[ovs-dev] [dpdk-howl PATCH v2] netdev-dpdk: Upgrade to dpdk v18.08.0
Eelco Chaudron
echaudro at redhat.com
Mon Sep 17 16:38:44 UTC 2018
Hi Ophir,
I did some basic PVP testing root vs non-root and it all seems to work
as expected (did not do any performance comparison).
I do see compiler warnings about usage of deprecated DPDK functions, I
guess these need to be fixed:
lib/netdev-dpdk.c: In function 'netdev_dpdk_destruct':
lib/netdev-dpdk.c:1336:9: error: 'rte_eth_dev_detach' is deprecated
(declared at
/home/root/OVS_dpdk-hwol_DPDK_v18.08/dpdk-stable/x86_64-native-linuxapp-gcc/include/rte_ethdev.h:1451)
[-Werror=deprecated-declarations]
if (rte_eth_dev_detach(dev->port_id, devname) < 0) {
^
lib/netdev-dpdk.c: In function 'netdev_dpdk_process_devargs':
lib/netdev-dpdk.c:1629:13: error: 'rte_eth_dev_attach' is deprecated
(declared at
/home/root/OVS_dpdk-hwol_DPDK_v18.08/dpdk-stable/x86_64-native-linuxapp-gcc/include/rte_ethdev.h:1435)
[-Werror=deprecated-declarations]
if (!rte_eth_dev_attach(devargs, &new_port_id)) {
^
lib/netdev-dpdk.c: In function 'netdev_dpdk_detach':
lib/netdev-dpdk.c:3159:5: error: 'rte_eth_dev_detach' is deprecated
(declared at
/home/root/OVS_dpdk-hwol_DPDK_v18.08/dpdk-stable/x86_64-native-linuxapp-gcc/include/rte_ethdev.h:1451)
[-Werror=deprecated-declarations]
ret = rte_eth_dev_detach(port_id, devname);
^
See more comments inline…
//Eelco
On 11 Sep 2018, at 1:04, Ophir Munk wrote:
> 1. Enable compilation and linkage with dpdk 18.08.0
> The following dpdk commits which were introduced after dpdk 17.11.x
> require OVS updates to accommodate to the dpdk changes.
> - ce17eddefc20 ("ethdev: introduce Rx queue offloads API")
> - ab3ce1e0c193 ("ethdev: remove old offload API")
> - c06ddf9698e0 ("meter: add configuration profile")
> - e58638c32411 ("ethdev: fix TPID handling in flow API")
> - cd8c7c7ce241 ("ethdev: replace bus specific struct with generic
> dev")
> - ac8d22de2394 ("ethdev: flatten RSS configuration in flow API")
>
> 2. Limit configured rss hash functions to only those supported
> by the eth device.
>
> 3. Update references to DPDK version in Documentation
>
> 4. Update DPDK version in travis' linux-build script
>
> Signed-off-by: Ophir Munk <ophirmu at mellanox.com>
> ---
> v1:
> First version
>
> v2:
> Avoid seg faults cases as described in
> https://patchwork.ozlabs.org/patch/965451/
> by using the patch in:
> https://github.com/kevintraynor/ovs-dpdk-master/commit/88f46cc5ab338eb4f3ca5db1eacd0effefe4fa0c
>
> .travis/linux-build.sh | 2 +-
> Documentation/faq/releases.rst | 2 +-
> lib/netdev-dpdk.c | 142
> +++++++++++++++++++++++++++++------------
> 3 files changed, 104 insertions(+), 42 deletions(-)
>
> diff --git a/.travis/linux-build.sh b/.travis/linux-build.sh
> index 4b9fc4a..c60ac71 100755
> --- a/.travis/linux-build.sh
> +++ b/.travis/linux-build.sh
> @@ -83,7 +83,7 @@ fi
>
> if [ "$DPDK" ]; then
> if [ -z "$DPDK_VER" ]; then
> - DPDK_VER="17.11.3"
> + DPDK_VER="18.08.0"
> fi
> install_dpdk $DPDK_VER
> if [ "$CC" = "clang" ]; then
> diff --git a/Documentation/faq/releases.rst
> b/Documentation/faq/releases.rst
> index 41d41e3..646ae09 100644
> --- a/Documentation/faq/releases.rst
> +++ b/Documentation/faq/releases.rst
> @@ -168,7 +168,7 @@ Q: What DPDK version does each Open vSwitch
> release work with?
> 2.7.x 16.11.7
> 2.8.x 17.05.2
> 2.9.x 17.11.3
> - 2.10.x 17.11.3
> + 2.10.x 18.08.0
> ============ =======
>
> Q: Are all the DPDK releases that OVS versions work with maintained?
> diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
> index f91aa27..6879e9a 100644
> --- a/lib/netdev-dpdk.c
> +++ b/lib/netdev-dpdk.c
> @@ -168,11 +168,7 @@ static const struct rte_eth_conf port_conf = {
> .rxmode = {
> .mq_mode = ETH_MQ_RX_RSS,
> .split_hdr_size = 0,
> - .header_split = 0, /* Header Split disabled */
> - .hw_ip_checksum = 0, /* IP checksum offload disabled */
> - .hw_vlan_filter = 0, /* VLAN filtering disabled */
> - .jumbo_frame = 0, /* Jumbo Frame Support disabled */
> - .hw_strip_crc = 0,
> + .offloads = 0,
> },
> .rx_adv_conf = {
> .rss_conf = {
> @@ -364,6 +360,7 @@ struct dpdk_ring {
> struct ingress_policer {
> struct rte_meter_srtcm_params app_srtcm_params;
> struct rte_meter_srtcm in_policer;
> + struct rte_meter_srtcm_profile in_prof;
> rte_spinlock_t policer_lock;
> };
> @@ -894,6 +891,8 @@ dpdk_eth_dev_port_config(struct netdev_dpdk *dev,
> int n_rxq, int n_txq)
> struct rte_eth_dev_info info;
> uint16_t conf_mtu;
> + rte_eth_dev_info_get(dev->port_id, &info);
> +
> /* As of DPDK 17.11.1 a few PMDs require to explicitly enable
> * scatter to support jumbo RX. Checking the offload capabilities
> * is not an option as PMDs are not required yet to report
> @@ -901,20 +900,24 @@ dpdk_eth_dev_port_config(struct netdev_dpdk
> *dev, int n_rxq, int n_txq)
> * (testing or code review). Listing all such PMDs feels harder
> * than highlighting the one known not to need scatter */
> if (dev->mtu > ETHER_MTU) {
> - rte_eth_dev_info_get(dev->port_id, &info);
> if (strncmp(info.driver_name, "net_nfp", 7)) {
> - conf.rxmode.enable_scatter = 1;
> + conf.rxmode.offloads |= DEV_RX_OFFLOAD_SCATTER;
> }
> }
EC> The above was supposed to be fixed in the drivers for 18+, was this
the case? If so we should be able to change this hard-coded nic name?
I’ve copied in Pablo who added the code above.
> conf.intr_conf.lsc = dev->lsc_interrupt_mode;
> - conf.rxmode.hw_ip_checksum = (dev->hw_ol_features &
> - NETDEV_RX_CHECKSUM_OFFLOAD) != 0;
> + conf.rxmode.offloads |= ((dev->hw_ol_features &
> + NETDEV_RX_CHECKSUM_OFFLOAD) != 0) ?
> + DEV_RX_OFFLOAD_CHECKSUM : 0;
EC> Maybe convert the above also to an if, as below, to make it more
clear?
> if (dev->hw_ol_features & NETDEV_RX_HW_CRC_STRIP) {
> - conf.rxmode.hw_strip_crc = 1;
> + conf.rxmode.offloads |= DEV_RX_OFFLOAD_CRC_STRIP;
> }
> + /* Limit configured rss hash functions to only those supported
> + * by the eth device. */
> + conf.rx_adv_conf.rss_conf.rss_hf &= info.flow_type_rss_offloads;
> +
> /* A device may report more queues than it makes available (this
> has
> * been observed for Intel xl710, which reserves some of them for
> * SRIOV): rte_eth_*_queue_setup will fail if a queue is not
> @@ -1932,16 +1935,18 @@ netdev_dpdk_eth_tx_burst(struct netdev_dpdk
> *dev, int qid,
> static inline bool
> netdev_dpdk_policer_pkt_handle(struct rte_meter_srtcm *meter,
> + struct rte_meter_srtcm_profile
> *profile,
> struct rte_mbuf *pkt, uint64_t time)
> {
> uint32_t pkt_len = rte_pktmbuf_pkt_len(pkt) - sizeof(struct
> ether_hdr);
> - return rte_meter_srtcm_color_blind_check(meter, time, pkt_len) ==
> + return rte_meter_srtcm_color_blind_check(meter, profile, time,
> pkt_len) ==
> e_RTE_METER_GREEN;
EC> Indentation looks odd
> }
> static int
> netdev_dpdk_policer_run(struct rte_meter_srtcm *meter,
> + struct rte_meter_srtcm_profile *profile,
> struct rte_mbuf **pkts, int pkt_cnt,
> bool should_steal)
> {
> @@ -1953,7 +1958,8 @@ netdev_dpdk_policer_run(struct rte_meter_srtcm
> *meter,
> for (i = 0; i < pkt_cnt; i++) {
> pkt = pkts[i];
> /* Handle current packet */
> - if (netdev_dpdk_policer_pkt_handle(meter, pkt, current_time))
> {
> + if (netdev_dpdk_policer_pkt_handle(meter, profile, pkt,
> + current_time)) {
EC> Indentation
>
> if (cnt != i) {
> pkts[cnt] = pkt;
> }
> @@ -1975,8 +1981,8 @@ ingress_policer_run(struct ingress_policer
> *policer, struct rte_mbuf **pkts,
> int cnt = 0;
> rte_spinlock_lock(&policer->policer_lock);
> - cnt = netdev_dpdk_policer_run(&policer->in_policer, pkts,
> - pkt_cnt, should_steal);
> + cnt = netdev_dpdk_policer_run(&policer->in_policer,
> &policer->in_prof,
> + pkts, pkt_cnt, should_steal);
> rte_spinlock_unlock(&policer->policer_lock);
> return cnt;
> @@ -2767,8 +2773,15 @@ netdev_dpdk_policer_construct(uint32_t rate,
> uint32_t burst)
> policer->app_srtcm_params.cir = rate_bytes;
> policer->app_srtcm_params.cbs = burst_bytes;
> policer->app_srtcm_params.ebs = 0;
> + err = rte_meter_srtcm_profile_config(&policer->in_prof,
> + &policer->app_srtcm_params);
> + if (err) {
> + VLOG_ERR("Could not create rte meter profile for ingress
> policer");
> + free(policer);
> + return NULL;
> + }
> err = rte_meter_srtcm_config(&policer->in_policer,
> - &policer->app_srtcm_params);
> + &policer->in_prof);
EC> indentation
> if (err) {
> VLOG_ERR("Could not create rte meter for ingress policer");
> free(policer);
> @@ -3043,13 +3056,18 @@ netdev_dpdk_get_status(const struct netdev
> *netdev, struct smap *args)
> smap_add_format(args, "if_descr", "%s %s", rte_version(),
> dev_info.driver_name);
> - if (dev_info.pci_dev) {
> - smap_add_format(args, "pci-vendor_id", "0x%x",
> - dev_info.pci_dev->id.vendor_id);
> - smap_add_format(args, "pci-device_id", "0x%x",
> - dev_info.pci_dev->id.device_id);
> + const struct rte_bus *bus;
> + const struct rte_pci_device *pci_dev;
> + bus = rte_bus_find_by_device(dev_info.device);
> + if (bus && !strcmp(bus->name, "pci")) {
> + pci_dev = RTE_DEV_TO_PCI(dev_info.device);
> + if (pci_dev) {
> + smap_add_format(args, "pci-vendor_id", "0x%u",
> + pci_dev->id.vendor_id);
> + smap_add_format(args, "pci-device_id", "0x%x",
> + pci_dev->id.device_id);
EC> Guess this should be %x, and odd indentation, but I see Ilya got
this also.
> + }
> }
> -
> return 0;
> }
> @@ -3727,6 +3745,7 @@ struct egress_policer {
> struct qos_conf qos_conf;
> struct rte_meter_srtcm_params app_srtcm_params;
> struct rte_meter_srtcm egress_meter;
> + struct rte_meter_srtcm_profile egress_prof;
> };
> static void
> @@ -3749,8 +3768,16 @@ egress_policer_qos_construct(const struct smap
> *details,
> policer = xmalloc(sizeof *policer);
> qos_conf_init(&policer->qos_conf, &egress_policer_ops);
> egress_policer_details_to_param(details,
> &policer->app_srtcm_params);
> + err = rte_meter_srtcm_profile_config(&policer->egress_prof,
> + &policer->app_srtcm_params);
> + if (err) {
> + free(policer);
> + *conf = NULL;
> + return -err;
> + }
> err = rte_meter_srtcm_config(&policer->egress_meter,
> - &policer->app_srtcm_params);
> + &policer->egress_prof);
> +
> if (!err) {
> *conf = &policer->qos_conf;
> } else {
> @@ -3803,7 +3830,8 @@ egress_policer_run(struct qos_conf *conf, struct
> rte_mbuf **pkts, int pkt_cnt,
> struct egress_policer *policer =
> CONTAINER_OF(conf, struct egress_policer, qos_conf);
> - cnt = netdev_dpdk_policer_run(&policer->egress_meter, pkts,
> + cnt = netdev_dpdk_policer_run(&policer->egress_meter,
> + &policer->egress_prof, pkts,
> pkt_cnt, should_steal);
> return cnt;
> @@ -4103,15 +4131,15 @@ dump_flow_pattern(struct rte_flow_item *item)
> VLOG_DBG("rte flow vlan pattern:\n");
> if (vlan_spec) {
> - VLOG_DBG(" Spec: tpid=0x%"PRIx16", tci=0x%"PRIx16"\n",
> - ntohs(vlan_spec->tpid), ntohs(vlan_spec->tci));
> + VLOG_DBG(" Spec: inner_type=0x%"PRIx16",
> tci=0x%"PRIx16"\n",
> + ntohs(vlan_spec->inner_type),
> ntohs(vlan_spec->tci));
> } else {
> VLOG_DBG(" Spec = null\n");
> }
> if (vlan_mask) {
> - VLOG_DBG(" Mask: tpid=0x%"PRIx16", tci=0x%"PRIx16"\n",
> - vlan_mask->tpid, vlan_mask->tci);
> + VLOG_DBG(" Mask: inner_type=0x%"PRIx16",
> tci=0x%"PRIx16"\n",
> + vlan_mask->inner_type, vlan_mask->tci);
EC> is the mask always in host order vs the value being in network
order? Or is this an existing bug?
> } else {
> VLOG_DBG(" Mask = null\n");
> }
> @@ -4281,27 +4309,56 @@ add_flow_action(struct flow_actions *actions,
> enum rte_flow_action_type type,
> actions->cnt++;
> }
> +/*
> + * Storage for struct rte_flow_action_rss
> + * including storage for key and queue array
> + */
> +#define MAX_RSS_HASH_KEY_LENGTH 128
> +#define MAX_ACTION_RSS_QUEUE_NUM 128
> +struct action_rss_data {
> + struct rte_flow_action_rss conf;
> + uint16_t queue[MAX_ACTION_RSS_QUEUE_NUM];
> + uint8_t key[MAX_RSS_HASH_KEY_LENGTH];
> +};
> +
> static struct rte_flow_action_rss *
> add_flow_rss_action(struct flow_actions *actions,
> struct netdev *netdev) {
> int i;
> - struct rte_flow_action_rss *rss;
> + struct action_rss_data *rss_data;
> - rss = xmalloc(sizeof(*rss) + sizeof(uint16_t) * netdev->n_rxq);
> - /*
> - * Setting it to NULL will let the driver use the default RSS
> - * configuration we have set: &port_conf.rx_adv_conf.rss_conf.
> - */
> - rss->rss_conf = NULL;
> - rss->num = netdev->n_rxq;
> + if (netdev->n_rxq > MAX_ACTION_RSS_QUEUE_NUM) {
> + VLOG_ERR("Num of rxq (%u) must not be greater " \
> + "than max rss num of queues (%u)",
> + netdev->n_rxq, MAX_ACTION_RSS_QUEUE_NUM);
> + return NULL;
> + }
> - for (i = 0; i < rss->num; i++) {
> - rss->queue[i] = i;
> + rss_data = xmalloc(sizeof(struct action_rss_data));
> + *rss_data = (struct action_rss_data){
> + .conf = (struct rte_flow_action_rss){
> + .func = RTE_ETH_HASH_FUNCTION_DEFAULT,
> + .level = 0,
> + .types = ETH_RSS_IP,
> + .key_len = 0,
> + .queue_num = netdev->n_rxq,
> + .queue = rss_data->queue,
> + .key = rss_data->key,
> + },
> + .key = { 0 },
> + .queue = { 0 },
> + };
> +
EC> See Ilya's comment on xzalloc() so only non-zero values are needed.
> + /* TODO: Override key with default */
EC> Any reason why this todo is left in?
> + /* Override queue array with default */
> + for (i = 0; i < rss_data->conf.queue_num; i++) {
> + rss_data->queue[i] = i;
> }
> - add_flow_action(actions, RTE_FLOW_ACTION_TYPE_RSS, rss);
> + add_flow_action(actions, RTE_FLOW_ACTION_TYPE_RSS,
> &rss_data->conf);
> - return rss;
> + return &rss_data->conf;
> }
> static int
> @@ -4365,7 +4422,7 @@ netdev_dpdk_add_rte_flow_offload(struct netdev
> *netdev,
> vlan_mask.tci = match->wc.masks.vlans[0].tci &
> ~htons(VLAN_CFI);
> /* match any protocols */
> - vlan_mask.tpid = 0;
> + vlan_mask.inner_type = 0;
> add_flow_pattern(&patterns, RTE_FLOW_ITEM_TYPE_VLAN,
> &vlan_spec, &vlan_mask);
> @@ -4516,6 +4573,11 @@ end_proto_check:
> struct rte_flow_action_rss *rss;
> rss = add_flow_rss_action(&actions, netdev);
> + if (!rss) {
> + VLOG_ERR("add_flow_rss_action error.\n");
> + ret = -1;
> + goto out;
> + }
> add_flow_action(&actions, RTE_FLOW_ACTION_TYPE_END, NULL);
> flow = rte_flow_create(dev->port_id, &flow_attr, patterns.items,
More information about the dev
mailing list