Claude review: drm/imagination: Restrict init_fw_trace_mask module param to read only

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

Patch Review

**Verdict: Good to merge.**

```c
-module_param_cb(init_fw_trace_mask, &pvr_fw_trace_init_mask_ops, &pvr_fw_trace_init_mask, 0600);
+module_param_cb(init_fw_trace_mask, &pvr_fw_trace_init_mask_ops, &pvr_fw_trace_init_mask, 0400);
```

This changes the sysfs permissions from `0600` (root read/write) to `0400` (root read-only). The rationale is sound: `pvr_fw_trace_init_mask` is only consumed once, during `pvr_fw_trace_init()`:

```c
fw_trace->group_mask = pvr_fw_trace_init_mask;
```

Writing to the sysfs parameter at runtime would update the variable but would **not** send a KCCB command to the firmware, so the change would be silently ignored. The debugfs `trace_mask` file (fixed in patch 1) calls `update_logtype()` which properly notifies the firmware. Making the parameter read-only prevents user confusion.

One minor note: the `pvr_fw_trace_init_mask_set` callback in `pvr_fw_trace_init_mask_ops` becomes dead code with `0400` permissions (the `.set` handler will never be called via sysfs). It's still used if someone passes the parameter on the module load command line, so it's not truly unused — but if the authors wanted to be thorough, they could set `.set = NULL` in the ops struct since `module_param_cb` with read-only permissions won't invoke it. This is a very minor observation, not a blocker.

---
Generated by Claude Code Patch Reviewer