Skip to contentSkip to conversation

multishot recvmsg

RFC PATCH Applied Dylan Yudaken 2 participants 5 messages Jul 14, 2022 io_uring

v2 of 3 Applied Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover multishot recvmsg

This series adds an API (patch 1) and a test (#2) for multishot recvmsg.

I have not included docs yet, but I want to get feedback on the API for handling the result (if there is any).

There was no feedback on v1, but in v2 I added more API to make it easier to deal with truncated payloads. Additionally there is extra testing for this.

Dylan Yudaken (2):
add multishot recvmsg API
add tests for multishot recvmsg

 src/include/liburing.h          |  66 ++++++++++++
 src/include/liburing/io_uring.h |   7 ++
 test/recv-multishot.c           | 180 ++++++++++++++++++++++++++++----
 3 files changed, 234 insertions(+), 19 deletions(-)

base-commit: a705a6b307adfa52ba6315df0b8d5964a3b898b2

Signature · 1 line
2.30.2
#1 lore raw

RFC PATCH v2 1/2 add multishot recvmsg API

This adds a new API to do multishot recvmsg. This is more complicated than multishot recv as it requires handling a well known data layout copied by the kernel.

---
 src/include/liburing.h          | 66 +++++++++++++++++++++++++++++++++
 src/include/liburing/io_uring.h |  7 ++++
 2 files changed, 73 insertions(+)
src/include/liburing.h +66 -0
index d35bfa9..c9e4e75 100644
@@ -419,6 +419,13 @@ static inline void io_uring_prep_recvmsg(struct io_uring_sqe *sqe, int fd,
 	sqe->msg_flags = flags;
 }
 
+static inline void io_uring_prep_recvmsg_multishot(struct io_uring_sqe *sqe, int fd,
+						   struct msghdr *msg, unsigned flags)
+{
+	io_uring_prep_recvmsg(sqe, fd, msg, flags);
+	sqe->ioprio |= IORING_RECV_MULTISHOT;
+}
+
 static inline void io_uring_prep_sendmsg(struct io_uring_sqe *sqe, int fd,
 					 const struct msghdr *msg,
 					 unsigned flags)
@@ -685,6 +692,65 @@ static inline void io_uring_prep_recv_multishot(struct io_uring_sqe *sqe,
 	sqe->ioprio |= IORING_RECV_MULTISHOT;
 }
 
+static inline struct io_uring_recvmsg_out *io_uring_recvmsg_validate(
+	void *buf, int buf_len, struct msghdr *m)
+{
+	size_t header = m->msg_controllen + m->msg_namelen + sizeof(struct io_uring_recvmsg_out);
+
+	if (buf_len < header)
+		return NULL;
+	return (struct io_uring_recvmsg_out *)buf;
+}
+
+static inline void *io_uring_recvmsg_name(struct io_uring_recvmsg_out *o)
+{
+	return (void *)&o[1];
+}
+
+static inline struct cmsghdr *io_uring_recvmsg_cmsg_firsthdr(struct io_uring_recvmsg_out *o,
+							     struct msghdr *m)
+{
+	if (o->controllen < sizeof(struct cmsghdr))
+		return NULL;
+	return (struct cmsghdr *)((unsigned char *)io_uring_recvmsg_name(o) + m->msg_namelen);
+}
+
+static inline void *io_uring_recvmsg_payload(struct io_uring_recvmsg_out *o,
+					     struct msghdr *m)
+{
+	return (void *)((unsigned char *)io_uring_recvmsg_name(o) +
+			m->msg_namelen + m->msg_controllen);
+}
+
+static inline size_t io_uring_recvmsg_payload_length(struct io_uring_recvmsg_out *o,
+						     int buf_length,
+						     struct msghdr *m)
+{
+	unsigned long payload_start = (unsigned long)io_uring_recvmsg_payload(o, m);
+	unsigned long payload_end = (unsigned long)o + buf_length;
+
+	return payload_end - payload_start;
+}
+
+static inline struct cmsghdr *io_uring_recvmsg_cmsg_nexthdr(struct io_uring_recvmsg_out *o,
+							    struct msghdr *m,
+							    struct cmsghdr *cmsg)
+{
+	unsigned char *end;
+
+	if (cmsg->cmsg_len < sizeof(struct cmsghdr))
+		return NULL;
+	end = (unsigned char *)io_uring_recvmsg_payload(o, m);
+	cmsg = (struct cmsghdr *)((unsigned char *)cmsg + CMSG_ALIGN(cmsg->cmsg_len));
+
+	if ((unsigned char *)(cmsg + 1) > end)
+		return NULL;
+	if (((unsigned char *)cmsg) + CMSG_ALIGN(cmsg->cmsg_len) > end)
+		return NULL;
+
+	return cmsg;
+}
+
 static inline void io_uring_prep_openat2(struct io_uring_sqe *sqe, int dfd,
 					const char *path, struct open_how *how)
 {
src/include/liburing/io_uring.h +7 -0
index 6b260e2..51126c4 100644
@@ -573,6 +573,13 @@ struct io_uring_file_index_range {
 	__u64	resv;
 };
 
+struct io_uring_recvmsg_out {
+	__u32 namelen;
+	__u32 controllen;
+	__u32 payloadlen;
+	__u32 flags;
+};
+
 /*
  * accept flags stored in sqe->ioprio
  */
Signature · 1 line
2.30.2
#2 lore raw

RFC PATCH v2 2/2 add tests for multishot recvmsg

In reply to Dylan Yudaken

Expand the multishot recv test to include recvmsg.
This also checks that sockaddr comes back, and that control messages work properly.

---
 test/recv-multishot.c | 180 +++++++++++++++++++++++++++++++++++++-----
 1 file changed, 161 insertions(+), 19 deletions(-)
test/recv-multishot.c +161 -19
index 9df8184..a322e43 100644
@@ -27,20 +27,45 @@ enum early_error_t {
 struct args {
 	bool stream;
 	bool wait_each;
+	bool recvmsg;
 	enum early_error_t early_error;
 };
 
+static int check_sockaddr(struct sockaddr_in *in)
+{
+	struct in_addr expected;
+
+	inet_pton(AF_INET, "127.0.0.1", &expected);
+	if (in->sin_family != AF_INET) {
+		fprintf(stderr, "bad family %d\n", (int)htons(in->sin_family));
+		return -1;
+	}
+	if (memcmp(&expected, &in->sin_addr, sizeof(in->sin_addr))) {
+		char buff[256];
+		const char *addr = inet_ntop(AF_INET, &in->sin_addr, buff, sizeof(buff));
+
+		fprintf(stderr, "unexpected address %s\n", addr ? addr : "INVALID");
+		return -1;
+	}
+	return 0;
+}
+
 static int test(struct args *args)
 {
 	int const N = 8;
 	int const N_BUFFS = N * 64;
 	int const N_CQE_OVERFLOW = 4;
 	int const min_cqes = 2;
+	int const NAME_LEN = sizeof(struct sockaddr_storage);
+	int const CONTROL_LEN = CMSG_ALIGN(sizeof(struct sockaddr_storage))
+					+ sizeof(struct cmsghdr);
 	struct io_uring ring;
 	struct io_uring_cqe *cqe;
 	struct io_uring_sqe *sqe;
-	int fds[2], ret, i, j, total_sent_bytes = 0, total_recv_bytes = 0;
+	int fds[2], ret, i, j;
+	int total_sent_bytes = 0, total_recv_bytes = 0, total_dropped_bytes = 0;
 	int send_buff[256];
+	int *sent_buffs[N_BUFFS];
 	int *recv_buffs[N_BUFFS];
 	int *at;
 	struct io_uring_cqe recv_cqe[N_BUFFS];
@@ -50,7 +75,7 @@ static int test(struct args *args)
 	struct __kernel_timespec timeout = {
 		.tv_sec = 1,
 	};
-
+	struct msghdr msg;
 
 	memset(recv_buffs, 0, sizeof(recv_buffs));
 
@@ -75,21 +100,42 @@ static int test(struct args *args)
 		return ret;
 	}
 
+	if (!args->stream) {
+		bool val = true;
+
+		/* force some cmsgs to come back to us */
+		ret = setsockopt(fds[0], IPPROTO_IP, IP_RECVORIGDSTADDR, &val,
+				 sizeof(val));
+		if (ret) {
+			fprintf(stderr, "setsockopt failed %d\n", errno);
+			goto cleanup;
+		}
+	}
+
 	for (i = 0; i < ARRAY_SIZE(send_buff); i++)
 		send_buff[i] = i;
 
 	for (i = 0; i < ARRAY_SIZE(recv_buffs); i++) {
 		/* prepare some different sized buffers */
-		int buffer_size = (i % 2 == 0 && args->stream) ? 1 : N * sizeof(int);
+		int buffer_size = (i % 2 == 0 && (args->stream || args->recvmsg)) ? 1 : N;
+
+		buffer_size *= sizeof(int);
+		if (args->recvmsg) {
+			buffer_size +=
+				sizeof(struct io_uring_recvmsg_out) +
+				NAME_LEN +
+				CONTROL_LEN;
+		}
 
-		recv_buffs[i] = malloc(sizeof(*at) * buffer_size);
+		recv_buffs[i] = malloc(buffer_size);
 
 		if (i > 2 && args->early_error == ERROR_NOT_ENOUGH_BUFFERS)
 			continue;
 
 		sqe = io_uring_get_sqe(&ring);
 		io_uring_prep_provide_buffers(sqe, recv_buffs[i],
-					buffer_size * sizeof(*recv_buffs[i]), 1, 7, i);
+					buffer_size, 1, 7, i);
+		memset(recv_buffs[i], 0xcc, buffer_size);
 		if (io_uring_submit_and_wait_timeout(&ring, &cqe, 1, &timeout, NULL) != 0) {
 			fprintf(stderr, "provide buffers failed: %d\n", ret);
 			ret = -1;
@@ -99,7 +145,19 @@ static int test(struct args *args)
 	}
 
 	sqe = io_uring_get_sqe(&ring);
-	io_uring_prep_recv_multishot(sqe, fds[0], NULL, 0, 0);
+	if (args->recvmsg) {
+		unsigned int flags = 0;
+
+		if (!args->stream)
+			flags |= MSG_TRUNC;
+
+		memset(&msg, 0, sizeof(msg));
+		msg.msg_namelen = NAME_LEN;
+		msg.msg_controllen = CONTROL_LEN;
+		io_uring_prep_recvmsg_multishot(sqe, fds[0], &msg, flags);
+	} else {
+		io_uring_prep_recv_multishot(sqe, fds[0], NULL, 0, 0);
+	}
 	sqe->flags |= IOSQE_BUFFER_SELECT;
 	sqe->buf_group = 7;
 	io_uring_sqe_set_data64(sqe, 1234);
@@ -111,6 +169,7 @@ static int test(struct args *args)
 		int to_send = sizeof(*at) * (i+1);
 
 		total_sent_bytes += to_send;
+		sent_buffs[i] = at;
 		if (send(fds[1], at, to_send, 0) != to_send) {
 			if (early_error_started)
 				break;
@@ -202,9 +261,12 @@ static int test(struct args *args)
 			(args->early_error == ERROR_EARLY_OVERFLOW &&
 			 !args->wait_each && i == N_CQE_OVERFLOW);
 		int *this_recv;
+		int orig_payload_size = cqe->res;
 
 
 		if (should_be_last) {
+			int used_res = cqe->res;
+
 			if (!is_last) {
 				fprintf(stderr, "not last cqe had error %d\n", i);
 				goto cleanup;
@@ -234,7 +296,22 @@ static int test(struct args *args)
 				break;
 			case ERROR_NONE:
 			case ERROR_EARLY_CLOSE_SENDER:
-				if (cqe->res != 0) {
+				if (args->recvmsg && (cqe->flags & IORING_CQE_F_BUFFER)) {
+					void *buff = recv_buffs[cqe->flags >> 16];
+					struct io_uring_recvmsg_out *o =
+						io_uring_recvmsg_validate(buff, cqe->res, &msg);
+
+					if (!o) {
+						fprintf(stderr, "invalid buff\n");
+						goto cleanup;
+					}
+					if (o->payloadlen != 0) {
+						fprintf(stderr, "expected 0 payloadlen, got %u\n",
+							o->payloadlen);
+						goto cleanup;
+					}
+					used_res = 0;
+				} else if (cqe->res != 0) {
 					fprintf(stderr, "early error: res %d\n", cqe->res);
 					goto cleanup;
 				}
@@ -254,7 +331,7 @@ static int test(struct args *args)
 				goto cleanup;
 			}
 
-			if (cqe->res <= 0)
+			if (used_res <= 0)
 				continue;
 		} else {
 			if (!(cqe->flags & IORING_CQE_F_MORE)) {
@@ -268,7 +345,61 @@ static int test(struct args *args)
 			goto cleanup;
 		}
 
+		this_recv = recv_buffs[cqe->flags >> 16];
+
+		if (args->recvmsg) {
+			struct io_uring_recvmsg_out *o = io_uring_recvmsg_validate(
+				this_recv, cqe->res, &msg);
+
+			if (!o) {
+				fprintf(stderr, "bad recvmsg\n");
+				goto cleanup;
+			}
+			orig_payload_size = o->payloadlen;
+
+			if (!args->stream) {
+				orig_payload_size = o->payloadlen;
+
+				struct cmsghdr *cmsg;
+
+				if (o->namelen < sizeof(struct sockaddr_in)) {
+					fprintf(stderr, "bad addr len %d",
+						o->namelen);
+					goto cleanup;
+				}
+				if (check_sockaddr((struct sockaddr_in *)io_uring_recvmsg_name(o)))
+					goto cleanup;
+
+				cmsg = io_uring_recvmsg_cmsg_firsthdr(o, &msg);
+				if (!cmsg ||
+				    cmsg->cmsg_level != IPPROTO_IP ||
+				    cmsg->cmsg_type != IP_RECVORIGDSTADDR) {
+					fprintf(stderr, "bad cmsg");
+					goto cleanup;
+				}
+				if (check_sockaddr((struct sockaddr_in *)CMSG_DATA(cmsg)))
+					goto cleanup;
+				cmsg = io_uring_recvmsg_cmsg_nexthdr(o, &msg, cmsg);
+				if (cmsg) {
+					fprintf(stderr, "unexpected extra cmsg\n");
+					goto cleanup;
+				}
+
+			}
+
+			this_recv = (int *)io_uring_recvmsg_payload(o, &msg);
+			cqe->res = io_uring_recvmsg_payload_length(o, cqe->res, &msg);
+			if (o->payloadlen != cqe->res) {
+				if (!(o->flags & MSG_TRUNC)) {
+					fprintf(stderr, "expected truncated flag\n");
+					goto cleanup;
+				}
+				total_dropped_bytes += (o->payloadlen - cqe->res);
+			}
+		}
+
 		total_recv_bytes += cqe->res;
+
 		if (cqe->res % 4 != 0) {
 			/*
 			 * doesn't seem to happen in practice, would need some
@@ -278,9 +409,20 @@ static int test(struct args *args)
 			goto cleanup;
 		}
 
-		/* check buffer arrived in order (for tcp) */
-		this_recv = recv_buffs[cqe->flags >> 16];
-		for (j = 0; args->stream && j < cqe->res / 4; j++) {
+		/*
+		 * for tcp: check buffer arrived in order
+		 * for udp: based on size validate data based on size
+		 */
+		if (!args->stream) {
+			int sent_idx = orig_payload_size / sizeof(*at) - 1;
+
+			if (sent_idx < 0 || sent_idx > N) {
+				fprintf(stderr, "Bad sent idx: %d\n", sent_idx);
+				goto cleanup;
+			}
+			at = sent_buffs[sent_idx];
+		}
+		for (j = 0; j < cqe->res / 4; j++) {
 			int sent = *at++;
 			int recv = *this_recv++;
 
@@ -291,15 +433,14 @@ static int test(struct args *args)
 		}
 	}
 
-	if (args->early_error == ERROR_NONE && total_recv_bytes < total_sent_bytes) {
+	if (args->early_error == ERROR_NONE &&
+	    total_recv_bytes + total_dropped_bytes < total_sent_bytes) {
 		fprintf(stderr,
-			"missing recv: recv=%d sent=%d\n", total_recv_bytes, total_sent_bytes);
+			"missing recv: recv=%d dropped=%d sent=%d\n",
+			total_recv_bytes, total_sent_bytes, total_dropped_bytes);
 		goto cleanup;
 	}
 
-	/* check the final one */
-	cqe = &recv_cqe[recv_cqes-1];
-
 	ret = 0;
 cleanup:
 	for (i = 0; i < ARRAY_SIZE(recv_buffs); i++)
@@ -320,18 +461,19 @@ int main(int argc, char *argv[])
 	if (argc > 1)
 		return T_EXIT_SKIP;
 
-	for (loop = 0; loop < 4; loop++) {
+	for (loop = 0; loop < 8; loop++) {
 		struct args a = {
 			.stream = loop & 0x01,
 			.wait_each = loop & 0x2,
+			.recvmsg = loop & 0x04,
 		};
 		for (early_error = 0; early_error < ERROR_EARLY_LAST; early_error++) {
 			a.early_error = (enum early_error_t)early_error;
 			ret = test(&a);
 			if (ret) {
 				fprintf(stderr,
-					"test stream=%d wait_each=%d early_error=%d failed\n",
-					a.stream, a.wait_each, a.early_error);
+					"test stream=%d wait_each=%d recvmsg=%d early_error=%d failed\n",
+					a.stream, a.wait_each, a.recvmsg, a.early_error);
 				return T_EXIT_FAIL;
 			}
 			if (no_recv_mshot)
Signature · 1 line
2.30.2
maintainer
#3 lore raw
In reply to Dylan Yudaken
Quoted 9 lines · Dylan Yudaken

This series adds an API (patch 1) and a test (#2) for multishot recvmsg.

I have not included docs yet, but I want to get feedback on the API for handling the result (if there is any).

There was no feedback on v1, but in v2 I added more API to make it easier to deal with truncated payloads. Additionally there is extra testing for this.

I'm going to shove this in for some testing. We have a bit of time to play with before the next release, might make sense to attempt to mature it in tree.

Signature · 1 line
Jens Axboe
maintainer Applied
#4 lore raw
In reply to Dylan Yudaken
Quoted 9 lines · Dylan Yudaken

This series adds an API (patch 1) and a test (#2) for multishot recvmsg.

I have not included docs yet, but I want to get feedback on the API for handling the result (if there is any).

There was no feedback on v1, but in v2 I added more API to make it easier to deal with truncated payloads. Additionally there is extra testing for this.

[...]

Applied, thanks!

[1/2] add multishot recvmsg API

      commit: 874406f7fb0942238a77041d84f07a480400ed81
[2/2] add tests for multishot recvmsg
      commit: 95dc4789e9dbacb7b3913b1a6e0dfef80fabb74d

Best regards,

Signature · 1 line
Jens Axboe