Claude review: drm/hyperv: validate resolution_count and fix WIN8 fallback

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

Patch Review

**Verdict: Good, minor nit.**

**Resolution count bounds check** — Correct and necessary. The existing code only checked for zero:

```c
if (msg->resolution_resp.resolution_count == 0 ||
    msg->resolution_resp.resolution_count >
    SYNTHVID_MAX_RESOLUTION_COUNT) {
```

Without this, a malicious or buggy host sending `resolution_count > 64` would cause the loop at line 405 to read past `supported_resolution[SYNTHVID_MAX_RESOLUTION_COUNT]`, which is an OOB read on `hv->init_buf`. Since `init_buf` is a flat `u8[0x4000]` inside `struct hyperv_drm_device`, this reads adjacent struct members rather than unmapped memory, but it's still a data-integrity bug that can corrupt `screen_width_max`/`screen_height_max` with garbage.

**WIN8 fallback** — Clean refactoring. The change from:

```c
} else {
    hv->screen_width_max = SYNTHVID_WIDTH_WIN8;
    hv->screen_height_max = SYNTHVID_HEIGHT_WIN8;
}
```

to:

```c
if (!hv->screen_width_max) {
    hv->screen_width_max = SYNTHVID_WIDTH_WIN8;
    hv->screen_height_max = SYNTHVID_HEIGHT_WIN8;
    hv->preferred_width = SYNTHVID_WIDTH_WIN8;
    hv->preferred_height = SYNTHVID_HEIGHT_WIN8;
}
```

This correctly converges three previously-distinct failure paths (pre-WIN10, WIN10 probe failure, WIN10 invalid count) into a single fallback. It also fixes the pre-WIN10 branch that never populated `preferred_width`/`preferred_height`, which would leave them at 0.

**Nit:** The error message uses `%d` for `resolution_count` which is `u8`:

```c
drm_err(dev, "Invalid resolution count: %d\n",
        msg->resolution_resp.resolution_count);
```

`%u` would be more semantically precise for an unsigned type, though `%d` works correctly for `u8` values and is common in the kernel. Not worth a respin.

---

---
Generated by Claude Code Patch Reviewer