]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/commitdiff
io_uring: used cached copies of sq->dropped and cq->overflow
authorJens Axboe <axboe@kernel.dk>
Fri, 25 Oct 2019 16:04:25 +0000 (10:04 -0600)
committerJens Axboe <axboe@kernel.dk>
Fri, 25 Oct 2019 16:58:45 +0000 (10:58 -0600)
We currently use the ring values directly, but that can lead to issues
if the application is malicious and changes these values on our behalf.
Created in-kernel cached versions of them, and just overwrite the user
side when we update them. This is similar to how we treat the sq/cq
ring tail/head updates.

Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
fs/io_uring.c

index 32f6598ecae9f5e5ecb2a82d88a438fe59b30586..292c4c733cbe0ab95e6896f2772ed3e738b0956a 100644 (file)
@@ -197,6 +197,7 @@ struct io_ring_ctx {
                unsigned                sq_entries;
                unsigned                sq_mask;
                unsigned                sq_thread_idle;
+               unsigned                cached_sq_dropped;
                struct io_uring_sqe     *sq_sqes;
 
                struct list_head        defer_list;
@@ -212,6 +213,7 @@ struct io_ring_ctx {
 
        struct {
                unsigned                cached_cq_tail;
+               atomic_t                cached_cq_overflow;
                unsigned                cq_entries;
                unsigned                cq_mask;
                struct wait_queue_head  cq_wait;
@@ -420,7 +422,8 @@ static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
 static inline bool __io_sequence_defer(struct io_ring_ctx *ctx,
                                       struct io_kiocb *req)
 {
-       return req->sequence != ctx->cached_cq_tail + ctx->rings->sq_dropped;
+       return req->sequence != ctx->cached_cq_tail + ctx->cached_sq_dropped
+                                       + atomic_read(&ctx->cached_cq_overflow);
 }
 
 static inline bool io_sequence_defer(struct io_ring_ctx *ctx,
@@ -567,9 +570,8 @@ static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data,
                WRITE_ONCE(cqe->res, res);
                WRITE_ONCE(cqe->flags, 0);
        } else {
-               unsigned overflow = READ_ONCE(ctx->rings->cq_overflow);
-
-               WRITE_ONCE(ctx->rings->cq_overflow, overflow + 1);
+               WRITE_ONCE(ctx->rings->cq_overflow,
+                               atomic_inc_return(&ctx->cached_cq_overflow));
        }
 }
 
@@ -2564,7 +2566,8 @@ static bool io_get_sqring(struct io_ring_ctx *ctx, struct sqe_submit *s)
 
        /* drop invalid entries */
        ctx->cached_sq_head++;
-       rings->sq_dropped++;
+       ctx->cached_sq_dropped++;
+       WRITE_ONCE(rings->sq_dropped, ctx->cached_sq_dropped);
        return false;
 }