io_uring: add completion locking for iopoll

There are pieces of code that may allow iopoll to race filling cqes,
temporarily add spinlocking around posting events.

Cc: stable@vger.kernel.org
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/84d86b5c117feda075471c5c9e65208e0dccf5d0.1669203009.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/rw.c b/io_uring/rw.c
index 1ce065709724b7aa246f6094ec734aeb72f513a9..61c326831949333adc45ca78d0527b6cf43c8b83 100644 (file)
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -1049,6 +1049,7 @@ int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin)
        else if (!pos)
                return 0;
 
+       spin_lock(&ctx->completion_lock);
        prev = start;
        wq_list_for_each_resume(pos, prev) {
                struct io_kiocb *req = container_of(pos, struct io_kiocb, comp_list);
@@ -1063,11 +1064,11 @@ int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin)
                req->cqe.flags = io_put_kbuf(req, 0);
                __io_fill_cqe_req(req->ctx, req);
        }
-
+       io_commit_cqring(ctx);
+       spin_unlock(&ctx->completion_lock);
        if (unlikely(!nr_events))
                return 0;
 
-       io_commit_cqring(ctx);
        io_cqring_ev_posted_iopoll(ctx);
        pos = start ? start->next : ctx->iopoll_list.first;
        wq_list_cut(&ctx->iopoll_list, prev, start);