#include <net/if.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
+#include <sys/utsname.h>
#include <netinet/in.h>
-#include <linux/if.h>
#include <linux/ethtool.h>
#include <linux/sockios.h>
+#include <linux/version.h>
#include <fcntl.h>
/* DPDK headers don't like -pedantic. */
#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-pedantic"
+#pragma GCC diagnostic ignored "-Wpedantic"
#endif
#include <rte_atomic.h>
#include <rte_ethdev.h>
#include <rte_common.h>
#include <rte_interrupts.h>
#include <rte_alarm.h>
+#include <rte_malloc.h>
#ifdef PEDANTIC
-#pragma GCC diagnostic error "-pedantic"
+#pragma GCC diagnostic error "-Wpedantic"
#endif
#include "mlx5.h"
#include "mlx5_rxtx.h"
#include "mlx5_utils.h"
+/* Add defines in case the running kernel is not the same as user headers. */
+#ifndef ETHTOOL_GLINKSETTINGS
+struct ethtool_link_settings {
+ uint32_t cmd;
+ uint32_t speed;
+ uint8_t duplex;
+ uint8_t port;
+ uint8_t phy_address;
+ uint8_t autoneg;
+ uint8_t mdio_support;
+ uint8_t eth_to_mdix;
+ uint8_t eth_tp_mdix_ctrl;
+ int8_t link_mode_masks_nwords;
+ uint32_t reserved[8];
+ uint32_t link_mode_masks[];
+};
+
+#define ETHTOOL_GLINKSETTINGS 0x0000004c
+#define ETHTOOL_LINK_MODE_1000baseT_Full_BIT 5
+#define ETHTOOL_LINK_MODE_Autoneg_BIT 6
+#define ETHTOOL_LINK_MODE_1000baseKX_Full_BIT 17
+#define ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT 18
+#define ETHTOOL_LINK_MODE_10000baseKR_Full_BIT 19
+#define ETHTOOL_LINK_MODE_10000baseR_FEC_BIT 20
+#define ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT 21
+#define ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT 22
+#define ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT 23
+#define ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT 24
+#define ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT 25
+#define ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT 26
+#define ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT 27
+#define ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT 28
+#define ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT 29
+#define ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT 30
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_25G
+#define ETHTOOL_LINK_MODE_25000baseCR_Full_BIT 31
+#define ETHTOOL_LINK_MODE_25000baseKR_Full_BIT 32
+#define ETHTOOL_LINK_MODE_25000baseSR_Full_BIT 33
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_50G
+#define ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT 34
+#define ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT 35
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_100G
+#define ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT 36
+#define ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT 37
+#define ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT 38
+#define ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT 39
+#endif
+#define ETHTOOL_LINK_MODE_MASK_MAX_KERNEL_NU32 (SCHAR_MAX)
+
+/**
+ * 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.
*
return 0;
}
+/**
+ * Check if the counter is located on ib counters file.
+ *
+ * @param[in] cntr
+ * Counter name.
+ *
+ * @return
+ * 1 if counter is located on ib counters file , 0 otherwise.
+ */
+int
+priv_is_ib_cntr(const char *cntr)
+{
+ if (!strcmp(cntr, "out_of_buffer"))
+ return 1;
+ return 0;
+}
+
/**
* Read from sysfs entry.
*
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 (priv_is_ib_cntr(entry)) {
+ MKSTR(path, "%s/ports/1/hw_counters/%s",
+ priv->ctx->device->ibdev_path, entry);
+ file = fopen(path, "rb");
+ } else {
+ 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);
return ret;
}
+/**
+ * Return the number of active VFs for the current device.
+ *
+ * @param[in] priv
+ * Pointer to private structure.
+ * @param[out] num_vfs
+ * Number of active VFs.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_num_vfs(struct priv *priv, uint16_t *num_vfs)
+{
+ /* The sysfs entry name depends on the operating system. */
+ const char **name = (const char *[]){
+ "device/sriov_numvfs",
+ "device/mlx5_num_vfs",
+ NULL,
+ };
+ int ret;
+
+ do {
+ unsigned long ulong_num_vfs;
+
+ ret = priv_get_sysfs_ulong(priv, *name, &ulong_num_vfs);
+ if (!ret)
+ *num_vfs = ulong_num_vfs;
+ } while (*(++name) && ret);
+ return ret;
+}
+
/**
* Get device MTU.
*
return 0;
}
+/**
+ * Read device counter from sysfs.
+ *
+ * @param priv
+ * Pointer to private structure.
+ * @param name
+ * Counter name.
+ * @param[out] cntr
+ * Counter output buffer.
+ *
+ * @return
+ * 0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_cntr_sysfs(struct priv *priv, const char *name, uint64_t *cntr)
+{
+ unsigned long ulong_ctr;
+
+ if (priv_get_sysfs_ulong(priv, name, &ulong_ctr) == -1)
+ return -1;
+ *cntr = ulong_ctr;
+ return 0;
+}
+
/**
* Set device MTU.
*
static int
priv_set_mtu(struct priv *priv, uint16_t mtu)
{
- return priv_set_sysfs_ulong(priv, "mtu", mtu);
+ uint16_t new_mtu;
+
+ if (priv_set_sysfs_ulong(priv, "mtu", mtu) ||
+ priv_get_mtu(priv, &new_mtu))
+ return -1;
+ if (new_mtu == mtu)
+ return 0;
+ errno = EINVAL;
+ return -1;
}
/**
if (priv_get_sysfs_ulong(priv, "flags", &tmp) == -1)
return -1;
tmp &= keep;
- tmp |= flags;
+ tmp |= (flags & (~keep));
return priv_set_sysfs_ulong(priv, "flags", tmp);
}
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) {
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);
void
mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
{
- struct priv *priv = dev->data->dev_private;
+ struct priv *priv = mlx5_get_priv(dev);
unsigned int max;
char ifname[IF_NAMESIZE];
+ info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
priv_lock(priv);
/* FIXME: we should ask the device for these values. */
info->min_rx_bufsize = 32;
(DEV_RX_OFFLOAD_IPV4_CKSUM |
DEV_RX_OFFLOAD_UDP_CKSUM |
DEV_RX_OFFLOAD_TCP_CKSUM) :
- 0);
- info->tx_offload_capa =
- (priv->hw_csum ?
- (DEV_TX_OFFLOAD_IPV4_CKSUM |
- DEV_TX_OFFLOAD_UDP_CKSUM |
- DEV_TX_OFFLOAD_TCP_CKSUM) :
- 0);
+ 0) |
+ (priv->hw_vlan_strip ? DEV_RX_OFFLOAD_VLAN_STRIP : 0);
+ if (!priv->mps)
+ 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->tso)
+ info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO;
+ if (priv->tunnel_en)
+ info->tx_offload_capa |= (DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
+ DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
+ DEV_TX_OFFLOAD_GRE_TNL_TSO);
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->reta_size = priv->reta_idx_n ?
+ priv->reta_idx_n : priv->ind_table_max_size;
+ info->hash_key_size = ((*priv->rss_conf) ?
+ (*priv->rss_conf)[0]->rss_key_len :
+ 0);
+ info->speed_capa = priv->link_speed_capa;
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_EXT_UNKNOWN,
+ RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
+ RTE_PTYPE_UNKNOWN
+
+ };
+
+ if (dev->rx_pkt_burst == mlx5_rx_burst ||
+ dev->rx_pkt_burst == mlx5_rx_burst_vec)
+ return ptypes;
+ return NULL;
+}
+
/**
- * DPDK callback to retrieve physical link information (unlocked version).
+ * DPDK callback to retrieve physical link information.
*
* @param dev
* Pointer to Ethernet device structure.
* Wait for request completion (ignored).
*/
static int
-mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
+mlx5_link_update_unlocked_gset(struct rte_eth_dev *dev, int wait_to_complete)
{
- struct priv *priv = dev->data->dev_private;
+ struct priv *priv = mlx5_get_priv(dev);
struct ethtool_cmd edata = {
- .cmd = ETHTOOL_GSET
+ .cmd = ETHTOOL_GSET /* Deprecated since Linux v4.5. */
};
struct ifreq ifr;
struct rte_eth_link dev_link;
int link_speed = 0;
+ /* priv_lock() is not taken to allow concurrent calls. */
+
(void)wait_to_complete;
if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) {
WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno));
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;
+ ifr.ifr_data = (void *)&edata;
if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
WARN("ioctl(SIOCETHTOOL, ETHTOOL_GSET) failed: %s",
strerror(errno));
dev_link.link_speed = 0;
else
dev_link.link_speed = link_speed;
+ priv->link_speed_capa = 0;
+ if (edata.supported & SUPPORTED_Autoneg)
+ priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG;
+ if (edata.supported & (SUPPORTED_1000baseT_Full |
+ SUPPORTED_1000baseKX_Full))
+ priv->link_speed_capa |= ETH_LINK_SPEED_1G;
+ if (edata.supported & SUPPORTED_10000baseKR_Full)
+ priv->link_speed_capa |= ETH_LINK_SPEED_10G;
+ if (edata.supported & (SUPPORTED_40000baseKR4_Full |
+ SUPPORTED_40000baseCR4_Full |
+ SUPPORTED_40000baseSR4_Full |
+ SUPPORTED_40000baseLR4_Full))
+ priv->link_speed_capa |= ETH_LINK_SPEED_40G;
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;
+}
+
+/**
+ * Retrieve physical link information (unlocked version using new ioctl).
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param wait_to_complete
+ * Wait for request completion (ignored).
+ */
+static int
+mlx5_link_update_unlocked_gs(struct rte_eth_dev *dev, int wait_to_complete)
+{
+ struct priv *priv = mlx5_get_priv(dev);
+ __extension__ struct {
+ struct ethtool_link_settings edata;
+ uint32_t link_mode_data[3 *
+ ETHTOOL_LINK_MODE_MASK_MAX_KERNEL_NU32];
+ } ecmd;
+
+ struct ifreq ifr;
+ struct rte_eth_link dev_link;
+ uint64_t sc;
+
+ (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));
+ memset(&ecmd, 0, sizeof(ecmd));
+ ecmd.edata.cmd = ETHTOOL_GLINKSETTINGS;
+ ifr.ifr_data = (void *)&ecmd;
+ if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+ DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s",
+ strerror(errno));
+ return -1;
+ }
+ ecmd.edata.link_mode_masks_nwords = -ecmd.edata.link_mode_masks_nwords;
+ if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+ DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s",
+ strerror(errno));
+ return -1;
+ }
+ dev_link.link_speed = ecmd.edata.speed;
+ sc = ecmd.edata.link_mode_masks[0] |
+ ((uint64_t)ecmd.edata.link_mode_masks[1] << 32);
+ priv->link_speed_capa = 0;
+ if (sc & ETHTOOL_LINK_MODE_Autoneg_BIT)
+ priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG;
+ if (sc & (ETHTOOL_LINK_MODE_1000baseT_Full_BIT |
+ ETHTOOL_LINK_MODE_1000baseKX_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_1G;
+ if (sc & (ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT |
+ ETHTOOL_LINK_MODE_10000baseKR_Full_BIT |
+ ETHTOOL_LINK_MODE_10000baseR_FEC_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_10G;
+ if (sc & (ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT |
+ ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_20G;
+ if (sc & (ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT |
+ ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT |
+ ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT |
+ ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_40G;
+ if (sc & (ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT |
+ ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT |
+ ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT |
+ ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_56G;
+ if (sc & (ETHTOOL_LINK_MODE_25000baseCR_Full_BIT |
+ ETHTOOL_LINK_MODE_25000baseKR_Full_BIT |
+ ETHTOOL_LINK_MODE_25000baseSR_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_25G;
+ if (sc & (ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT |
+ ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_50G;
+ if (sc & (ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT |
+ ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT |
+ ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT |
+ ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT))
+ priv->link_speed_capa |= ETH_LINK_SPEED_100G;
+ dev_link.link_duplex = ((ecmd.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;
int
mlx5_link_update(struct rte_eth_dev *dev, int wait_to_complete)
{
- struct priv *priv = dev->data->dev_private;
- int ret;
-
- priv_lock(priv);
- ret = mlx5_link_update_unlocked(dev, wait_to_complete);
- priv_unlock(priv);
- return ret;
+ struct utsname utsname;
+ int ver[3];
+
+ if (uname(&utsname) == -1 ||
+ sscanf(utsname.release, "%d.%d.%d",
+ &ver[0], &ver[1], &ver[2]) != 3 ||
+ KERNEL_VERSION(ver[0], ver[1], ver[2]) < KERNEL_VERSION(4, 9, 0))
+ return mlx5_link_update_unlocked_gset(dev, wait_to_complete);
+ return mlx5_link_update_unlocked_gs(dev, wait_to_complete);
}
/**
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;
+ unsigned int max_frame_len;
+ int rehash;
+ int restart = priv->started;
+
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
priv_lock(priv);
/* Set kernel interface MTU first. */
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
+ /* Make sure everyone has left dev->rx_pkt_burst() and uses
* removed_rx_burst() instead. */
rte_wmb();
usleep(1000);
+ /* MTU does not include header and CRC. */
+ max_frame_len = ETHER_HDR_LEN + mtu + ETHER_CRC_LEN;
+ /* Check if at least one queue is going to need a SGE update. */
+ for (i = 0; i != priv->rxqs_n; ++i) {
+ struct rxq *rxq = (*priv->rxqs)[i];
+ unsigned int mb_len;
+ unsigned int size = RTE_PKTMBUF_HEADROOM + max_frame_len;
+ unsigned int sges_n;
+
+ if (rxq == NULL)
+ continue;
+ mb_len = rte_pktmbuf_data_room_size(rxq->mp);
+ assert(mb_len >= RTE_PKTMBUF_HEADROOM);
+ /*
+ * Determine the number of SGEs needed for a full packet
+ * and round it to the next power of two.
+ */
+ sges_n = log2above((size / mb_len) + !!(size % mb_len));
+ if (sges_n != rxq->sges_n)
+ break;
+ }
+ /*
+ * If all queues have the right number of SGEs, a simple rehash
+ * of their buffers is enough, otherwise SGE information can only
+ * be updated in a queue by recreating it. All resources that depend
+ * on queues (flows, indirection tables) must be recreated as well in
+ * that case.
+ */
+ rehash = (i == priv->rxqs_n);
+ if (!rehash) {
+ /* Clean up everything as with mlx5_dev_stop(). */
+ 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);
+ }
+recover:
/* 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;
+ struct rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct rxq_ctrl, rxq);
+ unsigned int mb_len;
+ unsigned int tmp;
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));
+ mb_len = rte_pktmbuf_data_room_size(rxq->mp);
+ assert(mb_len >= RTE_PKTMBUF_HEADROOM);
/* Provide new values to rxq_setup(). */
- dev->data->dev_conf.rxmode.jumbo_frame = sp;
+ dev->data->dev_conf.rxmode.jumbo_frame =
+ (max_frame_len > ETHER_MAX_LEN);
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;
+ if (rehash)
+ ret = rxq_rehash(dev, rxq_ctrl);
+ else
+ ret = rxq_ctrl_setup(dev, rxq_ctrl, 1 << rxq->elts_n,
+ rxq_ctrl->socket, NULL, rxq->mp);
+ if (!ret)
+ continue;
+ /* Attempt to roll back in case of error. */
+ tmp = (mb_len << rxq->sges_n) - RTE_PKTMBUF_HEADROOM;
+ if (max_frame_len != tmp) {
+ max_frame_len = tmp;
+ goto recover;
}
- /* Scattered burst function takes priority. */
- if (rxq->sp)
- rx_func = mlx5_rx_burst_sp;
+ /* Double fault, disable RX. */
+ break;
}
+ /* Mimic mlx5_dev_start(). */
+ if (ret) {
+ ERROR("unable to reconfigure RX queues, RX disabled");
+ } else if (restart &&
+ !rehash &&
+ !priv_create_hash_rxqs(priv) &&
+ !priv_rehash_flows(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->mtu = mtu;
/* Burst functions can now be called again. */
rte_wmb();
- dev->rx_pkt_burst = rx_func;
+ /*
+ * Use a safe RX burst function in case of error, otherwise select RX
+ * burst function again.
+ */
+ if (!ret)
+ priv_select_rx_function(priv);
out:
priv_unlock(priv);
assert(ret >= 0);
};
int ret;
- ifr.ifr_data = ðpause;
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ ifr.ifr_data = (void *)ðpause;
priv_lock(priv);
if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
ret = errno;
};
int ret;
- ifr.ifr_data = ðpause;
+ if (mlx5_is_secondary())
+ return -E_RTE_SECONDARY;
+
+ ifr.ifr_data = (void *)ðpause;
ethpause.autoneg = fc_conf->autoneg;
if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) ||
(fc_conf->mode & RTE_FC_RX_PAUSE))
/* Extract information. */
if (sscanf(line,
"PCI_SLOT_NAME="
- "%" SCNx16 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n",
+ "%" SCNx32 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n",
&pci_addr->domain,
&pci_addr->bus,
&pci_addr->devid,
priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev)
{
struct ibv_async_event event;
- int port_change = 0;
+ struct rte_eth_link *link = &dev->data->dev_link;
int ret = 0;
/* Read all message and acknowledge them. */
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
+ if (event.event_type != IBV_EVENT_PORT_ACTIVE &&
+ event.event_type != IBV_EVENT_PORT_ERR)
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)) {
+ mlx5_link_update(dev, 0);
+ if (((link->link_speed == 0) && link->link_status) ||
+ ((link->link_speed != 0) && !link->link_status)) {
+ if (!priv->pending_alarm) {
/* 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;
+ }
+ } else {
+ ret = 1;
}
return ret;
}
priv_lock(priv);
assert(priv->pending_alarm == 1);
+ priv->pending_alarm = 0;
ret = priv_dev_link_status_handler(priv, dev);
priv_unlock(priv);
if (ret)
- _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL,
+ NULL);
}
/**
* Callback argument.
*/
void
-mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg)
+mlx5_dev_interrupt_handler(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);
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL,
+ NULL);
}
/**
rte_eal_alarm_cancel(mlx5_dev_link_status_handler, dev);
priv->pending_alarm = 0;
priv->intr_handle.fd = 0;
- priv->intr_handle.type = 0;
+ priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN;
}
/**
dev);
}
}
+
+/**
+ * Change the link state (UP / DOWN).
+ *
+ * @param priv
+ * 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;
+
+ if (up) {
+ err = priv_set_flags(priv, ~IFF_UP, IFF_UP);
+ if (err)
+ return err;
+ priv_select_tx_function(priv);
+ priv_select_rx_function(priv);
+ } 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_ctrl *primary_txq_ctrl;
+ struct txq_ctrl *txq_ctrl;
+
+ if (primary_txq == NULL)
+ continue;
+ primary_txq_ctrl = container_of(primary_txq,
+ struct txq_ctrl, txq);
+ txq_ctrl = rte_calloc_socket("TXQ", 1, sizeof(*txq_ctrl) +
+ (1 << primary_txq->elts_n) *
+ sizeof(struct rte_mbuf *), 0,
+ primary_txq_ctrl->socket);
+ if (txq_ctrl != NULL) {
+ if (txq_ctrl_setup(priv->dev,
+ txq_ctrl,
+ 1 << primary_txq->elts_n,
+ primary_txq_ctrl->socket,
+ NULL) == 0) {
+ txq_ctrl->txq.stats.idx =
+ primary_txq->stats.idx;
+ tx_queues[i] = &txq_ctrl->txq;
+ continue;
+ }
+ rte_free(txq_ctrl);
+ }
+ while (i) {
+ txq_ctrl = tx_queues[--i];
+ txq_cleanup(txq_ctrl);
+ rte_free(txq_ctrl);
+ }
+ goto error;
+ }
+ /* RX queues. */
+ for (i = 0; i != nb_rx_queues; ++i) {
+ struct rxq_ctrl *primary_rxq =
+ container_of((*sd->primary_priv->rxqs)[i],
+ struct rxq_ctrl, rxq);
+
+ 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_select_tx_function(priv);
+ priv_select_rx_function(priv);
+ priv_unlock(priv);
+end:
+ /* More sanity checks. */
+ 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;
+}
+
+/**
+ * Configure the TX function to use.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+void
+priv_select_tx_function(struct priv *priv)
+{
+ priv->dev->tx_pkt_burst = mlx5_tx_burst;
+ /* Select appropriate TX function. */
+ if (priv->mps == MLX5_MPW_ENHANCED) {
+ if (priv_check_vec_tx_support(priv) > 0) {
+ if (priv_check_raw_vec_tx_support(priv) > 0)
+ priv->dev->tx_pkt_burst = mlx5_tx_burst_raw_vec;
+ else
+ priv->dev->tx_pkt_burst = mlx5_tx_burst_vec;
+ DEBUG("selected Enhanced MPW TX vectorized function");
+ } else {
+ priv->dev->tx_pkt_burst = mlx5_tx_burst_empw;
+ DEBUG("selected Enhanced MPW TX function");
+ }
+ } else if (priv->mps && priv->txq_inline) {
+ priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw_inline;
+ DEBUG("selected MPW inline TX function");
+ } else if (priv->mps) {
+ priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw;
+ DEBUG("selected MPW TX function");
+ }
+}
+
+/**
+ * Configure the RX function to use.
+ *
+ * @param priv
+ * Pointer to private structure.
+ */
+void
+priv_select_rx_function(struct priv *priv)
+{
+ if (priv_check_vec_rx_support(priv) > 0) {
+ priv_prep_vec_rx_function(priv);
+ priv->dev->rx_pkt_burst = mlx5_rx_burst_vec;
+ DEBUG("selected RX vectorized function");
+ } else {
+ priv->dev->rx_pkt_burst = mlx5_rx_burst;
+ }
+}