summaryrefslogtreecommitdiffstats
path: root/drivers/net/mlx5
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/net/mlx5')
-rw-r--r--drivers/net/mlx5/Makefile155
-rw-r--r--drivers/net/mlx5/mlx5.c675
-rw-r--r--drivers/net/mlx5/mlx5.h257
-rw-r--r--drivers/net/mlx5/mlx5_defs.h98
-rw-r--r--drivers/net/mlx5/mlx5_ethdev.c1282
-rw-r--r--drivers/net/mlx5/mlx5_fdir.c980
-rw-r--r--drivers/net/mlx5/mlx5_mac.c510
-rw-r--r--drivers/net/mlx5/mlx5_rss.c367
-rw-r--r--drivers/net/mlx5/mlx5_rxmode.c481
-rw-r--r--drivers/net/mlx5/mlx5_rxq.c1555
-rw-r--r--drivers/net/mlx5/mlx5_rxtx.c1249
-rw-r--r--drivers/net/mlx5/mlx5_rxtx.h352
-rw-r--r--drivers/net/mlx5/mlx5_stats.c144
-rw-r--r--drivers/net/mlx5/mlx5_trigger.c126
-rw-r--r--drivers/net/mlx5/mlx5_txq.c605
-rw-r--r--drivers/net/mlx5/mlx5_utils.h184
-rw-r--r--drivers/net/mlx5/mlx5_vlan.c234
-rw-r--r--drivers/net/mlx5/rte_pmd_mlx5_version.map3
18 files changed, 9257 insertions, 0 deletions
diff --git a/drivers/net/mlx5/Makefile b/drivers/net/mlx5/Makefile
new file mode 100644
index 00000000..92bfa070
--- /dev/null
+++ b/drivers/net/mlx5/Makefile
@@ -0,0 +1,155 @@
+# BSD LICENSE
+#
+# Copyright 2015 6WIND S.A.
+# Copyright 2015 Mellanox.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#
+# * Redistributions of source code must retain the above copyright
+# notice, this list of conditions and the following disclaimer.
+# * Redistributions in binary form must reproduce the above copyright
+# notice, this list of conditions and the following disclaimer in
+# the documentation and/or other materials provided with the
+# distribution.
+# * Neither the name of 6WIND S.A. nor the names of its
+# contributors may be used to endorse or promote products derived
+# from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# Library name.
+LIB = librte_pmd_mlx5.a
+
+# Sources.
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_rxq.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_txq.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_rxtx.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_trigger.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_mac.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_rxmode.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_vlan.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_stats.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_rss.c
+SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_fdir.c
+
+# Dependencies.
+DEPDIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += lib/librte_mbuf
+DEPDIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += lib/librte_eal
+DEPDIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += lib/librte_mempool
+
+# Basic CFLAGS.
+CFLAGS += -O3
+CFLAGS += -std=gnu99 -Wall -Wextra
+CFLAGS += -g
+CFLAGS += -I.
+CFLAGS += -D_XOPEN_SOURCE=600
+CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -Wno-strict-prototypes
+LDLIBS += -libverbs
+
+# A few warnings cannot be avoided in external headers.
+CFLAGS += -Wno-error=cast-qual
+
+EXPORT_MAP := rte_pmd_mlx5_version.map
+LIBABIVER := 1
+
+# DEBUG which is usually provided on the command-line may enable
+# CONFIG_RTE_LIBRTE_MLX5_DEBUG.
+ifeq ($(DEBUG),1)
+CONFIG_RTE_LIBRTE_MLX5_DEBUG := y
+endif
+
+# User-defined CFLAGS.
+ifeq ($(CONFIG_RTE_LIBRTE_MLX5_DEBUG),y)
+CFLAGS += -pedantic -UNDEBUG -DPEDANTIC
+else
+CFLAGS += -DNDEBUG -UPEDANTIC
+endif
+
+ifdef CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N
+CFLAGS += -DMLX5_PMD_SGE_WR_N=$(CONFIG_RTE_LIBRTE_MLX5_SGE_WR_N)
+endif
+
+ifdef CONFIG_RTE_LIBRTE_MLX5_MAX_INLINE
+CFLAGS += -DMLX5_PMD_MAX_INLINE=$(CONFIG_RTE_LIBRTE_MLX5_MAX_INLINE)
+endif
+
+ifdef CONFIG_RTE_LIBRTE_MLX5_TX_MP_CACHE
+CFLAGS += -DMLX5_PMD_TX_MP_CACHE=$(CONFIG_RTE_LIBRTE_MLX5_TX_MP_CACHE)
+endif
+
+include $(RTE_SDK)/mk/rte.lib.mk
+
+# Generate and clean-up mlx5_autoconf.h.
+
+export CC CFLAGS CPPFLAGS EXTRA_CFLAGS EXTRA_CPPFLAGS
+export AUTO_CONFIG_CFLAGS = -Wno-error
+
+ifndef V
+AUTOCONF_OUTPUT := >/dev/null
+endif
+
+mlx5_autoconf.h: $(RTE_SDK)/scripts/auto-config-h.sh
+ $Q $(RM) -f -- '$@'
+ $Q sh -- '$<' '$@' \
+ HAVE_EXP_QUERY_DEVICE \
+ infiniband/verbs.h \
+ type 'struct ibv_exp_device_attr' $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_FLOW_SPEC_IPV6 \
+ infiniband/verbs.h \
+ type 'struct ibv_exp_flow_spec_ipv6' $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_EXP_QP_BURST_CREATE_ENABLE_MULTI_PACKET_SEND_WR \
+ infiniband/verbs.h \
+ enum IBV_EXP_QP_BURST_CREATE_ENABLE_MULTI_PACKET_SEND_WR \
+ $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS \
+ infiniband/verbs.h \
+ enum IBV_EXP_DEVICE_ATTR_VLAN_OFFLOADS \
+ $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_EXP_CQ_RX_TCP_PACKET \
+ infiniband/verbs.h \
+ enum IBV_EXP_CQ_RX_TCP_PACKET \
+ $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_VERBS_FCS \
+ infiniband/verbs.h \
+ enum IBV_EXP_CREATE_WQ_FLAG_SCATTER_FCS \
+ $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_VERBS_RX_END_PADDING \
+ infiniband/verbs.h \
+ enum IBV_EXP_CREATE_WQ_FLAG_RX_END_PADDING \
+ $(AUTOCONF_OUTPUT)
+ $Q sh -- '$<' '$@' \
+ HAVE_VERBS_VLAN_INSERTION \
+ infiniband/verbs.h \
+ enum IBV_EXP_RECEIVE_WQ_CVLAN_INSERTION \
+ $(AUTOCONF_OUTPUT)
+
+$(SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD):.c=.o): mlx5_autoconf.h
+
+clean_mlx5: FORCE
+ $Q rm -f -- mlx5_autoconf.h
+
+clean: clean_mlx5
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
new file mode 100644
index 00000000..041cfc33
--- /dev/null
+++ b/drivers/net/mlx5/mlx5.c
@@ -0,0 +1,675 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <unistd.h>
+#include <string.h>
+#include <assert.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <net/if.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_pci.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_utils.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+/**
+ * Retrieve integer value from environment variable.
+ *
+ * @param[in] name
+ * Environment variable name.
+ *
+ * @return
+ * Integer value, 0 if the variable is not set.
+ */
+int
+mlx5_getenv_int(const char *name)
+{
+ const char *val = getenv(name);
+
+ if (val == NULL)
+ return 0;
+ return atoi(val);
+}
+
+/**
+ * DPDK callback to close the device.
+ *
+ * Destroy all queues and objects, free memory.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+static void
+mlx5_dev_close(struct rte_eth_dev *dev)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ void *tmp;
+ unsigned int i;
+
+ priv_lock(priv);
+ DEBUG("%p: closing device \"%s\"",
+ (void *)dev,
+ ((priv->ctx != NULL) ? priv->ctx->device->name : ""));
+ /* In case mlx5_dev_stop() has not been called. */
+ priv_dev_interrupt_handler_uninstall(priv, dev);
+ priv_special_flow_disable_all(priv);
+ priv_mac_addrs_disable(priv);
+ priv_destroy_hash_rxqs(priv);
+
+ /* Remove flow director elements. */
+ priv_fdir_disable(priv);
+ priv_fdir_delete_filters_list(priv);
+
+ /* Prevent crashes when queues are still in use. */
+ dev->rx_pkt_burst = removed_rx_burst;
+ dev->tx_pkt_burst = removed_tx_burst;
+ if (priv->rxqs != NULL) {
+ /* XXX race condition if mlx5_rx_burst() is still running. */
+ usleep(1000);
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ tmp = (*priv->rxqs)[i];
+ if (tmp == NULL)
+ continue;
+ (*priv->rxqs)[i] = NULL;
+ rxq_cleanup(tmp);
+ rte_free(tmp);
+ }
+ priv->rxqs_n = 0;
+ priv->rxqs = NULL;
+ }
+ if (priv->txqs != NULL) {
+ /* XXX race condition if mlx5_tx_burst() is still running. */
+ usleep(1000);
+ for (i = 0; (i != priv->txqs_n); ++i) {
+ tmp = (*priv->txqs)[i];
+ if (tmp == NULL)
+ continue;
+ (*priv->txqs)[i] = NULL;
+ txq_cleanup(tmp);
+ rte_free(tmp);
+ }
+ priv->txqs_n = 0;
+ priv->txqs = NULL;
+ }
+ if (priv->pd != NULL) {
+ assert(priv->ctx != NULL);
+ claim_zero(ibv_dealloc_pd(priv->pd));
+ claim_zero(ibv_close_device(priv->ctx));
+ } else
+ assert(priv->ctx == NULL);
+ if (priv->rss_conf != NULL) {
+ for (i = 0; (i != hash_rxq_init_n); ++i)
+ rte_free((*priv->rss_conf)[i]);
+ rte_free(priv->rss_conf);
+ }
+ if (priv->reta_idx != NULL)
+ rte_free(priv->reta_idx);
+ priv_unlock(priv);
+ memset(priv, 0, sizeof(*priv));
+}
+
+static const struct eth_dev_ops mlx5_dev_ops = {
+ .dev_configure = mlx5_dev_configure,
+ .dev_start = mlx5_dev_start,
+ .dev_stop = mlx5_dev_stop,
+ .dev_set_link_down = mlx5_set_link_down,
+ .dev_set_link_up = mlx5_set_link_up,
+ .dev_close = mlx5_dev_close,
+ .promiscuous_enable = mlx5_promiscuous_enable,
+ .promiscuous_disable = mlx5_promiscuous_disable,
+ .allmulticast_enable = mlx5_allmulticast_enable,
+ .allmulticast_disable = mlx5_allmulticast_disable,
+ .link_update = mlx5_link_update,
+ .stats_get = mlx5_stats_get,
+ .stats_reset = mlx5_stats_reset,
+ .dev_infos_get = mlx5_dev_infos_get,
+ .dev_supported_ptypes_get = mlx5_dev_supported_ptypes_get,
+ .vlan_filter_set = mlx5_vlan_filter_set,
+ .rx_queue_setup = mlx5_rx_queue_setup,
+ .tx_queue_setup = mlx5_tx_queue_setup,
+ .rx_queue_release = mlx5_rx_queue_release,
+ .tx_queue_release = mlx5_tx_queue_release,
+ .flow_ctrl_get = mlx5_dev_get_flow_ctrl,
+ .flow_ctrl_set = mlx5_dev_set_flow_ctrl,
+ .mac_addr_remove = mlx5_mac_addr_remove,
+ .mac_addr_add = mlx5_mac_addr_add,
+ .mac_addr_set = mlx5_mac_addr_set,
+ .mtu_set = mlx5_dev_set_mtu,
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ .vlan_strip_queue_set = mlx5_vlan_strip_queue_set,
+ .vlan_offload_set = mlx5_vlan_offload_set,
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ .reta_update = mlx5_dev_rss_reta_update,
+ .reta_query = mlx5_dev_rss_reta_query,
+ .rss_hash_update = mlx5_rss_hash_update,
+ .rss_hash_conf_get = mlx5_rss_hash_conf_get,
+#ifdef MLX5_FDIR_SUPPORT
+ .filter_ctrl = mlx5_dev_filter_ctrl,
+#endif /* MLX5_FDIR_SUPPORT */
+};
+
+static struct {
+ struct rte_pci_addr pci_addr; /* associated PCI address */
+ uint32_t ports; /* physical ports bitfield. */
+} mlx5_dev[32];
+
+/**
+ * Get device index in mlx5_dev[] from PCI bus address.
+ *
+ * @param[in] pci_addr
+ * PCI bus address to look for.
+ *
+ * @return
+ * mlx5_dev[] index on success, -1 on failure.
+ */
+static int
+mlx5_dev_idx(struct rte_pci_addr *pci_addr)
+{
+ unsigned int i;
+ int ret = -1;
+
+ assert(pci_addr != NULL);
+ for (i = 0; (i != RTE_DIM(mlx5_dev)); ++i) {
+ if ((mlx5_dev[i].pci_addr.domain == pci_addr->domain) &&
+ (mlx5_dev[i].pci_addr.bus == pci_addr->bus) &&
+ (mlx5_dev[i].pci_addr.devid == pci_addr->devid) &&
+ (mlx5_dev[i].pci_addr.function == pci_addr->function))
+ return i;
+ if ((mlx5_dev[i].ports == 0) && (ret == -1))
+ ret = i;
+ }
+ return ret;
+}
+
+static struct eth_driver mlx5_driver;
+
+/**
+ * DPDK callback to register a PCI device.
+ *
+ * This function creates an Ethernet device for each port of a given
+ * PCI device.
+ *
+ * @param[in] pci_drv
+ * PCI driver structure (mlx5_driver).
+ * @param[in] pci_dev
+ * PCI device information.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+static int
+mlx5_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+{
+ struct ibv_device **list;
+ struct ibv_device *ibv_dev;
+ int err = 0;
+ struct ibv_context *attr_ctx = NULL;
+ struct ibv_device_attr device_attr;
+ unsigned int vf;
+ unsigned int mps;
+ int idx;
+ int i;
+
+ (void)pci_drv;
+ assert(pci_drv == &mlx5_driver.pci_drv);
+ /* Get mlx5_dev[] index. */
+ idx = mlx5_dev_idx(&pci_dev->addr);
+ if (idx == -1) {
+ ERROR("this driver cannot support any more adapters");
+ return -ENOMEM;
+ }
+ DEBUG("using driver device index %d", idx);
+
+ /* Save PCI address. */
+ mlx5_dev[idx].pci_addr = pci_dev->addr;
+ list = ibv_get_device_list(&i);
+ if (list == NULL) {
+ assert(errno);
+ if (errno == ENOSYS) {
+ WARN("cannot list devices, is ib_uverbs loaded?");
+ return 0;
+ }
+ return -errno;
+ }
+ assert(i >= 0);
+ /*
+ * For each listed device, check related sysfs entry against
+ * the provided PCI ID.
+ */
+ while (i != 0) {
+ struct rte_pci_addr pci_addr;
+
+ --i;
+ DEBUG("checking device \"%s\"", list[i]->name);
+ if (mlx5_ibv_device_to_pci_addr(list[i], &pci_addr))
+ continue;
+ if ((pci_dev->addr.domain != pci_addr.domain) ||
+ (pci_dev->addr.bus != pci_addr.bus) ||
+ (pci_dev->addr.devid != pci_addr.devid) ||
+ (pci_dev->addr.function != pci_addr.function))
+ continue;
+ vf = ((pci_dev->id.device_id ==
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4VF) ||
+ (pci_dev->id.device_id ==
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4LXVF));
+ /* Multi-packet send is only supported by ConnectX-4 Lx PF. */
+ mps = (pci_dev->id.device_id ==
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4LX);
+ INFO("PCI information matches, using device \"%s\" (VF: %s,"
+ " MPS: %s)",
+ list[i]->name,
+ vf ? "true" : "false",
+ mps ? "true" : "false");
+ attr_ctx = ibv_open_device(list[i]);
+ err = errno;
+ break;
+ }
+ if (attr_ctx == NULL) {
+ ibv_free_device_list(list);
+ switch (err) {
+ case 0:
+ WARN("cannot access device, is mlx5_ib loaded?");
+ return 0;
+ case EINVAL:
+ WARN("cannot use device, are drivers up to date?");
+ return 0;
+ }
+ assert(err > 0);
+ return -err;
+ }
+ ibv_dev = list[i];
+
+ DEBUG("device opened");
+ if (ibv_query_device(attr_ctx, &device_attr))
+ goto error;
+ INFO("%u port(s) detected", device_attr.phys_port_cnt);
+
+ for (i = 0; i < device_attr.phys_port_cnt; i++) {
+ uint32_t port = i + 1; /* ports are indexed from one */
+ uint32_t test = (1 << i);
+ struct ibv_context *ctx = NULL;
+ struct ibv_port_attr port_attr;
+ struct ibv_pd *pd = NULL;
+ struct priv *priv = NULL;
+ struct rte_eth_dev *eth_dev;
+#ifdef HAVE_EXP_QUERY_DEVICE
+ struct ibv_exp_device_attr exp_device_attr;
+#endif /* HAVE_EXP_QUERY_DEVICE */
+ struct ether_addr mac;
+
+#ifdef HAVE_EXP_QUERY_DEVICE
+ exp_device_attr.comp_mask =
+ IBV_EXP_DEVICE_ATTR_EXP_CAP_FLAGS |
+ IBV_EXP_DEVICE_ATTR_RX_HASH |
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ IBV_EXP_DEVICE_ATTR_VLAN_OFFLOADS |
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+#ifdef HAVE_VERBS_RX_END_PADDING
+ IBV_EXP_DEVICE_ATTR_RX_PAD_END_ALIGN |
+#endif /* HAVE_VERBS_RX_END_PADDING */
+ 0;
+#endif /* HAVE_EXP_QUERY_DEVICE */
+
+ DEBUG("using port %u (%08" PRIx32 ")", port, test);
+
+ ctx = ibv_open_device(ibv_dev);
+ if (ctx == NULL)
+ goto port_error;
+
+ /* Check port status. */
+ err = ibv_query_port(ctx, port, &port_attr);
+ if (err) {
+ ERROR("port query failed: %s", strerror(err));
+ goto port_error;
+ }
+
+ if (port_attr.link_layer != IBV_LINK_LAYER_ETHERNET) {
+ ERROR("port %d is not configured in Ethernet mode",
+ port);
+ goto port_error;
+ }
+
+ if (port_attr.state != IBV_PORT_ACTIVE)
+ DEBUG("port %d is not active: \"%s\" (%d)",
+ port, ibv_port_state_str(port_attr.state),
+ port_attr.state);
+
+ /* Allocate protection domain. */
+ pd = ibv_alloc_pd(ctx);
+ if (pd == NULL) {
+ ERROR("PD allocation failure");
+ err = ENOMEM;
+ goto port_error;
+ }
+
+ mlx5_dev[idx].ports |= test;
+
+ /* from rte_ethdev.c */
+ priv = rte_zmalloc("ethdev private structure",
+ sizeof(*priv),
+ RTE_CACHE_LINE_SIZE);
+ if (priv == NULL) {
+ ERROR("priv allocation failure");
+ err = ENOMEM;
+ goto port_error;
+ }
+
+ priv->ctx = ctx;
+ priv->device_attr = device_attr;
+ priv->port = port;
+ priv->pd = pd;
+ priv->mtu = ETHER_MTU;
+#ifdef HAVE_EXP_QUERY_DEVICE
+ if (ibv_exp_query_device(ctx, &exp_device_attr)) {
+ ERROR("ibv_exp_query_device() failed");
+ goto port_error;
+ }
+
+ priv->hw_csum =
+ ((exp_device_attr.exp_device_cap_flags &
+ IBV_EXP_DEVICE_RX_CSUM_TCP_UDP_PKT) &&
+ (exp_device_attr.exp_device_cap_flags &
+ IBV_EXP_DEVICE_RX_CSUM_IP_PKT));
+ DEBUG("checksum offloading is %ssupported",
+ (priv->hw_csum ? "" : "not "));
+
+ priv->hw_csum_l2tun = !!(exp_device_attr.exp_device_cap_flags &
+ IBV_EXP_DEVICE_VXLAN_SUPPORT);
+ DEBUG("L2 tunnel checksum offloads are %ssupported",
+ (priv->hw_csum_l2tun ? "" : "not "));
+
+ priv->ind_table_max_size = exp_device_attr.rx_hash_caps.max_rwq_indirection_table_size;
+ /* Remove this check once DPDK supports larger/variable
+ * indirection tables. */
+ if (priv->ind_table_max_size > (unsigned int)RSS_INDIRECTION_TABLE_SIZE)
+ priv->ind_table_max_size = RSS_INDIRECTION_TABLE_SIZE;
+ DEBUG("maximum RX indirection table size is %u",
+ priv->ind_table_max_size);
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ priv->hw_vlan_strip = !!(exp_device_attr.wq_vlan_offloads_cap &
+ IBV_EXP_RECEIVE_WQ_CVLAN_STRIP);
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ DEBUG("VLAN stripping is %ssupported",
+ (priv->hw_vlan_strip ? "" : "not "));
+
+#ifdef HAVE_VERBS_FCS
+ priv->hw_fcs_strip = !!(exp_device_attr.exp_device_cap_flags &
+ IBV_EXP_DEVICE_SCATTER_FCS);
+#endif /* HAVE_VERBS_FCS */
+ DEBUG("FCS stripping configuration is %ssupported",
+ (priv->hw_fcs_strip ? "" : "not "));
+
+#ifdef HAVE_VERBS_RX_END_PADDING
+ priv->hw_padding = !!exp_device_attr.rx_pad_end_addr_align;
+#endif /* HAVE_VERBS_RX_END_PADDING */
+ DEBUG("hardware RX end alignment padding is %ssupported",
+ (priv->hw_padding ? "" : "not "));
+
+#else /* HAVE_EXP_QUERY_DEVICE */
+ priv->ind_table_max_size = RSS_INDIRECTION_TABLE_SIZE;
+#endif /* HAVE_EXP_QUERY_DEVICE */
+
+ priv->vf = vf;
+ priv->mps = mps;
+ /* Allocate and register default RSS hash keys. */
+ priv->rss_conf = rte_calloc(__func__, hash_rxq_init_n,
+ sizeof((*priv->rss_conf)[0]), 0);
+ if (priv->rss_conf == NULL) {
+ err = ENOMEM;
+ goto port_error;
+ }
+ err = rss_hash_rss_conf_new_key(priv,
+ rss_hash_default_key,
+ rss_hash_default_key_len,
+ ETH_RSS_PROTO_MASK);
+ if (err)
+ goto port_error;
+ /* Configure the first MAC address by default. */
+ if (priv_get_mac(priv, &mac.addr_bytes)) {
+ ERROR("cannot get MAC address, is mlx5_en loaded?"
+ " (errno: %s)", strerror(errno));
+ goto port_error;
+ }
+ INFO("port %u MAC address is %02x:%02x:%02x:%02x:%02x:%02x",
+ priv->port,
+ mac.addr_bytes[0], mac.addr_bytes[1],
+ mac.addr_bytes[2], mac.addr_bytes[3],
+ mac.addr_bytes[4], mac.addr_bytes[5]);
+ /* Register MAC address. */
+ claim_zero(priv_mac_addr_add(priv, 0,
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ mac.addr_bytes));
+ /* Initialize FD filters list. */
+ err = fdir_init_filters_list(priv);
+ if (err)
+ goto port_error;
+#ifndef NDEBUG
+ {
+ char ifname[IF_NAMESIZE];
+
+ if (priv_get_ifname(priv, &ifname) == 0)
+ DEBUG("port %u ifname is \"%s\"",
+ priv->port, ifname);
+ else
+ DEBUG("port %u ifname is unknown", priv->port);
+ }
+#endif
+ /* Get actual MTU if possible. */
+ priv_get_mtu(priv, &priv->mtu);
+ DEBUG("port %u MTU is %u", priv->port, priv->mtu);
+
+ /* from rte_ethdev.c */
+ {
+ char name[RTE_ETH_NAME_MAX_LEN];
+
+ snprintf(name, sizeof(name), "%s port %u",
+ ibv_get_device_name(ibv_dev), port);
+ eth_dev = rte_eth_dev_allocate(name, RTE_ETH_DEV_PCI);
+ }
+ if (eth_dev == NULL) {
+ ERROR("can not allocate rte ethdev");
+ err = ENOMEM;
+ goto port_error;
+ }
+
+ /* Secondary processes have to use local storage for their
+ * private data as well as a copy of eth_dev->data, but this
+ * pointer must not be modified before burst functions are
+ * actually called. */
+ if (mlx5_is_secondary()) {
+ struct mlx5_secondary_data *sd =
+ &mlx5_secondary_data[eth_dev->data->port_id];
+ sd->primary_priv = eth_dev->data->dev_private;
+ if (sd->primary_priv == NULL) {
+ ERROR("no private data for port %u",
+ eth_dev->data->port_id);
+ err = EINVAL;
+ goto port_error;
+ }
+ sd->shared_dev_data = eth_dev->data;
+ rte_spinlock_init(&sd->lock);
+ memcpy(sd->data.name, sd->shared_dev_data->name,
+ sizeof(sd->data.name));
+ sd->data.dev_private = priv;
+ sd->data.rx_mbuf_alloc_failed = 0;
+ sd->data.mtu = ETHER_MTU;
+ sd->data.port_id = sd->shared_dev_data->port_id;
+ sd->data.mac_addrs = priv->mac;
+ eth_dev->tx_pkt_burst = mlx5_tx_burst_secondary_setup;
+ eth_dev->rx_pkt_burst = mlx5_rx_burst_secondary_setup;
+ } else {
+ eth_dev->data->dev_private = priv;
+ eth_dev->data->rx_mbuf_alloc_failed = 0;
+ eth_dev->data->mtu = ETHER_MTU;
+ eth_dev->data->mac_addrs = priv->mac;
+ }
+
+ eth_dev->pci_dev = pci_dev;
+ rte_eth_copy_pci_info(eth_dev, pci_dev);
+ eth_dev->driver = &mlx5_driver;
+ priv->dev = eth_dev;
+ eth_dev->dev_ops = &mlx5_dev_ops;
+
+ TAILQ_INIT(&eth_dev->link_intr_cbs);
+
+ /* Bring Ethernet device up. */
+ DEBUG("forcing Ethernet interface up");
+ priv_set_flags(priv, ~IFF_UP, IFF_UP);
+ continue;
+
+port_error:
+ if (priv) {
+ rte_free(priv->rss_conf);
+ rte_free(priv);
+ }
+ if (pd)
+ claim_zero(ibv_dealloc_pd(pd));
+ if (ctx)
+ claim_zero(ibv_close_device(ctx));
+ break;
+ }
+
+ /*
+ * XXX if something went wrong in the loop above, there is a resource
+ * leak (ctx, pd, priv, dpdk ethdev) but we can do nothing about it as
+ * long as the dpdk does not provide a way to deallocate a ethdev and a
+ * way to enumerate the registered ethdevs to free the previous ones.
+ */
+
+ /* no port found, complain */
+ if (!mlx5_dev[idx].ports) {
+ err = ENODEV;
+ goto error;
+ }
+
+error:
+ if (attr_ctx)
+ claim_zero(ibv_close_device(attr_ctx));
+ if (list)
+ ibv_free_device_list(list);
+ assert(err >= 0);
+ return -err;
+}
+
+static const struct rte_pci_id mlx5_pci_id_map[] = {
+ {
+ .vendor_id = PCI_VENDOR_ID_MELLANOX,
+ .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX4,
+ .subsystem_vendor_id = PCI_ANY_ID,
+ .subsystem_device_id = PCI_ANY_ID
+ },
+ {
+ .vendor_id = PCI_VENDOR_ID_MELLANOX,
+ .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX4VF,
+ .subsystem_vendor_id = PCI_ANY_ID,
+ .subsystem_device_id = PCI_ANY_ID
+ },
+ {
+ .vendor_id = PCI_VENDOR_ID_MELLANOX,
+ .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX4LX,
+ .subsystem_vendor_id = PCI_ANY_ID,
+ .subsystem_device_id = PCI_ANY_ID
+ },
+ {
+ .vendor_id = PCI_VENDOR_ID_MELLANOX,
+ .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX4LXVF,
+ .subsystem_vendor_id = PCI_ANY_ID,
+ .subsystem_device_id = PCI_ANY_ID
+ },
+ {
+ .vendor_id = 0
+ }
+};
+
+static struct eth_driver mlx5_driver = {
+ .pci_drv = {
+ .name = MLX5_DRIVER_NAME,
+ .id_table = mlx5_pci_id_map,
+ .devinit = mlx5_pci_devinit,
+ .drv_flags = RTE_PCI_DRV_INTR_LSC,
+ },
+ .dev_private_size = sizeof(struct priv)
+};
+
+/**
+ * Driver initialization routine.
+ */
+static int
+rte_mlx5_pmd_init(const char *name, const char *args)
+{
+ (void)name;
+ (void)args;
+ /*
+ * RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use
+ * huge pages. Calling ibv_fork_init() during init allows
+ * applications to use fork() safely for purposes other than
+ * using this PMD, which is not supported in forked processes.
+ */
+ setenv("RDMAV_HUGEPAGES_SAFE", "1", 1);
+ ibv_fork_init();
+ rte_eal_pci_register(&mlx5_driver.pci_drv);
+ return 0;
+}
+
+static struct rte_driver rte_mlx5_driver = {
+ .type = PMD_PDEV,
+ .name = MLX5_DRIVER_NAME,
+ .init = rte_mlx5_pmd_init,
+};
+
+PMD_REGISTER_DRIVER(rte_mlx5_driver)
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
new file mode 100644
index 00000000..24876625
--- /dev/null
+++ b/drivers/net/mlx5/mlx5.h
@@ -0,0 +1,257 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef RTE_PMD_MLX5_H_
+#define RTE_PMD_MLX5_H_
+
+#include <stddef.h>
+#include <stdint.h>
+#include <limits.h>
+#include <net/if.h>
+#include <netinet/in.h>
+#include <linux/if.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_spinlock.h>
+#include <rte_interrupts.h>
+#include <rte_errno.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5_utils.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+enum {
+ PCI_VENDOR_ID_MELLANOX = 0x15b3,
+};
+
+enum {
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4 = 0x1013,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4VF = 0x1014,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4LX = 0x1015,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX4LXVF = 0x1016,
+};
+
+struct priv {
+ struct rte_eth_dev *dev; /* Ethernet device. */
+ struct ibv_context *ctx; /* Verbs context. */
+ struct ibv_device_attr device_attr; /* Device properties. */
+ struct ibv_pd *pd; /* Protection Domain. */
+ /*
+ * MAC addresses array and configuration bit-field.
+ * An extra entry that cannot be modified by the DPDK is reserved
+ * for broadcast frames (destination MAC address ff:ff:ff:ff:ff:ff).
+ */
+ struct ether_addr mac[MLX5_MAX_MAC_ADDRESSES];
+ BITFIELD_DECLARE(mac_configured, uint32_t, MLX5_MAX_MAC_ADDRESSES);
+ uint16_t vlan_filter[MLX5_MAX_VLAN_IDS]; /* VLAN filters table. */
+ unsigned int vlan_filter_n; /* Number of configured VLAN filters. */
+ /* Device properties. */
+ uint16_t mtu; /* Configured MTU. */
+ uint8_t port; /* Physical port number. */
+ unsigned int started:1; /* Device started, flows enabled. */
+ unsigned int promisc_req:1; /* Promiscuous mode requested. */
+ unsigned int allmulti_req:1; /* All multicast mode requested. */
+ unsigned int hw_csum:1; /* Checksum offload is supported. */
+ unsigned int hw_csum_l2tun:1; /* Same for L2 tunnels. */
+ unsigned int hw_vlan_strip:1; /* VLAN stripping is supported. */
+ unsigned int hw_fcs_strip:1; /* FCS stripping is supported. */
+ unsigned int hw_padding:1; /* End alignment padding is supported. */
+ unsigned int vf:1; /* This is a VF device. */
+ unsigned int mps:1; /* Whether multi-packet send is supported. */
+ unsigned int pending_alarm:1; /* An alarm is pending. */
+ /* RX/TX queues. */
+ unsigned int rxqs_n; /* RX queues array size. */
+ unsigned int txqs_n; /* TX queues array size. */
+ struct rxq *(*rxqs)[]; /* RX queues. */
+ struct txq *(*txqs)[]; /* TX queues. */
+ /* Indirection tables referencing all RX WQs. */
+ struct ibv_exp_rwq_ind_table *(*ind_tables)[];
+ unsigned int ind_tables_n; /* Number of indirection tables. */
+ unsigned int ind_table_max_size; /* Maximum indirection table size. */
+ /* Hash RX QPs feeding the indirection table. */
+ struct hash_rxq (*hash_rxqs)[];
+ unsigned int hash_rxqs_n; /* Hash RX QPs array size. */
+ /* RSS configuration array indexed by hash RX queue type. */
+ struct rte_eth_rss_conf *(*rss_conf)[];
+ uint64_t rss_hf; /* RSS DPDK bit field of active RSS. */
+ struct rte_intr_handle intr_handle; /* Interrupt handler. */
+ unsigned int (*reta_idx)[]; /* RETA index table. */
+ unsigned int reta_idx_n; /* RETA index size. */
+ struct fdir_filter_list *fdir_filter_list; /* Flow director rules. */
+ rte_spinlock_t lock; /* Lock for control functions. */
+};
+
+/* Local storage for secondary process data. */
+struct mlx5_secondary_data {
+ struct rte_eth_dev_data data; /* Local device data. */
+ struct priv *primary_priv; /* Private structure from primary. */
+ struct rte_eth_dev_data *shared_dev_data; /* Shared device data. */
+ rte_spinlock_t lock; /* Port configuration lock. */
+} mlx5_secondary_data[RTE_MAX_ETHPORTS];
+
+/**
+ * Lock private structure to protect it from concurrent access in the
+ * control path.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+static inline void
+priv_lock(struct priv *priv)
+{
+ rte_spinlock_lock(&priv->lock);
+}
+
+/**
+ * Unlock private structure.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+static inline void
+priv_unlock(struct priv *priv)
+{
+ rte_spinlock_unlock(&priv->lock);
+}
+
+/* mlx5.c */
+
+int mlx5_getenv_int(const char *);
+
+/* mlx5_ethdev.c */
+
+struct priv *mlx5_get_priv(struct rte_eth_dev *dev);
+int mlx5_is_secondary(void);
+int priv_get_ifname(const struct priv *, char (*)[IF_NAMESIZE]);
+int priv_ifreq(const struct priv *, int req, struct ifreq *);
+int priv_get_mtu(struct priv *, uint16_t *);
+int priv_set_flags(struct priv *, unsigned int, unsigned int);
+int mlx5_dev_configure(struct rte_eth_dev *);
+void mlx5_dev_infos_get(struct rte_eth_dev *, struct rte_eth_dev_info *);
+const uint32_t *mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev);
+int mlx5_link_update(struct rte_eth_dev *, int);
+int mlx5_dev_set_mtu(struct rte_eth_dev *, uint16_t);
+int mlx5_dev_get_flow_ctrl(struct rte_eth_dev *, struct rte_eth_fc_conf *);
+int mlx5_dev_set_flow_ctrl(struct rte_eth_dev *, struct rte_eth_fc_conf *);
+int mlx5_ibv_device_to_pci_addr(const struct ibv_device *,
+ struct rte_pci_addr *);
+void mlx5_dev_link_status_handler(void *);
+void mlx5_dev_interrupt_handler(struct rte_intr_handle *, void *);
+void priv_dev_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *);
+void priv_dev_interrupt_handler_install(struct priv *, struct rte_eth_dev *);
+int mlx5_set_link_down(struct rte_eth_dev *dev);
+int mlx5_set_link_up(struct rte_eth_dev *dev);
+struct priv *mlx5_secondary_data_setup(struct priv *priv);
+
+/* mlx5_mac.c */
+
+int priv_get_mac(struct priv *, uint8_t (*)[ETHER_ADDR_LEN]);
+void hash_rxq_mac_addrs_del(struct hash_rxq *);
+void priv_mac_addrs_disable(struct priv *);
+void mlx5_mac_addr_remove(struct rte_eth_dev *, uint32_t);
+int hash_rxq_mac_addrs_add(struct hash_rxq *);
+int priv_mac_addr_add(struct priv *, unsigned int,
+ const uint8_t (*)[ETHER_ADDR_LEN]);
+int priv_mac_addrs_enable(struct priv *);
+void mlx5_mac_addr_add(struct rte_eth_dev *, struct ether_addr *, uint32_t,
+ uint32_t);
+void mlx5_mac_addr_set(struct rte_eth_dev *, struct ether_addr *);
+
+/* mlx5_rss.c */
+
+int rss_hash_rss_conf_new_key(struct priv *, const uint8_t *, unsigned int,
+ uint64_t);
+int mlx5_rss_hash_update(struct rte_eth_dev *, struct rte_eth_rss_conf *);
+int mlx5_rss_hash_conf_get(struct rte_eth_dev *, struct rte_eth_rss_conf *);
+int priv_rss_reta_index_resize(struct priv *, unsigned int);
+int mlx5_dev_rss_reta_query(struct rte_eth_dev *,
+ struct rte_eth_rss_reta_entry64 *, uint16_t);
+int mlx5_dev_rss_reta_update(struct rte_eth_dev *,
+ struct rte_eth_rss_reta_entry64 *, uint16_t);
+
+/* mlx5_rxmode.c */
+
+int priv_special_flow_enable(struct priv *, enum hash_rxq_flow_type);
+void priv_special_flow_disable(struct priv *, enum hash_rxq_flow_type);
+int priv_special_flow_enable_all(struct priv *);
+void priv_special_flow_disable_all(struct priv *);
+void mlx5_promiscuous_enable(struct rte_eth_dev *);
+void mlx5_promiscuous_disable(struct rte_eth_dev *);
+void mlx5_allmulticast_enable(struct rte_eth_dev *);
+void mlx5_allmulticast_disable(struct rte_eth_dev *);
+
+/* mlx5_stats.c */
+
+void mlx5_stats_get(struct rte_eth_dev *, struct rte_eth_stats *);
+void mlx5_stats_reset(struct rte_eth_dev *);
+
+/* mlx5_vlan.c */
+
+int mlx5_vlan_filter_set(struct rte_eth_dev *, uint16_t, int);
+void mlx5_vlan_offload_set(struct rte_eth_dev *, int);
+void mlx5_vlan_strip_queue_set(struct rte_eth_dev *, uint16_t, int);
+
+/* mlx5_trigger.c */
+
+int mlx5_dev_start(struct rte_eth_dev *);
+void mlx5_dev_stop(struct rte_eth_dev *);
+
+/* mlx5_fdir.c */
+
+int fdir_init_filters_list(struct priv *);
+void priv_fdir_delete_filters_list(struct priv *);
+void priv_fdir_disable(struct priv *);
+void priv_fdir_enable(struct priv *);
+int mlx5_dev_filter_ctrl(struct rte_eth_dev *, enum rte_filter_type,
+ enum rte_filter_op, void *);
+
+#endif /* RTE_PMD_MLX5_H_ */
diff --git a/drivers/net/mlx5/mlx5_defs.h b/drivers/net/mlx5/mlx5_defs.h
new file mode 100644
index 00000000..09207d9c
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_defs.h
@@ -0,0 +1,98 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef RTE_PMD_MLX5_DEFS_H_
+#define RTE_PMD_MLX5_DEFS_H_
+
+#include "mlx5_autoconf.h"
+
+/* Reported driver name. */
+#define MLX5_DRIVER_NAME "librte_pmd_mlx5"
+
+/* Maximum number of simultaneous MAC addresses. */
+#define MLX5_MAX_MAC_ADDRESSES 128
+
+/* Maximum number of simultaneous VLAN filters. */
+#define MLX5_MAX_VLAN_IDS 128
+
+/* Maximum number of special flows. */
+#define MLX5_MAX_SPECIAL_FLOWS 4
+
+/* Request send completion once in every 64 sends, might be less. */
+#define MLX5_PMD_TX_PER_COMP_REQ 64
+
+/* RSS Indirection table size. */
+#define RSS_INDIRECTION_TABLE_SIZE 256
+
+/* Maximum number of Scatter/Gather Elements per Work Request. */
+#ifndef MLX5_PMD_SGE_WR_N
+#define MLX5_PMD_SGE_WR_N 4
+#endif
+
+/* Maximum size for inline data. */
+#ifndef MLX5_PMD_MAX_INLINE
+#define MLX5_PMD_MAX_INLINE 0
+#endif
+
+/*
+ * Maximum number of cached Memory Pools (MPs) per TX queue. Each RTE MP
+ * from which buffers are to be transmitted will have to be mapped by this
+ * driver to their own Memory Region (MR). This is a slow operation.
+ *
+ * This value is always 1 for RX queues.
+ */
+#ifndef MLX5_PMD_TX_MP_CACHE
+#define MLX5_PMD_TX_MP_CACHE 8
+#endif
+
+/*
+ * If defined, only use software counters. The PMD will never ask the hardware
+ * for these, and many of them won't be available.
+ */
+#ifndef MLX5_PMD_SOFT_COUNTERS
+#define MLX5_PMD_SOFT_COUNTERS 1
+#endif
+
+/* Alarm timeout. */
+#define MLX5_ALARM_TIMEOUT_US 100000
+
+/*
+ * Extended flow priorities necessary to support flow director are available
+ * since MLNX_OFED 3.2. Considering this version adds support for VLAN
+ * offloads as well, their availability means flow director can be used.
+ */
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+#define MLX5_FDIR_SUPPORT 1
+#endif
+
+#endif /* RTE_PMD_MLX5_DEFS_H_ */
diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
new file mode 100644
index 00000000..36b369e7
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -0,0 +1,1282 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <assert.h>
+#include <unistd.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <dirent.h>
+#include <net/if.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <linux/if.h>
+#include <linux/ethtool.h>
+#include <linux/sockios.h>
+#include <fcntl.h>
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_atomic.h>
+#include <rte_ethdev.h>
+#include <rte_mbuf.h>
+#include <rte_common.h>
+#include <rte_interrupts.h>
+#include <rte_alarm.h>
+#include <rte_malloc.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_utils.h"
+
+/**
+ * Return private structure associated with an Ethernet device.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * Pointer to private structure.
+ */
+struct priv *
+mlx5_get_priv(struct rte_eth_dev *dev)
+{
+ struct mlx5_secondary_data *sd;
+
+ if (!mlx5_is_secondary())
+ return dev->data->dev_private;
+ sd = &mlx5_secondary_data[dev->data->port_id];
+ return sd->data.dev_private;
+}
+
+/**
+ * Check if running as a secondary process.
+ *
+ * @return
+ * Nonzero if running as a secondary process.
+ */
+inline int
+mlx5_is_secondary(void)
+{
+ return rte_eal_process_type() != RTE_PROC_PRIMARY;
+}
+
+/**
+ * Get interface name from private structure.
+ *
+ * @param[in] priv
+ * Pointer to private structure.
+ * @param[out] ifname
+ * Interface name output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_ifname(const struct priv *priv, char (*ifname)[IF_NAMESIZE])
+{
+ DIR *dir;
+ struct dirent *dent;
+ unsigned int dev_type = 0;
+ unsigned int dev_port_prev = ~0u;
+ char match[IF_NAMESIZE] = "";
+
+ {
+ MKSTR(path, "%s/device/net", priv->ctx->device->ibdev_path);
+
+ dir = opendir(path);
+ if (dir == NULL)
+ return -1;
+ }
+ while ((dent = readdir(dir)) != NULL) {
+ char *name = dent->d_name;
+ FILE *file;
+ unsigned int dev_port;
+ int r;
+
+ if ((name[0] == '.') &&
+ ((name[1] == '\0') ||
+ ((name[1] == '.') && (name[2] == '\0'))))
+ continue;
+
+ MKSTR(path, "%s/device/net/%s/%s",
+ priv->ctx->device->ibdev_path, name,
+ (dev_type ? "dev_id" : "dev_port"));
+
+ file = fopen(path, "rb");
+ if (file == NULL) {
+ if (errno != ENOENT)
+ continue;
+ /*
+ * Switch to dev_id when dev_port does not exist as
+ * is the case with Linux kernel versions < 3.15.
+ */
+try_dev_id:
+ match[0] = '\0';
+ if (dev_type)
+ break;
+ dev_type = 1;
+ dev_port_prev = ~0u;
+ rewinddir(dir);
+ continue;
+ }
+ r = fscanf(file, (dev_type ? "%x" : "%u"), &dev_port);
+ fclose(file);
+ if (r != 1)
+ continue;
+ /*
+ * Switch to dev_id when dev_port returns the same value for
+ * all ports. May happen when using a MOFED release older than
+ * 3.0 with a Linux kernel >= 3.15.
+ */
+ if (dev_port == dev_port_prev)
+ goto try_dev_id;
+ dev_port_prev = dev_port;
+ if (dev_port == (priv->port - 1u))
+ snprintf(match, sizeof(match), "%s", name);
+ }
+ closedir(dir);
+ if (match[0] == '\0')
+ return -1;
+ strncpy(*ifname, match, sizeof(*ifname));
+ return 0;
+}
+
+/**
+ * Read from sysfs entry.
+ *
+ * @param[in] priv
+ * Pointer to private structure.
+ * @param[in] entry
+ * Entry name relative to sysfs path.
+ * @param[out] buf
+ * Data output buffer.
+ * @param size
+ * Buffer size.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+static int
+priv_sysfs_read(const struct priv *priv, const char *entry,
+ char *buf, size_t size)
+{
+ char ifname[IF_NAMESIZE];
+ FILE *file;
+ int ret;
+ int err;
+
+ if (priv_get_ifname(priv, &ifname))
+ return -1;
+
+ MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path,
+ ifname, entry);
+
+ file = fopen(path, "rb");
+ if (file == NULL)
+ return -1;
+ ret = fread(buf, 1, size, file);
+ err = errno;
+ if (((size_t)ret < size) && (ferror(file)))
+ ret = -1;
+ else
+ ret = size;
+ fclose(file);
+ errno = err;
+ return ret;
+}
+
+/**
+ * Write to sysfs entry.
+ *
+ * @param[in] priv
+ * Pointer to private structure.
+ * @param[in] entry
+ * Entry name relative to sysfs path.
+ * @param[in] buf
+ * Data buffer.
+ * @param size
+ * Buffer size.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+static int
+priv_sysfs_write(const struct priv *priv, const char *entry,
+ char *buf, size_t size)
+{
+ char ifname[IF_NAMESIZE];
+ FILE *file;
+ int ret;
+ int err;
+
+ if (priv_get_ifname(priv, &ifname))
+ return -1;
+
+ MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path,
+ ifname, entry);
+
+ file = fopen(path, "wb");
+ if (file == NULL)
+ return -1;
+ ret = fwrite(buf, 1, size, file);
+ err = errno;
+ if (((size_t)ret < size) || (ferror(file)))
+ ret = -1;
+ else
+ ret = size;
+ fclose(file);
+ errno = err;
+ return ret;
+}
+
+/**
+ * Get unsigned long sysfs property.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[in] name
+ * Entry name relative to sysfs path.
+ * @param[out] value
+ * Value output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+static int
+priv_get_sysfs_ulong(struct priv *priv, const char *name, unsigned long *value)
+{
+ int ret;
+ unsigned long value_ret;
+ char value_str[32];
+
+ ret = priv_sysfs_read(priv, name, value_str, (sizeof(value_str) - 1));
+ if (ret == -1) {
+ DEBUG("cannot read %s value from sysfs: %s",
+ name, strerror(errno));
+ return -1;
+ }
+ value_str[ret] = '\0';
+ errno = 0;
+ value_ret = strtoul(value_str, NULL, 0);
+ if (errno) {
+ DEBUG("invalid %s value `%s': %s", name, value_str,
+ strerror(errno));
+ return -1;
+ }
+ *value = value_ret;
+ return 0;
+}
+
+/**
+ * Set unsigned long sysfs property.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[in] name
+ * Entry name relative to sysfs path.
+ * @param value
+ * Value to set.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+static int
+priv_set_sysfs_ulong(struct priv *priv, const char *name, unsigned long value)
+{
+ int ret;
+ MKSTR(value_str, "%lu", value);
+
+ ret = priv_sysfs_write(priv, name, value_str, (sizeof(value_str) - 1));
+ if (ret == -1) {
+ DEBUG("cannot write %s `%s' (%lu) to sysfs: %s",
+ name, value_str, value, strerror(errno));
+ return -1;
+ }
+ return 0;
+}
+
+/**
+ * Perform ifreq ioctl() on associated Ethernet device.
+ *
+ * @param[in] priv
+ * Pointer to private structure.
+ * @param req
+ * Request number to pass to ioctl().
+ * @param[out] ifr
+ * Interface request structure output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_ifreq(const struct priv *priv, int req, struct ifreq *ifr)
+{
+ int sock = socket(PF_INET, SOCK_DGRAM, IPPROTO_IP);
+ int ret = -1;
+
+ if (sock == -1)
+ return ret;
+ if (priv_get_ifname(priv, &ifr->ifr_name) == 0)
+ ret = ioctl(sock, req, ifr);
+ close(sock);
+ return ret;
+}
+
+/**
+ * Get device MTU.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[out] mtu
+ * MTU value output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_mtu(struct priv *priv, uint16_t *mtu)
+{
+ unsigned long ulong_mtu;
+
+ if (priv_get_sysfs_ulong(priv, "mtu", &ulong_mtu) == -1)
+ return -1;
+ *mtu = ulong_mtu;
+ return 0;
+}
+
+/**
+ * Set device MTU.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param mtu
+ * MTU value to set.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+static int
+priv_set_mtu(struct priv *priv, uint16_t mtu)
+{
+ return priv_set_sysfs_ulong(priv, "mtu", mtu);
+}
+
+/**
+ * Set device flags.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param keep
+ * Bitmask for flags that must remain untouched.
+ * @param flags
+ * Bitmask for flags to modify.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_set_flags(struct priv *priv, unsigned int keep, unsigned int flags)
+{
+ unsigned long tmp;
+
+ if (priv_get_sysfs_ulong(priv, "flags", &tmp) == -1)
+ return -1;
+ tmp &= keep;
+ tmp |= flags;
+ return priv_set_sysfs_ulong(priv, "flags", tmp);
+}
+
+/**
+ * Ethernet device configuration.
+ *
+ * Prepare the driver for a given number of TX and RX queues.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+dev_configure(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ unsigned int rxqs_n = dev->data->nb_rx_queues;
+ unsigned int txqs_n = dev->data->nb_tx_queues;
+ unsigned int i;
+ unsigned int j;
+ unsigned int reta_idx_n;
+
+ priv->rss_hf = dev->data->dev_conf.rx_adv_conf.rss_conf.rss_hf;
+ priv->rxqs = (void *)dev->data->rx_queues;
+ priv->txqs = (void *)dev->data->tx_queues;
+ if (txqs_n != priv->txqs_n) {
+ INFO("%p: TX queues number update: %u -> %u",
+ (void *)dev, priv->txqs_n, txqs_n);
+ priv->txqs_n = txqs_n;
+ }
+ if (rxqs_n > priv->ind_table_max_size) {
+ ERROR("cannot handle this many RX queues (%u)", rxqs_n);
+ return EINVAL;
+ }
+ if (rxqs_n == priv->rxqs_n)
+ return 0;
+ INFO("%p: RX queues number update: %u -> %u",
+ (void *)dev, priv->rxqs_n, rxqs_n);
+ priv->rxqs_n = rxqs_n;
+ /* If the requested number of RX queues is not a power of two, use the
+ * maximum indirection table size for better balancing.
+ * The result is always rounded to the next power of two. */
+ reta_idx_n = (1 << log2above((rxqs_n & (rxqs_n - 1)) ?
+ priv->ind_table_max_size :
+ rxqs_n));
+ if (priv_rss_reta_index_resize(priv, reta_idx_n))
+ return ENOMEM;
+ /* When the number of RX queues is not a power of two, the remaining
+ * table entries are padded with reused WQs and hashes are not spread
+ * uniformly. */
+ for (i = 0, j = 0; (i != reta_idx_n); ++i) {
+ (*priv->reta_idx)[i] = j;
+ if (++j == rxqs_n)
+ j = 0;
+ }
+ return 0;
+}
+
+/**
+ * DPDK callback for Ethernet device configuration.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_configure(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ priv_lock(priv);
+ ret = dev_configure(dev);
+ assert(ret >= 0);
+ priv_unlock(priv);
+ return -ret;
+}
+
+/**
+ * DPDK callback to get information about the device.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[out] info
+ * Info structure output buffer.
+ */
+void
+mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ unsigned int max;
+ char ifname[IF_NAMESIZE];
+
+ priv_lock(priv);
+ /* FIXME: we should ask the device for these values. */
+ info->min_rx_bufsize = 32;
+ info->max_rx_pktlen = 65536;
+ /*
+ * Since we need one CQ per QP, the limit is the minimum number
+ * between the two values.
+ */
+ max = ((priv->device_attr.max_cq > priv->device_attr.max_qp) ?
+ priv->device_attr.max_qp : priv->device_attr.max_cq);
+ /* If max >= 65535 then max = 0, max_rx_queues is uint16_t. */
+ if (max >= 65535)
+ max = 65535;
+ info->max_rx_queues = max;
+ info->max_tx_queues = max;
+ info->max_mac_addrs = RTE_DIM(priv->mac);
+ info->rx_offload_capa =
+ (priv->hw_csum ?
+ (DEV_RX_OFFLOAD_IPV4_CKSUM |
+ DEV_RX_OFFLOAD_UDP_CKSUM |
+ DEV_RX_OFFLOAD_TCP_CKSUM) :
+ 0);
+ info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT;
+ if (priv->hw_csum)
+ info->tx_offload_capa |=
+ (DEV_TX_OFFLOAD_IPV4_CKSUM |
+ DEV_TX_OFFLOAD_UDP_CKSUM |
+ DEV_TX_OFFLOAD_TCP_CKSUM);
+ if (priv_get_ifname(priv, &ifname) == 0)
+ info->if_index = if_nametoindex(ifname);
+ /* FIXME: RETA update/query API expects the callee to know the size of
+ * the indirection table, for this PMD the size varies depending on
+ * the number of RX queues, it becomes impossible to find the correct
+ * size if it is not fixed.
+ * The API should be updated to solve this problem. */
+ info->reta_size = priv->ind_table_max_size;
+ info->speed_capa =
+ ETH_LINK_SPEED_1G |
+ ETH_LINK_SPEED_10G |
+ ETH_LINK_SPEED_20G |
+ ETH_LINK_SPEED_25G |
+ ETH_LINK_SPEED_40G |
+ ETH_LINK_SPEED_50G |
+ ETH_LINK_SPEED_56G |
+ ETH_LINK_SPEED_100G;
+ priv_unlock(priv);
+}
+
+const uint32_t *
+mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev)
+{
+ static const uint32_t ptypes[] = {
+ /* refers to rxq_cq_to_pkt_type() */
+ RTE_PTYPE_L3_IPV4,
+ RTE_PTYPE_L3_IPV6,
+ RTE_PTYPE_INNER_L3_IPV4,
+ RTE_PTYPE_INNER_L3_IPV6,
+ RTE_PTYPE_UNKNOWN
+
+ };
+
+ if (dev->rx_pkt_burst == mlx5_rx_burst ||
+ dev->rx_pkt_burst == mlx5_rx_burst_sp)
+ return ptypes;
+ return NULL;
+}
+
+/**
+ * DPDK callback to retrieve physical link information (unlocked version).
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param wait_to_complete
+ * Wait for request completion (ignored).
+ */
+static int
+mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ struct ethtool_cmd edata = {
+ .cmd = ETHTOOL_GSET
+ };
+ struct ifreq ifr;
+ struct rte_eth_link dev_link;
+ int link_speed = 0;
+
+ (void)wait_to_complete;
+ if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) {
+ WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno));
+ return -1;
+ }
+ memset(&dev_link, 0, sizeof(dev_link));
+ dev_link.link_status = ((ifr.ifr_flags & IFF_UP) &&
+ (ifr.ifr_flags & IFF_RUNNING));
+ ifr.ifr_data = &edata;
+ if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+ WARN("ioctl(SIOCETHTOOL, ETHTOOL_GSET) failed: %s",
+ strerror(errno));
+ return -1;
+ }
+ link_speed = ethtool_cmd_speed(&edata);
+ if (link_speed == -1)
+ dev_link.link_speed = 0;
+ else
+ dev_link.link_speed = link_speed;
+ dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
+ ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+ dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+ ETH_LINK_SPEED_FIXED);
+ if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
+ /* Link status changed. */
+ dev->data->dev_link = dev_link;
+ return 0;
+ }
+ /* Link status is still the same. */
+ return -1;
+}
+
+/**
+ * DPDK callback to retrieve physical link information.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param wait_to_complete
+ * Wait for request completion (ignored).
+ */
+int
+mlx5_link_update(struct rte_eth_dev *dev, int wait_to_complete)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ int ret;
+
+ priv_lock(priv);
+ ret = mlx5_link_update_unlocked(dev, wait_to_complete);
+ priv_unlock(priv);
+ return ret;
+}
+
+/**
+ * DPDK callback to change the MTU.
+ *
+ * Setting the MTU affects hardware MRU (packets larger than the MTU cannot be
+ * received). Use this as a hint to enable/disable scattered packets support
+ * and improve performance when not needed.
+ * Since failure is not an option, reconfiguring queues on the fly is not
+ * recommended.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param in_mtu
+ * New MTU.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret = 0;
+ unsigned int i;
+ uint16_t (*rx_func)(void *, struct rte_mbuf **, uint16_t) =
+ mlx5_rx_burst;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ priv_lock(priv);
+ /* Set kernel interface MTU first. */
+ if (priv_set_mtu(priv, mtu)) {
+ ret = errno;
+ WARN("cannot set port %u MTU to %u: %s", priv->port, mtu,
+ strerror(ret));
+ goto out;
+ } else
+ DEBUG("adapter port %u MTU set to %u", priv->port, mtu);
+ priv->mtu = mtu;
+ /* Temporarily replace RX handler with a fake one, assuming it has not
+ * been copied elsewhere. */
+ dev->rx_pkt_burst = removed_rx_burst;
+ /* Make sure everyone has left mlx5_rx_burst() and uses
+ * removed_rx_burst() instead. */
+ rte_wmb();
+ usleep(1000);
+ /* Reconfigure each RX queue. */
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ struct rxq *rxq = (*priv->rxqs)[i];
+ unsigned int max_frame_len;
+ int sp;
+
+ if (rxq == NULL)
+ continue;
+ /* Calculate new maximum frame length according to MTU and
+ * toggle scattered support (sp) if necessary. */
+ max_frame_len = (priv->mtu + ETHER_HDR_LEN +
+ (ETHER_MAX_VLAN_FRAME_LEN - ETHER_MAX_LEN));
+ sp = (max_frame_len > (rxq->mb_len - RTE_PKTMBUF_HEADROOM));
+ /* Provide new values to rxq_setup(). */
+ dev->data->dev_conf.rxmode.jumbo_frame = sp;
+ dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame_len;
+ ret = rxq_rehash(dev, rxq);
+ if (ret) {
+ /* Force SP RX if that queue requires it and abort. */
+ if (rxq->sp)
+ rx_func = mlx5_rx_burst_sp;
+ break;
+ }
+ /* Scattered burst function takes priority. */
+ if (rxq->sp)
+ rx_func = mlx5_rx_burst_sp;
+ }
+ /* Burst functions can now be called again. */
+ rte_wmb();
+ dev->rx_pkt_burst = rx_func;
+out:
+ priv_unlock(priv);
+ assert(ret >= 0);
+ return -ret;
+}
+
+/**
+ * DPDK callback to get flow control status.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[out] fc_conf
+ * Flow control output buffer.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct ifreq ifr;
+ struct ethtool_pauseparam ethpause = {
+ .cmd = ETHTOOL_GPAUSEPARAM
+ };
+ int ret;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ ifr.ifr_data = &ethpause;
+ priv_lock(priv);
+ if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+ ret = errno;
+ WARN("ioctl(SIOCETHTOOL, ETHTOOL_GPAUSEPARAM)"
+ " failed: %s",
+ strerror(ret));
+ goto out;
+ }
+
+ fc_conf->autoneg = ethpause.autoneg;
+ if (ethpause.rx_pause && ethpause.tx_pause)
+ fc_conf->mode = RTE_FC_FULL;
+ else if (ethpause.rx_pause)
+ fc_conf->mode = RTE_FC_RX_PAUSE;
+ else if (ethpause.tx_pause)
+ fc_conf->mode = RTE_FC_TX_PAUSE;
+ else
+ fc_conf->mode = RTE_FC_NONE;
+ ret = 0;
+
+out:
+ priv_unlock(priv);
+ assert(ret >= 0);
+ return -ret;
+}
+
+/**
+ * DPDK callback to modify flow control parameters.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[in] fc_conf
+ * Flow control parameters.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct ifreq ifr;
+ struct ethtool_pauseparam ethpause = {
+ .cmd = ETHTOOL_SPAUSEPARAM
+ };
+ int ret;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ ifr.ifr_data = &ethpause;
+ ethpause.autoneg = fc_conf->autoneg;
+ if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) ||
+ (fc_conf->mode & RTE_FC_RX_PAUSE))
+ ethpause.rx_pause = 1;
+ else
+ ethpause.rx_pause = 0;
+
+ if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) ||
+ (fc_conf->mode & RTE_FC_TX_PAUSE))
+ ethpause.tx_pause = 1;
+ else
+ ethpause.tx_pause = 0;
+
+ priv_lock(priv);
+ if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+ ret = errno;
+ WARN("ioctl(SIOCETHTOOL, ETHTOOL_SPAUSEPARAM)"
+ " failed: %s",
+ strerror(ret));
+ goto out;
+ }
+ ret = 0;
+
+out:
+ priv_unlock(priv);
+ assert(ret >= 0);
+ return -ret;
+}
+
+/**
+ * Get PCI information from struct ibv_device.
+ *
+ * @param device
+ * Pointer to Ethernet device structure.
+ * @param[out] pci_addr
+ * PCI bus address output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+mlx5_ibv_device_to_pci_addr(const struct ibv_device *device,
+ struct rte_pci_addr *pci_addr)
+{
+ FILE *file;
+ char line[32];
+ MKSTR(path, "%s/device/uevent", device->ibdev_path);
+
+ file = fopen(path, "rb");
+ if (file == NULL)
+ return -1;
+ while (fgets(line, sizeof(line), file) == line) {
+ size_t len = strlen(line);
+ int ret;
+
+ /* Truncate long lines. */
+ if (len == (sizeof(line) - 1))
+ while (line[(len - 1)] != '\n') {
+ ret = fgetc(file);
+ if (ret == EOF)
+ break;
+ line[(len - 1)] = ret;
+ }
+ /* Extract information. */
+ if (sscanf(line,
+ "PCI_SLOT_NAME="
+ "%" SCNx16 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n",
+ &pci_addr->domain,
+ &pci_addr->bus,
+ &pci_addr->devid,
+ &pci_addr->function) == 4) {
+ ret = 0;
+ break;
+ }
+ }
+ fclose(file);
+ return 0;
+}
+
+/**
+ * Link status handler.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param dev
+ * Pointer to the rte_eth_dev structure.
+ *
+ * @return
+ * Nonzero if the callback process can be called immediately.
+ */
+static int
+priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev)
+{
+ struct ibv_async_event event;
+ int port_change = 0;
+ int ret = 0;
+
+ /* Read all message and acknowledge them. */
+ for (;;) {
+ if (ibv_get_async_event(priv->ctx, &event))
+ break;
+
+ if (event.event_type == IBV_EVENT_PORT_ACTIVE ||
+ event.event_type == IBV_EVENT_PORT_ERR)
+ port_change = 1;
+ else
+ DEBUG("event type %d on port %d not handled",
+ event.event_type, event.element.port_num);
+ ibv_ack_async_event(&event);
+ }
+
+ if (port_change ^ priv->pending_alarm) {
+ struct rte_eth_link *link = &dev->data->dev_link;
+
+ priv->pending_alarm = 0;
+ mlx5_link_update_unlocked(dev, 0);
+ if (((link->link_speed == 0) && link->link_status) ||
+ ((link->link_speed != 0) && !link->link_status)) {
+ /* Inconsistent status, check again later. */
+ priv->pending_alarm = 1;
+ rte_eal_alarm_set(MLX5_ALARM_TIMEOUT_US,
+ mlx5_dev_link_status_handler,
+ dev);
+ } else
+ ret = 1;
+ }
+ return ret;
+}
+
+/**
+ * Handle delayed link status event.
+ *
+ * @param arg
+ * Registered argument.
+ */
+void
+mlx5_dev_link_status_handler(void *arg)
+{
+ struct rte_eth_dev *dev = arg;
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ priv_lock(priv);
+ assert(priv->pending_alarm == 1);
+ ret = priv_dev_link_status_handler(priv, dev);
+ priv_unlock(priv);
+ if (ret)
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
+}
+
+/**
+ * Handle interrupts from the NIC.
+ *
+ * @param[in] intr_handle
+ * Interrupt handler.
+ * @param cb_arg
+ * Callback argument.
+ */
+void
+mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg)
+{
+ struct rte_eth_dev *dev = cb_arg;
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ (void)intr_handle;
+ priv_lock(priv);
+ ret = priv_dev_link_status_handler(priv, dev);
+ priv_unlock(priv);
+ if (ret)
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
+}
+
+/**
+ * Uninstall interrupt handler.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param dev
+ * Pointer to the rte_eth_dev structure.
+ */
+void
+priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev)
+{
+ if (!dev->data->dev_conf.intr_conf.lsc)
+ return;
+ rte_intr_callback_unregister(&priv->intr_handle,
+ mlx5_dev_interrupt_handler,
+ dev);
+ if (priv->pending_alarm)
+ rte_eal_alarm_cancel(mlx5_dev_link_status_handler, dev);
+ priv->pending_alarm = 0;
+ priv->intr_handle.fd = 0;
+ priv->intr_handle.type = 0;
+}
+
+/**
+ * Install interrupt handler.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param dev
+ * Pointer to the rte_eth_dev structure.
+ */
+void
+priv_dev_interrupt_handler_install(struct priv *priv, struct rte_eth_dev *dev)
+{
+ int rc, flags;
+
+ if (!dev->data->dev_conf.intr_conf.lsc)
+ return;
+ assert(priv->ctx->async_fd > 0);
+ flags = fcntl(priv->ctx->async_fd, F_GETFL);
+ rc = fcntl(priv->ctx->async_fd, F_SETFL, flags | O_NONBLOCK);
+ if (rc < 0) {
+ INFO("failed to change file descriptor async event queue");
+ dev->data->dev_conf.intr_conf.lsc = 0;
+ } else {
+ priv->intr_handle.fd = priv->ctx->async_fd;
+ priv->intr_handle.type = RTE_INTR_HANDLE_EXT;
+ rte_intr_callback_register(&priv->intr_handle,
+ mlx5_dev_interrupt_handler,
+ dev);
+ }
+}
+
+/**
+ * Change the link state (UP / DOWN).
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param up
+ * Nonzero for link up, otherwise link down.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_set_link(struct priv *priv, int up)
+{
+ struct rte_eth_dev *dev = priv->dev;
+ int err;
+ unsigned int i;
+
+ if (up) {
+ err = priv_set_flags(priv, ~IFF_UP, IFF_UP);
+ if (err)
+ return err;
+ for (i = 0; i < priv->rxqs_n; i++)
+ if ((*priv->rxqs)[i]->sp)
+ break;
+ /* Check if an sp queue exists.
+ * Note: Some old frames might be received.
+ */
+ if (i == priv->rxqs_n)
+ dev->rx_pkt_burst = mlx5_rx_burst;
+ else
+ dev->rx_pkt_burst = mlx5_rx_burst_sp;
+ dev->tx_pkt_burst = mlx5_tx_burst;
+ } else {
+ err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP);
+ if (err)
+ return err;
+ dev->rx_pkt_burst = removed_rx_burst;
+ dev->tx_pkt_burst = removed_tx_burst;
+ }
+ return 0;
+}
+
+/**
+ * DPDK callback to bring the link DOWN.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+mlx5_set_link_down(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int err;
+
+ priv_lock(priv);
+ err = priv_set_link(priv, 0);
+ priv_unlock(priv);
+ return err;
+}
+
+/**
+ * DPDK callback to bring the link UP.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+mlx5_set_link_up(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int err;
+
+ priv_lock(priv);
+ err = priv_set_link(priv, 1);
+ priv_unlock(priv);
+ return err;
+}
+
+/**
+ * Configure secondary process queues from a private data pointer (primary
+ * or secondary) and update burst callbacks. Can take place only once.
+ *
+ * All queues must have been previously created by the primary process to
+ * avoid undefined behavior.
+ *
+ * @param priv
+ * Private data pointer from either primary or secondary process.
+ *
+ * @return
+ * Private data pointer from secondary process, NULL in case of error.
+ */
+struct priv *
+mlx5_secondary_data_setup(struct priv *priv)
+{
+ unsigned int port_id = 0;
+ struct mlx5_secondary_data *sd;
+ void **tx_queues;
+ void **rx_queues;
+ unsigned int nb_tx_queues;
+ unsigned int nb_rx_queues;
+ unsigned int i;
+
+ /* priv must be valid at this point. */
+ assert(priv != NULL);
+ /* priv->dev must also be valid but may point to local memory from
+ * another process, possibly with the same address and must not
+ * be dereferenced yet. */
+ assert(priv->dev != NULL);
+ /* Determine port ID by finding out where priv comes from. */
+ while (1) {
+ sd = &mlx5_secondary_data[port_id];
+ rte_spinlock_lock(&sd->lock);
+ /* Primary process? */
+ if (sd->primary_priv == priv)
+ break;
+ /* Secondary process? */
+ if (sd->data.dev_private == priv)
+ break;
+ rte_spinlock_unlock(&sd->lock);
+ if (++port_id == RTE_DIM(mlx5_secondary_data))
+ port_id = 0;
+ }
+ /* Switch to secondary private structure. If private data has already
+ * been updated by another thread, there is nothing else to do. */
+ priv = sd->data.dev_private;
+ if (priv->dev->data == &sd->data)
+ goto end;
+ /* Sanity checks. Secondary private structure is supposed to point
+ * to local eth_dev, itself still pointing to the shared device data
+ * structure allocated by the primary process. */
+ assert(sd->shared_dev_data != &sd->data);
+ assert(sd->data.nb_tx_queues == 0);
+ assert(sd->data.tx_queues == NULL);
+ assert(sd->data.nb_rx_queues == 0);
+ assert(sd->data.rx_queues == NULL);
+ assert(priv != sd->primary_priv);
+ assert(priv->dev->data == sd->shared_dev_data);
+ assert(priv->txqs_n == 0);
+ assert(priv->txqs == NULL);
+ assert(priv->rxqs_n == 0);
+ assert(priv->rxqs == NULL);
+ nb_tx_queues = sd->shared_dev_data->nb_tx_queues;
+ nb_rx_queues = sd->shared_dev_data->nb_rx_queues;
+ /* Allocate local storage for queues. */
+ tx_queues = rte_zmalloc("secondary ethdev->tx_queues",
+ sizeof(sd->data.tx_queues[0]) * nb_tx_queues,
+ RTE_CACHE_LINE_SIZE);
+ rx_queues = rte_zmalloc("secondary ethdev->rx_queues",
+ sizeof(sd->data.rx_queues[0]) * nb_rx_queues,
+ RTE_CACHE_LINE_SIZE);
+ if (tx_queues == NULL || rx_queues == NULL)
+ goto error;
+ /* Lock to prevent control operations during setup. */
+ priv_lock(priv);
+ /* TX queues. */
+ for (i = 0; i != nb_tx_queues; ++i) {
+ struct txq *primary_txq = (*sd->primary_priv->txqs)[i];
+ struct txq *txq;
+
+ if (primary_txq == NULL)
+ continue;
+ txq = rte_calloc_socket("TXQ", 1, sizeof(*txq), 0,
+ primary_txq->socket);
+ if (txq != NULL) {
+ if (txq_setup(priv->dev,
+ txq,
+ primary_txq->elts_n * MLX5_PMD_SGE_WR_N,
+ primary_txq->socket,
+ NULL) == 0) {
+ txq->stats.idx = primary_txq->stats.idx;
+ tx_queues[i] = txq;
+ continue;
+ }
+ rte_free(txq);
+ }
+ while (i) {
+ txq = tx_queues[--i];
+ txq_cleanup(txq);
+ rte_free(txq);
+ }
+ goto error;
+ }
+ /* RX queues. */
+ for (i = 0; i != nb_rx_queues; ++i) {
+ struct rxq *primary_rxq = (*sd->primary_priv->rxqs)[i];
+
+ if (primary_rxq == NULL)
+ continue;
+ /* Not supported yet. */
+ rx_queues[i] = NULL;
+ }
+ /* Update everything. */
+ priv->txqs = (void *)tx_queues;
+ priv->txqs_n = nb_tx_queues;
+ priv->rxqs = (void *)rx_queues;
+ priv->rxqs_n = nb_rx_queues;
+ sd->data.rx_queues = rx_queues;
+ sd->data.tx_queues = tx_queues;
+ sd->data.nb_rx_queues = nb_rx_queues;
+ sd->data.nb_tx_queues = nb_tx_queues;
+ sd->data.dev_link = sd->shared_dev_data->dev_link;
+ sd->data.mtu = sd->shared_dev_data->mtu;
+ memcpy(sd->data.rx_queue_state, sd->shared_dev_data->rx_queue_state,
+ sizeof(sd->data.rx_queue_state));
+ memcpy(sd->data.tx_queue_state, sd->shared_dev_data->tx_queue_state,
+ sizeof(sd->data.tx_queue_state));
+ sd->data.dev_flags = sd->shared_dev_data->dev_flags;
+ /* Use local data from now on. */
+ rte_mb();
+ priv->dev->data = &sd->data;
+ rte_mb();
+ priv->dev->tx_pkt_burst = mlx5_tx_burst;
+ priv->dev->rx_pkt_burst = removed_rx_burst;
+ priv_unlock(priv);
+end:
+ /* More sanity checks. */
+ assert(priv->dev->tx_pkt_burst == mlx5_tx_burst);
+ assert(priv->dev->rx_pkt_burst == removed_rx_burst);
+ assert(priv->dev->data == &sd->data);
+ rte_spinlock_unlock(&sd->lock);
+ return priv;
+error:
+ priv_unlock(priv);
+ rte_free(tx_queues);
+ rte_free(rx_queues);
+ rte_spinlock_unlock(&sd->lock);
+ return NULL;
+}
diff --git a/drivers/net/mlx5/mlx5_fdir.c b/drivers/net/mlx5/mlx5_fdir.c
new file mode 100644
index 00000000..63e43ad9
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_fdir.c
@@ -0,0 +1,980 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <assert.h>
+#include <stdint.h>
+#include <string.h>
+#include <errno.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ether.h>
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+
+struct fdir_flow_desc {
+ uint16_t dst_port;
+ uint16_t src_port;
+ uint32_t src_ip[4];
+ uint32_t dst_ip[4];
+ uint8_t mac[6];
+ uint16_t vlan_tag;
+ enum hash_rxq_type type;
+};
+
+struct mlx5_fdir_filter {
+ LIST_ENTRY(mlx5_fdir_filter) next;
+ uint16_t queue; /* Queue assigned to if FDIR match. */
+ struct fdir_flow_desc desc;
+ struct ibv_exp_flow *flow;
+};
+
+LIST_HEAD(fdir_filter_list, mlx5_fdir_filter);
+
+/**
+ * Convert struct rte_eth_fdir_filter to mlx5 filter descriptor.
+ *
+ * @param[in] fdir_filter
+ * DPDK filter structure to convert.
+ * @param[out] desc
+ * Resulting mlx5 filter descriptor.
+ * @param mode
+ * Flow director mode.
+ */
+static void
+fdir_filter_to_flow_desc(const struct rte_eth_fdir_filter *fdir_filter,
+ struct fdir_flow_desc *desc, enum rte_fdir_mode mode)
+{
+ /* Initialize descriptor. */
+ memset(desc, 0, sizeof(*desc));
+
+ /* Set VLAN ID. */
+ desc->vlan_tag = fdir_filter->input.flow_ext.vlan_tci;
+
+ /* Set MAC address. */
+ if (mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
+ rte_memcpy(desc->mac,
+ fdir_filter->input.flow.mac_vlan_flow.mac_addr.
+ addr_bytes,
+ sizeof(desc->mac));
+ desc->type = HASH_RXQ_ETH;
+ return;
+ }
+
+ /* Set mode */
+ switch (fdir_filter->input.flow_type) {
+ case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
+ desc->type = HASH_RXQ_UDPV4;
+ break;
+ case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
+ desc->type = HASH_RXQ_TCPV4;
+ break;
+ case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
+ desc->type = HASH_RXQ_IPV4;
+ break;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
+ desc->type = HASH_RXQ_UDPV6;
+ break;
+ case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
+ desc->type = HASH_RXQ_TCPV6;
+ break;
+ case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
+ desc->type = HASH_RXQ_IPV6;
+ break;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ default:
+ break;
+ }
+
+ /* Set flow values */
+ switch (fdir_filter->input.flow_type) {
+ case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
+ case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
+ desc->src_port = fdir_filter->input.flow.udp4_flow.src_port;
+ desc->dst_port = fdir_filter->input.flow.udp4_flow.dst_port;
+ case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
+ desc->src_ip[0] = fdir_filter->input.flow.ip4_flow.src_ip;
+ desc->dst_ip[0] = fdir_filter->input.flow.ip4_flow.dst_ip;
+ break;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
+ case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
+ desc->src_port = fdir_filter->input.flow.udp6_flow.src_port;
+ desc->dst_port = fdir_filter->input.flow.udp6_flow.dst_port;
+ /* Fall through. */
+ case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
+ rte_memcpy(desc->src_ip,
+ fdir_filter->input.flow.ipv6_flow.src_ip,
+ sizeof(desc->src_ip));
+ rte_memcpy(desc->dst_ip,
+ fdir_filter->input.flow.ipv6_flow.dst_ip,
+ sizeof(desc->dst_ip));
+ break;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ default:
+ break;
+ }
+}
+
+/**
+ * Check if two flow descriptors overlap according to configured mask.
+ *
+ * @param priv
+ * Private structure that provides flow director mask.
+ * @param desc1
+ * First flow descriptor to compare.
+ * @param desc2
+ * Second flow descriptor to compare.
+ *
+ * @return
+ * Nonzero if descriptors overlap.
+ */
+static int
+priv_fdir_overlap(const struct priv *priv,
+ const struct fdir_flow_desc *desc1,
+ const struct fdir_flow_desc *desc2)
+{
+ const struct rte_eth_fdir_masks *mask =
+ &priv->dev->data->dev_conf.fdir_conf.mask;
+ unsigned int i;
+
+ if (desc1->type != desc2->type)
+ return 0;
+ /* Ignore non masked bits. */
+ for (i = 0; i != RTE_DIM(desc1->mac); ++i)
+ if ((desc1->mac[i] & mask->mac_addr_byte_mask) !=
+ (desc2->mac[i] & mask->mac_addr_byte_mask))
+ return 0;
+ if (((desc1->src_port & mask->src_port_mask) !=
+ (desc2->src_port & mask->src_port_mask)) ||
+ ((desc1->dst_port & mask->dst_port_mask) !=
+ (desc2->dst_port & mask->dst_port_mask)))
+ return 0;
+ switch (desc1->type) {
+ case HASH_RXQ_IPV4:
+ case HASH_RXQ_UDPV4:
+ case HASH_RXQ_TCPV4:
+ if (((desc1->src_ip[0] & mask->ipv4_mask.src_ip) !=
+ (desc2->src_ip[0] & mask->ipv4_mask.src_ip)) ||
+ ((desc1->dst_ip[0] & mask->ipv4_mask.dst_ip) !=
+ (desc2->dst_ip[0] & mask->ipv4_mask.dst_ip)))
+ return 0;
+ break;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ case HASH_RXQ_IPV6:
+ case HASH_RXQ_UDPV6:
+ case HASH_RXQ_TCPV6:
+ for (i = 0; i != RTE_DIM(desc1->src_ip); ++i)
+ if (((desc1->src_ip[i] & mask->ipv6_mask.src_ip[i]) !=
+ (desc2->src_ip[i] & mask->ipv6_mask.src_ip[i])) ||
+ ((desc1->dst_ip[i] & mask->ipv6_mask.dst_ip[i]) !=
+ (desc2->dst_ip[i] & mask->ipv6_mask.dst_ip[i])))
+ return 0;
+ break;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ default:
+ break;
+ }
+ return 1;
+}
+
+/**
+ * Create flow director steering rule for a specific filter.
+ *
+ * @param priv
+ * Private structure.
+ * @param mlx5_fdir_filter
+ * Filter to create a steering rule for.
+ * @param fdir_queue
+ * Flow director queue for matching packets.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_flow_add(struct priv *priv,
+ struct mlx5_fdir_filter *mlx5_fdir_filter,
+ struct fdir_queue *fdir_queue)
+{
+ struct ibv_exp_flow *flow;
+ struct fdir_flow_desc *desc = &mlx5_fdir_filter->desc;
+ enum rte_fdir_mode fdir_mode =
+ priv->dev->data->dev_conf.fdir_conf.mode;
+ struct rte_eth_fdir_masks *mask =
+ &priv->dev->data->dev_conf.fdir_conf.mask;
+ FLOW_ATTR_SPEC_ETH(data, priv_flow_attr(priv, NULL, 0, desc->type));
+ struct ibv_exp_flow_attr *attr = &data->attr;
+ uintptr_t spec_offset = (uintptr_t)&data->spec;
+ struct ibv_exp_flow_spec_eth *spec_eth;
+ struct ibv_exp_flow_spec_ipv4 *spec_ipv4;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ struct ibv_exp_flow_spec_ipv6 *spec_ipv6;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ struct ibv_exp_flow_spec_tcp_udp *spec_tcp_udp;
+ struct mlx5_fdir_filter *iter_fdir_filter;
+ unsigned int i;
+
+ /* Abort if an existing flow overlaps this one to avoid packet
+ * duplication, even if it targets another queue. */
+ LIST_FOREACH(iter_fdir_filter, priv->fdir_filter_list, next)
+ if ((iter_fdir_filter != mlx5_fdir_filter) &&
+ (iter_fdir_filter->flow != NULL) &&
+ (priv_fdir_overlap(priv,
+ &mlx5_fdir_filter->desc,
+ &iter_fdir_filter->desc)))
+ return EEXIST;
+
+ /*
+ * No padding must be inserted by the compiler between attr and spec.
+ * This layout is expected by libibverbs.
+ */
+ assert(((uint8_t *)attr + sizeof(*attr)) == (uint8_t *)spec_offset);
+ priv_flow_attr(priv, attr, sizeof(data), desc->type);
+
+ /* Set Ethernet spec */
+ spec_eth = (struct ibv_exp_flow_spec_eth *)spec_offset;
+
+ /* The first specification must be Ethernet. */
+ assert(spec_eth->type == IBV_EXP_FLOW_SPEC_ETH);
+ assert(spec_eth->size == sizeof(*spec_eth));
+
+ /* VLAN ID */
+ spec_eth->val.vlan_tag = desc->vlan_tag & mask->vlan_tci_mask;
+ spec_eth->mask.vlan_tag = mask->vlan_tci_mask;
+
+ /* Update priority */
+ attr->priority = 2;
+
+ if (fdir_mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
+ /* MAC Address */
+ for (i = 0; i != RTE_DIM(spec_eth->mask.dst_mac); ++i) {
+ spec_eth->val.dst_mac[i] =
+ desc->mac[i] & mask->mac_addr_byte_mask;
+ spec_eth->mask.dst_mac[i] = mask->mac_addr_byte_mask;
+ }
+ goto create_flow;
+ }
+
+ switch (desc->type) {
+ case HASH_RXQ_IPV4:
+ case HASH_RXQ_UDPV4:
+ case HASH_RXQ_TCPV4:
+ spec_offset += spec_eth->size;
+
+ /* Set IP spec */
+ spec_ipv4 = (struct ibv_exp_flow_spec_ipv4 *)spec_offset;
+
+ /* The second specification must be IP. */
+ assert(spec_ipv4->type == IBV_EXP_FLOW_SPEC_IPV4);
+ assert(spec_ipv4->size == sizeof(*spec_ipv4));
+
+ spec_ipv4->val.src_ip =
+ desc->src_ip[0] & mask->ipv4_mask.src_ip;
+ spec_ipv4->val.dst_ip =
+ desc->dst_ip[0] & mask->ipv4_mask.dst_ip;
+ spec_ipv4->mask.src_ip = mask->ipv4_mask.src_ip;
+ spec_ipv4->mask.dst_ip = mask->ipv4_mask.dst_ip;
+
+ /* Update priority */
+ attr->priority = 1;
+
+ if (desc->type == HASH_RXQ_IPV4)
+ goto create_flow;
+
+ spec_offset += spec_ipv4->size;
+ break;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ case HASH_RXQ_IPV6:
+ case HASH_RXQ_UDPV6:
+ case HASH_RXQ_TCPV6:
+ spec_offset += spec_eth->size;
+
+ /* Set IP spec */
+ spec_ipv6 = (struct ibv_exp_flow_spec_ipv6 *)spec_offset;
+
+ /* The second specification must be IP. */
+ assert(spec_ipv6->type == IBV_EXP_FLOW_SPEC_IPV6);
+ assert(spec_ipv6->size == sizeof(*spec_ipv6));
+
+ for (i = 0; i != RTE_DIM(desc->src_ip); ++i) {
+ ((uint32_t *)spec_ipv6->val.src_ip)[i] =
+ desc->src_ip[i] & mask->ipv6_mask.src_ip[i];
+ ((uint32_t *)spec_ipv6->val.dst_ip)[i] =
+ desc->dst_ip[i] & mask->ipv6_mask.dst_ip[i];
+ }
+ rte_memcpy(spec_ipv6->mask.src_ip,
+ mask->ipv6_mask.src_ip,
+ sizeof(spec_ipv6->mask.src_ip));
+ rte_memcpy(spec_ipv6->mask.dst_ip,
+ mask->ipv6_mask.dst_ip,
+ sizeof(spec_ipv6->mask.dst_ip));
+
+ /* Update priority */
+ attr->priority = 1;
+
+ if (desc->type == HASH_RXQ_IPV6)
+ goto create_flow;
+
+ spec_offset += spec_ipv6->size;
+ break;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ default:
+ ERROR("invalid flow attribute type");
+ return EINVAL;
+ }
+
+ /* Set TCP/UDP flow specification. */
+ spec_tcp_udp = (struct ibv_exp_flow_spec_tcp_udp *)spec_offset;
+
+ /* The third specification must be TCP/UDP. */
+ assert(spec_tcp_udp->type == IBV_EXP_FLOW_SPEC_TCP ||
+ spec_tcp_udp->type == IBV_EXP_FLOW_SPEC_UDP);
+ assert(spec_tcp_udp->size == sizeof(*spec_tcp_udp));
+
+ spec_tcp_udp->val.src_port = desc->src_port & mask->src_port_mask;
+ spec_tcp_udp->val.dst_port = desc->dst_port & mask->dst_port_mask;
+ spec_tcp_udp->mask.src_port = mask->src_port_mask;
+ spec_tcp_udp->mask.dst_port = mask->dst_port_mask;
+
+ /* Update priority */
+ attr->priority = 0;
+
+create_flow:
+
+ errno = 0;
+ flow = ibv_exp_create_flow(fdir_queue->qp, attr);
+ if (flow == NULL) {
+ /* It's not clear whether errno is always set in this case. */
+ ERROR("%p: flow director configuration failed, errno=%d: %s",
+ (void *)priv, errno,
+ (errno ? strerror(errno) : "Unknown error"));
+ if (errno)
+ return errno;
+ return EINVAL;
+ }
+
+ DEBUG("%p: added flow director rule (%p)", (void *)priv, (void *)flow);
+ mlx5_fdir_filter->flow = flow;
+ return 0;
+}
+
+/**
+ * Get flow director queue for a specific RX queue, create it in case
+ * it does not exist.
+ *
+ * @param priv
+ * Private structure.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * Related flow director queue on success, NULL otherwise.
+ */
+static struct fdir_queue *
+priv_get_fdir_queue(struct priv *priv, uint16_t idx)
+{
+ struct fdir_queue *fdir_queue = &(*priv->rxqs)[idx]->fdir_queue;
+ struct ibv_exp_rwq_ind_table *ind_table = NULL;
+ struct ibv_qp *qp = NULL;
+ struct ibv_exp_rwq_ind_table_init_attr ind_init_attr;
+ struct ibv_exp_rx_hash_conf hash_conf;
+ struct ibv_exp_qp_init_attr qp_init_attr;
+ int err = 0;
+
+ /* Return immediately if it has already been created. */
+ if (fdir_queue->qp != NULL)
+ return fdir_queue;
+
+ ind_init_attr = (struct ibv_exp_rwq_ind_table_init_attr){
+ .pd = priv->pd,
+ .log_ind_tbl_size = 0,
+ .ind_tbl = &((*priv->rxqs)[idx]->wq),
+ .comp_mask = 0,
+ };
+
+ errno = 0;
+ ind_table = ibv_exp_create_rwq_ind_table(priv->ctx,
+ &ind_init_attr);
+ if (ind_table == NULL) {
+ /* Not clear whether errno is set. */
+ err = (errno ? errno : EINVAL);
+ ERROR("RX indirection table creation failed with error %d: %s",
+ err, strerror(err));
+ goto error;
+ }
+
+ /* Create fdir_queue qp. */
+ hash_conf = (struct ibv_exp_rx_hash_conf){
+ .rx_hash_function = IBV_EXP_RX_HASH_FUNC_TOEPLITZ,
+ .rx_hash_key_len = rss_hash_default_key_len,
+ .rx_hash_key = rss_hash_default_key,
+ .rx_hash_fields_mask = 0,
+ .rwq_ind_tbl = ind_table,
+ };
+ qp_init_attr = (struct ibv_exp_qp_init_attr){
+ .max_inl_recv = 0, /* Currently not supported. */
+ .qp_type = IBV_QPT_RAW_PACKET,
+ .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD |
+ IBV_EXP_QP_INIT_ATTR_RX_HASH),
+ .pd = priv->pd,
+ .rx_hash_conf = &hash_conf,
+ .port_num = priv->port,
+ };
+
+ qp = ibv_exp_create_qp(priv->ctx, &qp_init_attr);
+ if (qp == NULL) {
+ err = (errno ? errno : EINVAL);
+ ERROR("hash RX QP creation failure: %s", strerror(err));
+ goto error;
+ }
+
+ fdir_queue->ind_table = ind_table;
+ fdir_queue->qp = qp;
+
+ return fdir_queue;
+
+error:
+ if (qp != NULL)
+ claim_zero(ibv_destroy_qp(qp));
+
+ if (ind_table != NULL)
+ claim_zero(ibv_exp_destroy_rwq_ind_table(ind_table));
+
+ return NULL;
+}
+
+/**
+ * Enable flow director filter and create steering rules.
+ *
+ * @param priv
+ * Private structure.
+ * @param mlx5_fdir_filter
+ * Filter to create steering rule for.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_filter_enable(struct priv *priv,
+ struct mlx5_fdir_filter *mlx5_fdir_filter)
+{
+ struct fdir_queue *fdir_queue;
+
+ /* Check if flow already exists. */
+ if (mlx5_fdir_filter->flow != NULL)
+ return 0;
+
+ /* Get fdir_queue for specific queue. */
+ fdir_queue = priv_get_fdir_queue(priv, mlx5_fdir_filter->queue);
+
+ if (fdir_queue == NULL) {
+ ERROR("failed to create flow director rxq for queue %d",
+ mlx5_fdir_filter->queue);
+ return EINVAL;
+ }
+
+ /* Create flow */
+ return priv_fdir_flow_add(priv, mlx5_fdir_filter, fdir_queue);
+}
+
+/**
+ * Initialize flow director filters list.
+ *
+ * @param priv
+ * Private structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+fdir_init_filters_list(struct priv *priv)
+{
+ /* Filter list initialization should be done only once. */
+ if (priv->fdir_filter_list)
+ return 0;
+
+ /* Create filters list. */
+ priv->fdir_filter_list =
+ rte_calloc(__func__, 1, sizeof(*priv->fdir_filter_list), 0);
+
+ if (priv->fdir_filter_list == NULL) {
+ int err = ENOMEM;
+
+ ERROR("cannot allocate flow director filter list: %s",
+ strerror(err));
+ return err;
+ }
+
+ LIST_INIT(priv->fdir_filter_list);
+
+ return 0;
+}
+
+/**
+ * Flush all filters.
+ *
+ * @param priv
+ * Private structure.
+ */
+static void
+priv_fdir_filter_flush(struct priv *priv)
+{
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+
+ while ((mlx5_fdir_filter = LIST_FIRST(priv->fdir_filter_list))) {
+ struct ibv_exp_flow *flow = mlx5_fdir_filter->flow;
+
+ DEBUG("%p: flushing flow director filter %p",
+ (void *)priv, (void *)mlx5_fdir_filter);
+ LIST_REMOVE(mlx5_fdir_filter, next);
+ if (flow != NULL)
+ claim_zero(ibv_exp_destroy_flow(flow));
+ rte_free(mlx5_fdir_filter);
+ }
+}
+
+/**
+ * Remove all flow director filters and delete list.
+ *
+ * @param priv
+ * Private structure.
+ */
+void
+priv_fdir_delete_filters_list(struct priv *priv)
+{
+ priv_fdir_filter_flush(priv);
+ rte_free(priv->fdir_filter_list);
+ priv->fdir_filter_list = NULL;
+}
+
+/**
+ * Disable flow director, remove all steering rules.
+ *
+ * @param priv
+ * Private structure.
+ */
+void
+priv_fdir_disable(struct priv *priv)
+{
+ unsigned int i;
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+ struct fdir_queue *fdir_queue;
+
+ /* Run on every flow director filter and destroy flow handle. */
+ LIST_FOREACH(mlx5_fdir_filter, priv->fdir_filter_list, next) {
+ struct ibv_exp_flow *flow;
+
+ /* Only valid elements should be in the list */
+ assert(mlx5_fdir_filter != NULL);
+ flow = mlx5_fdir_filter->flow;
+
+ /* Destroy flow handle */
+ if (flow != NULL) {
+ claim_zero(ibv_exp_destroy_flow(flow));
+ mlx5_fdir_filter->flow = NULL;
+ }
+ }
+
+ /* Run on every RX queue to destroy related flow director QP and
+ * indirection table. */
+ for (i = 0; (i != priv->rxqs_n); i++) {
+ fdir_queue = &(*priv->rxqs)[i]->fdir_queue;
+
+ if (fdir_queue->qp != NULL) {
+ claim_zero(ibv_destroy_qp(fdir_queue->qp));
+ fdir_queue->qp = NULL;
+ }
+
+ if (fdir_queue->ind_table != NULL) {
+ claim_zero(ibv_exp_destroy_rwq_ind_table
+ (fdir_queue->ind_table));
+ fdir_queue->ind_table = NULL;
+ }
+ }
+}
+
+/**
+ * Enable flow director, create steering rules.
+ *
+ * @param priv
+ * Private structure.
+ */
+void
+priv_fdir_enable(struct priv *priv)
+{
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+
+ /* Run on every fdir filter and create flow handle */
+ LIST_FOREACH(mlx5_fdir_filter, priv->fdir_filter_list, next) {
+ /* Only valid elements should be in the list */
+ assert(mlx5_fdir_filter != NULL);
+
+ priv_fdir_filter_enable(priv, mlx5_fdir_filter);
+ }
+}
+
+/**
+ * Find specific filter in list.
+ *
+ * @param priv
+ * Private structure.
+ * @param fdir_filter
+ * Flow director filter to find.
+ *
+ * @return
+ * Filter element if found, otherwise NULL.
+ */
+static struct mlx5_fdir_filter *
+priv_find_filter_in_list(struct priv *priv,
+ const struct rte_eth_fdir_filter *fdir_filter)
+{
+ struct fdir_flow_desc desc;
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+ enum rte_fdir_mode fdir_mode = priv->dev->data->dev_conf.fdir_conf.mode;
+
+ /* Get flow director filter to look for. */
+ fdir_filter_to_flow_desc(fdir_filter, &desc, fdir_mode);
+
+ /* Look for the requested element. */
+ LIST_FOREACH(mlx5_fdir_filter, priv->fdir_filter_list, next) {
+ /* Only valid elements should be in the list. */
+ assert(mlx5_fdir_filter != NULL);
+
+ /* Return matching filter. */
+ if (!memcmp(&desc, &mlx5_fdir_filter->desc, sizeof(desc)))
+ return mlx5_fdir_filter;
+ }
+
+ /* Filter not found */
+ return NULL;
+}
+
+/**
+ * Add new flow director filter and store it in list.
+ *
+ * @param priv
+ * Private structure.
+ * @param fdir_filter
+ * Flow director filter to add.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_filter_add(struct priv *priv,
+ const struct rte_eth_fdir_filter *fdir_filter)
+{
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+ enum rte_fdir_mode fdir_mode = priv->dev->data->dev_conf.fdir_conf.mode;
+ int err = 0;
+
+ /* Validate queue number. */
+ if (fdir_filter->action.rx_queue >= priv->rxqs_n) {
+ ERROR("invalid queue number %d", fdir_filter->action.rx_queue);
+ return EINVAL;
+ }
+
+ /* Duplicate filters are currently unsupported. */
+ mlx5_fdir_filter = priv_find_filter_in_list(priv, fdir_filter);
+ if (mlx5_fdir_filter != NULL) {
+ ERROR("filter already exists");
+ return EINVAL;
+ }
+
+ /* Create new flow director filter. */
+ mlx5_fdir_filter =
+ rte_calloc(__func__, 1, sizeof(*mlx5_fdir_filter), 0);
+ if (mlx5_fdir_filter == NULL) {
+ err = ENOMEM;
+ ERROR("cannot allocate flow director filter: %s",
+ strerror(err));
+ return err;
+ }
+
+ /* Set queue. */
+ mlx5_fdir_filter->queue = fdir_filter->action.rx_queue;
+
+ /* Convert to mlx5 filter descriptor. */
+ fdir_filter_to_flow_desc(fdir_filter,
+ &mlx5_fdir_filter->desc, fdir_mode);
+
+ /* Insert new filter into list. */
+ LIST_INSERT_HEAD(priv->fdir_filter_list, mlx5_fdir_filter, next);
+
+ DEBUG("%p: flow director filter %p added",
+ (void *)priv, (void *)mlx5_fdir_filter);
+
+ /* Enable filter immediately if device is started. */
+ if (priv->started)
+ err = priv_fdir_filter_enable(priv, mlx5_fdir_filter);
+
+ return err;
+}
+
+/**
+ * Update queue for specific filter.
+ *
+ * @param priv
+ * Private structure.
+ * @param fdir_filter
+ * Filter to be updated.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_filter_update(struct priv *priv,
+ const struct rte_eth_fdir_filter *fdir_filter)
+{
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+
+ /* Validate queue number. */
+ if (fdir_filter->action.rx_queue >= priv->rxqs_n) {
+ ERROR("invalid queue number %d", fdir_filter->action.rx_queue);
+ return EINVAL;
+ }
+
+ mlx5_fdir_filter = priv_find_filter_in_list(priv, fdir_filter);
+ if (mlx5_fdir_filter != NULL) {
+ struct ibv_exp_flow *flow = mlx5_fdir_filter->flow;
+ int err = 0;
+
+ /* Update queue number. */
+ mlx5_fdir_filter->queue = fdir_filter->action.rx_queue;
+
+ /* Destroy flow handle. */
+ if (flow != NULL) {
+ claim_zero(ibv_exp_destroy_flow(flow));
+ mlx5_fdir_filter->flow = NULL;
+ }
+ DEBUG("%p: flow director filter %p updated",
+ (void *)priv, (void *)mlx5_fdir_filter);
+
+ /* Enable filter if device is started. */
+ if (priv->started)
+ err = priv_fdir_filter_enable(priv, mlx5_fdir_filter);
+
+ return err;
+ }
+
+ /* Filter not found, create it. */
+ DEBUG("%p: filter not found for update, creating new filter",
+ (void *)priv);
+ return priv_fdir_filter_add(priv, fdir_filter);
+}
+
+/**
+ * Delete specific filter.
+ *
+ * @param priv
+ * Private structure.
+ * @param fdir_filter
+ * Filter to be deleted.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_filter_delete(struct priv *priv,
+ const struct rte_eth_fdir_filter *fdir_filter)
+{
+ struct mlx5_fdir_filter *mlx5_fdir_filter;
+
+ mlx5_fdir_filter = priv_find_filter_in_list(priv, fdir_filter);
+ if (mlx5_fdir_filter != NULL) {
+ struct ibv_exp_flow *flow = mlx5_fdir_filter->flow;
+
+ /* Remove element from list. */
+ LIST_REMOVE(mlx5_fdir_filter, next);
+
+ /* Destroy flow handle. */
+ if (flow != NULL) {
+ claim_zero(ibv_exp_destroy_flow(flow));
+ mlx5_fdir_filter->flow = NULL;
+ }
+
+ DEBUG("%p: flow director filter %p deleted",
+ (void *)priv, (void *)mlx5_fdir_filter);
+
+ /* Delete filter. */
+ rte_free(mlx5_fdir_filter);
+
+ return 0;
+ }
+
+ ERROR("%p: flow director delete failed, cannot find filter",
+ (void *)priv);
+ return EINVAL;
+}
+
+/**
+ * Get flow director information.
+ *
+ * @param priv
+ * Private structure.
+ * @param[out] fdir_info
+ * Resulting flow director information.
+ */
+static void
+priv_fdir_info_get(struct priv *priv, struct rte_eth_fdir_info *fdir_info)
+{
+ struct rte_eth_fdir_masks *mask =
+ &priv->dev->data->dev_conf.fdir_conf.mask;
+
+ fdir_info->mode = priv->dev->data->dev_conf.fdir_conf.mode;
+ fdir_info->guarant_spc = 0;
+
+ rte_memcpy(&fdir_info->mask, mask, sizeof(fdir_info->mask));
+
+ fdir_info->max_flexpayload = 0;
+ fdir_info->flow_types_mask[0] = 0;
+
+ fdir_info->flex_payload_unit = 0;
+ fdir_info->max_flex_payload_segment_num = 0;
+ fdir_info->flex_payload_limit = 0;
+ memset(&fdir_info->flex_conf, 0, sizeof(fdir_info->flex_conf));
+}
+
+/**
+ * Deal with flow director operations.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param filter_op
+ * Operation to perform.
+ * @param arg
+ * Pointer to operation-specific structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_fdir_ctrl_func(struct priv *priv, enum rte_filter_op filter_op, void *arg)
+{
+ enum rte_fdir_mode fdir_mode =
+ priv->dev->data->dev_conf.fdir_conf.mode;
+ int ret = 0;
+
+ if (filter_op == RTE_ETH_FILTER_NOP)
+ return 0;
+
+ if (fdir_mode != RTE_FDIR_MODE_PERFECT &&
+ fdir_mode != RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
+ ERROR("%p: flow director mode %d not supported",
+ (void *)priv, fdir_mode);
+ return EINVAL;
+ }
+
+ switch (filter_op) {
+ case RTE_ETH_FILTER_ADD:
+ ret = priv_fdir_filter_add(priv, arg);
+ break;
+ case RTE_ETH_FILTER_UPDATE:
+ ret = priv_fdir_filter_update(priv, arg);
+ break;
+ case RTE_ETH_FILTER_DELETE:
+ ret = priv_fdir_filter_delete(priv, arg);
+ break;
+ case RTE_ETH_FILTER_FLUSH:
+ priv_fdir_filter_flush(priv);
+ break;
+ case RTE_ETH_FILTER_INFO:
+ priv_fdir_info_get(priv, arg);
+ break;
+ default:
+ DEBUG("%p: unknown operation %u", (void *)priv, filter_op);
+ ret = EINVAL;
+ break;
+ }
+ return ret;
+}
+
+/**
+ * Manage filter operations.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param filter_type
+ * Filter type.
+ * @param filter_op
+ * Operation to perform.
+ * @param arg
+ * Pointer to operation-specific structure.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_filter_ctrl(struct rte_eth_dev *dev,
+ enum rte_filter_type filter_type,
+ enum rte_filter_op filter_op,
+ void *arg)
+{
+ int ret = -EINVAL;
+ struct priv *priv = dev->data->dev_private;
+
+ switch (filter_type) {
+ case RTE_ETH_FILTER_FDIR:
+ priv_lock(priv);
+ ret = priv_fdir_ctrl_func(priv, filter_op, arg);
+ priv_unlock(priv);
+ break;
+ default:
+ ERROR("%p: filter type (%d) not supported",
+ (void *)dev, filter_type);
+ break;
+ }
+
+ return ret;
+}
diff --git a/drivers/net/mlx5/mlx5_mac.c b/drivers/net/mlx5/mlx5_mac.c
new file mode 100644
index 00000000..c9cea485
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_mac.c
@@ -0,0 +1,510 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <assert.h>
+#include <stdint.h>
+#include <string.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <netinet/in.h>
+#include <linux/if.h>
+#include <sys/ioctl.h>
+#include <arpa/inet.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_utils.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_defs.h"
+
+/**
+ * Get MAC address by querying netdevice.
+ *
+ * @param[in] priv
+ * struct priv for the requested device.
+ * @param[out] mac
+ * MAC address output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN])
+{
+ struct ifreq request;
+
+ if (priv_ifreq(priv, SIOCGIFHWADDR, &request))
+ return -1;
+ memcpy(mac, request.ifr_hwaddr.sa_data, ETHER_ADDR_LEN);
+ return 0;
+}
+
+/**
+ * Delete MAC flow steering rule.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param mac_index
+ * MAC address index.
+ * @param vlan_index
+ * VLAN index to use.
+ */
+static void
+hash_rxq_del_mac_flow(struct hash_rxq *hash_rxq, unsigned int mac_index,
+ unsigned int vlan_index)
+{
+#ifndef NDEBUG
+ const uint8_t (*mac)[ETHER_ADDR_LEN] =
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ hash_rxq->priv->mac[mac_index].addr_bytes;
+#endif
+
+ assert(mac_index < RTE_DIM(hash_rxq->mac_flow));
+ assert(vlan_index < RTE_DIM(hash_rxq->mac_flow[mac_index]));
+ if (hash_rxq->mac_flow[mac_index][vlan_index] == NULL)
+ return;
+ DEBUG("%p: removing MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u"
+ " VLAN index %u",
+ (void *)hash_rxq,
+ (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5],
+ mac_index,
+ vlan_index);
+ claim_zero(ibv_exp_destroy_flow(hash_rxq->mac_flow
+ [mac_index][vlan_index]));
+ hash_rxq->mac_flow[mac_index][vlan_index] = NULL;
+}
+
+/**
+ * Unregister a MAC address from a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param mac_index
+ * MAC address index.
+ */
+static void
+hash_rxq_mac_addr_del(struct hash_rxq *hash_rxq, unsigned int mac_index)
+{
+ unsigned int i;
+
+ assert(mac_index < RTE_DIM(hash_rxq->mac_flow));
+ for (i = 0; (i != RTE_DIM(hash_rxq->mac_flow[mac_index])); ++i)
+ hash_rxq_del_mac_flow(hash_rxq, mac_index, i);
+}
+
+/**
+ * Unregister all MAC addresses from a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ */
+void
+hash_rxq_mac_addrs_del(struct hash_rxq *hash_rxq)
+{
+ unsigned int i;
+
+ for (i = 0; (i != RTE_DIM(hash_rxq->mac_flow)); ++i)
+ hash_rxq_mac_addr_del(hash_rxq, i);
+}
+
+/**
+ * Unregister a MAC address.
+ *
+ * This is done for each hash RX queue.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param mac_index
+ * MAC address index.
+ */
+static void
+priv_mac_addr_del(struct priv *priv, unsigned int mac_index)
+{
+ unsigned int i;
+
+ assert(mac_index < RTE_DIM(priv->mac));
+ if (!BITFIELD_ISSET(priv->mac_configured, mac_index))
+ return;
+ for (i = 0; (i != priv->hash_rxqs_n); ++i)
+ hash_rxq_mac_addr_del(&(*priv->hash_rxqs)[i], mac_index);
+ BITFIELD_RESET(priv->mac_configured, mac_index);
+}
+
+/**
+ * Unregister all MAC addresses from all hash RX queues.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+void
+priv_mac_addrs_disable(struct priv *priv)
+{
+ unsigned int i;
+
+ for (i = 0; (i != priv->hash_rxqs_n); ++i)
+ hash_rxq_mac_addrs_del(&(*priv->hash_rxqs)[i]);
+}
+
+/**
+ * DPDK callback to remove a MAC address.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param index
+ * MAC address index.
+ */
+void
+mlx5_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ DEBUG("%p: removing MAC address from index %" PRIu32,
+ (void *)dev, index);
+ if (index >= RTE_DIM(priv->mac))
+ goto end;
+ priv_mac_addr_del(priv, index);
+end:
+ priv_unlock(priv);
+}
+
+/**
+ * Add MAC flow steering rule.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param mac_index
+ * MAC address index to register.
+ * @param vlan_index
+ * VLAN index to use.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+hash_rxq_add_mac_flow(struct hash_rxq *hash_rxq, unsigned int mac_index,
+ unsigned int vlan_index)
+{
+ struct ibv_exp_flow *flow;
+ struct priv *priv = hash_rxq->priv;
+ const uint8_t (*mac)[ETHER_ADDR_LEN] =
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ priv->mac[mac_index].addr_bytes;
+ FLOW_ATTR_SPEC_ETH(data, priv_flow_attr(priv, NULL, 0, hash_rxq->type));
+ struct ibv_exp_flow_attr *attr = &data->attr;
+ struct ibv_exp_flow_spec_eth *spec = &data->spec;
+ unsigned int vlan_enabled = !!priv->vlan_filter_n;
+ unsigned int vlan_id = priv->vlan_filter[vlan_index];
+
+ assert(mac_index < RTE_DIM(hash_rxq->mac_flow));
+ assert(vlan_index < RTE_DIM(hash_rxq->mac_flow[mac_index]));
+ if (hash_rxq->mac_flow[mac_index][vlan_index] != NULL)
+ return 0;
+ /*
+ * No padding must be inserted by the compiler between attr and spec.
+ * This layout is expected by libibverbs.
+ */
+ assert(((uint8_t *)attr + sizeof(*attr)) == (uint8_t *)spec);
+ priv_flow_attr(priv, attr, sizeof(data), hash_rxq->type);
+ /* The first specification must be Ethernet. */
+ assert(spec->type == IBV_EXP_FLOW_SPEC_ETH);
+ assert(spec->size == sizeof(*spec));
+ *spec = (struct ibv_exp_flow_spec_eth){
+ .type = IBV_EXP_FLOW_SPEC_ETH,
+ .size = sizeof(*spec),
+ .val = {
+ .dst_mac = {
+ (*mac)[0], (*mac)[1], (*mac)[2],
+ (*mac)[3], (*mac)[4], (*mac)[5]
+ },
+ .vlan_tag = (vlan_enabled ? htons(vlan_id) : 0),
+ },
+ .mask = {
+ .dst_mac = "\xff\xff\xff\xff\xff\xff",
+ .vlan_tag = (vlan_enabled ? htons(0xfff) : 0),
+ },
+ };
+ DEBUG("%p: adding MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u"
+ " VLAN index %u filtering %s, ID %u",
+ (void *)hash_rxq,
+ (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5],
+ mac_index,
+ vlan_index,
+ (vlan_enabled ? "enabled" : "disabled"),
+ vlan_id);
+ /* Create related flow. */
+ errno = 0;
+ flow = ibv_exp_create_flow(hash_rxq->qp, attr);
+ if (flow == NULL) {
+ /* It's not clear whether errno is always set in this case. */
+ ERROR("%p: flow configuration failed, errno=%d: %s",
+ (void *)hash_rxq, errno,
+ (errno ? strerror(errno) : "Unknown error"));
+ if (errno)
+ return errno;
+ return EINVAL;
+ }
+ hash_rxq->mac_flow[mac_index][vlan_index] = flow;
+ return 0;
+}
+
+/**
+ * Register a MAC address in a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param mac_index
+ * MAC address index to register.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+hash_rxq_mac_addr_add(struct hash_rxq *hash_rxq, unsigned int mac_index)
+{
+ struct priv *priv = hash_rxq->priv;
+ unsigned int i = 0;
+ int ret;
+
+ assert(mac_index < RTE_DIM(hash_rxq->mac_flow));
+ assert(RTE_DIM(hash_rxq->mac_flow[mac_index]) ==
+ RTE_DIM(priv->vlan_filter));
+ /* Add a MAC address for each VLAN filter, or at least once. */
+ do {
+ ret = hash_rxq_add_mac_flow(hash_rxq, mac_index, i);
+ if (ret) {
+ /* Failure, rollback. */
+ while (i != 0)
+ hash_rxq_del_mac_flow(hash_rxq, mac_index,
+ --i);
+ return ret;
+ }
+ } while (++i < priv->vlan_filter_n);
+ return 0;
+}
+
+/**
+ * Register all MAC addresses in a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+hash_rxq_mac_addrs_add(struct hash_rxq *hash_rxq)
+{
+ struct priv *priv = hash_rxq->priv;
+ unsigned int i;
+ int ret;
+
+ assert(RTE_DIM(priv->mac) == RTE_DIM(hash_rxq->mac_flow));
+ for (i = 0; (i != RTE_DIM(priv->mac)); ++i) {
+ if (!BITFIELD_ISSET(priv->mac_configured, i))
+ continue;
+ ret = hash_rxq_mac_addr_add(hash_rxq, i);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0)
+ hash_rxq_mac_addr_del(hash_rxq, --i);
+ assert(ret > 0);
+ return ret;
+ }
+ return 0;
+}
+
+/**
+ * Register a MAC address.
+ *
+ * This is done for each hash RX queue.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param mac_index
+ * MAC address index to use.
+ * @param mac
+ * MAC address to register.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_mac_addr_add(struct priv *priv, unsigned int mac_index,
+ const uint8_t (*mac)[ETHER_ADDR_LEN])
+{
+ unsigned int i;
+ int ret;
+
+ assert(mac_index < RTE_DIM(priv->mac));
+ /* First, make sure this address isn't already configured. */
+ for (i = 0; (i != RTE_DIM(priv->mac)); ++i) {
+ /* Skip this index, it's going to be reconfigured. */
+ if (i == mac_index)
+ continue;
+ if (!BITFIELD_ISSET(priv->mac_configured, i))
+ continue;
+ if (memcmp(priv->mac[i].addr_bytes, *mac, sizeof(*mac)))
+ continue;
+ /* Address already configured elsewhere, return with error. */
+ return EADDRINUSE;
+ }
+ if (BITFIELD_ISSET(priv->mac_configured, mac_index))
+ priv_mac_addr_del(priv, mac_index);
+ priv->mac[mac_index] = (struct ether_addr){
+ {
+ (*mac)[0], (*mac)[1], (*mac)[2],
+ (*mac)[3], (*mac)[4], (*mac)[5]
+ }
+ };
+ if (!priv_allow_flow_type(priv, HASH_RXQ_FLOW_TYPE_MAC))
+ goto end;
+ for (i = 0; (i != priv->hash_rxqs_n); ++i) {
+ ret = hash_rxq_mac_addr_add(&(*priv->hash_rxqs)[i], mac_index);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0)
+ hash_rxq_mac_addr_del(&(*priv->hash_rxqs)[--i],
+ mac_index);
+ return ret;
+ }
+end:
+ BITFIELD_SET(priv->mac_configured, mac_index);
+ return 0;
+}
+
+/**
+ * Register all MAC addresses in all hash RX queues.
+ *
+ * @param priv
+ * Pointer to private structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_mac_addrs_enable(struct priv *priv)
+{
+ unsigned int i;
+ int ret;
+
+ if (!priv_allow_flow_type(priv, HASH_RXQ_FLOW_TYPE_MAC))
+ return 0;
+ for (i = 0; (i != priv->hash_rxqs_n); ++i) {
+ ret = hash_rxq_mac_addrs_add(&(*priv->hash_rxqs)[i]);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0)
+ hash_rxq_mac_addrs_del(&(*priv->hash_rxqs)[--i]);
+ assert(ret > 0);
+ return ret;
+ }
+ return 0;
+}
+
+/**
+ * DPDK callback to add a MAC address.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param mac_addr
+ * MAC address to register.
+ * @param index
+ * MAC address index.
+ * @param vmdq
+ * VMDq pool index to associate address with (ignored).
+ */
+void
+mlx5_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
+ uint32_t index, uint32_t vmdq)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ if (mlx5_is_secondary())
+ return;
+
+ (void)vmdq;
+ priv_lock(priv);
+ DEBUG("%p: adding MAC address at index %" PRIu32,
+ (void *)dev, index);
+ if (index >= RTE_DIM(priv->mac))
+ goto end;
+ priv_mac_addr_add(priv, index,
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ mac_addr->addr_bytes);
+end:
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback to set primary MAC address.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param mac_addr
+ * MAC address to register.
+ */
+void
+mlx5_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
+{
+ DEBUG("%p: setting primary MAC address", (void *)dev);
+ mlx5_mac_addr_remove(dev, 0);
+ mlx5_mac_addr_add(dev, mac_addr, 0, 0);
+}
diff --git a/drivers/net/mlx5/mlx5_rss.c b/drivers/net/mlx5/mlx5_rss.c
new file mode 100644
index 00000000..639e935b
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_rss.c
@@ -0,0 +1,367 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include <errno.h>
+#include <string.h>
+#include <assert.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+
+/**
+ * Get a RSS configuration hash key.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param rss_hf
+ * RSS hash functions configuration must be retrieved for.
+ *
+ * @return
+ * Pointer to a RSS configuration structure or NULL if rss_hf cannot
+ * be matched.
+ */
+static struct rte_eth_rss_conf *
+rss_hash_get(struct priv *priv, uint64_t rss_hf)
+{
+ unsigned int i;
+
+ for (i = 0; (i != hash_rxq_init_n); ++i) {
+ uint64_t dpdk_rss_hf = hash_rxq_init[i].dpdk_rss_hf;
+
+ if (!(dpdk_rss_hf & rss_hf))
+ continue;
+ return (*priv->rss_conf)[i];
+ }
+ return NULL;
+}
+
+/**
+ * Register a RSS key.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param key
+ * Hash key to register.
+ * @param key_len
+ * Hash key length in bytes.
+ * @param rss_hf
+ * RSS hash functions the provided key applies to.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+rss_hash_rss_conf_new_key(struct priv *priv, const uint8_t *key,
+ unsigned int key_len, uint64_t rss_hf)
+{
+ unsigned int i;
+
+ for (i = 0; (i != hash_rxq_init_n); ++i) {
+ struct rte_eth_rss_conf *rss_conf;
+ uint64_t dpdk_rss_hf = hash_rxq_init[i].dpdk_rss_hf;
+
+ if (!(dpdk_rss_hf & rss_hf))
+ continue;
+ rss_conf = rte_realloc((*priv->rss_conf)[i],
+ (sizeof(*rss_conf) + key_len),
+ 0);
+ if (!rss_conf)
+ return ENOMEM;
+ rss_conf->rss_key = (void *)(rss_conf + 1);
+ rss_conf->rss_key_len = key_len;
+ rss_conf->rss_hf = dpdk_rss_hf;
+ memcpy(rss_conf->rss_key, key, key_len);
+ (*priv->rss_conf)[i] = rss_conf;
+ }
+ return 0;
+}
+
+/**
+ * DPDK callback to update the RSS hash configuration.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[in] rss_conf
+ * RSS configuration data.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_rss_hash_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf)
+{
+ struct priv *priv = dev->data->dev_private;
+ int err = 0;
+
+ priv_lock(priv);
+
+ assert(priv->rss_conf != NULL);
+
+ /* Apply configuration. */
+ if (rss_conf->rss_key)
+ err = rss_hash_rss_conf_new_key(priv,
+ rss_conf->rss_key,
+ rss_conf->rss_key_len,
+ rss_conf->rss_hf);
+ /* Store protocols for which RSS is enabled. */
+ priv->rss_hf = rss_conf->rss_hf;
+ priv_unlock(priv);
+ assert(err >= 0);
+ return -err;
+}
+
+/**
+ * DPDK callback to get the RSS hash configuration.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[in, out] rss_conf
+ * RSS configuration data.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_rss_hash_conf_get(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct rte_eth_rss_conf *priv_rss_conf;
+
+ priv_lock(priv);
+
+ assert(priv->rss_conf != NULL);
+
+ priv_rss_conf = rss_hash_get(priv, rss_conf->rss_hf);
+ if (!priv_rss_conf) {
+ rss_conf->rss_hf = 0;
+ priv_unlock(priv);
+ return -EINVAL;
+ }
+ if (rss_conf->rss_key &&
+ rss_conf->rss_key_len >= priv_rss_conf->rss_key_len)
+ memcpy(rss_conf->rss_key,
+ priv_rss_conf->rss_key,
+ priv_rss_conf->rss_key_len);
+ rss_conf->rss_key_len = priv_rss_conf->rss_key_len;
+ rss_conf->rss_hf = priv_rss_conf->rss_hf;
+
+ priv_unlock(priv);
+ return 0;
+}
+
+/**
+ * Allocate/reallocate RETA index table.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @praram reta_size
+ * The size of the array to allocate.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_rss_reta_index_resize(struct priv *priv, unsigned int reta_size)
+{
+ void *mem;
+ unsigned int old_size = priv->reta_idx_n;
+
+ if (priv->reta_idx_n == reta_size)
+ return 0;
+
+ mem = rte_realloc(priv->reta_idx,
+ reta_size * sizeof((*priv->reta_idx)[0]), 0);
+ if (!mem)
+ return ENOMEM;
+ priv->reta_idx = mem;
+ priv->reta_idx_n = reta_size;
+
+ if (old_size < reta_size)
+ memset(&(*priv->reta_idx)[old_size], 0,
+ (reta_size - old_size) *
+ sizeof((*priv->reta_idx)[0]));
+ return 0;
+}
+
+/**
+ * Query RETA table.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[in, out] reta_conf
+ * Pointer to the first RETA configuration structure.
+ * @param reta_size
+ * Number of entries.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_dev_rss_reta_query(struct priv *priv,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ unsigned int reta_size)
+{
+ unsigned int idx;
+ unsigned int i;
+ int ret;
+
+ /* See RETA comment in mlx5_dev_infos_get(). */
+ ret = priv_rss_reta_index_resize(priv, priv->ind_table_max_size);
+ if (ret)
+ return ret;
+
+ /* Fill each entry of the table even if its bit is not set. */
+ for (idx = 0, i = 0; (i != reta_size); ++i) {
+ idx = i / RTE_RETA_GROUP_SIZE;
+ reta_conf[idx].reta[i % RTE_RETA_GROUP_SIZE] =
+ (*priv->reta_idx)[i];
+ }
+ return 0;
+}
+
+/**
+ * Update RETA table.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[in] reta_conf
+ * Pointer to the first RETA configuration structure.
+ * @param reta_size
+ * Number of entries.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+priv_dev_rss_reta_update(struct priv *priv,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ unsigned int reta_size)
+{
+ unsigned int idx;
+ unsigned int i;
+ unsigned int pos;
+ int ret;
+
+ /* See RETA comment in mlx5_dev_infos_get(). */
+ ret = priv_rss_reta_index_resize(priv, priv->ind_table_max_size);
+ if (ret)
+ return ret;
+
+ for (idx = 0, i = 0; (i != reta_size); ++i) {
+ idx = i / RTE_RETA_GROUP_SIZE;
+ pos = i % RTE_RETA_GROUP_SIZE;
+ if (((reta_conf[idx].mask >> i) & 0x1) == 0)
+ continue;
+ assert(reta_conf[idx].reta[pos] < priv->rxqs_n);
+ (*priv->reta_idx)[i] = reta_conf[idx].reta[pos];
+ }
+ return 0;
+}
+
+/**
+ * DPDK callback to get the RETA indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param reta_conf
+ * Pointer to RETA configuration structure array.
+ * @param reta_size
+ * Size of the RETA table.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_rss_reta_query(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size)
+{
+ int ret;
+ struct priv *priv = dev->data->dev_private;
+
+ priv_lock(priv);
+ ret = priv_dev_rss_reta_query(priv, reta_conf, reta_size);
+ priv_unlock(priv);
+ return -ret;
+}
+
+/**
+ * DPDK callback to update the RETA indirection table.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param reta_conf
+ * Pointer to RETA configuration structure array.
+ * @param reta_size
+ * Size of the RETA table.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_rss_reta_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size)
+{
+ int ret;
+ struct priv *priv = dev->data->dev_private;
+
+ priv_lock(priv);
+ ret = priv_dev_rss_reta_update(priv, reta_conf, reta_size);
+ priv_unlock(priv);
+ return -ret;
+}
diff --git a/drivers/net/mlx5/mlx5_rxmode.c b/drivers/net/mlx5/mlx5_rxmode.c
new file mode 100644
index 00000000..3a55f633
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_rxmode.c
@@ -0,0 +1,481 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <errno.h>
+#include <string.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ethdev.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_utils.h"
+
+/* Initialization data for special flows. */
+static const struct special_flow_init special_flow_init[] = {
+ [HASH_RXQ_FLOW_TYPE_PROMISC] = {
+ .dst_mac_val = "\x00\x00\x00\x00\x00\x00",
+ .dst_mac_mask = "\x00\x00\x00\x00\x00\x00",
+ .hash_types =
+ 1 << HASH_RXQ_TCPV4 |
+ 1 << HASH_RXQ_UDPV4 |
+ 1 << HASH_RXQ_IPV4 |
+#ifdef HAVE_FLOW_SPEC_IPV6
+ 1 << HASH_RXQ_TCPV6 |
+ 1 << HASH_RXQ_UDPV6 |
+ 1 << HASH_RXQ_IPV6 |
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ 1 << HASH_RXQ_ETH |
+ 0,
+ .per_vlan = 0,
+ },
+ [HASH_RXQ_FLOW_TYPE_ALLMULTI] = {
+ .dst_mac_val = "\x01\x00\x00\x00\x00\x00",
+ .dst_mac_mask = "\x01\x00\x00\x00\x00\x00",
+ .hash_types =
+ 1 << HASH_RXQ_UDPV4 |
+ 1 << HASH_RXQ_IPV4 |
+#ifdef HAVE_FLOW_SPEC_IPV6
+ 1 << HASH_RXQ_UDPV6 |
+ 1 << HASH_RXQ_IPV6 |
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ 1 << HASH_RXQ_ETH |
+ 0,
+ .per_vlan = 0,
+ },
+ [HASH_RXQ_FLOW_TYPE_BROADCAST] = {
+ .dst_mac_val = "\xff\xff\xff\xff\xff\xff",
+ .dst_mac_mask = "\xff\xff\xff\xff\xff\xff",
+ .hash_types =
+ 1 << HASH_RXQ_UDPV4 |
+ 1 << HASH_RXQ_IPV4 |
+#ifdef HAVE_FLOW_SPEC_IPV6
+ 1 << HASH_RXQ_UDPV6 |
+ 1 << HASH_RXQ_IPV6 |
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ 1 << HASH_RXQ_ETH |
+ 0,
+ .per_vlan = 1,
+ },
+#ifdef HAVE_FLOW_SPEC_IPV6
+ [HASH_RXQ_FLOW_TYPE_IPV6MULTI] = {
+ .dst_mac_val = "\x33\x33\x00\x00\x00\x00",
+ .dst_mac_mask = "\xff\xff\x00\x00\x00\x00",
+ .hash_types =
+ 1 << HASH_RXQ_UDPV6 |
+ 1 << HASH_RXQ_IPV6 |
+ 1 << HASH_RXQ_ETH |
+ 0,
+ .per_vlan = 1,
+ },
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+};
+
+/**
+ * Enable a special flow in a hash RX queue for a given VLAN index.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param flow_type
+ * Special flow type.
+ * @param vlan_index
+ * VLAN index to use.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+hash_rxq_special_flow_enable_vlan(struct hash_rxq *hash_rxq,
+ enum hash_rxq_flow_type flow_type,
+ unsigned int vlan_index)
+{
+ struct priv *priv = hash_rxq->priv;
+ struct ibv_exp_flow *flow;
+ FLOW_ATTR_SPEC_ETH(data, priv_flow_attr(priv, NULL, 0, hash_rxq->type));
+ struct ibv_exp_flow_attr *attr = &data->attr;
+ struct ibv_exp_flow_spec_eth *spec = &data->spec;
+ const uint8_t *mac;
+ const uint8_t *mask;
+ unsigned int vlan_enabled = (priv->vlan_filter_n &&
+ special_flow_init[flow_type].per_vlan);
+ unsigned int vlan_id = priv->vlan_filter[vlan_index];
+
+ /* Check if flow is relevant for this hash_rxq. */
+ if (!(special_flow_init[flow_type].hash_types & (1 << hash_rxq->type)))
+ return 0;
+ /* Check if flow already exists. */
+ if (hash_rxq->special_flow[flow_type][vlan_index] != NULL)
+ return 0;
+
+ /*
+ * No padding must be inserted by the compiler between attr and spec.
+ * This layout is expected by libibverbs.
+ */
+ assert(((uint8_t *)attr + sizeof(*attr)) == (uint8_t *)spec);
+ priv_flow_attr(priv, attr, sizeof(data), hash_rxq->type);
+ /* The first specification must be Ethernet. */
+ assert(spec->type == IBV_EXP_FLOW_SPEC_ETH);
+ assert(spec->size == sizeof(*spec));
+
+ mac = special_flow_init[flow_type].dst_mac_val;
+ mask = special_flow_init[flow_type].dst_mac_mask;
+ *spec = (struct ibv_exp_flow_spec_eth){
+ .type = IBV_EXP_FLOW_SPEC_ETH,
+ .size = sizeof(*spec),
+ .val = {
+ .dst_mac = {
+ mac[0], mac[1], mac[2],
+ mac[3], mac[4], mac[5],
+ },
+ .vlan_tag = (vlan_enabled ? htons(vlan_id) : 0),
+ },
+ .mask = {
+ .dst_mac = {
+ mask[0], mask[1], mask[2],
+ mask[3], mask[4], mask[5],
+ },
+ .vlan_tag = (vlan_enabled ? htons(0xfff) : 0),
+ },
+ };
+
+ errno = 0;
+ flow = ibv_exp_create_flow(hash_rxq->qp, attr);
+ if (flow == NULL) {
+ /* It's not clear whether errno is always set in this case. */
+ ERROR("%p: flow configuration failed, errno=%d: %s",
+ (void *)hash_rxq, errno,
+ (errno ? strerror(errno) : "Unknown error"));
+ if (errno)
+ return errno;
+ return EINVAL;
+ }
+ hash_rxq->special_flow[flow_type][vlan_index] = flow;
+ DEBUG("%p: special flow %s (index %d) VLAN %u (index %u) enabled",
+ (void *)hash_rxq, hash_rxq_flow_type_str(flow_type), flow_type,
+ vlan_id, vlan_index);
+ return 0;
+}
+
+/**
+ * Disable a special flow in a hash RX queue for a given VLAN index.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param flow_type
+ * Special flow type.
+ * @param vlan_index
+ * VLAN index to use.
+ */
+static void
+hash_rxq_special_flow_disable_vlan(struct hash_rxq *hash_rxq,
+ enum hash_rxq_flow_type flow_type,
+ unsigned int vlan_index)
+{
+ struct ibv_exp_flow *flow =
+ hash_rxq->special_flow[flow_type][vlan_index];
+
+ if (flow == NULL)
+ return;
+ claim_zero(ibv_exp_destroy_flow(flow));
+ hash_rxq->special_flow[flow_type][vlan_index] = NULL;
+ DEBUG("%p: special flow %s (index %d) VLAN %u (index %u) disabled",
+ (void *)hash_rxq, hash_rxq_flow_type_str(flow_type), flow_type,
+ hash_rxq->priv->vlan_filter[vlan_index], vlan_index);
+}
+
+/**
+ * Enable a special flow in a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param flow_type
+ * Special flow type.
+ * @param vlan_index
+ * VLAN index to use.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+hash_rxq_special_flow_enable(struct hash_rxq *hash_rxq,
+ enum hash_rxq_flow_type flow_type)
+{
+ struct priv *priv = hash_rxq->priv;
+ unsigned int i = 0;
+ int ret;
+
+ assert((unsigned int)flow_type < RTE_DIM(hash_rxq->special_flow));
+ assert(RTE_DIM(hash_rxq->special_flow[flow_type]) ==
+ RTE_DIM(priv->vlan_filter));
+ /* Add a special flow for each VLAN filter when relevant. */
+ do {
+ ret = hash_rxq_special_flow_enable_vlan(hash_rxq, flow_type, i);
+ if (ret) {
+ /* Failure, rollback. */
+ while (i != 0)
+ hash_rxq_special_flow_disable_vlan(hash_rxq,
+ flow_type,
+ --i);
+ return ret;
+ }
+ } while (special_flow_init[flow_type].per_vlan &&
+ ++i < priv->vlan_filter_n);
+ return 0;
+}
+
+/**
+ * Disable a special flow in a hash RX queue.
+ *
+ * @param hash_rxq
+ * Pointer to hash RX queue structure.
+ * @param flow_type
+ * Special flow type.
+ */
+static void
+hash_rxq_special_flow_disable(struct hash_rxq *hash_rxq,
+ enum hash_rxq_flow_type flow_type)
+{
+ unsigned int i;
+
+ assert((unsigned int)flow_type < RTE_DIM(hash_rxq->special_flow));
+ for (i = 0; (i != RTE_DIM(hash_rxq->special_flow[flow_type])); ++i)
+ hash_rxq_special_flow_disable_vlan(hash_rxq, flow_type, i);
+}
+
+/**
+ * Enable a special flow in all hash RX queues.
+ *
+ * @param priv
+ * Private structure.
+ * @param flow_type
+ * Special flow type.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_special_flow_enable(struct priv *priv, enum hash_rxq_flow_type flow_type)
+{
+ unsigned int i;
+
+ if (!priv_allow_flow_type(priv, flow_type))
+ return 0;
+ for (i = 0; (i != priv->hash_rxqs_n); ++i) {
+ struct hash_rxq *hash_rxq = &(*priv->hash_rxqs)[i];
+ int ret;
+
+ ret = hash_rxq_special_flow_enable(hash_rxq, flow_type);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0) {
+ hash_rxq = &(*priv->hash_rxqs)[--i];
+ hash_rxq_special_flow_disable(hash_rxq, flow_type);
+ }
+ return ret;
+ }
+ return 0;
+}
+
+/**
+ * Disable a special flow in all hash RX queues.
+ *
+ * @param priv
+ * Private structure.
+ * @param flow_type
+ * Special flow type.
+ */
+void
+priv_special_flow_disable(struct priv *priv, enum hash_rxq_flow_type flow_type)
+{
+ unsigned int i;
+
+ for (i = 0; (i != priv->hash_rxqs_n); ++i) {
+ struct hash_rxq *hash_rxq = &(*priv->hash_rxqs)[i];
+
+ hash_rxq_special_flow_disable(hash_rxq, flow_type);
+ }
+}
+
+/**
+ * Enable all special flows in all hash RX queues.
+ *
+ * @param priv
+ * Private structure.
+ */
+int
+priv_special_flow_enable_all(struct priv *priv)
+{
+ enum hash_rxq_flow_type flow_type;
+
+ for (flow_type = 0; flow_type != HASH_RXQ_FLOW_TYPE_MAC; ++flow_type) {
+ int ret;
+
+ ret = priv_special_flow_enable(priv, flow_type);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (flow_type)
+ priv_special_flow_disable(priv, --flow_type);
+ return ret;
+ }
+ return 0;
+}
+
+/**
+ * Disable all special flows in all hash RX queues.
+ *
+ * @param priv
+ * Private structure.
+ */
+void
+priv_special_flow_disable_all(struct priv *priv)
+{
+ enum hash_rxq_flow_type flow_type;
+
+ for (flow_type = 0; flow_type != HASH_RXQ_FLOW_TYPE_MAC; ++flow_type)
+ priv_special_flow_disable(priv, flow_type);
+}
+
+/**
+ * DPDK callback to enable promiscuous mode.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_promiscuous_enable(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ priv->promisc_req = 1;
+ ret = priv_rehash_flows(priv);
+ if (ret)
+ ERROR("error while enabling promiscuous mode: %s",
+ strerror(ret));
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback to disable promiscuous mode.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_promiscuous_disable(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ priv->promisc_req = 0;
+ ret = priv_rehash_flows(priv);
+ if (ret)
+ ERROR("error while disabling promiscuous mode: %s",
+ strerror(ret));
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback to enable allmulti mode.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_allmulticast_enable(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ priv->allmulti_req = 1;
+ ret = priv_rehash_flows(priv);
+ if (ret)
+ ERROR("error while enabling allmulticast mode: %s",
+ strerror(ret));
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback to disable allmulti mode.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_allmulticast_disable(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ priv->allmulti_req = 0;
+ ret = priv_rehash_flows(priv);
+ if (ret)
+ ERROR("error while disabling allmulticast mode: %s",
+ strerror(ret));
+ priv_unlock(priv);
+}
diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
new file mode 100644
index 00000000..cbb017bb
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_rxq.c
@@ -0,0 +1,1555 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <assert.h>
+#include <errno.h>
+#include <string.h>
+#include <stdint.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_mbuf.h>
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_utils.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+/* Initialization data for hash RX queues. */
+const struct hash_rxq_init hash_rxq_init[] = {
+ [HASH_RXQ_TCPV4] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
+ IBV_EXP_RX_HASH_DST_IPV4 |
+ IBV_EXP_RX_HASH_SRC_PORT_TCP |
+ IBV_EXP_RX_HASH_DST_PORT_TCP),
+ .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_TCP,
+ .flow_priority = 0,
+ .flow_spec.tcp_udp = {
+ .type = IBV_EXP_FLOW_SPEC_TCP,
+ .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
+ },
+ [HASH_RXQ_UDPV4] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
+ IBV_EXP_RX_HASH_DST_IPV4 |
+ IBV_EXP_RX_HASH_SRC_PORT_UDP |
+ IBV_EXP_RX_HASH_DST_PORT_UDP),
+ .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_UDP,
+ .flow_priority = 0,
+ .flow_spec.tcp_udp = {
+ .type = IBV_EXP_FLOW_SPEC_UDP,
+ .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
+ },
+ [HASH_RXQ_IPV4] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV4 |
+ IBV_EXP_RX_HASH_DST_IPV4),
+ .dpdk_rss_hf = (ETH_RSS_IPV4 |
+ ETH_RSS_FRAG_IPV4),
+ .flow_priority = 1,
+ .flow_spec.ipv4 = {
+ .type = IBV_EXP_FLOW_SPEC_IPV4,
+ .size = sizeof(hash_rxq_init[0].flow_spec.ipv4),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_ETH],
+ },
+#ifdef HAVE_FLOW_SPEC_IPV6
+ [HASH_RXQ_TCPV6] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
+ IBV_EXP_RX_HASH_DST_IPV6 |
+ IBV_EXP_RX_HASH_SRC_PORT_TCP |
+ IBV_EXP_RX_HASH_DST_PORT_TCP),
+ .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_TCP,
+ .flow_priority = 0,
+ .flow_spec.tcp_udp = {
+ .type = IBV_EXP_FLOW_SPEC_TCP,
+ .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
+ },
+ [HASH_RXQ_UDPV6] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
+ IBV_EXP_RX_HASH_DST_IPV6 |
+ IBV_EXP_RX_HASH_SRC_PORT_UDP |
+ IBV_EXP_RX_HASH_DST_PORT_UDP),
+ .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_UDP,
+ .flow_priority = 0,
+ .flow_spec.tcp_udp = {
+ .type = IBV_EXP_FLOW_SPEC_UDP,
+ .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
+ },
+ [HASH_RXQ_IPV6] = {
+ .hash_fields = (IBV_EXP_RX_HASH_SRC_IPV6 |
+ IBV_EXP_RX_HASH_DST_IPV6),
+ .dpdk_rss_hf = (ETH_RSS_IPV6 |
+ ETH_RSS_FRAG_IPV6),
+ .flow_priority = 1,
+ .flow_spec.ipv6 = {
+ .type = IBV_EXP_FLOW_SPEC_IPV6,
+ .size = sizeof(hash_rxq_init[0].flow_spec.ipv6),
+ },
+ .underlayer = &hash_rxq_init[HASH_RXQ_ETH],
+ },
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ [HASH_RXQ_ETH] = {
+ .hash_fields = 0,
+ .dpdk_rss_hf = 0,
+ .flow_priority = 2,
+ .flow_spec.eth = {
+ .type = IBV_EXP_FLOW_SPEC_ETH,
+ .size = sizeof(hash_rxq_init[0].flow_spec.eth),
+ },
+ .underlayer = NULL,
+ },
+};
+
+/* Number of entries in hash_rxq_init[]. */
+const unsigned int hash_rxq_init_n = RTE_DIM(hash_rxq_init);
+
+/* Initialization data for hash RX queue indirection tables. */
+static const struct ind_table_init ind_table_init[] = {
+ {
+ .max_size = -1u, /* Superseded by HW limitations. */
+ .hash_types =
+ 1 << HASH_RXQ_TCPV4 |
+ 1 << HASH_RXQ_UDPV4 |
+ 1 << HASH_RXQ_IPV4 |
+#ifdef HAVE_FLOW_SPEC_IPV6
+ 1 << HASH_RXQ_TCPV6 |
+ 1 << HASH_RXQ_UDPV6 |
+ 1 << HASH_RXQ_IPV6 |
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ 0,
+#ifdef HAVE_FLOW_SPEC_IPV6
+ .hash_types_n = 6,
+#else /* HAVE_FLOW_SPEC_IPV6 */
+ .hash_types_n = 3,
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ },
+ {
+ .max_size = 1,
+ .hash_types = 1 << HASH_RXQ_ETH,
+ .hash_types_n = 1,
+ },
+};
+
+#define IND_TABLE_INIT_N RTE_DIM(ind_table_init)
+
+/* Default RSS hash key also used for ConnectX-3. */
+uint8_t rss_hash_default_key[] = {
+ 0x2c, 0xc6, 0x81, 0xd1,
+ 0x5b, 0xdb, 0xf4, 0xf7,
+ 0xfc, 0xa2, 0x83, 0x19,
+ 0xdb, 0x1a, 0x3e, 0x94,
+ 0x6b, 0x9e, 0x38, 0xd9,
+ 0x2c, 0x9c, 0x03, 0xd1,
+ 0xad, 0x99, 0x44, 0xa7,
+ 0xd9, 0x56, 0x3d, 0x59,
+ 0x06, 0x3c, 0x25, 0xf3,
+ 0xfc, 0x1f, 0xdc, 0x2a,
+};
+
+/* Length of the default RSS hash key. */
+const size_t rss_hash_default_key_len = sizeof(rss_hash_default_key);
+
+/**
+ * Populate flow steering rule for a given hash RX queue type using
+ * information from hash_rxq_init[]. Nothing is written to flow_attr when
+ * flow_attr_size is not large enough, but the required size is still returned.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[out] flow_attr
+ * Pointer to flow attribute structure to fill. Note that the allocated
+ * area must be larger and large enough to hold all flow specifications.
+ * @param flow_attr_size
+ * Entire size of flow_attr and trailing room for flow specifications.
+ * @param type
+ * Hash RX queue type to use for flow steering rule.
+ *
+ * @return
+ * Total size of the flow attribute buffer. No errors are defined.
+ */
+size_t
+priv_flow_attr(struct priv *priv, struct ibv_exp_flow_attr *flow_attr,
+ size_t flow_attr_size, enum hash_rxq_type type)
+{
+ size_t offset = sizeof(*flow_attr);
+ const struct hash_rxq_init *init = &hash_rxq_init[type];
+
+ assert(priv != NULL);
+ assert((size_t)type < RTE_DIM(hash_rxq_init));
+ do {
+ offset += init->flow_spec.hdr.size;
+ init = init->underlayer;
+ } while (init != NULL);
+ if (offset > flow_attr_size)
+ return offset;
+ flow_attr_size = offset;
+ init = &hash_rxq_init[type];
+ *flow_attr = (struct ibv_exp_flow_attr){
+ .type = IBV_EXP_FLOW_ATTR_NORMAL,
+#ifdef MLX5_FDIR_SUPPORT
+ /* Priorities < 3 are reserved for flow director. */
+ .priority = init->flow_priority + 3,
+#else /* MLX5_FDIR_SUPPORT */
+ .priority = init->flow_priority,
+#endif /* MLX5_FDIR_SUPPORT */
+ .num_of_specs = 0,
+ .port = priv->port,
+ .flags = 0,
+ };
+ do {
+ offset -= init->flow_spec.hdr.size;
+ memcpy((void *)((uintptr_t)flow_attr + offset),
+ &init->flow_spec,
+ init->flow_spec.hdr.size);
+ ++flow_attr->num_of_specs;
+ init = init->underlayer;
+ } while (init != NULL);
+ return flow_attr_size;
+}
+
+/**
+ * Convert hash type position in indirection table initializer to
+ * hash RX queue type.
+ *
+ * @param table
+ * Indirection table initializer.
+ * @param pos
+ * Hash type position.
+ *
+ * @return
+ * Hash RX queue type.
+ */
+static enum hash_rxq_type
+hash_rxq_type_from_pos(const struct ind_table_init *table, unsigned int pos)
+{
+ enum hash_rxq_type type = 0;
+
+ assert(pos < table->hash_types_n);
+ do {
+ if ((table->hash_types & (1 << type)) && (pos-- == 0))
+ break;
+ ++type;
+ } while (1);
+ return type;
+}
+
+/**
+ * Filter out disabled hash RX queue types from ind_table_init[].
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param[out] table
+ * Output table.
+ *
+ * @return
+ * Number of table entries.
+ */
+static unsigned int
+priv_make_ind_table_init(struct priv *priv,
+ struct ind_table_init (*table)[IND_TABLE_INIT_N])
+{
+ uint64_t rss_hf;
+ unsigned int i;
+ unsigned int j;
+ unsigned int table_n = 0;
+ /* Mandatory to receive frames not handled by normal hash RX queues. */
+ unsigned int hash_types_sup = 1 << HASH_RXQ_ETH;
+
+ rss_hf = priv->rss_hf;
+ /* Process other protocols only if more than one queue. */
+ if (priv->rxqs_n > 1)
+ for (i = 0; (i != hash_rxq_init_n); ++i)
+ if (rss_hf & hash_rxq_init[i].dpdk_rss_hf)
+ hash_types_sup |= (1 << i);
+
+ /* Filter out entries whose protocols are not in the set. */
+ for (i = 0, j = 0; (i != IND_TABLE_INIT_N); ++i) {
+ unsigned int nb;
+ unsigned int h;
+
+ /* j is increased only if the table has valid protocols. */
+ assert(j <= i);
+ (*table)[j] = ind_table_init[i];
+ (*table)[j].hash_types &= hash_types_sup;
+ for (h = 0, nb = 0; (h != hash_rxq_init_n); ++h)
+ if (((*table)[j].hash_types >> h) & 0x1)
+ ++nb;
+ (*table)[i].hash_types_n = nb;
+ if (nb) {
+ ++table_n;
+ ++j;
+ }
+ }
+ return table_n;
+}
+
+/**
+ * Initialize hash RX queues and indirection table.
+ *
+ * @param priv
+ * Pointer to private structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_create_hash_rxqs(struct priv *priv)
+{
+ struct ibv_exp_wq *wqs[priv->reta_idx_n];
+ struct ind_table_init ind_table_init[IND_TABLE_INIT_N];
+ unsigned int ind_tables_n =
+ priv_make_ind_table_init(priv, &ind_table_init);
+ unsigned int hash_rxqs_n = 0;
+ struct hash_rxq (*hash_rxqs)[] = NULL;
+ struct ibv_exp_rwq_ind_table *(*ind_tables)[] = NULL;
+ unsigned int i;
+ unsigned int j;
+ unsigned int k;
+ int err = 0;
+
+ assert(priv->ind_tables == NULL);
+ assert(priv->ind_tables_n == 0);
+ assert(priv->hash_rxqs == NULL);
+ assert(priv->hash_rxqs_n == 0);
+ assert(priv->pd != NULL);
+ assert(priv->ctx != NULL);
+ if (priv->rxqs_n == 0)
+ return EINVAL;
+ assert(priv->rxqs != NULL);
+ if (ind_tables_n == 0) {
+ ERROR("all hash RX queue types have been filtered out,"
+ " indirection table cannot be created");
+ return EINVAL;
+ }
+ if (priv->rxqs_n & (priv->rxqs_n - 1)) {
+ INFO("%u RX queues are configured, consider rounding this"
+ " number to the next power of two for better balancing",
+ priv->rxqs_n);
+ DEBUG("indirection table extended to assume %u WQs",
+ priv->reta_idx_n);
+ }
+ for (i = 0; (i != priv->reta_idx_n); ++i)
+ wqs[i] = (*priv->rxqs)[(*priv->reta_idx)[i]]->wq;
+ /* Get number of hash RX queues to configure. */
+ for (i = 0, hash_rxqs_n = 0; (i != ind_tables_n); ++i)
+ hash_rxqs_n += ind_table_init[i].hash_types_n;
+ DEBUG("allocating %u hash RX queues for %u WQs, %u indirection tables",
+ hash_rxqs_n, priv->rxqs_n, ind_tables_n);
+ /* Create indirection tables. */
+ ind_tables = rte_calloc(__func__, ind_tables_n,
+ sizeof((*ind_tables)[0]), 0);
+ if (ind_tables == NULL) {
+ err = ENOMEM;
+ ERROR("cannot allocate indirection tables container: %s",
+ strerror(err));
+ goto error;
+ }
+ for (i = 0; (i != ind_tables_n); ++i) {
+ struct ibv_exp_rwq_ind_table_init_attr ind_init_attr = {
+ .pd = priv->pd,
+ .log_ind_tbl_size = 0, /* Set below. */
+ .ind_tbl = wqs,
+ .comp_mask = 0,
+ };
+ unsigned int ind_tbl_size = ind_table_init[i].max_size;
+ struct ibv_exp_rwq_ind_table *ind_table;
+
+ if (priv->reta_idx_n < ind_tbl_size)
+ ind_tbl_size = priv->reta_idx_n;
+ ind_init_attr.log_ind_tbl_size = log2above(ind_tbl_size);
+ errno = 0;
+ ind_table = ibv_exp_create_rwq_ind_table(priv->ctx,
+ &ind_init_attr);
+ if (ind_table != NULL) {
+ (*ind_tables)[i] = ind_table;
+ continue;
+ }
+ /* Not clear whether errno is set. */
+ err = (errno ? errno : EINVAL);
+ ERROR("RX indirection table creation failed with error %d: %s",
+ err, strerror(err));
+ goto error;
+ }
+ /* Allocate array that holds hash RX queues and related data. */
+ hash_rxqs = rte_calloc(__func__, hash_rxqs_n,
+ sizeof((*hash_rxqs)[0]), 0);
+ if (hash_rxqs == NULL) {
+ err = ENOMEM;
+ ERROR("cannot allocate hash RX queues container: %s",
+ strerror(err));
+ goto error;
+ }
+ for (i = 0, j = 0, k = 0;
+ ((i != hash_rxqs_n) && (j != ind_tables_n));
+ ++i) {
+ struct hash_rxq *hash_rxq = &(*hash_rxqs)[i];
+ enum hash_rxq_type type =
+ hash_rxq_type_from_pos(&ind_table_init[j], k);
+ struct rte_eth_rss_conf *priv_rss_conf =
+ (*priv->rss_conf)[type];
+ struct ibv_exp_rx_hash_conf hash_conf = {
+ .rx_hash_function = IBV_EXP_RX_HASH_FUNC_TOEPLITZ,
+ .rx_hash_key_len = (priv_rss_conf ?
+ priv_rss_conf->rss_key_len :
+ rss_hash_default_key_len),
+ .rx_hash_key = (priv_rss_conf ?
+ priv_rss_conf->rss_key :
+ rss_hash_default_key),
+ .rx_hash_fields_mask = hash_rxq_init[type].hash_fields,
+ .rwq_ind_tbl = (*ind_tables)[j],
+ };
+ struct ibv_exp_qp_init_attr qp_init_attr = {
+ .max_inl_recv = 0, /* Currently not supported. */
+ .qp_type = IBV_QPT_RAW_PACKET,
+ .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD |
+ IBV_EXP_QP_INIT_ATTR_RX_HASH),
+ .pd = priv->pd,
+ .rx_hash_conf = &hash_conf,
+ .port_num = priv->port,
+ };
+
+ DEBUG("using indirection table %u for hash RX queue %u type %d",
+ j, i, type);
+ *hash_rxq = (struct hash_rxq){
+ .priv = priv,
+ .qp = ibv_exp_create_qp(priv->ctx, &qp_init_attr),
+ .type = type,
+ };
+ if (hash_rxq->qp == NULL) {
+ err = (errno ? errno : EINVAL);
+ ERROR("Hash RX QP creation failure: %s",
+ strerror(err));
+ goto error;
+ }
+ if (++k < ind_table_init[j].hash_types_n)
+ continue;
+ /* Switch to the next indirection table and reset hash RX
+ * queue type array index. */
+ ++j;
+ k = 0;
+ }
+ priv->ind_tables = ind_tables;
+ priv->ind_tables_n = ind_tables_n;
+ priv->hash_rxqs = hash_rxqs;
+ priv->hash_rxqs_n = hash_rxqs_n;
+ assert(err == 0);
+ return 0;
+error:
+ if (hash_rxqs != NULL) {
+ for (i = 0; (i != hash_rxqs_n); ++i) {
+ struct ibv_qp *qp = (*hash_rxqs)[i].qp;
+
+ if (qp == NULL)
+ continue;
+ claim_zero(ibv_destroy_qp(qp));
+ }
+ rte_free(hash_rxqs);
+ }
+ if (ind_tables != NULL) {
+ for (j = 0; (j != ind_tables_n); ++j) {
+ struct ibv_exp_rwq_ind_table *ind_table =
+ (*ind_tables)[j];
+
+ if (ind_table == NULL)
+ continue;
+ claim_zero(ibv_exp_destroy_rwq_ind_table(ind_table));
+ }
+ rte_free(ind_tables);
+ }
+ return err;
+}
+
+/**
+ * Clean up hash RX queues and indirection table.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+void
+priv_destroy_hash_rxqs(struct priv *priv)
+{
+ unsigned int i;
+
+ DEBUG("destroying %u hash RX queues", priv->hash_rxqs_n);
+ if (priv->hash_rxqs_n == 0) {
+ assert(priv->hash_rxqs == NULL);
+ assert(priv->ind_tables == NULL);
+ return;
+ }
+ for (i = 0; (i != priv->hash_rxqs_n); ++i) {
+ struct hash_rxq *hash_rxq = &(*priv->hash_rxqs)[i];
+ unsigned int j, k;
+
+ assert(hash_rxq->priv == priv);
+ assert(hash_rxq->qp != NULL);
+ /* Also check that there are no remaining flows. */
+ for (j = 0; (j != RTE_DIM(hash_rxq->special_flow)); ++j)
+ for (k = 0;
+ (k != RTE_DIM(hash_rxq->special_flow[j]));
+ ++k)
+ assert(hash_rxq->special_flow[j][k] == NULL);
+ for (j = 0; (j != RTE_DIM(hash_rxq->mac_flow)); ++j)
+ for (k = 0; (k != RTE_DIM(hash_rxq->mac_flow[j])); ++k)
+ assert(hash_rxq->mac_flow[j][k] == NULL);
+ claim_zero(ibv_destroy_qp(hash_rxq->qp));
+ }
+ priv->hash_rxqs_n = 0;
+ rte_free(priv->hash_rxqs);
+ priv->hash_rxqs = NULL;
+ for (i = 0; (i != priv->ind_tables_n); ++i) {
+ struct ibv_exp_rwq_ind_table *ind_table =
+ (*priv->ind_tables)[i];
+
+ assert(ind_table != NULL);
+ claim_zero(ibv_exp_destroy_rwq_ind_table(ind_table));
+ }
+ priv->ind_tables_n = 0;
+ rte_free(priv->ind_tables);
+ priv->ind_tables = NULL;
+}
+
+/**
+ * Check whether a given flow type is allowed.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param type
+ * Flow type to check.
+ *
+ * @return
+ * Nonzero if the given flow type is allowed.
+ */
+int
+priv_allow_flow_type(struct priv *priv, enum hash_rxq_flow_type type)
+{
+ /* Only FLOW_TYPE_PROMISC is allowed when promiscuous mode
+ * has been requested. */
+ if (priv->promisc_req)
+ return type == HASH_RXQ_FLOW_TYPE_PROMISC;
+ switch (type) {
+ case HASH_RXQ_FLOW_TYPE_PROMISC:
+ return !!priv->promisc_req;
+ case HASH_RXQ_FLOW_TYPE_ALLMULTI:
+ return !!priv->allmulti_req;
+ case HASH_RXQ_FLOW_TYPE_BROADCAST:
+#ifdef HAVE_FLOW_SPEC_IPV6
+ case HASH_RXQ_FLOW_TYPE_IPV6MULTI:
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ /* If allmulti is enabled, broadcast and ipv6multi
+ * are unnecessary. */
+ return !priv->allmulti_req;
+ case HASH_RXQ_FLOW_TYPE_MAC:
+ return 1;
+ default:
+ /* Unsupported flow type is not allowed. */
+ return 0;
+ }
+ return 0;
+}
+
+/**
+ * Automatically enable/disable flows according to configuration.
+ *
+ * @param priv
+ * Private structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+priv_rehash_flows(struct priv *priv)
+{
+ unsigned int i;
+
+ for (i = 0; (i != RTE_DIM((*priv->hash_rxqs)[0].special_flow)); ++i)
+ if (!priv_allow_flow_type(priv, i)) {
+ priv_special_flow_disable(priv, i);
+ } else {
+ int ret = priv_special_flow_enable(priv, i);
+
+ if (ret)
+ return ret;
+ }
+ if (priv_allow_flow_type(priv, HASH_RXQ_FLOW_TYPE_MAC))
+ return priv_mac_addrs_enable(priv);
+ priv_mac_addrs_disable(priv);
+ return 0;
+}
+
+/**
+ * Allocate RX queue elements with scattered packets support.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param elts_n
+ * Number of elements to allocate.
+ * @param[in] pool
+ * If not NULL, fetch buffers from this array instead of allocating them
+ * with rte_pktmbuf_alloc().
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+rxq_alloc_elts_sp(struct rxq *rxq, unsigned int elts_n,
+ struct rte_mbuf **pool)
+{
+ unsigned int i;
+ struct rxq_elt_sp (*elts)[elts_n] =
+ rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0,
+ rxq->socket);
+ int ret = 0;
+
+ if (elts == NULL) {
+ ERROR("%p: can't allocate packets array", (void *)rxq);
+ ret = ENOMEM;
+ goto error;
+ }
+ /* For each WR (packet). */
+ for (i = 0; (i != elts_n); ++i) {
+ unsigned int j;
+ struct rxq_elt_sp *elt = &(*elts)[i];
+ struct ibv_sge (*sges)[RTE_DIM(elt->sges)] = &elt->sges;
+
+ /* These two arrays must have the same size. */
+ assert(RTE_DIM(elt->sges) == RTE_DIM(elt->bufs));
+ /* For each SGE (segment). */
+ for (j = 0; (j != RTE_DIM(elt->bufs)); ++j) {
+ struct ibv_sge *sge = &(*sges)[j];
+ struct rte_mbuf *buf;
+
+ if (pool != NULL) {
+ buf = *(pool++);
+ assert(buf != NULL);
+ rte_pktmbuf_reset(buf);
+ } else
+ buf = rte_pktmbuf_alloc(rxq->mp);
+ if (buf == NULL) {
+ assert(pool == NULL);
+ ERROR("%p: empty mbuf pool", (void *)rxq);
+ ret = ENOMEM;
+ goto error;
+ }
+ elt->bufs[j] = buf;
+ /* Headroom is reserved by rte_pktmbuf_alloc(). */
+ assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
+ /* Buffer is supposed to be empty. */
+ assert(rte_pktmbuf_data_len(buf) == 0);
+ assert(rte_pktmbuf_pkt_len(buf) == 0);
+ /* sge->addr must be able to store a pointer. */
+ assert(sizeof(sge->addr) >= sizeof(uintptr_t));
+ if (j == 0) {
+ /* The first SGE keeps its headroom. */
+ sge->addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ sge->length = (buf->buf_len -
+ RTE_PKTMBUF_HEADROOM);
+ } else {
+ /* Subsequent SGEs lose theirs. */
+ assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
+ SET_DATA_OFF(buf, 0);
+ sge->addr = (uintptr_t)buf->buf_addr;
+ sge->length = buf->buf_len;
+ }
+ sge->lkey = rxq->mr->lkey;
+ /* Redundant check for tailroom. */
+ assert(sge->length == rte_pktmbuf_tailroom(buf));
+ }
+ }
+ DEBUG("%p: allocated and configured %u WRs (%zu segments)",
+ (void *)rxq, elts_n, (elts_n * RTE_DIM((*elts)[0].sges)));
+ rxq->elts_n = elts_n;
+ rxq->elts_head = 0;
+ rxq->elts.sp = elts;
+ assert(ret == 0);
+ return 0;
+error:
+ if (elts != NULL) {
+ assert(pool == NULL);
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ unsigned int j;
+ struct rxq_elt_sp *elt = &(*elts)[i];
+
+ for (j = 0; (j != RTE_DIM(elt->bufs)); ++j) {
+ struct rte_mbuf *buf = elt->bufs[j];
+
+ if (buf != NULL)
+ rte_pktmbuf_free_seg(buf);
+ }
+ }
+ rte_free(elts);
+ }
+ DEBUG("%p: failed, freed everything", (void *)rxq);
+ assert(ret > 0);
+ return ret;
+}
+
+/**
+ * Free RX queue elements with scattered packets support.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ */
+static void
+rxq_free_elts_sp(struct rxq *rxq)
+{
+ unsigned int i;
+ unsigned int elts_n = rxq->elts_n;
+ struct rxq_elt_sp (*elts)[elts_n] = rxq->elts.sp;
+
+ DEBUG("%p: freeing WRs", (void *)rxq);
+ rxq->elts_n = 0;
+ rxq->elts.sp = NULL;
+ if (elts == NULL)
+ return;
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ unsigned int j;
+ struct rxq_elt_sp *elt = &(*elts)[i];
+
+ for (j = 0; (j != RTE_DIM(elt->bufs)); ++j) {
+ struct rte_mbuf *buf = elt->bufs[j];
+
+ if (buf != NULL)
+ rte_pktmbuf_free_seg(buf);
+ }
+ }
+ rte_free(elts);
+}
+
+/**
+ * Allocate RX queue elements.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param elts_n
+ * Number of elements to allocate.
+ * @param[in] pool
+ * If not NULL, fetch buffers from this array instead of allocating them
+ * with rte_pktmbuf_alloc().
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n, struct rte_mbuf **pool)
+{
+ unsigned int i;
+ struct rxq_elt (*elts)[elts_n] =
+ rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0,
+ rxq->socket);
+ int ret = 0;
+
+ if (elts == NULL) {
+ ERROR("%p: can't allocate packets array", (void *)rxq);
+ ret = ENOMEM;
+ goto error;
+ }
+ /* For each WR (packet). */
+ for (i = 0; (i != elts_n); ++i) {
+ struct rxq_elt *elt = &(*elts)[i];
+ struct ibv_sge *sge = &(*elts)[i].sge;
+ struct rte_mbuf *buf;
+
+ if (pool != NULL) {
+ buf = *(pool++);
+ assert(buf != NULL);
+ rte_pktmbuf_reset(buf);
+ } else
+ buf = rte_pktmbuf_alloc(rxq->mp);
+ if (buf == NULL) {
+ assert(pool == NULL);
+ ERROR("%p: empty mbuf pool", (void *)rxq);
+ ret = ENOMEM;
+ goto error;
+ }
+ elt->buf = buf;
+ /* Headroom is reserved by rte_pktmbuf_alloc(). */
+ assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
+ /* Buffer is supposed to be empty. */
+ assert(rte_pktmbuf_data_len(buf) == 0);
+ assert(rte_pktmbuf_pkt_len(buf) == 0);
+ /* sge->addr must be able to store a pointer. */
+ assert(sizeof(sge->addr) >= sizeof(uintptr_t));
+ /* SGE keeps its headroom. */
+ sge->addr = (uintptr_t)
+ ((uint8_t *)buf->buf_addr + RTE_PKTMBUF_HEADROOM);
+ sge->length = (buf->buf_len - RTE_PKTMBUF_HEADROOM);
+ sge->lkey = rxq->mr->lkey;
+ /* Redundant check for tailroom. */
+ assert(sge->length == rte_pktmbuf_tailroom(buf));
+ }
+ DEBUG("%p: allocated and configured %u single-segment WRs",
+ (void *)rxq, elts_n);
+ rxq->elts_n = elts_n;
+ rxq->elts_head = 0;
+ rxq->elts.no_sp = elts;
+ assert(ret == 0);
+ return 0;
+error:
+ if (elts != NULL) {
+ assert(pool == NULL);
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ struct rxq_elt *elt = &(*elts)[i];
+ struct rte_mbuf *buf = elt->buf;
+
+ if (buf != NULL)
+ rte_pktmbuf_free_seg(buf);
+ }
+ rte_free(elts);
+ }
+ DEBUG("%p: failed, freed everything", (void *)rxq);
+ assert(ret > 0);
+ return ret;
+}
+
+/**
+ * Free RX queue elements.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ */
+static void
+rxq_free_elts(struct rxq *rxq)
+{
+ unsigned int i;
+ unsigned int elts_n = rxq->elts_n;
+ struct rxq_elt (*elts)[elts_n] = rxq->elts.no_sp;
+
+ DEBUG("%p: freeing WRs", (void *)rxq);
+ rxq->elts_n = 0;
+ rxq->elts.no_sp = NULL;
+ if (elts == NULL)
+ return;
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ struct rxq_elt *elt = &(*elts)[i];
+ struct rte_mbuf *buf = elt->buf;
+
+ if (buf != NULL)
+ rte_pktmbuf_free_seg(buf);
+ }
+ rte_free(elts);
+}
+
+/**
+ * Clean up a RX queue.
+ *
+ * Destroy objects, free allocated memory and reset the structure for reuse.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ */
+void
+rxq_cleanup(struct rxq *rxq)
+{
+ struct ibv_exp_release_intf_params params;
+
+ DEBUG("cleaning up %p", (void *)rxq);
+ if (rxq->sp)
+ rxq_free_elts_sp(rxq);
+ else
+ rxq_free_elts(rxq);
+ rxq->poll = NULL;
+ rxq->recv = NULL;
+ if (rxq->if_wq != NULL) {
+ assert(rxq->priv != NULL);
+ assert(rxq->priv->ctx != NULL);
+ assert(rxq->wq != NULL);
+ params = (struct ibv_exp_release_intf_params){
+ .comp_mask = 0,
+ };
+ claim_zero(ibv_exp_release_intf(rxq->priv->ctx,
+ rxq->if_wq,
+ &params));
+ }
+ if (rxq->if_cq != NULL) {
+ assert(rxq->priv != NULL);
+ assert(rxq->priv->ctx != NULL);
+ assert(rxq->cq != NULL);
+ params = (struct ibv_exp_release_intf_params){
+ .comp_mask = 0,
+ };
+ claim_zero(ibv_exp_release_intf(rxq->priv->ctx,
+ rxq->if_cq,
+ &params));
+ }
+ if (rxq->wq != NULL)
+ claim_zero(ibv_exp_destroy_wq(rxq->wq));
+ if (rxq->cq != NULL)
+ claim_zero(ibv_destroy_cq(rxq->cq));
+ if (rxq->rd != NULL) {
+ struct ibv_exp_destroy_res_domain_attr attr = {
+ .comp_mask = 0,
+ };
+
+ assert(rxq->priv != NULL);
+ assert(rxq->priv->ctx != NULL);
+ claim_zero(ibv_exp_destroy_res_domain(rxq->priv->ctx,
+ rxq->rd,
+ &attr));
+ }
+ if (rxq->mr != NULL)
+ claim_zero(ibv_dereg_mr(rxq->mr));
+ memset(rxq, 0, sizeof(*rxq));
+}
+
+/**
+ * Reconfigure a RX queue with new parameters.
+ *
+ * rxq_rehash() does not allocate mbufs, which, if not done from the right
+ * thread (such as a control thread), may corrupt the pool.
+ * In case of failure, the queue is left untouched.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rxq
+ * RX queue pointer.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq)
+{
+ struct priv *priv = rxq->priv;
+ struct rxq tmpl = *rxq;
+ unsigned int mbuf_n;
+ unsigned int desc_n;
+ struct rte_mbuf **pool;
+ unsigned int i, k;
+ struct ibv_exp_wq_attr mod;
+ int err;
+
+ DEBUG("%p: rehashing queue %p", (void *)dev, (void *)rxq);
+ /* Number of descriptors and mbufs currently allocated. */
+ desc_n = (tmpl.elts_n * (tmpl.sp ? MLX5_PMD_SGE_WR_N : 1));
+ mbuf_n = desc_n;
+ /* Toggle RX checksum offload if hardware supports it. */
+ if (priv->hw_csum) {
+ tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum;
+ rxq->csum = tmpl.csum;
+ }
+ if (priv->hw_csum_l2tun) {
+ tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum;
+ rxq->csum_l2tun = tmpl.csum_l2tun;
+ }
+ /* Enable scattered packets support for this queue if necessary. */
+ if ((dev->data->dev_conf.rxmode.jumbo_frame) &&
+ (dev->data->dev_conf.rxmode.max_rx_pkt_len >
+ (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) {
+ tmpl.sp = 1;
+ desc_n /= MLX5_PMD_SGE_WR_N;
+ } else
+ tmpl.sp = 0;
+ DEBUG("%p: %s scattered packets support (%u WRs)",
+ (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc_n);
+ /* If scatter mode is the same as before, nothing to do. */
+ if (tmpl.sp == rxq->sp) {
+ DEBUG("%p: nothing to do", (void *)dev);
+ return 0;
+ }
+ /* From now on, any failure will render the queue unusable.
+ * Reinitialize WQ. */
+ mod = (struct ibv_exp_wq_attr){
+ .attr_mask = IBV_EXP_WQ_ATTR_STATE,
+ .wq_state = IBV_EXP_WQS_RESET,
+ };
+ err = ibv_exp_modify_wq(tmpl.wq, &mod);
+ if (err) {
+ ERROR("%p: cannot reset WQ: %s", (void *)dev, strerror(err));
+ assert(err > 0);
+ return err;
+ }
+ /* Allocate pool. */
+ pool = rte_malloc(__func__, (mbuf_n * sizeof(*pool)), 0);
+ if (pool == NULL) {
+ ERROR("%p: cannot allocate memory", (void *)dev);
+ return ENOBUFS;
+ }
+ /* Snatch mbufs from original queue. */
+ k = 0;
+ if (rxq->sp) {
+ struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ struct rxq_elt_sp *elt = &(*elts)[i];
+ unsigned int j;
+
+ for (j = 0; (j != RTE_DIM(elt->bufs)); ++j) {
+ assert(elt->bufs[j] != NULL);
+ pool[k++] = elt->bufs[j];
+ }
+ }
+ } else {
+ struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ struct rxq_elt *elt = &(*elts)[i];
+ struct rte_mbuf *buf = elt->buf;
+
+ pool[k++] = buf;
+ }
+ }
+ assert(k == mbuf_n);
+ tmpl.elts_n = 0;
+ tmpl.elts.sp = NULL;
+ assert((void *)&tmpl.elts.sp == (void *)&tmpl.elts.no_sp);
+ err = ((tmpl.sp) ?
+ rxq_alloc_elts_sp(&tmpl, desc_n, pool) :
+ rxq_alloc_elts(&tmpl, desc_n, pool));
+ if (err) {
+ ERROR("%p: cannot reallocate WRs, aborting", (void *)dev);
+ rte_free(pool);
+ assert(err > 0);
+ return err;
+ }
+ assert(tmpl.elts_n == desc_n);
+ assert(tmpl.elts.sp != NULL);
+ rte_free(pool);
+ /* Clean up original data. */
+ rxq->elts_n = 0;
+ rte_free(rxq->elts.sp);
+ rxq->elts.sp = NULL;
+ /* Change queue state to ready. */
+ mod = (struct ibv_exp_wq_attr){
+ .attr_mask = IBV_EXP_WQ_ATTR_STATE,
+ .wq_state = IBV_EXP_WQS_RDY,
+ };
+ err = ibv_exp_modify_wq(tmpl.wq, &mod);
+ if (err) {
+ ERROR("%p: WQ state to IBV_EXP_WQS_RDY failed: %s",
+ (void *)dev, strerror(err));
+ goto error;
+ }
+ /* Post SGEs. */
+ assert(tmpl.if_wq != NULL);
+ if (tmpl.sp) {
+ struct rxq_elt_sp (*elts)[tmpl.elts_n] = tmpl.elts.sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ err = tmpl.if_wq->recv_sg_list
+ (tmpl.wq,
+ (*elts)[i].sges,
+ RTE_DIM((*elts)[i].sges));
+ if (err)
+ break;
+ }
+ } else {
+ struct rxq_elt (*elts)[tmpl.elts_n] = tmpl.elts.no_sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ err = tmpl.if_wq->recv_burst(
+ tmpl.wq,
+ &(*elts)[i].sge,
+ 1);
+ if (err)
+ break;
+ }
+ }
+ if (err) {
+ ERROR("%p: failed to post SGEs with error %d",
+ (void *)dev, err);
+ /* Set err because it does not contain a valid errno value. */
+ err = EIO;
+ goto error;
+ }
+ if (tmpl.sp)
+ tmpl.recv = tmpl.if_wq->recv_sg_list;
+ else
+ tmpl.recv = tmpl.if_wq->recv_burst;
+error:
+ *rxq = tmpl;
+ assert(err >= 0);
+ return err;
+}
+
+/**
+ * Configure a RX queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
+ * @param[in] conf
+ * Thresholds parameters.
+ * @param mp
+ * Memory pool for buffer allocations.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc,
+ unsigned int socket, const struct rte_eth_rxconf *conf,
+ struct rte_mempool *mp)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct rxq tmpl = {
+ .priv = priv,
+ .mp = mp,
+ .socket = socket
+ };
+ struct ibv_exp_wq_attr mod;
+ union {
+ struct ibv_exp_query_intf_params params;
+ struct ibv_exp_cq_init_attr cq;
+ struct ibv_exp_res_domain_init_attr rd;
+ struct ibv_exp_wq_init_attr wq;
+ } attr;
+ enum ibv_exp_query_intf_status status;
+ struct rte_mbuf *buf;
+ int ret = 0;
+ unsigned int i;
+ unsigned int cq_size = desc;
+
+ (void)conf; /* Thresholds configuration (ignored). */
+ if ((desc == 0) || (desc % MLX5_PMD_SGE_WR_N)) {
+ ERROR("%p: invalid number of RX descriptors (must be a"
+ " multiple of %d)", (void *)dev, MLX5_PMD_SGE_WR_N);
+ return EINVAL;
+ }
+ /* Get mbuf length. */
+ buf = rte_pktmbuf_alloc(mp);
+ if (buf == NULL) {
+ ERROR("%p: unable to allocate mbuf", (void *)dev);
+ return ENOMEM;
+ }
+ tmpl.mb_len = buf->buf_len;
+ assert((rte_pktmbuf_headroom(buf) +
+ rte_pktmbuf_tailroom(buf)) == tmpl.mb_len);
+ assert(rte_pktmbuf_headroom(buf) == RTE_PKTMBUF_HEADROOM);
+ rte_pktmbuf_free(buf);
+ /* Toggle RX checksum offload if hardware supports it. */
+ if (priv->hw_csum)
+ tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum;
+ if (priv->hw_csum_l2tun)
+ tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum;
+ /* Enable scattered packets support for this queue if necessary. */
+ if ((dev->data->dev_conf.rxmode.jumbo_frame) &&
+ (dev->data->dev_conf.rxmode.max_rx_pkt_len >
+ (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) {
+ tmpl.sp = 1;
+ desc /= MLX5_PMD_SGE_WR_N;
+ }
+ DEBUG("%p: %s scattered packets support (%u WRs)",
+ (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc);
+ /* Use the entire RX mempool as the memory region. */
+ tmpl.mr = mlx5_mp2mr(priv->pd, mp);
+ if (tmpl.mr == NULL) {
+ ret = EINVAL;
+ ERROR("%p: MR creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.rd = (struct ibv_exp_res_domain_init_attr){
+ .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL |
+ IBV_EXP_RES_DOMAIN_MSG_MODEL),
+ .thread_model = IBV_EXP_THREAD_SINGLE,
+ .msg_model = IBV_EXP_MSG_HIGH_BW,
+ };
+ tmpl.rd = ibv_exp_create_res_domain(priv->ctx, &attr.rd);
+ if (tmpl.rd == NULL) {
+ ret = ENOMEM;
+ ERROR("%p: RD creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.cq = (struct ibv_exp_cq_init_attr){
+ .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN,
+ .res_domain = tmpl.rd,
+ };
+ tmpl.cq = ibv_exp_create_cq(priv->ctx, cq_size, NULL, NULL, 0,
+ &attr.cq);
+ if (tmpl.cq == NULL) {
+ ret = ENOMEM;
+ ERROR("%p: CQ creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ DEBUG("priv->device_attr.max_qp_wr is %d",
+ priv->device_attr.max_qp_wr);
+ DEBUG("priv->device_attr.max_sge is %d",
+ priv->device_attr.max_sge);
+ /* Configure VLAN stripping. */
+ tmpl.vlan_strip = dev->data->dev_conf.rxmode.hw_vlan_strip;
+ attr.wq = (struct ibv_exp_wq_init_attr){
+ .wq_context = NULL, /* Could be useful in the future. */
+ .wq_type = IBV_EXP_WQT_RQ,
+ /* Max number of outstanding WRs. */
+ .max_recv_wr = ((priv->device_attr.max_qp_wr < (int)cq_size) ?
+ priv->device_attr.max_qp_wr :
+ (int)cq_size),
+ /* Max number of scatter/gather elements in a WR. */
+ .max_recv_sge = ((priv->device_attr.max_sge <
+ MLX5_PMD_SGE_WR_N) ?
+ priv->device_attr.max_sge :
+ MLX5_PMD_SGE_WR_N),
+ .pd = priv->pd,
+ .cq = tmpl.cq,
+ .comp_mask =
+ IBV_EXP_CREATE_WQ_RES_DOMAIN |
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ IBV_EXP_CREATE_WQ_VLAN_OFFLOADS |
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ 0,
+ .res_domain = tmpl.rd,
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ .vlan_offloads = (tmpl.vlan_strip ?
+ IBV_EXP_RECEIVE_WQ_CVLAN_STRIP :
+ 0),
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ };
+
+#ifdef HAVE_VERBS_FCS
+ /* By default, FCS (CRC) is stripped by hardware. */
+ if (dev->data->dev_conf.rxmode.hw_strip_crc) {
+ tmpl.crc_present = 0;
+ } else if (priv->hw_fcs_strip) {
+ /* Ask HW/Verbs to leave CRC in place when supported. */
+ attr.wq.flags |= IBV_EXP_CREATE_WQ_FLAG_SCATTER_FCS;
+ attr.wq.comp_mask |= IBV_EXP_CREATE_WQ_FLAGS;
+ tmpl.crc_present = 1;
+ } else {
+ WARN("%p: CRC stripping has been disabled but will still"
+ " be performed by hardware, make sure MLNX_OFED and"
+ " firmware are up to date",
+ (void *)dev);
+ tmpl.crc_present = 0;
+ }
+ DEBUG("%p: CRC stripping is %s, %u bytes will be subtracted from"
+ " incoming frames to hide it",
+ (void *)dev,
+ tmpl.crc_present ? "disabled" : "enabled",
+ tmpl.crc_present << 2);
+#endif /* HAVE_VERBS_FCS */
+
+#ifdef HAVE_VERBS_RX_END_PADDING
+ if (!mlx5_getenv_int("MLX5_PMD_ENABLE_PADDING"))
+ ; /* Nothing else to do. */
+ else if (priv->hw_padding) {
+ INFO("%p: enabling packet padding on queue %p",
+ (void *)dev, (void *)rxq);
+ attr.wq.flags |= IBV_EXP_CREATE_WQ_FLAG_RX_END_PADDING;
+ attr.wq.comp_mask |= IBV_EXP_CREATE_WQ_FLAGS;
+ } else
+ WARN("%p: packet padding has been requested but is not"
+ " supported, make sure MLNX_OFED and firmware are"
+ " up to date",
+ (void *)dev);
+#endif /* HAVE_VERBS_RX_END_PADDING */
+
+ tmpl.wq = ibv_exp_create_wq(priv->ctx, &attr.wq);
+ if (tmpl.wq == NULL) {
+ ret = (errno ? errno : EINVAL);
+ ERROR("%p: WQ creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ if (tmpl.sp)
+ ret = rxq_alloc_elts_sp(&tmpl, desc, NULL);
+ else
+ ret = rxq_alloc_elts(&tmpl, desc, NULL);
+ if (ret) {
+ ERROR("%p: RXQ allocation failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ /* Save port ID. */
+ tmpl.port_id = dev->data->port_id;
+ DEBUG("%p: RTE port ID: %u", (void *)rxq, tmpl.port_id);
+ attr.params = (struct ibv_exp_query_intf_params){
+ .intf_scope = IBV_EXP_INTF_GLOBAL,
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ .intf_version = 1,
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ .intf = IBV_EXP_INTF_CQ,
+ .obj = tmpl.cq,
+ };
+ tmpl.if_cq = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
+ if (tmpl.if_cq == NULL) {
+ ERROR("%p: CQ interface family query failed with status %d",
+ (void *)dev, status);
+ goto error;
+ }
+ attr.params = (struct ibv_exp_query_intf_params){
+ .intf_scope = IBV_EXP_INTF_GLOBAL,
+ .intf = IBV_EXP_INTF_WQ,
+ .obj = tmpl.wq,
+ };
+ tmpl.if_wq = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
+ if (tmpl.if_wq == NULL) {
+ ERROR("%p: WQ interface family query failed with status %d",
+ (void *)dev, status);
+ goto error;
+ }
+ /* Change queue state to ready. */
+ mod = (struct ibv_exp_wq_attr){
+ .attr_mask = IBV_EXP_WQ_ATTR_STATE,
+ .wq_state = IBV_EXP_WQS_RDY,
+ };
+ ret = ibv_exp_modify_wq(tmpl.wq, &mod);
+ if (ret) {
+ ERROR("%p: WQ state to IBV_EXP_WQS_RDY failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ /* Post SGEs. */
+ if (tmpl.sp) {
+ struct rxq_elt_sp (*elts)[tmpl.elts_n] = tmpl.elts.sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ ret = tmpl.if_wq->recv_sg_list
+ (tmpl.wq,
+ (*elts)[i].sges,
+ RTE_DIM((*elts)[i].sges));
+ if (ret)
+ break;
+ }
+ } else {
+ struct rxq_elt (*elts)[tmpl.elts_n] = tmpl.elts.no_sp;
+
+ for (i = 0; (i != RTE_DIM(*elts)); ++i) {
+ ret = tmpl.if_wq->recv_burst(
+ tmpl.wq,
+ &(*elts)[i].sge,
+ 1);
+ if (ret)
+ break;
+ }
+ }
+ if (ret) {
+ ERROR("%p: failed to post SGEs with error %d",
+ (void *)dev, ret);
+ /* Set ret because it does not contain a valid errno value. */
+ ret = EIO;
+ goto error;
+ }
+ /* Clean up rxq in case we're reinitializing it. */
+ DEBUG("%p: cleaning-up old rxq just in case", (void *)rxq);
+ rxq_cleanup(rxq);
+ *rxq = tmpl;
+ DEBUG("%p: rxq updated with %p", (void *)rxq, (void *)&tmpl);
+ assert(ret == 0);
+ /* Assign function in queue. */
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ rxq->poll = rxq->if_cq->poll_length_flags_cvlan;
+#else /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ rxq->poll = rxq->if_cq->poll_length_flags;
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ if (rxq->sp)
+ rxq->recv = rxq->if_wq->recv_sg_list;
+ else
+ rxq->recv = rxq->if_wq->recv_burst;
+ return 0;
+error:
+ rxq_cleanup(&tmpl);
+ assert(ret > 0);
+ return ret;
+}
+
+/**
+ * DPDK callback to configure a RX queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
+ * @param[in] conf
+ * Thresholds parameters.
+ * @param mp
+ * Memory pool for buffer allocations.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ unsigned int socket, const struct rte_eth_rxconf *conf,
+ struct rte_mempool *mp)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct rxq *rxq = (*priv->rxqs)[idx];
+ int ret;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ priv_lock(priv);
+ DEBUG("%p: configuring queue %u for %u descriptors",
+ (void *)dev, idx, desc);
+ if (idx >= priv->rxqs_n) {
+ ERROR("%p: queue index out of range (%u >= %u)",
+ (void *)dev, idx, priv->rxqs_n);
+ priv_unlock(priv);
+ return -EOVERFLOW;
+ }
+ if (rxq != NULL) {
+ DEBUG("%p: reusing already allocated queue index %u (%p)",
+ (void *)dev, idx, (void *)rxq);
+ if (priv->started) {
+ priv_unlock(priv);
+ return -EEXIST;
+ }
+ (*priv->rxqs)[idx] = NULL;
+ rxq_cleanup(rxq);
+ } else {
+ rxq = rte_calloc_socket("RXQ", 1, sizeof(*rxq), 0, socket);
+ if (rxq == NULL) {
+ ERROR("%p: unable to allocate queue index %u",
+ (void *)dev, idx);
+ priv_unlock(priv);
+ return -ENOMEM;
+ }
+ }
+ ret = rxq_setup(dev, rxq, desc, socket, conf, mp);
+ if (ret)
+ rte_free(rxq);
+ else {
+ rxq->stats.idx = idx;
+ DEBUG("%p: adding RX queue %p to list",
+ (void *)dev, (void *)rxq);
+ (*priv->rxqs)[idx] = rxq;
+ /* Update receive callback. */
+ if (rxq->sp)
+ dev->rx_pkt_burst = mlx5_rx_burst_sp;
+ else
+ dev->rx_pkt_burst = mlx5_rx_burst;
+ }
+ priv_unlock(priv);
+ return -ret;
+}
+
+/**
+ * DPDK callback to release a RX queue.
+ *
+ * @param dpdk_rxq
+ * Generic RX queue pointer.
+ */
+void
+mlx5_rx_queue_release(void *dpdk_rxq)
+{
+ struct rxq *rxq = (struct rxq *)dpdk_rxq;
+ struct priv *priv;
+ unsigned int i;
+
+ if (mlx5_is_secondary())
+ return;
+
+ if (rxq == NULL)
+ return;
+ priv = rxq->priv;
+ priv_lock(priv);
+ for (i = 0; (i != priv->rxqs_n); ++i)
+ if ((*priv->rxqs)[i] == rxq) {
+ DEBUG("%p: removing RX queue %p from list",
+ (void *)priv->dev, (void *)rxq);
+ (*priv->rxqs)[i] = NULL;
+ break;
+ }
+ rxq_cleanup(rxq);
+ rte_free(rxq);
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback for RX in secondary processes.
+ *
+ * This function configures all queues from primary process information
+ * if necessary before reverting to the normal RX burst callback.
+ *
+ * @param dpdk_rxq
+ * Generic pointer to RX queue structure.
+ * @param[out] pkts
+ * Array to store received packets.
+ * @param pkts_n
+ * Maximum number of packets in array.
+ *
+ * @return
+ * Number of packets successfully received (<= pkts_n).
+ */
+uint16_t
+mlx5_rx_burst_secondary_setup(void *dpdk_rxq, struct rte_mbuf **pkts,
+ uint16_t pkts_n)
+{
+ struct rxq *rxq = dpdk_rxq;
+ struct priv *priv = mlx5_secondary_data_setup(rxq->priv);
+ struct priv *primary_priv;
+ unsigned int index;
+
+ if (priv == NULL)
+ return 0;
+ primary_priv =
+ mlx5_secondary_data[priv->dev->data->port_id].primary_priv;
+ /* Look for queue index in both private structures. */
+ for (index = 0; index != priv->rxqs_n; ++index)
+ if (((*primary_priv->rxqs)[index] == rxq) ||
+ ((*priv->rxqs)[index] == rxq))
+ break;
+ if (index == priv->rxqs_n)
+ return 0;
+ rxq = (*priv->rxqs)[index];
+ return priv->dev->rx_pkt_burst(rxq, pkts, pkts_n);
+}
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
new file mode 100644
index 00000000..9d1380a0
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -0,0 +1,1249 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <assert.h>
+#include <stdint.h>
+#include <string.h>
+#include <stdlib.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_mbuf.h>
+#include <rte_mempool.h>
+#include <rte_prefetch.h>
+#include <rte_common.h>
+#include <rte_branch_prediction.h>
+#include <rte_memory.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_utils.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+/**
+ * Manage TX completions.
+ *
+ * When sending a burst, mlx5_tx_burst() posts several WRs.
+ * To improve performance, a completion event is only required once every
+ * MLX5_PMD_TX_PER_COMP_REQ sends. Doing so discards completion information
+ * for other WRs, but this information would not be used anyway.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ *
+ * @return
+ * 0 on success, -1 on failure.
+ */
+static int
+txq_complete(struct txq *txq)
+{
+ unsigned int elts_comp = txq->elts_comp;
+ unsigned int elts_tail = txq->elts_tail;
+ unsigned int elts_free = txq->elts_tail;
+ const unsigned int elts_n = txq->elts_n;
+ int wcs_n;
+
+ if (unlikely(elts_comp == 0))
+ return 0;
+#ifdef DEBUG_SEND
+ DEBUG("%p: processing %u work requests completions",
+ (void *)txq, elts_comp);
+#endif
+ wcs_n = txq->poll_cnt(txq->cq, elts_comp);
+ if (unlikely(wcs_n == 0))
+ return 0;
+ if (unlikely(wcs_n < 0)) {
+ DEBUG("%p: ibv_poll_cq() failed (wcs_n=%d)",
+ (void *)txq, wcs_n);
+ return -1;
+ }
+ elts_comp -= wcs_n;
+ assert(elts_comp <= txq->elts_comp);
+ /*
+ * Assume WC status is successful as nothing can be done about it
+ * anyway.
+ */
+ elts_tail += wcs_n * txq->elts_comp_cd_init;
+ if (elts_tail >= elts_n)
+ elts_tail -= elts_n;
+
+ while (elts_free != elts_tail) {
+ struct txq_elt *elt = &(*txq->elts)[elts_free];
+ unsigned int elts_free_next =
+ (((elts_free + 1) == elts_n) ? 0 : elts_free + 1);
+ struct rte_mbuf *tmp = elt->buf;
+ struct txq_elt *elt_next = &(*txq->elts)[elts_free_next];
+
+#ifndef NDEBUG
+ /* Poisoning. */
+ memset(elt, 0x66, sizeof(*elt));
+#endif
+ RTE_MBUF_PREFETCH_TO_FREE(elt_next->buf);
+ /* Faster than rte_pktmbuf_free(). */
+ do {
+ struct rte_mbuf *next = NEXT(tmp);
+
+ rte_pktmbuf_free_seg(tmp);
+ tmp = next;
+ } while (tmp != NULL);
+ elts_free = elts_free_next;
+ }
+
+ txq->elts_tail = elts_tail;
+ txq->elts_comp = elts_comp;
+ return 0;
+}
+
+/* For best performance, this function should not be inlined. */
+struct ibv_mr *mlx5_mp2mr(struct ibv_pd *, const struct rte_mempool *)
+ __attribute__((noinline));
+
+/**
+ * Register mempool as a memory region.
+ *
+ * @param pd
+ * Pointer to protection domain.
+ * @param mp
+ * Pointer to memory pool.
+ *
+ * @return
+ * Memory region pointer, NULL in case of error.
+ */
+struct ibv_mr *
+mlx5_mp2mr(struct ibv_pd *pd, const struct rte_mempool *mp)
+{
+ const struct rte_memseg *ms = rte_eal_get_physmem_layout();
+ uintptr_t start = mp->elt_va_start;
+ uintptr_t end = mp->elt_va_end;
+ unsigned int i;
+
+ DEBUG("mempool %p area start=%p end=%p size=%zu",
+ (const void *)mp, (void *)start, (void *)end,
+ (size_t)(end - start));
+ /* Round start and end to page boundary if found in memory segments. */
+ for (i = 0; (i < RTE_MAX_MEMSEG) && (ms[i].addr != NULL); ++i) {
+ uintptr_t addr = (uintptr_t)ms[i].addr;
+ size_t len = ms[i].len;
+ unsigned int align = ms[i].hugepage_sz;
+
+ if ((start > addr) && (start < addr + len))
+ start = RTE_ALIGN_FLOOR(start, align);
+ if ((end > addr) && (end < addr + len))
+ end = RTE_ALIGN_CEIL(end, align);
+ }
+ DEBUG("mempool %p using start=%p end=%p size=%zu for MR",
+ (const void *)mp, (void *)start, (void *)end,
+ (size_t)(end - start));
+ return ibv_reg_mr(pd,
+ (void *)start,
+ end - start,
+ IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_REMOTE_WRITE);
+}
+
+/**
+ * Get Memory Pool (MP) from mbuf. If mbuf is indirect, the pool from which
+ * the cloned mbuf is allocated is returned instead.
+ *
+ * @param buf
+ * Pointer to mbuf.
+ *
+ * @return
+ * Memory pool where data is located for given mbuf.
+ */
+static struct rte_mempool *
+txq_mb2mp(struct rte_mbuf *buf)
+{
+ if (unlikely(RTE_MBUF_INDIRECT(buf)))
+ return rte_mbuf_from_indirect(buf)->pool;
+ return buf->pool;
+}
+
+/**
+ * Get Memory Region (MR) <-> Memory Pool (MP) association from txq->mp2mr[].
+ * Add MP to txq->mp2mr[] if it's not registered yet. If mp2mr[] is full,
+ * remove an entry first.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param[in] mp
+ * Memory Pool for which a Memory Region lkey must be returned.
+ *
+ * @return
+ * mr->lkey on success, (uint32_t)-1 on failure.
+ */
+static uint32_t
+txq_mp2mr(struct txq *txq, const struct rte_mempool *mp)
+{
+ unsigned int i;
+ struct ibv_mr *mr;
+
+ for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) {
+ if (unlikely(txq->mp2mr[i].mp == NULL)) {
+ /* Unknown MP, add a new MR for it. */
+ break;
+ }
+ if (txq->mp2mr[i].mp == mp) {
+ assert(txq->mp2mr[i].lkey != (uint32_t)-1);
+ assert(txq->mp2mr[i].mr->lkey == txq->mp2mr[i].lkey);
+ return txq->mp2mr[i].lkey;
+ }
+ }
+ /* Add a new entry, register MR first. */
+ DEBUG("%p: discovered new memory pool \"%s\" (%p)",
+ (void *)txq, mp->name, (const void *)mp);
+ mr = mlx5_mp2mr(txq->priv->pd, mp);
+ if (unlikely(mr == NULL)) {
+ DEBUG("%p: unable to configure MR, ibv_reg_mr() failed.",
+ (void *)txq);
+ return (uint32_t)-1;
+ }
+ if (unlikely(i == RTE_DIM(txq->mp2mr))) {
+ /* Table is full, remove oldest entry. */
+ DEBUG("%p: MR <-> MP table full, dropping oldest entry.",
+ (void *)txq);
+ --i;
+ claim_zero(ibv_dereg_mr(txq->mp2mr[0].mr));
+ memmove(&txq->mp2mr[0], &txq->mp2mr[1],
+ (sizeof(txq->mp2mr) - sizeof(txq->mp2mr[0])));
+ }
+ /* Store the new entry. */
+ txq->mp2mr[i].mp = mp;
+ txq->mp2mr[i].mr = mr;
+ txq->mp2mr[i].lkey = mr->lkey;
+ DEBUG("%p: new MR lkey for MP \"%s\" (%p): 0x%08" PRIu32,
+ (void *)txq, mp->name, (const void *)mp, txq->mp2mr[i].lkey);
+ return txq->mp2mr[i].lkey;
+}
+
+struct txq_mp2mr_mbuf_check_data {
+ const struct rte_mempool *mp;
+ int ret;
+};
+
+/**
+ * Callback function for rte_mempool_obj_iter() to check whether a given
+ * mempool object looks like a mbuf.
+ *
+ * @param[in, out] arg
+ * Context data (struct txq_mp2mr_mbuf_check_data). Contains mempool pointer
+ * and return value.
+ * @param[in] start
+ * Object start address.
+ * @param[in] end
+ * Object end address.
+ * @param index
+ * Unused.
+ *
+ * @return
+ * Nonzero value when object is not a mbuf.
+ */
+static void
+txq_mp2mr_mbuf_check(void *arg, void *start, void *end,
+ uint32_t index __rte_unused)
+{
+ struct txq_mp2mr_mbuf_check_data *data = arg;
+ struct rte_mbuf *buf =
+ (void *)((uintptr_t)start + data->mp->header_size);
+
+ (void)index;
+ /* Check whether mbuf structure fits element size and whether mempool
+ * pointer is valid. */
+ if (((uintptr_t)end >= (uintptr_t)(buf + 1)) &&
+ (buf->pool == data->mp))
+ data->ret = 0;
+ else
+ data->ret = -1;
+}
+
+/**
+ * Iterator function for rte_mempool_walk() to register existing mempools and
+ * fill the MP to MR cache of a TX queue.
+ *
+ * @param[in] mp
+ * Memory Pool to register.
+ * @param *arg
+ * Pointer to TX queue structure.
+ */
+void
+txq_mp2mr_iter(const struct rte_mempool *mp, void *arg)
+{
+ struct txq *txq = arg;
+ struct txq_mp2mr_mbuf_check_data data = {
+ .mp = mp,
+ .ret = -1,
+ };
+
+ /* Discard empty mempools. */
+ if (mp->size == 0)
+ return;
+ /* Register mempool only if the first element looks like a mbuf. */
+ rte_mempool_obj_iter((void *)mp->elt_va_start,
+ 1,
+ mp->header_size + mp->elt_size + mp->trailer_size,
+ 1,
+ mp->elt_pa,
+ mp->pg_num,
+ mp->pg_shift,
+ txq_mp2mr_mbuf_check,
+ &data);
+ if (data.ret)
+ return;
+ txq_mp2mr(txq, mp);
+}
+
+/**
+ * Insert VLAN using mbuf headroom space.
+ *
+ * @param buf
+ * Buffer for VLAN insertion.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static inline int
+insert_vlan_sw(struct rte_mbuf *buf)
+{
+ uintptr_t addr;
+ uint32_t vlan;
+ uint16_t head_room_len = rte_pktmbuf_headroom(buf);
+
+ if (head_room_len < 4)
+ return EINVAL;
+
+ addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ vlan = htonl(0x81000000 | buf->vlan_tci);
+ memmove((void *)(addr - 4), (void *)addr, 12);
+ memcpy((void *)(addr + 8), &vlan, sizeof(vlan));
+
+ SET_DATA_OFF(buf, head_room_len - 4);
+ DATA_LEN(buf) += 4;
+
+ return 0;
+}
+
+#if MLX5_PMD_SGE_WR_N > 1
+
+/**
+ * Copy scattered mbuf contents to a single linear buffer.
+ *
+ * @param[out] linear
+ * Linear output buffer.
+ * @param[in] buf
+ * Scattered input buffer.
+ *
+ * @return
+ * Number of bytes copied to the output buffer or 0 if not large enough.
+ */
+static unsigned int
+linearize_mbuf(linear_t *linear, struct rte_mbuf *buf)
+{
+ unsigned int size = 0;
+ unsigned int offset;
+
+ do {
+ unsigned int len = DATA_LEN(buf);
+
+ offset = size;
+ size += len;
+ if (unlikely(size > sizeof(*linear)))
+ return 0;
+ memcpy(&(*linear)[offset],
+ rte_pktmbuf_mtod(buf, uint8_t *),
+ len);
+ buf = NEXT(buf);
+ } while (buf != NULL);
+ return size;
+}
+
+/**
+ * Handle scattered buffers for mlx5_tx_burst().
+ *
+ * @param txq
+ * TX queue structure.
+ * @param segs
+ * Number of segments in buf.
+ * @param elt
+ * TX queue element to fill.
+ * @param[in] buf
+ * Buffer to process.
+ * @param elts_head
+ * Index of the linear buffer to use if necessary (normally txq->elts_head).
+ * @param[out] sges
+ * Array filled with SGEs on success.
+ *
+ * @return
+ * A structure containing the processed packet size in bytes and the
+ * number of SGEs. Both fields are set to (unsigned int)-1 in case of
+ * failure.
+ */
+static struct tx_burst_sg_ret {
+ unsigned int length;
+ unsigned int num;
+}
+tx_burst_sg(struct txq *txq, unsigned int segs, struct txq_elt *elt,
+ struct rte_mbuf *buf, unsigned int elts_head,
+ struct ibv_sge (*sges)[MLX5_PMD_SGE_WR_N])
+{
+ unsigned int sent_size = 0;
+ unsigned int j;
+ int linearize = 0;
+
+ /* When there are too many segments, extra segments are
+ * linearized in the last SGE. */
+ if (unlikely(segs > RTE_DIM(*sges))) {
+ segs = (RTE_DIM(*sges) - 1);
+ linearize = 1;
+ }
+ /* Update element. */
+ elt->buf = buf;
+ /* Register segments as SGEs. */
+ for (j = 0; (j != segs); ++j) {
+ struct ibv_sge *sge = &(*sges)[j];
+ uint32_t lkey;
+
+ /* Retrieve Memory Region key for this memory pool. */
+ lkey = txq_mp2mr(txq, txq_mb2mp(buf));
+ if (unlikely(lkey == (uint32_t)-1)) {
+ /* MR does not exist. */
+ DEBUG("%p: unable to get MP <-> MR association",
+ (void *)txq);
+ /* Clean up TX element. */
+ elt->buf = NULL;
+ goto stop;
+ }
+ /* Update SGE. */
+ sge->addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ if (txq->priv->vf)
+ rte_prefetch0((volatile void *)
+ (uintptr_t)sge->addr);
+ sge->length = DATA_LEN(buf);
+ sge->lkey = lkey;
+ sent_size += sge->length;
+ buf = NEXT(buf);
+ }
+ /* If buf is not NULL here and is not going to be linearized,
+ * nb_segs is not valid. */
+ assert(j == segs);
+ assert((buf == NULL) || (linearize));
+ /* Linearize extra segments. */
+ if (linearize) {
+ struct ibv_sge *sge = &(*sges)[segs];
+ linear_t *linear = &(*txq->elts_linear)[elts_head];
+ unsigned int size = linearize_mbuf(linear, buf);
+
+ assert(segs == (RTE_DIM(*sges) - 1));
+ if (size == 0) {
+ /* Invalid packet. */
+ DEBUG("%p: packet too large to be linearized.",
+ (void *)txq);
+ /* Clean up TX element. */
+ elt->buf = NULL;
+ goto stop;
+ }
+ /* If MLX5_PMD_SGE_WR_N is 1, free mbuf immediately. */
+ if (RTE_DIM(*sges) == 1) {
+ do {
+ struct rte_mbuf *next = NEXT(buf);
+
+ rte_pktmbuf_free_seg(buf);
+ buf = next;
+ } while (buf != NULL);
+ elt->buf = NULL;
+ }
+ /* Update SGE. */
+ sge->addr = (uintptr_t)&(*linear)[0];
+ sge->length = size;
+ sge->lkey = txq->mr_linear->lkey;
+ sent_size += size;
+ /* Include last segment. */
+ segs++;
+ }
+ return (struct tx_burst_sg_ret){
+ .length = sent_size,
+ .num = segs,
+ };
+stop:
+ return (struct tx_burst_sg_ret){
+ .length = -1,
+ .num = -1,
+ };
+}
+
+#endif /* MLX5_PMD_SGE_WR_N > 1 */
+
+/**
+ * DPDK callback for TX.
+ *
+ * @param dpdk_txq
+ * Generic pointer to TX queue structure.
+ * @param[in] pkts
+ * Packets to transmit.
+ * @param pkts_n
+ * Number of packets in array.
+ *
+ * @return
+ * Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ struct txq *txq = (struct txq *)dpdk_txq;
+ unsigned int elts_head = txq->elts_head;
+ const unsigned int elts_n = txq->elts_n;
+ unsigned int elts_comp_cd = txq->elts_comp_cd;
+ unsigned int elts_comp = 0;
+ unsigned int i;
+ unsigned int max;
+ int err;
+ struct rte_mbuf *buf = pkts[0];
+
+ assert(elts_comp_cd != 0);
+ /* Prefetch first packet cacheline. */
+ rte_prefetch0(buf);
+ txq_complete(txq);
+ max = (elts_n - (elts_head - txq->elts_tail));
+ if (max > elts_n)
+ max -= elts_n;
+ assert(max >= 1);
+ assert(max <= elts_n);
+ /* Always leave one free entry in the ring. */
+ --max;
+ if (max == 0)
+ return 0;
+ if (max > pkts_n)
+ max = pkts_n;
+ for (i = 0; (i != max); ++i) {
+ struct rte_mbuf *buf_next = pkts[i + 1];
+ unsigned int elts_head_next =
+ (((elts_head + 1) == elts_n) ? 0 : elts_head + 1);
+ struct txq_elt *elt = &(*txq->elts)[elts_head];
+ unsigned int segs = NB_SEGS(buf);
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ unsigned int sent_size = 0;
+#endif
+ uint32_t send_flags = 0;
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ int insert_vlan = 0;
+#endif /* HAVE_VERBS_VLAN_INSERTION */
+
+ if (i + 1 < max)
+ rte_prefetch0(buf_next);
+ /* Request TX completion. */
+ if (unlikely(--elts_comp_cd == 0)) {
+ elts_comp_cd = txq->elts_comp_cd_init;
+ ++elts_comp;
+ send_flags |= IBV_EXP_QP_BURST_SIGNALED;
+ }
+ /* Should we enable HW CKSUM offload */
+ if (buf->ol_flags &
+ (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) {
+ send_flags |= IBV_EXP_QP_BURST_IP_CSUM;
+ /* HW does not support checksum offloads at arbitrary
+ * offsets but automatically recognizes the packet
+ * type. For inner L3/L4 checksums, only VXLAN (UDP)
+ * tunnels are currently supported. */
+ if (RTE_ETH_IS_TUNNEL_PKT(buf->packet_type))
+ send_flags |= IBV_EXP_QP_BURST_TUNNEL;
+ }
+ if (buf->ol_flags & PKT_TX_VLAN_PKT) {
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ if (!txq->priv->mps)
+ insert_vlan = 1;
+ else
+#endif /* HAVE_VERBS_VLAN_INSERTION */
+ {
+ err = insert_vlan_sw(buf);
+ if (unlikely(err))
+ goto stop;
+ }
+ }
+ if (likely(segs == 1)) {
+ uintptr_t addr;
+ uint32_t length;
+ uint32_t lkey;
+ uintptr_t buf_next_addr;
+
+ /* Retrieve buffer information. */
+ addr = rte_pktmbuf_mtod(buf, uintptr_t);
+ length = DATA_LEN(buf);
+ /* Update element. */
+ elt->buf = buf;
+ if (txq->priv->vf)
+ rte_prefetch0((volatile void *)
+ (uintptr_t)addr);
+ /* Prefetch next buffer data. */
+ if (i + 1 < max) {
+ buf_next_addr =
+ rte_pktmbuf_mtod(buf_next, uintptr_t);
+ rte_prefetch0((volatile void *)
+ (uintptr_t)buf_next_addr);
+ }
+ /* Put packet into send queue. */
+#if MLX5_PMD_MAX_INLINE > 0
+ if (length <= txq->max_inline) {
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ if (insert_vlan)
+ err = txq->send_pending_inline_vlan
+ (txq->qp,
+ (void *)addr,
+ length,
+ send_flags,
+ &buf->vlan_tci);
+ else
+#endif /* HAVE_VERBS_VLAN_INSERTION */
+ err = txq->send_pending_inline
+ (txq->qp,
+ (void *)addr,
+ length,
+ send_flags);
+ } else
+#endif
+ {
+ /* Retrieve Memory Region key for this
+ * memory pool. */
+ lkey = txq_mp2mr(txq, txq_mb2mp(buf));
+ if (unlikely(lkey == (uint32_t)-1)) {
+ /* MR does not exist. */
+ DEBUG("%p: unable to get MP <-> MR"
+ " association", (void *)txq);
+ /* Clean up TX element. */
+ elt->buf = NULL;
+ goto stop;
+ }
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ if (insert_vlan)
+ err = txq->send_pending_vlan
+ (txq->qp,
+ addr,
+ length,
+ lkey,
+ send_flags,
+ &buf->vlan_tci);
+ else
+#endif /* HAVE_VERBS_VLAN_INSERTION */
+ err = txq->send_pending
+ (txq->qp,
+ addr,
+ length,
+ lkey,
+ send_flags);
+ }
+ if (unlikely(err))
+ goto stop;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ sent_size += length;
+#endif
+ } else {
+#if MLX5_PMD_SGE_WR_N > 1
+ struct ibv_sge sges[MLX5_PMD_SGE_WR_N];
+ struct tx_burst_sg_ret ret;
+
+ ret = tx_burst_sg(txq, segs, elt, buf, elts_head,
+ &sges);
+ if (ret.length == (unsigned int)-1)
+ goto stop;
+ /* Put SG list into send queue. */
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ if (insert_vlan)
+ err = txq->send_pending_sg_list_vlan
+ (txq->qp,
+ sges,
+ ret.num,
+ send_flags,
+ &buf->vlan_tci);
+ else
+#endif /* HAVE_VERBS_VLAN_INSERTION */
+ err = txq->send_pending_sg_list
+ (txq->qp,
+ sges,
+ ret.num,
+ send_flags);
+ if (unlikely(err))
+ goto stop;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ sent_size += ret.length;
+#endif
+#else /* MLX5_PMD_SGE_WR_N > 1 */
+ DEBUG("%p: TX scattered buffers support not"
+ " compiled in", (void *)txq);
+ goto stop;
+#endif /* MLX5_PMD_SGE_WR_N > 1 */
+ }
+ elts_head = elts_head_next;
+ buf = buf_next;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent bytes counter. */
+ txq->stats.obytes += sent_size;
+#endif
+ }
+stop:
+ /* Take a shortcut if nothing must be sent. */
+ if (unlikely(i == 0))
+ return 0;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment sent packets counter. */
+ txq->stats.opackets += i;
+#endif
+ /* Ring QP doorbell. */
+ err = txq->send_flush(txq->qp);
+ if (unlikely(err)) {
+ /* A nonzero value is not supposed to be returned.
+ * Nothing can be done about it. */
+ DEBUG("%p: send_flush() failed with error %d",
+ (void *)txq, err);
+ }
+ txq->elts_head = elts_head;
+ txq->elts_comp += elts_comp;
+ txq->elts_comp_cd = elts_comp_cd;
+ return i;
+}
+
+/**
+ * Translate RX completion flags to packet type.
+ *
+ * @param flags
+ * RX completion flags returned by poll_length_flags().
+ *
+ * @note: fix mlx5_dev_supported_ptypes_get() if any change here.
+ *
+ * @return
+ * Packet type for struct rte_mbuf.
+ */
+static inline uint32_t
+rxq_cq_to_pkt_type(uint32_t flags)
+{
+ uint32_t pkt_type;
+
+ if (flags & IBV_EXP_CQ_RX_TUNNEL_PACKET)
+ pkt_type =
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_OUTER_IPV4_PACKET,
+ RTE_PTYPE_L3_IPV4) |
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_OUTER_IPV6_PACKET,
+ RTE_PTYPE_L3_IPV6) |
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_IPV4_PACKET,
+ RTE_PTYPE_INNER_L3_IPV4) |
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_IPV6_PACKET,
+ RTE_PTYPE_INNER_L3_IPV6);
+ else
+ pkt_type =
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_IPV4_PACKET,
+ RTE_PTYPE_L3_IPV4) |
+ TRANSPOSE(flags,
+ IBV_EXP_CQ_RX_IPV6_PACKET,
+ RTE_PTYPE_L3_IPV6);
+ return pkt_type;
+}
+
+/**
+ * Translate RX completion flags to offload flags.
+ *
+ * @param[in] rxq
+ * Pointer to RX queue structure.
+ * @param flags
+ * RX completion flags returned by poll_length_flags().
+ *
+ * @return
+ * Offload flags (ol_flags) for struct rte_mbuf.
+ */
+static inline uint32_t
+rxq_cq_to_ol_flags(const struct rxq *rxq, uint32_t flags)
+{
+ uint32_t ol_flags = 0;
+
+ if (rxq->csum) {
+ /* Set IP checksum flag only for IPv4/IPv6 packets. */
+ if (flags &
+ (IBV_EXP_CQ_RX_IPV4_PACKET | IBV_EXP_CQ_RX_IPV6_PACKET))
+ ol_flags |=
+ TRANSPOSE(~flags,
+ IBV_EXP_CQ_RX_IP_CSUM_OK,
+ PKT_RX_IP_CKSUM_BAD);
+#ifdef HAVE_EXP_CQ_RX_TCP_PACKET
+ /* Set L4 checksum flag only for TCP/UDP packets. */
+ if (flags &
+ (IBV_EXP_CQ_RX_TCP_PACKET | IBV_EXP_CQ_RX_UDP_PACKET))
+#endif /* HAVE_EXP_CQ_RX_TCP_PACKET */
+ ol_flags |=
+ TRANSPOSE(~flags,
+ IBV_EXP_CQ_RX_TCP_UDP_CSUM_OK,
+ PKT_RX_L4_CKSUM_BAD);
+ }
+ /*
+ * PKT_RX_IP_CKSUM_BAD and PKT_RX_L4_CKSUM_BAD are used in place
+ * of PKT_RX_EIP_CKSUM_BAD because the latter is not functional
+ * (its value is 0).
+ */
+ if ((flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) && (rxq->csum_l2tun))
+ ol_flags |=
+ TRANSPOSE(~flags,
+ IBV_EXP_CQ_RX_OUTER_IP_CSUM_OK,
+ PKT_RX_IP_CKSUM_BAD) |
+ TRANSPOSE(~flags,
+ IBV_EXP_CQ_RX_OUTER_TCP_UDP_CSUM_OK,
+ PKT_RX_L4_CKSUM_BAD);
+ return ol_flags;
+}
+
+/**
+ * DPDK callback for RX with scattered packets support.
+ *
+ * @param dpdk_rxq
+ * Generic pointer to RX queue structure.
+ * @param[out] pkts
+ * Array to store received packets.
+ * @param pkts_n
+ * Maximum number of packets in array.
+ *
+ * @return
+ * Number of packets successfully received (<= pkts_n).
+ */
+uint16_t
+mlx5_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ struct rxq *rxq = (struct rxq *)dpdk_rxq;
+ struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp;
+ const unsigned int elts_n = rxq->elts_n;
+ unsigned int elts_head = rxq->elts_head;
+ unsigned int i;
+ unsigned int pkts_ret = 0;
+ int ret;
+
+ if (unlikely(!rxq->sp))
+ return mlx5_rx_burst(dpdk_rxq, pkts, pkts_n);
+ if (unlikely(elts == NULL)) /* See RTE_DEV_CMD_SET_MTU. */
+ return 0;
+ for (i = 0; (i != pkts_n); ++i) {
+ struct rxq_elt_sp *elt = &(*elts)[elts_head];
+ unsigned int len;
+ unsigned int pkt_buf_len;
+ struct rte_mbuf *pkt_buf = NULL; /* Buffer returned in pkts. */
+ struct rte_mbuf **pkt_buf_next = &pkt_buf;
+ unsigned int seg_headroom = RTE_PKTMBUF_HEADROOM;
+ unsigned int j = 0;
+ uint32_t flags;
+ uint16_t vlan_tci;
+
+ /* Sanity checks. */
+ assert(elts_head < rxq->elts_n);
+ assert(rxq->elts_head < rxq->elts_n);
+ ret = rxq->poll(rxq->cq, NULL, NULL, &flags, &vlan_tci);
+ if (unlikely(ret < 0)) {
+ struct ibv_wc wc;
+ int wcs_n;
+
+ DEBUG("rxq=%p, poll_length() failed (ret=%d)",
+ (void *)rxq, ret);
+ /* ibv_poll_cq() must be used in case of failure. */
+ wcs_n = ibv_poll_cq(rxq->cq, 1, &wc);
+ if (unlikely(wcs_n == 0))
+ break;
+ if (unlikely(wcs_n < 0)) {
+ DEBUG("rxq=%p, ibv_poll_cq() failed (wcs_n=%d)",
+ (void *)rxq, wcs_n);
+ break;
+ }
+ assert(wcs_n == 1);
+ if (unlikely(wc.status != IBV_WC_SUCCESS)) {
+ /* Whatever, just repost the offending WR. */
+ DEBUG("rxq=%p, wr_id=%" PRIu64 ": bad work"
+ " completion status (%d): %s",
+ (void *)rxq, wc.wr_id, wc.status,
+ ibv_wc_status_str(wc.status));
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment dropped packets counter. */
+ ++rxq->stats.idropped;
+#endif
+ goto repost;
+ }
+ ret = wc.byte_len;
+ }
+ if (ret == 0)
+ break;
+ assert(ret >= (rxq->crc_present << 2));
+ len = ret - (rxq->crc_present << 2);
+ pkt_buf_len = len;
+ /*
+ * Replace spent segments with new ones, concatenate and
+ * return them as pkt_buf.
+ */
+ while (1) {
+ struct ibv_sge *sge = &elt->sges[j];
+ struct rte_mbuf *seg = elt->bufs[j];
+ struct rte_mbuf *rep;
+ unsigned int seg_tailroom;
+
+ assert(seg != NULL);
+ /*
+ * Fetch initial bytes of packet descriptor into a
+ * cacheline while allocating rep.
+ */
+ rte_prefetch0(seg);
+ rep = __rte_mbuf_raw_alloc(rxq->mp);
+ if (unlikely(rep == NULL)) {
+ /*
+ * Unable to allocate a replacement mbuf,
+ * repost WR.
+ */
+ DEBUG("rxq=%p: can't allocate a new mbuf",
+ (void *)rxq);
+ if (pkt_buf != NULL) {
+ *pkt_buf_next = NULL;
+ rte_pktmbuf_free(pkt_buf);
+ }
+ /* Increment out of memory counters. */
+ ++rxq->stats.rx_nombuf;
+ ++rxq->priv->dev->data->rx_mbuf_alloc_failed;
+ goto repost;
+ }
+#ifndef NDEBUG
+ /* Poison user-modifiable fields in rep. */
+ NEXT(rep) = (void *)((uintptr_t)-1);
+ SET_DATA_OFF(rep, 0xdead);
+ DATA_LEN(rep) = 0xd00d;
+ PKT_LEN(rep) = 0xdeadd00d;
+ NB_SEGS(rep) = 0x2a;
+ PORT(rep) = 0x2a;
+ rep->ol_flags = -1;
+#endif
+ assert(rep->buf_len == seg->buf_len);
+ assert(rep->buf_len == rxq->mb_len);
+ /* Reconfigure sge to use rep instead of seg. */
+ assert(sge->lkey == rxq->mr->lkey);
+ sge->addr = ((uintptr_t)rep->buf_addr + seg_headroom);
+ elt->bufs[j] = rep;
+ ++j;
+ /* Update pkt_buf if it's the first segment, or link
+ * seg to the previous one and update pkt_buf_next. */
+ *pkt_buf_next = seg;
+ pkt_buf_next = &NEXT(seg);
+ /* Update seg information. */
+ seg_tailroom = (seg->buf_len - seg_headroom);
+ assert(sge->length == seg_tailroom);
+ SET_DATA_OFF(seg, seg_headroom);
+ if (likely(len <= seg_tailroom)) {
+ /* Last segment. */
+ DATA_LEN(seg) = len;
+ PKT_LEN(seg) = len;
+ /* Sanity check. */
+ assert(rte_pktmbuf_headroom(seg) ==
+ seg_headroom);
+ assert(rte_pktmbuf_tailroom(seg) ==
+ (seg_tailroom - len));
+ break;
+ }
+ DATA_LEN(seg) = seg_tailroom;
+ PKT_LEN(seg) = seg_tailroom;
+ /* Sanity check. */
+ assert(rte_pktmbuf_headroom(seg) == seg_headroom);
+ assert(rte_pktmbuf_tailroom(seg) == 0);
+ /* Fix len and clear headroom for next segments. */
+ len -= seg_tailroom;
+ seg_headroom = 0;
+ }
+ /* Update head and tail segments. */
+ *pkt_buf_next = NULL;
+ assert(pkt_buf != NULL);
+ assert(j != 0);
+ NB_SEGS(pkt_buf) = j;
+ PORT(pkt_buf) = rxq->port_id;
+ PKT_LEN(pkt_buf) = pkt_buf_len;
+ if (rxq->csum | rxq->csum_l2tun | rxq->vlan_strip) {
+ pkt_buf->packet_type = rxq_cq_to_pkt_type(flags);
+ pkt_buf->ol_flags = rxq_cq_to_ol_flags(rxq, flags);
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ if (flags & IBV_EXP_CQ_RX_CVLAN_STRIPPED_V1) {
+ pkt_buf->ol_flags |= PKT_RX_VLAN_PKT;
+ pkt_buf->vlan_tci = vlan_tci;
+ }
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ }
+
+ /* Return packet. */
+ *(pkts++) = pkt_buf;
+ ++pkts_ret;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment bytes counter. */
+ rxq->stats.ibytes += pkt_buf_len;
+#endif
+repost:
+ ret = rxq->recv(rxq->wq, elt->sges, RTE_DIM(elt->sges));
+ if (unlikely(ret)) {
+ /* Inability to repost WRs is fatal. */
+ DEBUG("%p: recv_sg_list(): failed (ret=%d)",
+ (void *)rxq->priv,
+ ret);
+ abort();
+ }
+ if (++elts_head >= elts_n)
+ elts_head = 0;
+ continue;
+ }
+ if (unlikely(i == 0))
+ return 0;
+ rxq->elts_head = elts_head;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment packets counter. */
+ rxq->stats.ipackets += pkts_ret;
+#endif
+ return pkts_ret;
+}
+
+/**
+ * DPDK callback for RX.
+ *
+ * The following function is the same as mlx5_rx_burst_sp(), except it doesn't
+ * manage scattered packets. Improves performance when MRU is lower than the
+ * size of the first segment.
+ *
+ * @param dpdk_rxq
+ * Generic pointer to RX queue structure.
+ * @param[out] pkts
+ * Array to store received packets.
+ * @param pkts_n
+ * Maximum number of packets in array.
+ *
+ * @return
+ * Number of packets successfully received (<= pkts_n).
+ */
+uint16_t
+mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ struct rxq *rxq = (struct rxq *)dpdk_rxq;
+ struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp;
+ const unsigned int elts_n = rxq->elts_n;
+ unsigned int elts_head = rxq->elts_head;
+ struct ibv_sge sges[pkts_n];
+ unsigned int i;
+ unsigned int pkts_ret = 0;
+ int ret;
+
+ if (unlikely(rxq->sp))
+ return mlx5_rx_burst_sp(dpdk_rxq, pkts, pkts_n);
+ for (i = 0; (i != pkts_n); ++i) {
+ struct rxq_elt *elt = &(*elts)[elts_head];
+ unsigned int len;
+ struct rte_mbuf *seg = elt->buf;
+ struct rte_mbuf *rep;
+ uint32_t flags;
+ uint16_t vlan_tci;
+
+ /* Sanity checks. */
+ assert(seg != NULL);
+ assert(elts_head < rxq->elts_n);
+ assert(rxq->elts_head < rxq->elts_n);
+ /*
+ * Fetch initial bytes of packet descriptor into a
+ * cacheline while allocating rep.
+ */
+ rte_prefetch0(seg);
+ rte_prefetch0(&seg->cacheline1);
+ ret = rxq->poll(rxq->cq, NULL, NULL, &flags, &vlan_tci);
+ if (unlikely(ret < 0)) {
+ struct ibv_wc wc;
+ int wcs_n;
+
+ DEBUG("rxq=%p, poll_length() failed (ret=%d)",
+ (void *)rxq, ret);
+ /* ibv_poll_cq() must be used in case of failure. */
+ wcs_n = ibv_poll_cq(rxq->cq, 1, &wc);
+ if (unlikely(wcs_n == 0))
+ break;
+ if (unlikely(wcs_n < 0)) {
+ DEBUG("rxq=%p, ibv_poll_cq() failed (wcs_n=%d)",
+ (void *)rxq, wcs_n);
+ break;
+ }
+ assert(wcs_n == 1);
+ if (unlikely(wc.status != IBV_WC_SUCCESS)) {
+ /* Whatever, just repost the offending WR. */
+ DEBUG("rxq=%p, wr_id=%" PRIu64 ": bad work"
+ " completion status (%d): %s",
+ (void *)rxq, wc.wr_id, wc.status,
+ ibv_wc_status_str(wc.status));
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment dropped packets counter. */
+ ++rxq->stats.idropped;
+#endif
+ /* Add SGE to array for repost. */
+ sges[i] = elt->sge;
+ goto repost;
+ }
+ ret = wc.byte_len;
+ }
+ if (ret == 0)
+ break;
+ assert(ret >= (rxq->crc_present << 2));
+ len = ret - (rxq->crc_present << 2);
+ rep = __rte_mbuf_raw_alloc(rxq->mp);
+ if (unlikely(rep == NULL)) {
+ /*
+ * Unable to allocate a replacement mbuf,
+ * repost WR.
+ */
+ DEBUG("rxq=%p: can't allocate a new mbuf",
+ (void *)rxq);
+ /* Increment out of memory counters. */
+ ++rxq->stats.rx_nombuf;
+ ++rxq->priv->dev->data->rx_mbuf_alloc_failed;
+ goto repost;
+ }
+
+ /* Reconfigure sge to use rep instead of seg. */
+ elt->sge.addr = (uintptr_t)rep->buf_addr + RTE_PKTMBUF_HEADROOM;
+ assert(elt->sge.lkey == rxq->mr->lkey);
+ elt->buf = rep;
+
+ /* Add SGE to array for repost. */
+ sges[i] = elt->sge;
+
+ /* Update seg information. */
+ SET_DATA_OFF(seg, RTE_PKTMBUF_HEADROOM);
+ NB_SEGS(seg) = 1;
+ PORT(seg) = rxq->port_id;
+ NEXT(seg) = NULL;
+ PKT_LEN(seg) = len;
+ DATA_LEN(seg) = len;
+ if (rxq->csum | rxq->csum_l2tun | rxq->vlan_strip) {
+ seg->packet_type = rxq_cq_to_pkt_type(flags);
+ seg->ol_flags = rxq_cq_to_ol_flags(rxq, flags);
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ if (flags & IBV_EXP_CQ_RX_CVLAN_STRIPPED_V1) {
+ seg->ol_flags |= PKT_RX_VLAN_PKT;
+ seg->vlan_tci = vlan_tci;
+ }
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ }
+ /* Return packet. */
+ *(pkts++) = seg;
+ ++pkts_ret;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment bytes counter. */
+ rxq->stats.ibytes += len;
+#endif
+repost:
+ if (++elts_head >= elts_n)
+ elts_head = 0;
+ continue;
+ }
+ if (unlikely(i == 0))
+ return 0;
+ /* Repost WRs. */
+#ifdef DEBUG_RECV
+ DEBUG("%p: reposting %u WRs", (void *)rxq, i);
+#endif
+ ret = rxq->recv(rxq->wq, sges, i);
+ if (unlikely(ret)) {
+ /* Inability to repost WRs is fatal. */
+ DEBUG("%p: recv_burst(): failed (ret=%d)",
+ (void *)rxq->priv,
+ ret);
+ abort();
+ }
+ rxq->elts_head = elts_head;
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ /* Increment packets counter. */
+ rxq->stats.ipackets += pkts_ret;
+#endif
+ return pkts_ret;
+}
+
+/**
+ * Dummy DPDK callback for TX.
+ *
+ * This function is used to temporarily replace the real callback during
+ * unsafe control operations on the queue, or in case of error.
+ *
+ * @param dpdk_txq
+ * Generic pointer to TX queue structure.
+ * @param[in] pkts
+ * Packets to transmit.
+ * @param pkts_n
+ * Number of packets in array.
+ *
+ * @return
+ * Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+removed_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ (void)dpdk_txq;
+ (void)pkts;
+ (void)pkts_n;
+ return 0;
+}
+
+/**
+ * Dummy DPDK callback for RX.
+ *
+ * This function is used to temporarily replace the real callback during
+ * unsafe control operations on the queue, or in case of error.
+ *
+ * @param dpdk_rxq
+ * Generic pointer to RX queue structure.
+ * @param[out] pkts
+ * Array to store received packets.
+ * @param pkts_n
+ * Maximum number of packets in array.
+ *
+ * @return
+ * Number of packets successfully received (<= pkts_n).
+ */
+uint16_t
+removed_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
+{
+ (void)dpdk_rxq;
+ (void)pkts;
+ (void)pkts_n;
+ return 0;
+}
diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
new file mode 100644
index 00000000..0e2b607d
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -0,0 +1,352 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef RTE_PMD_MLX5_RXTX_H_
+#define RTE_PMD_MLX5_RXTX_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_mbuf.h>
+#include <rte_mempool.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5_utils.h"
+#include "mlx5.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+struct mlx5_rxq_stats {
+ unsigned int idx; /**< Mapping index. */
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ uint64_t ipackets; /**< Total of successfully received packets. */
+ uint64_t ibytes; /**< Total of successfully received bytes. */
+#endif
+ uint64_t idropped; /**< Total of packets dropped when RX ring full. */
+ uint64_t rx_nombuf; /**< Total of RX mbuf allocation failures. */
+};
+
+struct mlx5_txq_stats {
+ unsigned int idx; /**< Mapping index. */
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ uint64_t opackets; /**< Total of successfully sent packets. */
+ uint64_t obytes; /**< Total of successfully sent bytes. */
+#endif
+ uint64_t odropped; /**< Total of packets not sent when TX ring full. */
+};
+
+/* RX element (scattered packets). */
+struct rxq_elt_sp {
+ struct ibv_sge sges[MLX5_PMD_SGE_WR_N]; /* Scatter/Gather Elements. */
+ struct rte_mbuf *bufs[MLX5_PMD_SGE_WR_N]; /* SGEs buffers. */
+};
+
+/* RX element. */
+struct rxq_elt {
+ struct ibv_sge sge; /* Scatter/Gather Element. */
+ struct rte_mbuf *buf; /* SGE buffer. */
+};
+
+/* Flow director queue structure. */
+struct fdir_queue {
+ struct ibv_qp *qp; /* Associated RX QP. */
+ struct ibv_exp_rwq_ind_table *ind_table; /* Indirection table. */
+};
+
+struct priv;
+
+/* RX queue descriptor. */
+struct rxq {
+ struct priv *priv; /* Back pointer to private data. */
+ struct rte_mempool *mp; /* Memory Pool for allocations. */
+ struct ibv_cq *cq; /* Completion Queue. */
+ struct ibv_exp_wq *wq; /* Work Queue. */
+ int32_t (*poll)(); /* Verbs poll function. */
+ int32_t (*recv)(); /* Verbs receive function. */
+ unsigned int port_id; /* Port ID for incoming packets. */
+ unsigned int elts_n; /* (*elts)[] length. */
+ unsigned int elts_head; /* Current index in (*elts)[]. */
+ unsigned int sp:1; /* Use scattered RX elements. */
+ unsigned int csum:1; /* Enable checksum offloading. */
+ unsigned int csum_l2tun:1; /* Same for L2 tunnels. */
+ unsigned int vlan_strip:1; /* Enable VLAN stripping. */
+ unsigned int crc_present:1; /* CRC must be subtracted. */
+ union {
+ struct rxq_elt_sp (*sp)[]; /* Scattered RX elements. */
+ struct rxq_elt (*no_sp)[]; /* RX elements. */
+ } elts;
+ uint32_t mb_len; /* Length of a mp-issued mbuf. */
+ unsigned int socket; /* CPU socket ID for allocations. */
+ struct mlx5_rxq_stats stats; /* RX queue counters. */
+ struct ibv_exp_res_domain *rd; /* Resource Domain. */
+ struct fdir_queue fdir_queue; /* Flow director queue. */
+ struct ibv_mr *mr; /* Memory Region (for mp). */
+ struct ibv_exp_wq_family *if_wq; /* WQ burst interface. */
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ struct ibv_exp_cq_family_v1 *if_cq; /* CQ interface. */
+#else /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+ struct ibv_exp_cq_family *if_cq; /* CQ interface. */
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+};
+
+/* Hash RX queue types. */
+enum hash_rxq_type {
+ HASH_RXQ_TCPV4,
+ HASH_RXQ_UDPV4,
+ HASH_RXQ_IPV4,
+#ifdef HAVE_FLOW_SPEC_IPV6
+ HASH_RXQ_TCPV6,
+ HASH_RXQ_UDPV6,
+ HASH_RXQ_IPV6,
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ HASH_RXQ_ETH,
+};
+
+/* Flow structure with Ethernet specification. It is packed to prevent padding
+ * between attr and spec as this layout is expected by libibverbs. */
+struct flow_attr_spec_eth {
+ struct ibv_exp_flow_attr attr;
+ struct ibv_exp_flow_spec_eth spec;
+} __attribute__((packed));
+
+/* Define a struct flow_attr_spec_eth object as an array of at least
+ * "size" bytes. Room after the first index is normally used to store
+ * extra flow specifications. */
+#define FLOW_ATTR_SPEC_ETH(name, size) \
+ struct flow_attr_spec_eth name \
+ [((size) / sizeof(struct flow_attr_spec_eth)) + \
+ !!((size) % sizeof(struct flow_attr_spec_eth))]
+
+/* Initialization data for hash RX queue. */
+struct hash_rxq_init {
+ uint64_t hash_fields; /* Fields that participate in the hash. */
+ uint64_t dpdk_rss_hf; /* Matching DPDK RSS hash fields. */
+ unsigned int flow_priority; /* Flow priority to use. */
+ union {
+ struct {
+ enum ibv_exp_flow_spec_type type;
+ uint16_t size;
+ } hdr;
+ struct ibv_exp_flow_spec_tcp_udp tcp_udp;
+ struct ibv_exp_flow_spec_ipv4 ipv4;
+#ifdef HAVE_FLOW_SPEC_IPV6
+ struct ibv_exp_flow_spec_ipv6 ipv6;
+#endif /* HAVE_FLOW_SPEC_IPV6 */
+ struct ibv_exp_flow_spec_eth eth;
+ } flow_spec; /* Flow specification template. */
+ const struct hash_rxq_init *underlayer; /* Pointer to underlayer. */
+};
+
+/* Initialization data for indirection table. */
+struct ind_table_init {
+ unsigned int max_size; /* Maximum number of WQs. */
+ /* Hash RX queues using this table. */
+ unsigned int hash_types;
+ unsigned int hash_types_n;
+};
+
+/* Initialization data for special flows. */
+struct special_flow_init {
+ uint8_t dst_mac_val[6];
+ uint8_t dst_mac_mask[6];
+ unsigned int hash_types;
+ unsigned int per_vlan:1;
+};
+
+enum hash_rxq_flow_type {
+ HASH_RXQ_FLOW_TYPE_PROMISC,
+ HASH_RXQ_FLOW_TYPE_ALLMULTI,
+ HASH_RXQ_FLOW_TYPE_BROADCAST,
+ HASH_RXQ_FLOW_TYPE_IPV6MULTI,
+ HASH_RXQ_FLOW_TYPE_MAC,
+};
+
+#ifndef NDEBUG
+static inline const char *
+hash_rxq_flow_type_str(enum hash_rxq_flow_type flow_type)
+{
+ switch (flow_type) {
+ case HASH_RXQ_FLOW_TYPE_PROMISC:
+ return "promiscuous";
+ case HASH_RXQ_FLOW_TYPE_ALLMULTI:
+ return "allmulticast";
+ case HASH_RXQ_FLOW_TYPE_BROADCAST:
+ return "broadcast";
+ case HASH_RXQ_FLOW_TYPE_IPV6MULTI:
+ return "IPv6 multicast";
+ case HASH_RXQ_FLOW_TYPE_MAC:
+ return "MAC";
+ }
+ return NULL;
+}
+#endif /* NDEBUG */
+
+struct hash_rxq {
+ struct priv *priv; /* Back pointer to private data. */
+ struct ibv_qp *qp; /* Hash RX QP. */
+ enum hash_rxq_type type; /* Hash RX queue type. */
+ /* MAC flow steering rules, one per VLAN ID. */
+ struct ibv_exp_flow *mac_flow[MLX5_MAX_MAC_ADDRESSES][MLX5_MAX_VLAN_IDS];
+ struct ibv_exp_flow *special_flow
+ [MLX5_MAX_SPECIAL_FLOWS][MLX5_MAX_VLAN_IDS];
+};
+
+/* TX element. */
+struct txq_elt {
+ struct rte_mbuf *buf;
+};
+
+/* Linear buffer type. It is used when transmitting buffers with too many
+ * segments that do not fit the hardware queue (see max_send_sge).
+ * Extra segments are copied (linearized) in such buffers, replacing the
+ * last SGE during TX.
+ * The size is arbitrary but large enough to hold a jumbo frame with
+ * 8 segments considering mbuf.buf_len is about 2048 bytes. */
+typedef uint8_t linear_t[16384];
+
+/* TX queue descriptor. */
+struct txq {
+ struct priv *priv; /* Back pointer to private data. */
+ int32_t (*poll_cnt)(struct ibv_cq *cq, uint32_t max);
+ int (*send_pending)();
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ int (*send_pending_vlan)();
+#endif
+#if MLX5_PMD_MAX_INLINE > 0
+ int (*send_pending_inline)();
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ int (*send_pending_inline_vlan)();
+#endif
+#endif
+#if MLX5_PMD_SGE_WR_N > 1
+ int (*send_pending_sg_list)();
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ int (*send_pending_sg_list_vlan)();
+#endif
+#endif
+ int (*send_flush)(struct ibv_qp *qp);
+ struct ibv_cq *cq; /* Completion Queue. */
+ struct ibv_qp *qp; /* Queue Pair. */
+ struct txq_elt (*elts)[]; /* TX elements. */
+#if MLX5_PMD_MAX_INLINE > 0
+ uint32_t max_inline; /* Max inline send size <= MLX5_PMD_MAX_INLINE. */
+#endif
+ unsigned int elts_n; /* (*elts)[] length. */
+ unsigned int elts_head; /* Current index in (*elts)[]. */
+ unsigned int elts_tail; /* First element awaiting completion. */
+ unsigned int elts_comp; /* Number of completion requests. */
+ unsigned int elts_comp_cd; /* Countdown for next completion request. */
+ unsigned int elts_comp_cd_init; /* Initial value for countdown. */
+ struct {
+ const struct rte_mempool *mp; /* Cached Memory Pool. */
+ struct ibv_mr *mr; /* Memory Region (for mp). */
+ uint32_t lkey; /* mr->lkey */
+ } mp2mr[MLX5_PMD_TX_MP_CACHE]; /* MP to MR translation table. */
+ struct mlx5_txq_stats stats; /* TX queue counters. */
+ /* Elements used only for init part are here. */
+ linear_t (*elts_linear)[]; /* Linearized buffers. */
+ struct ibv_mr *mr_linear; /* Memory Region for linearized buffers. */
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ struct ibv_exp_qp_burst_family_v1 *if_qp; /* QP burst interface. */
+#else
+ struct ibv_exp_qp_burst_family *if_qp; /* QP burst interface. */
+#endif
+ struct ibv_exp_cq_family *if_cq; /* CQ interface. */
+ struct ibv_exp_res_domain *rd; /* Resource Domain. */
+ unsigned int socket; /* CPU socket ID for allocations. */
+};
+
+/* mlx5_rxq.c */
+
+extern const struct hash_rxq_init hash_rxq_init[];
+extern const unsigned int hash_rxq_init_n;
+
+extern uint8_t rss_hash_default_key[];
+extern const size_t rss_hash_default_key_len;
+
+size_t priv_flow_attr(struct priv *, struct ibv_exp_flow_attr *,
+ size_t, enum hash_rxq_type);
+int priv_create_hash_rxqs(struct priv *);
+void priv_destroy_hash_rxqs(struct priv *);
+int priv_allow_flow_type(struct priv *, enum hash_rxq_flow_type);
+int priv_rehash_flows(struct priv *);
+void rxq_cleanup(struct rxq *);
+int rxq_rehash(struct rte_eth_dev *, struct rxq *);
+int rxq_setup(struct rte_eth_dev *, struct rxq *, uint16_t, unsigned int,
+ const struct rte_eth_rxconf *, struct rte_mempool *);
+int mlx5_rx_queue_setup(struct rte_eth_dev *, uint16_t, uint16_t, unsigned int,
+ const struct rte_eth_rxconf *, struct rte_mempool *);
+void mlx5_rx_queue_release(void *);
+uint16_t mlx5_rx_burst_secondary_setup(void *dpdk_rxq, struct rte_mbuf **pkts,
+ uint16_t pkts_n);
+
+
+/* mlx5_txq.c */
+
+void txq_cleanup(struct txq *);
+int txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc,
+ unsigned int socket, const struct rte_eth_txconf *conf);
+
+int mlx5_tx_queue_setup(struct rte_eth_dev *, uint16_t, uint16_t, unsigned int,
+ const struct rte_eth_txconf *);
+void mlx5_tx_queue_release(void *);
+uint16_t mlx5_tx_burst_secondary_setup(void *dpdk_txq, struct rte_mbuf **pkts,
+ uint16_t pkts_n);
+
+/* mlx5_rxtx.c */
+
+struct ibv_mr *mlx5_mp2mr(struct ibv_pd *, const struct rte_mempool *);
+void txq_mp2mr_iter(const struct rte_mempool *, void *);
+uint16_t mlx5_tx_burst(void *, struct rte_mbuf **, uint16_t);
+uint16_t mlx5_rx_burst_sp(void *, struct rte_mbuf **, uint16_t);
+uint16_t mlx5_rx_burst(void *, struct rte_mbuf **, uint16_t);
+uint16_t removed_tx_burst(void *, struct rte_mbuf **, uint16_t);
+uint16_t removed_rx_burst(void *, struct rte_mbuf **, uint16_t);
+
+#endif /* RTE_PMD_MLX5_RXTX_H_ */
diff --git a/drivers/net/mlx5/mlx5_stats.c b/drivers/net/mlx5/mlx5_stats.c
new file mode 100644
index 00000000..2d3cb519
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_stats.c
@@ -0,0 +1,144 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ethdev.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_defs.h"
+
+/**
+ * DPDK callback to get device statistics.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param[out] stats
+ * Stats structure output buffer.
+ */
+void
+mlx5_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ struct rte_eth_stats tmp = {0};
+ unsigned int i;
+ unsigned int idx;
+
+ priv_lock(priv);
+ /* Add software counters. */
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ struct rxq *rxq = (*priv->rxqs)[i];
+
+ if (rxq == NULL)
+ continue;
+ idx = rxq->stats.idx;
+ if (idx < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ tmp.q_ipackets[idx] += rxq->stats.ipackets;
+ tmp.q_ibytes[idx] += rxq->stats.ibytes;
+#endif
+ tmp.q_errors[idx] += (rxq->stats.idropped +
+ rxq->stats.rx_nombuf);
+ }
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ tmp.ipackets += rxq->stats.ipackets;
+ tmp.ibytes += rxq->stats.ibytes;
+#endif
+ tmp.ierrors += rxq->stats.idropped;
+ tmp.rx_nombuf += rxq->stats.rx_nombuf;
+ }
+ for (i = 0; (i != priv->txqs_n); ++i) {
+ struct txq *txq = (*priv->txqs)[i];
+
+ if (txq == NULL)
+ continue;
+ idx = txq->stats.idx;
+ if (idx < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ tmp.q_opackets[idx] += txq->stats.opackets;
+ tmp.q_obytes[idx] += txq->stats.obytes;
+#endif
+ tmp.q_errors[idx] += txq->stats.odropped;
+ }
+#ifdef MLX5_PMD_SOFT_COUNTERS
+ tmp.opackets += txq->stats.opackets;
+ tmp.obytes += txq->stats.obytes;
+#endif
+ tmp.oerrors += txq->stats.odropped;
+ }
+#ifndef MLX5_PMD_SOFT_COUNTERS
+ /* FIXME: retrieve and add hardware counters. */
+#endif
+ *stats = tmp;
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback to clear device statistics.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_stats_reset(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ unsigned int i;
+ unsigned int idx;
+
+ priv_lock(priv);
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ if ((*priv->rxqs)[i] == NULL)
+ continue;
+ idx = (*priv->rxqs)[i]->stats.idx;
+ (*priv->rxqs)[i]->stats =
+ (struct mlx5_rxq_stats){ .idx = idx };
+ }
+ for (i = 0; (i != priv->txqs_n); ++i) {
+ if ((*priv->txqs)[i] == NULL)
+ continue;
+ idx = (*priv->txqs)[i]->stats.idx;
+ (*priv->txqs)[i]->stats =
+ (struct mlx5_txq_stats){ .idx = idx };
+ }
+#ifndef MLX5_PMD_SOFT_COUNTERS
+ /* FIXME: reset hardware counters. */
+#endif
+ priv_unlock(priv);
+}
diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
new file mode 100644
index 00000000..e9b9a293
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -0,0 +1,126 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_interrupts.h>
+#include <rte_alarm.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_utils.h"
+
+/**
+ * DPDK callback to start the device.
+ *
+ * Simulate device start by attaching all configured flows.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_dev_start(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+ int err;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ priv_lock(priv);
+ if (priv->started) {
+ priv_unlock(priv);
+ return 0;
+ }
+ DEBUG("%p: allocating and configuring hash RX queues", (void *)dev);
+ err = priv_create_hash_rxqs(priv);
+ if (!err)
+ err = priv_rehash_flows(priv);
+ if (!err)
+ priv->started = 1;
+ else {
+ ERROR("%p: an error occurred while configuring hash RX queues:"
+ " %s",
+ (void *)priv, strerror(err));
+ /* Rollback. */
+ priv_special_flow_disable_all(priv);
+ priv_mac_addrs_disable(priv);
+ priv_destroy_hash_rxqs(priv);
+ }
+ if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE)
+ priv_fdir_enable(priv);
+ priv_dev_interrupt_handler_install(priv, dev);
+ priv_unlock(priv);
+ return -err;
+}
+
+/**
+ * DPDK callback to stop the device.
+ *
+ * Simulate device stop by detaching all configured flows.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ */
+void
+mlx5_dev_stop(struct rte_eth_dev *dev)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ if (mlx5_is_secondary())
+ return;
+
+ priv_lock(priv);
+ if (!priv->started) {
+ priv_unlock(priv);
+ return;
+ }
+ DEBUG("%p: cleaning up and destroying hash RX queues", (void *)dev);
+ priv_special_flow_disable_all(priv);
+ priv_mac_addrs_disable(priv);
+ priv_destroy_hash_rxqs(priv);
+ priv_fdir_disable(priv);
+ priv_dev_interrupt_handler_uninstall(priv, dev);
+ priv->started = 0;
+ priv_unlock(priv);
+}
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
new file mode 100644
index 00000000..31ce53ad
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_txq.c
@@ -0,0 +1,605 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <assert.h>
+#include <errno.h>
+#include <string.h>
+#include <stdint.h>
+
+/* Verbs header. */
+/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <infiniband/verbs.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_mbuf.h>
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5_utils.h"
+#include "mlx5.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_autoconf.h"
+#include "mlx5_defs.h"
+
+/**
+ * Allocate TX queue elements.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param elts_n
+ * Number of elements to allocate.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+txq_alloc_elts(struct txq *txq, unsigned int elts_n)
+{
+ unsigned int i;
+ struct txq_elt (*elts)[elts_n] =
+ rte_calloc_socket("TXQ", 1, sizeof(*elts), 0, txq->socket);
+ linear_t (*elts_linear)[elts_n] =
+ rte_calloc_socket("TXQ", 1, sizeof(*elts_linear), 0,
+ txq->socket);
+ struct ibv_mr *mr_linear = NULL;
+ int ret = 0;
+
+ if ((elts == NULL) || (elts_linear == NULL)) {
+ ERROR("%p: can't allocate packets array", (void *)txq);
+ ret = ENOMEM;
+ goto error;
+ }
+ mr_linear =
+ ibv_reg_mr(txq->priv->pd, elts_linear, sizeof(*elts_linear),
+ (IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_REMOTE_WRITE));
+ if (mr_linear == NULL) {
+ ERROR("%p: unable to configure MR, ibv_reg_mr() failed",
+ (void *)txq);
+ ret = EINVAL;
+ goto error;
+ }
+ for (i = 0; (i != elts_n); ++i) {
+ struct txq_elt *elt = &(*elts)[i];
+
+ elt->buf = NULL;
+ }
+ DEBUG("%p: allocated and configured %u WRs", (void *)txq, elts_n);
+ txq->elts_n = elts_n;
+ txq->elts = elts;
+ txq->elts_head = 0;
+ txq->elts_tail = 0;
+ txq->elts_comp = 0;
+ /* Request send completion every MLX5_PMD_TX_PER_COMP_REQ packets or
+ * at least 4 times per ring. */
+ txq->elts_comp_cd_init =
+ ((MLX5_PMD_TX_PER_COMP_REQ < (elts_n / 4)) ?
+ MLX5_PMD_TX_PER_COMP_REQ : (elts_n / 4));
+ txq->elts_comp_cd = txq->elts_comp_cd_init;
+ txq->elts_linear = elts_linear;
+ txq->mr_linear = mr_linear;
+ assert(ret == 0);
+ return 0;
+error:
+ if (mr_linear != NULL)
+ claim_zero(ibv_dereg_mr(mr_linear));
+
+ rte_free(elts_linear);
+ rte_free(elts);
+
+ DEBUG("%p: failed, freed everything", (void *)txq);
+ assert(ret > 0);
+ return ret;
+}
+
+/**
+ * Free TX queue elements.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ */
+static void
+txq_free_elts(struct txq *txq)
+{
+ unsigned int elts_n = txq->elts_n;
+ unsigned int elts_head = txq->elts_head;
+ unsigned int elts_tail = txq->elts_tail;
+ struct txq_elt (*elts)[elts_n] = txq->elts;
+ linear_t (*elts_linear)[elts_n] = txq->elts_linear;
+ struct ibv_mr *mr_linear = txq->mr_linear;
+
+ DEBUG("%p: freeing WRs", (void *)txq);
+ txq->elts_n = 0;
+ txq->elts_head = 0;
+ txq->elts_tail = 0;
+ txq->elts_comp = 0;
+ txq->elts_comp_cd = 0;
+ txq->elts_comp_cd_init = 0;
+ txq->elts = NULL;
+ txq->elts_linear = NULL;
+ txq->mr_linear = NULL;
+ if (mr_linear != NULL)
+ claim_zero(ibv_dereg_mr(mr_linear));
+
+ rte_free(elts_linear);
+ if (elts == NULL)
+ return;
+ while (elts_tail != elts_head) {
+ struct txq_elt *elt = &(*elts)[elts_tail];
+
+ assert(elt->buf != NULL);
+ rte_pktmbuf_free(elt->buf);
+#ifndef NDEBUG
+ /* Poisoning. */
+ memset(elt, 0x77, sizeof(*elt));
+#endif
+ if (++elts_tail == elts_n)
+ elts_tail = 0;
+ }
+ rte_free(elts);
+}
+
+/**
+ * Clean up a TX queue.
+ *
+ * Destroy objects, free allocated memory and reset the structure for reuse.
+ *
+ * @param txq
+ * Pointer to TX queue structure.
+ */
+void
+txq_cleanup(struct txq *txq)
+{
+ struct ibv_exp_release_intf_params params;
+ size_t i;
+
+ DEBUG("cleaning up %p", (void *)txq);
+ txq_free_elts(txq);
+ txq->poll_cnt = NULL;
+#if MLX5_PMD_MAX_INLINE > 0
+ txq->send_pending_inline = NULL;
+#endif
+ txq->send_flush = NULL;
+ if (txq->if_qp != NULL) {
+ assert(txq->priv != NULL);
+ assert(txq->priv->ctx != NULL);
+ assert(txq->qp != NULL);
+ params = (struct ibv_exp_release_intf_params){
+ .comp_mask = 0,
+ };
+ claim_zero(ibv_exp_release_intf(txq->priv->ctx,
+ txq->if_qp,
+ &params));
+ }
+ if (txq->if_cq != NULL) {
+ assert(txq->priv != NULL);
+ assert(txq->priv->ctx != NULL);
+ assert(txq->cq != NULL);
+ params = (struct ibv_exp_release_intf_params){
+ .comp_mask = 0,
+ };
+ claim_zero(ibv_exp_release_intf(txq->priv->ctx,
+ txq->if_cq,
+ &params));
+ }
+ if (txq->qp != NULL)
+ claim_zero(ibv_destroy_qp(txq->qp));
+ if (txq->cq != NULL)
+ claim_zero(ibv_destroy_cq(txq->cq));
+ if (txq->rd != NULL) {
+ struct ibv_exp_destroy_res_domain_attr attr = {
+ .comp_mask = 0,
+ };
+
+ assert(txq->priv != NULL);
+ assert(txq->priv->ctx != NULL);
+ claim_zero(ibv_exp_destroy_res_domain(txq->priv->ctx,
+ txq->rd,
+ &attr));
+ }
+ for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) {
+ if (txq->mp2mr[i].mp == NULL)
+ break;
+ assert(txq->mp2mr[i].mr != NULL);
+ claim_zero(ibv_dereg_mr(txq->mp2mr[i].mr));
+ }
+ memset(txq, 0, sizeof(*txq));
+}
+
+/**
+ * Configure a TX queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param txq
+ * Pointer to TX queue structure.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
+ * @param[in] conf
+ * Thresholds parameters.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc,
+ unsigned int socket, const struct rte_eth_txconf *conf)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ struct txq tmpl = {
+ .priv = priv,
+ .socket = socket
+ };
+ union {
+ struct ibv_exp_query_intf_params params;
+ struct ibv_exp_qp_init_attr init;
+ struct ibv_exp_res_domain_init_attr rd;
+ struct ibv_exp_cq_init_attr cq;
+ struct ibv_exp_qp_attr mod;
+ } attr;
+ enum ibv_exp_query_intf_status status;
+ int ret = 0;
+
+ (void)conf; /* Thresholds configuration (ignored). */
+ if ((desc == 0) || (desc % MLX5_PMD_SGE_WR_N)) {
+ ERROR("%p: invalid number of TX descriptors (must be a"
+ " multiple of %d)", (void *)dev, MLX5_PMD_SGE_WR_N);
+ return EINVAL;
+ }
+ desc /= MLX5_PMD_SGE_WR_N;
+ /* MRs will be registered in mp2mr[] later. */
+ attr.rd = (struct ibv_exp_res_domain_init_attr){
+ .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL |
+ IBV_EXP_RES_DOMAIN_MSG_MODEL),
+ .thread_model = IBV_EXP_THREAD_SINGLE,
+ .msg_model = IBV_EXP_MSG_HIGH_BW,
+ };
+ tmpl.rd = ibv_exp_create_res_domain(priv->ctx, &attr.rd);
+ if (tmpl.rd == NULL) {
+ ret = ENOMEM;
+ ERROR("%p: RD creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.cq = (struct ibv_exp_cq_init_attr){
+ .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN,
+ .res_domain = tmpl.rd,
+ };
+ tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, NULL, 0, &attr.cq);
+ if (tmpl.cq == NULL) {
+ ret = ENOMEM;
+ ERROR("%p: CQ creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ DEBUG("priv->device_attr.max_qp_wr is %d",
+ priv->device_attr.max_qp_wr);
+ DEBUG("priv->device_attr.max_sge is %d",
+ priv->device_attr.max_sge);
+ attr.init = (struct ibv_exp_qp_init_attr){
+ /* CQ to be associated with the send queue. */
+ .send_cq = tmpl.cq,
+ /* CQ to be associated with the receive queue. */
+ .recv_cq = tmpl.cq,
+ .cap = {
+ /* Max number of outstanding WRs. */
+ .max_send_wr = ((priv->device_attr.max_qp_wr < desc) ?
+ priv->device_attr.max_qp_wr :
+ desc),
+ /* Max number of scatter/gather elements in a WR. */
+ .max_send_sge = ((priv->device_attr.max_sge <
+ MLX5_PMD_SGE_WR_N) ?
+ priv->device_attr.max_sge :
+ MLX5_PMD_SGE_WR_N),
+#if MLX5_PMD_MAX_INLINE > 0
+ .max_inline_data = MLX5_PMD_MAX_INLINE,
+#endif
+ },
+ .qp_type = IBV_QPT_RAW_PACKET,
+ /* Do *NOT* enable this, completions events are managed per
+ * TX burst. */
+ .sq_sig_all = 0,
+ .pd = priv->pd,
+ .res_domain = tmpl.rd,
+ .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD |
+ IBV_EXP_QP_INIT_ATTR_RES_DOMAIN),
+ };
+ tmpl.qp = ibv_exp_create_qp(priv->ctx, &attr.init);
+ if (tmpl.qp == NULL) {
+ ret = (errno ? errno : EINVAL);
+ ERROR("%p: QP creation failure: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+#if MLX5_PMD_MAX_INLINE > 0
+ /* ibv_create_qp() updates this value. */
+ tmpl.max_inline = attr.init.cap.max_inline_data;
+#endif
+ attr.mod = (struct ibv_exp_qp_attr){
+ /* Move the QP to this state. */
+ .qp_state = IBV_QPS_INIT,
+ /* Primary port number. */
+ .port_num = priv->port
+ };
+ ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod,
+ (IBV_EXP_QP_STATE | IBV_EXP_QP_PORT));
+ if (ret) {
+ ERROR("%p: QP state to IBV_QPS_INIT failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ ret = txq_alloc_elts(&tmpl, desc);
+ if (ret) {
+ ERROR("%p: TXQ allocation failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.mod = (struct ibv_exp_qp_attr){
+ .qp_state = IBV_QPS_RTR
+ };
+ ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod, IBV_EXP_QP_STATE);
+ if (ret) {
+ ERROR("%p: QP state to IBV_QPS_RTR failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.mod.qp_state = IBV_QPS_RTS;
+ ret = ibv_exp_modify_qp(tmpl.qp, &attr.mod, IBV_EXP_QP_STATE);
+ if (ret) {
+ ERROR("%p: QP state to IBV_QPS_RTS failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ attr.params = (struct ibv_exp_query_intf_params){
+ .intf_scope = IBV_EXP_INTF_GLOBAL,
+ .intf = IBV_EXP_INTF_CQ,
+ .obj = tmpl.cq,
+ };
+ tmpl.if_cq = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
+ if (tmpl.if_cq == NULL) {
+ ret = EINVAL;
+ ERROR("%p: CQ interface family query failed with status %d",
+ (void *)dev, status);
+ goto error;
+ }
+ attr.params = (struct ibv_exp_query_intf_params){
+ .intf_scope = IBV_EXP_INTF_GLOBAL,
+ .intf = IBV_EXP_INTF_QP_BURST,
+ .obj = tmpl.qp,
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ .intf_version = 1,
+#endif
+#ifdef HAVE_EXP_QP_BURST_CREATE_ENABLE_MULTI_PACKET_SEND_WR
+ /* Enable multi-packet send if supported. */
+ .family_flags =
+ (priv->mps ?
+ IBV_EXP_QP_BURST_CREATE_ENABLE_MULTI_PACKET_SEND_WR :
+ 0),
+#endif
+ };
+ tmpl.if_qp = ibv_exp_query_intf(priv->ctx, &attr.params, &status);
+ if (tmpl.if_qp == NULL) {
+ ret = EINVAL;
+ ERROR("%p: QP interface family query failed with status %d",
+ (void *)dev, status);
+ goto error;
+ }
+ /* Clean up txq in case we're reinitializing it. */
+ DEBUG("%p: cleaning-up old txq just in case", (void *)txq);
+ txq_cleanup(txq);
+ *txq = tmpl;
+ txq->poll_cnt = txq->if_cq->poll_cnt;
+#if MLX5_PMD_MAX_INLINE > 0
+ txq->send_pending_inline = txq->if_qp->send_pending_inline;
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ txq->send_pending_inline_vlan = txq->if_qp->send_pending_inline_vlan;
+#endif
+#endif
+#if MLX5_PMD_SGE_WR_N > 1
+ txq->send_pending_sg_list = txq->if_qp->send_pending_sg_list;
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ txq->send_pending_sg_list_vlan = txq->if_qp->send_pending_sg_list_vlan;
+#endif
+#endif
+ txq->send_pending = txq->if_qp->send_pending;
+#ifdef HAVE_VERBS_VLAN_INSERTION
+ txq->send_pending_vlan = txq->if_qp->send_pending_vlan;
+#endif
+ txq->send_flush = txq->if_qp->send_flush;
+ DEBUG("%p: txq updated with %p", (void *)txq, (void *)&tmpl);
+ /* Pre-register known mempools. */
+ rte_mempool_walk(txq_mp2mr_iter, txq);
+ assert(ret == 0);
+ return 0;
+error:
+ txq_cleanup(&tmpl);
+ assert(ret > 0);
+ return ret;
+}
+
+/**
+ * DPDK callback to configure a TX queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * TX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
+ * @param[in] conf
+ * Thresholds parameters.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ unsigned int socket, const struct rte_eth_txconf *conf)
+{
+ struct priv *priv = dev->data->dev_private;
+ struct txq *txq = (*priv->txqs)[idx];
+ int ret;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ priv_lock(priv);
+ DEBUG("%p: configuring queue %u for %u descriptors",
+ (void *)dev, idx, desc);
+ if (idx >= priv->txqs_n) {
+ ERROR("%p: queue index out of range (%u >= %u)",
+ (void *)dev, idx, priv->txqs_n);
+ priv_unlock(priv);
+ return -EOVERFLOW;
+ }
+ if (txq != NULL) {
+ DEBUG("%p: reusing already allocated queue index %u (%p)",
+ (void *)dev, idx, (void *)txq);
+ if (priv->started) {
+ priv_unlock(priv);
+ return -EEXIST;
+ }
+ (*priv->txqs)[idx] = NULL;
+ txq_cleanup(txq);
+ } else {
+ txq = rte_calloc_socket("TXQ", 1, sizeof(*txq), 0, socket);
+ if (txq == NULL) {
+ ERROR("%p: unable to allocate queue index %u",
+ (void *)dev, idx);
+ priv_unlock(priv);
+ return -ENOMEM;
+ }
+ }
+ ret = txq_setup(dev, txq, desc, socket, conf);
+ if (ret)
+ rte_free(txq);
+ else {
+ txq->stats.idx = idx;
+ DEBUG("%p: adding TX queue %p to list",
+ (void *)dev, (void *)txq);
+ (*priv->txqs)[idx] = txq;
+ /* Update send callback. */
+ dev->tx_pkt_burst = mlx5_tx_burst;
+ }
+ priv_unlock(priv);
+ return -ret;
+}
+
+/**
+ * DPDK callback to release a TX queue.
+ *
+ * @param dpdk_txq
+ * Generic TX queue pointer.
+ */
+void
+mlx5_tx_queue_release(void *dpdk_txq)
+{
+ struct txq *txq = (struct txq *)dpdk_txq;
+ struct priv *priv;
+ unsigned int i;
+
+ if (mlx5_is_secondary())
+ return;
+
+ if (txq == NULL)
+ return;
+ priv = txq->priv;
+ priv_lock(priv);
+ for (i = 0; (i != priv->txqs_n); ++i)
+ if ((*priv->txqs)[i] == txq) {
+ DEBUG("%p: removing TX queue %p from list",
+ (void *)priv->dev, (void *)txq);
+ (*priv->txqs)[i] = NULL;
+ break;
+ }
+ txq_cleanup(txq);
+ rte_free(txq);
+ priv_unlock(priv);
+}
+
+/**
+ * DPDK callback for TX in secondary processes.
+ *
+ * This function configures all queues from primary process information
+ * if necessary before reverting to the normal TX burst callback.
+ *
+ * @param dpdk_txq
+ * Generic pointer to TX queue structure.
+ * @param[in] pkts
+ * Packets to transmit.
+ * @param pkts_n
+ * Number of packets in array.
+ *
+ * @return
+ * Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+mlx5_tx_burst_secondary_setup(void *dpdk_txq, struct rte_mbuf **pkts,
+ uint16_t pkts_n)
+{
+ struct txq *txq = dpdk_txq;
+ struct priv *priv = mlx5_secondary_data_setup(txq->priv);
+ struct priv *primary_priv;
+ unsigned int index;
+
+ if (priv == NULL)
+ return 0;
+ primary_priv =
+ mlx5_secondary_data[priv->dev->data->port_id].primary_priv;
+ /* Look for queue index in both private structures. */
+ for (index = 0; index != priv->txqs_n; ++index)
+ if (((*primary_priv->txqs)[index] == txq) ||
+ ((*priv->txqs)[index] == txq))
+ break;
+ if (index == priv->txqs_n)
+ return 0;
+ txq = (*priv->txqs)[index];
+ return priv->dev->tx_pkt_burst(txq, pkts, pkts_n);
+}
diff --git a/drivers/net/mlx5/mlx5_utils.h b/drivers/net/mlx5/mlx5_utils.h
new file mode 100644
index 00000000..a824787f
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_utils.h
@@ -0,0 +1,184 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef RTE_PMD_MLX5_UTILS_H_
+#define RTE_PMD_MLX5_UTILS_H_
+
+#include <stddef.h>
+#include <stdio.h>
+#include <limits.h>
+#include <assert.h>
+#include <errno.h>
+
+#include "mlx5_defs.h"
+
+/* Bit-field manipulation. */
+#define BITFIELD_DECLARE(bf, type, size) \
+ type bf[(((size_t)(size) / (sizeof(type) * CHAR_BIT)) + \
+ !!((size_t)(size) % (sizeof(type) * CHAR_BIT)))]
+#define BITFIELD_DEFINE(bf, type, size) \
+ BITFIELD_DECLARE((bf), type, (size)) = { 0 }
+#define BITFIELD_SET(bf, b) \
+ (assert((size_t)(b) < (sizeof(bf) * CHAR_BIT)), \
+ (void)((bf)[((b) / (sizeof((bf)[0]) * CHAR_BIT))] |= \
+ ((size_t)1 << ((b) % (sizeof((bf)[0]) * CHAR_BIT)))))
+#define BITFIELD_RESET(bf, b) \
+ (assert((size_t)(b) < (sizeof(bf) * CHAR_BIT)), \
+ (void)((bf)[((b) / (sizeof((bf)[0]) * CHAR_BIT))] &= \
+ ~((size_t)1 << ((b) % (sizeof((bf)[0]) * CHAR_BIT)))))
+#define BITFIELD_ISSET(bf, b) \
+ (assert((size_t)(b) < (sizeof(bf) * CHAR_BIT)), \
+ !!(((bf)[((b) / (sizeof((bf)[0]) * CHAR_BIT))] & \
+ ((size_t)1 << ((b) % (sizeof((bf)[0]) * CHAR_BIT))))))
+
+/* Save and restore errno around argument evaluation. */
+#define ERRNO_SAFE(x) ((errno = (int []){ errno, ((x), 0) }[0]))
+
+/*
+ * Helper macros to work around __VA_ARGS__ limitations in a C99 compliant
+ * manner.
+ */
+#define PMD_DRV_LOG_STRIP(a, b) a
+#define PMD_DRV_LOG_OPAREN (
+#define PMD_DRV_LOG_CPAREN )
+#define PMD_DRV_LOG_COMMA ,
+
+/* Return the file name part of a path. */
+static inline const char *
+pmd_drv_log_basename(const char *s)
+{
+ const char *n = s;
+
+ while (*n)
+ if (*(n++) == '/')
+ s = n;
+ return s;
+}
+
+/*
+ * When debugging is enabled (NDEBUG not defined), file, line and function
+ * information replace the driver name (MLX5_DRIVER_NAME) in log messages.
+ */
+#ifndef NDEBUG
+
+#define PMD_DRV_LOG___(level, ...) \
+ ERRNO_SAFE(RTE_LOG(level, PMD, __VA_ARGS__))
+#define PMD_DRV_LOG__(level, ...) \
+ PMD_DRV_LOG___(level, "%s:%u: %s(): " __VA_ARGS__)
+#define PMD_DRV_LOG_(level, s, ...) \
+ PMD_DRV_LOG__(level, \
+ s "\n" PMD_DRV_LOG_COMMA \
+ pmd_drv_log_basename(__FILE__) PMD_DRV_LOG_COMMA \
+ __LINE__ PMD_DRV_LOG_COMMA \
+ __func__, \
+ __VA_ARGS__)
+
+#else /* NDEBUG */
+
+#define PMD_DRV_LOG___(level, ...) \
+ ERRNO_SAFE(RTE_LOG(level, PMD, MLX5_DRIVER_NAME ": " __VA_ARGS__))
+#define PMD_DRV_LOG__(level, ...) \
+ PMD_DRV_LOG___(level, __VA_ARGS__)
+#define PMD_DRV_LOG_(level, s, ...) \
+ PMD_DRV_LOG__(level, s "\n", __VA_ARGS__)
+
+#endif /* NDEBUG */
+
+/* Generic printf()-like logging macro with automatic line feed. */
+#define PMD_DRV_LOG(level, ...) \
+ PMD_DRV_LOG_(level, \
+ __VA_ARGS__ PMD_DRV_LOG_STRIP PMD_DRV_LOG_OPAREN, \
+ PMD_DRV_LOG_CPAREN)
+
+/*
+ * Like assert(), DEBUG() becomes a no-op and claim_zero() does not perform
+ * any check when debugging is disabled.
+ */
+#ifndef NDEBUG
+
+#define DEBUG(...) PMD_DRV_LOG(DEBUG, __VA_ARGS__)
+#define claim_zero(...) assert((__VA_ARGS__) == 0)
+
+#else /* NDEBUG */
+
+#define DEBUG(...) (void)0
+#define claim_zero(...) (__VA_ARGS__)
+
+#endif /* NDEBUG */
+
+#define INFO(...) PMD_DRV_LOG(INFO, __VA_ARGS__)
+#define WARN(...) PMD_DRV_LOG(WARNING, __VA_ARGS__)
+#define ERROR(...) PMD_DRV_LOG(ERR, __VA_ARGS__)
+
+/* Convenience macros for accessing mbuf fields. */
+#define NEXT(m) ((m)->next)
+#define DATA_LEN(m) ((m)->data_len)
+#define PKT_LEN(m) ((m)->pkt_len)
+#define DATA_OFF(m) ((m)->data_off)
+#define SET_DATA_OFF(m, o) ((m)->data_off = (o))
+#define NB_SEGS(m) ((m)->nb_segs)
+#define PORT(m) ((m)->port)
+
+/* Transpose flags. Useful to convert IBV to DPDK flags. */
+#define TRANSPOSE(val, from, to) \
+ (((from) >= (to)) ? \
+ (((val) & (from)) / ((from) / (to))) : \
+ (((val) & (from)) * ((to) / (from))))
+
+/* Allocate a buffer on the stack and fill it with a printf format string. */
+#define MKSTR(name, ...) \
+ char name[snprintf(NULL, 0, __VA_ARGS__) + 1]; \
+ \
+ snprintf(name, sizeof(name), __VA_ARGS__)
+
+/**
+ * Return nearest power of two above input value.
+ *
+ * @param v
+ * Input value.
+ *
+ * @return
+ * Nearest power of two above input value.
+ */
+static inline unsigned int
+log2above(unsigned int v)
+{
+ unsigned int l;
+ unsigned int r;
+
+ for (l = 0, r = 0; (v >> 1); ++l, v >>= 1)
+ r |= (v & 1);
+ return l + r;
+}
+
+#endif /* RTE_PMD_MLX5_UTILS_H_ */
diff --git a/drivers/net/mlx5/mlx5_vlan.c b/drivers/net/mlx5/mlx5_vlan.c
new file mode 100644
index 00000000..ea7af1e4
--- /dev/null
+++ b/drivers/net/mlx5/mlx5_vlan.c
@@ -0,0 +1,234 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of 6WIND S.A. nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stddef.h>
+#include <errno.h>
+#include <assert.h>
+#include <stdint.h>
+
+/* DPDK headers don't like -pedantic. */
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-pedantic"
+#endif
+#include <rte_ethdev.h>
+#include <rte_common.h>
+#ifdef PEDANTIC
+#pragma GCC diagnostic error "-pedantic"
+#endif
+
+#include "mlx5_utils.h"
+#include "mlx5.h"
+#include "mlx5_autoconf.h"
+
+/**
+ * Configure a VLAN filter.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param vlan_id
+ * VLAN ID to filter.
+ * @param on
+ * Toggle filter.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+ struct priv *priv = dev->data->dev_private;
+ unsigned int i;
+
+ DEBUG("%p: %s VLAN filter ID %" PRIu16,
+ (void *)dev, (on ? "enable" : "disable"), vlan_id);
+ assert(priv->vlan_filter_n <= RTE_DIM(priv->vlan_filter));
+ for (i = 0; (i != priv->vlan_filter_n); ++i)
+ if (priv->vlan_filter[i] == vlan_id)
+ break;
+ /* Check if there's room for another VLAN filter. */
+ if (i == RTE_DIM(priv->vlan_filter))
+ return ENOMEM;
+ if (i < priv->vlan_filter_n) {
+ assert(priv->vlan_filter_n != 0);
+ /* Enabling an existing VLAN filter has no effect. */
+ if (on)
+ return 0;
+ /* Remove VLAN filter from list. */
+ --priv->vlan_filter_n;
+ memmove(&priv->vlan_filter[i],
+ &priv->vlan_filter[i + 1],
+ priv->vlan_filter_n - i);
+ priv->vlan_filter[priv->vlan_filter_n] = 0;
+ } else {
+ assert(i == priv->vlan_filter_n);
+ /* Disabling an unknown VLAN filter has no effect. */
+ if (!on)
+ return 0;
+ /* Add new VLAN filter. */
+ priv->vlan_filter[priv->vlan_filter_n] = vlan_id;
+ ++priv->vlan_filter_n;
+ }
+ /* Rehash flows in all hash RX queues. */
+ priv_mac_addrs_disable(priv);
+ priv_special_flow_disable_all(priv);
+ return priv_rehash_flows(priv);
+}
+
+/**
+ * DPDK callback to configure a VLAN filter.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param vlan_id
+ * VLAN ID to filter.
+ * @param on
+ * Toggle filter.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+ struct priv *priv = dev->data->dev_private;
+ int ret;
+
+ priv_lock(priv);
+ ret = vlan_filter_set(dev, vlan_id, on);
+ priv_unlock(priv);
+ assert(ret >= 0);
+ return -ret;
+}
+
+/**
+ * Set/reset VLAN stripping for a specific queue.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param idx
+ * RX queue index.
+ * @param on
+ * Enable/disable VLAN stripping.
+ */
+static void
+priv_vlan_strip_queue_set(struct priv *priv, uint16_t idx, int on)
+{
+ struct rxq *rxq = (*priv->rxqs)[idx];
+#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
+ struct ibv_exp_wq_attr mod;
+ uint16_t vlan_offloads =
+ (on ? IBV_EXP_RECEIVE_WQ_CVLAN_STRIP : 0) |
+ 0;
+ int err;
+
+ DEBUG("set VLAN offloads 0x%x for port %d queue %d",
+ vlan_offloads, rxq->port_id, idx);
+ mod = (struct ibv_exp_wq_attr){
+ .attr_mask = IBV_EXP_WQ_ATTR_VLAN_OFFLOADS,
+ .vlan_offloads = vlan_offloads,
+ };
+
+ err = ibv_exp_modify_wq(rxq->wq, &mod);
+ if (err) {
+ ERROR("%p: failed to modified stripping mode: %s",
+ (void *)priv, strerror(err));
+ return;
+ }
+
+#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
+
+ /* Update related bits in RX queue. */
+ rxq->vlan_strip = !!on;
+}
+
+/**
+ * Callback to set/reset VLAN stripping for a specific queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param queue
+ * RX queue index.
+ * @param on
+ * Enable/disable VLAN stripping.
+ */
+void
+mlx5_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ /* Validate hw support */
+ if (!priv->hw_vlan_strip) {
+ ERROR("VLAN stripping is not supported");
+ return;
+ }
+
+ /* Validate queue number */
+ if (queue >= priv->rxqs_n) {
+ ERROR("VLAN stripping, invalid queue number %d", queue);
+ return;
+ }
+
+ priv_lock(priv);
+ priv_vlan_strip_queue_set(priv, queue, on);
+ priv_unlock(priv);
+}
+
+/**
+ * Callback to set/reset VLAN offloads for a port.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param mask
+ * VLAN offload bit mask.
+ */
+void
+mlx5_vlan_offload_set(struct rte_eth_dev *dev, int mask)
+{
+ struct priv *priv = dev->data->dev_private;
+ unsigned int i;
+
+ if (mask & ETH_VLAN_STRIP_MASK) {
+ int hw_vlan_strip = dev->data->dev_conf.rxmode.hw_vlan_strip;
+
+ if (!priv->hw_vlan_strip) {
+ ERROR("VLAN stripping is not supported");
+ return;
+ }
+
+ /* Run on every RX queue and set/reset VLAN stripping. */
+ priv_lock(priv);
+ for (i = 0; (i != priv->rxqs_n); i++)
+ priv_vlan_strip_queue_set(priv, i, hw_vlan_strip);
+ priv_unlock(priv);
+ }
+}
diff --git a/drivers/net/mlx5/rte_pmd_mlx5_version.map b/drivers/net/mlx5/rte_pmd_mlx5_version.map
new file mode 100644
index 00000000..ad607bbe
--- /dev/null
+++ b/drivers/net/mlx5/rte_pmd_mlx5_version.map
@@ -0,0 +1,3 @@
+DPDK_2.2 {
+ local: *;
+};