[PATCH 5.15 058/106] io_uring: fix races with buffer table unregister

From: Greg Kroah-Hartman
Date: Mon Jun 20 2022 - 09:21:38 EST


From: Pavel Begunkov <asml.silence@xxxxxxxxx>

[ Upstream commit d11d31fc5d8a96f707facee0babdcffaafa38de2 ]

Fixed buffer table quiesce might unlock ->uring_lock, potentially
letting new requests to be submitted, don't allow those requests to
use the table as they will race with unregistration.

Reported-and-tested-by: van fantasy <g1042620637@xxxxxxxxx>
Fixes: bd54b6fe3316ec ("io_uring: implement fixed buffers registration similar to fixed files")
Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
fs/io_uring.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 5f111a660fff..be2176575353 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -8905,12 +8905,19 @@ static void __io_sqe_buffers_unregister(struct io_ring_ctx *ctx)

static int io_sqe_buffers_unregister(struct io_ring_ctx *ctx)
{
+ unsigned nr = ctx->nr_user_bufs;
int ret;

if (!ctx->buf_data)
return -ENXIO;

+ /*
+ * Quiesce may unlock ->uring_lock, and while it's not held
+ * prevent new requests using the table.
+ */
+ ctx->nr_user_bufs = 0;
ret = io_rsrc_ref_quiesce(ctx->buf_data, ctx);
+ ctx->nr_user_bufs = nr;
if (!ret)
__io_sqe_buffers_unregister(ctx);
return ret;
--
2.35.1