* [PATCHv4 3/6] qemu/net: add raw backend
[not found] <cover.1257200517.git.mst@redhat.com>
2009-11-02 22:23 ` [PATCHv4 1/6] qemu/virtio: move features to an inline function Michael S. Tsirkin
2009-11-02 22:23 ` [PATCHv4 2/6] qemu/net: routines to get tap fd Michael S. Tsirkin
@ 2009-11-02 22:23 ` Or Gerlitz
2009-11-02 22:24 ` [PATCHv4 4/6] qemu/net: move typedef to qemu-common.h Michael S. Tsirkin
` (2 subsequent siblings)
5 siblings, 0 replies; 11+ messages in thread
From: Or Gerlitz @ 2009-11-02 22:23 UTC (permalink / raw)
To: avi, kvm, virtualization; +Cc: gregory.haskins
Add raw network backend option which uses a packet socket to provide
raw networking access. Once the socket is opened it's bound to a
provided host interface, such that packets received on the interface
are delivered to the VM and packets sent by the VM are sent to the
interface.
This is functionally similar to the existing pcap network
backend, with the same advantages and problems.
Differences from pcap:
- can get an open socket from the monitor,
which allows running without NET_ADMIN priviledges
- support iovec sends with writev, saving one data copy
- one less dependency on an external library
- we have access to the underlying file descriptor
which makes it possible to connect to vhost net
- don't support polling all interfaces, always bind to a specific one
Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
hw/virtio-net.c | 3 +-
net.c | 192 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
qemu-options.hx | 4 +
3 files changed, 198 insertions(+), 1 deletions(-)
diff --git a/hw/virtio-net.c b/hw/virtio-net.c
index 469c6e3..2e51a6a 100644
--- a/hw/virtio-net.c
+++ b/hw/virtio-net.c
@@ -531,7 +531,8 @@ static ssize_t virtio_net_receive2(VLANClientState *vc, const uint8_t *buf, size
virtqueue_pop(n->rx_vq, &elem) == 0) {
if (i == 0)
return -1;
- fprintf(stderr, "virtio-net truncating packet\n");
+ fprintf(stderr, "virtio-net truncating packet. offset %zd size %zd\n",
+ offset, size);
exit(1);
}
diff --git a/net.c b/net.c
index 8ac639b..1fb2f2f 100644
--- a/net.c
+++ b/net.c
@@ -93,6 +93,9 @@
#endif
#endif
+#include <netpacket/packet.h>
+#include <net/ethernet.h>
+
#if defined(__OpenBSD__)
#include <util.h>
#endif
@@ -1870,6 +1873,158 @@ static TAPState *net_tap_init(VLANState *vlan, const char *model,
#endif /* !_WIN32 */
+typedef struct RAWState {
+ VLANClientState *vc;
+ int fd;
+ uint8_t buf[4096];
+ int promisc;
+} RAWState;
+
+static int net_raw_fd_init(Monitor *mon, const char *ifname, int promisc)
+{
+ int fd, ret;
+ struct ifreq req;
+ struct sockaddr_ll lladdr;
+
+ fd = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
+ if (fd < 0)
+ fprintf(stderr, "packet socket failed\n");
+
+ memset(&req, 0, sizeof(req));
+ strncpy(req.ifr_name, ifname, IFNAMSIZ-1);
+ ret = ioctl(fd, SIOCGIFINDEX, &req);
+ if (ret < 0)
+ fprintf(stderr, "SIOCGIFINDEX failed\n");
+
+ memset(&lladdr, 0, sizeof(lladdr));
+ lladdr.sll_family = AF_PACKET;
+ lladdr.sll_protocol = htons(ETH_P_ALL);
+ lladdr.sll_ifindex = req.ifr_ifindex;
+ ret = bind(fd, (const struct sockaddr *)&lladdr, sizeof(lladdr));
+ if (ret < 0)
+ fprintf(stderr, "bind failed\n");
+
+ /* set iface to promiscuous mode (packets sent to the VM MAC) */
+ if (promisc) {
+ ret = ioctl(fd, SIOCGIFFLAGS, &req);
+ if (ret < 0)
+ perror("SIOCGIFFLAGS failed\n");
+ req.ifr_flags |= IFF_PROMISC;
+ ret = ioctl(fd, SIOCSIFFLAGS, &req);
+ if (ret < 0)
+ fprintf(stderr, "SIOCSIFFLAGS to promiscous failed\n");
+ }
+
+ ret = fcntl(fd, F_SETFL, fcntl(fd, F_GETFL) | O_NONBLOCK);
+ if (ret < 0)
+ fprintf(stderr, "O_NONBLOCK set failed\n");
+
+ return fd;
+}
+
+static void raw_cleanup(VLANClientState *vc)
+{
+ struct ifreq req;
+ RAWState *s = vc->opaque;
+
+ qemu_set_fd_handler2(s->fd, NULL, NULL, NULL, NULL);
+ if (s->promisc) {
+ ioctl(s->fd, SIOCGIFFLAGS, &req);
+ req.ifr_flags &= ~IFF_PROMISC;
+ ioctl(s->fd, SIOCSIFFLAGS, &req);
+ }
+ close(s->fd);
+ qemu_free(s);
+}
+
+static void raw_send(void *opaque);
+
+static int raw_can_send(void *opaque)
+{
+ RAWState *s = opaque;
+
+ return qemu_can_send_packet(s->vc);
+}
+
+static void raw_send_completed(VLANClientState *vc, ssize_t len)
+{
+ RAWState *s = vc->opaque;
+
+ qemu_set_fd_handler2(s->fd, raw_can_send, raw_send, NULL, s);
+}
+
+static void raw_send(void *opaque)
+{
+ RAWState *s = opaque;
+ int size;
+
+ do {
+ size = recv(s->fd, s->buf, sizeof(s->buf), MSG_TRUNC);
+ if (size <= 0)
+ break;
+
+ size = qemu_send_packet_async(s->vc, s->buf, size,
+ raw_send_completed);
+ if (size == 0)
+ qemu_set_fd_handler2(s->fd, NULL, NULL, NULL, NULL);
+
+ } while (size > 0);
+}
+
+static ssize_t raw_receive_iov(VLANClientState *vc, const struct iovec *iov,
+ int iovcnt)
+{
+ ssize_t len;
+ RAWState *s = vc->opaque;
+
+ do {
+ len = writev(s->fd, iov, iovcnt);
+ } while (len == -1 && (errno == EINTR || errno == EAGAIN));
+
+ return len;
+}
+
+static ssize_t raw_receive(VLANClientState *vc, const uint8_t *buf, size_t size)
+{
+ struct iovec iov[1];
+
+ iov[0].iov_base = (char *)buf;
+ iov[0].iov_len = size;
+
+ return raw_receive_iov(vc, iov, 1);
+}
+
+static int net_raw_init(Monitor *mon, VLANState *vlan, const char *model,
+ const char *name, const char *ifname,
+ int promisc, int fd)
+{
+ RAWState *s;
+
+ s = qemu_mallocz(sizeof(RAWState));
+
+ if (fd == -1) {
+ s->fd = net_raw_fd_init(mon, ifname, promisc);
+ s->promisc = promisc;
+ } else
+ s->fd = fd;
+
+ fcntl(s->fd, F_SETFL, O_NONBLOCK);
+
+ s->vc = qemu_new_vlan_client(vlan, model, name, NULL, raw_receive,
+ raw_receive_iov, raw_cleanup, s);
+ qemu_set_fd_handler2(s->fd, raw_can_send, raw_send, NULL, s);
+
+ if (fd == -1)
+ snprintf(s->vc->info_str, sizeof(s->vc->info_str),
+ "raw: ifname=%s, promisc=%d", ifname, promisc);
+ else
+ snprintf(s->vc->info_str, sizeof(s->vc->info_str),
+ "raw: fd=%d", fd);
+
+ vlan->nb_host_devs++;
+ return 0;
+}
+
#if defined(CONFIG_VDE)
typedef struct VDEState {
VLANClientState *vc;
@@ -2632,6 +2787,23 @@ static int net_init_nic(QemuOpts *opts, Monitor *mon)
return idx;
}
+static int net_init_raw(QemuOpts *opts, Monitor *mon)
+{
+ VLANState *vlan;
+ int fd = -1;
+ vlan = qemu_find_vlan(qemu_opt_get_number(opts, "vlan", 0), 1);
+ if (qemu_opt_get(opts, "fd")) {
+ fd = net_handle_fd_param(mon, qemu_opt_get(opts, "fd"));
+ if (fd < 0)
+ return -EINVAL;
+ }
+ return net_raw_init(mon, vlan, "raw",
+ qemu_opt_get(opts, "name"),
+ qemu_opt_get(opts, "ifname"),
+ qemu_opt_get_bool(opts, "promisc", 0),
+ fd);
+}
+
static int net_init_slirp_configs(const char *name, const char *value, void *opaque)
{
struct slirp_config_str *config;
@@ -3136,6 +3308,26 @@ static struct {
},
{ /* end of list */ }
},
+ }, {
+ .type = "raw",
+ .init = net_init_raw,
+ .desc = {
+ NET_COMMON_PARAMS_DESC,
+ {
+ .name = "fd",
+ .type = QEMU_OPT_STRING,
+ .help = "file descriptor of an already opened raw socket",
+ }, {
+ .name = "ifname",
+ .type = QEMU_OPT_STRING,
+ .help = "interface name",
+ }, {
+ .name = "promisc",
+ .type = QEMU_OPT_BOOL,
+ .help = "enable promiscious mode at startup",
+ },
+ { /* end of list */ }
+ },
#ifdef CONFIG_VDE
}, {
.type = "vde",
diff --git a/qemu-options.hx b/qemu-options.hx
index bde3e3f..0d5440f 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -825,6 +825,10 @@ DEF("net", HAS_ARG, QEMU_OPTION_net,
" default of 'sndbuf=1048576' can be disabled using 'sndbuf=0'\n"
#endif
#endif
+ "-net raw[,vlan=n][,name=str],ifname=name[,promisc=m]\n"
+ " bound the host network interface to VLAN 'n' in a raw manner:\n"
+ " packets received on the interface are delivered to the vlan and\n"
+ " packets delivered on the vlan are sent to the interface\n"
"-net socket[,vlan=n][,name=str][,fd=h][,listen=[host]:port][,connect=host:port]\n"
" connect the vlan 'n' to another VLAN using a socket connection\n"
"-net socket[,vlan=n][,name=str][,fd=h][,mcast=maddr:port]\n"
--
1.6.5.2.143.g8cc62
^ permalink raw reply related [flat|nested] 11+ messages in thread* [PATCHv4 6/6] qemu-kvm: vhost-net implementation
[not found] <cover.1257200517.git.mst@redhat.com>
` (4 preceding siblings ...)
2009-11-02 22:24 ` [PATCHv4 5/6] qemu/raw: add API to get raw socket Michael S. Tsirkin
@ 2009-11-02 22:24 ` Michael S. Tsirkin
2009-11-05 0:22 ` Sridhar Samudrala
5 siblings, 1 reply; 11+ messages in thread
From: Michael S. Tsirkin @ 2009-11-02 22:24 UTC (permalink / raw)
To: avi, kvm, virtualization; +Cc: gregory.haskins
This adds support for vhost-net virtio kernel backend.
This patch is not intended to being merged yet.
I'm posting it for the benefit of people testing
the backend.
Usage instructions:
vhost currently requires MSI-X support in guest virtio.
This means guests kernel version should be >= 2.6.31.
To enable vhost, simply add ",vhost" flag to nic options.
Example with tap backend:
qemu-system-x86_64 -m 1G disk-c.qcow2 \
-net tap,ifname=msttap0,script=/home/mst/ifup,downscript=no \
-net nic,model=virtio,vhost
Example with raw socket backend:
ifconfig eth3 promisc
qemu-system-x86_64 -m 1G disk-c.qcow2 \
-net raw,ifname=eth3 \
-net nic,model=virtio,vhost
This patchset is RFC, but works without issues for me.
TODO:
* migration support
* level triggered interrupts
* fix driver unloading/hotplug
* general cleanup and upstreaming
It still needs to be split up, tested and benchmarked properly,
but posting it here in case people want to test drive
the kernel bits I posted.
Some further info, performance etc:
http://www.linux-kvm.org/page/VhostNet
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
Makefile.target | 3 +-
hw/vhost_net.c | 251 +++++++++++++++++++++++++++++++++++++++++++++
hw/vhost_net.h | 38 +++++++
hw/virtio-net.c | 67 ++++++++++--
hw/virtio-pci.c | 40 +++++++
hw/virtio.c | 19 ----
hw/virtio.h | 28 +++++-
kvm/include/linux/vhost.h | 126 +++++++++++++++++++++++
net.c | 7 ++
net.h | 1 +
qemu-kvm.c | 8 --
qemu-kvm.h | 9 ++
12 files changed, 555 insertions(+), 42 deletions(-)
create mode 100644 hw/vhost_net.c
create mode 100644 hw/vhost_net.h
create mode 100644 kvm/include/linux/vhost.h
diff --git a/Makefile.target b/Makefile.target
index acee285..0d8e688 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -160,7 +160,8 @@ obj-y = vl.o monitor.o pci.o isa_mmio.o machine.o \
gdbstub.o gdbstub-xml.o
# virtio has to be here due to weird dependency between PCI and virtio-net.
# need to fix this properly
-obj-y += virtio-blk.o virtio-balloon.o virtio-net.o virtio-console.o virtio-pci.o
+obj-y += virtio-blk.o virtio-balloon.o virtio-net.o virtio-console.o virtio-pci.o \
+ vhost_net.o
obj-$(CONFIG_KVM) += kvm.o kvm-all.o
# MSI-X depends on kvm for interrupt injection,
# so moved it from Makefile.hw to Makefile.target for now
diff --git a/hw/vhost_net.c b/hw/vhost_net.c
new file mode 100644
index 0000000..bc179ab
--- /dev/null
+++ b/hw/vhost_net.c
@@ -0,0 +1,251 @@
+#include <sys/eventfd.h>
+#include <sys/socket.h>
+#include <linux/kvm.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <linux/vhost.h>
+#include <linux/virtio_ring.h>
+#include <netpacket/packet.h>
+#include <net/ethernet.h>
+#include <net/if.h>
+#include <netinet/in.h>
+
+#include <stdio.h>
+
+#include "net.h"
+#include "qemu-kvm.h"
+
+#include "vhost_net.h"
+
+static int vhost_virtqueue_init(struct vhost_dev *dev,
+ struct VirtIODevice *vdev,
+ struct vhost_virtqueue *vq,
+ struct VirtQueue *q,
+ unsigned idx)
+{
+ target_phys_addr_t s, l;
+ int r;
+ struct vhost_vring_addr addr = {
+ .index = idx,
+ };
+ struct vhost_vring_file file = {
+ .index = idx,
+ };
+ struct vhost_vring_state size = {
+ .index = idx,
+ };
+
+ size.num = q->vring.num;
+ r = ioctl(dev->control, VHOST_SET_VRING_NUM, &size);
+ if (r)
+ return -errno;
+
+ file.fd = vq->kick = eventfd(0, 0);
+ r = ioctl(dev->control, VHOST_SET_VRING_KICK, &file);
+ if (r)
+ return -errno;
+ file.fd = vq->call = eventfd(0, 0);
+ r = ioctl(dev->control, VHOST_SET_VRING_CALL, &file);
+ if (r)
+ return -errno;
+
+ s = l = sizeof(struct vring_desc) * q->vring.num;
+ vq->desc = cpu_physical_memory_map(q->vring.desc, &l, 0);
+ if (!vq->desc || l != s)
+ return -ENOMEM;
+ addr.user_addr = (u_int64_t)(unsigned long)vq->desc;
+ r = ioctl(dev->control, VHOST_SET_VRING_DESC, &addr);
+ if (r < 0)
+ return -errno;
+ s = l = offsetof(struct vring_avail, ring) +
+ sizeof(u_int64_t) * q->vring.num;
+ vq->avail = cpu_physical_memory_map(q->vring.avail, &l, 0);
+ if (!vq->avail || l != s)
+ return -ENOMEM;
+ addr.user_addr = (u_int64_t)(unsigned long)vq->avail;
+ r = ioctl(dev->control, VHOST_SET_VRING_AVAIL, &addr);
+ if (r < 0)
+ return -errno;
+ s = l = offsetof(struct vring_used, ring) +
+ sizeof(struct vring_used_elem) * q->vring.num;
+ vq->used = cpu_physical_memory_map(q->vring.used, &l, 1);
+ if (!vq->used || l != s)
+ return -ENOMEM;
+ addr.user_addr = (u_int64_t)(unsigned long)vq->used;
+ r = ioctl(dev->control, VHOST_SET_VRING_USED, &addr);
+ if (r < 0)
+ return -errno;
+
+ r = vdev->binding->irqfd(vdev->binding_opaque, q->vector, vq->call);
+ if (r < 0)
+ return -errno;
+
+ r = vdev->binding->queuefd(vdev->binding_opaque, idx, vq->kick);
+ if (r < 0)
+ return -errno;
+
+ return 0;
+}
+
+static int vhost_dev_init(struct vhost_dev *hdev)
+{
+ uint64_t features;
+ int r;
+ hdev->control = open("/dev/vhost-net", O_RDWR);
+ if (hdev->control < 0)
+ return -errno;
+ r = ioctl(hdev->control, VHOST_SET_OWNER, NULL);
+ if (r < 0)
+ return -errno;
+
+ r = ioctl(hdev->control, VHOST_GET_FEATURES, &features);
+ if (r < 0)
+ return -errno;
+ hdev->features = features;
+ return 0;
+}
+
+static void vhost_dev_cleanup(struct vhost_dev *hdev)
+{
+ close(hdev->control);
+}
+
+static int vhost_dev_start(struct vhost_dev *hdev,
+ VirtIODevice *vdev)
+{
+ int i, r, n = 0;
+ struct vhost_memory *mem;
+
+ r = ioctl(hdev->control, VHOST_ACK_FEATURES, &hdev->acked_features);
+ if (r < 0)
+ return -errno;
+
+ for (i = 0; i < KVM_MAX_NUM_MEM_REGIONS; ++i) {
+ if (!slots[i].len || (slots[i].flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+ continue;
+ }
+ ++n;
+ }
+
+ mem = qemu_mallocz(offsetof(struct vhost_memory, regions) +
+ n * sizeof(struct vhost_memory_region));
+ if (!mem)
+ return -ENOMEM;
+ mem->nregions = n;
+ n = 0;
+ for (i = 0; i < KVM_MAX_NUM_MEM_REGIONS; ++i) {
+ if (!slots[i].len || (slots[i].flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+ continue;
+ }
+ mem->regions[n].guest_phys_addr = slots[i].phys_addr;
+ mem->regions[n].memory_size = slots[i].len;
+ mem->regions[n].userspace_addr = slots[i].userspace_addr;
+ ++n;
+ }
+
+ r = ioctl(hdev->control, VHOST_SET_MEM_TABLE, mem);
+ if (r < 0)
+ return -errno;
+
+ for (i = 0; i < hdev->nvqs; ++i) {
+ r = vhost_virtqueue_init(hdev,
+ vdev,
+ hdev->vqs + i,
+ vdev->vq + i,
+ i);
+ if (r < 0)
+ return r;
+ }
+
+ return 0;
+}
+
+unsigned vhost_net_get_features(struct vhost_net *net)
+{
+ unsigned features = 0;
+ if (net->dev.features & (1 << VIRTIO_F_NOTIFY_ON_EMPTY))
+ features |= VIRTIO_F_NOTIFY_ON_EMPTY;
+ if (net->dev.features & (1 << VIRTIO_RING_F_INDIRECT_DESC))
+ features |= VIRTIO_RING_F_INDIRECT_DESC;
+ return features;
+}
+
+void vhost_net_ack_features(struct vhost_net *net, unsigned features)
+{
+ net->dev.acked_features = net->dev.backend_features;
+ if (features & (1 << VIRTIO_F_NOTIFY_ON_EMPTY))
+ net->dev.acked_features |= VIRTIO_F_NOTIFY_ON_EMPTY;
+ if (features & (1 << VIRTIO_RING_F_INDIRECT_DESC))
+ net->dev.acked_features |= VIRTIO_RING_F_INDIRECT_DESC;
+}
+
+static int vhost_net_get_fd(VLANClientState *backend,
+ unsigned long long *backend_features)
+{
+ int r;
+ r = raw_get_fd(backend);
+ if (r >= 0) {
+ *backend_features = (1 << VHOST_NET_F_VIRTIO_NET_HDR);
+ return r;
+ }
+ r = tap_get_fd(backend);
+ if (r >= 0) {
+ *backend_features = 0;
+ return r;
+ }
+ fprintf(stderr, "vhost requires raw socket or tap backend\n");
+ return -EBADFD;
+}
+
+int vhost_net_init(struct vhost_net *net, VLANClientState *backend)
+{
+ int r;
+
+ if (!backend) {
+ fprintf(stderr, "vhost requires backend to be setup\n");
+ return -EINVAL;
+ }
+ r = vhost_net_get_fd(backend, &net->dev.backend_features);
+ if (r < 0)
+ return r;
+ net->backend = r;
+
+ r = vhost_dev_init(&net->dev);
+ if (r < 0)
+ return r;
+ if (~net->dev.features & net->dev.backend_features) {
+ fprintf(stderr, "vhost lacks feature mask %llu for backend\n",
+ ~net->dev.features & net->dev.backend_features);
+ vhost_dev_cleanup(&net->dev);
+ return -EINVAL;
+ }
+
+ /* Set sane init value. Override when guest acks. */
+ vhost_net_ack_features(net, 0);
+ return 0;
+}
+
+int vhost_net_start(struct vhost_net *net,
+ VirtIODevice *dev)
+{
+ struct vhost_vring_file file = { };
+ int r;
+
+ net->dev.nvqs = 2;
+ net->dev.vqs = net->vqs;
+ r = vhost_dev_start(&net->dev, dev);
+ if (r < 0)
+ return r;
+
+ /* Stop polling backend from qemu. */
+ qemu_set_fd_handler(net->backend, NULL, NULL, NULL);
+ file.fd = net->backend;
+ for (file.index = 0; file.index < net->dev.nvqs; ++file.index) {
+ r = ioctl(net->dev.control, VHOST_NET_SET_BACKEND, &file);
+ if (r < 0) {
+ /* TODO: cleanup on error. */
+ return -errno;
+ }
+ }
+ return 0;
+}
diff --git a/hw/vhost_net.h b/hw/vhost_net.h
new file mode 100644
index 0000000..65720e1
--- /dev/null
+++ b/hw/vhost_net.h
@@ -0,0 +1,38 @@
+#ifndef VHOST_NET_H
+#define VHOST_NET_H
+
+#include "hw/virtio.h"
+
+struct vhost_virtqueue {
+ int kick;
+ int call;
+ void *desc;
+ void *avail;
+ void *used;
+};
+
+struct vhost_dev {
+ int control;
+ struct vhost_virtqueue *vqs;
+ int nvqs;
+ unsigned long long features;
+ unsigned long long acked_features;
+ unsigned long long backend_features;
+};
+
+struct vhost_net {
+ struct vhost_dev dev;
+ struct vhost_virtqueue vqs[2];
+ int backend;
+};
+
+int vhost_net_init(struct vhost_net *net,
+ VLANClientState *backend);
+
+int vhost_net_start(struct vhost_net *net,
+ VirtIODevice *dev);
+
+unsigned vhost_net_get_features(struct vhost_net *net);
+void vhost_net_ack_features(struct vhost_net *net, unsigned features);
+
+#endif
diff --git a/hw/virtio-net.c b/hw/virtio-net.c
index 2e51a6a..3b0b947 100644
--- a/hw/virtio-net.c
+++ b/hw/virtio-net.c
@@ -19,6 +19,8 @@
#include "qemu-kvm.h"
#endif
+#include "vhost_net.h"
+
#define TAP_VNET_HDR
#define VIRTIO_NET_VM_VERSION 10
@@ -56,6 +58,8 @@ typedef struct VirtIONet
uint8_t *macs;
} mac_table;
uint32_t *vlans;
+ int vhost_device;
+ struct vhost_net vhost;
} VirtIONet;
/* TODO
@@ -127,16 +131,10 @@ static void virtio_net_reset(VirtIODevice *vdev)
static uint32_t virtio_net_get_features(VirtIODevice *vdev)
{
- uint32_t features = (1 << VIRTIO_NET_F_MAC) |
- (1 << VIRTIO_NET_F_MRG_RXBUF) |
- (1 << VIRTIO_NET_F_STATUS) |
- (1 << VIRTIO_NET_F_CTRL_VQ) |
- (1 << VIRTIO_NET_F_CTRL_RX) |
- (1 << VIRTIO_NET_F_CTRL_VLAN) |
- (1 << VIRTIO_NET_F_CTRL_RX_EXTRA);
+ uint32_t features = 0;
+ VirtIONet *n = to_virtio_net(vdev);
#ifdef TAP_VNET_HDR
- VirtIONet *n = to_virtio_net(vdev);
VLANClientState *host = n->vc->vlan->first_client;
if (tap_has_vnet_hdr(host)) {
@@ -149,12 +147,23 @@ static uint32_t virtio_net_get_features(VirtIODevice *vdev)
features |= (1 << VIRTIO_NET_F_HOST_TSO4);
features |= (1 << VIRTIO_NET_F_HOST_TSO6);
features |= (1 << VIRTIO_NET_F_HOST_ECN);
- features |= (1 << VIRTIO_NET_F_MRG_RXBUF);
/* Kernel can't actually handle UFO in software currently. */
}
#endif
- return features | virtio_common_features();
+ if (n->vhost_device)
+ features |= (1 << VIRTIO_NET_F_MAC) | vhost_net_get_features(&n->vhost);
+ else
+ features |= virtio_common_features() |
+ (1 << VIRTIO_NET_F_MAC) |
+ (1 << VIRTIO_NET_F_MRG_RXBUF) |
+ (1 << VIRTIO_NET_F_STATUS) |
+ (1 << VIRTIO_NET_F_CTRL_VQ) |
+ (1 << VIRTIO_NET_F_CTRL_RX) |
+ (1 << VIRTIO_NET_F_CTRL_VLAN) |
+ (1 << VIRTIO_NET_F_CTRL_RX_EXTRA);
+
+ return features;
}
static uint32_t virtio_net_bad_features(VirtIODevice *vdev)
@@ -175,11 +184,15 @@ static uint32_t virtio_net_bad_features(VirtIODevice *vdev)
static void virtio_net_set_features(VirtIODevice *vdev, uint32_t features)
{
VirtIONet *n = to_virtio_net(vdev);
+ /* vhost net supports no features */
#ifdef TAP_VNET_HDR
VLANClientState *host = n->vc->vlan->first_client;
#endif
n->mergeable_rx_bufs = !!(features & (1 << VIRTIO_NET_F_MRG_RXBUF));
+ if (n->vhost_device) {
+ vhost_net_ack_features(&n->vhost, features);
+ }
#ifdef TAP_VNET_HDR
if (!tap_has_vnet_hdr(host) || !host->set_offload)
@@ -351,6 +364,9 @@ static void virtio_net_handle_rx(VirtIODevice *vdev, VirtQueue *vq)
static int do_virtio_net_can_receive(VirtIONet *n, int bufsize)
{
+ if (n->vhost_device)
+ return 0;
+
if (!virtio_queue_ready(n->rx_vq) ||
!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK))
return 0;
@@ -411,6 +427,7 @@ static int iov_fill(struct iovec *iov, int iovcnt, const void *buf, int count)
while (offset < count && i < iovcnt) {
int len = MIN(iov[i].iov_len, count - offset);
memcpy(iov[i].iov_base, buf + offset, len);
+
offset += len;
i++;
}
@@ -611,6 +628,8 @@ static void virtio_net_flush_tx(VirtIONet *n, VirtQueue *vq)
#else
int has_vnet_hdr = 0;
#endif
+ if (n->vhost_device)
+ return;
if (!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK))
return;
@@ -810,6 +829,8 @@ static void virtio_net_cleanup(VLANClientState *vc)
{
VirtIONet *n = vc->opaque;
+ /* TODO: vhost device cleanup */
+
qemu_purge_queued_packets(vc);
unregister_savevm("virtio-net", n);
@@ -823,6 +844,21 @@ static void virtio_net_cleanup(VLANClientState *vc)
virtio_cleanup(&n->vdev);
}
+static void virtio_net_driver_ok(VirtIODevice *vdev)
+{
+ VirtIONet *n = to_virtio_net(vdev);
+ int r;
+
+ if (!n->vhost_device)
+ return;
+
+ r = vhost_net_start(&n->vhost, vdev);
+ if (r) {
+ fprintf(stderr, "\nvhost_net_init returned %d\n", r);
+ exit(-r);
+ }
+}
+
VirtIODevice *virtio_net_init(DeviceState *dev)
{
VirtIONet *n;
@@ -831,6 +867,15 @@ VirtIODevice *virtio_net_init(DeviceState *dev)
n = (VirtIONet *)virtio_common_init("virtio-net", VIRTIO_ID_NET,
sizeof(struct virtio_net_config),
sizeof(VirtIONet));
+ n->vhost_device = dev->nd->vhost_device;
+ if (n->vhost_device) {
+ int r = vhost_net_init(&n->vhost, dev->nd->vlan->first_client);
+ if (r) {
+ fprintf(stderr, "Unable to initialize vhost device: %d\n", r);
+ virtio_cleanup(&n->vdev);
+ return NULL;
+ }
+ }
n->vdev.get_config = virtio_net_get_config;
n->vdev.set_config = virtio_net_set_config;
@@ -838,6 +883,7 @@ VirtIODevice *virtio_net_init(DeviceState *dev)
n->vdev.set_features = virtio_net_set_features;
n->vdev.bad_features = virtio_net_bad_features;
n->vdev.reset = virtio_net_reset;
+ n->vdev.driver_ok = virtio_net_driver_ok;
n->rx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_rx);
n->tx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_tx);
n->ctrl_vq = virtio_add_queue(&n->vdev, 64, virtio_net_handle_ctrl);
@@ -864,7 +910,6 @@ VirtIODevice *virtio_net_init(DeviceState *dev)
n->vdev.nvectors = 3;
else
n->vdev.nvectors = dev->nd->nvectors;
-
register_savevm("virtio-net", virtio_net_id++, VIRTIO_NET_VM_VERSION,
virtio_net_save, virtio_net_load, n);
diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index 0716f6f..b7f073b 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -15,11 +15,13 @@
#include <inttypes.h>
+#include <linux/kvm.h>
#include "virtio.h"
#include "pci.h"
#include "sysemu.h"
#include "msix.h"
#include "net.h"
+#include "qemu-kvm.h"
/* from Linux's linux/virtio_pci.h */
@@ -199,6 +201,8 @@ static void virtio_ioport_write(void *opaque, uint32_t addr, uint32_t val)
vdev->status = val & 0xFF;
if (vdev->status == 0)
virtio_pci_reset(&proxy->pci_dev.qdev);
+ if ((val & VIRTIO_CONFIG_S_DRIVER_OK) && vdev->driver_ok)
+ vdev->driver_ok(vdev);
break;
case VIRTIO_MSI_CONFIG_VECTOR:
msix_vector_unuse(&proxy->pci_dev, vdev->config_vector);
@@ -373,12 +377,48 @@ static void virtio_write_config(PCIDevice *pci_dev, uint32_t address,
msix_write_config(pci_dev, address, val, len);
}
+static int virtio_pci_irqfd(void * opaque, uint16_t vector, int fd)
+{
+ VirtIOPCIProxy *proxy = opaque;
+ struct kvm_irqfd call = { };
+ int r;
+
+ if (vector >= proxy->pci_dev.msix_entries_nr)
+ return -EINVAL;
+ if (!proxy->pci_dev.msix_entry_used[vector])
+ return -ENOENT;
+ call.fd = fd;
+ call.gsi = proxy->pci_dev.msix_irq_entries[vector].gsi;
+ r = kvm_vm_ioctl(kvm_state, KVM_IRQFD, &call);
+ if (r < 0)
+ return r;
+ return 0;
+}
+
+static int virtio_pci_queuefd(void * opaque, int n, int fd)
+{
+ VirtIOPCIProxy *proxy = opaque;
+ struct kvm_ioeventfd kick = {
+ .datamatch = n,
+ .addr = proxy->addr + VIRTIO_PCI_QUEUE_NOTIFY,
+ .len = 2,
+ .flags = KVM_IOEVENTFD_FLAG_DATAMATCH | KVM_IOEVENTFD_FLAG_PIO,
+ .fd = fd,
+ };
+ int r = kvm_vm_ioctl(kvm_state, KVM_IOEVENTFD, &kick);
+ if (r < 0)
+ return r;
+ return 0;
+}
+
static const VirtIOBindings virtio_pci_bindings = {
.notify = virtio_pci_notify,
.save_config = virtio_pci_save_config,
.load_config = virtio_pci_load_config,
.save_queue = virtio_pci_save_queue,
.load_queue = virtio_pci_load_queue,
+ .irqfd = virtio_pci_irqfd,
+ .queuefd = virtio_pci_queuefd,
};
static void virtio_init_pci(VirtIOPCIProxy *proxy, VirtIODevice *vdev,
diff --git a/hw/virtio.c b/hw/virtio.c
index 337ff27..cc5c205 100644
--- a/hw/virtio.c
+++ b/hw/virtio.c
@@ -54,24 +54,6 @@ typedef struct VRingUsed
VRingUsedElem ring[0];
} VRingUsed;
-typedef struct VRing
-{
- unsigned int num;
- target_phys_addr_t desc;
- target_phys_addr_t avail;
- target_phys_addr_t used;
-} VRing;
-
-struct VirtQueue
-{
- VRing vring;
- target_phys_addr_t pa;
- uint16_t last_avail_idx;
- int inuse;
- uint16_t vector;
- void (*handle_output)(VirtIODevice *vdev, VirtQueue *vq);
-};
-
#define VIRTIO_PCI_QUEUE_MAX 16
/* virt queue functions */
@@ -401,7 +383,6 @@ int virtqueue_pop(VirtQueue *vq, VirtQueueElement *elem)
sg->iov_base = cpu_physical_memory_map(vring_desc_addr(desc_pa, i),
&len, is_write);
-
if (sg->iov_base == NULL || len != sg->iov_len) {
fprintf(stderr, "virtio: trying to map MMIO memory\n");
exit(1);
diff --git a/hw/virtio.h b/hw/virtio.h
index 799e608..12792da 100644
--- a/hw/virtio.h
+++ b/hw/virtio.h
@@ -54,15 +54,34 @@
struct VirtQueue;
+typedef struct VRing
+{
+ unsigned int num;
+ target_phys_addr_t desc;
+ target_phys_addr_t avail;
+ target_phys_addr_t used;
+} VRing;
+
+typedef struct VirtQueue VirtQueue;
+struct VirtIODevice;
+typedef struct VirtIODevice VirtIODevice;
+
+struct VirtQueue
+{
+ VRing vring;
+ target_phys_addr_t pa;
+ uint16_t last_avail_idx;
+ int inuse;
+ uint16_t vector;
+ void (*handle_output)(VirtIODevice *vdev, VirtQueue *vq);
+};
+
static inline target_phys_addr_t vring_align(target_phys_addr_t addr,
unsigned long align)
{
return (addr + align - 1) & ~(align - 1);
}
-typedef struct VirtQueue VirtQueue;
-typedef struct VirtIODevice VirtIODevice;
-
#define VIRTQUEUE_MAX_SIZE 1024
typedef struct VirtQueueElement
@@ -81,6 +100,8 @@ typedef struct {
void (*save_queue)(void * opaque, int n, QEMUFile *f);
int (*load_config)(void * opaque, QEMUFile *f);
int (*load_queue)(void * opaque, int n, QEMUFile *f);
+ int (*irqfd)(void * opaque, uint16_t vector, int fd);
+ int (*queuefd)(void * opaque, int n, int fd);
} VirtIOBindings;
#define VIRTIO_PCI_QUEUE_MAX 16
@@ -104,6 +125,7 @@ struct VirtIODevice
void (*get_config)(VirtIODevice *vdev, uint8_t *config);
void (*set_config)(VirtIODevice *vdev, const uint8_t *config);
void (*reset)(VirtIODevice *vdev);
+ void (*driver_ok)(VirtIODevice *vdev);
VirtQueue *vq;
const VirtIOBindings *binding;
void *binding_opaque;
diff --git a/kvm/include/linux/vhost.h b/kvm/include/linux/vhost.h
new file mode 100644
index 0000000..aa4ff24
--- /dev/null
+++ b/kvm/include/linux/vhost.h
@@ -0,0 +1,126 @@
+#ifndef _LINUX_VHOST_H
+#define _LINUX_VHOST_H
+/* Userspace interface for in-kernel virtio accelerators. */
+
+/* vhost is used to reduce the number of system calls involved in virtio.
+ *
+ * Existing virtio net code is used in the guest without modification.
+ *
+ * This header includes interface used by userspace hypervisor for
+ * device configuration.
+ */
+
+#include <linux/types.h>
+
+#include <linux/ioctl.h>
+#include <linux/virtio_config.h>
+#include <linux/virtio_ring.h>
+
+struct vhost_vring_state {
+ unsigned int index;
+ unsigned int num;
+};
+
+struct vhost_vring_file {
+ unsigned int index;
+ int fd; /* Pass -1 to unbind from file. */
+
+};
+
+struct vhost_vring_addr {
+ unsigned int index;
+ unsigned int padding;
+ __u64 user_addr;
+};
+
+struct vhost_memory_region {
+ __u64 guest_phys_addr;
+ __u64 memory_size; /* bytes */
+ __u64 userspace_addr;
+ __u64 flags_padding; /* No flags are currently specified. */
+};
+
+/* All region addresses and sizes must be 4K aligned. */
+#define VHOST_PAGE_SIZE 0x1000
+
+struct vhost_memory {
+ __u32 nregions;
+ __u32 padding;
+ struct vhost_memory_region regions[0];
+};
+
+/* ioctls */
+
+#define VHOST_VIRTIO 0xAF
+
+/* Features bitmask for forward compatibility. Transport bits are used for
+ * vhost specific features. */
+#define VHOST_GET_FEATURES _IOR(VHOST_VIRTIO, 0x00, __u64)
+#define VHOST_ACK_FEATURES _IOW(VHOST_VIRTIO, 0x00, __u64)
+
+/* Set current process as the (exclusive) owner of this file descriptor. This
+ * must be called before any other vhost command. Further calls to
+ * VHOST_OWNER_SET fail until VHOST_OWNER_RESET is called. */
+#define VHOST_SET_OWNER _IO(VHOST_VIRTIO, 0x01)
+/* Give up ownership, and reset the device to default values.
+ * Allows subsequent call to VHOST_OWNER_SET to succeed. */
+#define VHOST_RESET_OWNER _IO(VHOST_VIRTIO, 0x02)
+
+/* Set up/modify memory layout */
+#define VHOST_SET_MEM_TABLE _IOW(VHOST_VIRTIO, 0x03, struct vhost_memory)
+
+/* Write logging setup. */
+/* Memory writes can optionally be logged by setting bit at an offset
+ * (calculated from the physical address) from specified log base.
+ * The bit is set using an atomic 32 bit operation. */
+/* Set base address for logging. */
+#define VHOST_SET_LOG_BASE _IOW(VHOST_VIRTIO, 0x04, __u64)
+/* Specify an eventfd file descriptor to signal on log write. */
+#define VHOST_SET_LOG_FD _IOW(VHOST_VIRTIO, 0x07, int)
+
+/* Ring setup. These parameters can not be modified while ring is running
+ * (bound to a device). */
+/* Set number of descriptors in ring */
+#define VHOST_SET_VRING_NUM _IOW(VHOST_VIRTIO, 0x10, struct vhost_vring_state)
+/* Start of array of descriptors (virtually contiguous) */
+#define VHOST_SET_VRING_DESC _IOW(VHOST_VIRTIO, 0x11, struct vhost_vring_addr)
+/* Used structure address. Must be 32 bit aligned */
+#define VHOST_SET_VRING_USED _IOW(VHOST_VIRTIO, 0x12, struct vhost_vring_addr)
+/* Available structure address. Must be 16 bit aligned */
+#define VHOST_SET_VRING_AVAIL _IOW(VHOST_VIRTIO, 0x13, struct vhost_vring_addr)
+/* Base value where queue looks for available descriptors */
+#define VHOST_SET_VRING_BASE _IOW(VHOST_VIRTIO, 0x14, struct vhost_vring_state)
+/* Get accessor: reads index, writes value in num */
+#define VHOST_GET_VRING_BASE _IOWR(VHOST_VIRTIO, 0x14, struct vhost_vring_state)
+
+/* Logging support. Can be modified while ring is running. */
+/* Log writes to used structure, at offset calculated from specified address.
+ * Address must be 32 bit aligned. Pass 0x1 to disable logging. */
+#define VHOST_SET_VRING_LOG _IOW(VHOST_VIRTIO, 0x18, struct vhost_vring_addr)
+#define VHOST_VRING_LOG_DISABLE (0x1)
+
+/* The following ioctls use eventfd file descriptors to signal and poll
+ * for events. */
+
+/* Set eventfd to poll for added buffers */
+#define VHOST_SET_VRING_KICK _IOW(VHOST_VIRTIO, 0x20, struct vhost_vring_file)
+/* Set eventfd to signal when buffers have beed used */
+#define VHOST_SET_VRING_CALL _IOW(VHOST_VIRTIO, 0x21, struct vhost_vring_file)
+/* Set eventfd to signal an error */
+#define VHOST_SET_VRING_ERR _IOW(VHOST_VIRTIO, 0x22, struct vhost_vring_file)
+
+/* VHOST_NET specific defines */
+
+/* Attach virtio net ring to a raw socket, or tap device.
+ * The socket must be already bound to an ethernet device, this device will be
+ * used for transmit. Pass fd -1 to unbind from the socket and the transmit
+ * device. This can be used to stop the ring (e.g. for migration). */
+#define VHOST_NET_SET_BACKEND _IOW(VHOST_VIRTIO, 0x30, struct vhost_vring_file)
+
+/* Feature bits */
+/* Log all write descriptors. Can be changed while device is active. */
+#define VHOST_F_LOG_ALL 26
+/* vhost-net should add virtio_net_hdr for RX, and strip for TX packets. */
+#define VHOST_NET_F_VIRTIO_NET_HDR 27
+
+#endif
diff --git a/net.c b/net.c
index 9168460..5d98e90 100644
--- a/net.c
+++ b/net.c
@@ -2767,6 +2767,9 @@ static int net_init_nic(QemuOpts *opts, Monitor *mon)
if (qemu_opt_get(opts, "addr")) {
nd->devaddr = qemu_strdup(qemu_opt_get(opts, "addr"));
}
+ if (qemu_opt_get(opts, "vhost")) {
+ nd->vhost_device = qemu_opt_get_bool(opts, "vhost", 0);
+ }
nd->macaddr[0] = 0x52;
nd->macaddr[1] = 0x54;
@@ -3182,6 +3185,10 @@ static struct {
.name = "vectors",
.type = QEMU_OPT_NUMBER,
.help = "number of MSI-x vectors, 0 to disable MSI-X",
+ }, {
+ .name = "vhost",
+ .type = QEMU_OPT_BOOL,
+ .help = "enable vhost backend",
},
{ /* end of list */ }
},
diff --git a/net.h b/net.h
index 932b50d..adcd5c6 100644
--- a/net.h
+++ b/net.h
@@ -115,6 +115,7 @@ struct NICInfo {
int used;
int bootable;
int nvectors;
+ int vhost_device;
};
extern int nb_nics;
diff --git a/qemu-kvm.c b/qemu-kvm.c
index 62ca050..a547975 100644
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -150,14 +150,6 @@ static inline void clear_gsi(kvm_context_t kvm, unsigned int gsi)
DPRINTF("Invalid GSI %d\n");
}
-struct slot_info {
- unsigned long phys_addr;
- unsigned long len;
- unsigned long userspace_addr;
- unsigned flags;
- int logging_count;
-};
-
struct slot_info slots[KVM_MAX_NUM_MEM_REGIONS];
static void init_slots(void)
diff --git a/qemu-kvm.h b/qemu-kvm.h
index d6748c7..2ab6c33 100644
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -1240,6 +1240,15 @@ int kvm_ioctl(KVMState *s, int type, ...);
int kvm_vm_ioctl(KVMState *s, int type, ...);
int kvm_check_extension(KVMState *s, unsigned int ext);
+struct slot_info {
+ unsigned long phys_addr;
+ unsigned long len;
+ unsigned long userspace_addr;
+ unsigned flags;
+ int logging_count;
+};
+
+extern struct slot_info slots[KVM_MAX_NUM_MEM_REGIONS];
#endif
#endif
--
1.6.5.2.143.g8cc62
^ permalink raw reply related [flat|nested] 11+ messages in thread