2
0
mirror of https://github.com/openvswitch/ovs synced 2025-08-22 01:51:26 +00:00

netdev-dpdk: Add new 'dpdkvhostuserclient' port type

The 'dpdkvhostuser' port type no longer supports both server and client
mode. Instead, 'dpdkvhostuser' ports are always 'server' mode and
'dpdkvhostuserclient' ports are always 'client' mode.

Suggested-by: Daniele Di Proietto <diproiettod@vmware.com>
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
This commit is contained in:
Ciara Loftus 2016-08-19 10:22:30 +01:00 committed by Daniele Di Proietto
parent 70ec021ad9
commit 2d24d165d6
4 changed files with 161 additions and 128 deletions

View File

@ -461,6 +461,21 @@ For users wanting to do packet forwarding using kernel stack below are the steps
```
## <a name="vhost"></a> 6. Vhost Walkthrough
Two types of vHost User ports are available in OVS:
1. vhost-user (dpdkvhostuser ports)
2. vhost-user-client (dpdkvhostuserclient ports)
vHost User uses a client-server model. The server creates/manages/destroys the
vHost User sockets, and the client connects to the server. Depending on which
port type you use, dpdkvhostuser or dpdkvhostuserclient, a different
configuration of the client-server model is used.
For vhost-user ports, OVS DPDK acts as the server and QEMU the client.
For vhost-user-client ports, OVS DPDK acts as the client and QEMU the server.
### 6.1 vhost-user
- Prerequisites:
@ -570,49 +585,6 @@ For users wanting to do packet forwarding using kernel stack below are the steps
where `-L`: Changes the numbers of channels of the specified network device
and `combined`: Changes the number of multi-purpose channels.
4. OVS vHost client-mode & vHost reconnect (OPTIONAL)
By default, OVS DPDK acts as the vHost socket server for dpdkvhostuser
ports and QEMU acts as the vHost client. This means OVS creates and
manages the vHost socket and QEMU is the client which connects to the
vHost server (OVS). In QEMU v2.7 the option is available for QEMU to act
as the vHost server meaning the roles can be reversed and OVS can become
the vHost client. To enable client mode for a given dpdkvhostuserport,
one must specify a valid 'vhost-server-path' like so:
```
ovs-vsctl set Interface dpdkvhostuser0 options:vhost-server-path=/path/to/socket
```
Setting this value automatically switches the port to client mode (from
OVS' perspective). 'vhost-server-path' reflects the full path of the
socket that has been or will be created by QEMU for the given vHost User
port. Once a path is specified, the port will remain in 'client' mode
for the remainder of it's lifetime ie. it cannot be reverted back to
server mode.
One must append ',server' to the 'chardev' arguments on the QEMU command
line, to instruct QEMU to use vHost server mode for a given interface,
like so:
````
-chardev socket,id=char0,path=/path/to/socket,server
````
If the corresponding dpdkvhostuser port has not yet been configured in
OVS with vhost-server-path=/path/to/socket, QEMU will print a log
similar to the following:
`QEMU waiting for connection on: disconnected:unix:/path/to/socket,server`
QEMU will wait until the port is created sucessfully in OVS to boot the
VM.
One benefit of using this mode is the ability for vHost ports to
'reconnect' in event of the switch crashing or being brought down. Once
it is brought back up, the vHost ports will reconnect automatically and
normal service will resume.
- VM Configuration with libvirt
* change the user/group, access control policty and restart libvirtd.
@ -657,7 +629,49 @@ For users wanting to do packet forwarding using kernel stack below are the steps
Note: For information on libvirt and further tuning refer [libvirt].
### 6.2 DPDK backend inside VM
### 6.2 vhost-user-client
- Prerequisites:
QEMU version >= 2.7
- Adding vhost-user-client ports to Switch
```
ovs-vsctl add-port br0 vhost-client-1 -- set Interface vhost-client-1
type=dpdkvhostuserclient options:vhost-server-path=/path/to/socket
```
Unlike vhost-user ports, the name given to port does not govern the name of
the socket device. 'vhost-server-path' reflects the full path of the socket
that has been or will be created by QEMU for the given vHost User client
port.
- Adding vhost-user-client ports to VM
The same QEMU parameters as vhost-user ports described in section 6.1 can
be used, with one change necessary. One must append ',server' to the
'chardev' arguments on the QEMU command line, to instruct QEMU to use vHost
server mode for a given interface, like so:
````
-chardev socket,id=char0,path=/path/to/socket,server
````
If the corresponding dpdkvhostuserclient port has not yet been configured
in OVS with vhost-server-path=/path/to/socket, QEMU will print a log
similar to the following:
`QEMU waiting for connection on: disconnected:unix:/path/to/socket,server`
QEMU will wait until the port is created sucessfully in OVS to boot the VM.
One benefit of using this mode is the ability for vHost ports to
'reconnect' in event of the switch crashing or being brought down. Once it
is brought back up, the vHost ports will reconnect automatically and normal
service will resume.
### 6.3 DPDK backend inside VM
Please note that additional configuration is required if you want to run
ovs-vswitchd with DPDK backend inside a QEMU virtual machine. Ovs-vswitchd

1
NEWS
View File

@ -124,6 +124,7 @@ v2.6.0 - xx xxx xxxx
* Jumbo frame support
* Remove dpdkvhostcuse port type.
* OVS client mode for vHost and vHost reconnect (Requires QEMU 2.7)
* 'dpdkvhostuserclient' port type.
- Increase number of registers to 16.
- ovs-benchmark: This utility has been removed due to lack of use and
bitrot.

View File

@ -356,9 +356,8 @@ struct netdev_dpdk {
/* True if vHost device is 'up' and has been reconfigured at least once */
bool vhost_reconfigured;
/* Identifiers used to distinguish vhost devices from each other. */
char vhost_server_id[PATH_MAX];
char vhost_client_id[PATH_MAX];
/* Identifier used to distinguish vhost devices from each other. */
char vhost_id[PATH_MAX];
/* In dpdk_list. */
struct ovs_list list_node OVS_GUARDED_BY(dpdk_mutex);
@ -814,8 +813,6 @@ netdev_dpdk_init(struct netdev *netdev, unsigned int port_no,
dev->max_packet_len = MTU_TO_FRAME_LEN(dev->mtu);
ovsrcu_index_init(&dev->vid, -1);
dev->vhost_reconfigured = false;
/* initialise vHost port in server mode */
dev->vhost_driver_flags &= ~RTE_VHOST_USER_CLIENT;
err = netdev_dpdk_mempool_configure(dev);
if (err) {
@ -878,16 +875,6 @@ dpdk_dev_parse_name(const char dev_name[], const char prefix[],
}
}
/* Returns a pointer to the relevant vHost socket ID depending on the mode in
* use */
static char *
get_vhost_id(struct netdev_dpdk *dev)
OVS_REQUIRES(dev->mutex)
{
return dev->vhost_driver_flags & RTE_VHOST_USER_CLIENT ?
dev->vhost_client_id : dev->vhost_server_id;
}
static int
netdev_dpdk_vhost_construct(struct netdev *netdev)
{
@ -911,27 +898,38 @@ netdev_dpdk_vhost_construct(struct netdev *netdev)
ovs_mutex_lock(&dpdk_mutex);
/* Take the name of the vhost-user port and append it to the location where
* the socket is to be created, then register the socket. Sockets are
* registered initially in 'server' mode.
* the socket is to be created, then register the socket.
*/
snprintf(dev->vhost_server_id, sizeof dev->vhost_server_id, "%s/%s",
snprintf(dev->vhost_id, sizeof dev->vhost_id, "%s/%s",
vhost_sock_dir, name);
err = rte_vhost_driver_register(dev->vhost_server_id,
dev->vhost_driver_flags);
dev->vhost_driver_flags &= ~RTE_VHOST_USER_CLIENT;
err = rte_vhost_driver_register(dev->vhost_id, dev->vhost_driver_flags);
if (err) {
VLOG_ERR("vhost-user socket device setup failure for socket %s\n",
dev->vhost_server_id);
dev->vhost_id);
} else {
if (!(dev->vhost_driver_flags & RTE_VHOST_USER_CLIENT)) {
/* OVS server mode - add this socket to list for deletion */
fatal_signal_add_file_to_unlink(dev->vhost_server_id);
VLOG_INFO("Socket %s created for vhost-user port %s\n",
dev->vhost_server_id, name);
}
err = netdev_dpdk_init(netdev, -1, DPDK_DEV_VHOST);
fatal_signal_add_file_to_unlink(dev->vhost_id);
VLOG_INFO("Socket %s created for vhost-user port %s\n",
dev->vhost_id, name);
}
err = netdev_dpdk_init(netdev, -1, DPDK_DEV_VHOST);
ovs_mutex_unlock(&dpdk_mutex);
return err;
}
static int
netdev_dpdk_vhost_client_construct(struct netdev *netdev)
{
int err;
if (rte_eal_init_ret) {
return rte_eal_init_ret;
}
ovs_mutex_lock(&dpdk_mutex);
err = netdev_dpdk_init(netdev, -1, DPDK_DEV_VHOST);
ovs_mutex_unlock(&dpdk_mutex);
return err;
}
@ -1005,8 +1003,7 @@ netdev_dpdk_vhost_destruct(struct netdev *netdev)
VLOG_ERR("Removing port '%s' while vhost device still attached.",
netdev->name);
VLOG_ERR("To restore connectivity after re-adding of port, VM on socket"
" '%s' must be restarted.",
get_vhost_id(dev));
" '%s' must be restarted.", dev->vhost_id);
}
free(ovsrcu_get_protected(struct ingress_policer *,
@ -1016,7 +1013,7 @@ netdev_dpdk_vhost_destruct(struct netdev *netdev)
ovs_list_remove(&dev->list_node);
dpdk_mp_put(dev->dpdk_mp);
vhost_id = xstrdup(get_vhost_id(dev));
vhost_id = xstrdup(dev->vhost_id);
ovs_mutex_unlock(&dev->mutex);
ovs_mutex_unlock(&dpdk_mutex);
@ -1108,15 +1105,16 @@ netdev_dpdk_ring_set_config(struct netdev *netdev, const struct smap *args)
}
static int
netdev_dpdk_vhost_set_config(struct netdev *netdev, const struct smap *args)
netdev_dpdk_vhost_client_set_config(struct netdev *netdev,
const struct smap *args)
{
struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
const char *path;
if (!(dev->vhost_driver_flags & RTE_VHOST_USER_CLIENT)) {
path = smap_get(args, "vhost-server-path");
if (path && strcmp(path, dev->vhost_client_id)) {
strcpy(dev->vhost_client_id, path);
if (path && strcmp(path, dev->vhost_id)) {
strcpy(dev->vhost_id, path);
netdev_request_reconfigure(netdev);
}
}
@ -2302,7 +2300,7 @@ netdev_dpdk_remap_txqs(struct netdev_dpdk *dev)
}
}
VLOG_DBG("TX queue mapping for %s\n", get_vhost_id(dev));
VLOG_DBG("TX queue mapping for %s\n", dev->vhost_id);
for (i = 0; i < total_txqs; i++) {
VLOG_DBG("%2d --> %2d", i, dev->tx_q[i].map);
}
@ -2327,7 +2325,7 @@ new_device(int vid)
/* Add device to the vhost port with the same name as that passed down. */
LIST_FOR_EACH(dev, list_node, &dpdk_list) {
ovs_mutex_lock(&dev->mutex);
if (strncmp(ifname, get_vhost_id(dev), IF_NAME_SZ) == 0) {
if (strncmp(ifname, dev->vhost_id, IF_NAME_SZ) == 0) {
uint32_t qp_num = rte_vhost_get_queue_num(vid);
/* Get NUMA information */
@ -2456,7 +2454,7 @@ vring_state_changed(int vid, uint16_t queue_id, int enable)
ovs_mutex_lock(&dpdk_mutex);
LIST_FOR_EACH (dev, list_node, &dpdk_list) {
ovs_mutex_lock(&dev->mutex);
if (strncmp(ifname, get_vhost_id(dev), IF_NAME_SZ) == 0) {
if (strncmp(ifname, dev->vhost_id, IF_NAME_SZ) == 0) {
if (enable) {
dev->tx_q[qid].map = qid;
} else {
@ -2949,17 +2947,13 @@ out:
return err;
}
static int
netdev_dpdk_vhost_reconfigure(struct netdev *netdev)
static void
dpdk_vhost_reconfigure_helper(struct netdev_dpdk *dev)
OVS_REQUIRES(dpdk_mutex)
OVS_REQUIRES(dev->mutex)
{
struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
int err = 0;
ovs_mutex_lock(&dpdk_mutex);
ovs_mutex_lock(&dev->mutex);
netdev->n_txq = dev->requested_n_txq;
netdev->n_rxq = dev->requested_n_rxq;
dev->up.n_txq = dev->requested_n_txq;
dev->up.n_rxq = dev->requested_n_rxq;
/* Enable TX queue 0 by default if it wasn't disabled. */
if (dev->tx_q[0].map == OVS_VHOST_QUEUE_MAP_UNKNOWN) {
@ -2971,50 +2965,61 @@ netdev_dpdk_vhost_reconfigure(struct netdev *netdev)
if (dev->requested_socket_id != dev->socket_id
|| dev->requested_mtu != dev->mtu) {
if (!netdev_dpdk_mempool_configure(dev)) {
netdev_change_seq_changed(netdev);
netdev_change_seq_changed(&dev->up);
}
}
if (netdev_dpdk_get_vid(dev) >= 0) {
dev->vhost_reconfigured = true;
}
}
static int
netdev_dpdk_vhost_reconfigure(struct netdev *netdev)
{
struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
ovs_mutex_lock(&dpdk_mutex);
ovs_mutex_lock(&dev->mutex);
dpdk_vhost_reconfigure_helper(dev);
ovs_mutex_unlock(&dev->mutex);
ovs_mutex_unlock(&dpdk_mutex);
return 0;
}
static int
netdev_dpdk_vhost_client_reconfigure(struct netdev *netdev)
{
struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
int err = 0;
ovs_mutex_lock(&dpdk_mutex);
ovs_mutex_lock(&dev->mutex);
dpdk_vhost_reconfigure_helper(dev);
/* Configure vHost client mode if requested and if the following criteria
* are met:
* 1. Device is currently in 'server' mode.
* 2. Device is currently not active.
* 3. A path has been specified.
* 1. Device hasn't been registered yet.
* 2. A path has been specified.
*/
if (!(dev->vhost_driver_flags & RTE_VHOST_USER_CLIENT)
&& !(netdev_dpdk_get_vid(dev) >= 0)
&& strlen(dev->vhost_client_id)) {
/* Unregister server-mode device */
char *vhost_id = xstrdup(get_vhost_id(dev));
ovs_mutex_unlock(&dev->mutex);
ovs_mutex_unlock(&dpdk_mutex);
err = dpdk_vhost_driver_unregister(dev, vhost_id);
free(vhost_id);
ovs_mutex_lock(&dpdk_mutex);
ovs_mutex_lock(&dev->mutex);
&& strlen(dev->vhost_id)) {
/* Register client-mode device */
err = rte_vhost_driver_register(dev->vhost_id,
RTE_VHOST_USER_CLIENT);
if (err) {
VLOG_ERR("Unable to remove vhost-user socket %s",
get_vhost_id(dev));
VLOG_ERR("vhost-user device setup failure for device %s\n",
dev->vhost_id);
} else {
fatal_signal_remove_file_to_unlink(get_vhost_id(dev));
/* Register client-mode device */
err = rte_vhost_driver_register(dev->vhost_client_id,
RTE_VHOST_USER_CLIENT);
if (err) {
VLOG_ERR("vhost-user device setup failure for device %s\n",
dev->vhost_client_id);
} else {
/* Configuration successful */
dev->vhost_driver_flags |= RTE_VHOST_USER_CLIENT;
VLOG_INFO("vHost User device '%s' changed to 'client' mode, "
"using client socket '%s'",
dev->up.name, get_vhost_id(dev));
}
/* Configuration successful */
dev->vhost_driver_flags |= RTE_VHOST_USER_CLIENT;
VLOG_INFO("vHost User device '%s' created in 'client' mode, "
"using client socket '%s'",
dev->up.name, dev->vhost_id);
}
}
@ -3521,7 +3526,7 @@ static const struct netdev_class dpdk_vhost_class =
"dpdkvhostuser",
netdev_dpdk_vhost_construct,
netdev_dpdk_vhost_destruct,
netdev_dpdk_vhost_set_config,
NULL,
NULL,
netdev_dpdk_vhost_send,
netdev_dpdk_vhost_get_carrier,
@ -3530,6 +3535,20 @@ static const struct netdev_class dpdk_vhost_class =
NULL,
netdev_dpdk_vhost_reconfigure,
netdev_dpdk_vhost_rxq_recv);
static const struct netdev_class dpdk_vhost_client_class =
NETDEV_DPDK_CLASS(
"dpdkvhostuserclient",
netdev_dpdk_vhost_client_construct,
netdev_dpdk_vhost_destruct,
netdev_dpdk_vhost_client_set_config,
NULL,
netdev_dpdk_vhost_send,
netdev_dpdk_vhost_get_carrier,
netdev_dpdk_vhost_get_stats,
NULL,
NULL,
netdev_dpdk_vhost_client_reconfigure,
netdev_dpdk_vhost_rxq_recv);
void
netdev_dpdk_register(void)
@ -3538,6 +3557,7 @@ netdev_dpdk_register(void)
netdev_register_provider(&dpdk_class);
netdev_register_provider(&dpdk_ring_class);
netdev_register_provider(&dpdk_vhost_class);
netdev_register_provider(&dpdk_vhost_client_class);
}
void

View File

@ -2370,11 +2370,9 @@
<column name="options" key="vhost-server-path"
type='{"type": "string"}'>
<p>
When specified, switches the given port permanently to 'client'
mode. The value specifies the path to the socket associated with a
vHost User client mode device that has been or will be created by
QEMU.
Only supported by DPDK vHost interfaces.
The value specifies the path to the socket associated with a vHost
User client mode device that has been or will be created by QEMU.
Only supported by dpdkvhostuserclient interfaces.
</p>
</column>
</group>