Claude review: dma-buf: set SB_I_NOEXEC on the pseudo filesystem

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

Patch Review

**Correctness: Good.** The one-line change is placed at the right location — inside `dma_buf_fs_init_context()`, after `init_pseudo()` succeeds and before returning. The field `fc->s_iflags` is defined in `struct fs_context` (at `include/linux/fs_context.h:106`) and is documented as being OR'd into `sb->s_iflags` during superblock creation, so this will propagate correctly.

```c
 	ctx->dops = &dma_buf_dentry_ops;
+	fc->s_iflags |= SB_I_NOEXEC;
 	return 0;
```

This exactly mirrors the pattern in `anon_inodefs_init_fs_context()` at `fs/anon_inodes.c:89`:
```c
	fc->s_iflags |= SB_I_NOEXEC;
```

**Commit message: Good.** The commit message clearly explains:
- What the bug is (missing `SB_I_NOEXEC` on dmabuf superblock)
- What triggers it (the new `path_noexec()` assertion from 6.17)
- The exact warning stack trace
- A concrete reproducer using the dma_heap selftest
- Why the fix is correct (dma-bufs have no business being executable)

**Fixes tag: Correct.** Points to `1e7ab6f67824` which introduced the assertion that now fires. This is the right commit to cite since dma-buf was working before that assertion was added.

**Minor nit (non-blocking):** The placement of `fc->s_iflags |= SB_I_NOEXEC` after `ctx->dops = &dma_buf_dentry_ops` works but conventionally other filesystems tend to set the iflags before other context setup (as in `anon_inodes.c`). This is purely cosmetic and doesn't affect correctness at all.

**Verdict: Looks good.** Clean, minimal, correct fix. Ready to merge.

---
Generated by Claude Code Patch Reviewer