vhost: support inflight share memory protocol feature

This patch introduces two new messages VHOST_USER_GET_INFLIGHT_FD and VHOST_USER_SET_INFLIGHT_FD to support transferring a shared buffer between qemu and backend.

Signed-off-by Lin Li <[email protected]>

Signed-off-by Xun Ni <[email protected]>

Signed-off-by Yu Zhang <[email protected]>

Signed-off-by Jin Yu <[email protected]>

---
 lib/librte_vhost/vhost.h      |   7 +
 lib/librte_vhost/vhost_user.c | 260 +++++++++++++++++++++++++++++++++-
 lib/librte_vhost/vhost_user.h |   4 +-
 3 files changed, 269 insertions(+), 2 deletions(-)

lib/librte_vhost/vhost.h +7 -0

index 884befa85..d67ba849a 100644
@@ -286,6 +286,12 @@ struct guest_page {
 	uint64_t size;
 };
 
+struct inflight_mem_info {
+	int		fd;
+	void		*addr;
+	uint64_t	size;
+};
+
 /**
  * Device structure contains all configuration information relating
  * to the device.

@@ -303,6 +309,7 @@ struct virtio_net {
 	uint32_t		nr_vring;
 	int			dequeue_zero_copy;
 	struct vhost_virtqueue	*virtqueue[VHOST_MAX_QUEUE_PAIRS * 2];
+	struct inflight_mem_info *inflight_info;
 #define IF_NAME_SZ (PATH_MAX > IFNAMSIZ ? PATH_MAX : IFNAMSIZ)
 	char			ifname[IF_NAME_SZ];
 	uint64_t		log_size;

lib/librte_vhost/vhost_user.c +4 -0

index c9e29ece8..66582a30f 100644
@@ -37,6 +37,10 @@
 #ifdef RTE_LIBRTE_VHOST_POSTCOPY
 #include <linux/userfaultfd.h>
 #endif
+#ifdef F_ADD_SEALS /* if file sealing is supported, so is memfd */
+#include <linux/memfd.h>
+#define MEMFD_SUPPORTED
+#endif
 
 #include <rte_common.h>
 #include <rte_malloc.h>

lib/librte_vhost/vhost_user.c +13 -0

+#define VIRTQUEUE_MAX_SIZE	1024
+
+#define CLOEXEC		0x0001U
+
+#define ALIGN_DOWN(n, m) ((n) / (m) * (m))
+#define ALIGN_UP(n, m) ALIGN_DOWN((n) + (m) - 1, (m))
+
 static const char *vhost_message_str[VHOST_USER_MAX] = {
 	[VHOST_USER_NONE] = "VHOST_USER_NONE",
 	[VHOST_USER_GET_FEATURES] = "VHOST_USER_GET_FEATURES",

@@ -78,6 +91,8 @@ static const char *vhost_message_str[VHOST_USER_MAX] = {
 	[VHOST_USER_POSTCOPY_ADVISE]  = "VHOST_USER_POSTCOPY_ADVISE",
 	[VHOST_USER_POSTCOPY_LISTEN]  = "VHOST_USER_POSTCOPY_LISTEN",
 	[VHOST_USER_POSTCOPY_END]  = "VHOST_USER_POSTCOPY_END",
+	[VHOST_USER_GET_INFLIGHT_FD] = "VHOST_USER_GET_INFLIGHT_FD",
+	[VHOST_USER_SET_INFLIGHT_FD] = "VHOST_USER_SET_INFLIGHT_FD",
 };
 
 static int send_vhost_reply(int sockfd, struct VhostUserMsg *msg);

@@ -160,6 +175,21 @@ vhost_backend_cleanup(struct virtio_net *dev)
 		dev->log_addr = 0;
 	}
 
+	if (dev->inflight_info->addr) {
+		munmap(dev->inflight_info->addr, dev->inflight_info->size);
+		dev->inflight_info->addr = NULL;
+	}

lib/librte_vhost/vhost_user.c +1 -0

+
 	if (dev->slave_req_fd >= 0) {
 		close(dev->slave_req_fd);
 		dev->slave_req_fd = -1;

@@ -1165,6 +1195,233 @@ virtio_is_ready(struct virtio_net *dev)
 	return 1;
 }

lib/librte_vhost/vhost_user.c +76 -0

+	return -1;
+#endif
+}
+
+static void *
+inflight_mem_alloc(const char *name, size_t size, int *fd)
+{
+	void *ptr;
+	int mfd = -1;
+	char fname[20] = "/tmp/memfd-XXXXXX";
+
+	*fd = -1;
+	mfd = mem_create(name, CLOEXEC);
+	if (mfd != -1) {
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	} else {
+		mfd = mkstemp(fname);
+		unlink(fname);
+
+		if (mfd == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"mkstemp fail for alloc inflight buffer\n");
+			return NULL;
+		}
+
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	}
+
+	ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
+	if (ptr == MAP_FAILED) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"mmap fail for alloc inflight buffer\n");
+		close(mfd);
+		return NULL;
+	}
+
+	*fd = mfd;
+	return ptr;
+}
+
+static uint32_t
+get_pervq_shm_size_split(uint16_t queue_size)
+{
+	return ALIGN_UP(sizeof(struct rte_vhost_inflight_desc_split) *
+			queue_size + sizeof(uint64_t) + sizeof(uint16_t) * 4,
+			INFLIGHT_ALIGNMENT);
+}
+
+static uint32_t
+get_pervq_shm_size_packed(uint16_t queue_size)
+{
+	return ALIGN_UP(sizeof(struct rte_vhost_inflight_desc_packed) *
+			queue_size + sizeof(uint64_t) + sizeof(uint16_t) * 6 +
+			sizeof(uint8_t) * 9, INFLIGHT_ALIGNMENT);
+}
+
+static int
+vhost_user_get_inflight_fd(struct virtio_net **pdev,
+				     VhostUserMsg *msg,
+				     int main_fd __rte_unused)
+{
+	int fd, i, j;
+	uint64_t pervq_inflight_size, mmap_size;
+	void *addr;
+	uint16_t num_queues, queue_size;
+	struct virtio_net *dev = *pdev;

lib/librte_vhost/vhost_user.c +104 -0

+	if (!dev->inflight_info) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Failed to alloc dev inflight area");
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	num_queues = msg->payload.inflight.num_queues;
+	queue_size = msg->payload.inflight.queue_size;
+
+	RTE_LOG(INFO, VHOST_CONFIG, "get_inflight_fd num_queues: %u\n",
+		msg->payload.inflight.num_queues);
+	RTE_LOG(INFO, VHOST_CONFIG, "get_inflight_fd queue_size: %u\n",
+		msg->payload.inflight.queue_size);
+
+	if (vq_is_packed(dev))
+		pervq_inflight_size = get_pervq_shm_size_packed(queue_size);
+	else
+		pervq_inflight_size = get_pervq_shm_size_split(queue_size);
+
+	mmap_size = num_queues * pervq_inflight_size;
+	addr = inflight_mem_alloc("vhost-inflight", mmap_size, &fd);
+	if (!addr) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Failed to alloc vhost inflight area");
+			msg->payload.inflight.mmap_size = 0;
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+	memset(addr, 0, mmap_size);
+
+	dev->inflight_info->addr = addr;
+	dev->inflight_info->size = msg->payload.inflight.mmap_size = mmap_size;
+	dev->inflight_info->fd = msg->fds[0] = fd;
+	msg->payload.inflight.mmap_offset = 0;
+	msg->fd_num = 1;
+
+	if (vq_is_packed(dev)) {
+		for (i = 0; i < num_queues; i++) {
+			inflight_packed =
+				(struct rte_vhost_inflight_info_packed *)addr;
+			inflight_packed->used_wrap_counter = 1;
+			inflight_packed->old_used_wrap_counter = 1;
+			for (j = 0; j < queue_size; j++)
+				inflight_packed->desc[j].next = j + 1;
+			addr = (void *)((char *)addr + pervq_inflight_size);
+		}
+	}
+
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight mmap_size: %"PRIu64"\n",
+		msg->payload.inflight.mmap_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight mmap_offset: %"PRIu64"\n",
+		msg->payload.inflight.mmap_offset);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight fd: %d\n", msg->fds[0]);
+
+	return RTE_VHOST_MSG_RESULT_REPLY;
+}
+
+static int
+vhost_user_set_inflight_fd(struct virtio_net **pdev, VhostUserMsg *msg,
+		int main_fd __rte_unused)
+{
+	int fd;
+	uint64_t mmap_size, mmap_offset;
+	uint16_t num_queues, queue_size;
+	uint32_t pervq_inflight_size;
+	void *addr;
+	struct virtio_net *dev = *pdev;
+
+	fd = msg->fds[0];
+	if (msg->size != sizeof(msg->payload.inflight) || fd < 0) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Invalid set_inflight_fd message size is %d,fd is %d\n",
+			msg->size, fd);
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	mmap_size = msg->payload.inflight.mmap_size;
+	mmap_offset = msg->payload.inflight.mmap_offset;
+	num_queues = msg->payload.inflight.num_queues;
+	queue_size = msg->payload.inflight.queue_size;
+
+	if (vq_is_packed(dev))
+		pervq_inflight_size = get_pervq_shm_size_packed(queue_size);
+	else
+		pervq_inflight_size = get_pervq_shm_size_split(queue_size);
+
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd mmap_size: %"PRIu64"\n", mmap_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd mmap_offset: %"PRIu64"\n", mmap_offset);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd num_queues: %u\n", num_queues);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd queue_size: %u\n", queue_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd fd: %d\n", fd);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd pervq_inflight_size: %d\n",
+		pervq_inflight_size);
+
+	if (!dev->inflight_info) {
+		dev->inflight_info = calloc(1,

lib/librte_vhost/vhost_user.c +20 -1

+
+	addr = mmap(0, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED,
+			fd, mmap_offset);
+	if (addr == MAP_FAILED) {
+		RTE_LOG(ERR, VHOST_CONFIG, "failed to mmap share memory.\n");
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	if (dev->inflight_info->fd)
+		close(dev->inflight_info->fd);
+
+	dev->inflight_info->fd = fd;
+	dev->inflight_info->addr = addr;
+	dev->inflight_info->size = mmap_size;
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
 static int
 vhost_user_set_vring_call(struct virtio_net **pdev, struct VhostUserMsg *msg,
 			int main_fd __rte_unused)

@@ -1762,9 +2019,10 @@ static vhost_message_handler_t vhost_message_handlers[VHOST_USER_MAX] = {
 	[VHOST_USER_POSTCOPY_ADVISE] = vhost_user_set_postcopy_advise,
 	[VHOST_USER_POSTCOPY_LISTEN] = vhost_user_set_postcopy_listen,
 	[VHOST_USER_POSTCOPY_END] = vhost_user_postcopy_end,
+	[VHOST_USER_GET_INFLIGHT_FD] = vhost_user_get_inflight_fd,
+	[VHOST_USER_SET_INFLIGHT_FD] = vhost_user_set_inflight_fd,
 };
 
-
 /* return bytes# of read on success or negative val on failure. */
 static int
 read_vhost_message(int sockfd, struct VhostUserMsg *msg)

lib/librte_vhost/vhost_user.h +3 -1

index 17a1d7bca..6563f7315 100644
@@ -54,7 +54,9 @@ typedef enum VhostUserRequest {
 	VHOST_USER_POSTCOPY_ADVISE = 28,
 	VHOST_USER_POSTCOPY_LISTEN = 29,
 	VHOST_USER_POSTCOPY_END = 30,
-	VHOST_USER_MAX = 31
+	VHOST_USER_GET_INFLIGHT_FD = 31,
+	VHOST_USER_SET_INFLIGHT_FD = 32,
+	VHOST_USER_MAX = 33
 } VhostUserRequest;
 
 typedef enum VhostUserSlaveRequest {

Signature · 1 line

2.17.2

-----Original Message-----
From: Bie, Tiwei
Sent: Thursday, September 26, 2019 3:39 PM
To: Yu, Jin <[email protected]>
Cc: [email protected]; Liu, Changpeng <[email protected]>; [email protected]; Wang, Zhihong <[email protected]>; Lin Li <[email protected]>; Xun Ni <[email protected]>; Yu Zhang <[email protected]>
Subject: Re: [PATCH v7 04/10] vhost: add two new messages to support a shared buffer

Quoted 53 lines · Jin Yu

This patch introduces two new messages VHOST_USER_GET_INFLIGHT_FD and VHOST_USER_SET_INFLIGHT_FD to support transferring a shared buffer between qemu and backend.

Signed-off-by Lin Li <[email protected]>

Signed-off-by Xun Ni <[email protected]>

Signed-off-by Yu Zhang <[email protected]>

Signed-off-by Jin Yu <[email protected]>

---
 lib/librte_vhost/vhost.h      |   7 +
 lib/librte_vhost/vhost_user.c | 260 +++++++++++++++++++++++++++++++++-
 lib/librte_vhost/vhost_user.h |   4 +-
 3 files changed, 269 insertions(+), 2 deletions(-)

lib/librte_vhost/vhost.h index renamed from lib/librte_vhost/vhost.h +0 -0

884befa85..d67ba849a 100644

lib/librte_vhost/vhost.h +7 -0

@@ -286,6 +286,12 @@ struct guest_page {
 	uint64_t size;
 };
 
+struct inflight_mem_info {
+	int		fd;
+	void		*addr;
+	uint64_t	size;
+};
+
 /**
  * Device structure contains all configuration information relating
  * to the device.

@@ -303,6 +309,7 @@ struct virtio_net {
 	uint32_t		nr_vring;
 	int			dequeue_zero_copy;
 	struct vhost_virtqueue	*virtqueue[VHOST_MAX_QUEUE_PAIRS * 2];
+	struct inflight_mem_info *inflight_info;
 #define IF_NAME_SZ (PATH_MAX > IFNAMSIZ ? PATH_MAX : IFNAMSIZ)
 	char			ifname[IF_NAME_SZ];
 	uint64_t		log_size;

(quoted diff) +0 -0

b/lib/librte_vhost/vhost_user.c index c9e29ece8..66582a30f 100644

lib/librte_vhost/vhost_user.c +2 -0

@@ -37,6 +37,10 @@
 #ifdef RTE_LIBRTE_VHOST_POSTCOPY
 #include <linux/userfaultfd.h>
 #endif
+#ifdef F_ADD_SEALS /* if file sealing is supported, so is memfd */
+#include <linux/memfd.h> #define MEMFD_SUPPORTED #endif
 
 #include <rte_common.h>
 #include <rte_malloc.h>

lib/librte_vhost/vhost_user.c line 57

Quoted 5 lines

lib/librte_vhost/vhost_user.c +2 -0

@@ -49,6 +53,15 @@
 #define VIRTIO_MIN_MTU 68
 #define VIRTIO_MAX_MTU 65535
 
+#define INFLIGHT_ALIGNMENT	64
+#define INFLIGHT_VERSION	0xabcd

Why the version is 0xabcd? Shouldn't it be 1?

https://git.qemu.org/?p=qemu.git;a=blob;f=docs/interop/vhost-
user.rst;h=7827b710aa0a223f6f184b229f4b5ced6420b638;hb=HEAD#l551
https://git.qemu.org/?p=qemu.git;a=blob;f=docs/interop/vhost-
user.rst;h=7827b710aa0a223f6f184b229f4b5ced6420b638;hb=HEAD#l658

lib/librte_vhost/vhost_user.c

Quoted 11 lines

lib/librte_vhost/vhost_user.c +7 -0

+#define VIRTQUEUE_MAX_SIZE	1024
+
+#define CLOEXEC		0x0001U
+
+#define ALIGN_DOWN(n, m) ((n) / (m) * (m)) #define ALIGN_UP(n, m)
+ALIGN_DOWN((n) + (m) - 1, (m))
+
 static const char *vhost_message_str[VHOST_USER_MAX] = {
 	[VHOST_USER_NONE] = "VHOST_USER_NONE",
 	[VHOST_USER_GET_FEATURES] = "VHOST_USER_GET_FEATURES", @@

-78,6

Quoted 2 lines

+91,8 @@ static const char *vhost_message_str[VHOST_USER_MAX] = {
[VHOST_USER_POSTCOPY_ADVISE] =

"VHOST_USER_POSTCOPY_ADVISE",

Quoted 1 line

[VHOST_USER_POSTCOPY_LISTEN] =

"VHOST_USER_POSTCOPY_LISTEN",

Quoted 10 lines

lib/librte_vhost/vhost_user.c +2 -1

 	[VHOST_USER_POSTCOPY_END]  = "VHOST_USER_POSTCOPY_END",
+	[VHOST_USER_GET_INFLIGHT_FD] = "VHOST_USER_GET_INFLIGHT_FD",
+	[VHOST_USER_SET_INFLIGHT_FD] = "VHOST_USER_SET_INFLIGHT_FD",
 };
 
 static int send_vhost_reply(int sockfd, struct VhostUserMsg *msg); @@
-160,6 +175,21 @@ vhost_backend_cleanup(struct virtio_net *dev)
 		dev->log_addr = 0;
 	}

Quoted 4 lines

lib/librte_vhost/vhost_user.c +4 -0

+	if (dev->inflight_info->addr) {
+		munmap(dev->inflight_info->addr, dev->inflight_info->size);
+		dev->inflight_info->addr = NULL;
+	}

lib/librte_vhost/vhost_user.c vhost_backend_cleanup(struct virtio_net *dev) · line 191

Quoted 10 lines

lib/librte_vhost/vhost_user.c +10 -0

+
+	if (dev->inflight_info->fd > 0) {
+		close(dev->inflight_info->fd);
+		dev->inflight_info->fd = -1;
+	}
+
+	if (dev->inflight_info) {
+		free(dev->inflight_info);
+		dev->inflight_info = NULL;
+	}

You should check dev->inflight_info first before dereferencing it.

Quoted 9 lines

lib/librte_vhost/vhost_user.c +1 -0

+
 	if (dev->slave_req_fd >= 0) {
 		close(dev->slave_req_fd);
 		dev->slave_req_fd = -1;

@@ -1165,6 +1195,233 @@ virtio_is_ready(struct virtio_net *dev)
 	return 1;
 }
 

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev) · line 1206

Quoted 8 lines

lib/librte_vhost/vhost_user.c +8 -0

+static int
+mem_create(const char *name, unsigned int flags) { #ifdef
+MEMFD_SUPPORTED
+	return memfd_create(name, flags);
+#else
+	RTE_LOG(ERR, VHOST_CONFIG,
+		"doesn't support memfd--name:%s and flag:%x\n",
+		name, flags);

You probably don't want to always print an error when memfd isn't available.

Quoted 78 lines

lib/librte_vhost/vhost_user.c +77 -0

+	return -1;
+#endif
+}
+
+static void *
+inflight_mem_alloc(const char *name, size_t size, int *fd) {
+	void *ptr;
+	int mfd = -1;
+	char fname[20] = "/tmp/memfd-XXXXXX";
+
+	*fd = -1;
+	mfd = mem_create(name, CLOEXEC);
+	if (mfd != -1) {
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	} else {
+		mfd = mkstemp(fname);
+		unlink(fname);
+
+		if (mfd == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"mkstemp fail for alloc inflight buffer\n");
+			return NULL;
+		}
+
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	}
+
+	ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
+	if (ptr == MAP_FAILED) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"mmap fail for alloc inflight buffer\n");
+		close(mfd);
+		return NULL;
+	}
+
+	*fd = mfd;
+	return ptr;
+}
+
+static uint32_t
+get_pervq_shm_size_split(uint16_t queue_size) {
+	return ALIGN_UP(sizeof(struct rte_vhost_inflight_desc_split) *
+			queue_size + sizeof(uint64_t) + sizeof(uint16_t) * 4,
+			INFLIGHT_ALIGNMENT);
+}
+
+static uint32_t
+get_pervq_shm_size_packed(uint16_t queue_size) {
+	return ALIGN_UP(sizeof(struct rte_vhost_inflight_desc_packed) *
+			queue_size + sizeof(uint64_t) + sizeof(uint16_t) * 6 +
+			sizeof(uint8_t) * 9, INFLIGHT_ALIGNMENT); }
+
+static int
+vhost_user_get_inflight_fd(struct virtio_net **pdev,
+				     VhostUserMsg *msg,
+				     int main_fd __rte_unused)
+{
+	int fd, i, j;
+	uint64_t pervq_inflight_size, mmap_size;
+	void *addr;
+	uint16_t num_queues, queue_size;
+	struct virtio_net *dev = *pdev;
+	struct rte_vhost_inflight_info_packed *inflight_packed = NULL;
+
+	if (msg->size != sizeof(msg->payload.inflight)) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Invalid get_inflight_fd message size is %d",

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev) · line 1292

Quoted 5 lines

lib/librte_vhost/vhost_user.c +5 -0

+			msg->size);
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	dev->inflight_info = calloc(1, sizeof(struct inflight_mem_info));

You need to check whether dev->inflight_info has been allocated, otherwise memory may leak.

Quoted 22 lines

lib/librte_vhost/vhost_user.c +21 -0

+	if (!dev->inflight_info) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Failed to alloc dev inflight area");
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	num_queues = msg->payload.inflight.num_queues;
+	queue_size = msg->payload.inflight.queue_size;
+
+	RTE_LOG(INFO, VHOST_CONFIG, "get_inflight_fd num_queues: %u\n",
+		msg->payload.inflight.num_queues);
+	RTE_LOG(INFO, VHOST_CONFIG, "get_inflight_fd queue_size: %u\n",
+		msg->payload.inflight.queue_size);
+
+	if (vq_is_packed(dev))
+		pervq_inflight_size = get_pervq_shm_size_packed(queue_size);
+	else
+		pervq_inflight_size = get_pervq_shm_size_split(queue_size);
+
+	mmap_size = num_queues * pervq_inflight_size;
+	addr = inflight_mem_alloc("vhost-inflight", mmap_size, &fd);

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev)

Quoted 10 lines

lib/librte_vhost/vhost_user.c +10 -0

+	if (!addr) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Failed to alloc vhost inflight area");
+			msg->payload.inflight.mmap_size = 0;
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+	memset(addr, 0, mmap_size);
+
+	dev->inflight_info->addr = addr;
+	dev->inflight_info->size = msg->payload.inflight.mmap_size =

mmap_size;

Quoted 77 lines

lib/librte_vhost/vhost_user.c +77 -0

+	dev->inflight_info->fd = msg->fds[0] = fd;
+	msg->payload.inflight.mmap_offset = 0;
+	msg->fd_num = 1;
+
+	if (vq_is_packed(dev)) {
+		for (i = 0; i < num_queues; i++) {
+			inflight_packed =
+				(struct rte_vhost_inflight_info_packed *)addr;
+			inflight_packed->used_wrap_counter = 1;
+			inflight_packed->old_used_wrap_counter = 1;
+			for (j = 0; j < queue_size; j++)
+				inflight_packed->desc[j].next = j + 1;
+			addr = (void *)((char *)addr + pervq_inflight_size);
+		}
+	}
+
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight mmap_size: %"PRIu64"\n",
+		msg->payload.inflight.mmap_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight mmap_offset: %"PRIu64"\n",
+		msg->payload.inflight.mmap_offset);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"send inflight fd: %d\n", msg->fds[0]);
+
+	return RTE_VHOST_MSG_RESULT_REPLY;
+}
+
+static int
+vhost_user_set_inflight_fd(struct virtio_net **pdev, VhostUserMsg *msg,
+		int main_fd __rte_unused)
+{
+	int fd;
+	uint64_t mmap_size, mmap_offset;
+	uint16_t num_queues, queue_size;
+	uint32_t pervq_inflight_size;
+	void *addr;
+	struct virtio_net *dev = *pdev;
+
+	fd = msg->fds[0];
+	if (msg->size != sizeof(msg->payload.inflight) || fd < 0) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"Invalid set_inflight_fd message size is %d,fd is %d\n",
+			msg->size, fd);
+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	mmap_size = msg->payload.inflight.mmap_size;
+	mmap_offset = msg->payload.inflight.mmap_offset;
+	num_queues = msg->payload.inflight.num_queues;
+	queue_size = msg->payload.inflight.queue_size;
+
+	if (vq_is_packed(dev))
+		pervq_inflight_size = get_pervq_shm_size_packed(queue_size);
+	else
+		pervq_inflight_size = get_pervq_shm_size_split(queue_size);
+
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd mmap_size: %"PRIu64"\n", mmap_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd mmap_offset: %"PRIu64"\n", mmap_offset);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd num_queues: %u\n", num_queues);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd queue_size: %u\n", queue_size);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd fd: %d\n", fd);
+	RTE_LOG(INFO, VHOST_CONFIG,
+		"set_inflight_fd pervq_inflight_size: %d\n",
+		pervq_inflight_size);
+
+	if (!dev->inflight_info) {
+		dev->inflight_info = calloc(1,
+					    sizeof(struct inflight_mem_info));
+		if (dev->inflight_info == NULL) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"Failed to alloc dev inflight area");

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev) · line 1406

Quoted 6 lines

lib/librte_vhost/vhost_user.c +6 -0

+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+	}
+
+	if (dev->inflight_info->addr)
+		munmap(dev->inflight_info->addr, dev->inflight_info->size);

dev->inflight_info->addr won't be zeroed when below mmap() fails.

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev)

Quoted 6 lines

lib/librte_vhost/vhost_user.c +5 -0

+
+	addr = mmap(0, mmap_size, PROT_READ | PROT_WRITE, MAP_SHARED,
+			fd, mmap_offset);
+	if (addr == MAP_FAILED) {
+		RTE_LOG(ERR, VHOST_CONFIG, "failed to mmap share

memory.\n");

Quoted 5 lines

lib/librte_vhost/vhost_user.c +5 -0

+		return RTE_VHOST_MSG_RESULT_ERR;
+	}
+
+	if (dev->inflight_info->fd)
+		close(dev->inflight_info->fd);

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev)

Quoted 10 lines

lib/librte_vhost/vhost_user.c +8 -0

+
+	dev->inflight_info->fd = fd;
+	dev->inflight_info->addr = addr;
+	dev->inflight_info->size = mmap_size;
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
 static int
 vhost_user_set_vring_call(struct virtio_net **pdev, struct VhostUserMsg

*msg,

lib/librte_vhost/vhost_user.c static vhost_message_handler_t

Quoted 2 lines

lib/librte_vhost/vhost_user.c +0 -0

 			int main_fd __rte_unused)

@@ -1762,9 +2019,10 @@ static vhost_message_handler_t

vhost_message_handlers[VHOST_USER_MAX] = {

Quoted 28 lines

lib/librte_vhost/vhost_user.c +2 -1

 	[VHOST_USER_POSTCOPY_ADVISE] = vhost_user_set_postcopy_advise,
 	[VHOST_USER_POSTCOPY_LISTEN] = vhost_user_set_postcopy_listen,
 	[VHOST_USER_POSTCOPY_END] = vhost_user_postcopy_end,
+	[VHOST_USER_GET_INFLIGHT_FD] = vhost_user_get_inflight_fd,
+	[VHOST_USER_SET_INFLIGHT_FD] = vhost_user_set_inflight_fd,
 };
 
-
 /* return bytes# of read on success or negative val on failure. */
 static int  read_vhost_message(int sockfd, struct VhostUserMsg *msg)

(quoted diff) +0 -0

b/lib/librte_vhost/vhost_user.h index 17a1d7bca..6563f7315 100644

lib/librte_vhost/vhost_user.h +3 -2

@@ -54,7 +54,9 @@ typedef enum VhostUserRequest {
 	VHOST_USER_POSTCOPY_ADVISE = 28,
 	VHOST_USER_POSTCOPY_LISTEN = 29,
 	VHOST_USER_POSTCOPY_END = 30,
-	VHOST_USER_MAX = 31
+	VHOST_USER_GET_INFLIGHT_FD = 31,
+	VHOST_USER_SET_INFLIGHT_FD = 32,
+	VHOST_USER_MAX = 33
 } VhostUserRequest;
 
 typedef enum VhostUserSlaveRequest {
--

2.17.2

Quoted 11 lines

lib/librte_vhost/vhost_user.c virtio_is_ready(struct virtio_net *dev) · line 1206

Quoted 9 lines

lib/librte_vhost/vhost_user.c +8 -0

+static int
+mem_create(const char *name, unsigned int flags) { #ifdef
+MEMFD_SUPPORTED
+	return memfd_create(name, flags);
+#else
+	RTE_LOG(ERR, VHOST_CONFIG,
+		"doesn't support memfd--name:%s and flag:%x\n",
+		name, flags);

You probably don't want to always print an error when memfd isn't available.

I saw below error message from ci/Intel-compilation. ../lib/librte_vhost/vhost_user.c:1202:24: error: unused parameter ‘name’ [-Werror=unused-parameter]
mem_create(const char *name, unsigned int flags)

                        ^~~~

../lib/librte_vhost/vhost_user.c:1202:43: error: unused parameter ‘flags’ [-Werror=unused-parameter]
mem_create(const char *name, unsigned int flags)

That's why I add this print.Should I ignore this?

Thanks.

Quoted 49 lines

Quoted 49 lines

lib/librte_vhost/vhost_user.c +48 -0

+	return -1;
+#endif
+}
+
+static void *
+inflight_mem_alloc(const char *name, size_t size, int *fd) {
+	void *ptr;
+	int mfd = -1;
+	char fname[20] = "/tmp/memfd-XXXXXX";
+
+	*fd = -1;
+	mfd = mem_create(name, CLOEXEC);
+	if (mfd != -1) {
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	} else {
+		mfd = mkstemp(fname);
+		unlink(fname);
+
+		if (mfd == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"mkstemp fail for alloc inflight buffer\n");
+			return NULL;
+		}
+
+		if (ftruncate(mfd, size) == -1) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"ftruncate fail for alloc inflight buffer\n");
+			close(mfd);
+			return NULL;
+		}
+	}
+
+	ptr = mmap(0, size, PROT_READ | PROT_WRITE, MAP_SHARED, mfd, 0);
+	if (ptr == MAP_FAILED) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"mmap fail for alloc inflight buffer\n");
+		close(mfd);
+		return NULL;
+	}
+
+	*fd = mfd;
+	return ptr;
+}

lib/librte_vhost/vhost_user.c +0 -0

@@ -1458,6 +1472,188 @@ static int vhost_user_set_vring_err(struct virtio_net **pdev __rte_unused,
 	return RTE_VHOST_MSG_RESULT_OK;
 }
 

lib/librte_vhost/vhost_user.c +19 -0

+		return 1;
+
+	return -1;
+}
+
+static int
+vhost_check_queue_inflights_split(struct virtio_net *dev,
+	struct vhost_virtqueue *vq)
+{
+	uint16_t i = 0;
+	uint16_t resubmit_num = 0, last_io, num;
+	struct vring_used *used = vq->used;
+	struct rte_vhost_resubmit_info *resubmit = NULL;
+	struct rte_vhost_inflight_info_split *inflight_split;
+
+	if (!(dev->protocol_features &
+		(1ULL << VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD)))
+		return RTE_VHOST_MSG_RESULT_OK;
+

lib/librte_vhost/vhost_user.c +68 -0

+	vq->global_counter = 0;
+	last_io = inflight_split->last_inflight_io;
+
+	if (inflight_split->used_idx != used->idx) {
+		inflight_split->desc[last_io].inflight = 0;
+		rte_compiler_barrier();
+		inflight_split->used_idx = used->idx;
+	}
+
+	for (i = 0; i < inflight_split->desc_num; i++) {
+		if (inflight_split->desc[i].inflight == 1)
+			resubmit_num++;
+	}
+
+	vq->last_avail_idx += resubmit_num;
+
+	if (resubmit_num) {
+		resubmit  = calloc(1, sizeof(struct rte_vhost_resubmit_info));
+		if (!resubmit) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit info.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		resubmit->resubmit_list = calloc(resubmit_num,
+			sizeof(struct rte_vhost_resubmit_desc));
+		if (!resubmit->resubmit_list) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for inflight desc.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		num = 0;
+		for (i = 0; i < vq->inflight_split->desc_num; i++) {
+			if (vq->inflight_split->desc[i].inflight == 1) {
+				resubmit->resubmit_list[num].index = i;
+				resubmit->resubmit_list[num].counter =
+					inflight_split->desc[i].counter;
+				num++;
+			}
+		}
+		resubmit->resubmit_num = num;
+
+		if (resubmit->resubmit_num > 1)
+			qsort(resubmit->resubmit_list, resubmit->resubmit_num,
+				sizeof(struct rte_vhost_resubmit_desc),
+				resubmit_desc_compare);
+
+		vq->global_counter = resubmit->resubmit_list[0].counter + 1;
+		vq->resubmit_inflight = resubmit;
+	}
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
+static int
+vhost_check_queue_inflights_packed(struct virtio_net *dev,
+						 struct vhost_virtqueue *vq)
+{
+	uint16_t i = 0;
+	uint16_t resubmit_num = 0, old_used_idx, num;
+	struct rte_vhost_resubmit_info *resubmit = NULL;
+	struct rte_vhost_inflight_info_packed *inflight_packed;
+
+	if (!(dev->protocol_features &
+		(1ULL << VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD)))
+		return RTE_VHOST_MSG_RESULT_OK;
+

lib/librte_vhost/vhost_user.c +79 -0

+	vq->global_counter = 0;
+	old_used_idx = inflight_packed->old_used_idx;
+
+	if (inflight_packed->used_idx != old_used_idx) {
+		if (inflight_packed->desc[old_used_idx].inflight == 0) {
+			inflight_packed->old_used_idx =
+				inflight_packed->used_idx;
+			inflight_packed->old_used_wrap_counter =
+				inflight_packed->used_wrap_counter;
+			inflight_packed->old_free_head =
+				inflight_packed->free_head;
+		} else {
+			inflight_packed->used_idx =
+				inflight_packed->old_used_idx;
+			inflight_packed->used_wrap_counter =
+				inflight_packed->old_used_wrap_counter;
+			inflight_packed->free_head =
+				inflight_packed->old_free_head;
+		}
+	}
+
+	for (i = 0; i < inflight_packed->desc_num; i++) {
+		if (inflight_packed->desc[i].inflight == 1)
+			resubmit_num++;
+	}
+
+	if (resubmit_num) {
+		resubmit = calloc(1, sizeof(struct rte_vhost_resubmit_info));
+		if (resubmit == NULL) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit info.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		resubmit->resubmit_list = calloc(resubmit_num,
+			sizeof(struct rte_vhost_resubmit_desc));
+		if (resubmit->resubmit_list == NULL) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit desc.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		num = 0;
+		for (i = 0; i < inflight_packed->desc_num; i++) {
+			if (vq->inflight_packed->desc[i].inflight == 1) {
+				resubmit->resubmit_list[num].index = i;
+				resubmit->resubmit_list[num].counter =
+					inflight_packed->desc[i].counter;
+				num++;
+			}
+		}
+		resubmit->resubmit_num = num;
+
+		if (resubmit->resubmit_num > 1)
+			qsort(resubmit->resubmit_list, resubmit->resubmit_num,
+				sizeof(struct rte_vhost_resubmit_desc),
+				resubmit_desc_compare);
+
+		vq->global_counter = resubmit->resubmit_list[0].counter + 1;
+		vq->resubmit_inflight = resubmit;
+	}
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
 static int
 vhost_user_set_vring_kick(struct virtio_net **pdev, struct VhostUserMsg *msg,
 			int main_fd __rte_unused)

@@ -1499,6 +1695,20 @@ vhost_user_set_vring_kick(struct virtio_net **pdev, struct VhostUserMsg *msg,
 		close(vq->kickfd);
 	vq->kickfd = file.fd;
 
+	if (vq_is_packed(dev)) {
+		if (vhost_check_queue_inflights_packed(dev, vq)) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"Failed to inflights for vq: %d\n", file.index);
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+	} else {
+		if (vhost_check_queue_inflights_split(dev, vq)) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"Failed to inflights for vq: %d\n", file.index);
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+	}
+
 	return RTE_VHOST_MSG_RESULT_OK;
 }
 

Signature · 1 line

2.17.2

-----Original Message-----
From: Bie, Tiwei
Sent: Thursday, September 26, 2019 4:05 PM
To: Yu, Jin <[email protected]>
Cc: [email protected]; Liu, Changpeng <[email protected]>; [email protected]; Wang, Zhihong <[email protected]>; Lin Li <[email protected]>; Xun Ni <[email protected]>; Yu Zhang <[email protected]>
Subject: Re: [PATCH v7 05/10] vhost: checkout the resubmit inflight information

Review static int vhost_user_set_vring_err(struct

Quoted 1 line · Jin Yu

(quoted diff) +0 -0

@@ -1458,6 +1472,188 @@ static int vhost_user_set_vring_err(struct

virtio_net **pdev __rte_unused,

Quoted 3 lines

lib/librte_vhost/vhost_user.c +0 -0

 	return RTE_VHOST_MSG_RESULT_OK;
 }
 

lib/librte_vhost/vhost_user.c static int vhost_user_set_vring_err(struct virtio_net **pdev __rte_unused, · line 1484

Quoted 9 lines

lib/librte_vhost/vhost_user.c +9 -0

+static int
+resubmit_desc_compare(const void *a, const void *b) {
+	const struct rte_vhost_resubmit_desc *desc0 =
+		(const struct rte_vhost_resubmit_desc *)a;
+	const struct rte_vhost_resubmit_desc *desc1 =
+		(const struct rte_vhost_resubmit_desc *)b;
+
+	if (desc1->counter > desc0->counter &&
+		(desc1->counter - desc0->counter) < VIRTQUEUE_MAX_SIZE * 2)

Why is this hardcoded to 1024 * 2 here?

Quoted 25 lines

lib/librte_vhost/vhost_user.c +24 -0

+		return 1;
+
+	return -1;
+}
+
+static int
+vhost_check_queue_inflights_split(struct virtio_net *dev,
+	struct vhost_virtqueue *vq)
+{
+	uint16_t i = 0;
+	uint16_t resubmit_num = 0, last_io, num;
+	struct vring_used *used = vq->used;
+	struct rte_vhost_resubmit_info *resubmit = NULL;
+	struct rte_vhost_inflight_info_split *inflight_split;
+
+	if (!(dev->protocol_features &
+		(1ULL << VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD)))
+		return RTE_VHOST_MSG_RESULT_OK;
+
+	if ((!vq->inflight_split))
+		return RTE_VHOST_MSG_RESULT_ERR;
+
+	if (!vq->inflight_split->version) {
+		vq->inflight_split->version = INFLIGHT_VERSION;

lib/librte_vhost/vhost_user.c static int vhost_user_set_vring_err(struct virtio_net **pdev __rte_unused, · line 1513

Quoted 5 lines

lib/librte_vhost/vhost_user.c +5 -0

+		return RTE_VHOST_MSG_RESULT_OK;
+	}
+
+	inflight_split = vq->inflight_split;
+	vq->resubmit_inflight = NULL;

This is invoked in vhost message handler, you need to check whether it has been allocated first.

Quoted 71 lines

lib/librte_vhost/vhost_user.c +68 -0

+	vq->global_counter = 0;
+	last_io = inflight_split->last_inflight_io;
+
+	if (inflight_split->used_idx != used->idx) {
+		inflight_split->desc[last_io].inflight = 0;
+		rte_compiler_barrier();
+		inflight_split->used_idx = used->idx;
+	}
+
+	for (i = 0; i < inflight_split->desc_num; i++) {
+		if (inflight_split->desc[i].inflight == 1)
+			resubmit_num++;
+	}
+
+	vq->last_avail_idx += resubmit_num;
+
+	if (resubmit_num) {
+		resubmit  = calloc(1, sizeof(struct rte_vhost_resubmit_info));
+		if (!resubmit) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit info.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		resubmit->resubmit_list = calloc(resubmit_num,
+			sizeof(struct rte_vhost_resubmit_desc));
+		if (!resubmit->resubmit_list) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for inflight desc.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		num = 0;
+		for (i = 0; i < vq->inflight_split->desc_num; i++) {
+			if (vq->inflight_split->desc[i].inflight == 1) {
+				resubmit->resubmit_list[num].index = i;
+				resubmit->resubmit_list[num].counter =
+					inflight_split->desc[i].counter;
+				num++;
+			}
+		}
+		resubmit->resubmit_num = num;
+
+		if (resubmit->resubmit_num > 1)
+			qsort(resubmit->resubmit_list, resubmit-
 resubmit_num,
+				sizeof(struct rte_vhost_resubmit_desc),
+				resubmit_desc_compare);
+
+		vq->global_counter = resubmit->resubmit_list[0].counter + 1;
+		vq->resubmit_inflight = resubmit;
+	}
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
+static int
+vhost_check_queue_inflights_packed(struct virtio_net *dev,
+						 struct vhost_virtqueue *vq)
+{
+	uint16_t i = 0;
+	uint16_t resubmit_num = 0, old_used_idx, num;
+	struct rte_vhost_resubmit_info *resubmit = NULL;
+	struct rte_vhost_inflight_info_packed *inflight_packed;
+
+	if (!(dev->protocol_features &
+		(1ULL << VHOST_USER_PROTOCOL_F_INFLIGHT_SHMFD)))
+		return RTE_VHOST_MSG_RESULT_OK;
+

lib/librte_vhost/vhost_user.c static int vhost_user_set_vring_err(struct virtio_net **pdev __rte_unused, · line 1591

Quoted 10 lines

lib/librte_vhost/vhost_user.c +10 -0

+	if ((!vq->inflight_packed))
+		return RTE_VHOST_MSG_RESULT_ERR;
+
+	if (!vq->inflight_packed->version) {
+		vq->inflight_packed->version = INFLIGHT_VERSION;
+		return RTE_VHOST_MSG_RESULT_OK;
+	}
+
+	inflight_packed = vq->inflight_packed;
+	vq->resubmit_inflight = NULL;

Ditto.

Quoted 58 lines

lib/librte_vhost/vhost_user.c +57 -0

+	vq->global_counter = 0;
+	old_used_idx = inflight_packed->old_used_idx;
+
+	if (inflight_packed->used_idx != old_used_idx) {
+		if (inflight_packed->desc[old_used_idx].inflight == 0) {
+			inflight_packed->old_used_idx =
+				inflight_packed->used_idx;
+			inflight_packed->old_used_wrap_counter =
+				inflight_packed->used_wrap_counter;
+			inflight_packed->old_free_head =
+				inflight_packed->free_head;
+		} else {
+			inflight_packed->used_idx =
+				inflight_packed->old_used_idx;
+			inflight_packed->used_wrap_counter =
+				inflight_packed->old_used_wrap_counter;
+			inflight_packed->free_head =
+				inflight_packed->old_free_head;
+		}
+	}
+
+	for (i = 0; i < inflight_packed->desc_num; i++) {
+		if (inflight_packed->desc[i].inflight == 1)
+			resubmit_num++;
+	}
+
+	if (resubmit_num) {
+		resubmit = calloc(1, sizeof(struct rte_vhost_resubmit_info));
+		if (resubmit == NULL) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit info.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		resubmit->resubmit_list = calloc(resubmit_num,
+			sizeof(struct rte_vhost_resubmit_desc));
+		if (resubmit->resubmit_list == NULL) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+			 "Failed to allocate memory for resubmit desc.\n");
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+
+		num = 0;
+		for (i = 0; i < inflight_packed->desc_num; i++) {
+			if (vq->inflight_packed->desc[i].inflight == 1) {
+				resubmit->resubmit_list[num].index = i;
+				resubmit->resubmit_list[num].counter =
+					inflight_packed->desc[i].counter;
+				num++;
+			}
+		}
+		resubmit->resubmit_num = num;
+
+		if (resubmit->resubmit_num > 1)
+			qsort(resubmit->resubmit_list, resubmit-
 resubmit_num,
+				sizeof(struct rte_vhost_resubmit_desc),
+				resubmit_desc_compare);

lib/librte_vhost/vhost_user.c static int vhost_user_set_vring_err(struct virtio_net **pdev __rte_unused,

Quoted 10 lines

lib/librte_vhost/vhost_user.c +8 -0

+
+		vq->global_counter = resubmit->resubmit_list[0].counter + 1;
+		vq->resubmit_inflight = resubmit;
+	}
+
+	return RTE_VHOST_MSG_RESULT_OK;
+}
+
 static int
 vhost_user_set_vring_kick(struct virtio_net **pdev, struct VhostUserMsg

*msg,

Review vhost_user_set_vring_kick(struct virtio_net

Quoted 2 lines

(quoted diff) +0 -0

 			int main_fd __rte_unused)

@@ -1499,6 +1695,20 @@ vhost_user_set_vring_kick(struct virtio_net

**pdev, struct VhostUserMsg *msg,

Quoted 23 lines

lib/librte_vhost/vhost_user.c +14 -1

 		close(vq->kickfd);
 	vq->kickfd = file.fd;
 
+	if (vq_is_packed(dev)) {
+		if (vhost_check_queue_inflights_packed(dev, vq)) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"Failed to inflights for vq: %d\n", file.index);
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+	} else {
+		if (vhost_check_queue_inflights_split(dev, vq)) {
+			RTE_LOG(ERR, VHOST_CONFIG,
+				"Failed to inflights for vq: %d\n", file.index);
+			return RTE_VHOST_MSG_RESULT_ERR;
+		}
+	}
+
 	return RTE_VHOST_MSG_RESULT_OK;
 }
 
--

2.17.2

A vhost-user-blk example that support inflight feature. It uses the new APIs that introduced in the first patch, so It can show how there

APIs work to support inflight feature.

Signed-off-by Jin Yu <[email protected]>

---
V1 - add the case.
V2 - add the rte_vhost prefix.
V3 - add packed ring support

---
 examples/vhost_blk/Makefile           |   67 ++
 examples/vhost_blk/blk.c              |  125 +++
 examples/vhost_blk/blk_spec.h         |   95 ++
 examples/vhost_blk/meson.build        |   20 +
 examples/vhost_blk/vhost_blk.c        | 1313 +++++++++++++++++++++++++
 examples/vhost_blk/vhost_blk.h        |  116 +++
 examples/vhost_blk/vhost_blk_compat.c |  195 ++++
 7 files changed, 1931 insertions(+)

-----Original Message-----
From: Bie, Tiwei
Sent: Wednesday, September 25, 2019 10:46 PM
To: Yu, Jin <[email protected]>
Cc: [email protected]; Liu, Changpeng <[email protected]>; [email protected]; Wang, Zhihong <[email protected]>
Subject: Re: [PATCH v7 10/10] vhost: add vhost-user-blk example which support inflight

Quoted 2 lines · Jin Yu

A vhost-user-blk example that support inflight feature. It uses the new APIs that introduced in the first patch, so It can show how there

s/It/it/
s/there/these/

Quoted 8 lines

APIs work to support inflight feature.

Signed-off-by Jin Yu <[email protected]>

---
V1 - add the case.
V2 - add the rte_vhost prefix.
V3 - add packed ring support

Quoted 9 lines

---
 examples/vhost_blk/Makefile           |   67 ++
 examples/vhost_blk/blk.c              |  125 +++
 examples/vhost_blk/blk_spec.h         |   95 ++
 examples/vhost_blk/meson.build        |   20 +
 examples/vhost_blk/vhost_blk.c        | 1313 +++++++++++++++++++++++++
 examples/vhost_blk/vhost_blk.h        |  116 +++
 examples/vhost_blk/vhost_blk_compat.c |  195 ++++
 7 files changed, 1931 insertions(+)

I met some build issues when trying this example.

examples/vhost_blk/vhost_blk.c: In function ‘descriptor_get_next_packed’:
examples/vhost_blk/vhost_blk.c:71:21: error: invalid use of undefined type ‘struct vring_packed_desc’

  if (vq->desc_packed[*idx % vq->size].flags & VIRTQ_DESC_F_NEXT) {
                     ^

examples/vhost_blk/vhost_blk.c:71:21: error: dereferencing pointer to incomplete type ‘struct vring_packed_desc’ examples/vhost_blk/vhost_blk.c:73:26: error: invalid use of undefined type ‘struct vring_packed_desc’
return &vq->desc_packed[*idx % vq->size];

                          ^

examples/vhost_blk/vhost_blk.c: In function ‘inflight_submit_completion_packed’:
examples/vhost_blk/vhost_blk.c:165:2: warning:
‘rte_vhost_set_last_inflight_io_packed’ is deprecated: Symbol is not yet part of stable ABI [-Wdeprecated-declarations]

ret = rte_vhost_set_last_inflight_io_packed(ctrlr->bdev->vid, q_idx, ^~~
In file included from examples/vhost_blk/vhost_blk.c:17:0:
x86_64-native-linuxapp-gcc/include/rte_vhost.h:810:1: note: declared here rte_vhost_set_last_inflight_io_packed(int vid, ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

                         ^

...

Quoted 9 lines

-----Original Message-----
From: Bie, Tiwei
Sent: Wednesday, September 25, 2019 10:46 PM
To: Yu, Jin <[email protected]>
Cc: [email protected]; Liu, Changpeng <[email protected]>; [email protected]; Wang, Zhihong <[email protected]>
Subject: Re: [PATCH v7 10/10] vhost: add vhost-user-blk example which support inflight

Quoted 6 lines

Quoted 2 lines · Jin Yu

A vhost-user-blk example that support inflight feature. It uses the new APIs that introduced in the first patch, so It can show how there

s/It/it/
s/there/these/

Got it. Thanks.

Quoted 23 lines

Quoted 8 lines

APIs work to support inflight feature.

Signed-off-by Jin Yu <[email protected]>

---
V1 - add the case.
V2 - add the rte_vhost prefix.
V3 - add packed ring support

Quoted 9 lines

---
 examples/vhost_blk/Makefile           |   67 ++
 examples/vhost_blk/blk.c              |  125 +++
 examples/vhost_blk/blk_spec.h         |   95 ++
 examples/vhost_blk/meson.build        |   20 +
 examples/vhost_blk/vhost_blk.c        | 1313 +++++++++++++++++++++++++
 examples/vhost_blk/vhost_blk.h        |  116 +++
 examples/vhost_blk/vhost_blk_compat.c |  195 ++++
 7 files changed, 1931 insertions(+)

I met some build issues when trying this example.

examples/vhost_blk/vhost_blk.c: In function ‘descriptor_get_next_packed’:
examples/vhost_blk/vhost_blk.c:71:21: error: invalid use of undefined type ‘struct vring_packed_desc’

Quoted 7 lines

  if (vq->desc_packed[*idx % vq->size].flags & VIRTQ_DESC_F_NEXT) {
                     ^

examples/vhost_blk/vhost_blk.c:71:21: error: dereferencing pointer to incomplete type ‘struct vring_packed_desc’ examples/vhost_blk/vhost_blk.c:73:26: error: invalid use of undefined type ‘struct vring_packed_desc’
return &vq->desc_packed[*idx % vq->size];

The new version of virtio_ring.h have defined the vring_packed_desc. For the compatibility， I will add the definition in the rte_vhost.h.

Thanks.

Quoted 6 lines

                          ^

examples/vhost_blk/vhost_blk.c: In function ‘inflight_submit_completion_packed’:
examples/vhost_blk/vhost_blk.c:165:2: warning:
‘rte_vhost_set_last_inflight_io_packed’ is deprecated: Symbol is not yet part of stable ABI [-Wdeprecated-declarations]

Should I ignore this warning？ I'm not sure about this. The reason is the _rte_experiment keyword?

Thanks.

Quoted 8 lines

ret = rte_vhost_set_last_inflight_io_packed(ctrlr->bdev->vid, q_idx, ^~~
In file included from examples/vhost_blk/vhost_blk.c:17:0:
x86_64-native-linuxapp-gcc/include/rte_vhost.h:810:1: note: declared here rte_vhost_set_last_inflight_io_packed(int vid, ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

                         ^

...

[PATCH v7 00/10] vhost: support inflight share memory protocol feature

This patches introduces two new messages VHOST_USER_GET_INFLIGHT_FD and VHOST_USER_SET_INFLIGHT_FD to support transferring a shared buffer between qemu and backend.
Now It can both support split and packed ring. The example code show how these API work. The test has passed.

How to test the example:
1, Qemu need two patches.
https://patchwork.kernel.org/patch/10766813/
https://patchwork.kernel.org/patch/10861411/(QEMU merged) it also needs some manual modifications:
we should confirm that before we send get inflight we have already sent the set features but it seems Qemu didn't do like this. So we manually revise this, we can add below code in vhost_dev_get_inflight

    int r;

    r = vhost_dev_set_features(dev, dev->log_enabled);
    if (r < 0) {
        return -errno;
    }

before get_inflight_fd.
2, Guest OS version >= 5.0
3, run the example
4, run the qemu with vhost-user-blk-pci.
eg:

        -chardev socket,id=spdk_vhost_blk0,reconnect=1,path=xxxx\
        -device vhost-user-blk-pci,ring_packed=on,chardev=spdk_vhost_blk0,num-queues=1\

5, run fio in the guest
6, kill the example and run again.
7, the fio in the guest should continue run without errors.

Jin Yu (10):
vhost: add the inflight description
vhost: add packed ring
vhost: add the inflight structure
vhost: add two new messages to support a shared buffer
vhost: checkout the resubmit inflight information
vhost: add the APIs to operate inflight ring
vhost: add APIs for user getting inflight ring
vhost: add vring functions packed ring support
vhost: add an API for judging vq format
vhost: add vhost-user-blk example which support inflight

 examples/vhost_blk/Makefile            |   67 ++
 examples/vhost_blk/blk.c               |  125 +++
 examples/vhost_blk/blk_spec.h          |   95 ++
 examples/vhost_blk/meson.build         |   20 +
 examples/vhost_blk/vhost_blk.c         | 1313 ++++++++++++++++++++++++
 examples/vhost_blk/vhost_blk.h         |  116 +++
 examples/vhost_blk/vhost_blk_compat.c  |  195 ++++
 lib/librte_vhost/rte_vhost.h           |  250 ++++-
 lib/librte_vhost/rte_vhost_version.map |    9 +
 lib/librte_vhost/vhost.c               |  416 +++++++-
 lib/librte_vhost/vhost.h               |   16 +
 lib/librte_vhost/vhost_user.c          |  470 ++++++++-
 lib/librte_vhost/vhost_user.h          |   12 +-
 13 files changed, 3091 insertions(+), 13 deletions(-)
 create mode 100644 examples/vhost_blk/Makefile
 create mode 100644 examples/vhost_blk/blk.c
 create mode 100644 examples/vhost_blk/blk_spec.h
 create mode 100644 examples/vhost_blk/meson.build
 create mode 100644 examples/vhost_blk/vhost_blk.c
 create mode 100644 examples/vhost_blk/vhost_blk.h
 create mode 100644 examples/vhost_blk/vhost_blk_compat.c

Signature · 1 line

2.17.2

-----Original Message-----
From: Bie, Tiwei
Sent: Wednesday, September 25, 2019 10:26 PM
To: Yu, Jin <[email protected]>
Cc: [email protected]; Liu, Changpeng <[email protected]>; [email protected]; Wang, Zhihong <[email protected]>
Subject: Re: [PATCH v7 00/10] vhost: support inflight share memory protocol feature

Quoted 1 line

[PATCH v7 00/10] vhost: support inflight share memory protocol feature

This should be v8. Please increase the version number when sending a new series.

Thanks,
Tiwei

Quoted 70 lines · Jin Yu

This patches introduces two new messages VHOST_USER_GET_INFLIGHT_FD and VHOST_USER_SET_INFLIGHT_FD to support transferring a shared buffer between qemu and backend.
Now It can both support split and packed ring. The example code show how these API work. The test has passed.

How to test the example:
1, Qemu need two patches.
https://patchwork.kernel.org/patch/10766813/
https://patchwork.kernel.org/patch/10861411/(QEMU merged) it also needs some manual modifications:
we should confirm that before we send get inflight we have already sent the set features but it seems Qemu didn't do like this. So we manually revise this, we can add below code in vhost_dev_get_inflight

    int r;

    r = vhost_dev_set_features(dev, dev->log_enabled);
    if (r < 0) {
        return -errno;
    }

before get_inflight_fd.
2, Guest OS version >= 5.0
3, run the example
4, run the qemu with vhost-user-blk-pci.
eg:

        -chardev socket,id=spdk_vhost_blk0,reconnect=1,path=xxxx\
        -device

vhost-user-blk-pci,ring_packed=on,chardev=spdk_vhost_blk0,num-queues=1 \
5, run fio in the guest
6, kill the example and run again.
7, the fio in the guest should continue run without errors.

Jin Yu (10):
vhost: add the inflight description
vhost: add packed ring
vhost: add the inflight structure
vhost: add two new messages to support a shared buffer
vhost: checkout the resubmit inflight information
vhost: add the APIs to operate inflight ring
vhost: add APIs for user getting inflight ring
vhost: add vring functions packed ring support
vhost: add an API for judging vq format
vhost: add vhost-user-blk example which support inflight

 examples/vhost_blk/Makefile            |   67 ++
 examples/vhost_blk/blk.c               |  125 +++
 examples/vhost_blk/blk_spec.h          |   95 ++
 examples/vhost_blk/meson.build         |   20 +
 examples/vhost_blk/vhost_blk.c         | 1313 ++++++++++++++++++++++++
 examples/vhost_blk/vhost_blk.h         |  116 +++
 examples/vhost_blk/vhost_blk_compat.c  |  195 ++++
 lib/librte_vhost/rte_vhost.h           |  250 ++++-
 lib/librte_vhost/rte_vhost_version.map |    9 +
 lib/librte_vhost/vhost.c               |  416 +++++++-
 lib/librte_vhost/vhost.h               |   16 +
 lib/librte_vhost/vhost_user.c          |  470 ++++++++-
 lib/librte_vhost/vhost_user.h          |   12 +-
 13 files changed, 3091 insertions(+), 13 deletions(-)  create mode
100644 examples/vhost_blk/Makefile  create mode 100644
examples/vhost_blk/blk.c  create mode 100644
examples/vhost_blk/blk_spec.h  create mode 100644
examples/vhost_blk/meson.build  create mode 100644
examples/vhost_blk/vhost_blk.c  create mode 100644
examples/vhost_blk/vhost_blk.h  create mode 100644

examples/vhost_blk/vhost_blk_compat.c

--
2.17.2