Update API

- io_uring_sqe added a data field that's passed back at completion
- io_uring_sqe added an index field, for fixed buffer locations
- io_uring_setup(2) system call added a 'nr_iovecs' field

Signed-off-by: Jens Axboe <axboe@kernel.dk>

6 files changed, 40 insertions, 30 deletions

diff --git a/src/io_uring.c b/src/io_uring.c
index 6168b56..d9c1511 100644
--- a/src/io_uring.c
+++ b/src/io_uring.c
@@ -189,11 +189,12 @@ err:
  * contains the necessary information to read/write to the rings.
  */
 int io_uring_queue_init(unsigned entries, struct io_uring_params *p,
-			struct iovec *iovecs, struct io_uring *ring)
+			struct iovec *iovecs, unsigned nr_iovecs,
+			struct io_uring *ring)
 {
 	int fd, ret;
 
-	fd = io_uring_setup(entries, iovecs, p);
+	fd = io_uring_setup(entries, iovecs, nr_iovecs, p);
 	if (fd < 0)
 		return fd;
 
diff --git a/src/io_uring.h b/src/io_uring.h
index 20e4c22..b07bbbb 100644
--- a/src/io_uring.h
+++ b/src/io_uring.h
@@ -15,20 +15,23 @@
  * IO submission data structure (Submission Queue Entry)
  */
 struct io_uring_sqe {
-	__u8	opcode;
-	__u8	flags;
-	__u16	ioprio;
-	__s32	fd;
-	__u64	off;
+	__u8	opcode;		/* type of operation for this sqe */
+	__u8	flags;		/* as of now unused */
+	__u16	ioprio;		/* ioprio for the request */
+	__s32	fd;		/* file descriptor to do IO on */
+	__u64	off;		/* offset into file */
 	union {
-		void	*addr;
+		void	*addr;	/* buffer or iovecs */
 		__u64	__pad;
 	};
-	__u32	len;
+	__u32	len;		/* buffer size or number of iovecs */
 	union {
 		__kernel_rwf_t	rw_flags;
 		__u32		__resv;
 	};
+	__u16	index;		/* index into fixed buffers, if used */
+	__u16	__pad2[3];
+	__u64	data;		/* data to be passed back at completion time */
 };
 
 /*
@@ -50,7 +53,7 @@ struct io_uring_sqe {
  * IO completion data structure (Completion Queue Entry)
  */
 struct io_uring_cqe {
-	__u64	index;		/* what sqe this event came from */
+	__u64	data;		/* sqe->data submission passed back */
 	__s32	res;		/* result code for this event */
 	__u32	flags;
 };
diff --git a/src/liburing.h b/src/liburing.h
index 8adc9ec..ac645f1 100644
--- a/src/liburing.h
+++ b/src/liburing.h
@@ -44,7 +44,7 @@ struct io_uring {
  * System calls
  */
 extern int io_uring_setup(unsigned entries, struct iovec *iovecs,
-	struct io_uring_params *p);
+	unsigned nr_iovecs, struct io_uring_params *p);
 extern int io_uring_enter(unsigned fd, unsigned to_submit,
 	unsigned min_complete, unsigned flags);
 
@@ -52,7 +52,7 @@ extern int io_uring_enter(unsigned fd, unsigned to_submit,
  * Library interface
  */
 extern int io_uring_queue_init(unsigned entries, struct io_uring_params *p,
-	struct iovec *iovecs, struct io_uring *ring);
+	struct iovec *iovecs, unsigned nr_iovecs, struct io_uring *ring);
 extern void io_uring_queue_exit(struct io_uring *ring);
 extern int io_uring_get_completion(struct io_uring *ring,
 	struct io_uring_cqe **cqe_ptr);
@@ -61,10 +61,4 @@ extern int io_uring_wait_completion(struct io_uring *ring,
 extern int io_uring_submit(struct io_uring *ring);
 extern struct io_uring_sqe *io_uring_get_sqe(struct io_uring *ring);
 
-static inline struct io_uring_sqe *
-io_uring_sqe_from_cqe(struct io_uring *ring, struct io_uring_cqe *cqe)
-{
-	return &ring->sq.sqes[cqe->index];
-}
-
 #endif
diff --git a/src/syscall.c b/src/syscall.c
index eafdd39..d06d59c 100644
--- a/src/syscall.c
+++ b/src/syscall.c
@@ -18,9 +18,9 @@
 #endif
 
 int io_uring_setup(unsigned int entries, struct iovec *iovecs,
-		   struct io_uring_params *p)
+		   unsigned nr_iovecs, struct io_uring_params *p)
 {
-	return syscall(__NR_sys_io_uring_setup, entries, iovecs, p);
+	return syscall(__NR_sys_io_uring_setup, entries, iovecs, nr_iovecs, p);
 }
 
 int io_uring_enter(int fd, unsigned int to_submit, unsigned int min_complete,
diff --git a/test/io_uring-cp.c b/test/io_uring-cp.c
index 176d9de..9d7c56c 100644
--- a/test/io_uring-cp.c
+++ b/test/io_uring-cp.c
@@ -18,6 +18,11 @@ static struct io_uring in_ring;
 static struct io_uring out_ring;
 static struct iovec iovecs[QD];
 
+struct io_data {
+	off_t offset;
+	struct iovec *iov;
+};
+
 static int setup_context(unsigned entries, struct io_uring *ring, int offload)
 {
 	struct io_uring_params p;
@@ -27,7 +32,7 @@ static int setup_context(unsigned entries, struct io_uring *ring, int offload)
 	if (offload)
 		p.flags = IORING_SETUP_SQWQ;
 
-	ret = io_uring_queue_init(entries, &p, NULL, ring);
+	ret = io_uring_queue_init(entries, &p, NULL, 0, ring);
 	if (ret < 0) {
 		fprintf(stderr, "queue_init: %s\n", strerror(-ret));
 		return -1;
@@ -58,17 +63,23 @@ static unsigned sqe_index(struct io_uring_sqe *sqe)
 static int queue_read(int fd, off_t size, off_t offset)
 {
 	struct io_uring_sqe *sqe;
+	struct io_data *data;
 
 	sqe = io_uring_get_sqe(&in_ring);
 	if (!sqe)
 		return 1;
 
+	data = malloc(sizeof(*data));
+	data->offset = offset;
+	data->iov = &iovecs[sqe_index(sqe)];
+
 	sqe->opcode = IORING_OP_READV;
 	sqe->flags = 0;
 	sqe->ioprio = 0;
 	sqe->fd = fd;
 	sqe->off = offset;
-	sqe->addr = &iovecs[sqe_index(sqe)];
+	sqe->addr = data->iov;
+	sqe->data = (unsigned long) data;
 	iovecs[sqe_index(sqe)].iov_len = size;
 	sqe->len = 1;
 	return 0;
@@ -105,8 +116,9 @@ static int complete_writes(unsigned *writes)
 	return 0;
 }
 
-static void queue_write(int fd, off_t size, off_t offset, unsigned index)
+static void queue_write(int fd, struct io_uring_cqe *cqe)
 {
+	struct io_data *data = (struct io_data *) cqe->data;
 	struct io_uring_sqe *sqe;
 
 	sqe = io_uring_get_sqe(&out_ring);
@@ -114,10 +126,12 @@ static void queue_write(int fd, off_t size, off_t offset, unsigned index)
 	sqe->flags = 0;
 	sqe->ioprio = 0;
 	sqe->fd = fd;
-	sqe->off = offset;
-	sqe->addr = &iovecs[index];
-	iovecs[index].iov_len = size;
+	sqe->off = data->offset;
+	sqe->addr = data->iov;
+	sqe->data = 0;
+	data->iov->iov_len = cqe->res;
 	sqe->len = 1;
+	free(data);
 }
 
 int main(int argc, char *argv[])
@@ -164,7 +178,6 @@ int main(int argc, char *argv[])
 	write_left = read_left;
 	while (read_left || write_left) {
 		off_t this_size = read_left;
-		struct io_uring_sqe *sqe;
 
 		if (this_size > BS)
 			this_size = BS;
@@ -209,8 +222,7 @@ skip_read:
 						strerror(-cqe->res));
 				return 1;
 			}
-			sqe = io_uring_sqe_from_cqe(&in_ring, cqe);
-			queue_write(outfd, cqe->res, sqe->off, cqe->index);
+			queue_write(outfd, cqe);
 			write_left -= cqe->res;
 			writes++;
 		};
diff --git a/test/io_uring-test.c b/test/io_uring-test.c
index aaf7e66..caf61ed 100644
--- a/test/io_uring-test.c
+++ b/test/io_uring-test.c
@@ -32,7 +32,7 @@ int main(int argc, char *argv[])
 	memset(&p, 0, sizeof(p));
 	p.flags = IORING_SETUP_IOPOLL;
 
-	ret = io_uring_queue_init(QD, &p, NULL, &ring);
+	ret = io_uring_queue_init(QD, &p, NULL, 0, &ring);
 	if (ret < 0) {
 		fprintf(stderr, "queue_init: %s\n", strerror(-ret));
 		return 1;