X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fvirtio%2Fvirtio_user%2Fvhost_kernel.c;h=58e66bb7b4ae01cde92a07215051ea9f0fb92c72;hb=55c1238246d53d196420fd2768dd3d4210ab654b;hp=b79dcad17949f1a2751e1200e3abe2fb31163f6c;hpb=06856cabb89b0db000a065c17e32cdf2d2f58470;p=dpdk.git

diff --git a/drivers/net/virtio/virtio_user/vhost_kernel.c b/drivers/net/virtio/virtio_user/vhost_kernel.c
index b79dcad179..58e66bb7b4 100644
--- a/drivers/net/virtio/virtio_user/vhost_kernel.c
+++ b/drivers/net/virtio/virtio_user/vhost_kernel.c
@@ -14,6 +14,11 @@
 #include "virtio_user_dev.h"
 #include "vhost_kernel_tap.h"
 
+struct vhost_kernel_data {
+	int *vhostfds;
+	int *tapfds;
+};
+
 struct vhost_memory_kernel {
 	uint32_t nregions;
 	uint32_t padding;
@@ -38,6 +43,28 @@ struct vhost_memory_kernel {
 #define VHOST_SET_VRING_ERR _IOW(VHOST_VIRTIO, 0x22, struct vhost_vring_file)
 #define VHOST_NET_SET_BACKEND _IOW(VHOST_VIRTIO, 0x30, struct vhost_vring_file)
 
+/* with below features, vhost kernel does not need to do the checksum and TSO,
+ * these info will be passed to virtio_user through virtio net header.
+ */
+#define VHOST_KERNEL_GUEST_OFFLOADS_MASK	\
+	((1ULL << VIRTIO_NET_F_GUEST_CSUM) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_TSO4) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_TSO6) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_ECN)  |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_UFO))
+
+/* with below features, when flows from virtio_user to vhost kernel
+ * (1) if flows goes up through the kernel networking stack, it does not need
+ * to verify checksum, which can save CPU cycles;
+ * (2) if flows goes through a Linux bridge and outside from an interface
+ * (kernel driver), checksum and TSO will be done by GSO in kernel or even
+ * offloaded into real physical device.
+ */
+#define VHOST_KERNEL_HOST_OFFLOADS_MASK		\
+	((1ULL << VIRTIO_NET_F_HOST_TSO4) |	\
+	 (1ULL << VIRTIO_NET_F_HOST_TSO6) |	\
+	 (1ULL << VIRTIO_NET_F_CSUM))
+
 static uint64_t max_regions = 64;
 
 static void
@@ -74,21 +101,73 @@ vhost_kernel_ioctl(int fd, uint64_t request, void *arg)
 static int
 vhost_kernel_set_owner(struct virtio_user_dev *dev)
 {
-	return vhost_kernel_ioctl(dev->vhostfds[0], VHOST_SET_OWNER, NULL);
+	int ret;
+	uint32_t i;
+	struct vhost_kernel_data *data = dev->backend_data;
+
+	for (i = 0; i < dev->max_queue_pairs; ++i) {
+		if (data->vhostfds[i] < 0)
+			continue;
+
+		ret = vhost_kernel_ioctl(data->vhostfds[i], VHOST_SET_OWNER, NULL);
+		if (ret < 0)
+			return ret;
+	}
+
+	return 0;
 }
 
-static uint64_t vhost_req_user_to_kernel[] = {
-	[VHOST_USER_RESET_OWNER] = VHOST_RESET_OWNER,
-	[VHOST_USER_SET_FEATURES] = VHOST_SET_FEATURES,
-	[VHOST_USER_GET_FEATURES] = VHOST_GET_FEATURES,
-	[VHOST_USER_SET_VRING_CALL] = VHOST_SET_VRING_CALL,
-	[VHOST_USER_SET_VRING_NUM] = VHOST_SET_VRING_NUM,
-	[VHOST_USER_SET_VRING_BASE] = VHOST_SET_VRING_BASE,
-	[VHOST_USER_GET_VRING_BASE] = VHOST_GET_VRING_BASE,
-	[VHOST_USER_SET_VRING_ADDR] = VHOST_SET_VRING_ADDR,
-	[VHOST_USER_SET_VRING_KICK] = VHOST_SET_VRING_KICK,
-	[VHOST_USER_SET_MEM_TABLE] = VHOST_SET_MEM_TABLE,
-};
+static int
+vhost_kernel_get_features(struct virtio_user_dev *dev, uint64_t *features)
+{
+	int ret;
+	unsigned int tap_features;
+	struct vhost_kernel_data *data = dev->backend_data;
+
+	ret = vhost_kernel_ioctl(data->vhostfds[0], VHOST_GET_FEATURES, features);
+	if (ret < 0) {
+		PMD_DRV_LOG(ERR, "Failed to get features");
+		return -1;
+	}
+
+	ret = tap_support_features(&tap_features);
+	if (ret < 0) {
+		PMD_DRV_LOG(ERR, "Failed to get TAP features");
+		return -1;
+	}
+
+	/* with tap as the backend, all these features are supported
+	 * but not claimed by vhost-net, so we add them back when
+	 * reporting to upper layer.
+	 */
+	if (tap_features & IFF_VNET_HDR) {
+		*features |= VHOST_KERNEL_GUEST_OFFLOADS_MASK;
+		*features |= VHOST_KERNEL_HOST_OFFLOADS_MASK;
+	}
+
+	/* vhost_kernel will not declare this feature, but it does
+	 * support multi-queue.
+	 */
+	if (tap_features & IFF_MULTI_QUEUE)
+		*features |= (1ull << VIRTIO_NET_F_MQ);
+
+	return 0;
+}
+
+static int
+vhost_kernel_set_features(struct virtio_user_dev *dev, uint64_t features)
+{
+	struct vhost_kernel_data *data = dev->backend_data;
+
+	/* We don't need memory protection here */
+	features &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+	/* VHOST kernel does not know about below flags */
+	features &= ~VHOST_KERNEL_GUEST_OFFLOADS_MASK;
+	features &= ~VHOST_KERNEL_HOST_OFFLOADS_MASK;
+	features &= ~(1ULL << VIRTIO_NET_F_MQ);
+
+	return vhost_kernel_ioctl(data->vhostfds[0], VHOST_SET_FEATURES, &features);
+}
 
 static int
 add_memseg_list(const struct rte_memseg_list *msl, void *arg)
@@ -124,16 +203,18 @@ add_memseg_list(const struct rte_memseg_list *msl, void *arg)
  * have much more memory regions. Below function will treat each
  * contiguous memory space reserved by DPDK as one region.
  */
-static struct vhost_memory_kernel *
-prepare_vhost_memory_kernel(void)
+static int
+vhost_kernel_set_memory_table(struct virtio_user_dev *dev)
 {
+	struct vhost_kernel_data *data = dev->backend_data;
 	struct vhost_memory_kernel *vm;
+	int ret;
 
 	vm = malloc(sizeof(struct vhost_memory_kernel) +
 			max_regions *
 			sizeof(struct vhost_memory_region));
 	if (!vm)
-		return NULL;
+		goto err;
 
 	vm->nregions = 0;
 	vm->padding = 0;
@@ -142,149 +223,134 @@ prepare_vhost_memory_kernel(void)
 	 * The memory lock has already been taken by memory subsystem
 	 * or virtio_user_start_device().
 	 */
-	if (rte_memseg_list_walk_thread_unsafe(add_memseg_list, vm) < 0) {
-		free(vm);
-		return NULL;
-	}
+	ret = rte_memseg_list_walk_thread_unsafe(add_memseg_list, vm);
+	if (ret < 0)
+		goto err_free;
 
-	return vm;
-}
+	ret = vhost_kernel_ioctl(data->vhostfds[0], VHOST_SET_MEM_TABLE, vm);
+	if (ret < 0)
+		goto err_free;
 
-/* with below features, vhost kernel does not need to do the checksum and TSO,
- * these info will be passed to virtio_user through virtio net header.
- */
-#define VHOST_KERNEL_GUEST_OFFLOADS_MASK	\
-	((1ULL << VIRTIO_NET_F_GUEST_CSUM) |	\
-	 (1ULL << VIRTIO_NET_F_GUEST_TSO4) |	\
-	 (1ULL << VIRTIO_NET_F_GUEST_TSO6) |	\
-	 (1ULL << VIRTIO_NET_F_GUEST_ECN)  |	\
-	 (1ULL << VIRTIO_NET_F_GUEST_UFO))
+	free(vm);
 
-/* with below features, when flows from virtio_user to vhost kernel
- * (1) if flows goes up through the kernel networking stack, it does not need
- * to verify checksum, which can save CPU cycles;
- * (2) if flows goes through a Linux bridge and outside from an interface
- * (kernel driver), checksum and TSO will be done by GSO in kernel or even
- * offloaded into real physical device.
- */
-#define VHOST_KERNEL_HOST_OFFLOADS_MASK		\
-	((1ULL << VIRTIO_NET_F_HOST_TSO4) |	\
-	 (1ULL << VIRTIO_NET_F_HOST_TSO6) |	\
-	 (1ULL << VIRTIO_NET_F_CSUM))
+	return 0;
+err_free:
+	free(vm);
+err:
+	PMD_DRV_LOG(ERR, "Failed to set memory table");
+	return -1;
+}
 
-static unsigned int
-tap_support_features(void)
+static int
+vhost_kernel_set_vring(struct virtio_user_dev *dev, uint64_t req, struct vhost_vring_state *state)
 {
-	int tapfd;
-	unsigned int tap_features;
+	int ret, fd;
+	unsigned int index = state->index;
+	struct vhost_kernel_data *data = dev->backend_data;
 
-	tapfd = open(PATH_NET_TUN, O_RDWR);
-	if (tapfd < 0) {
-		PMD_DRV_LOG(ERR, "fail to open %s: %s",
-			    PATH_NET_TUN, strerror(errno));
-		return -1;
-	}
+	/* Convert from queue index to queue-pair & offset */
+	fd = data->vhostfds[state->index / 2];
+	state->index %= 2;
 
-	if (ioctl(tapfd, TUNGETFEATURES, &tap_features) == -1) {
-		PMD_DRV_LOG(ERR, "TUNGETFEATURES failed: %s", strerror(errno));
-		close(tapfd);
+	ret = vhost_kernel_ioctl(fd, req, state);
+	if (ret < 0) {
+		PMD_DRV_LOG(ERR, "Failed to set vring (request %" PRIu64 ")", req);
 		return -1;
 	}
 
-	close(tapfd);
-	return tap_features;
+	/* restore index back to queue index */
+	state->index = index;
+
+	return 0;
 }
 
 static int
-vhost_kernel_send_request(struct virtio_user_dev *dev,
-		   enum vhost_user_request req,
-		   void *arg)
+vhost_kernel_set_vring_num(struct virtio_user_dev *dev, struct vhost_vring_state *state)
 {
-	int ret = -1;
-	unsigned int i;
-	uint64_t req_kernel;
-	struct vhost_memory_kernel *vm = NULL;
-	int vhostfd;
-	unsigned int queue_sel;
-	unsigned int features;
+	return vhost_kernel_set_vring(dev, VHOST_SET_VRING_NUM, state);
+}
+
+static int
+vhost_kernel_set_vring_base(struct virtio_user_dev *dev, struct vhost_vring_state *state)
+{
+	return vhost_kernel_set_vring(dev, VHOST_SET_VRING_BASE, state);
+}
 
-	PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
+static int
+vhost_kernel_get_vring_base(struct virtio_user_dev *dev, struct vhost_vring_state *state)
+{
+	return vhost_kernel_set_vring(dev, VHOST_GET_VRING_BASE, state);
+}
 
-	req_kernel = vhost_req_user_to_kernel[req];
+static int
+vhost_kernel_set_vring_file(struct virtio_user_dev *dev, uint64_t req,
+		struct vhost_vring_file *file)
+{
+	int ret, fd;
+	unsigned int index = file->index;
+	struct vhost_kernel_data *data = dev->backend_data;
 
-	if (req_kernel == VHOST_SET_MEM_TABLE) {
-		vm = prepare_vhost_memory_kernel();
-		if (!vm)
-			return -1;
-		arg = (void *)vm;
+	/* Convert from queue index to queue-pair & offset */
+	fd = data->vhostfds[file->index / 2];
+	file->index %= 2;
+
+	ret = vhost_kernel_ioctl(fd, req, file);
+	if (ret < 0) {
+		PMD_DRV_LOG(ERR, "Failed to set vring file (request %" PRIu64 ")", req);
+		return -1;
 	}
 
-	if (req_kernel == VHOST_SET_FEATURES) {
-		/* We don't need memory protection here */
-		*(uint64_t *)arg &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+	/* restore index back to queue index */
+	file->index = index;
 
-		/* VHOST kernel does not know about below flags */
-		*(uint64_t *)arg &= ~VHOST_KERNEL_GUEST_OFFLOADS_MASK;
-		*(uint64_t *)arg &= ~VHOST_KERNEL_HOST_OFFLOADS_MASK;
+	return 0;
+}
 
-		*(uint64_t *)arg &= ~(1ULL << VIRTIO_NET_F_MQ);
-	}
+static int
+vhost_kernel_set_vring_kick(struct virtio_user_dev *dev, struct vhost_vring_file *file)
+{
+	return vhost_kernel_set_vring_file(dev, VHOST_SET_VRING_KICK, file);
+}
 
-	switch (req_kernel) {
-	case VHOST_SET_VRING_NUM:
-	case VHOST_SET_VRING_ADDR:
-	case VHOST_SET_VRING_BASE:
-	case VHOST_GET_VRING_BASE:
-	case VHOST_SET_VRING_KICK:
-	case VHOST_SET_VRING_CALL:
-		queue_sel = *(unsigned int *)arg;
-		vhostfd = dev->vhostfds[queue_sel / 2];
-		*(unsigned int *)arg = queue_sel % 2;
-		PMD_DRV_LOG(DEBUG, "vhostfd=%d, index=%u",
-			    vhostfd, *(unsigned int *)arg);
-		break;
-	default:
-		vhostfd = -1;
-	}
-	if (vhostfd == -1) {
-		for (i = 0; i < dev->max_queue_pairs; ++i) {
-			if (dev->vhostfds[i] < 0)
-				continue;
-
-			ret = ioctl(dev->vhostfds[i], req_kernel, arg);
-			if (ret < 0)
-				break;
-		}
-	} else {
-		ret = ioctl(vhostfd, req_kernel, arg);
-	}
+static int
+vhost_kernel_set_vring_call(struct virtio_user_dev *dev, struct vhost_vring_file *file)
+{
+	return vhost_kernel_set_vring_file(dev, VHOST_SET_VRING_CALL, file);
+}
 
-	if (!ret && req_kernel == VHOST_GET_FEATURES) {
-		features = tap_support_features();
-		/* with tap as the backend, all these features are supported
-		 * but not claimed by vhost-net, so we add them back when
-		 * reporting to upper layer.
-		 */
-		if (features & IFF_VNET_HDR) {
-			*((uint64_t *)arg) |= VHOST_KERNEL_GUEST_OFFLOADS_MASK;
-			*((uint64_t *)arg) |= VHOST_KERNEL_HOST_OFFLOADS_MASK;
-		}
+static int
+vhost_kernel_set_vring_addr(struct virtio_user_dev *dev, struct vhost_vring_addr *addr)
+{
+	int ret, fd;
+	unsigned int index = addr->index;
+	struct vhost_kernel_data *data = dev->backend_data;
+
+	/* Convert from queue index to queue-pair & offset */
+	fd = data->vhostfds[addr->index / 2];
+	addr->index %= 2;
 
-		/* vhost_kernel will not declare this feature, but it does
-		 * support multi-queue.
-		 */
-		if (features & IFF_MULTI_QUEUE)
-			*(uint64_t *)arg |= (1ull << VIRTIO_NET_F_MQ);
+	ret = vhost_kernel_ioctl(fd, VHOST_SET_VRING_ADDR, addr);
+	if (ret < 0) {
+		PMD_DRV_LOG(ERR, "Failed to set vring address");
+		return -1;
 	}
 
-	if (vm)
-		free(vm);
+	/* restore index back to queue index */
+	addr->index = index;
 
-	if (ret < 0)
-		PMD_DRV_LOG(ERR, "%s failed: %s",
-			    vhost_msg_strings[req], strerror(errno));
+	return 0;
+}
 
-	return ret;
+static int
+vhost_kernel_get_status(struct virtio_user_dev *dev __rte_unused, uint8_t *status __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+static int
+vhost_kernel_set_status(struct virtio_user_dev *dev __rte_unused, uint8_t status __rte_unused)
+{
+	return -ENOTSUP;
 }
 
 /**
@@ -298,21 +364,82 @@ static int
 vhost_kernel_setup(struct virtio_user_dev *dev)
 {
 	int vhostfd;
-	uint32_t i;
+	uint32_t q, i;
+	struct vhost_kernel_data *data;
+
+	data = malloc(sizeof(*data));
+	if (!data) {
+		PMD_INIT_LOG(ERR, "(%s) Failed to allocate Vhost-kernel data", dev->path);
+		return -1;
+	}
+
+	data->vhostfds = malloc(dev->max_queue_pairs * sizeof(int));
+	if (!data->vhostfds) {
+		PMD_INIT_LOG(ERR, "(%s) Failed to allocate Vhost FDs", dev->path);
+		goto err_data;
+	}
+	data->tapfds = malloc(dev->max_queue_pairs * sizeof(int));
+	if (!data->tapfds) {
+		PMD_INIT_LOG(ERR, "(%s) Failed to allocate TAP FDs", dev->path);
+		goto err_vhostfds;
+	}
+
+	for (q = 0; q < dev->max_queue_pairs; ++q) {
+		data->vhostfds[q] = -1;
+		data->tapfds[q] = -1;
+	}
 
 	get_vhost_kernel_max_regions();
 
 	for (i = 0; i < dev->max_queue_pairs; ++i) {
 		vhostfd = open(dev->path, O_RDWR);
 		if (vhostfd < 0) {
-			PMD_DRV_LOG(ERR, "fail to open %s, %s",
-				    dev->path, strerror(errno));
-			return -1;
+			PMD_DRV_LOG(ERR, "fail to open %s, %s", dev->path, strerror(errno));
+			goto err_tapfds;
 		}
 
-		dev->vhostfds[i] = vhostfd;
+		data->vhostfds[i] = vhostfd;
 	}
 
+	dev->backend_data = data;
+
+	return 0;
+
+err_tapfds:
+	for (i = 0; i < dev->max_queue_pairs; i++)
+		if (data->vhostfds[i] >= 0)
+			close(data->vhostfds[i]);
+
+	free(data->tapfds);
+err_vhostfds:
+	free(data->vhostfds);
+err_data:
+	free(data);
+
+	return -1;
+}
+
+static int
+vhost_kernel_destroy(struct virtio_user_dev *dev)
+{
+	struct vhost_kernel_data *data = dev->backend_data;
+	uint32_t i;
+
+	if (!data)
+		return 0;
+
+	for (i = 0; i < dev->max_queue_pairs; ++i) {
+		if (data->vhostfds[i] >= 0)
+			close(data->vhostfds[i]);
+		if (data->tapfds[i] >= 0)
+			close(data->tapfds[i]);
+	}
+
+	free(data->vhostfds);
+	free(data->tapfds);
+	free(data);
+	dev->backend_data = NULL;
+
 	return 0;
 }
 
@@ -348,14 +475,15 @@ vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
 	int vhostfd;
 	int tapfd;
 	int req_mq = (dev->max_queue_pairs > 1);
+	struct vhost_kernel_data *data = dev->backend_data;
 
-	vhostfd = dev->vhostfds[pair_idx];
+	vhostfd = data->vhostfds[pair_idx];
 
 	if (dev->qp_enabled[pair_idx] == enable)
 		return 0;
 
 	if (!enable) {
-		tapfd = dev->tapfds[pair_idx];
+		tapfd = data->tapfds[pair_idx];
 		if (vhost_kernel_set_backend(vhostfd, -1) < 0) {
 			PMD_DRV_LOG(ERR, "fail to set backend for vhost kernel");
 			return -1;
@@ -368,8 +496,8 @@ vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
 		return 0;
 	}
 
-	if (dev->tapfds[pair_idx] >= 0) {
-		tapfd = dev->tapfds[pair_idx];
+	if (data->tapfds[pair_idx] >= 0) {
+		tapfd = data->tapfds[pair_idx];
 		if (vhost_kernel_tap_set_offload(tapfd, dev->features) == -1)
 			return -1;
 		if (req_mq && vhost_kernel_tap_set_queue(tapfd, true) < 0) {
@@ -392,7 +520,7 @@ vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
 		return -1;
 	}
 
-	dev->tapfds[pair_idx] = tapfd;
+	data->tapfds[pair_idx] = tapfd;
 
 set_backend:
 	if (vhost_kernel_set_backend(vhostfd, tapfd) < 0) {
@@ -404,9 +532,45 @@ set_backend:
 	return 0;
 }
 
+static int
+vhost_kernel_get_backend_features(uint64_t *features)
+{
+	*features = 0;
+
+	return 0;
+}
+
+static int
+vhost_kernel_update_link_state(struct virtio_user_dev *dev __rte_unused)
+{
+	/* Nothing to update (Maybe get TAP interface link state?) */
+	return 0;
+}
+
+static int
+vhost_kernel_get_intr_fd(struct virtio_user_dev *dev __rte_unused)
+{
+	/* No link state interrupt with Vhost-kernel */
+	return -1;
+}
+
 struct virtio_user_backend_ops virtio_ops_kernel = {
 	.setup = vhost_kernel_setup,
+	.destroy = vhost_kernel_destroy,
+	.get_backend_features = vhost_kernel_get_backend_features,
 	.set_owner = vhost_kernel_set_owner,
-	.send_request = vhost_kernel_send_request,
-	.enable_qp = vhost_kernel_enable_queue_pair
+	.get_features = vhost_kernel_get_features,
+	.set_features = vhost_kernel_set_features,
+	.set_memory_table = vhost_kernel_set_memory_table,
+	.set_vring_num = vhost_kernel_set_vring_num,
+	.set_vring_base = vhost_kernel_set_vring_base,
+	.get_vring_base = vhost_kernel_get_vring_base,
+	.set_vring_call = vhost_kernel_set_vring_call,
+	.set_vring_kick = vhost_kernel_set_vring_kick,
+	.set_vring_addr = vhost_kernel_set_vring_addr,
+	.get_status = vhost_kernel_get_status,
+	.set_status = vhost_kernel_set_status,
+	.enable_qp = vhost_kernel_enable_queue_pair,
+	.update_link_state = vhost_kernel_update_link_state,
+	.get_intr_fd = vhost_kernel_get_intr_fd,
 };