@@ -159,6 +159,10 @@ function install_dpdk()
# Disable building DPDK unit tests. Not needed for OVS build or tests.
DPDK_OPTS="$DPDK_OPTS -Dtests=false"
+ # Disable DPDK developer mode, this results in less build checks and less
+ # meson verbose outputs.
+ DPDK_OPTS="$DPDK_OPTS -Ddeveloper_mode=disabled"
+
# Install DPDK using prefix.
DPDK_OPTS="$DPDK_OPTS --prefix=$(pwd)/build"
@@ -216,7 +220,7 @@ fi
if [ "$DPDK" ] || [ "$DPDK_SHARED" ]; then
if [ -z "$DPDK_VER" ]; then
- DPDK_VER="20.11.1"
+ DPDK_VER="21.11"
fi
install_dpdk $DPDK_VER
if [ "$CC" = "clang" ]; then
@@ -21,8 +21,8 @@ make -j4 HAVE_LLVM= HAVE_SQLITE= install
cd ..
pip3 install --disable-pip-version-check --user \
- flake8 hacking sphinx wheel setuptools
-pip3 install --user 'meson==0.47.1'
+ flake8 hacking sphinx wheel setuptools pyelftools
+pip3 install --user 'meson==0.49.2'
if [ "$M32" ]; then
# Installing 32-bit libraries.
@@ -225,7 +225,7 @@ Q: Are all the DPDK releases that OVS versions work with maintained?
The latest information about DPDK stable and LTS releases can be found
at `DPDK stable`_.
-.. _DPDK stable: http://doc.dpdk.org/guides-20.11/contributing/stable.html
+.. _DPDK stable: http://doc.dpdk.org/guides-21.11/contributing/stable.html
Q: I get an error like this when I configure Open vSwitch:
@@ -42,7 +42,7 @@ Build requirements
In addition to the requirements described in :doc:`general`, building Open
vSwitch with DPDK will require the following:
-- DPDK 20.11.1
+- DPDK 21.11
- A `DPDK supported NIC`_
@@ -59,8 +59,8 @@ vSwitch with DPDK will require the following:
Detailed system requirements can be found at `DPDK requirements`_.
-.. _DPDK supported NIC: https://doc.dpdk.org/guides-20.11/nics/index.html
-.. _DPDK requirements: https://doc.dpdk.org/guides-20.11/linux_gsg/sys_reqs.html
+.. _DPDK supported NIC: https://doc.dpdk.org/guides-21.11/nics/index.html
+.. _DPDK requirements: https://doc.dpdk.org/guides-21.11/linux_gsg/sys_reqs.html
.. _dpdk-install:
@@ -73,9 +73,9 @@ Install DPDK
#. Download the `DPDK sources`_, extract the file and set ``DPDK_DIR``::
$ cd /usr/src/
- $ wget https://fast.dpdk.org/rel/dpdk-20.11.1.tar.xz
- $ tar xf dpdk-20.11.1.tar.xz
- $ export DPDK_DIR=/usr/src/dpdk-stable-20.11.1
+ $ wget https://fast.dpdk.org/rel/dpdk-21.11.tar.xz
+ $ tar xf dpdk-21.11.tar.xz
+ $ export DPDK_DIR=/usr/src/dpdk-21.11
$ cd $DPDK_DIR
#. Configure and install DPDK using Meson
@@ -121,7 +121,7 @@ Install DPDK
.. _DPDK sources: http://dpdk.org/rel
.. _DPDK documentation:
- https://doc.dpdk.org/guides-20.11/linux_gsg/build_dpdk.html
+ https://doc.dpdk.org/guides-21.11/linux_gsg/build_dpdk.html
Install OVS
~~~~~~~~~~~
@@ -722,7 +722,7 @@ Limitations
release notes`_.
.. _DPDK release notes:
- https://doc.dpdk.org/guides-20.11/rel_notes/release_20_11.html
+ https://doc.dpdk.org/guides-21.11/rel_notes/release_21_11.html
- Upper bound MTU: DPDK device drivers differ in how the L2 frame for a
given MTU value is calculated e.g. i40e driver includes 2 x vlan headers in
@@ -117,7 +117,7 @@ tool::
For more information, refer to the `DPDK documentation <dpdk-drivers>`__.
-.. _dpdk-drivers: https://doc.dpdk.org/guides-20.11/linux_gsg/linux_drivers.html
+.. _dpdk-drivers: https://doc.dpdk.org/guides-21.11/linux_gsg/linux_drivers.html
.. _dpdk-phy-multiqueue:
@@ -235,7 +235,7 @@ To hotplug a port with igb_uio in this case, DPDK must be configured to use
physical addressing for IOVA mode. For more information regarding IOVA modes
in DPDK please refer to the `DPDK IOVA Mode Detection`__.
-__ https://doc.dpdk.org/guides-20.11/prog_guide/env_abstraction_layer.html#iova-mode-detection
+__ https://doc.dpdk.org/guides-21.11/prog_guide/env_abstraction_layer.html#iova-mode-detection
To configure OVS DPDK to use physical addressing for IOVA::
@@ -267,7 +267,7 @@ Representors are multi devices created on top of one PF.
For more information, refer to the `DPDK documentation`__.
-__ https://doc.dpdk.org/guides-20.11/prog_guide/switch_representation.html
+__ https://doc.dpdk.org/guides-21.11/prog_guide/switch_representation.html
Prior to port representors there was a one-to-one relationship between the PF
and the eth device. With port representors the relationship becomes one PF to
@@ -401,7 +401,7 @@ in the ``options`` column of the ``Interface`` table.
kernel netdevice, and be inherited from it when Open vSwitch is restarted,
even if the options described in this section are unset from Open vSwitch.
-.. _bifurcated-drivers: https://doc.dpdk.org/guides-20.11/linux_gsg/linux_drivers.html#bifurcated-driver
+.. _bifurcated-drivers: https://doc.dpdk.org/guides-21.11/linux_gsg/linux_drivers.html#bifurcated-driver
- Configure the VF MAC address::
@@ -63,4 +63,4 @@ run::
More information on the different types of virtual DPDK PMDs can be found in
the `DPDK documentation`__.
-__ https://doc.dpdk.org/guides-20.11/nics/overview.html
+__ https://doc.dpdk.org/guides-21.11/nics/overview.html
@@ -539,4 +539,4 @@ shown with::
Further information can be found in the
`DPDK documentation
-<https://doc.dpdk.org/guides-20.11/prog_guide/vhost_lib.html>`__
+<https://doc.dpdk.org/guides-21.11/prog_guide/vhost_lib.html>`__
@@ -353,7 +353,7 @@ All tests are skipped if no hugepages are configured. User must look into the DP
manual to figure out how to `Configure hugepages`_.
The phy test will skip if no compatible physical device is available.
-.. _Configure hugepages: https://doc.dpdk.org/guides-20.11/linux_gsg/sys_reqs.html
+.. _Configure hugepages: https://doc.dpdk.org/guides-21.11/linux_gsg/sys_reqs.html
All the features documented under `Unit Tests`_ are available for the DPDK
datapath testsuite.
@@ -10,6 +10,7 @@ Post-v2.16.0
limiting behavior.
* Add hardware offload support for matching IPv4/IPv6 frag types
(experimental).
+ * Add support for DPDK 21.11.
- Python:
* For SSL support, the use of the pyOpenSSL library has been replaced
with the native 'ssl' module.
@@ -58,29 +58,31 @@ enum OVS_PACKED_ENUM dp_packet_source {
enum dp_packet_offload_mask {
/* Value 0 is not used. */
/* Is the 'rss_hash' valid? */
- DEF_OL_FLAG(DP_PACKET_OL_RSS_HASH, PKT_RX_RSS_HASH, 0x1),
+ DEF_OL_FLAG(DP_PACKET_OL_RSS_HASH, RTE_MBUF_F_RX_RSS_HASH, 0x1),
/* Is the 'flow_mark' valid? */
- DEF_OL_FLAG(DP_PACKET_OL_FLOW_MARK, PKT_RX_FDIR_ID, 0x2),
+ DEF_OL_FLAG(DP_PACKET_OL_FLOW_MARK, RTE_MBUF_F_RX_FDIR_ID, 0x2),
/* Bad L4 checksum in the packet. */
- DEF_OL_FLAG(DP_PACKET_OL_RX_L4_CKSUM_BAD, PKT_RX_L4_CKSUM_BAD, 0x4),
+ DEF_OL_FLAG(DP_PACKET_OL_RX_L4_CKSUM_BAD, RTE_MBUF_F_RX_L4_CKSUM_BAD, 0x4),
/* Bad IP checksum in the packet. */
- DEF_OL_FLAG(DP_PACKET_OL_RX_IP_CKSUM_BAD, PKT_RX_IP_CKSUM_BAD, 0x8),
+ DEF_OL_FLAG(DP_PACKET_OL_RX_IP_CKSUM_BAD, RTE_MBUF_F_RX_IP_CKSUM_BAD, 0x8),
/* Valid L4 checksum in the packet. */
- DEF_OL_FLAG(DP_PACKET_OL_RX_L4_CKSUM_GOOD, PKT_RX_L4_CKSUM_GOOD, 0x10),
+ DEF_OL_FLAG(DP_PACKET_OL_RX_L4_CKSUM_GOOD, RTE_MBUF_F_RX_L4_CKSUM_GOOD,
+ 0x10),
/* Valid IP checksum in the packet. */
- DEF_OL_FLAG(DP_PACKET_OL_RX_IP_CKSUM_GOOD, PKT_RX_IP_CKSUM_GOOD, 0x20),
+ DEF_OL_FLAG(DP_PACKET_OL_RX_IP_CKSUM_GOOD, RTE_MBUF_F_RX_IP_CKSUM_GOOD,
+ 0x20),
/* TCP Segmentation Offload. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_TCP_SEG, PKT_TX_TCP_SEG, 0x40),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_TCP_SEG, RTE_MBUF_F_TX_TCP_SEG, 0x40),
/* Offloaded packet is IPv4. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_IPV4, PKT_TX_IPV4, 0x80),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_IPV4, RTE_MBUF_F_TX_IPV4, 0x80),
/* Offloaded packet is IPv6. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_IPV6, PKT_TX_IPV6, 0x100),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_IPV6, RTE_MBUF_F_TX_IPV6, 0x100),
/* Offload TCP checksum. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_TCP_CKSUM, PKT_TX_TCP_CKSUM, 0x200),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_TCP_CKSUM, RTE_MBUF_F_TX_TCP_CKSUM, 0x200),
/* Offload UDP checksum. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_UDP_CKSUM, PKT_TX_UDP_CKSUM, 0x400),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_UDP_CKSUM, RTE_MBUF_F_TX_UDP_CKSUM, 0x400),
/* Offload SCTP checksum. */
- DEF_OL_FLAG(DP_PACKET_OL_TX_SCTP_CKSUM, PKT_TX_SCTP_CKSUM, 0x800),
+ DEF_OL_FLAG(DP_PACKET_OL_TX_SCTP_CKSUM, RTE_MBUF_F_TX_SCTP_CKSUM, 0x800),
/* Adding new field requires adding to DP_PACKET_OL_SUPPORTED_MASK. */
};
@@ -181,7 +181,7 @@ static int vring_state_changed(int vid, uint16_t queue_id, int enable);
static void destroy_connection(int vid);
static void vhost_guest_notified(int vid);
-static const struct vhost_device_ops virtio_net_device_ops =
+static const struct rte_vhost_device_ops virtio_net_device_ops =
{
.new_device = new_device,
.destroy_device = destroy_device,
@@ -1279,26 +1279,6 @@ common_construct(struct netdev *netdev, dpdk_port_t port_no,
return 0;
}
-/* Get the number of OVS interfaces which have the same DPDK
- * rte device (e.g. same pci bus address).
- * FIXME: avoid direct access to DPDK internal array rte_eth_devices.
- */
-static int
-netdev_dpdk_get_num_ports(struct rte_device *device)
- OVS_REQUIRES(dpdk_mutex)
-{
- struct netdev_dpdk *dev;
- int count = 0;
-
- LIST_FOR_EACH (dev, list_node, &dpdk_list) {
- if (rte_eth_devices[dev->port_id].device == device
- && rte_eth_devices[dev->port_id].state != RTE_ETH_DEV_UNUSED) {
- count++;
- }
- }
- return count;
-}
-
static int
vhost_common_construct(struct netdev *netdev)
OVS_REQUIRES(dpdk_mutex)
@@ -1452,8 +1432,6 @@ static void
netdev_dpdk_destruct(struct netdev *netdev)
{
struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
- struct rte_device *rte_dev;
- struct rte_eth_dev *eth_dev;
ovs_mutex_lock(&dpdk_mutex);
@@ -1461,20 +1439,43 @@ netdev_dpdk_destruct(struct netdev *netdev)
dev->started = false;
if (dev->attached) {
- /* Retrieve eth device data before closing it.
- * FIXME: avoid direct access to DPDK internal array rte_eth_devices.
- */
- eth_dev = &rte_eth_devices[dev->port_id];
- rte_dev = eth_dev->device;
+ bool dpdk_resources_still_used = false;
+ struct rte_eth_dev_info dev_info;
+ dpdk_port_t sibling_port_id;
+
+ /* Check if this netdev has siblings (i.e. shares DPDK resources) among
+ * other OVS netdevs. */
+ RTE_ETH_FOREACH_DEV_SIBLING (sibling_port_id, dev->port_id) {
+ struct netdev_dpdk *sibling;
+
+ /* RTE_ETH_FOREACH_DEV_SIBLING lists dev->port_id as part of the
+ * loop. */
+ if (sibling_port_id == dev->port_id) {
+ continue;
+ }
+ LIST_FOR_EACH (sibling, list_node, &dpdk_list) {
+ if (sibling->port_id != sibling_port_id) {
+ continue;
+ }
+ dpdk_resources_still_used = true;
+ break;
+ }
+ if (dpdk_resources_still_used) {
+ break;
+ }
+ }
+
+ /* Retrieve eth device data before closing it. */
+ rte_eth_dev_info_get(dev->port_id, &dev_info);
/* Remove the eth device. */
rte_eth_dev_close(dev->port_id);
- /* Remove this rte device and all its eth devices if all the eth
- * devices belonging to the rte device are closed.
- */
- if (!netdev_dpdk_get_num_ports(rte_dev)) {
- int ret = rte_dev_remove(rte_dev);
+ /* Remove the rte device if no associated eth device is used by OVS.
+ * Note: any remaining eth devices associated to this rte device are
+ * closed by DPDK ethdev layer. */
+ if (!dpdk_resources_still_used) {
+ int ret = rte_dev_remove(dev_info.device);
if (ret < 0) {
VLOG_ERR("Device '%s' can not be detached: %s.",
@@ -2165,14 +2166,14 @@ netdev_dpdk_prep_hwol_packet(struct netdev_dpdk *dev, struct rte_mbuf *mbuf)
{
struct dp_packet *pkt = CONTAINER_OF(mbuf, struct dp_packet, mbuf);
- if (mbuf->ol_flags & PKT_TX_L4_MASK) {
+ if (mbuf->ol_flags & RTE_MBUF_F_TX_L4_MASK) {
mbuf->l2_len = (char *)dp_packet_l3(pkt) - (char *)dp_packet_eth(pkt);
mbuf->l3_len = (char *)dp_packet_l4(pkt) - (char *)dp_packet_l3(pkt);
mbuf->outer_l2_len = 0;
mbuf->outer_l3_len = 0;
}
- if (mbuf->ol_flags & PKT_TX_TCP_SEG) {
+ if (mbuf->ol_flags & RTE_MBUF_F_TX_TCP_SEG) {
struct tcp_header *th = dp_packet_l4(pkt);
if (!th) {
@@ -2182,11 +2183,11 @@ netdev_dpdk_prep_hwol_packet(struct netdev_dpdk *dev, struct rte_mbuf *mbuf)
}
mbuf->l4_len = TCP_OFFSET(th->tcp_ctl) * 4;
- mbuf->ol_flags |= PKT_TX_TCP_CKSUM;
+ mbuf->ol_flags |= RTE_MBUF_F_TX_TCP_CKSUM;
mbuf->tso_segsz = dev->mtu - mbuf->l3_len - mbuf->l4_len;
- if (mbuf->ol_flags & PKT_TX_IPV4) {
- mbuf->ol_flags |= PKT_TX_IP_CKSUM;
+ if (mbuf->ol_flags & RTE_MBUF_F_TX_IPV4) {
+ mbuf->ol_flags |= RTE_MBUF_F_TX_IP_CKSUM;
}
}
return true;
@@ -2533,7 +2534,7 @@ netdev_dpdk_filter_packet_len(struct netdev_dpdk *dev, struct rte_mbuf **pkts,
for (i = 0; i < pkt_cnt; i++) {
pkt = pkts[i];
if (OVS_UNLIKELY((pkt->pkt_len > dev->max_packet_len)
- && !(pkt->ol_flags & PKT_TX_TCP_SEG))) {
+ && !(pkt->ol_flags & RTE_MBUF_F_TX_TCP_SEG))) {
VLOG_WARN_RL(&rl, "%s: Too big size %" PRIu32 " "
"max_packet_len %d", dev->up.name, pkt->pkt_len,
dev->max_packet_len);
@@ -2755,12 +2756,12 @@ dpdk_copy_dp_packet_to_mbuf(struct rte_mempool *mp, struct dp_packet *pkt_orig)
mbuf_dest->tx_offload = pkt_orig->mbuf.tx_offload;
mbuf_dest->packet_type = pkt_orig->mbuf.packet_type;
mbuf_dest->ol_flags |= (pkt_orig->mbuf.ol_flags &
- ~(EXT_ATTACHED_MBUF | IND_ATTACHED_MBUF));
+ ~(RTE_MBUF_F_EXTERNAL | RTE_MBUF_F_INDIRECT));
memcpy(&pkt_dest->l2_pad_size, &pkt_orig->l2_pad_size,
sizeof(struct dp_packet) - offsetof(struct dp_packet, l2_pad_size));
- if (mbuf_dest->ol_flags & PKT_TX_L4_MASK) {
+ if (mbuf_dest->ol_flags & RTE_MBUF_F_TX_L4_MASK) {
mbuf_dest->l2_len = (char *)dp_packet_l3(pkt_dest)
- (char *)dp_packet_eth(pkt_dest);
mbuf_dest->l3_len = (char *)dp_packet_l4(pkt_dest)
@@ -2805,7 +2806,7 @@ dpdk_do_tx_copy(struct netdev *netdev, int qid, struct dp_packet_batch *batch)
uint32_t size = dp_packet_size(packet);
if (size > dev->max_packet_len
- && !(packet->mbuf.ol_flags & PKT_TX_TCP_SEG)) {
+ && !(packet->mbuf.ol_flags & RTE_MBUF_F_TX_TCP_SEG)) {
VLOG_WARN_RL(&rl, "Too big size %u max_packet_len %d", size,
dev->max_packet_len);
mtu_drops++;
@@ -3679,8 +3680,8 @@ netdev_dpdk_get_status(const struct netdev *netdev, struct smap *args)
ovs_mutex_unlock(&dev->mutex);
const struct rte_bus *bus;
const struct rte_pci_device *pci_dev;
- uint16_t vendor_id = PCI_ANY_ID;
- uint16_t device_id = PCI_ANY_ID;
+ uint16_t vendor_id = RTE_PCI_ANY_ID;
+ uint16_t device_id = RTE_PCI_ANY_ID;
bus = rte_bus_find_by_device(dev_info.device);
if (bus && !strcmp(bus->name, "pci")) {
pci_dev = RTE_DEV_TO_PCI(dev_info.device);
@@ -3793,12 +3794,12 @@ static void
netdev_dpdk_detach(struct unixctl_conn *conn, int argc OVS_UNUSED,
const char *argv[], void *aux OVS_UNUSED)
{
- char *response;
- dpdk_port_t port_id;
- struct netdev_dpdk *dev;
- struct rte_device *rte_dev;
struct ds used_interfaces = DS_EMPTY_INITIALIZER;
+ struct rte_eth_dev_info dev_info;
+ dpdk_port_t sibling_port_id;
+ dpdk_port_t port_id;
bool used = false;
+ char *response;
ovs_mutex_lock(&dpdk_mutex);
@@ -3808,18 +3809,21 @@ netdev_dpdk_detach(struct unixctl_conn *conn, int argc OVS_UNUSED,
goto error;
}
- rte_dev = rte_eth_devices[port_id].device;
ds_put_format(&used_interfaces,
"Device '%s' is being used by the following interfaces:",
argv[1]);
- LIST_FOR_EACH (dev, list_node, &dpdk_list) {
- /* FIXME: avoid direct access to DPDK array rte_eth_devices. */
- if (rte_eth_devices[dev->port_id].device == rte_dev
- && rte_eth_devices[dev->port_id].state != RTE_ETH_DEV_UNUSED) {
+ RTE_ETH_FOREACH_DEV_SIBLING (sibling_port_id, port_id) {
+ struct netdev_dpdk *dev;
+
+ LIST_FOR_EACH (dev, list_node, &dpdk_list) {
+ if (dev->port_id != sibling_port_id) {
+ continue;
+ }
used = true;
ds_put_format(&used_interfaces, " %s",
netdev_get_name(&dev->up));
+ break;
}
}
@@ -3831,8 +3835,9 @@ netdev_dpdk_detach(struct unixctl_conn *conn, int argc OVS_UNUSED,
}
ds_destroy(&used_interfaces);
+ rte_eth_dev_info_get(port_id, &dev_info);
rte_eth_dev_close(port_id);
- if (rte_dev_remove(rte_dev) < 0) {
+ if (rte_dev_remove(dev_info.device) < 0) {
response = xasprintf("Device '%s' can not be detached", argv[1]);
goto error;
}
@@ -5272,7 +5277,7 @@ netdev_dpdk_rte_flow_query_count(struct netdev *netdev,
struct rte_flow_query_count *query,
struct rte_flow_error *error)
{
- struct rte_flow_action_count count = { .shared = 0, .id = 0 };
+ struct rte_flow_action_count count = { .id = 0, };
const struct rte_flow_action actions[] = {
{
.type = RTE_FLOW_ACTION_TYPE_COUNT,