1 files changed, 248 insertions, 160 deletions

diff --git a/fs/io_uring.c b/fs/io_uring.c
index b42dfa0243bf..86dac2b2e276 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -205,6 +205,7 @@ struct fixed_file_ref_node {
 	struct list_head		file_list;
 	struct fixed_file_data		*file_data;
 	struct llist_node		llist;
+	bool				done;
 };
 
 struct fixed_file_data {
@@ -478,6 +479,7 @@ struct io_sr_msg {
 struct io_open {
 	struct file			*file;
 	int				dfd;
+	bool				ignore_nonblock;
 	struct filename			*filename;
 	struct open_how			how;
 	unsigned long			nofile;
@@ -995,20 +997,33 @@ static void io_sq_thread_drop_mm(void)
 	if (mm) {
 		kthread_unuse_mm(mm);
 		mmput(mm);
+		current->mm = NULL;
 	}
 }
 
 static int __io_sq_thread_acquire_mm(struct io_ring_ctx *ctx)
 {
-	if (!current->mm) {
-		if (unlikely(!(ctx->flags & IORING_SETUP_SQPOLL) ||
-			     !ctx->sqo_task->mm ||
-			     !mmget_not_zero(ctx->sqo_task->mm)))
-			return -EFAULT;
-		kthread_use_mm(ctx->sqo_task->mm);
+	struct mm_struct *mm;
+
+	if (current->mm)
+		return 0;
+
+	/* Should never happen */
+	if (unlikely(!(ctx->flags & IORING_SETUP_SQPOLL)))
+		return -EFAULT;
+
+	task_lock(ctx->sqo_task);
+	mm = ctx->sqo_task->mm;
+	if (unlikely(!mm || !mmget_not_zero(mm)))
+		mm = NULL;
+	task_unlock(ctx->sqo_task);
+
+	if (mm) {
+		kthread_use_mm(mm);
+		return 0;
 	}
 
-	return 0;
+	return -EFAULT;
 }
 
 static int io_sq_thread_acquire_mm(struct io_ring_ctx *ctx,
@@ -1269,14 +1284,17 @@ static bool io_identity_cow(struct io_kiocb *req)
 	 */
 	io_init_identity(id);
 	if (creds)
-		req->work.identity->creds = creds;
+		id->creds = creds;
 
 	/* add one for this request */
 	refcount_inc(&id->count);
 
-	/* drop old identity, assign new one. one ref for req, one for tctx */
-	if (req->work.identity != tctx->identity &&
-	    refcount_sub_and_test(2, &req->work.identity->count))
+	/* drop tctx and req identity references, if needed */
+	if (tctx->identity != &tctx->__identity &&
+	    refcount_dec_and_test(&tctx->identity->count))
+		kfree(tctx->identity);
+	if (req->work.identity != &tctx->__identity &&
+	    refcount_dec_and_test(&req->work.identity->count))
 		kfree(req->work.identity);
 
 	req->work.identity = id;
@@ -1295,22 +1313,6 @@ static bool io_grab_identity(struct io_kiocb *req)
 			return false;
 		req->work.flags |= IO_WQ_WORK_FSIZE;
 	}
-
-	if (!(req->work.flags & IO_WQ_WORK_FILES) &&
-	    (def->work_flags & IO_WQ_WORK_FILES) &&
-	    !(req->flags & REQ_F_NO_FILE_TABLE)) {
-		if (id->files != current->files ||
-		    id->nsproxy != current->nsproxy)
-			return false;
-		atomic_inc(&id->files->count);
-		get_nsproxy(id->nsproxy);
-		req->flags |= REQ_F_INFLIGHT;
-
-		spin_lock_irq(&ctx->inflight_lock);
-		list_add(&req->inflight_entry, &ctx->inflight_list);
-		spin_unlock_irq(&ctx->inflight_lock);
-		req->work.flags |= IO_WQ_WORK_FILES;
-	}
 #ifdef CONFIG_BLK_CGROUP
 	if (!(req->work.flags & IO_WQ_WORK_BLKCG) &&
 	    (def->work_flags & IO_WQ_WORK_BLKCG)) {
@@ -1352,6 +1354,21 @@ static bool io_grab_identity(struct io_kiocb *req)
 		}
 		spin_unlock(&current->fs->lock);
 	}
+	if (!(req->work.flags & IO_WQ_WORK_FILES) &&
+	    (def->work_flags & IO_WQ_WORK_FILES) &&
+	    !(req->flags & REQ_F_NO_FILE_TABLE)) {
+		if (id->files != current->files ||
+		    id->nsproxy != current->nsproxy)
+			return false;
+		atomic_inc(&id->files->count);
+		get_nsproxy(id->nsproxy);
+		req->flags |= REQ_F_INFLIGHT;
+
+		spin_lock_irq(&ctx->inflight_lock);
+		list_add(&req->inflight_entry, &ctx->inflight_list);
+		spin_unlock_irq(&ctx->inflight_lock);
+		req->work.flags |= IO_WQ_WORK_FILES;
+	}
 
 	return true;
 }
@@ -1365,6 +1382,9 @@ static void io_prep_async_work(struct io_kiocb *req)
 	io_req_init_async(req);
 	id = req->work.identity;
 
+	if (req->flags & REQ_F_FORCE_ASYNC)
+		req->work.flags |= IO_WQ_WORK_CONCURRENT;
+
 	if (req->flags & REQ_F_ISREG) {
 		if (def->hash_reg_file || (ctx->flags & IORING_SETUP_IOPOLL))
 			io_wq_hash_work(&req->work, file_inode(req->file));
@@ -1574,14 +1594,29 @@ static void io_cqring_mark_overflow(struct io_ring_ctx *ctx)
 	}
 }
 
-static inline bool io_match_files(struct io_kiocb *req,
-				       struct files_struct *files)
+static inline bool __io_match_files(struct io_kiocb *req,
+				    struct files_struct *files)
+{
+	return ((req->flags & REQ_F_WORK_INITIALIZED) &&
+	        (req->work.flags & IO_WQ_WORK_FILES)) &&
+		req->work.identity->files == files;
+}
+
+static bool io_match_files(struct io_kiocb *req,
+			   struct files_struct *files)
 {
+	struct io_kiocb *link;
+
 	if (!files)
 		return true;
-	if ((req->flags & REQ_F_WORK_INITIALIZED) &&
-	    (req->work.flags & IO_WQ_WORK_FILES))
-		return req->work.identity->files == files;
+	if (__io_match_files(req, files))
+		return true;
+	if (req->flags & REQ_F_LINK_HEAD) {
+		list_for_each_entry(link, &req->link_list, link_list) {
+			if (__io_match_files(link, files))
+				return true;
+		}
+	}
 	return false;
 }
 
@@ -1665,7 +1700,8 @@ static void __io_cqring_fill_event(struct io_kiocb *req, long res, long cflags)
 		WRITE_ONCE(cqe->user_data, req->user_data);
 		WRITE_ONCE(cqe->res, res);
 		WRITE_ONCE(cqe->flags, cflags);
-	} else if (ctx->cq_overflow_flushed || req->task->io_uring->in_idle) {
+	} else if (ctx->cq_overflow_flushed ||
+		   atomic_read(&req->task->io_uring->in_idle)) {
 		/*
 		 * If we're in ring overflow flush mode, or in task cancel mode,
 		 * then we cannot store the request for later flushing, we need
@@ -1835,7 +1871,7 @@ static void __io_free_req(struct io_kiocb *req)
 	io_dismantle_req(req);
 
 	percpu_counter_dec(&tctx->inflight);
-	if (tctx->in_idle)
+	if (atomic_read(&tctx->in_idle))
 		wake_up(&tctx->wait);
 	put_task_struct(req->task);
 
@@ -1846,59 +1882,39 @@ static void __io_free_req(struct io_kiocb *req)
 	percpu_ref_put(&ctx->refs);
 }
 
-static bool io_link_cancel_timeout(struct io_kiocb *req)
+static void io_kill_linked_timeout(struct io_kiocb *req)
 {
-	struct io_timeout_data *io = req->async_data;
 	struct io_ring_ctx *ctx = req->ctx;
-	int ret;
-
-	ret = hrtimer_try_to_cancel(&io->timer);
-	if (ret != -1) {
-		io_cqring_fill_event(req, -ECANCELED);
-		io_commit_cqring(ctx);
-		req->flags &= ~REQ_F_LINK_HEAD;
-		io_put_req_deferred(req, 1);
-		return true;
-	}
-
-	return false;
-}
-
-static bool __io_kill_linked_timeout(struct io_kiocb *req)
-{
 	struct io_kiocb *link;
-	bool wake_ev;
+	bool cancelled = false;
+	unsigned long flags;
 
-	if (list_empty(&req->link_list))
-		return false;
-	link = list_first_entry(&req->link_list, struct io_kiocb, link_list);
-	if (link->opcode != IORING_OP_LINK_TIMEOUT)
-		return false;
+	spin_lock_irqsave(&ctx->completion_lock, flags);
+	link = list_first_entry_or_null(&req->link_list, struct io_kiocb,
+					link_list);
 	/*
 	 * Can happen if a linked timeout fired and link had been like
 	 * req -> link t-out -> link t-out [-> ...]
 	 */
-	if (!(link->flags & REQ_F_LTIMEOUT_ACTIVE))
-		return false;
+	if (link && (link->flags & REQ_F_LTIMEOUT_ACTIVE)) {
+		struct io_timeout_data *io = link->async_data;
+		int ret;
 
-	list_del_init(&link->link_list);
-	wake_ev = io_link_cancel_timeout(link);
+		list_del_init(&link->link_list);
+		ret = hrtimer_try_to_cancel(&io->timer);
+		if (ret != -1) {
+			io_cqring_fill_event(link, -ECANCELED);
+			io_commit_cqring(ctx);
+			cancelled = true;
+		}
+	}
 	req->flags &= ~REQ_F_LINK_TIMEOUT;
-	return wake_ev;
-}
-
-static void io_kill_linked_timeout(struct io_kiocb *req)
-{
-	struct io_ring_ctx *ctx = req->ctx;
-	unsigned long flags;
-	bool wake_ev;
-
-	spin_lock_irqsave(&ctx->completion_lock, flags);
-	wake_ev = __io_kill_linked_timeout(req);
 	spin_unlock_irqrestore(&ctx->completion_lock, flags);
 
-	if (wake_ev)
+	if (cancelled) {
 		io_cqring_ev_posted(ctx);
+		io_put_req(link);
+	}
 }
 
 static struct io_kiocb *io_req_link_next(struct io_kiocb *req)
@@ -2562,7 +2578,6 @@ static bool io_resubmit_prep(struct io_kiocb *req, int error)
 	}
 end_req:
 	req_set_fail_links(req);
-	io_req_complete(req, ret);
 	return false;
 }
 #endif
@@ -3177,7 +3192,7 @@ static void io_req_map_rw(struct io_kiocb *req, const struct iovec *iovec,
 	rw->free_iovec = iovec;
 	rw->bytes_done = 0;
 	/* can only be fixed buffers, no need to do anything */
-	if (iter->type == ITER_BVEC)
+	if (iov_iter_is_bvec(iter))
 		return;
 	if (!iovec) {
 		unsigned iov_off = 0;
@@ -3532,8 +3547,7 @@ static int io_write(struct io_kiocb *req, bool force_nonblock,
 	 * we return to userspace.
 	 */
 	if (req->flags & REQ_F_ISREG) {
-		__sb_start_write(file_inode(req->file)->i_sb,
-					SB_FREEZE_WRITE, true);
+		sb_start_write(file_inode(req->file)->i_sb);
 		__sb_writers_release(file_inode(req->file)->i_sb,
 					SB_FREEZE_WRITE);
 	}
@@ -3781,6 +3795,7 @@ static int __io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe
 		return ret;
 	}
 	req->open.nofile = rlimit(RLIMIT_NOFILE);
+	req->open.ignore_nonblock = false;
 	req->flags |= REQ_F_NEED_CLEANUP;
 	return 0;
 }
@@ -3824,7 +3839,7 @@ static int io_openat2(struct io_kiocb *req, bool force_nonblock)
 	struct file *file;
 	int ret;
 
-	if (force_nonblock)
+	if (force_nonblock && !req->open.ignore_nonblock)
 		return -EAGAIN;
 
 	ret = build_open_flags(&req->open.how, &op);
@@ -3839,6 +3854,21 @@ static int io_openat2(struct io_kiocb *req, bool force_nonblock)
 	if (IS_ERR(file)) {
 		put_unused_fd(ret);
 		ret = PTR_ERR(file);
+		/*
+		 * A work-around to ensure that /proc/self works that way
+		 * that it should - if we get -EOPNOTSUPP back, then assume
+		 * that proc_self_get_link() failed us because we're in async
+		 * context. We should be safe to retry this from the task
+		 * itself with force_nonblock == false set, as it should not
+		 * block on lookup. Would be nice to know this upfront and
+		 * avoid the async dance, but doesn't seem feasible.
+		 */
+		if (ret == -EOPNOTSUPP && io_wq_current_is_worker()) {
+			req->open.ignore_nonblock = true;
+			refcount_inc(&req->refs);
+			io_req_task_queue(req);
+			return 0;
+		}
 	} else {
 		fsnotify_open(file);
 		fd_install(ret, file);
@@ -4469,7 +4499,8 @@ static int __io_compat_recvmsg_copy_hdr(struct io_kiocb *req,
 			return -EFAULT;
 		if (clen < 0)
 			return -EINVAL;
-		sr->len = iomsg->iov[0].iov_len;
+		sr->len = clen;
+		iomsg->iov[0].iov_len = clen;
 		iomsg->iov = NULL;
 	} else {
 		ret = __import_iovec(READ, (struct iovec __user *)uiov, len,
@@ -4977,8 +5008,10 @@ static int io_poll_double_wake(struct wait_queue_entry *wait, unsigned mode,
 		/* make sure double remove sees this as being gone */
 		wait->private = NULL;
 		spin_unlock(&poll->head->lock);
-		if (!done)
-			__io_async_wake(req, poll, mask, io_poll_task_func);
+		if (!done) {
+			/* use wait func handler, so it matches the rq type */
+			poll->wait.func(&poll->wait, mode, sync, key);
+		}
 	}
 	refcount_dec(&req->refs);
 	return 1;
@@ -6180,7 +6213,6 @@ static struct io_kiocb *io_prep_linked_timeout(struct io_kiocb *req)
 static void __io_queue_sqe(struct io_kiocb *req, struct io_comp_state *cs)
 {
 	struct io_kiocb *linked_timeout;
-	struct io_kiocb *nxt;
 	const struct cred *old_creds = NULL;
 	int ret;
 
@@ -6206,7 +6238,6 @@ again:
 	 */
 	if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
 		if (!io_arm_poll_handler(req)) {
-punt:
 			/*
 			 * Queued up for async execution, worker will release
 			 * submit reference when the iocb is actually submitted.
@@ -6216,33 +6247,25 @@ punt:
 
 		if (linked_timeout)
 			io_queue_linked_timeout(linked_timeout);
-		goto exit;
-	}
+	} else if (likely(!ret)) {
+		/* drop submission reference */
+		req = io_put_req_find_next(req);
+		if (linked_timeout)
+			io_queue_linked_timeout(linked_timeout);
 
-	if (unlikely(ret)) {
+		if (req) {
+			if (!(req->flags & REQ_F_FORCE_ASYNC))
+				goto again;
+			io_queue_async_work(req);
+		}
+	} else {
 		/* un-prep timeout, so it'll be killed as any other linked */
 		req->flags &= ~REQ_F_LINK_TIMEOUT;
 		req_set_fail_links(req);
 		io_put_req(req);
 		io_req_complete(req, ret);
-		goto exit;
 	}
 
-	/* drop submission reference */
-	nxt = io_put_req_find_next(req);
-	if (linked_timeout)
-		io_queue_linked_timeout(linked_timeout);
-
-	if (nxt) {
-		req = nxt;
-
-		if (req->flags & REQ_F_FORCE_ASYNC) {
-			linked_timeout = NULL;
-			goto punt;
-		}
-		goto again;
-	}
-exit:
 	if (old_creds)
 		revert_creds(old_creds);
 }
@@ -6266,13 +6289,6 @@ fail_req:
 			if (unlikely(ret))
 				goto fail_req;
 		}
-
-		/*
-		 * Never try inline submit of IOSQE_ASYNC is set, go straight
-		 * to async execution.
-		 */
-		io_req_init_async(req);
-		req->work.flags |= IO_WQ_WORK_CONCURRENT;
 		io_queue_async_work(req);
 	} else {
 		if (sqe) {
@@ -6958,9 +6974,7 @@ static int io_sqe_files_unregister(struct io_ring_ctx *ctx)
 		return -ENXIO;
 
 	spin_lock(&data->lock);
-	if (!list_empty(&data->ref_list))
-		ref_node = list_first_entry(&data->ref_list,
-				struct fixed_file_ref_node, node);
+	ref_node = data->node;
 	spin_unlock(&data->lock);
 	if (ref_node)
 		percpu_ref_kill(&ref_node->refs);
@@ -7309,10 +7323,6 @@ static void __io_file_put_work(struct fixed_file_ref_node *ref_node)
 		kfree(pfile);
 	}
 
-	spin_lock(&file_data->lock);
-	list_del(&ref_node->node);
-	spin_unlock(&file_data->lock);
-
 	percpu_ref_exit(&ref_node->refs);
 	kfree(ref_node);
 	percpu_ref_put(&file_data->refs);
@@ -7339,17 +7349,32 @@ static void io_file_put_work(struct work_struct *work)
 static void io_file_data_ref_zero(struct percpu_ref *ref)
 {
 	struct fixed_file_ref_node *ref_node;
+	struct fixed_file_data *data;
 	struct io_ring_ctx *ctx;
-	bool first_add;
+	bool first_add = false;
 	int delay = HZ;
 
 	ref_node = container_of(ref, struct fixed_file_ref_node, refs);
-	ctx = ref_node->file_data->ctx;
+	data = ref_node->file_data;
+	ctx = data->ctx;
+
+	spin_lock(&data->lock);
+	ref_node->done = true;
 
-	if (percpu_ref_is_dying(&ctx->file_data->refs))
+	while (!list_empty(&data->ref_list)) {
+		ref_node = list_first_entry(&data->ref_list,
+					struct fixed_file_ref_node, node);
+		/* recycle ref nodes in order */
+		if (!ref_node->done)
+			break;
+		list_del(&ref_node->node);
+		first_add |= llist_add(&ref_node->llist, &ctx->file_put_llist);
+	}
+	spin_unlock(&data->lock);
+
+	if (percpu_ref_is_dying(&data->refs))
 		delay = 0;
 
-	first_add = llist_add(&ref_node->llist, &ctx->file_put_llist);
 	if (!delay)
 		mod_delayed_work(system_wq, &ctx->file_put_work, 0);
 	else if (first_add)
@@ -7373,6 +7398,7 @@ static struct fixed_file_ref_node *alloc_fixed_file_ref_node(
 	INIT_LIST_HEAD(&ref_node->node);
 	INIT_LIST_HEAD(&ref_node->file_list);
 	ref_node->file_data = ctx->file_data;
+	ref_node->done = false;
 	return ref_node;
 }
 
@@ -7468,7 +7494,7 @@ static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
 
 	file_data->node = ref_node;
 	spin_lock(&file_data->lock);
-	list_add(&ref_node->node, &file_data->ref_list);
+	list_add_tail(&ref_node->node, &file_data->ref_list);
 	spin_unlock(&file_data->lock);
 	percpu_ref_get(&file_data->refs);
 	return ret;
@@ -7627,7 +7653,7 @@ static int __io_sqe_files_update(struct io_ring_ctx *ctx,
 	if (needs_switch) {
 		percpu_ref_kill(&data->node->refs);
 		spin_lock(&data->lock);
-		list_add(&ref_node->node, &data->ref_list);
+		list_add_tail(&ref_node->node, &data->ref_list);
 		data->node = ref_node;
 		spin_unlock(&data->lock);
 		percpu_ref_get(&ctx->file_data->refs);
@@ -7727,7 +7753,8 @@ static int io_uring_alloc_task_context(struct task_struct *task)
 	xa_init(&tctx->xa);
 	init_waitqueue_head(&tctx->wait);
 	tctx->last = NULL;
-	tctx->in_idle = 0;
+	atomic_set(&tctx->in_idle, 0);
+	tctx->sqpoll = false;
 	io_init_identity(&tctx->__identity);
 	tctx->identity = &tctx->__identity;
 	task->io_uring = tctx;
@@ -8420,22 +8447,6 @@ static bool io_match_link(struct io_kiocb *preq, struct io_kiocb *req)
 	return false;
 }
 
-static bool io_match_link_files(struct io_kiocb *req,
-				struct files_struct *files)
-{
-	struct io_kiocb *link;
-
-	if (io_match_files(req, files))
-		return true;
-	if (req->flags & REQ_F_LINK_HEAD) {
-		list_for_each_entry(link, &req->link_list, link_list) {
-			if (io_match_files(link, files))
-				return true;
-		}
-	}
-	return false;
-}
-
 /*
  * We're looking to cancel 'req' because it's holding on to our files, but
  * 'req' could be a link to another request. See if it is, and cancel that
@@ -8485,7 +8496,21 @@ static bool io_timeout_remove_link(struct io_ring_ctx *ctx,
 
 static bool io_cancel_link_cb(struct io_wq_work *work, void *data)
 {
-	return io_match_link(container_of(work, struct io_kiocb, work), data);
+	struct io_kiocb *req = container_of(work, struct io_kiocb, work);
+	bool ret;
+
+	if (req->flags & REQ_F_LINK_TIMEOUT) {
+		unsigned long flags;
+		struct io_ring_ctx *ctx = req->ctx;
+
+		/* protect against races with linked timeouts */
+		spin_lock_irqsave(&ctx->completion_lock, flags);
+		ret = io_match_link(req, data);
+		spin_unlock_irqrestore(&ctx->completion_lock, flags);
+	} else {
+		ret = io_match_link(req, data);
+	}
+	return ret;
 }
 
 static void io_attempt_cancel(struct io_ring_ctx *ctx, struct io_kiocb *req)
@@ -8511,6 +8536,7 @@ static void io_attempt_cancel(struct io_ring_ctx *ctx, struct io_kiocb *req)
 }
 
 static void io_cancel_defer_files(struct io_ring_ctx *ctx,
+				  struct task_struct *task,
 				  struct files_struct *files)
 {
 	struct io_defer_entry *de = NULL;
@@ -8518,7 +8544,8 @@ static void io_cancel_defer_files(struct io_ring_ctx *ctx,
 
 	spin_lock_irq(&ctx->completion_lock);
 	list_for_each_entry_reverse(de, &ctx->defer_list, list) {
-		if (io_match_link_files(de->req, files)) {
+		if (io_task_match(de->req, task) &&
+		    io_match_files(de->req, files)) {
 			list_cut_position(&list, &ctx->defer_list, &de->list);
 			break;
 		}
@@ -8544,7 +8571,6 @@ static bool io_uring_cancel_files(struct io_ring_ctx *ctx,
 	if (list_empty_careful(&ctx->inflight_list))
 		return false;
 
-	io_cancel_defer_files(ctx, files);
 	/* cancel all at once, should be faster than doing it one by one*/
 	io_wq_cancel_cb(ctx->io_wq, io_wq_files_match, files, true);
 
@@ -8630,8 +8656,16 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
 {
 	struct task_struct *task = current;
 
-	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data)
+	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
 		task = ctx->sq_data->thread;
+		atomic_inc(&task->io_uring->in_idle);
+		io_sq_thread_park(ctx->sq_data);
+	}
+
+	if (files)
+		io_cancel_defer_files(ctx, NULL, files);
+	else
+		io_cancel_defer_files(ctx, task, NULL);
 
 	io_cqring_overflow_flush(ctx, true, task, files);
 
@@ -8639,12 +8673,23 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
 		io_run_task_work();
 		cond_resched();
 	}
+
+	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
+		atomic_dec(&task->io_uring->in_idle);
+		/*
+		 * If the files that are going away are the ones in the thread
+		 * identity, clear them out.
+		 */
+		if (task->io_uring->identity->files == files)
+			task->io_uring->identity->files = NULL;
+		io_sq_thread_unpark(ctx->sq_data);
+	}
 }
 
 /*
  * Note that this task has used io_uring. We use it for cancelation purposes.
  */
-static int io_uring_add_task_file(struct file *file)
+static int io_uring_add_task_file(struct io_ring_ctx *ctx, struct file *file)
 {
 	struct io_uring_task *tctx = current->io_uring;
 
@@ -8666,6 +8711,14 @@ static int io_uring_add_task_file(struct file *file)
 		tctx->last = file;
 	}
 
+	/*
+	 * This is race safe in that the task itself is doing this, hence it
+	 * cannot be going through the exit/cancel paths at the same time.
+	 * This cannot be modified while exit/cancel is running.
+	 */
+	if (!tctx->sqpoll && (ctx->flags & IORING_SETUP_SQPOLL))
+		tctx->sqpoll = true;
+
 	return 0;
 }
 
@@ -8707,7 +8760,7 @@ void __io_uring_files_cancel(struct files_struct *files)
 	unsigned long index;
 
 	/* make sure overflow events are dropped */
-	tctx->in_idle = true;
+	atomic_inc(&tctx->in_idle);
 
 	xa_for_each(&tctx->xa, index, file) {
 		struct io_ring_ctx *ctx = file->private_data;
@@ -8716,6 +8769,35 @@ void __io_uring_files_cancel(struct files_struct *files)
 		if (files)
 			io_uring_del_task_file(file);
 	}
+
+	atomic_dec(&tctx->in_idle);
+}
+
+static s64 tctx_inflight(struct io_uring_task *tctx)
+{
+	unsigned long index;
+	struct file *file;
+	s64 inflight;
+
+	inflight = percpu_counter_sum(&tctx->inflight);
+	if (!tctx->sqpoll)
+		return inflight;
+
+	/*
+	 * If we have SQPOLL rings, then we need to iterate and find them, and
+	 * add the pending count for those.
+	 */
+	xa_for_each(&tctx->xa, index, file) {
+		struct io_ring_ctx *ctx = file->private_data;
+
+		if (ctx->flags & IORING_SETUP_SQPOLL) {
+			struct io_uring_task *__tctx = ctx->sqo_task->io_uring;
+
+			inflight += percpu_counter_sum(&__tctx->inflight);
+		}
+	}
+
+	return inflight;
 }
 
 /*
@@ -8729,11 +8811,11 @@ void __io_uring_task_cancel(void)
 	s64 inflight;
 
 	/* make sure overflow events are dropped */
-	tctx->in_idle = true;
+	atomic_inc(&tctx->in_idle);
 
 	do {
 		/* read completions before cancelations */
-		inflight = percpu_counter_sum(&tctx->inflight);
+		inflight = tctx_inflight(tctx);
 		if (!inflight)
 			break;
 		__io_uring_files_cancel(NULL);
@@ -8744,13 +8826,13 @@ void __io_uring_task_cancel(void)
 		 * If we've seen completions, retry. This avoids a race where
 		 * a completion comes in before we did prepare_to_wait().
 		 */
-		if (inflight != percpu_counter_sum(&tctx->inflight))
+		if (inflight != tctx_inflight(tctx))
 			continue;
 		schedule();
 	} while (1);
 
 	finish_wait(&tctx->wait, &wait);
-	tctx->in_idle = false;
+	atomic_dec(&tctx->in_idle);
 }
 
 static int io_uring_flush(struct file *file, void *data)
@@ -8895,7 +8977,7 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 			io_sqpoll_wait_sq(ctx);
 		submitted = to_submit;
 	} else if (to_submit) {
-		ret = io_uring_add_task_file(f.file);
+		ret = io_uring_add_task_file(ctx, f.file);
 		if (unlikely(ret))
 			goto out;
 		mutex_lock(&ctx->uring_lock);
@@ -8932,7 +9014,8 @@ out_fput:
 #ifdef CONFIG_PROC_FS
 static int io_uring_show_cred(int id, void *p, void *data)
 {
-	const struct cred *cred = p;
+	struct io_identity *iod = p;
+	const struct cred *cred = iod->creds;
 	struct seq_file *m = data;
 	struct user_namespace *uns = seq_user_ns(m);
 	struct group_info *gi;
@@ -9100,6 +9183,7 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 {
 	struct file *file;
 	int ret;
+	int fd;
 
 #if defined(CONFIG_UNIX)
 	ret = sock_create_kern(&init_net, PF_UNIX, SOCK_RAW, IPPROTO_IP,
@@ -9111,12 +9195,12 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 	ret = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
 	if (ret < 0)
 		goto err;
+	fd = ret;
 
 	file = anon_inode_getfile("[io_uring]", &io_uring_fops, ctx,
 					O_RDWR | O_CLOEXEC);
 	if (IS_ERR(file)) {
-err_fd:
-		put_unused_fd(ret);
+		put_unused_fd(fd);
 		ret = PTR_ERR(file);
 		goto err;
 	}
@@ -9124,12 +9208,14 @@ err_fd:
 #if defined(CONFIG_UNIX)
 	ctx->ring_sock->file = file;
 #endif
-	if (unlikely(io_uring_add_task_file(file))) {
-		file = ERR_PTR(-ENOMEM);
-		goto err_fd;
+	ret = io_uring_add_task_file(ctx, file);
+	if (ret) {
+		fput(file);
+		put_unused_fd(fd);
+		goto err;
 	}
-	fd_install(ret, file);
-	return ret;
+	fd_install(fd, file);
+	return fd;
 err:
 #if defined(CONFIG_UNIX)
 	sock_release(ctx->ring_sock);
@@ -9169,7 +9255,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p,
 		 * to a power-of-two, if it isn't already. We do NOT impose
 		 * any cq vs sq ring sizing.
 		 */
-		if (p->cq_entries < p->sq_entries)
+		if (!p->cq_entries)
 			return -EINVAL;
 		if (p->cq_entries > IORING_MAX_CQ_ENTRIES) {
 			if (!(p->flags & IORING_SETUP_CLAMP))
@@ -9177,6 +9263,8 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p,
 			p->cq_entries = IORING_MAX_CQ_ENTRIES;
 		}
 		p->cq_entries = roundup_pow_of_two(p->cq_entries);
+		if (p->cq_entries < p->sq_entries)
+			return -EINVAL;
 	} else {
 		p->cq_entries = 2 * p->sq_entries;
 	}