Claude review: drm/msm/dsi: support DSC configurations with slice_per_pkt > 1

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Concept**: Replaces the hardcoded `slice_per_pkt = 1` assumption in the MSM DSI host with support for configurable values. A `dsc_slice_per_pkt` field is added to `struct drm_dsc_config`.

**Issues:**

1. **Division by zero risk**: In `dsi_update_dsc_timing()`, the new code does:
   ```c
   pkt_per_line = slice_per_intf / msm_host->dsc_slice_per_pkt;
   ```
   If `dsc_slice_per_pkt` is 0 (e.g. a panel driver that sets `dsc` but doesn't set this new field, and the `?: 1` fallback in `dsi_host_attach` is somehow bypassed or a different code path sets the DSC config), this is an OOPS. The fallback at attach time (`dsi->dsc->dsc_slice_per_pkt ?: 1`) is good but fragile — consider also clamping or validating in `dsi_update_dsc_timing` itself.

2. **Questionable placement in `drm_dsc_config`**: The comment says "This is not part of DSC standard, and only used in some DSI panels so far." Adding transport-layer packetization config to a struct representing the DSC standard parameter set conflates concerns. The existing `msm_dsi_host` already has a local `dsc_slice_per_pkt` field — you could keep it purely there, and pass it from `mipi_dsi_device` via a separate field rather than polluting `drm_dsc_config`. This needs buy-in from DRM core maintainers (Maxime, Thomas, etc.).

3. **Naming redundancy**: The field is called `dsc_slice_per_pkt` inside `struct drm_dsc_config`. Since it's already in a `dsc` struct, just `slice_per_pkt` would be sufficient and consistent with how other fields like `slice_count`, `slice_width` etc. are named.

4. **The `pkt_per_line` log2 encoding check**: The existing code at line 964 warns if `pkt_per_line > 4` because the register field is log2-encoded and only supports 1, 2, 4. With `slice_per_pkt > 1`, `pkt_per_line` can now be smaller (e.g., 4 slices / 4 slice_per_pkt = 1), but there's no validation that `slice_per_intf` is evenly divisible by `dsc_slice_per_pkt`, which could give non-power-of-2 results.

---
Generated by Claude Code Patch Reviewer