[PATCH 5/6] io_uring: add parameter region registration

[Pavel Begunkov <asml.silence@gmail.com>]

Allow the user to pre-register a region for passing various paramteres.

To use it for passing wait loop arguments, which is wired in the
following commit, the region has to be registered with the
IORING_PARAM_REGION_WAIT_ARG flag set. The flag also requires the
context to be currently disabled, i.e. IORING_SETUP_R_DISABLED, to avoid
races with otherwise potentially running waiters.

This will also be useful in the future for various request / SQE
arguments like iovec, the meta read/write API, and also for BPF.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 include/linux/io_uring_types.h |  6 ++++
 include/uapi/linux/io_uring.h  | 13 ++++++++
 io_uring/io_uring.c            |  1 +
 io_uring/register.c            | 59 ++++++++++++++++++++++++++++++++++
 4 files changed, 79 insertions(+)

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 1d3a37234ace..aa5f5ea98076 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -324,6 +324,9 @@ struct io_ring_ctx {
 		unsigned		cq_entries;
 		struct io_ev_fd	__rcu	*io_ev_fd;
 		unsigned		cq_extra;
+
+		void			*cq_wait_arg;
+		size_t			cq_wait_size;
 	} ____cacheline_aligned_in_smp;
 
 	/*
@@ -429,6 +432,9 @@ struct io_ring_ctx {
 	unsigned short			n_sqe_pages;
 	struct page			**ring_pages;
 	struct page			**sqe_pages;
+
+	/* used for optimised request parameter and wait argument passing  */
+	struct io_mapped_region		param_region;
 };
 
 struct io_tw_state {
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 7ceeccbbf4cb..49b94029c137 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -627,6 +627,8 @@ enum io_uring_register_op {
 	/* resize CQ ring */
 	IORING_REGISTER_RESIZE_RINGS		= 33,
 
+	IORING_REGISTER_PARAM_REGION		= 34,
+
 	/* this goes last */
 	IORING_REGISTER_LAST,
 
@@ -660,6 +662,17 @@ struct io_uring_region_desc {
 	__u64 __resv[4];
 };
 
+enum {
+	/* expose the region as registered wait arguments */
+	IORING_PARAM_REGION_WAIT_ARG			= 1,
+};
+
+struct io_uring_param_region_reg {
+	__u64 region_uptr; /* struct io_uring_region_desc * */
+	__u64 flags;
+	__u64 __resv[2];
+};
+
 /*
  * Register a fully sparse file space, rather than pass in an array of all
  * -1 file descriptors.
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 286b7bb73978..c640b8a4ceee 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -2709,6 +2709,7 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
 	io_alloc_cache_free(&ctx->msg_cache, io_msg_cache_free);
 	io_futex_cache_free(ctx);
 	io_destroy_buffers(ctx);
+	io_free_region(ctx, &ctx->param_region);
 	mutex_unlock(&ctx->uring_lock);
 	if (ctx->sq_creds)
 		put_cred(ctx->sq_creds);
diff --git a/io_uring/register.c b/io_uring/register.c
index 3c5a3cfb186b..d1ba14da37ea 100644
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -570,6 +570,59 @@ static int io_register_resize_rings(struct io_ring_ctx *ctx, void __user *arg)
 	return ret;
 }
 
+/*
+ * Register a page holding N entries of struct io_uring_reg_wait, which can
+ * be used via io_uring_enter(2) if IORING_GETEVENTS_EXT_ARG_REG is set.
+ * If that is set with IORING_GETEVENTS_EXT_ARG, then instead of passing
+ * in a pointer for a struct io_uring_getevents_arg, an index into this
+ * registered array is passed, avoiding two (arg + timeout) copies per
+ * invocation.
+ */
+static int io_register_mapped_heap(struct io_ring_ctx *ctx, void __user *uarg)
+{
+	struct io_uring_param_region_reg __user *reg_uptr = uarg;
+	struct io_uring_param_region_reg reg;
+	struct io_uring_region_desc __user *rd_uptr;
+	struct io_uring_region_desc rd;
+	int ret;
+
+	if (io_region_is_set(&ctx->param_region))
+		return -EBUSY;
+	if (copy_from_user(&reg, reg_uptr, sizeof(reg)))
+		return -EFAULT;
+	rd_uptr = u64_to_user_ptr(reg.region_uptr);
+	if (copy_from_user(&rd, rd_uptr, sizeof(rd)))
+		return -EFAULT;
+
+	if (memchr_inv(&reg.__resv, 0, sizeof(reg.__resv)))
+		return -EINVAL;
+	if (reg.flags != IORING_PARAM_REGION_WAIT_ARG)
+		return -EINVAL;
+
+	/*
+	 * This ensures there are no waiters. Waiters are unlocked and it's
+	 * hard to synchronise with them, especially if we need to initialise
+	 * the region.
+	 */
+	if ((reg.flags & IORING_PARAM_REGION_WAIT_ARG) &&
+	    !(ctx->flags & IORING_SETUP_R_DISABLED))
+		return -EINVAL;
+
+	ret = io_create_region(ctx, &ctx->param_region, &rd);
+	if (ret)
+		return ret;
+	if (copy_to_user(rd_uptr, &rd, sizeof(rd))) {
+		io_free_region(ctx, &ctx->param_region);
+		return -EFAULT;
+	}
+
+	if (reg.flags & IORING_PARAM_REGION_WAIT_ARG) {
+		ctx->cq_wait_arg = io_region_get_ptr(&ctx->param_region);
+		ctx->cq_wait_size = rd.size;
+	}
+	return 0;
+}
+
 static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 			       void __user *arg, unsigned nr_args)
 	__releases(ctx->uring_lock)
@@ -764,6 +817,12 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 			break;
 		ret = io_register_resize_rings(ctx, arg);
 		break;
+	case IORING_REGISTER_PARAM_REGION:
+		ret = -EINVAL;
+		if (!arg || nr_args != 1)
+			break;
+		ret = io_register_mapped_heap(ctx, arg);
+		break;
 	default:
 		ret = -EINVAL;
 		break;
-- 
2.46.0