io_uring/net: limit inline multishot retries

If we have multiple clients and some/all are flooding the receives to
such an extent that we can retry a LOT handling multishot receives, then
we can be starving some clients and hence serving traffic in an
imbalanced fashion.

Limit multishot retry attempts to some arbitrary value, whose only
purpose serves to ensure that we don't keep serving a single connection
for way too long. We default to 32 retries, which should be more than
enough to provide fairness, yet not so small that we'll spend too much
time requeuing rather than handling traffic.

Cc: stable@vger.kernel.org
Depends-on: 704ea888d646 ("io_uring/poll: add requeue return code from poll multishot handling")
Depends-on: 1e5d765a82f ("io_uring/net: un-indent mshot retry path in io_recv_finish()")
Depends-on: e84b01a880f6 ("io_uring/poll: move poll execution helpers higher up")
Fixes: b3fdea6ecb55 ("io_uring: multishot recv")
Fixes: 9bb66906f23e ("io_uring: support multishot in recvmsg")
Link: https://github.com/axboe/liburing/issues/1043
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/net.c b/io_uring/net.c
index 740c6bfa5b5900034824ae115b1041935d762813..a12ff69e684348513335a43b105b821eb588139b 100644 (file)
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -60,6 +60,7 @@ struct io_sr_msg {
        unsigned                        len;
        unsigned                        done_io;
        unsigned                        msg_flags;
+       unsigned                        nr_multishot_loops;
        u16                             flags;
        /* initialised and used only by !msg send variants */
        u16                             addr_len;
@@ -70,6 +71,13 @@ struct io_sr_msg {
        struct io_kiocb                 *notif;
 };
 
+/*
+ * Number of times we'll try and do receives if there's more data. If we
+ * exceed this limit, then add us to the back of the queue and retry from
+ * there. This helps fairness between flooding clients.
+ */
+#define MULTISHOT_MAX_RETRY    32
+
 static inline bool io_check_multishot(struct io_kiocb *req,
                                      unsigned int issue_flags)
 {
@@ -611,6 +619,7 @@ int io_recvmsg_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
                sr->msg_flags |= MSG_CMSG_COMPAT;
 #endif
        sr->done_io = 0;
+       sr->nr_multishot_loops = 0;
        return 0;
 }
 
@@ -654,12 +663,20 @@ static inline bool io_recv_finish(struct io_kiocb *req, int *ret,
         */
        if (io_fill_cqe_req_aux(req, issue_flags & IO_URING_F_COMPLETE_DEFER,
                                *ret, cflags | IORING_CQE_F_MORE)) {
+               struct io_sr_msg *sr = io_kiocb_to_cmd(req, struct io_sr_msg);
+               int mshot_retry_ret = IOU_ISSUE_SKIP_COMPLETE;
+
                io_recv_prep_retry(req);
                /* Known not-empty or unknown state, retry */
-               if (cflags & IORING_CQE_F_SOCK_NONEMPTY || msg->msg_inq == -1)
-                       return false;
+               if (cflags & IORING_CQE_F_SOCK_NONEMPTY || msg->msg_inq == -1) {
+                       if (sr->nr_multishot_loops++ < MULTISHOT_MAX_RETRY)
+                               return false;
+                       /* mshot retries exceeded, force a requeue */
+                       sr->nr_multishot_loops = 0;
+                       mshot_retry_ret = IOU_REQUEUE;
+               }
                if (issue_flags & IO_URING_F_MULTISHOT)
-                       *ret = IOU_ISSUE_SKIP_COMPLETE;
+                       *ret = mshot_retry_ret;
                else
                        *ret = -EAGAIN;
                return true;