io_uring: Use WRITE_ONCE() when writing to sq_flags

[ Upstream commit 20c0b38 ]

The compiler should be forbidden from any strange optimization for async
writes to user visible data-structures. Without proper protection, the
compiler can cause write-tearing or invent writes that would confuse the
userspace.

However, there are writes to sq_flags which are not protected by
WRITE_ONCE(). Use WRITE_ONCE() for these writes.

This is purely a theoretical issue. Presumably, any compiler is very
unlikely to do such optimizations.

Fixes: 75b28af ("io_uring: allocate the two rings together")
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Link: https://lore.kernel.org/r/20210808001342.964634-3-namit@vmware.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fs/io_uring.c

@@ -1482,7 +1482,8 @@ static bool __io_cqring_overflow_flush(struct io_ring_ctx *ctx, bool force)
 	if (all_flushed) {
 		clear_bit(0, &ctx->sq_check_overflow);
 		clear_bit(0, &ctx->cq_check_overflow);
-		ctx->rings->sq_flags &= ~IORING_SQ_CQ_OVERFLOW;
+		WRITE_ONCE(ctx->rings->sq_flags,
+			   ctx->rings->sq_flags & ~IORING_SQ_CQ_OVERFLOW);
 	}
 
 	if (posted)
@@ -1562,7 +1563,9 @@ static bool io_cqring_event_overflow(struct io_ring_ctx *ctx, u64 user_data,
 	if (list_empty(&ctx->cq_overflow_list)) {
 		set_bit(0, &ctx->sq_check_overflow);
 		set_bit(0, &ctx->cq_check_overflow);
-		ctx->rings->sq_flags |= IORING_SQ_CQ_OVERFLOW;
+		WRITE_ONCE(ctx->rings->sq_flags,
+			   ctx->rings->sq_flags | IORING_SQ_CQ_OVERFLOW);
+
 	}
 	ocqe->cqe.user_data = user_data;
 	ocqe->cqe.res = res;
@@ -6790,14 +6793,16 @@ static inline void io_ring_set_wakeup_flag(struct io_ring_ctx *ctx)
 {
 	/* Tell userspace we may need a wakeup call */
 	spin_lock_irq(&ctx->completion_lock);
-	ctx->rings->sq_flags |= IORING_SQ_NEED_WAKEUP;
+	WRITE_ONCE(ctx->rings->sq_flags,
+		   ctx->rings->sq_flags | IORING_SQ_NEED_WAKEUP);
 	spin_unlock_irq(&ctx->completion_lock);
 }
 
 static inline void io_ring_clear_wakeup_flag(struct io_ring_ctx *ctx)
 {
 	spin_lock_irq(&ctx->completion_lock);
-	ctx->rings->sq_flags &= ~IORING_SQ_NEED_WAKEUP;
+	WRITE_ONCE(ctx->rings->sq_flags,
+		   ctx->rings->sq_flags & ~IORING_SQ_NEED_WAKEUP);
 	spin_unlock_irq(&ctx->completion_lock);
 }