1 files changed, 54 insertions, 67 deletions

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index cd9a137ad6ce..adf944bb5a2f 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1176,12 +1176,11 @@ static void ctx_flush_and_put(struct io_ring_ctx *ctx, struct io_tw_state *ts)
 
 static unsigned int handle_tw_list(struct llist_node *node,
 				   struct io_ring_ctx **ctx,
-				   struct io_tw_state *ts,
-				   struct llist_node *last)
+				   struct io_tw_state *ts)
 {
 	unsigned int count = 0;
 
-	while (node && node != last) {
+	do {
 		struct llist_node *next = node->next;
 		struct io_kiocb *req = container_of(node, struct io_kiocb,
 						    io_task_work.node);
@@ -1205,7 +1204,7 @@ static unsigned int handle_tw_list(struct llist_node *node,
 			*ctx = NULL;
 			cond_resched();
 		}
-	}
+	} while (node);
 
 	return count;
 }
@@ -1224,22 +1223,6 @@ static inline struct llist_node *io_llist_xchg(struct llist_head *head,
 	return xchg(&head->first, new);
 }
 
-/**
- * io_llist_cmpxchg - possibly swap all entries in a lock-less list
- * @head:	the head of lock-less list to delete all entries
- * @old:	expected old value of the first entry of the list
- * @new:	new entry as the head of the list
- *
- * perform a cmpxchg on the first entry of the list.
- */
-
-static inline struct llist_node *io_llist_cmpxchg(struct llist_head *head,
-						  struct llist_node *old,
-						  struct llist_node *new)
-{
-	return cmpxchg(&head->first, old, new);
-}
-
 static __cold void io_fallback_tw(struct io_uring_task *tctx, bool sync)
 {
 	struct llist_node *node = llist_del_all(&tctx->task_list);
@@ -1274,9 +1257,7 @@ void tctx_task_work(struct callback_head *cb)
 	struct io_ring_ctx *ctx = NULL;
 	struct io_uring_task *tctx = container_of(cb, struct io_uring_task,
 						  task_work);
-	struct llist_node fake = {};
 	struct llist_node *node;
-	unsigned int loops = 0;
 	unsigned int count = 0;
 
 	if (unlikely(current->flags & PF_EXITING)) {
@@ -1284,21 +1265,9 @@ void tctx_task_work(struct callback_head *cb)
 		return;
 	}
 
-	do {
-		loops++;
-		node = io_llist_xchg(&tctx->task_list, &fake);
-		count += handle_tw_list(node, &ctx, &ts, &fake);
-
-		/* skip expensive cmpxchg if there are items in the list */
-		if (READ_ONCE(tctx->task_list.first) != &fake)
-			continue;
-		if (ts.locked && !wq_list_empty(&ctx->submit_state.compl_reqs)) {
-			io_submit_flush_completions(ctx);
-			if (READ_ONCE(tctx->task_list.first) != &fake)
-				continue;
-		}
-		node = io_llist_cmpxchg(&tctx->task_list, &fake, NULL);
-	} while (node != &fake);
+	node = llist_del_all(&tctx->task_list);
+	if (node)
+		count = handle_tw_list(node, &ctx, &ts);
 
 	ctx_flush_and_put(ctx, &ts);
 
@@ -1306,7 +1275,7 @@ void tctx_task_work(struct callback_head *cb)
 	if (unlikely(atomic_read(&tctx->in_cancel)))
 		io_uring_drop_tctx_refs(current);
 
-	trace_io_uring_task_work_run(tctx, count, loops);
+	trace_io_uring_task_work_run(tctx, count, 1);
 }
 
 static inline void io_req_local_work_add(struct io_kiocb *req, unsigned flags)
@@ -1420,7 +1389,20 @@ static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx)
 	}
 }
 
-static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts)
+static bool io_run_local_work_continue(struct io_ring_ctx *ctx, int events,
+				       int min_events)
+{
+	if (llist_empty(&ctx->work_llist))
+		return false;
+	if (events < min_events)
+		return true;
+	if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
+		atomic_or(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
+	return false;
+}
+
+static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts,
+			       int min_events)
 {
 	struct llist_node *node;
 	unsigned int loops = 0;
@@ -1449,18 +1431,20 @@ again:
 	}
 	loops++;
 
-	if (!llist_empty(&ctx->work_llist))
+	if (io_run_local_work_continue(ctx, ret, min_events))
 		goto again;
 	if (ts->locked) {
 		io_submit_flush_completions(ctx);
-		if (!llist_empty(&ctx->work_llist))
+		if (io_run_local_work_continue(ctx, ret, min_events))
 			goto again;
 	}
+
 	trace_io_uring_local_work_run(ctx, ret, loops);
 	return ret;
 }
 
-static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
+static inline int io_run_local_work_locked(struct io_ring_ctx *ctx,
+					   int min_events)
 {
 	struct io_tw_state ts = { .locked = true, };
 	int ret;
@@ -1468,20 +1452,20 @@ static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
 	if (llist_empty(&ctx->work_llist))
 		return 0;
 
-	ret = __io_run_local_work(ctx, &ts);
+	ret = __io_run_local_work(ctx, &ts, min_events);
 	/* shouldn't happen! */
 	if (WARN_ON_ONCE(!ts.locked))
 		mutex_lock(&ctx->uring_lock);
 	return ret;
 }
 
-static int io_run_local_work(struct io_ring_ctx *ctx)
+static int io_run_local_work(struct io_ring_ctx *ctx, int min_events)
 {
 	struct io_tw_state ts = {};
 	int ret;
 
 	ts.locked = mutex_trylock(&ctx->uring_lock);
-	ret = __io_run_local_work(ctx, &ts);
+	ret = __io_run_local_work(ctx, &ts, min_events);
 	if (ts.locked)
 		mutex_unlock(&ctx->uring_lock);
 
@@ -1677,7 +1661,7 @@ static int io_iopoll_check(struct io_ring_ctx *ctx, long min)
 		    io_task_work_pending(ctx)) {
 			u32 tail = ctx->cached_cq_tail;
 
-			(void) io_run_local_work_locked(ctx);
+			(void) io_run_local_work_locked(ctx, min);
 
 			if (task_work_pending(current) ||
 			    wq_list_empty(&ctx->iopoll_list)) {
@@ -2520,7 +2504,7 @@ int io_run_task_work_sig(struct io_ring_ctx *ctx)
 {
 	if (!llist_empty(&ctx->work_llist)) {
 		__set_current_state(TASK_RUNNING);
-		if (io_run_local_work(ctx) > 0)
+		if (io_run_local_work(ctx, INT_MAX) > 0)
 			return 0;
 	}
 	if (io_run_task_work() > 0)
@@ -2543,7 +2527,7 @@ static bool current_pending_io(void)
 static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
 					  struct io_wait_queue *iowq)
 {
-	int io_wait, ret;
+	int ret;
 
 	if (unlikely(READ_ONCE(ctx->check_cq)))
 		return 1;
@@ -2561,7 +2545,6 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
 	 * can take into account that the task is waiting for IO - turns out
 	 * to be important for low QD IO.
 	 */
-	io_wait = current->in_iowait;
 	if (current_pending_io())
 		current->in_iowait = 1;
 	ret = 0;
@@ -2569,7 +2552,7 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
 		schedule();
 	else if (!schedule_hrtimeout(&iowq->timeout, HRTIMER_MODE_ABS))
 		ret = -ETIME;
-	current->in_iowait = io_wait;
+	current->in_iowait = 0;
 	return ret;
 }
 
@@ -2588,7 +2571,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 	if (!io_allowed_run_tw(ctx))
 		return -EEXIST;
 	if (!llist_empty(&ctx->work_llist))
-		io_run_local_work(ctx);
+		io_run_local_work(ctx, min_events);
 	io_run_task_work();
 	io_cqring_overflow_flush(ctx);
 	/* if user messes with these they will just get an early return */
@@ -2626,11 +2609,10 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 
 	trace_io_uring_cqring_wait(ctx, min_events);
 	do {
+		int nr_wait = (int) iowq.cq_tail - READ_ONCE(ctx->rings->cq.tail);
 		unsigned long check_cq;
 
 		if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
-			int nr_wait = (int) iowq.cq_tail - READ_ONCE(ctx->rings->cq.tail);
-
 			atomic_set(&ctx->cq_wait_nr, nr_wait);
 			set_current_state(TASK_INTERRUPTIBLE);
 		} else {
@@ -2649,7 +2631,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 		 */
 		io_run_task_work();
 		if (!llist_empty(&ctx->work_llist))
-			io_run_local_work(ctx);
+			io_run_local_work(ctx, nr_wait);
 
 		/*
 		 * Non-local task_work will be run on exit to userspace, but
@@ -2720,7 +2702,7 @@ static void *__io_uaddr_map(struct page ***pages, unsigned short *npages,
 	struct page **page_array;
 	unsigned int nr_pages;
 	void *page_addr;
-	int ret, i;
+	int ret, i, pinned;
 
 	*npages = 0;
 
@@ -2734,12 +2716,12 @@ static void *__io_uaddr_map(struct page ***pages, unsigned short *npages,
 	if (!page_array)
 		return ERR_PTR(-ENOMEM);
 
-	ret = pin_user_pages_fast(uaddr, nr_pages, FOLL_WRITE | FOLL_LONGTERM,
-					page_array);
-	if (ret != nr_pages) {
-err:
-		io_pages_free(&page_array, ret > 0 ? ret : 0);
-		return ret < 0 ? ERR_PTR(ret) : ERR_PTR(-EFAULT);
+
+	pinned = pin_user_pages_fast(uaddr, nr_pages, FOLL_WRITE | FOLL_LONGTERM,
+				     page_array);
+	if (pinned != nr_pages) {
+		ret = (pinned < 0) ? pinned : -EFAULT;
+		goto free_pages;
 	}
 
 	page_addr = page_address(page_array[0]);
@@ -2753,7 +2735,7 @@ err:
 		 * didn't support this feature.
 		 */
 		if (PageHighMem(page_array[i]))
-			goto err;
+			goto free_pages;
 
 		/*
 		 * No support for discontig pages for now, should either be a
@@ -2762,13 +2744,17 @@ err:
 		 * just fail them with EINVAL.
 		 */
 		if (page_address(page_array[i]) != page_addr)
-			goto err;
+			goto free_pages;
 		page_addr += PAGE_SIZE;
 	}
 
 	*pages = page_array;
 	*npages = nr_pages;
 	return page_to_virt(page_array[0]);
+
+free_pages:
+	io_pages_free(&page_array, pinned > 0 ? pinned : 0);
+	return ERR_PTR(ret);
 }
 
 static void *io_rings_map(struct io_ring_ctx *ctx, unsigned long uaddr,
@@ -2790,14 +2776,15 @@ static void io_rings_free(struct io_ring_ctx *ctx)
 	if (!(ctx->flags & IORING_SETUP_NO_MMAP)) {
 		io_mem_free(ctx->rings);
 		io_mem_free(ctx->sq_sqes);
-		ctx->rings = NULL;
-		ctx->sq_sqes = NULL;
 	} else {
 		io_pages_free(&ctx->ring_pages, ctx->n_ring_pages);
 		ctx->n_ring_pages = 0;
 		io_pages_free(&ctx->sqe_pages, ctx->n_sqe_pages);
 		ctx->n_sqe_pages = 0;
 	}
+
+	ctx->rings = NULL;
+	ctx->sq_sqes = NULL;
 }
 
 void *io_mem_alloc(size_t size)
@@ -3304,7 +3291,7 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 
 	if ((ctx->flags & IORING_SETUP_DEFER_TASKRUN) &&
 	    io_allowed_defer_tw_run(ctx))
-		ret |= io_run_local_work(ctx) > 0;
+		ret |= io_run_local_work(ctx, INT_MAX) > 0;
 	ret |= io_cancel_defer_files(ctx, task, cancel_all);
 	mutex_lock(&ctx->uring_lock);
 	ret |= io_poll_remove_all(ctx, task, cancel_all);
@@ -3666,7 +3653,7 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 			 * it should handle ownership problems if any.
 			 */
 			if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
-				(void)io_run_local_work_locked(ctx);
+				(void)io_run_local_work_locked(ctx, min_complete);
 		}
 		mutex_unlock(&ctx->uring_lock);
 	}