[RFC PATCH 04/12] io_uring, fs: plumb support to register/unregister raw-queue

[Kanchan Joshi]

From: Anuj Gupta <anuj20.g at samsung.com>

Extend io_uring's registration interface with

- IORING_REGISTER_QUEUE: to ask for a queue. It goes down via
fops->register_queue and returns identifier of the queue. This qid is
stored in ring's ctx.

- IORING_UNREGISTER_QUEUE: to return the previously registered queue.

At max one queue is allowed to be attached with the io_uring's ring.
The file for which queue is requested is expected to be in
registered file-set.

Signed-off-by: Kanchan Joshi <joshi.k at samsung.com>
Signed-off-by: Anuj Gupta <anuj20.g at samsung.com>
---
 fs/file.c                      | 14 ++++++++
 include/linux/fs.h             |  2 ++
 include/linux/io_uring_types.h |  3 ++
 include/uapi/linux/io_uring.h  |  4 +++
 io_uring/io_uring.c            | 60 ++++++++++++++++++++++++++++++++++
 5 files changed, 83 insertions(+)

diff --git a/fs/file.c b/fs/file.c
index 7893ea161d77..7dada9cd0911 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -1318,3 +1318,17 @@ int iterate_fd(struct files_struct *files, unsigned n,
 	return res;
 }
 EXPORT_SYMBOL(iterate_fd);
+
+int file_register_queue(struct file *file)
+{
+	if (file->f_op->register_queue)
+		return file->f_op->register_queue(file);
+	return -EINVAL;
+}
+
+int file_unregister_queue(struct file *file, int qid)
+{
+	if (file->f_op->unregister_queue)
+		return file->f_op->unregister_queue(file, qid);
+	return -EINVAL;
+}
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 79acccc5e7d4..0a82aac6868b 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -3190,5 +3190,7 @@ extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
 		       int advice);
 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
 			   int advice);
+int file_register_queue(struct file *file);
+int file_unregister_queue(struct file *file, int qid);
 
 #endif /* _LINUX_FS_H */
diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 1b2a20a42413..8d4e721493d6 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -364,6 +364,9 @@ struct io_ring_ctx {
 	unsigned			sq_thread_idle;
 	/* protected by ->completion_lock */
 	unsigned			evfd_last_cq_tail;
+	/* for io_uring attached device queue */
+	int				dev_qid;
+	int				dev_fd;
 };
 
 struct io_tw_state {
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 0716cb17e436..a9d59bfd26f7 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -523,6 +523,10 @@ enum {
 	/* register a range of fixed file slots for automatic slot allocation */
 	IORING_REGISTER_FILE_ALLOC_RANGE	= 25,
 
+	/* register a device-queue with the ring */
+	IORING_REGISTER_QUEUE			= 26,
+	IORING_UNREGISTER_QUEUE			= 27,
+
 	/* this goes last */
 	IORING_REGISTER_LAST,
 
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 3bca7a79efda..5a9b7adf438e 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -337,6 +337,8 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
 	INIT_WQ_LIST(&ctx->locked_free_list);
 	INIT_DELAYED_WORK(&ctx->fallback_work, io_fallback_req_func);
 	INIT_WQ_LIST(&ctx->submit_state.compl_reqs);
+	/* -EINVAL implies nothing is registered with this ring */
+	ctx->dev_qid = -EINVAL;
 	return ctx;
 err:
 	kfree(ctx->dummy_ubuf);
@@ -2822,6 +2824,51 @@ static void io_req_caches_free(struct io_ring_ctx *ctx)
 	mutex_unlock(&ctx->uring_lock);
 }
 
+static int io_register_queue(struct io_ring_ctx *ctx, void __user *arg)
+{
+	struct io_fixed_file *file_slot;
+	struct file *file;
+	__s32 __user *fds = arg;
+	int fd, qid;
+
+	if (ctx->dev_qid != -EINVAL)
+		return -EINVAL;
+	if (copy_from_user(&fd, fds, sizeof(*fds)))
+		return -EFAULT;
+	file_slot = io_fixed_file_slot(&ctx->file_table,
+			array_index_nospec(fd, ctx->nr_user_files));
+	if (!file_slot->file_ptr)
+		return -EBADF;
+	file = (struct file *)(file_slot->file_ptr & FFS_MASK);
+	qid = file_register_queue(file);
+	if (qid < 0)
+		return qid;
+	ctx->dev_fd = fd;
+	ctx->dev_qid = qid;
+	return 0;
+}
+
+static int io_unregister_queue(struct io_ring_ctx *ctx)
+{
+	struct io_fixed_file *file_slot;
+	struct file *file;
+	int ret;
+
+	if (ctx->dev_qid == -EINVAL)
+		return 0;
+	file_slot = io_fixed_file_slot(&ctx->file_table,
+			array_index_nospec(ctx->dev_fd, ctx->nr_user_files));
+	if (!file_slot)
+		return -EBADF;
+	if (!file_slot->file_ptr)
+		return -EBADF;
+	file = (struct file *)(file_slot->file_ptr & FFS_MASK);
+	ret = file_unregister_queue(file, ctx->dev_qid);
+	if (!ret)
+		ctx->dev_qid = -EINVAL;
+	return ret;
+}
+
 static void io_rsrc_node_cache_free(struct io_cache_entry *entry)
 {
 	kfree(container_of(entry, struct io_rsrc_node, cache));
@@ -2835,6 +2882,7 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
 		return;
 
 	mutex_lock(&ctx->uring_lock);
+	io_unregister_queue(ctx);
 	if (ctx->buf_data)
 		__io_sqe_buffers_unregister(ctx);
 	if (ctx->file_data)
@@ -4418,6 +4466,18 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 			break;
 		ret = io_register_file_alloc_range(ctx, arg);
 		break;
+	case IORING_REGISTER_QUEUE:
+		ret = -EINVAL;
+		if (!arg || nr_args != 1)
+			break;
+		ret = io_register_queue(ctx, arg);
+		break;
+	case IORING_UNREGISTER_QUEUE:
+		ret = -EINVAL;
+		if (arg || nr_args)
+			break;
+		ret = io_unregister_queue(ctx);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
-- 
2.25.1