Claude review: drm/dumb-buffers: document that it's only for linear FB

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

Patch Review

**Content review:**

The added text:
```c
 * a hardware-specific ioctl to allocate suitable buffer objects. They are
 * also currently meant for only linear buffers, and using them with any
 * modifier other than DRM_FORMAT_MOD_LINEAR is undefined behavior. There
 * exist some KMS drivers abusing dumb objects for AFBC framebuffers, but this
 * behavior is discouraged, only exists as a hack now and shouldn't be
 * replicated.
```

A few observations:

1. **"undefined behavior" is misleading.** In the kernel context, "undefined behavior" has a very specific C-language connotation. Dumb buffers don't take a modifier argument at all — the `drm_mode_create_dumb` ioctl has `height`, `width`, `bpp`, `flags`, `handle`, `pitch`, and `size`. There's no modifier field. The issue isn't that passing a non-linear modifier is "undefined behavior"; it's that the interface has no mechanism to request or communicate a non-linear layout. Consider wording like: "The dumb buffer interface has no way to specify or negotiate buffer modifiers, so buffers are implicitly linear (DRM_FORMAT_MOD_LINEAR)."

2. **"They" is ambiguous.** The sentence "They are also currently meant for only linear buffers" follows a sentence about "Such drivers", so "They" could be read as referring to those specific ARM drivers rather than dumb buffers in general. Starting a new sentence more explicitly would help, e.g., "Dumb buffers are also currently meant for..."

3. **"abusing" and "hack" are strong language for kernel documentation.** While technically accurate, kernel documentation typically uses more neutral phrasing. Something like "Some KMS drivers use dumb objects for AFBC framebuffers, but this usage is non-standard and should not be replicated in new drivers" would be more appropriate for documentation that ends up in the built kernel docs.

4. **"only exists as a hack now"** — the word "now" is time-relative and will become stale. Consider "only exists for legacy reasons" or "only exists for backward compatibility" instead.

5. **Minor style nit:** "meant for only linear buffers" reads more naturally as "only meant for linear buffers."

**Suggested rewrite** of the added text:

```
 * a hardware-specific ioctl to allocate suitable buffer objects.
 *
 * Dumb buffers are only meant for linear framebuffers. The dumb buffer
 * interface has no way to specify or negotiate buffer modifiers, so
 * buffers are implicitly DRM_FORMAT_MOD_LINEAR. Some existing KMS
 * drivers use dumb objects for AFBC framebuffers, but this usage is
 * non-standard and should not be replicated in new drivers.
```

Overall, the patch addresses a real documentation gap and is welcome. With the wording cleaned up, it would be a useful addition.

---
Generated by Claude Code Patch Reviewer