreno enablement on other board dts', which don't enable `adreno_smmu` explicitly because of the previous point/convention