Commit 39e7d095 authored by David S. Miller's avatar David S. Miller
Browse files

Merge branch 'mlx4-next'

Or Gerlitz says:

====================
mlx4: Add SRIOV support for RoCE

This series adds SRIOV support for RoCE (RDMA over Ethernet) to the mlx4 driver.

The patches are against net-next, as of commit 2d8d40af

 "pkt_sched: fq:
do not hold qdisc lock while allocating memory"

changes from V1:
 - addressed feedback from Dave on patch #3 and changed get_real_sgid_index()
   to be called fill_in_real_sgid_index() and be a void  function.
 - removed some checkpatch warnings on long lines

changes from V0:
  - always check the return code of mlx4_get_roce_gid_from_slave().
    The call we fixed is introduced in patch #1 and later removed by
    patch #3 that allows guests to have multiple GIDS. The 1..3
    separation was done for proper division of patches to logical changes.
====================
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 36f6fdb7 aa9a2d51
......@@ -61,6 +61,11 @@ struct cm_generic_msg {
__be32 remote_comm_id;
};
struct cm_sidr_generic_msg {
struct ib_mad_hdr hdr;
__be32 request_id;
};
struct cm_req_msg {
unsigned char unused[0x60];
union ib_gid primary_path_sgid;
......@@ -69,28 +74,62 @@ struct cm_req_msg {
static void set_local_comm_id(struct ib_mad *mad, u32 cm_id)
{
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
msg->local_comm_id = cpu_to_be32(cm_id);
if (mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
struct cm_sidr_generic_msg *msg =
(struct cm_sidr_generic_msg *)mad;
msg->request_id = cpu_to_be32(cm_id);
} else if (mad->mad_hdr.attr_id == CM_SIDR_REP_ATTR_ID) {
pr_err("trying to set local_comm_id in SIDR_REP\n");
return;
} else {
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
msg->local_comm_id = cpu_to_be32(cm_id);
}
}
static u32 get_local_comm_id(struct ib_mad *mad)
{
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
return be32_to_cpu(msg->local_comm_id);
if (mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
struct cm_sidr_generic_msg *msg =
(struct cm_sidr_generic_msg *)mad;
return be32_to_cpu(msg->request_id);
} else if (mad->mad_hdr.attr_id == CM_SIDR_REP_ATTR_ID) {
pr_err("trying to set local_comm_id in SIDR_REP\n");
return -1;
} else {
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
return be32_to_cpu(msg->local_comm_id);
}
}
static void set_remote_comm_id(struct ib_mad *mad, u32 cm_id)
{
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
msg->remote_comm_id = cpu_to_be32(cm_id);
if (mad->mad_hdr.attr_id == CM_SIDR_REP_ATTR_ID) {
struct cm_sidr_generic_msg *msg =
(struct cm_sidr_generic_msg *)mad;
msg->request_id = cpu_to_be32(cm_id);
} else if (mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
pr_err("trying to set remote_comm_id in SIDR_REQ\n");
return;
} else {
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
msg->remote_comm_id = cpu_to_be32(cm_id);
}
}
static u32 get_remote_comm_id(struct ib_mad *mad)
{
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
return be32_to_cpu(msg->remote_comm_id);
if (mad->mad_hdr.attr_id == CM_SIDR_REP_ATTR_ID) {
struct cm_sidr_generic_msg *msg =
(struct cm_sidr_generic_msg *)mad;
return be32_to_cpu(msg->request_id);
} else if (mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
pr_err("trying to set remote_comm_id in SIDR_REQ\n");
return -1;
} else {
struct cm_generic_msg *msg = (struct cm_generic_msg *)mad;
return be32_to_cpu(msg->remote_comm_id);
}
}
static union ib_gid gid_from_req_msg(struct ib_device *ibdev, struct ib_mad *mad)
......@@ -282,19 +321,21 @@ int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id
u32 sl_cm_id;
int pv_cm_id = -1;
sl_cm_id = get_local_comm_id(mad);
if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID ||
mad->mad_hdr.attr_id == CM_REP_ATTR_ID) {
mad->mad_hdr.attr_id == CM_REP_ATTR_ID ||
mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
sl_cm_id = get_local_comm_id(mad);
id = id_map_alloc(ibdev, slave_id, sl_cm_id);
if (IS_ERR(id)) {
mlx4_ib_warn(ibdev, "%s: id{slave: %d, sl_cm_id: 0x%x} Failed to id_map_alloc\n",
__func__, slave_id, sl_cm_id);
return PTR_ERR(id);
}
} else if (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID) {
} else if (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID ||
mad->mad_hdr.attr_id == CM_SIDR_REP_ATTR_ID) {
return 0;
} else {
sl_cm_id = get_local_comm_id(mad);
id = id_map_get(ibdev, &pv_cm_id, slave_id, sl_cm_id);
}
......@@ -315,14 +356,18 @@ int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id
}
int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
struct ib_mad *mad)
struct ib_mad *mad)
{
u32 pv_cm_id;
struct id_map_entry *id;
if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID) {
if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID ||
mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
union ib_gid gid;
if (!slave)
return 0;
gid = gid_from_req_msg(ibdev, mad);
*slave = mlx4_ib_find_real_gid(ibdev, port, gid.global.interface_id);
if (*slave < 0) {
......@@ -341,7 +386,8 @@ int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
return -ENOENT;
}
*slave = id->slave_id;
if (slave)
*slave = id->slave_id;
set_remote_comm_id(mad, id->sl_cm_id);
if (mad->mad_hdr.attr_id == CM_DREQ_ATTR_ID)
......
......@@ -564,7 +564,7 @@ static int mlx4_ib_ipoib_csum_ok(__be16 status, __be16 checksum)
}
static int use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct ib_wc *wc,
unsigned tail, struct mlx4_cqe *cqe)
unsigned tail, struct mlx4_cqe *cqe, int is_eth)
{
struct mlx4_ib_proxy_sqp_hdr *hdr;
......@@ -574,12 +574,20 @@ static int use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct
DMA_FROM_DEVICE);
hdr = (struct mlx4_ib_proxy_sqp_hdr *) (qp->sqp_proxy_rcv[tail].addr);
wc->pkey_index = be16_to_cpu(hdr->tun.pkey_index);
wc->slid = be16_to_cpu(hdr->tun.slid_mac_47_32);
wc->sl = (u8) (be16_to_cpu(hdr->tun.sl_vid) >> 12);
wc->src_qp = be32_to_cpu(hdr->tun.flags_src_qp) & 0xFFFFFF;
wc->wc_flags |= (hdr->tun.g_ml_path & 0x80) ? (IB_WC_GRH) : 0;
wc->dlid_path_bits = 0;
if (is_eth) {
wc->vlan_id = be16_to_cpu(hdr->tun.sl_vid);
memcpy(&(wc->smac[0]), (char *)&hdr->tun.mac_31_0, 4);
memcpy(&(wc->smac[4]), (char *)&hdr->tun.slid_mac_47_32, 2);
wc->wc_flags |= (IB_WC_WITH_VLAN | IB_WC_WITH_SMAC);
} else {
wc->slid = be16_to_cpu(hdr->tun.slid_mac_47_32);
wc->sl = (u8) (be16_to_cpu(hdr->tun.sl_vid) >> 12);
}
return 0;
}
......@@ -594,6 +602,7 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
struct mlx4_srq *msrq = NULL;
int is_send;
int is_error;
int is_eth;
u32 g_mlpath_rqpn;
u16 wqe_ctr;
unsigned tail = 0;
......@@ -778,11 +787,15 @@ repoll:
break;
}
is_eth = (rdma_port_get_link_layer(wc->qp->device,
(*cur_qp)->port) ==
IB_LINK_LAYER_ETHERNET);
if (mlx4_is_mfunc(to_mdev(cq->ibcq.device)->dev)) {
if ((*cur_qp)->mlx4_ib_qp_type &
(MLX4_IB_QPT_PROXY_SMI_OWNER |
MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI))
return use_tunnel_data(*cur_qp, cq, wc, tail, cqe);
return use_tunnel_data(*cur_qp, cq, wc, tail,
cqe, is_eth);
}
wc->slid = be16_to_cpu(cqe->rlid);
......@@ -793,20 +806,21 @@ repoll:
wc->pkey_index = be32_to_cpu(cqe->immed_rss_invalid) & 0x7f;
wc->wc_flags |= mlx4_ib_ipoib_csum_ok(cqe->status,
cqe->checksum) ? IB_WC_IP_CSUM_OK : 0;
if (rdma_port_get_link_layer(wc->qp->device,
(*cur_qp)->port) == IB_LINK_LAYER_ETHERNET)
if (is_eth) {
wc->sl = be16_to_cpu(cqe->sl_vid) >> 13;
else
wc->sl = be16_to_cpu(cqe->sl_vid) >> 12;
if (be32_to_cpu(cqe->vlan_my_qpn) & MLX4_CQE_VLAN_PRESENT_MASK) {
wc->vlan_id = be16_to_cpu(cqe->sl_vid) &
MLX4_CQE_VID_MASK;
if (be32_to_cpu(cqe->vlan_my_qpn) &
MLX4_CQE_VLAN_PRESENT_MASK) {
wc->vlan_id = be16_to_cpu(cqe->sl_vid) &
MLX4_CQE_VID_MASK;
} else {
wc->vlan_id = 0xffff;
}
memcpy(wc->smac, cqe->smac, ETH_ALEN);
wc->wc_flags |= (IB_WC_WITH_VLAN | IB_WC_WITH_SMAC);
} else {
wc->sl = be16_to_cpu(cqe->sl_vid) >> 12;
wc->vlan_id = 0xffff;
}
wc->wc_flags |= IB_WC_WITH_VLAN;
memcpy(wc->smac, cqe->smac, ETH_ALEN);
wc->wc_flags |= IB_WC_WITH_SMAC;
}
return 0;
......
......@@ -467,6 +467,7 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
int ret = 0;
u16 tun_pkey_ix;
u16 cached_pkey;
u8 is_eth = dev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
if (dest_qpt > IB_QPT_GSI)
return -EINVAL;
......@@ -509,6 +510,10 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
* The driver will set the force loopback bit in post_send */
memset(&attr, 0, sizeof attr);
attr.port_num = port;
if (is_eth) {
memcpy(&attr.grh.dgid.raw[0], &grh->dgid.raw[0], 16);
attr.ah_flags = IB_AH_GRH;
}
ah = ib_create_ah(tun_ctx->pd, &attr);
if (IS_ERR(ah))
return -ENOMEM;
......@@ -540,11 +545,36 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
/* adjust tunnel data */
tun_mad->hdr.pkey_index = cpu_to_be16(tun_pkey_ix);
tun_mad->hdr.sl_vid = cpu_to_be16(((u16)(wc->sl)) << 12);
tun_mad->hdr.slid_mac_47_32 = cpu_to_be16(wc->slid);
tun_mad->hdr.flags_src_qp = cpu_to_be32(wc->src_qp & 0xFFFFFF);
tun_mad->hdr.g_ml_path = (grh && (wc->wc_flags & IB_WC_GRH)) ? 0x80 : 0;
if (is_eth) {
u16 vlan = 0;
if (mlx4_get_slave_default_vlan(dev->dev, port, slave, &vlan,
NULL)) {
/* VST mode */
if (vlan != wc->vlan_id)
/* Packet vlan is not the VST-assigned vlan.
* Drop the packet.
*/
goto out;
else
/* Remove the vlan tag before forwarding
* the packet to the VF.
*/
vlan = 0xffff;
} else {
vlan = wc->vlan_id;
}
tun_mad->hdr.sl_vid = cpu_to_be16(vlan);
memcpy((char *)&tun_mad->hdr.mac_31_0, &(wc->smac[0]), 4);
memcpy((char *)&tun_mad->hdr.slid_mac_47_32, &(wc->smac[4]), 2);
} else {
tun_mad->hdr.sl_vid = cpu_to_be16(((u16)(wc->sl)) << 12);
tun_mad->hdr.slid_mac_47_32 = cpu_to_be16(wc->slid);
}
ib_dma_sync_single_for_device(&dev->ib_dev,
tun_qp->tx_ring[tun_tx_ix].buf.map,
sizeof (struct mlx4_rcv_tunnel_mad),
......@@ -580,6 +610,41 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port,
int err;
int slave;
u8 *slave_id;
int is_eth = 0;
if (rdma_port_get_link_layer(ibdev, port) == IB_LINK_LAYER_INFINIBAND)
is_eth = 0;
else
is_eth = 1;
if (is_eth) {
if (!(wc->wc_flags & IB_WC_GRH)) {
mlx4_ib_warn(ibdev, "RoCE grh not present.\n");
return -EINVAL;
}
if (mad->mad_hdr.mgmt_class != IB_MGMT_CLASS_CM) {
mlx4_ib_warn(ibdev, "RoCE mgmt class is not CM\n");
return -EINVAL;
}
if (mlx4_get_slave_from_roce_gid(dev->dev, port, grh->dgid.raw, &slave)) {
mlx4_ib_warn(ibdev, "failed matching grh\n");
return -ENOENT;
}
if (slave >= dev->dev->caps.sqp_demux) {
mlx4_ib_warn(ibdev, "slave id: %d is bigger than allowed:%d\n",
slave, dev->dev->caps.sqp_demux);
return -ENOENT;
}
if (mlx4_ib_demux_cm_handler(ibdev, port, NULL, mad))
return 0;
err = mlx4_ib_send_to_slave(dev, slave, port, wc->qp->qp_type, wc, grh, mad);
if (err)
pr_debug("failed sending to slave %d via tunnel qp (%d)\n",
slave, err);
return 0;
}
/* Initially assume that this mad is for us */
slave = mlx4_master_func_num(dev->dev);
......@@ -1076,8 +1141,9 @@ static int is_proxy_qp0(struct mlx4_ib_dev *dev, int qpn, int slave)
int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
enum ib_qp_type dest_qpt, u16 pkey_index, u32 remote_qpn,
u32 qkey, struct ib_ah_attr *attr, struct ib_mad *mad)
enum ib_qp_type dest_qpt, u16 pkey_index,
u32 remote_qpn, u32 qkey, struct ib_ah_attr *attr,
u8 *s_mac, struct ib_mad *mad)
{
struct ib_sge list;
struct ib_send_wr wr, *bad_wr;
......@@ -1166,6 +1232,9 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
wr.num_sge = 1;
wr.opcode = IB_WR_SEND;
wr.send_flags = IB_SEND_SIGNALED;
if (s_mac)
memcpy(to_mah(ah)->av.eth.s_mac, s_mac, 6);
ret = ib_post_send(send_qp, &wr, &bad_wr);
out:
......@@ -1174,6 +1243,34 @@ out:
return ret;
}
static int get_slave_base_gid_ix(struct mlx4_ib_dev *dev, int slave, int port)
{
int gids;
int vfs;
if (rdma_port_get_link_layer(&dev->ib_dev, port) == IB_LINK_LAYER_INFINIBAND)
return slave;
gids = MLX4_ROCE_MAX_GIDS - MLX4_ROCE_PF_GIDS;
vfs = dev->dev->num_vfs;
if (slave == 0)
return 0;
if (slave <= gids % vfs)
return MLX4_ROCE_PF_GIDS + ((gids / vfs) + 1) * (slave - 1);
return MLX4_ROCE_PF_GIDS + (gids % vfs) + ((gids / vfs) * (slave - 1));
}
static void fill_in_real_sgid_index(struct mlx4_ib_dev *dev, int slave, int port,
struct ib_ah_attr *ah_attr)
{
if (rdma_port_get_link_layer(&dev->ib_dev, port) == IB_LINK_LAYER_INFINIBAND)
ah_attr->grh.sgid_index = slave;
else
ah_attr->grh.sgid_index += get_slave_base_gid_ix(dev, slave, port);
}
static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc *wc)
{
struct mlx4_ib_dev *dev = to_mdev(ctx->ib_dev);
......@@ -1260,12 +1357,14 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
memcpy(&ah.av, &tunnel->hdr.av, sizeof (struct mlx4_av));
ah.ibah.device = ctx->ib_dev;
mlx4_ib_query_ah(&ah.ibah, &ah_attr);
if ((ah_attr.ah_flags & IB_AH_GRH) &&
(ah_attr.grh.sgid_index != slave)) {
mlx4_ib_warn(ctx->ib_dev, "slave:%d accessed invalid sgid_index:%d\n",
slave, ah_attr.grh.sgid_index);
return;
}
if (ah_attr.ah_flags & IB_AH_GRH)
fill_in_real_sgid_index(dev, slave, ctx->port, &ah_attr);
memcpy(ah_attr.dmac, tunnel->hdr.mac, 6);
ah_attr.vlan_id = be16_to_cpu(tunnel->hdr.vlan);
/* if slave have default vlan use it */
mlx4_get_slave_default_vlan(dev->dev, ctx->port, slave,
&ah_attr.vlan_id, &ah_attr.sl);
mlx4_ib_send_to_wire(dev, slave, ctx->port,
is_proxy_qp0(dev, wc->src_qp, slave) ?
......@@ -1273,7 +1372,7 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
be16_to_cpu(tunnel->hdr.pkey_index),
be32_to_cpu(tunnel->hdr.remote_qpn),
be32_to_cpu(tunnel->hdr.qkey),
&ah_attr, &tunnel->mad);
&ah_attr, wc->smac, &tunnel->mad);
}
static int mlx4_ib_alloc_pv_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
......
......@@ -1888,14 +1888,6 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
pr_info_once("%s", mlx4_ib_version);
mlx4_foreach_non_ib_transport_port(i, dev)
num_ports++;
if (mlx4_is_mfunc(dev) && num_ports) {
dev_err(&dev->pdev->dev, "RoCE is not supported over SRIOV as yet\n");
return NULL;
}
num_ports = 0;
mlx4_foreach_ib_transport_port(i, dev)
num_ports++;
......
......@@ -215,8 +215,9 @@ static int send_mad_to_wire(struct mlx4_ib_demux_ctx *ctx, struct ib_mad *mad)
}
mlx4_ib_query_ah(dev->sm_ah[ctx->port - 1], &ah_attr);
spin_unlock(&dev->sm_lock);
return mlx4_ib_send_to_wire(dev, mlx4_master_func_num(dev->dev), ctx->port,
IB_QPT_GSI, 0, 1, IB_QP1_QKEY, &ah_attr, mad);
return mlx4_ib_send_to_wire(dev, mlx4_master_func_num(dev->dev),
ctx->port, IB_QPT_GSI, 0, 1, IB_QP1_QKEY,
&ah_attr, NULL, mad);
}
static int send_mad_to_slave(int slave, struct mlx4_ib_demux_ctx *ctx,
......
......@@ -241,6 +241,22 @@ struct mlx4_ib_proxy_sqp_hdr {
struct mlx4_rcv_tunnel_hdr tun;
} __packed;
struct mlx4_roce_smac_vlan_info {
u64 smac;
int smac_index;
int smac_port;
u64 candidate_smac;
int candidate_smac_index;
int candidate_smac_port;
u16 vid;
int vlan_index;
int vlan_port;
u16 candidate_vid;
int candidate_vlan_index;
int candidate_vlan_port;
int update_vid;
};
struct mlx4_ib_qp {
struct ib_qp ibqp;
struct mlx4_qp mqp;
......@@ -273,8 +289,9 @@ struct mlx4_ib_qp {
struct list_head gid_list;
struct list_head steering_rules;
struct mlx4_ib_buf *sqp_proxy_rcv;
struct mlx4_roce_smac_vlan_info pri;
struct mlx4_roce_smac_vlan_info alt;
u64 reg_id;
};
struct mlx4_ib_srq {
......@@ -720,9 +737,12 @@ void mlx4_ib_tunnels_update_work(struct work_struct *work);
int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
enum ib_qp_type qpt, struct ib_wc *wc,
struct ib_grh *grh, struct ib_mad *mad);
int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
enum ib_qp_type dest_qpt, u16 pkey_index, u32 remote_qpn,
u32 qkey, struct ib_ah_attr *attr, struct ib_mad *mad);
u32 qkey, struct ib_ah_attr *attr, u8 *s_mac,
struct ib_mad *mad);
__be64 mlx4_ib_get_new_demux_tid(struct mlx4_ib_demux_ctx *ctx);
int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
......
......@@ -662,10 +662,14 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
if (!sqp)
return -ENOMEM;
qp = &sqp->qp;
qp->pri.vid = 0xFFFF;
qp->alt.vid = 0xFFFF;
} else {
qp = kzalloc(sizeof (struct mlx4_ib_qp), GFP_KERNEL);
if (!qp)
return -ENOMEM;
qp->pri.vid = 0xFFFF;
qp->alt.vid = 0xFFFF;
}
} else
qp = *caller_qp;
......@@ -940,11 +944,32 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
{
struct mlx4_ib_cq *send_cq, *recv_cq;
if (qp->state != IB_QPS_RESET)
if (qp->state != IB_QPS_RESET) {
if (mlx4_qp_modify(dev->dev, NULL, to_mlx4_state(qp->state),
MLX4_QP_STATE_RST, NULL, 0, 0, &qp->mqp))
pr_warn("modify QP %06x to RESET failed.\n",
qp->mqp.qpn);
if (qp->pri.smac) {
mlx4_unregister_mac(dev->dev, qp->pri.smac_port, qp->pri.smac);
qp->pri.smac = 0;
}
if (qp->alt.smac) {
mlx4_unregister_mac(dev->dev, qp->alt.smac_port, qp->alt.smac);
qp->alt.smac = 0;
}
if (qp->pri.vid < 0x1000) {
mlx4_unregister_vlan(dev->dev, qp->pri.vlan_port, qp->pri.vid);
qp->pri.vid = 0xFFFF;
qp->pri.candidate_vid = 0xFFFF;
qp->pri.update_vid = 0;
}
if (qp->alt.vid < 0x1000) {
mlx4_unregister_vlan(dev->dev, qp->alt.vlan_port, qp->alt.vid);
qp->alt.vid = 0xFFFF;
qp->alt.candidate_vid = 0xFFFF;
qp->alt.update_vid = 0;
}
}
get_cqs(qp, &send_cq, &recv_cq);
......@@ -1057,6 +1082,8 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
qp = kzalloc(sizeof *qp, GFP_KERNEL);
if (!qp)
return ERR_PTR(-ENOMEM);
qp->pri.vid = 0xFFFF;
qp->alt.vid = 0xFFFF;
/* fall through */
case IB_QPT_UD:
{
......@@ -1188,12 +1215,13 @@ static void mlx4_set_sched(struct mlx4_qp_path *path, u8 port)
static int _mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah,
u64 smac, u16 vlan_tag, struct mlx4_qp_path *path,
u8 port)
struct mlx4_roce_smac_vlan_info *smac_info, u8 port)
{
int is_eth = rdma_port_get_link_layer(&dev->ib_dev, port) ==
IB_LINK_LAYER_ETHERNET;
int vidx;
int smac_index;
int err;
path->grh_mylmc = ah->src_path_bits & 0x7f;
......@@ -1223,61 +1251,103 @@ static int _mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah,
}
if (is_eth) {
path->sched_queue = MLX4_IB_DEFAULT_SCHED_QUEUE |
((port - 1) << 6) | ((ah->sl & 7) << 3);
if (!(ah->ah_flags & IB_AH_GRH))
return -1;
memcpy(path->dmac, ah->dmac, ETH_ALEN);
path->ackto = MLX4_IB_LINK_TYPE_ETH;
/* find the index into MAC table for IBoE */
if (!is_zero_ether_addr((const u8 *)&smac)) {
if (mlx4_find_cached_mac(dev->dev, port, smac,
&smac_index))
return -ENOENT;
} else {
smac_index = 0;
}
path->grh_mylmc &= 0x80 | smac_index;
path->sched_queue = MLX4_IB_DEFAULT_SCHED_QUEUE |
((port - 1) << 6) | ((ah->sl & 7) << 3);
path->feup |= MLX4_FEUP_FORCE_ETH_UP;
if (vlan_tag < 0x1000) {
if (mlx4_find_cached_vlan(dev->dev, port, vlan_tag, &vidx))
return -ENOENT;
path->vlan_index = vidx;
path->fl = 1 << 6;
if (smac_info->vid < 0x1000) {
/* both valid vlan ids */
if (smac_info->vid != vlan_tag) {
/* different VIDs. unreg old and reg new */
err = mlx4_register_vlan(dev->dev, port, vlan_tag, &vidx);
if (err)
return err;