#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>
/* Verbs header. */
/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-pedantic"
+#pragma GCC diagnostic ignored "-Wpedantic"
#endif
#include <infiniband/verbs.h>
#ifdef PEDANTIC
-#pragma GCC diagnostic error "-pedantic"
+#pragma GCC diagnostic error "-Wpedantic"
#endif
/* DPDK headers don't like -pedantic. */
#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-pedantic"
+#pragma GCC diagnostic ignored "-Wpedantic"
#endif
#include <rte_ether.h>
#include <rte_ethdev.h>
#include <rte_alarm.h>
#include <rte_memory.h>
#ifdef PEDANTIC
-#pragma GCC diagnostic error "-pedantic"
+#pragma GCC diagnostic error "-Wpedantic"
#endif
/* Generated configuration header. */
unsigned int sp:1; /* Use scattered RX elements. */
unsigned int csum:1; /* Enable checksum offloading. */
unsigned int csum_l2tun:1; /* Same for L2 tunnels. */
- uint32_t mb_len; /* Length of a mp-issued mbuf. */
struct mlx4_rxq_stats stats; /* RX queue counters. */
unsigned int socket; /* CPU socket ID for allocations. */
struct ibv_exp_res_domain *rd; /* Resource Domain. */
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);
}
}
mr_linear =
ibv_reg_mr(txq->priv->pd, elts_linear, sizeof(*elts_linear),
- (IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_REMOTE_WRITE));
+ IBV_ACCESS_LOCAL_WRITE);
if (mr_linear == NULL) {
ERROR("%p: unable to configure MR, ibv_reg_mr() failed",
(void *)txq);
return 0;
}
+struct mlx4_check_mempool_data {
+ int ret;
+ char *start;
+ char *end;
+};
+
+/* Called by mlx4_check_mempool() when iterating the memory chunks. */
+static void mlx4_check_mempool_cb(struct rte_mempool *mp,
+ void *opaque, struct rte_mempool_memhdr *memhdr,
+ unsigned mem_idx)
+{
+ struct mlx4_check_mempool_data *data = opaque;
+
+ (void)mp;
+ (void)mem_idx;
+
+ /* It already failed, skip the next chunks. */
+ if (data->ret != 0)
+ return;
+ /* It is the first chunk. */
+ if (data->start == NULL && data->end == NULL) {
+ data->start = memhdr->addr;
+ data->end = data->start + memhdr->len;
+ return;
+ }
+ if (data->end == memhdr->addr) {
+ data->end += memhdr->len;
+ return;
+ }
+ if (data->start == (char *)memhdr->addr + memhdr->len) {
+ data->start -= memhdr->len;
+ return;
+ }
+ /* Error, mempool is not virtually contigous. */
+ data->ret = -1;
+}
+
+/**
+ * Check if a mempool can be used: it must be virtually contiguous.
+ *
+ * @param[in] mp
+ * Pointer to memory pool.
+ * @param[out] start
+ * Pointer to the start address of the mempool virtual memory area
+ * @param[out] end
+ * Pointer to the end address of the mempool virtual memory area
+ *
+ * @return
+ * 0 on success (mempool is virtually contiguous), -1 on error.
+ */
+static int mlx4_check_mempool(struct rte_mempool *mp, uintptr_t *start,
+ uintptr_t *end)
+{
+ struct mlx4_check_mempool_data data;
+
+ memset(&data, 0, sizeof(data));
+ rte_mempool_mem_iter(mp, mlx4_check_mempool_cb, &data);
+ *start = (uintptr_t)data.start;
+ *end = (uintptr_t)data.end;
+
+ return data.ret;
+}
+
/* For best performance, this function should not be inlined. */
-static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, const struct rte_mempool *)
+static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, struct rte_mempool *)
__attribute__((noinline));
/**
* Memory region pointer, NULL in case of error.
*/
static struct ibv_mr *
-mlx4_mp2mr(struct ibv_pd *pd, const struct rte_mempool *mp)
+mlx4_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp)
{
const struct rte_memseg *ms = rte_eal_get_physmem_layout();
- uintptr_t start = (uintptr_t)STAILQ_FIRST(&mp->mem_list)->addr;
- uintptr_t end = start + STAILQ_FIRST(&mp->mem_list)->len;
+ uintptr_t start;
+ uintptr_t end;
unsigned int i;
+ if (mlx4_check_mempool(mp, &start, &end) != 0) {
+ ERROR("mempool %p: not virtually contiguous",
+ (void *)mp);
+ return NULL;
+ }
+
DEBUG("mempool %p area start=%p end=%p size=%zu",
- (const void *)mp, (void *)start, (void *)end,
+ (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) {
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,
+ (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);
+ IBV_ACCESS_LOCAL_WRITE);
}
/**
* mr->lkey on success, (uint32_t)-1 on failure.
*/
static uint32_t
-txq_mp2mr(struct txq *txq, const struct rte_mempool *mp)
+txq_mp2mr(struct txq *txq, struct rte_mempool *mp)
{
unsigned int i;
struct ibv_mr *mr;
}
/* Add a new entry, register MR first. */
DEBUG("%p: discovered new memory pool \"%s\" (%p)",
- (void *)txq, mp->name, (const void *)mp);
+ (void *)txq, mp->name, (void *)mp);
mr = mlx4_mp2mr(txq->priv->pd, mp);
if (unlikely(mr == NULL)) {
DEBUG("%p: unable to configure MR, ibv_reg_mr() failed.",
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);
+ (void *)txq, mp->name, (void *)mp, txq->mp2mr[i].lkey);
return txq->mp2mr[i].lkey;
}
if (rxq->csum)
ol_flags |=
- TRANSPOSE(~flags,
+ TRANSPOSE(flags,
IBV_EXP_CQ_RX_IP_CSUM_OK,
- PKT_RX_IP_CKSUM_BAD) |
- TRANSPOSE(~flags,
+ PKT_RX_IP_CKSUM_GOOD) |
+ 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).
- */
+ PKT_RX_L4_CKSUM_GOOD);
if ((flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) && (rxq->csum_l2tun))
ol_flags |=
- TRANSPOSE(~flags,
+ TRANSPOSE(flags,
IBV_EXP_CQ_RX_OUTER_IP_CSUM_OK,
- PKT_RX_IP_CKSUM_BAD) |
- TRANSPOSE(~flags,
+ PKT_RX_IP_CKSUM_GOOD) |
+ TRANSPOSE(flags,
IBV_EXP_CQ_RX_OUTER_TCP_UDP_CSUM_OK,
- PKT_RX_L4_CKSUM_BAD);
+ PKT_RX_L4_CKSUM_GOOD);
return ol_flags;
}
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);
* Fetch initial bytes of packet descriptor into a
* cacheline while allocating rep.
*/
- rte_prefetch0(seg);
- rte_prefetch0(&seg->cacheline1);
+ rte_mbuf_prefetch_part1(seg);
+ rte_mbuf_prefetch_part2(seg);
ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL,
&flags);
if (unlikely(ret < 0)) {
unsigned int i, k;
struct ibv_exp_qp_attr mod;
struct ibv_recv_wr *bad_wr;
+ unsigned int mb_len;
int err;
int parent = (rxq == &priv->rxq_parent);
(void *)dev, (void *)rxq);
return EINVAL;
}
+ mb_len = rte_pktmbuf_data_room_size(rxq->mp);
DEBUG("%p: rehashing queue %p", (void *)dev, (void *)rxq);
/* Number of descriptors and mbufs currently allocated. */
desc_n = (tmpl.elts_n * (tmpl.sp ? MLX4_PMD_SGE_WR_N : 1));
rxq->csum_l2tun = tmpl.csum_l2tun;
}
/* Enable scattered packets support for this queue if necessary. */
+ assert(mb_len >= RTE_PKTMBUF_HEADROOM);
if ((dev->data->dev_conf.rxmode.jumbo_frame) &&
(dev->data->dev_conf.rxmode.max_rx_pkt_len >
- (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) {
+ (mb_len - RTE_PKTMBUF_HEADROOM))) {
tmpl.sp = 1;
desc_n /= MLX4_PMD_SGE_WR_N;
} else
} attr;
enum ibv_exp_query_intf_status status;
struct ibv_recv_wr *bad_wr;
- struct rte_mbuf *buf;
+ unsigned int mb_len;
int ret = 0;
int parent = (rxq == &priv->rxq_parent);
desc = 1;
goto skip_mr;
}
+ mb_len = rte_pktmbuf_data_room_size(mp);
if ((desc == 0) || (desc % MLX4_PMD_SGE_WR_N)) {
ERROR("%p: invalid number of RX descriptors (must be a"
" multiple of %d)", (void *)dev, MLX4_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. */
+ assert(mb_len >= RTE_PKTMBUF_HEADROOM);
if ((dev->data->dev_conf.rxmode.jumbo_frame) &&
(dev->data->dev_conf.rxmode.max_rx_pkt_len >
- (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) {
+ (mb_len - RTE_PKTMBUF_HEADROOM))) {
tmpl.sp = 1;
desc /= MLX4_PMD_SGE_WR_N;
}
memset(priv, 0, sizeof(*priv));
}
+/**
+ * Change the link state (UP / DOWN).
+ *
+ * @param priv
+ * Pointer to Ethernet device private data.
+ * @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 = mlx4_rx_burst;
+ else
+ dev->rx_pkt_burst = mlx4_rx_burst_sp;
+ dev->tx_pkt_burst = mlx4_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.
+ */
+static int
+mlx4_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.
+ */
+static int
+mlx4_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;
+}
/**
* DPDK callback to get information about the device.
*
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));
/* Reconfigure each RX queue. */
for (i = 0; (i != priv->rxqs_n); ++i) {
struct rxq *rxq = (*priv->rxqs)[i];
+ unsigned int mb_len;
unsigned int max_frame_len;
int sp;
* 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);
+ sp = (max_frame_len > (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;
if (mlx4_is_secondary())
return -E_RTE_SECONDARY;
- ifr.ifr_data = ðpause;
+ ifr.ifr_data = (void *)ðpause;
priv_lock(priv);
if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
ret = errno;
if (mlx4_is_secondary())
return -E_RTE_SECONDARY;
- ifr.ifr_data = ðpause;
+ 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))
.dev_configure = mlx4_dev_configure,
.dev_start = mlx4_dev_start,
.dev_stop = mlx4_dev_stop,
+ .dev_set_link_down = mlx4_set_link_down,
+ .dev_set_link_up = mlx4_set_link_up,
.dev_close = mlx4_dev_close,
.promiscuous_enable = mlx4_promiscuous_enable,
.promiscuous_disable = mlx4_promiscuous_disable,
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);
}
/**
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);
}
/**
rte_eal_alarm_cancel(mlx4_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;
}
/**
* 0 on success, negative errno value on failure.
*/
static int
-mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
+mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
{
struct ibv_device **list;
struct ibv_device *ibv_dev;
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);
+ eth_dev = rte_eth_dev_allocate(name);
}
if (eth_dev == NULL) {
ERROR("can not allocate rte ethdev");
static const struct rte_pci_id mlx4_pci_id_map[] = {
{
- .vendor_id = PCI_VENDOR_ID_MELLANOX,
- .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3,
- .subsystem_vendor_id = PCI_ANY_ID,
- .subsystem_device_id = PCI_ANY_ID
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX3)
},
{
- .vendor_id = PCI_VENDOR_ID_MELLANOX,
- .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3PRO,
- .subsystem_vendor_id = PCI_ANY_ID,
- .subsystem_device_id = PCI_ANY_ID
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX3PRO)
},
{
- .vendor_id = PCI_VENDOR_ID_MELLANOX,
- .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3VF,
- .subsystem_vendor_id = PCI_ANY_ID,
- .subsystem_device_id = PCI_ANY_ID
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX,
+ PCI_DEVICE_ID_MELLANOX_CONNECTX3VF)
},
{
.vendor_id = 0
static struct eth_driver mlx4_driver = {
.pci_drv = {
- .name = MLX4_DRIVER_NAME,
+ .driver = {
+ .name = MLX4_DRIVER_NAME
+ },
.id_table = mlx4_pci_id_map,
- .devinit = mlx4_pci_devinit,
+ .probe = mlx4_pci_probe,
.drv_flags = RTE_PCI_DRV_INTR_LSC,
},
.dev_private_size = sizeof(struct priv)
/**
* Driver initialization routine.
*/
-static int
-rte_mlx4_pmd_init(const char *name, const char *args)
+RTE_INIT(rte_mlx4_pmd_init);
+static void
+rte_mlx4_pmd_init(void)
{
- (void)name;
- (void)args;
-
RTE_BUILD_BUG_ON(sizeof(wr_id_t) != sizeof(uint64_t));
/*
* RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use
setenv("RDMAV_HUGEPAGES_SAFE", "1", 1);
ibv_fork_init();
rte_eal_pci_register(&mlx4_driver.pci_drv);
- return 0;
}
-static struct rte_driver rte_mlx4_driver = {
- .type = PMD_PDEV,
- .name = MLX4_DRIVER_NAME,
- .init = rte_mlx4_pmd_init,
-};
-
-PMD_REGISTER_DRIVER(rte_mlx4_driver)
+RTE_PMD_EXPORT_NAME(net_mlx4, __COUNTER__);
+RTE_PMD_REGISTER_PCI_TABLE(net_mlx4, mlx4_pci_id_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_mlx4,
+ "* ib_uverbs & mlx4_en & mlx4_core & mlx4_ib");
git.droids-corp.org / dpdk.git / blobdiff