[ovs-dev] [PATCH] netdev-dpdk: Add vHost User PMD
Loftus, Ciara
ciara.loftus at intel.com
Tue May 10 09:22:21 UTC 2016
> On 21/04/2016 13:20, Ciara Loftus wrote:
> > DPDK 16.04 introduces the vHost PMD which allows 'dpdkvhostuser' ports
> > to be controlled by the librte_ether API, like physical 'dpdk' ports.
> > The commit integrates this functionality into OVS, and refactors some
> > of the existing vhost code such that it is vhost-cuse specific.
> > Similarly, there is now some overlap between dpdk and vhost-user port
> > code.
> >
> > Signed-off-by: Ciara Loftus <ciara.loftus at intel.com>
> > ---
> > INSTALL.DPDK.md | 12 ++
> > NEWS | 2 +
> > lib/netdev-dpdk.c | 515 +++++++++++++++++++++++++---------------------
> --------
>
> Hi Ciara, there's a lot of churn in this file. It might be worth
> considering to see if it could be split through a few commits commits to
> help reviewers. e.g. new features like adding get_features, get_status
> for vhost could be a separate patch at least.
I've split into 3:
- remove watchdog
- add pmd
- add get_stats & get_features
Couldn't quite find a way to split it up more.
>
> > 3 files changed, 254 insertions(+), 275 deletions(-)
> > mode change 100644 => 100755 lib/netdev-dpdk.c
>
> file permission change.
Woops. Fixed in v2.
>
> >
> > diff --git a/INSTALL.DPDK.md b/INSTALL.DPDK.md
> > index 7f76df8..5006812 100644
> > --- a/INSTALL.DPDK.md
> > +++ b/INSTALL.DPDK.md
> > @@ -945,6 +945,18 @@ Restrictions:
> > increased to the desired number of queues. Both DPDK and OVS must
> be
> > recompiled for this change to take effect.
> >
> > + DPDK 'eth' type ports:
> > + - dpdk, dpdkr and dpdkvhostuser ports are 'eth' type ports in the context
> of
> > + DPDK as they are all managed by the rte_ether API. This means that
> they
> > + adhere to the DPDK configuration option CONFIG_RTE_MAX_ETHPORTS
> which by
> > + default is set to 32. This means by default the combined total number of
> > + dpdk, dpdkr and dpdkvhostuser ports allowable in OVS with DPDK is 32.
> This
> > + value can be changed if desired by modifying the configuration file in
> > + DPDK, or by overriding the default value on the command line when
> building
> > + DPDK. eg.
> > +
> > + `make install CONFIG_RTE_MAX_ETHPORTS=64`
>
> format is not registering right for this in my md viewer.
It's looking ok on mine. What doesn't look right?
>
> > +
> > Bug Reporting:
> > --------------
> >
> > diff --git a/NEWS b/NEWS
> > index ea7f3a1..4dc0201 100644
> > --- a/NEWS
> > +++ b/NEWS
> > @@ -26,6 +26,8 @@ Post-v2.5.0
> > assignment.
> > * Type of log messages from PMD threads changed from INFO to DBG.
> > * QoS functionality with sample egress-policer implementation.
> > + * vHost PMD integration brings vhost-user ports under control of the
> > + rte_ether DPDK API.
> > - ovs-benchmark: This utility has been removed due to lack of use and
> > bitrot.
> > - ovs-appctl:
> > diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
> > old mode 100644
> > new mode 100755
> > index 208c5f5..4fccd63
> > --- a/lib/netdev-dpdk.c
> > +++ b/lib/netdev-dpdk.c
> > @@ -56,6 +56,7 @@
> > #include "rte_mbuf.h"
> > #include "rte_meter.h"
> > #include "rte_virtio_net.h"
> > +#include "rte_eth_vhost.h"
>
> nit: generally these go in alphabetical order.
Ok
>
> >
> > VLOG_DEFINE_THIS_MODULE(dpdk);
> > static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
> > @@ -109,6 +110,8 @@ BUILD_ASSERT_DECL((MAX_NB_MBUF /
> ROUND_DOWN_POW2(MAX_NB_MBUF/MIN_NB_MBUF))
> >
> > static char *cuse_dev_name = NULL; /* Character device
> cuse_dev_name. */
> > static char *vhost_sock_dir = NULL; /* Location of vhost-user sockets */
> > +/* Array that tracks the used & unused vHost user driver IDs */
> > +static unsigned int vhost_user_drv_ids[RTE_MAX_ETHPORTS];
> >
> > /*
> > * Maximum amount of time in micro seconds to try and enqueue to
> vhost.
> > @@ -143,7 +146,8 @@ enum { DRAIN_TSC = 200000ULL };
> >
> > enum dpdk_dev_type {
> > DPDK_DEV_ETH = 0,
> > - DPDK_DEV_VHOST = 1,
> > + DPDK_DEV_VHOST_USER = 1,
> > + DPDK_DEV_VHOST_CUSE = 2,
> > };
> >
> > static int rte_eal_init_ret = ENODEV;
> > @@ -275,8 +279,6 @@ struct dpdk_tx_queue {
> > * from concurrent access. It is used only
> > * if the queue is shared among different
> > * pmd threads (see 'txq_needs_locking'). */
> > - int map; /* Mapping of configured vhost-user queues
> > - * to enabled by guest. */
> > uint64_t tsc;
> > struct rte_mbuf *burst_pkts[MAX_TX_QUEUE_LEN];
> > };
> > @@ -329,12 +331,22 @@ struct netdev_dpdk {
> > int real_n_rxq;
> > bool txq_needs_locking;
> >
> > - /* virtio-net structure for vhost device */
> > + /* Spinlock for vhost cuse transmission. Other DPDK devices use
> spinlocks
> > + * in dpdk_tx_queue */
> > + rte_spinlock_t vhost_cuse_tx_lock;
>
> Why can't we continue to use the lock in dpdk_tx_queue? rather than
> adding a cuse specific lock.
This logic is taken from netdev-dpdk pre-multiqueue. I assumed it was the safest way to go since it's tried and tested.
>
> > +
> > + /* virtio-net structure for vhost cuse device */
> > OVSRCU_TYPE(struct virtio_net *) virtio_dev;
> >
> > /* Identifier used to distinguish vhost devices from each other */
> > char vhost_id[PATH_MAX];
> >
> > + /* ID of vhost user port given to the PMD driver */
> > + unsigned int vhost_pmd_id;
> > +
> > + /* Number of virtqueue pairs reported by the guest */
> > + uint32_t reported_queues;
>
> Is this useful? we could just use the real_n_*xq's directly.
I had a look at doing that, but it has some undesired knock-on effects. We need to initialise the var that tracks virtqueue numbers to zero but the above are inited to 1 and used by the phy ports. Think it's safer to maintain a separate variable. I've renamed this vhost_qp_nb to make it a bit clearer.
>
> > +
> > /* In dpdk_list. */
> > struct ovs_list list_node OVS_GUARDED_BY(dpdk_mutex);
> >
> > @@ -352,13 +364,20 @@ struct netdev_rxq_dpdk {
> > static bool dpdk_thread_is_pmd(void);
> >
> > static int netdev_dpdk_construct(struct netdev *);
> > +static int netdev_dpdk_vhost_user_construct(struct netdev *);
> >
> > struct virtio_net * netdev_dpdk_get_virtio(const struct netdev_dpdk
> *dev);
> >
> > +void vring_state_changed_callback(uint8_t port_id,
> > + enum rte_eth_event_type type OVS_UNUSED, void *param
> OVS_UNUSED);
> > +void device_state_changed_callback(uint8_t port_id,
> > + enum rte_eth_event_type type OVS_UNUSED, void *param
> OVS_UNUSED);
> > +
> > static bool
> > is_dpdk_class(const struct netdev_class *class)
> > {
> > - return class->construct == netdev_dpdk_construct;
> > + return ((class->construct == netdev_dpdk_construct) ||
> > + (class->construct == netdev_dpdk_vhost_user_construct));
>
> You should probably change the name of the function to something more
> intuitive now that it's not just for a "dpdk" netdev class.
Done
>
> > }
> >
> > /* DPDK NIC drivers allocate RX buffers at a particular granularity, typically
> > @@ -581,7 +600,9 @@ dpdk_eth_dev_queue_setup(struct netdev_dpdk
> *dev, int n_rxq, int n_txq)
> > }
> >
> > dev->up.n_rxq = n_rxq;
> > - dev->real_n_txq = n_txq;
> > + if (dev->type == DPDK_DEV_ETH) {
> > + dev->real_n_txq = n_txq;
> > + }
>
> This needs a comment to explain why you've added this check.
Done
>
> >
> > return 0;
> > }
> > @@ -672,11 +693,72 @@ netdev_dpdk_alloc_txq(struct netdev_dpdk
> *dev, unsigned int n_txqs)
> > /* Queues are shared among CPUs. Always flush */
> > dev->tx_q[i].flush_tx = true;
> > }
> > + }
> > +}
> >
> > - /* Initialize map for vhost devices. */
> > - dev->tx_q[i].map = OVS_VHOST_QUEUE_MAP_UNKNOWN;
> > - rte_spinlock_init(&dev->tx_q[i].tx_lock);
> > +void
> > +vring_state_changed_callback(uint8_t port_id,
> > + enum rte_eth_event_type type OVS_UNUSED,
> > + void *param OVS_UNUSED)
> > +{
> > + struct netdev_dpdk *dev;
> > + struct rte_eth_vhost_queue_event event;
> > + int err = 0;
> > +
> > + err = rte_eth_vhost_get_queue_event(port_id, &event);
> > + if (err || (event.rx == 1)) {
> > + return;
> > }
> > +
> > + ovs_mutex_lock(&dpdk_mutex);
> > + LIST_FOR_EACH (dev, list_node, &dpdk_list) {
> > + if (port_id == dev->port_id) {
> > + ovs_mutex_lock(&dev->mutex);
> > + if (event.enable) {
> > + dev->reported_queues++;
> > + } else {
> > + dev->reported_queues--;
> > + }
> > + dev->real_n_rxq = dev->reported_queues;
> > + dev->real_n_txq = dev->reported_queues;
> > + dev->txq_needs_locking = true;
>
> How do you know locking is needed at this point? Oh I just realised this
> is in the existing code. Seems like we should put in a check against
> netdev->n_txq like elsewhere in the code.
Done
>
> > + netdev_dpdk_alloc_txq(dev, dev->real_n_txq);
> > + ovs_mutex_unlock(&dev->mutex);
> > + break;
> > + }
> > + }
> > + ovs_mutex_unlock(&dpdk_mutex);
> > +
> > + return;
> > +}
> > +
> > +void
> > +device_state_changed_callback(uint8_t port_id,
> > + enum rte_eth_event_type type OVS_UNUSED,
> > + void *param OVS_UNUSED)
> > +{
> > + struct netdev_dpdk *dev;
> > +
> > + ovs_mutex_lock(&dpdk_mutex);
> > + LIST_FOR_EACH (dev, list_node, &dpdk_list) {
> > + if (port_id == dev->port_id) {
> > + ovs_mutex_lock(&dev->mutex);
> > + check_link_status(dev);
> > + if (dev->link.link_status == ETH_LINK_UP) {
> > + /* new device */
> > + VLOG_INFO("vHost Device '%s' has been added", dev->vhost_id);
> > +
> > + } else {
> > + /* destroy device */
> > + VLOG_INFO("vHost Device '%s' has been removed", dev-
> >vhost_id);
> > + }
> > + ovs_mutex_unlock(&dev->mutex);
> > + break;
> > + }
> > + }
> > + ovs_mutex_unlock(&dpdk_mutex);
> > +
> > + return;
> > }
> >
> > static int
> > @@ -688,6 +770,7 @@ netdev_dpdk_init(struct netdev *netdev, unsigned
> int port_no,
> > int sid;
> > int err = 0;
> > uint32_t buf_size;
> > + unsigned int n_queues = 0;
> >
> > ovs_mutex_init(&dev->mutex);
> > ovs_mutex_lock(&dev->mutex);
> > @@ -697,7 +780,7 @@ netdev_dpdk_init(struct netdev *netdev, unsigned
> int port_no,
> > /* If the 'sid' is negative, it means that the kernel fails
> > * to obtain the pci numa info. In that situation, always
> > * use 'SOCKET0'. */
> > - if (type == DPDK_DEV_ETH) {
> > + if (type != DPDK_DEV_VHOST_CUSE) {
> > sid = rte_eth_dev_socket_id(port_no);
>
> Are you trying to sneak in your NUMA changes for vhostuser ;-)
:-) Stay tuned!
>
> > } else {
> > sid = rte_lcore_to_socket_id(rte_get_master_lcore());
> > @@ -709,6 +792,8 @@ netdev_dpdk_init(struct netdev *netdev, unsigned
> int port_no,
> > dev->flags = 0;
> > dev->mtu = ETHER_MTU;
> > dev->max_packet_len = MTU_TO_FRAME_LEN(dev->mtu);
> > + dev->vhost_pmd_id = 0;
>
> 0 is a valid vhost_pmd_id number. Seen as it's used in the destruct it
> might be better to set to an invalid value like MAX and check for that
> in destruct, otherwise you're relying on higher layer to ensure that
> destruct is never called before construct. edit: actually i've another
> comment wrt vhost_pmd_id below.
>
> > + dev->reported_queues = 0;
> >
> > buf_size = dpdk_buf_size(dev->mtu);
> > dev->dpdk_mp = dpdk_mp_get(dev->socket_id,
> FRAME_LEN_TO_MTU(buf_size));
> > @@ -726,14 +811,23 @@ netdev_dpdk_init(struct netdev *netdev,
> unsigned int port_no,
> > netdev->requested_n_rxq = NR_QUEUE;
> > dev->real_n_txq = NR_QUEUE;
> >
> > - if (type == DPDK_DEV_ETH) {
> > - netdev_dpdk_alloc_txq(dev, NR_QUEUE);
> > + n_queues = (type == DPDK_DEV_VHOST_CUSE) ?
> OVS_VHOST_MAX_QUEUE_NUM :
> > + NR_QUEUE;
>
> No need for special cuse case above.
Ok
>
> > + netdev_dpdk_alloc_txq(dev, n_queues);
> > + if (type != DPDK_DEV_VHOST_CUSE) {
> > err = dpdk_eth_dev_init(dev);
> > if (err) {
> > goto unlock;
> > }
> > - } else {
> > - netdev_dpdk_alloc_txq(dev, OVS_VHOST_MAX_QUEUE_NUM);
> > + }
> > +
> > + if (type == DPDK_DEV_VHOST_USER) {
> > + rte_eth_dev_callback_register(port_no,
> RTE_ETH_EVENT_QUEUE_STATE,
> > + (void*)vring_state_changed_callback,
> > + NULL);
> > + rte_eth_dev_callback_register(port_no, RTE_ETH_EVENT_INTR_LSC,
> > + (void*)device_state_changed_callback,
>
> Seems like this using this callback is a different way to do the same
> thing as the watchdog timer - could we consolidate them and simplify?
Good suggestion. This is part one of the v2 patchset.
>
> > + NULL);
> > }
> >
> > ovs_list_push_back(&dpdk_list, &dev->list_node);
> > @@ -768,26 +862,37 @@ dpdk_dev_parse_name(const char dev_name[],
> const char prefix[],
> > }
> >
> > static int
> > -vhost_construct_helper(struct netdev *netdev)
> OVS_REQUIRES(dpdk_mutex)
> > +netdev_dpdk_vhost_cuse_construct(struct netdev *netdev)
> > {
> > + struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > + int err;
> > +
> > + ovs_mutex_lock(&dpdk_mutex);
> > + strncpy(dev->vhost_id, netdev->name, sizeof(dev->vhost_id));
> > if (rte_eal_init_ret) {
> > return rte_eal_init_ret;
> > }
> >
> > - return netdev_dpdk_init(netdev, -1, DPDK_DEV_VHOST);
> > + rte_spinlock_init(&dev->vhost_cuse_tx_lock);
> > +
> > + err = netdev_dpdk_init(netdev, -1, DPDK_DEV_VHOST_CUSE);
> > +
> > + ovs_mutex_unlock(&dpdk_mutex);
> > + return err;
> > }
> >
> > static int
> > -netdev_dpdk_vhost_cuse_construct(struct netdev *netdev)
> > +get_vhost_user_drv_id(void)
> > {
> > - struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > - int err;
> > + int i = 0;
> >
> > - ovs_mutex_lock(&dpdk_mutex);
> > - strncpy(dev->vhost_id, netdev->name, sizeof(dev->vhost_id));
> > - err = vhost_construct_helper(netdev);
> > - ovs_mutex_unlock(&dpdk_mutex);
> > - return err;
> > + for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
> > + if (vhost_user_drv_ids[i] == 0) {
>
> This is treating RTE_MAX_ETHPORTS as the max vhost ports. I didn't check
> but it sounds like it's for all DPDK ports and not just vhost. If true
> it means the code would allow >RTE_MAX_ETHPORTS.
It's for all DPDK ports under control of the ether API yes.
The theoretical max vhost ports == RTE_MAX_ETHPORTS though.
If we try register too many vhosts we will fail anyway on rte_eth_dev_attach.
>
> > + return i;
> > + }
> > + }
> > +
> > + return -1;
> > }
> >
> > static int
> > @@ -796,6 +901,9 @@ netdev_dpdk_vhost_user_construct(struct netdev
> *netdev)
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > const char *name = netdev->name;
> > int err;
> > + uint8_t port_no = 0;
> > + char devargs[4096];
>
> magic number - any define we can use?
It's not pretty but I tried to tidy it into a #define:
#define DEVARGS_MAX (RTE_ETH_NAME_MAX_LEN + PATH_MAX + 18)
>
> > + int driver_id = 0;
> >
> > /* 'name' is appended to 'vhost_sock_dir' and used to create a socket in
> > * the file system. '/' or '\' would traverse directories, so they're not
> > @@ -807,22 +915,35 @@ netdev_dpdk_vhost_user_construct(struct
> netdev *netdev)
> > return EINVAL;
> > }
> >
> > + if (rte_eal_init_ret) {
> > + return rte_eal_init_ret;
> > + }
> > +
>
> it's more consistent to check this before the name.
Ok
>
> > 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.
> > */
> > snprintf(dev->vhost_id, sizeof(dev->vhost_id), "%s/%s",
> > vhost_sock_dir, name);
> > + driver_id = get_vhost_user_drv_id();
> > + if (driver_id == -1) {
> > + VLOG_ERR("Too many vhost-user devices registered with PMD");
> > + err = ENODEV;
> > + } else {
> > + snprintf(devargs, sizeof(devargs),
> "eth_vhost%u,iface=%s,queues=%i",
> > + driver_id, dev->vhost_id, RTE_MAX_QUEUES_PER_PORT);
> > + err = rte_eth_dev_attach(devargs, &port_no);
> > + }
> >
> > - err = rte_vhost_driver_register(dev->vhost_id);
> > if (err) {
> > - VLOG_ERR("vhost-user socket device setup failure for socket %s\n",
> > - dev->vhost_id);
> > + VLOG_ERR("Failed to attach vhost-user device to DPDK");
>
> It might be a bit confusing getting this error message in some cases
> where you have tried to attach and failed and in other cases where it
> has not been tried because of too many vhost-user devices. Also the
> socket name would be handy.
I've tidied this up in the v2.
>
> > } else {
> > 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 = vhost_construct_helper(netdev);
> > + dev->vhost_pmd_id = driver_id;
>
> You're setting this but then it gets reset when you call
> netdev_dpdk_init(). So when you destruct you'll set the reset the wrong
> index in vhost_user_drv_ids[] ?
You're right. The re-init in netdev-dpdk_init is removed in v2.
>
> > + vhost_user_drv_ids[dev->vhost_pmd_id] = 1;
> > + err = netdev_dpdk_init(netdev, port_no, DPDK_DEV_VHOST_USER);
> > }
> >
> > ovs_mutex_unlock(&dpdk_mutex);
> > @@ -857,6 +978,12 @@ netdev_dpdk_destruct(struct netdev *netdev)
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> >
> > ovs_mutex_lock(&dev->mutex);
> > +
> > + if (dev->type == DPDK_DEV_VHOST_USER) {
> > + rte_eth_dev_detach(dev->port_id, dev->vhost_id);
> > + vhost_user_drv_ids[dev->vhost_pmd_id] = 0;
> > + }
> > +
> > rte_eth_dev_stop(dev->port_id);
> > ovs_mutex_unlock(&dev->mutex);
> >
> > @@ -868,7 +995,7 @@ netdev_dpdk_destruct(struct netdev *netdev)
> > }
> >
> > static void
> > -netdev_dpdk_vhost_destruct(struct netdev *netdev)
> > +netdev_dpdk_vhost_cuse_destruct(struct netdev *netdev)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> >
> > @@ -876,15 +1003,8 @@ netdev_dpdk_vhost_destruct(struct netdev
> *netdev)
> > if (netdev_dpdk_get_virtio(dev) != NULL) {
> > 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.",
> > - dev->vhost_id);
> > - }
> > -
> > - if (rte_vhost_driver_unregister(dev->vhost_id)) {
> > - VLOG_ERR("Unable to remove vhost-user socket %s", dev-
> >vhost_id);
> > - } else {
> > - fatal_signal_remove_file_to_unlink(dev->vhost_id);
> > + VLOG_ERR("To restore connectivity after re-adding of port, VM with"
> > + "port '%s' must be restarted.", dev->vhost_id);
> > }
> >
> > ovs_mutex_lock(&dpdk_mutex);
> > @@ -1000,30 +1120,6 @@ netdev_dpdk_vhost_cuse_set_multiq(struct
> netdev *netdev, unsigned int n_txq,
> > netdev->n_txq = n_txq;
> > dev->real_n_txq = 1;
> > netdev->n_rxq = 1;
> > - dev->txq_needs_locking = dev->real_n_txq != netdev->n_txq;
> > -
> > - ovs_mutex_unlock(&dev->mutex);
> > - ovs_mutex_unlock(&dpdk_mutex);
> > -
> > - return err;
> > -}
> > -
> > -static int
> > -netdev_dpdk_vhost_set_multiq(struct netdev *netdev, unsigned int
> n_txq,
> > - unsigned int n_rxq)
> > -{
> > - struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > - int err = 0;
> > -
> > - if (netdev->n_txq == n_txq && netdev->n_rxq == n_rxq) {
> > - return err;
> > - }
> > -
> > - ovs_mutex_lock(&dpdk_mutex);
> > - ovs_mutex_lock(&dev->mutex);
> > -
> > - netdev->n_txq = n_txq;
> > - netdev->n_rxq = n_rxq;
> >
> > ovs_mutex_unlock(&dev->mutex);
> > ovs_mutex_unlock(&dpdk_mutex);
> > @@ -1118,13 +1214,13 @@ dpdk_queue_flush(struct netdev_dpdk *dev,
> int qid)
> > }
> >
> > static bool
> > -is_vhost_running(struct virtio_net *virtio_dev)
> > +is_vhost_cuse_running(struct virtio_net *virtio_dev)
> > {
> > return (virtio_dev != NULL && (virtio_dev->flags &
> VIRTIO_DEV_RUNNING));
> > }
> >
> > static inline void
> > -netdev_dpdk_vhost_update_rx_counters(struct netdev_stats *stats,
> > +netdev_dpdk_vhost_cuse_update_rx_counters(struct netdev_stats
> *stats,
> > struct dp_packet **packets, int count)
> > {
> > int i;
> > @@ -1153,26 +1249,22 @@
> netdev_dpdk_vhost_update_rx_counters(struct netdev_stats *stats,
> > }
> >
> > /*
> > - * The receive path for the vhost port is the TX path out from guest.
> > + * The receive path for the vhost cuse port is the TX path out from guest.
> > */
> > static int
> > -netdev_dpdk_vhost_rxq_recv(struct netdev_rxq *rxq,
> > +netdev_dpdk_vhost_cuse_rxq_recv(struct netdev_rxq *rxq,
> > struct dp_packet **packets, int *c)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(rxq->netdev);
> > struct virtio_net *virtio_dev = netdev_dpdk_get_virtio(dev);
> > - int qid = rxq->queue_id;
> > + int qid = 1;
> > uint16_t nb_rx = 0;
> >
> > - if (OVS_UNLIKELY(!is_vhost_running(virtio_dev))) {
> > + if (OVS_UNLIKELY(!is_vhost_cuse_running(virtio_dev))) {
> > return EAGAIN;
> > }
> >
> > - if (rxq->queue_id >= dev->real_n_rxq) {
> > - return EOPNOTSUPP;
> > - }
> > -
> > - nb_rx = rte_vhost_dequeue_burst(virtio_dev, qid * VIRTIO_QNUM +
> VIRTIO_TXQ,
> > + nb_rx = rte_vhost_dequeue_burst(virtio_dev, qid,
> > dev->dpdk_mp->mp,
> > (struct rte_mbuf **)packets,
> > NETDEV_MAX_BURST);
> > @@ -1181,7 +1273,7 @@ netdev_dpdk_vhost_rxq_recv(struct netdev_rxq
> *rxq,
> > }
> >
> > rte_spinlock_lock(&dev->stats_lock);
> > - netdev_dpdk_vhost_update_rx_counters(&dev->stats, packets,
> nb_rx);
> > + netdev_dpdk_vhost_cuse_update_rx_counters(&dev->stats, packets,
> nb_rx);
> > rte_spinlock_unlock(&dev->stats_lock);
> >
> > *c = (int) nb_rx;
> > @@ -1217,6 +1309,18 @@ netdev_dpdk_rxq_recv(struct netdev_rxq *rxq,
> struct dp_packet **packets,
> > return 0;
> > }
> >
> > +static int
> > +netdev_dpdk_vhost_user_rxq_recv(struct netdev_rxq *rxq,
> > + struct dp_packet **packets, int *c)
> > +{
> > + struct netdev_dpdk *dev = netdev_dpdk_cast(rxq->netdev);
> > +
> > + if (rxq->queue_id >= dev->real_n_rxq) {
> > + return EOPNOTSUPP;
> > + }
> > +
> > + return netdev_dpdk_rxq_recv(rxq, packets, c);
> > +}
> > static inline int
> > netdev_dpdk_qos_run__(struct netdev_dpdk *dev, struct rte_mbuf
> **pkts,
> > int cnt)
> > @@ -1235,7 +1339,7 @@ netdev_dpdk_qos_run__(struct netdev_dpdk
> *dev, struct rte_mbuf **pkts,
> > }
> >
> > static inline void
> > -netdev_dpdk_vhost_update_tx_counters(struct netdev_stats *stats,
> > +netdev_dpdk_vhost_cuse_update_tx_counters(struct netdev_stats
> *stats,
> > struct dp_packet **packets,
> > int attempted,
> > int dropped)
> > @@ -1252,9 +1356,8 @@ netdev_dpdk_vhost_update_tx_counters(struct
> netdev_stats *stats,
> > }
> >
> > static void
> > -__netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
> > - struct dp_packet **pkts, int cnt,
> > - bool may_steal)
> > +__netdev_dpdk_vhost_cuse_send(struct netdev *netdev, struct
> dp_packet **pkts,
> > + int cnt, bool may_steal)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > struct virtio_net *virtio_dev = netdev_dpdk_get_virtio(dev);
> > @@ -1263,26 +1366,24 @@ __netdev_dpdk_vhost_send(struct netdev
> *netdev, int qid,
> > unsigned int qos_pkts = cnt;
> > uint64_t start = 0;
> >
> > - qid = dev->tx_q[qid % dev->real_n_txq].map;
> > -
> > - if (OVS_UNLIKELY(!is_vhost_running(virtio_dev) || qid < 0)) {
> > + if (OVS_UNLIKELY(!is_vhost_cuse_running(virtio_dev))) {
> > rte_spinlock_lock(&dev->stats_lock);
> > dev->stats.tx_dropped+= cnt;
> > rte_spinlock_unlock(&dev->stats_lock);
> > goto out;
> > }
> >
> > - rte_spinlock_lock(&dev->tx_q[qid].tx_lock);
> > + /* There is a single vhost-cuse TX queue, So we need to lock it for TX. */
> > + rte_spinlock_lock(&dev->vhost_cuse_tx_lock);
> >
> > /* Check has QoS has been configured for the netdev */
> > cnt = netdev_dpdk_qos_run__(dev, cur_pkts, cnt);
> > qos_pkts -= cnt;
> >
> > do {
> > - int vhost_qid = qid * VIRTIO_QNUM + VIRTIO_RXQ;
> > unsigned int tx_pkts;
> >
> > - tx_pkts = rte_vhost_enqueue_burst(virtio_dev, vhost_qid,
> > + tx_pkts = rte_vhost_enqueue_burst(virtio_dev, VIRTIO_RXQ,
> > cur_pkts, cnt);
> > if (OVS_LIKELY(tx_pkts)) {
> > /* Packets have been sent.*/
> > @@ -1301,7 +1402,7 @@ __netdev_dpdk_vhost_send(struct netdev
> *netdev, int qid,
> > * Unable to enqueue packets to vhost interface.
> > * Check available entries before retrying.
> > */
> > - while (!rte_vring_available_entries(virtio_dev, vhost_qid)) {
> > + while (!rte_vring_available_entries(virtio_dev, VIRTIO_RXQ)) {
> > if (OVS_UNLIKELY((rte_get_timer_cycles() - start) > timeout)) {
> > expired = 1;
> > break;
> > @@ -1313,12 +1414,12 @@ __netdev_dpdk_vhost_send(struct netdev
> *netdev, int qid,
> > }
> > }
> > } while (cnt);
> > -
> > - rte_spinlock_unlock(&dev->tx_q[qid].tx_lock);
> > + rte_spinlock_unlock(&dev->vhost_cuse_tx_lock);
> >
> > rte_spinlock_lock(&dev->stats_lock);
> > cnt += qos_pkts;
> > - netdev_dpdk_vhost_update_tx_counters(&dev->stats, pkts,
> total_pkts, cnt);
> > + netdev_dpdk_vhost_cuse_update_tx_counters(&dev->stats, pkts,
> total_pkts,
> > + cnt);
> > rte_spinlock_unlock(&dev->stats_lock);
> >
> > out:
> > @@ -1412,8 +1513,9 @@ dpdk_do_tx_copy(struct netdev *netdev, int qid,
> struct dp_packet **pkts,
> > newcnt++;
> > }
> >
> > - if (dev->type == DPDK_DEV_VHOST) {
> > - __netdev_dpdk_vhost_send(netdev, qid, (struct dp_packet **)
> mbufs, newcnt, true);
> > + if (dev->type == DPDK_DEV_VHOST_CUSE) {
> > + __netdev_dpdk_vhost_cuse_send(netdev, (struct dp_packet **)
> mbufs,
> > + newcnt, true);
> > } else {
> > unsigned int qos_pkts = newcnt;
> >
> > @@ -1437,8 +1539,8 @@ dpdk_do_tx_copy(struct netdev *netdev, int qid,
> struct dp_packet **pkts,
> > }
> >
> > static int
> > -netdev_dpdk_vhost_send(struct netdev *netdev, int qid, struct
> dp_packet **pkts,
> > - int cnt, bool may_steal)
> > +netdev_dpdk_vhost_cuse_send(struct netdev *netdev, int qid
> OVS_UNUSED,
> > + struct dp_packet **pkts, int cnt, bool may_steal)
> > {
> > if (OVS_UNLIKELY(pkts[0]->source != DPBUF_DPDK)) {
> > int i;
> > @@ -1450,7 +1552,7 @@ netdev_dpdk_vhost_send(struct netdev
> *netdev, int qid, struct dp_packet **pkts,
> > }
> > }
> > } else {
> > - __netdev_dpdk_vhost_send(netdev, qid, pkts, cnt, may_steal);
> > + __netdev_dpdk_vhost_cuse_send(netdev, pkts, cnt, may_steal);
> > }
> > return 0;
> > }
> > @@ -1461,11 +1563,6 @@ netdev_dpdk_send__(struct netdev_dpdk
> *dev, int qid,
> > {
>
> Now that this handles buffers for vhost, how is may_steal on/off catered
> for?
Looking into this.
>
> > int i;
> >
> > - if (OVS_UNLIKELY(dev->txq_needs_locking)) {
> > - qid = qid % dev->real_n_txq;
> > - rte_spinlock_lock(&dev->tx_q[qid].tx_lock);
> > - }
> > -
>
> why is the possible locking removed?
Error - re instantiated in the v2.
>
> > if (OVS_UNLIKELY(!may_steal ||
> > pkts[0]->source != DPBUF_DPDK)) {
> > struct netdev *netdev = &dev->up;
> > @@ -1541,6 +1638,18 @@ netdev_dpdk_eth_send(struct netdev *netdev,
> int qid,
> > }
> >
> > static int
> > +netdev_dpdk_vhost_user_send(struct netdev *netdev, int qid,
> > + struct dp_packet **pkts, int cnt, bool may_steal)
> > +{
> > + struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > +
> > + qid = qid % dev->real_n_txq;
>
> Due to the modulo, 2 cores can be trying to send on the same queue which
> makes for lock contention. The code was changed to avoid that but now
> the code to avoid and the locking seems to be gone so it could cause
> corruption.
Reintroduced the locking in v2.
>
> > +
> > + netdev_dpdk_send__(dev, qid, pkts, cnt, may_steal);
> > + return 0;
> > +}
> > +
> > +static int
> > netdev_dpdk_set_etheraddr(struct netdev *netdev, const struct
> eth_addr mac)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > @@ -1634,7 +1743,7 @@ static int
> > netdev_dpdk_get_carrier(const struct netdev *netdev, bool *carrier);
> >
> > static int
> > -netdev_dpdk_vhost_get_stats(const struct netdev *netdev,
> > +netdev_dpdk_vhost_cuse_get_stats(const struct netdev *netdev,
> > struct netdev_stats *stats)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > @@ -1796,14 +1905,14 @@ netdev_dpdk_get_carrier(const struct netdev
> *netdev, bool *carrier)
> > }
> >
> > static int
> > -netdev_dpdk_vhost_get_carrier(const struct netdev *netdev, bool
> *carrier)
> > +netdev_dpdk_vhost_cuse_get_carrier(const struct netdev *netdev, bool
> *carrier)
> > {
> > struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
> > struct virtio_net *virtio_dev = netdev_dpdk_get_virtio(dev);
> >
> > ovs_mutex_lock(&dev->mutex);
> >
> > - if (is_vhost_running(virtio_dev)) {
> > + if (is_vhost_cuse_running(virtio_dev)) {
> > *carrier = 1;
> > } else {
> > *carrier = 0;
> > @@ -1853,7 +1962,7 @@ netdev_dpdk_update_flags__(struct
> netdev_dpdk *dev,
> > return 0;
> > }
> >
> > - if (dev->type == DPDK_DEV_ETH) {
> > + if (dev->type != DPDK_DEV_VHOST_CUSE) {
> > if (dev->flags & NETDEV_UP) {
> > err = rte_eth_dev_start(dev->port_id);
> > if (err)
> > @@ -1998,75 +2107,7 @@ set_irq_status(struct virtio_net *virtio_dev)
> > }
> >
> > /*
> > - * Fixes mapping for vhost-user tx queues. Must be called after each
> > - * enabling/disabling of queues and real_n_txq modifications.
> > - */
> > -static void
> > -netdev_dpdk_remap_txqs(struct netdev_dpdk *dev)
> > - OVS_REQUIRES(dev->mutex)
> > -{
> > - int *enabled_queues, n_enabled = 0;
> > - int i, k, total_txqs = dev->real_n_txq;
> > -
> > - enabled_queues = dpdk_rte_mzalloc(total_txqs * sizeof
> *enabled_queues);
> > -
> > - for (i = 0; i < total_txqs; i++) {
> > - /* Enabled queues always mapped to themselves. */
> > - if (dev->tx_q[i].map == i) {
> > - enabled_queues[n_enabled++] = i;
> > - }
> > - }
> > -
> > - if (n_enabled == 0 && total_txqs != 0) {
> > - enabled_queues[0] = OVS_VHOST_QUEUE_DISABLED;
> > - n_enabled = 1;
> > - }
> > -
> > - k = 0;
> > - for (i = 0; i < total_txqs; i++) {
> > - if (dev->tx_q[i].map != i) {
> > - dev->tx_q[i].map = enabled_queues[k];
> > - k = (k + 1) % n_enabled;
> > - }
> > - }
> > -
> > - 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);
> > - }
> > -
> > - rte_free(enabled_queues);
> > -}
> > -
> > -static int
> > -netdev_dpdk_vhost_set_queues(struct netdev_dpdk *dev, struct
> virtio_net *virtio_dev)
> > - OVS_REQUIRES(dev->mutex)
> > -{
> > - uint32_t qp_num;
> > -
> > - qp_num = virtio_dev->virt_qp_nb;
> > - if (qp_num > dev->up.n_rxq) {
> > - VLOG_ERR("vHost Device '%s' %"PRIu64" can't be added - "
> > - "too many queues %d > %d", virtio_dev->ifname, virtio_dev-
> >device_fh,
> > - qp_num, dev->up.n_rxq);
> > - return -1;
> > - }
> > -
> > - dev->real_n_rxq = qp_num;
> > - dev->real_n_txq = qp_num;
> > - dev->txq_needs_locking = true;
> > - /* Enable TX queue 0 by default if it wasn't disabled. */
> > - if (dev->tx_q[0].map == OVS_VHOST_QUEUE_MAP_UNKNOWN) {
> > - dev->tx_q[0].map = 0;
> > - }
> > -
> > - netdev_dpdk_remap_txqs(dev);
> > -
> > - return 0;
> > -}
> > -
> > -/*
> > - * A new virtio-net device is added to a vhost port.
> > + * A new virtio-net device is added to a vhost cuse port.
> > */
> > static int
> > new_device(struct virtio_net *virtio_dev)
> > @@ -2079,11 +2120,6 @@ new_device(struct virtio_net *virtio_dev)
> > LIST_FOR_EACH(dev, list_node, &dpdk_list) {
> > if (strncmp(virtio_dev->ifname, dev->vhost_id, IF_NAME_SZ) == 0) {
> > ovs_mutex_lock(&dev->mutex);
> > - if (netdev_dpdk_vhost_set_queues(dev, virtio_dev)) {
> > - ovs_mutex_unlock(&dev->mutex);
> > - ovs_mutex_unlock(&dpdk_mutex);
> > - return -1;
> > - }
> > ovsrcu_set(&dev->virtio_dev, virtio_dev);
> > exists = true;
> > virtio_dev->flags |= VIRTIO_DEV_RUNNING;
> > @@ -2107,23 +2143,11 @@ new_device(struct virtio_net *virtio_dev)
> > return 0;
> > }
> >
> > -/* Clears mapping for all available queues of vhost interface. */
> > -static void
> > -netdev_dpdk_txq_map_clear(struct netdev_dpdk *dev)
> > - OVS_REQUIRES(dev->mutex)
> > -{
> > - int i;
> > -
> > - for (i = 0; i < dev->real_n_txq; i++) {
> > - dev->tx_q[i].map = OVS_VHOST_QUEUE_MAP_UNKNOWN;
> > - }
> > -}
> > -
> > /*
> > - * Remove a virtio-net device from the specific vhost port. Use dev-
> >remove
> > - * flag to stop any more packets from being sent or received to/from a VM
> and
> > - * ensure all currently queued packets have been sent/received before
> removing
> > - * the device.
> > + * Remove a virtio-net device from the specific vhost cuse port. Use
> > + * dev->remove flag to stop any more packets from being sent or
> received
> > + * to/from a VM and ensure all currently queued packets have been
> sent/received
> > + * before removing the device.
> > */
> > static void
> > destroy_device(volatile struct virtio_net *virtio_dev)
> > @@ -2138,7 +2162,6 @@ destroy_device(volatile struct virtio_net
> *virtio_dev)
> > ovs_mutex_lock(&dev->mutex);
> > virtio_dev->flags &= ~VIRTIO_DEV_RUNNING;
> > ovsrcu_set(&dev->virtio_dev, NULL);
> > - netdev_dpdk_txq_map_clear(dev);
> > exists = true;
> > ovs_mutex_unlock(&dev->mutex);
> > break;
> > @@ -2166,49 +2189,6 @@ destroy_device(volatile struct virtio_net
> *virtio_dev)
> > }
> > }
> >
> > -static int
> > -vring_state_changed(struct virtio_net *virtio_dev, uint16_t queue_id,
> > - int enable)
> > -{
> > - struct netdev_dpdk *dev;
> > - bool exists = false;
> > - int qid = queue_id / VIRTIO_QNUM;
> > -
> > - if (queue_id % VIRTIO_QNUM == VIRTIO_TXQ) {
> > - return 0;
> > - }
> > -
> > - ovs_mutex_lock(&dpdk_mutex);
> > - LIST_FOR_EACH (dev, list_node, &dpdk_list) {
> > - if (strncmp(virtio_dev->ifname, dev->vhost_id, IF_NAME_SZ) == 0) {
> > - ovs_mutex_lock(&dev->mutex);
> > - if (enable) {
> > - dev->tx_q[qid].map = qid;
> > - } else {
> > - dev->tx_q[qid].map = OVS_VHOST_QUEUE_DISABLED;
> > - }
> > - netdev_dpdk_remap_txqs(dev);
> > - exists = true;
> > - ovs_mutex_unlock(&dev->mutex);
> > - break;
> > - }
> > - }
> > - ovs_mutex_unlock(&dpdk_mutex);
> > -
> > - if (exists) {
> > - VLOG_INFO("State of queue %d ( tx_qid %d ) of vhost device '%s' %"
> > - PRIu64" changed to \'%s\'", queue_id, qid,
> > - virtio_dev->ifname, virtio_dev->device_fh,
> > - (enable == 1) ? "enabled" : "disabled");
> > - } else {
> > - VLOG_INFO("vHost Device '%s' %"PRIu64" not found", virtio_dev-
> >ifname,
> > - virtio_dev->device_fh);
> > - return -1;
> > - }
> > -
> > - return 0;
> > -}
> > -
> > struct virtio_net *
> > netdev_dpdk_get_virtio(const struct netdev_dpdk *dev)
> > {
> > @@ -2216,18 +2196,17 @@ netdev_dpdk_get_virtio(const struct
> netdev_dpdk *dev)
> > }
> >
> > /*
> > - * These callbacks allow virtio-net devices to be added to vhost ports when
> > - * configuration has been fully complete.
> > + * These callbacks allow virtio-net devices to be added to vhost cuse ports
> > + * when configuration has been fully complete.
> > */
> > static const struct virtio_net_device_ops virtio_net_device_ops =
> > {
> > .new_device = new_device,
> > .destroy_device = destroy_device,
> > - .vring_state_changed = vring_state_changed
> > };
> >
> > static void *
> > -start_vhost_loop(void *dummy OVS_UNUSED)
> > +start_vhost_cuse_loop(void *dummy OVS_UNUSED)
> > {
> > pthread_detach(pthread_self());
> > /* Put the cuse thread into quiescent state. */
> > @@ -2237,19 +2216,7 @@ start_vhost_loop(void *dummy OVS_UNUSED)
> > }
> >
> > static int
> > -dpdk_vhost_class_init(void)
> > -{
> > - rte_vhost_driver_callback_register(&virtio_net_device_ops);
> > - rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO4
> > - | 1ULL << VIRTIO_NET_F_HOST_TSO6
> > - | 1ULL << VIRTIO_NET_F_CSUM);
>
> I think you'll still want to set these for vhostuser.
I didn't think there was an equivalent function to expose this in the PMD but there is! Used in the v2.
>
> > -
> > - ovs_thread_create("vhost_thread", start_vhost_loop, NULL);
> > - return 0;
> > -}
> > -
> > -static int
> > -dpdk_vhost_cuse_class_init(void)
> > +netdev_dpdk_vhost_cuse_class_init(void)
> > {
> > int err = -1;
> >
> > @@ -2265,14 +2232,12 @@ dpdk_vhost_cuse_class_init(void)
> > return -1;
> > }
> >
> > - dpdk_vhost_class_init();
> > - return 0;
> > -}
> > + rte_vhost_driver_callback_register(&virtio_net_device_ops);
> > + rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO4
> > + | 1ULL << VIRTIO_NET_F_HOST_TSO6
> > + | 1ULL << VIRTIO_NET_F_CSUM);
> > + ovs_thread_create("vhost_cuse_thread", start_vhost_cuse_loop,
> NULL);
> >
> > -static int
> > -dpdk_vhost_user_class_init(void)
> > -{
> > - dpdk_vhost_class_init();
> > return 0;
> > }
> >
> > @@ -2859,30 +2824,30 @@ static const struct netdev_class dpdk_ring_class
> =
> > static const struct netdev_class OVS_UNUSED dpdk_vhost_cuse_class =
> > NETDEV_DPDK_CLASS(
> > "dpdkvhostcuse",
> > - dpdk_vhost_cuse_class_init,
> > + netdev_dpdk_vhost_cuse_class_init,
> > netdev_dpdk_vhost_cuse_construct,
> > - netdev_dpdk_vhost_destruct,
> > + netdev_dpdk_vhost_cuse_destruct,
> > netdev_dpdk_vhost_cuse_set_multiq,
> > - netdev_dpdk_vhost_send,
> > - netdev_dpdk_vhost_get_carrier,
> > - netdev_dpdk_vhost_get_stats,
> > + netdev_dpdk_vhost_cuse_send,
> > + netdev_dpdk_vhost_cuse_get_carrier,
> > + netdev_dpdk_vhost_cuse_get_stats,
> > NULL,
> > NULL,
> > - netdev_dpdk_vhost_rxq_recv);
> > + netdev_dpdk_vhost_cuse_rxq_recv);
> >
> > static const struct netdev_class OVS_UNUSED dpdk_vhost_user_class =
> > NETDEV_DPDK_CLASS(
> > "dpdkvhostuser",
> > - dpdk_vhost_user_class_init,
> > - netdev_dpdk_vhost_user_construct,
> > - netdev_dpdk_vhost_destruct,
> > - netdev_dpdk_vhost_set_multiq,
> > - netdev_dpdk_vhost_send,
> > - netdev_dpdk_vhost_get_carrier,
> > - netdev_dpdk_vhost_get_stats,
> > - NULL,
> > NULL,
> > - netdev_dpdk_vhost_rxq_recv);
> > + netdev_dpdk_vhost_user_construct,
> > + netdev_dpdk_destruct,
> > + netdev_dpdk_set_multiq,
> > + netdev_dpdk_vhost_user_send,
> > + netdev_dpdk_get_carrier,
> > + netdev_dpdk_get_stats,
>
> Are all the phy type stats available for vhost? or do you need to check
> the netdev type in that function and report specific stats. I didn't
> think DPDK would report things like rx_dropped for vhost but I'll be
> glad to be wrong.
Fixed in v2.
>
> > + netdev_dpdk_get_features,
> > + netdev_dpdk_get_status,
> > + netdev_dpdk_vhost_user_rxq_recv);
> >
> > void
> > netdev_dpdk_register(void)
> >
Thanks for the thorough review Kevin. Hopefully I've addressed your comments. The v2 will be posted shortly.
Thanks,
Ciara
More information about the dev
mailing list