[PATCH v2] io_uring/rsrc: Simplify buffer cloning by locking both rings

[PATCH v2] io_uring/rsrc: Simplify buffer cloning by locking both rings

[Jann Horn]

The locking in the buffer cloning code is somewhat complex because it goes
back and forth between locking the source ring and the destination ring.

Make it easier to reason about by locking both rings at the same time.
To avoid ABBA deadlocks, lock the rings in ascending kernel address order,
just like in lock_two_nondirectories().

Signed-off-by: Jann Horn <jannh@google.com>
---
(patch is based on top of Jens' for-next tree)
---
Changes in v2:
- fold in suggested changes from Jens (https://lore.kernel.org/r/2439336d-b6ae-4d08-a1e8-2372fc6df383@kernel.dk)
- Link to v1: https://lore.kernel.org/r/20250115-uring-clone-refactor-v1-1-b2d951577201@google.com
---
 io_uring/rsrc.c | 73 +++++++++++++++++++++++++++++++--------------------------
 1 file changed, 40 insertions(+), 33 deletions(-)

diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index e32ac58533914706c8e031754432ed634d7c0354..a1c7c8db55455e6db862a17ed611f3d6ce885b3d 100644
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -921,6 +921,16 @@ int io_import_fixed(int ddir, struct iov_iter *iter,
 	return 0;
 }
 
+/* Lock two rings at once. The rings must be different! */
+static void lock_two_rings(struct io_ring_ctx *ctx1, struct io_ring_ctx *ctx2)
+{
+	if (ctx1 > ctx2)
+		swap(ctx1, ctx2);
+	mutex_lock(&ctx1->uring_lock);
+	mutex_lock_nested(&ctx2->uring_lock, SINGLE_DEPTH_NESTING);
+}
+
+/* Both rings are locked by the caller. */
 static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx,
 			    struct io_uring_clone_buffers *arg)
 {
@@ -928,6 +938,9 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 	int i, ret, off, nr;
 	unsigned int nbufs;
 
+	lockdep_assert_held(&ctx->uring_lock);
+	lockdep_assert_held(&src_ctx->uring_lock);
+
 	/*
 	 * Accounting state is shared between the two rings; that only works if
 	 * both rings are accounted towards the same counters.
@@ -942,7 +955,7 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 	if (ctx->buf_table.nr && !(arg->flags & IORING_REGISTER_DST_REPLACE))
 		return -EBUSY;
 
-	nbufs = READ_ONCE(src_ctx->buf_table.nr);
+	nbufs = src_ctx->buf_table.nr;
 	if (!arg->nr)
 		arg->nr = nbufs;
 	else if (arg->nr > nbufs)
@@ -966,27 +979,20 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 		}
 	}
 
-	/*
-	 * Drop our own lock here. We'll setup the data we need and reference
-	 * the source buffers, then re-grab, check, and assign at the end.
-	 */
-	mutex_unlock(&ctx->uring_lock);
-
-	mutex_lock(&src_ctx->uring_lock);
 	ret = -ENXIO;
 	nbufs = src_ctx->buf_table.nr;
 	if (!nbufs)
-		goto out_unlock;
+		goto out_free;
 	ret = -EINVAL;
 	if (!arg->nr)
 		arg->nr = nbufs;
 	else if (arg->nr > nbufs)
-		goto out_unlock;
+		goto out_free;
 	ret = -EOVERFLOW;
 	if (check_add_overflow(arg->nr, arg->src_off, &off))
-		goto out_unlock;
+		goto out_free;
 	if (off > nbufs)
-		goto out_unlock;
+		goto out_free;
 
 	off = arg->dst_off;
 	i = arg->src_off;
@@ -1001,7 +1007,7 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 			dst_node = io_rsrc_node_alloc(ctx, IORING_RSRC_BUFFER);
 			if (!dst_node) {
 				ret = -ENOMEM;
-				goto out_unlock;
+				goto out_free;
 			}
 
 			refcount_inc(&src_node->buf->refs);
@@ -1011,10 +1017,6 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 		i++;
 	}
 
-	/* Have a ref on the bufs now, drop src lock and re-grab our own lock */
-	mutex_unlock(&src_ctx->uring_lock);
-	mutex_lock(&ctx->uring_lock);
-
 	/*
 	 * If asked for replace, put the old table. data->nodes[] holds both
 	 * old and new nodes at this point.
@@ -1023,24 +1025,17 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 		io_rsrc_data_free(ctx, &ctx->buf_table);
 
 	/*
-	 * ctx->buf_table should be empty now - either the contents are being
-	 * replaced and we just freed the table, or someone raced setting up
-	 * a buffer table while the clone was happening. If not empty, fall
-	 * through to failure handling.
+	 * ctx->buf_table must be empty now - either the contents are being
+	 * replaced and we just freed the table, or the contents are being
+	 * copied to a ring that does not have buffers yet (checked at function
+	 * entry).
 	 */
-	if (!ctx->buf_table.nr) {
-		ctx->buf_table = data;
-		return 0;
-	}
+	WARN_ON_ONCE(ctx->buf_table.nr);
+	ctx->buf_table = data;
+	return 0;
 
-	mutex_unlock(&ctx->uring_lock);
-	mutex_lock(&src_ctx->uring_lock);
-	/* someone raced setting up buffers, dump ours */
-	ret = -EBUSY;
-out_unlock:
+out_free:
 	io_rsrc_data_free(ctx, &data);
-	mutex_unlock(&src_ctx->uring_lock);
-	mutex_lock(&ctx->uring_lock);
 	return ret;
 }
 
@@ -1054,6 +1049,7 @@ static int io_clone_buffers(struct io_ring_ctx *ctx, struct io_ring_ctx *src_ctx
 int io_register_clone_buffers(struct io_ring_ctx *ctx, void __user *arg)
 {
 	struct io_uring_clone_buffers buf;
+	struct io_ring_ctx *src_ctx;
 	bool registered_src;
 	struct file *file;
 	int ret;
@@ -1071,7 +1067,18 @@ int io_register_clone_buffers(struct io_ring_ctx *ctx, void __user *arg)
 	file = io_uring_register_get_file(buf.src_fd, registered_src);
 	if (IS_ERR(file))
 		return PTR_ERR(file);
-	ret = io_clone_buffers(ctx, file->private_data, &buf);
+
+	src_ctx = file->private_data;
+	if (src_ctx != ctx) {
+		mutex_unlock(&ctx->uring_lock);
+		lock_two_rings(ctx, src_ctx);
+	}
+
+	ret = io_clone_buffers(ctx, src_ctx, &buf);
+
+	if (src_ctx != ctx)
+		mutex_unlock(&src_ctx->uring_lock);
+
 	if (!registered_src)
 		fput(file);
 	return ret;

---
base-commit: 1ac3ba2b3cc41645a30e49f93d3e09bd05e6d2c7
change-id: 20250115-uring-clone-refactor-06ddceddeb00

-- 
Jann Horn <jannh@google.com>

Re: [PATCH v2] io_uring/rsrc: Simplify buffer cloning by locking both rings

[Jens Axboe]

On 1/15/25 1:26 PM, Jann Horn wrote:
> The locking in the buffer cloning code is somewhat complex because it goes
> back and forth between locking the source ring and the destination ring.
> 
> Make it easier to reason about by locking both rings at the same time.
> To avoid ABBA deadlocks, lock the rings in ascending kernel address order,
> just like in lock_two_nondirectories().

I'm going to let this one sit for review for a day or two, and since it
depends changes slated for 6.13, I'll queue it up in a post-merge branch
for 6.14 if all goes well.

Thanks!

-- 
Jens Axboe

Re: [PATCH v2] io_uring/rsrc: Simplify buffer cloning by locking both rings

[Jens Axboe]

On Wed, 15 Jan 2025 21:26:03 +0100, Jann Horn wrote:
> The locking in the buffer cloning code is somewhat complex because it goes
> back and forth between locking the source ring and the destination ring.
> 
> Make it easier to reason about by locking both rings at the same time.
> To avoid ABBA deadlocks, lock the rings in ascending kernel address order,
> just like in lock_two_nondirectories().
> 
> [...]

Applied, thanks!

[1/1] io_uring/rsrc: Simplify buffer cloning by locking both rings
      commit: 8865af703c087b8bcc2fdd04b6a93d3cc0fb0e9f

Best regards,
-- 
Jens Axboe