[PATCH] Revert "i915: don't use a vma that didn't match the context VM"

[PATCH] Revert "i915: don't use a vma that didn't match the context VM"

[Joonas Lahtinen]

This reverts commit 5401b9adebc9e5f68df58226f51493ef0e6ceb4d.

Commit is no-op as per discussion in the linked mail thread.
The debated UAF scenario has not been possible in mainline kernel
since 2021.

Link: https://lore.kernel.org/intel-gfx/CAHk-=whfUM8y3PoFfT21+guKWK-mJmAE=8uLzOT+7HGv5NtqSw@mail.gmail.com/
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Tvrtko Ursulin <tursulin@ursulin.net>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 05997e8bbb29..1f303d4eaa4d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -898,8 +898,6 @@ static struct i915_vma *eb_lookup_vma(struct i915_execbuffer *eb, u32 handle)
 		vma = radix_tree_lookup(&eb->gem_context->handles_vma, handle);
 		if (likely(vma))
 			vma = i915_vma_tryget(vma);
-		else
-			vma = NULL;
 		rcu_read_unlock();
 		if (likely(vma))
 			return vma;
-- 
2.54.0

Re: [PATCH] Revert "i915: don't use a vma that didn't match the context VM"

[Linus Torvalds]

On Tue, 12 May 2026 at 01:59, Joonas Lahtinen
<joonas.lahtinen@linux.intel.com> wrote:
>
> Commit is no-op as per discussion in the linked mail thread.
> The debated UAF scenario has not been possible in mainline kernel
> since 2021.

That was never the issue.

The issue was that the code used to do this:

                if (likely(vma && vma->vm == vm))
                        vma = i915_vma_tryget(vma);

and the vma->vm == vm condition was nonsensical - and if it ever was
false, that code was all horribly buggy.

So the "since 2021" is complete nonsense.

It's only since April this year - commit a13edf9b92fc ("drm/i915/gem:
Drop check for changed VM in EXECBUF") that the "else" side is now
pointless and the revert makes sense.

Stop writing misleading commit logs and excuses for bad code. There is
no "since 2021".

                Linus

Claude review: Revert "i915: don't use a vma that didn't match the context VM"

[Claude Code Review Bot]

Patch Review

**Correctness: OK**

The patch removes:
```c
		else
			vma = NULL;
```
from the `eb_lookup_vma()` function in `i915_gem_execbuffer.c`.

This is correct because:

1. `radix_tree_lookup()` returns `void *` and yields `NULL` when the entry is not found (declared in `include/linux/radix-tree.h`). So when the `if (likely(vma))` test falls through, `vma` is already `NULL`.

2. `i915_vma_tryget()` (defined in `i915_vma.h:194`) also returns `NULL` on failure (when `kref_get_unless_zero` fails), so even after the `if (likely(vma))` → `i915_vma_tryget(vma)` path, a failure still leaves `vma` as `NULL`.

3. After the removed `else` branch, the code proceeds to `if (likely(vma)) return vma;` — which correctly handles the `NULL` case regardless.

**Commit message: Good.** References the original commit hash, links the relevant lore discussion with Linus, and explains why the original commit was a no-op. The Cc list is appropriate — includes the original participants in the discussion.

**Style: Fine.** A straightforward revert with `git revert` formatting.

**No concerns.** This is a clean, well-justified removal of dead code.

---
Generated by Claude Code Patch Reviewer

Claude review: Revert "i915: don't use a vma that didn't match the context VM"

[Claude Code Review Bot]

Overall Series Review

Subject: Revert "i915: don't use a vma that didn't match the context VM"
Author: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Patches: 2
Reviewed: 2026-05-16T13:59:28.286020

---

This is a single-patch revert of commit `5401b9adebc9`, which had added an explicit `else vma = NULL` branch to `eb_lookup_vma()`. The patch is correct: the removed code is indeed a no-op, and the revert is clean and well-justified.

**Verdict: Looks good.**

The `radix_tree_lookup()` function already returns `NULL` when the key is not found, so the `else vma = NULL` assignment is redundant — `vma` is already `NULL` in that branch. The only remaining question is whether the explicit assignment served a defensive/clarity purpose, but as the commit message explains (with a link to the Linus discussion), the original commit was motivated by a UAF concern that hasn't been possible since 2021. Removing dead defensive code that was added under a false premise is reasonable.

---
Generated by Claude Code Patch Reviewer