aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/mlx4
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband/hw/mlx4')
-rw-r--r--drivers/infiniband/hw/mlx4/ah.c2
-rw-r--r--drivers/infiniband/hw/mlx4/alias_GUID.c2
-rw-r--r--drivers/infiniband/hw/mlx4/cm.c80
-rw-r--r--drivers/infiniband/hw/mlx4/cq.c48
-rw-r--r--drivers/infiniband/hw/mlx4/doorbell.c4
-rw-r--r--drivers/infiniband/hw/mlx4/mad.c162
-rw-r--r--drivers/infiniband/hw/mlx4/main.c139
-rw-r--r--drivers/infiniband/hw/mlx4/mcg.c5
-rw-r--r--drivers/infiniband/hw/mlx4/mlx4_ib.h28
-rw-r--r--drivers/infiniband/hw/mlx4/mr.c39
-rw-r--r--drivers/infiniband/hw/mlx4/qp.c431
-rw-r--r--drivers/infiniband/hw/mlx4/srq.c7
-rw-r--r--drivers/infiniband/hw/mlx4/sysfs.c110
13 files changed, 839 insertions, 218 deletions
diff --git a/drivers/infiniband/hw/mlx4/ah.c b/drivers/infiniband/hw/mlx4/ah.c
index 170dca60804..2d8c3397774 100644
--- a/drivers/infiniband/hw/mlx4/ah.c
+++ b/drivers/infiniband/hw/mlx4/ah.c
@@ -73,7 +73,7 @@ static struct ib_ah *create_iboe_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr
{
struct mlx4_ib_dev *ibdev = to_mdev(pd->device);
struct mlx4_dev *dev = ibdev->dev;
- int is_mcast;
+ int is_mcast = 0;
struct in6_addr in6;
u16 vlan_tag;
diff --git a/drivers/infiniband/hw/mlx4/alias_GUID.c b/drivers/infiniband/hw/mlx4/alias_GUID.c
index 2f215b93db6..0eb141c4141 100644
--- a/drivers/infiniband/hw/mlx4/alias_GUID.c
+++ b/drivers/infiniband/hw/mlx4/alias_GUID.c
@@ -154,7 +154,7 @@ void mlx4_ib_notify_slaves_on_guid_change(struct mlx4_ib_dev *dev,
continue;
slave_id = (block_num * NUM_ALIAS_GUID_IN_REC) + i ;
- if (slave_id >= dev->dev->num_slaves)
+ if (slave_id >= dev->dev->num_vfs + 1)
return;
tmp_cur_ag = *(__be64 *)&p_data[i * GUID_REC_SIZE];
form_cache_ag = get_cached_alias_guid(dev, port_num,
diff --git a/drivers/infiniband/hw/mlx4/cm.c b/drivers/infiniband/hw/mlx4/cm.c
index d1f5f1dd77b..56a593e0ae5 100644
--- a/drivers/infiniband/hw/mlx4/cm.c
+++ b/drivers/infiniband/hw/mlx4/cm.c
@@ -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)
diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c
index cc40f08ca8f..1066eec854a 100644
--- a/drivers/infiniband/hw/mlx4/cq.c
+++ b/drivers/infiniband/hw/mlx4/cq.c
@@ -102,7 +102,7 @@ static int mlx4_ib_alloc_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *
int err;
err = mlx4_buf_alloc(dev->dev, nent * dev->dev->caps.cqe_size,
- PAGE_SIZE * 2, &buf->buf);
+ PAGE_SIZE * 2, &buf->buf, GFP_KERNEL);
if (err)
goto out;
@@ -113,7 +113,7 @@ static int mlx4_ib_alloc_cq_buf(struct mlx4_ib_dev *dev, struct mlx4_ib_cq_buf *
if (err)
goto err_buf;
- err = mlx4_buf_write_mtt(dev->dev, &buf->mtt, &buf->buf);
+ err = mlx4_buf_write_mtt(dev->dev, &buf->mtt, &buf->buf, GFP_KERNEL);
if (err)
goto err_mtt;
@@ -209,7 +209,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector
uar = &to_mucontext(context)->uar;
} else {
- err = mlx4_db_alloc(dev->dev, &cq->db, 1);
+ err = mlx4_db_alloc(dev->dev, &cq->db, 1, GFP_KERNEL);
if (err)
goto err_cq;
@@ -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;
diff --git a/drivers/infiniband/hw/mlx4/doorbell.c b/drivers/infiniband/hw/mlx4/doorbell.c
index 8aee4233b38..c5174098636 100644
--- a/drivers/infiniband/hw/mlx4/doorbell.c
+++ b/drivers/infiniband/hw/mlx4/doorbell.c
@@ -45,7 +45,6 @@ int mlx4_ib_db_map_user(struct mlx4_ib_ucontext *context, unsigned long virt,
struct mlx4_db *db)
{
struct mlx4_ib_user_db_page *page;
- struct ib_umem_chunk *chunk;
int err = 0;
mutex_lock(&context->db_page_mutex);
@@ -73,8 +72,7 @@ int mlx4_ib_db_map_user(struct mlx4_ib_ucontext *context, unsigned long virt,
list_add(&page->list, &context->db_page_list);
found:
- chunk = list_entry(page->umem->chunk_list.next, struct ib_umem_chunk, list);
- db->dma = sg_dma_address(chunk->page_list) + (virt & ~PAGE_MASK);
+ db->dma = sg_dma_address(page->umem->sg_head.sgl) + (virt & ~PAGE_MASK);
db->u.user_page = page;
++page->refcnt;
diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
index f2a3f48107e..287ad0564ac 100644
--- a/drivers/infiniband/hw/mlx4/mad.c
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -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;
@@ -477,10 +478,6 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
if (!tun_ctx || tun_ctx->state != DEMUX_PV_STATE_ACTIVE)
return -EAGAIN;
- /* QP0 forwarding only for Dom0 */
- if (!dest_qpt && (mlx4_master_func_num(dev->dev) != slave))
- return -EINVAL;
-
if (!dest_qpt)
tun_qp = &tun_ctx->qp[0];
else
@@ -509,6 +506,10 @@ 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 +541,36 @@ 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 +606,41 @@ 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);
@@ -602,6 +663,21 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port,
}
/* Class-specific handling */
switch (mad->mad_hdr.mgmt_class) {
+ case IB_MGMT_CLASS_SUBN_LID_ROUTED:
+ case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
+ /* 255 indicates the dom0 */
+ if (slave != 255 && slave != mlx4_master_func_num(dev->dev)) {
+ if (!mlx4_vf_smi_enabled(dev->dev, slave, port))
+ return -EPERM;
+ /* for a VF. drop unsolicited MADs */
+ if (!(mad->mad_hdr.method & IB_MGMT_METHOD_RESP)) {
+ mlx4_ib_warn(ibdev, "demux QP0. rejecting unsolicited mad for slave %d class 0x%x, method 0x%x\n",
+ slave, mad->mad_hdr.mgmt_class,
+ mad->mad_hdr.method);
+ return -EINVAL;
+ }
+ }
+ break;
case IB_MGMT_CLASS_SUBN_ADM:
if (mlx4_ib_demux_sa_handler(ibdev, port, slave,
(struct ib_sa_mad *) mad))
@@ -1076,8 +1152,9 @@ 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;
@@ -1099,10 +1176,6 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
if (!sqp_ctx || sqp_ctx->state != DEMUX_PV_STATE_ACTIVE)
return -EAGAIN;
- /* QP0 forwarding only for Dom0 */
- if (dest_qpt == IB_QPT_SMI && (mlx4_master_func_num(dev->dev) != slave))
- return -EINVAL;
-
if (dest_qpt == IB_QPT_SMI) {
src_qpnum = 0;
sqp = &sqp_ctx->qp[0];
@@ -1166,6 +1239,9 @@ 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 +1250,22 @@ out:
return ret;
}
+static int get_slave_base_gid_ix(struct mlx4_ib_dev *dev, int slave, int port)
+{
+ if (rdma_port_get_link_layer(&dev->ib_dev, port) == IB_LINK_LAYER_INFINIBAND)
+ return slave;
+ return mlx4_get_base_gid_ix(dev->dev, slave, port);
+}
+
+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);
@@ -1184,6 +1276,7 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
struct ib_ah_attr ah_attr;
u8 *slave_id;
int slave;
+ int port;
/* Get slave that sent this packet */
if (wc->src_qp < dev->dev->phys_caps.base_proxy_sqpn ||
@@ -1199,11 +1292,6 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
"belongs to another slave\n", wc->src_qp);
return;
}
- if (slave != mlx4_master_func_num(dev->dev) && !(wc->src_qp & 0x2)) {
- mlx4_ib_warn(ctx->ib_dev, "can't multiplex bad sqp:%d: "
- "non-master trying to send QP0 packets\n", wc->src_qp);
- return;
- }
/* Map transaction ID */
ib_dma_sync_single_for_cpu(ctx->ib_dev, tun_qp->ring[wr_ix].map,
@@ -1231,6 +1319,12 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
/* Class-specific handling */
switch (tunnel->mad.mad_hdr.mgmt_class) {
+ case IB_MGMT_CLASS_SUBN_LID_ROUTED:
+ case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
+ if (slave != mlx4_master_func_num(dev->dev) &&
+ !mlx4_vf_smi_enabled(dev->dev, slave, ctx->port))
+ return;
+ break;
case IB_MGMT_CLASS_SUBN_ADM:
if (mlx4_ib_multiplex_sa_handler(ctx->ib_dev, ctx->port, slave,
(struct ib_sa_mad *) &tunnel->mad))
@@ -1260,12 +1354,18 @@ 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);
+ if (ah_attr.ah_flags & IB_AH_GRH)
+ fill_in_real_sgid_index(dev, slave, ctx->port, &ah_attr);
+
+ port = mlx4_slave_convert_port(dev->dev, slave, ah_attr.port_num);
+ if (port < 0)
return;
- }
+ ah_attr.port_num = port;
+ 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 +1373,7 @@ 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,
@@ -1657,9 +1757,9 @@ static int create_pv_resources(struct ib_device *ibdev, int slave, int port,
return -EEXIST;
ctx->state = DEMUX_PV_STATE_STARTING;
- /* have QP0 only on port owner, and only if link layer is IB */
- if (ctx->slave == mlx4_master_func_num(to_mdev(ctx->ib_dev)->dev) &&
- rdma_port_get_link_layer(ibdev, ctx->port) == IB_LINK_LAYER_INFINIBAND)
+ /* have QP0 only if link layer is IB */
+ if (rdma_port_get_link_layer(ibdev, ctx->port) ==
+ IB_LINK_LAYER_INFINIBAND)
ctx->has_smi = 1;
if (ctx->has_smi) {
@@ -1850,7 +1950,15 @@ static int mlx4_ib_alloc_demux_ctx(struct mlx4_ib_dev *dev,
ctx->port = port;
ctx->ib_dev = &dev->ib_dev;
- for (i = 0; i < dev->dev->caps.sqp_demux; i++) {
+ for (i = 0;
+ i < min(dev->dev->caps.sqp_demux, (u16)(dev->dev->num_vfs + 1));
+ i++) {
+ struct mlx4_active_ports actv_ports =
+ mlx4_get_active_ports(dev->dev, i);
+
+ if (!test_bit(port - 1, actv_ports.ports))
+ continue;
+
ret = alloc_pv_object(dev, i, port, &ctx->tun[i]);
if (ret) {
ret = -ENOMEM;
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index e81c5547e64..0f7027e7db1 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -48,13 +48,14 @@
#include <linux/mlx4/driver.h>
#include <linux/mlx4/cmd.h>
+#include <linux/mlx4/qp.h>
#include "mlx4_ib.h"
#include "user.h"
#define DRV_NAME MLX4_IB_DRV_NAME
-#define DRV_VERSION "1.0"
-#define DRV_RELDATE "April 4, 2008"
+#define DRV_VERSION "2.2-1"
+#define DRV_RELDATE "Feb 2014"
#define MLX4_IB_FLOW_MAX_PRIO 0xFFF
#define MLX4_IB_FLOW_QPN_MASK 0xFFFFFF
@@ -544,12 +545,11 @@ static int mlx4_ib_modify_device(struct ib_device *ibdev, int mask,
return 0;
}
-static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
- u32 cap_mask)
+static int mlx4_ib_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
+ u32 cap_mask)
{
struct mlx4_cmd_mailbox *mailbox;
int err;
- u8 is_eth = dev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
mailbox = mlx4_alloc_cmd_mailbox(dev->dev);
if (IS_ERR(mailbox))
@@ -563,8 +563,8 @@ static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
((__be32 *) mailbox->buf)[1] = cpu_to_be32(cap_mask);
}
- err = mlx4_cmd(dev->dev, mailbox->dma, port, is_eth, MLX4_CMD_SET_PORT,
- MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE);
+ err = mlx4_cmd(dev->dev, mailbox->dma, port, 0, MLX4_CMD_SET_PORT,
+ MLX4_CMD_TIME_CLASS_B, MLX4_CMD_WRAPPED);
mlx4_free_cmd_mailbox(dev->dev, mailbox);
return err;
@@ -573,11 +573,20 @@ static int mlx4_SET_PORT(struct mlx4_ib_dev *dev, u8 port, int reset_qkey_viols,
static int mlx4_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
struct ib_port_modify *props)
{
+ struct mlx4_ib_dev *mdev = to_mdev(ibdev);
+ u8 is_eth = mdev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
struct ib_port_attr attr;
u32 cap_mask;
int err;
- mutex_lock(&to_mdev(ibdev)->cap_mask_mutex);
+ /* return OK if this is RoCE. CM calls ib_modify_port() regardless
+ * of whether port link layer is ETH or IB. For ETH ports, qkey
+ * violations and port capabilities are not meaningful.
+ */
+ if (is_eth)
+ return 0;
+
+ mutex_lock(&mdev->cap_mask_mutex);
err = mlx4_ib_query_port(ibdev, port, &attr);
if (err)
@@ -586,9 +595,9 @@ static int mlx4_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
cap_mask = (attr.port_cap_flags | props->set_port_cap_mask) &
~props->clr_port_cap_mask;
- err = mlx4_SET_PORT(to_mdev(ibdev), port,
- !!(mask & IB_PORT_RESET_QKEY_CNTR),
- cap_mask);
+ err = mlx4_ib_SET_PORT(mdev, port,
+ !!(mask & IB_PORT_RESET_QKEY_CNTR),
+ cap_mask);
out:
mutex_unlock(&to_mdev(ibdev)->cap_mask_mutex);
@@ -1546,7 +1555,7 @@ static int mlx4_ib_addr_event(int event, struct net_device *event_netdev,
iboe = &ibdev->iboe;
spin_lock(&iboe->lock);
- for (port = 1; port <= MLX4_MAX_PORTS; ++port)
+ for (port = 1; port <= ibdev->dev->caps.num_ports; ++port)
if ((netif_is_bond_master(real_dev) &&
(real_dev == iboe->masters[port - 1])) ||
(!netif_is_bond_master(real_dev) &&
@@ -1569,14 +1578,14 @@ static u8 mlx4_ib_get_dev_port(struct net_device *dev,
iboe = &ibdev->iboe;
- for (port = 1; port <= MLX4_MAX_PORTS; ++port)
+ for (port = 1; port <= ibdev->dev->caps.num_ports; ++port)
if ((netif_is_bond_master(real_dev) &&
(real_dev == iboe->masters[port - 1])) ||
(!netif_is_bond_master(real_dev) &&
(real_dev == iboe->netdevs[port - 1])))
break;
- if ((port == 0) || (port > MLX4_MAX_PORTS))
+ if ((port == 0) || (port > ibdev->dev->caps.num_ports))
return 0;
else
return port;
@@ -1614,6 +1623,53 @@ static int mlx4_ib_inet6_event(struct notifier_block *this, unsigned long event,
}
#endif
+#define MLX4_IB_INVALID_MAC ((u64)-1)
+static void mlx4_ib_update_qps(struct mlx4_ib_dev *ibdev,
+ struct net_device *dev,
+ int port)
+{
+ u64 new_smac = 0;
+ u64 release_mac = MLX4_IB_INVALID_MAC;
+ struct mlx4_ib_qp *qp;
+
+ read_lock(&dev_base_lock);
+ new_smac = mlx4_mac_to_u64(dev->dev_addr);
+ read_unlock(&dev_base_lock);
+
+ mutex_lock(&ibdev->qp1_proxy_lock[port - 1]);
+ qp = ibdev->qp1_proxy[port - 1];
+ if (qp) {
+ int new_smac_index;
+ u64 old_smac = qp->pri.smac;
+ struct mlx4_update_qp_params update_params;
+
+ if (new_smac == old_smac)
+ goto unlock;
+
+ new_smac_index = mlx4_register_mac(ibdev->dev, port, new_smac);
+
+ if (new_smac_index < 0)
+ goto unlock;
+
+ update_params.smac_index = new_smac_index;
+ if (mlx4_update_qp(ibdev->dev, &qp->mqp, MLX4_UPDATE_QP_SMAC,
+ &update_params)) {
+ release_mac = new_smac;
+ goto unlock;
+ }
+
+ qp->pri.smac = new_smac;
+ qp->pri.smac_index = new_smac_index;
+
+ release_mac = old_smac;
+ }
+
+unlock:
+ mutex_unlock(&ibdev->qp1_proxy_lock[port - 1]);
+ if (release_mac != MLX4_IB_INVALID_MAC)
+ mlx4_unregister_mac(ibdev->dev, port, release_mac);
+}
+
static void mlx4_ib_get_dev_addr(struct net_device *dev,
struct mlx4_ib_dev *ibdev, u8 port)
{
@@ -1626,7 +1682,7 @@ static void mlx4_ib_get_dev_addr(struct net_device *dev,
union ib_gid gid;
- if ((port == 0) || (port > MLX4_MAX_PORTS))
+ if ((port == 0) || (port > ibdev->dev->caps.num_ports))
return;
/* IPv4 gids */
@@ -1689,9 +1745,13 @@ static int mlx4_ib_init_gid_table(struct mlx4_ib_dev *ibdev)
return 0;
}
-static void mlx4_ib_scan_netdevs(struct mlx4_ib_dev *ibdev)
+static void mlx4_ib_scan_netdevs(struct mlx4_ib_dev *ibdev,
+ struct net_device *dev,
+ unsigned long event)
+
{
struct mlx4_ib_iboe *iboe;
+ int update_qps_port = -1;
int port;
iboe = &ibdev->iboe;
@@ -1719,6 +1779,11 @@ static void mlx4_ib_scan_netdevs(struct mlx4_ib_dev *ibdev)
}
curr_master = iboe->masters[port - 1];
+ if (dev == iboe->netdevs[port - 1] &&
+ (event == NETDEV_CHANGEADDR || event == NETDEV_REGISTER ||
+ event == NETDEV_UP || event == NETDEV_CHANGE))
+ update_qps_port = port;
+
if (curr_netdev) {
port_state = (netif_running(curr_netdev) && netif_carrier_ok(curr_netdev)) ?
IB_PORT_ACTIVE : IB_PORT_DOWN;
@@ -1752,6 +1817,9 @@ static void mlx4_ib_scan_netdevs(struct mlx4_ib_dev *ibdev)
}
spin_unlock(&iboe->lock);
+
+ if (update_qps_port > 0)
+ mlx4_ib_update_qps(ibdev, dev, update_qps_port);
}
static int mlx4_ib_netdev_event(struct notifier_block *this,
@@ -1764,7 +1832,7 @@ static int mlx4_ib_netdev_event(struct notifier_block *this,
return NOTIFY_DONE;
ibdev = container_of(this, struct mlx4_ib_dev, iboe.nb);
- mlx4_ib_scan_netdevs(ibdev);
+ mlx4_ib_scan_netdevs(ibdev, dev, event);
return NOTIFY_DONE;
}
@@ -1803,7 +1871,7 @@ static void init_pkeys(struct mlx4_ib_dev *ibdev)
static void mlx4_ib_alloc_eqs(struct mlx4_dev *dev, struct mlx4_ib_dev *ibdev)
{
- char name[32];
+ char name[80];
int eq_per_port = 0;
int added_eqs = 0;
int total_eqs = 0;
@@ -1833,8 +1901,8 @@ static void mlx4_ib_alloc_eqs(struct mlx4_dev *dev, struct mlx4_ib_dev *ibdev)
eq = 0;
mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_IB) {
for (j = 0; j < eq_per_port; j++) {
- sprintf(name, "mlx4-ib-%d-%d@%s",
- i, j, dev->pdev->bus->name);
+ snprintf(name, sizeof(name), "mlx4-ib-%d-%d@%s",
+ i, j, dev->pdev->bus->name);
/* Set IRQ for specific name (per ring) */
if (mlx4_assign_eq(dev, name, NULL,
&ibdev->eq_table[eq])) {
@@ -1888,14 +1956,6 @@ 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++;
@@ -2051,13 +2111,15 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
goto err_map;
for (i = 0; i < ibdev->num_ports; ++i) {
+ mutex_init(&ibdev->qp1_proxy_lock[i]);
if (mlx4_ib_port_link_layer(&ibdev->ib_dev, i + 1) ==
IB_LINK_LAYER_ETHERNET) {
err = mlx4_counter_alloc(ibdev->dev, &ibdev->counters[i]);
if (err)
ibdev->counters[i] = -1;
- } else
- ibdev->counters[i] = -1;
+ } else {
+ ibdev->counters[i] = -1;
+ }
}
mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_IB)
@@ -2133,7 +2195,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
for (i = 1 ; i <= ibdev->num_ports ; ++i)
reset_gid_table(ibdev, i);
rtnl_lock();
- mlx4_ib_scan_netdevs(ibdev);
+ mlx4_ib_scan_netdevs(ibdev, NULL, 0);
rtnl_unlock();
mlx4_ib_init_gid_table(ibdev);
}
@@ -2331,17 +2393,24 @@ static void do_slave_init(struct mlx4_ib_dev *ibdev, int slave, int do_init)
struct mlx4_dev *dev = ibdev->dev;
int i;
unsigned long flags;
+ struct mlx4_active_ports actv_ports;
+ unsigned int ports;
+ unsigned int first_port;
if (!mlx4_is_master(dev))
return;
- dm = kcalloc(dev->caps.num_ports, sizeof *dm, GFP_ATOMIC);
+ actv_ports = mlx4_get_active_ports(dev, slave);
+ ports = bitmap_weight(actv_ports.ports, dev->caps.num_ports);
+ first_port = find_first_bit(actv_ports.ports, dev->caps.num_ports);
+
+ dm = kcalloc(ports, sizeof(*dm), GFP_ATOMIC);
if (!dm) {
pr_err("failed to allocate memory for tunneling qp update\n");
goto out;
}
- for (i = 0; i < dev->caps.num_ports; i++) {
+ for (i = 0; i < ports; i++) {
dm[i] = kmalloc(sizeof (struct mlx4_ib_demux_work), GFP_ATOMIC);
if (!dm[i]) {
pr_err("failed to allocate memory for tunneling qp update work struct\n");
@@ -2353,9 +2422,9 @@ static void do_slave_init(struct mlx4_ib_dev *ibdev, int slave, int do_init)
}
}
/* initialize or tear down tunnel QPs for the slave */
- for (i = 0; i < dev->caps.num_ports; i++) {
+ for (i = 0; i < ports; i++) {
INIT_WORK(&dm[i]->work, mlx4_ib_tunnels_update_work);
- dm[i]->port = i + 1;
+ dm[i]->port = first_port + i + 1;
dm[i]->slave = slave;
dm[i]->do_init = do_init;
dm[i]->dev = ibdev;
diff --git a/drivers/infiniband/hw/mlx4/mcg.c b/drivers/infiniband/hw/mlx4/mcg.c
index 25b2cdff00f..ed327e6c8fd 100644
--- a/drivers/infiniband/hw/mlx4/mcg.c
+++ b/drivers/infiniband/hw/mlx4/mcg.c
@@ -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,
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index a230683af94..369da3ca5d6 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -156,6 +156,7 @@ enum mlx4_ib_qp_flags {
MLX4_IB_QP_LSO = IB_QP_CREATE_IPOIB_UD_LSO,
MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK = IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK,
MLX4_IB_QP_NETIF = IB_QP_CREATE_NETIF_QP,
+ MLX4_IB_QP_CREATE_USE_GFP_NOIO = IB_QP_CREATE_USE_GFP_NOIO,
MLX4_IB_SRIOV_TUNNEL_QP = 1 << 30,
MLX4_IB_SRIOV_SQP = 1 << 31,
};
@@ -241,6 +242,22 @@ 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 +290,9 @@ 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 {
@@ -505,6 +523,9 @@ struct mlx4_ib_dev {
int steer_qpn_count;
int steer_qpn_base;
int steering_support;
+ struct mlx4_ib_qp *qp1_proxy[MLX4_MAX_PORTS];
+ /* lock when destroying qp1_proxy and getting netdev events */
+ struct mutex qp1_proxy_lock[MLX4_MAX_PORTS];
};
struct ib_event_work {
@@ -720,9 +741,12 @@ 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,
diff --git a/drivers/infiniband/hw/mlx4/mr.c b/drivers/infiniband/hw/mlx4/mr.c
index e471f089ff0..cb2a8727f3f 100644
--- a/drivers/infiniband/hw/mlx4/mr.c
+++ b/drivers/infiniband/hw/mlx4/mr.c
@@ -90,11 +90,11 @@ int mlx4_ib_umem_write_mtt(struct mlx4_ib_dev *dev, struct mlx4_mtt *mtt,
struct ib_umem *umem)
{
u64 *pages;
- struct ib_umem_chunk *chunk;
- int i, j, k;
+ int i, k, entry;
int n;
int len;
int err = 0;
+ struct scatterlist *sg;
pages = (u64 *) __get_free_page(GFP_KERNEL);
if (!pages)
@@ -102,26 +102,25 @@ int mlx4_ib_umem_write_mtt(struct mlx4_ib_dev *dev, struct mlx4_mtt *mtt,
i = n = 0;
- list_for_each_entry(chunk, &umem->chunk_list, list)
- for (j = 0; j < chunk->nmap; ++j) {
- len = sg_dma_len(&chunk->page_list[j]) >> mtt->page_shift;
- for (k = 0; k < len; ++k) {
- pages[i++] = sg_dma_address(&chunk->page_list[j]) +
- umem->page_size * k;
- /*
- * Be friendly to mlx4_write_mtt() and
- * pass it chunks of appropriate size.
- */
- if (i == PAGE_SIZE / sizeof (u64)) {
- err = mlx4_write_mtt(dev->dev, mtt, n,
- i, pages);
- if (err)
- goto out;
- n += i;
- i = 0;
- }
+ for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
+ len = sg_dma_len(sg) >> mtt->page_shift;
+ for (k = 0; k < len; ++k) {
+ pages[i++] = sg_dma_address(sg) +
+ umem->page_size * k;
+ /*
+ * Be friendly to mlx4_write_mtt() and
+ * pass it chunks of appropriate size.
+ */
+ if (i == PAGE_SIZE / sizeof (u64)) {
+ err = mlx4_write_mtt(dev->dev, mtt, n,
+ i, pages);
+ if (err)
+ goto out;
+ n += i;
+ i = 0;
}
}
+ }
if (i)
err = mlx4_write_mtt(dev->dev, mtt, n, i, pages);
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index d8f4d1fe849..67780452f0c 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -608,9 +608,20 @@ static int qp_has_rq(struct ib_qp_init_attr *attr)
return !attr->srq;
}
+static int qp0_enabled_vf(struct mlx4_dev *dev, int qpn)
+{
+ int i;
+ for (i = 0; i < dev->caps.num_ports; i++) {
+ if (qpn == dev->caps.qp0_proxy[i])
+ return !!dev->caps.qp0_qkey[i];
+ }
+ return 0;
+}
+
static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
struct ib_qp_init_attr *init_attr,
- struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp)
+ struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp,
+ gfp_t gfp)
{
int qpn;
int err;
@@ -625,10 +636,13 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
!(init_attr->create_flags & MLX4_IB_SRIOV_SQP))) {
if (init_attr->qp_type == IB_QPT_GSI)
qp_type = MLX4_IB_QPT_PROXY_GSI;
- else if (mlx4_is_master(dev->dev))
- qp_type = MLX4_IB_QPT_PROXY_SMI_OWNER;
- else
- qp_type = MLX4_IB_QPT_PROXY_SMI;
+ else {
+ if (mlx4_is_master(dev->dev) ||
+ qp0_enabled_vf(dev->dev, sqpn))
+ qp_type = MLX4_IB_QPT_PROXY_SMI_OWNER;
+ else
+ qp_type = MLX4_IB_QPT_PROXY_SMI;
+ }
}
qpn = sqpn;
/* add extra sg entry for tunneling */
@@ -643,7 +657,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
return -EINVAL;
if (tnl_init->proxy_qp_type == IB_QPT_GSI)
qp_type = MLX4_IB_QPT_TUN_GSI;
- else if (tnl_init->slave == mlx4_master_func_num(dev->dev))
+ else if (tnl_init->slave == mlx4_master_func_num(dev->dev) ||
+ mlx4_vf_smi_enabled(dev->dev, tnl_init->slave,
+ tnl_init->port))
qp_type = MLX4_IB_QPT_TUN_SMI_OWNER;
else
qp_type = MLX4_IB_QPT_TUN_SMI;
@@ -658,14 +674,18 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
if (qp_type == MLX4_IB_QPT_SMI || qp_type == MLX4_IB_QPT_GSI ||
(qp_type & (MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_SMI_OWNER |
MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER))) {
- sqp = kzalloc(sizeof (struct mlx4_ib_sqp), GFP_KERNEL);
+ sqp = kzalloc(sizeof (struct mlx4_ib_sqp), gfp);
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);
+ qp = kzalloc(sizeof (struct mlx4_ib_qp), gfp);
if (!qp)
return -ENOMEM;
+ qp->pri.vid = 0xFFFF;
+ qp->alt.vid = 0xFFFF;
}
} else
qp = *caller_qp;
@@ -744,14 +764,14 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
goto err;
if (qp_has_rq(init_attr)) {
- err = mlx4_db_alloc(dev->dev, &qp->db, 0);
+ err = mlx4_db_alloc(dev->dev, &qp->db, 0, gfp);
if (err)
goto err;
*qp->db.db = 0;
}
- if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf)) {
+ if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf, gfp)) {
err = -ENOMEM;
goto err_db;
}
@@ -761,13 +781,12 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
if (err)
goto err_buf;
- err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf);
+ err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf, gfp);
if (err)
goto err_mtt;
- qp->sq.wrid = kmalloc(qp->sq.wqe_cnt * sizeof (u64), GFP_KERNEL);
- qp->rq.wrid = kmalloc(qp->rq.wqe_cnt * sizeof (u64), GFP_KERNEL);
-
+ qp->sq.wrid = kmalloc(qp->sq.wqe_cnt * sizeof (u64), gfp);
+ qp->rq.wrid = kmalloc(qp->rq.wqe_cnt * sizeof (u64), gfp);
if (!qp->sq.wrid || !qp->rq.wrid) {
err = -ENOMEM;
goto err_wrid;
@@ -797,7 +816,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
goto err_proxy;
}
- err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp);
+ err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp, gfp);
if (err)
goto err_qpn;
@@ -940,11 +959,32 @@ 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);
@@ -1015,7 +1055,10 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
struct mlx4_ib_qp *qp = NULL;
int err;
u16 xrcdn = 0;
+ gfp_t gfp;
+ gfp = (init_attr->create_flags & MLX4_IB_QP_CREATE_USE_GFP_NOIO) ?
+ GFP_NOIO : GFP_KERNEL;
/*
* We only support LSO, vendor flag1, and multicast loopback blocking,
* and only for kernel UD QPs.
@@ -1024,7 +1067,8 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK |
MLX4_IB_SRIOV_TUNNEL_QP |
MLX4_IB_SRIOV_SQP |
- MLX4_IB_QP_NETIF))
+ MLX4_IB_QP_NETIF |
+ MLX4_IB_QP_CREATE_USE_GFP_NOIO))
return ERR_PTR(-EINVAL);
if (init_attr->create_flags & IB_QP_CREATE_NETIF_QP) {
@@ -1034,7 +1078,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
if (init_attr->create_flags &&
(udata ||
- ((init_attr->create_flags & ~MLX4_IB_SRIOV_SQP) &&
+ ((init_attr->create_flags & ~(MLX4_IB_SRIOV_SQP | MLX4_IB_QP_CREATE_USE_GFP_NOIO)) &&
init_attr->qp_type != IB_QPT_UD) ||
((init_attr->create_flags & MLX4_IB_SRIOV_SQP) &&
init_attr->qp_type > IB_QPT_GSI)))
@@ -1054,14 +1098,16 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
case IB_QPT_RC:
case IB_QPT_UC:
case IB_QPT_RAW_PACKET:
- qp = kzalloc(sizeof *qp, GFP_KERNEL);
+ qp = kzalloc(sizeof *qp, gfp);
if (!qp)
return ERR_PTR(-ENOMEM);
+ qp->pri.vid = 0xFFFF;
+ qp->alt.vid = 0xFFFF;
/* fall through */
case IB_QPT_UD:
{
err = create_qp_common(to_mdev(pd->device), pd, init_attr,
- udata, 0, &qp);
+ udata, 0, &qp, gfp);
if (err)
return ERR_PTR(err);
@@ -1079,7 +1125,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
err = create_qp_common(to_mdev(pd->device), pd, init_attr, udata,
get_sqp_num(to_mdev(pd->device), init_attr),
- &qp);
+ &qp, gfp);
if (err)
return ERR_PTR(err);
@@ -1105,6 +1151,12 @@ int mlx4_ib_destroy_qp(struct ib_qp *qp)
if (is_qp0(dev, mqp))
mlx4_CLOSE_PORT(dev->dev, mqp->port);
+ if (dev->qp1_proxy[mqp->port - 1] == mqp) {
+ mutex_lock(&dev->qp1_proxy_lock[mqp->port - 1]);
+ dev->qp1_proxy[mqp->port - 1] = NULL;
+ mutex_unlock(&dev->qp1_proxy_lock[mqp->port - 1]);
+ }
+
pd = get_pd(mqp);
destroy_qp_common(dev, mqp, !!pd->ibpd.uobject);
@@ -1188,12 +1240,13 @@ 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 +1276,103 @@ 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;
+ smac_info->candidate_vid = vlan_tag;
+ smac_info->candidate_vlan_index = vidx;
+ smac_info->candidate_vlan_port = port;
+ smac_info->update_vid = 1;
+ path->vlan_index = vidx;
+ } else {
+ path->vlan_index = smac_info->vlan_index;
+ }
+ } else {
+ /* no current vlan tag in qp */
+ err = mlx4_register_vlan(dev->dev, port, vlan_tag, &vidx);
+ if (err)
+ return err;
+ smac_info->candidate_vid = vlan_tag;
+ smac_info->candidate_vlan_index = vidx;
+ smac_info->candidate_vlan_port = port;
+ smac_info->update_vid = 1;
+ path->vlan_index = vidx;
+ }
path->feup |= MLX4_FVL_FORCE_ETH_VLAN;
+ path->fl = 1 << 6;
+ } else {
+ /* have current vlan tag. unregister it at modify-qp success */
+ if (smac_info->vid < 0x1000) {
+ smac_info->candidate_vid = 0xFFFF;
+ smac_info->update_vid = 1;
+ }
}
- } else
+
+ /* get smac_index for RoCE use.
+ * If no smac was yet assigned, register one.
+ * If one was already assigned, but the new mac differs,
+ * unregister the old one and register the new one.
+ */
+ if (!smac_info->smac || smac_info->smac != smac) {
+ /* register candidate now, unreg if needed, after success */
+ smac_index = mlx4_register_mac(dev->dev, port, smac);
+ if (smac_index >= 0) {
+ smac_info->candidate_smac_index = smac_index;
+ smac_info->candidate_smac = smac;
+ smac_info->candidate_smac_port = port;
+ } else {
+ return -EINVAL;
+ }
+ } else {
+ smac_index = smac_info->smac_index;
+ }
+
+ memcpy(path->dmac, ah->dmac, 6);
+ path->ackto = MLX4_IB_LINK_TYPE_ETH;
+ /* put MAC table smac index for IBoE */
+ path->grh_mylmc = (u8) (smac_index) | 0x80;
+ } else {
path->sched_queue = MLX4_IB_DEFAULT_SCHED_QUEUE |
((port - 1) << 6) | ((ah->sl & 0xf) << 2);
+ }
return 0;
}
static int mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_qp_attr *qp,
enum ib_qp_attr_mask qp_attr_mask,
+ struct mlx4_ib_qp *mqp,
struct mlx4_qp_path *path, u8 port)
{
return _mlx4_set_path(dev, &qp->ah_attr,
mlx4_mac_to_u64((u8 *)qp->smac),
(qp_attr_mask & IB_QP_VID) ? qp->vlan_id : 0xffff,
- path, port);
+ path, &mqp->pri, port);
}
static int mlx4_set_alt_path(struct mlx4_ib_dev *dev,
const struct ib_qp_attr *qp,
enum ib_qp_attr_mask qp_attr_mask,
+ struct mlx4_ib_qp *mqp,
struct mlx4_qp_path *path, u8 port)
{
return _mlx4_set_path(dev, &qp->alt_ah_attr,
mlx4_mac_to_u64((u8 *)qp->alt_smac),
(qp_attr_mask & IB_QP_ALT_VID) ?
qp->alt_vlan_id : 0xffff,
- path, port);
+ path, &mqp->alt, port);
}
static void update_mcg_macs(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
@@ -1292,6 +1387,37 @@ static void update_mcg_macs(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
}
}
+static int handle_eth_ud_smac_index(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, u8 *smac,
+ struct mlx4_qp_context *context)
+{
+ struct net_device *ndev;
+ u64 u64_mac;
+ int smac_index;
+
+
+ ndev = dev->iboe.netdevs[qp->port - 1];
+ if (ndev) {
+ smac = ndev->dev_addr;
+ u64_mac = mlx4_mac_to_u64(smac);
+ } else {
+ u64_mac = dev->dev->caps.def_mac[qp->port];
+ }
+
+ context->pri_path.sched_queue = MLX4_IB_DEFAULT_SCHED_QUEUE | ((qp->port - 1) << 6);
+ if (!qp->pri.smac) {
+ smac_index = mlx4_register_mac(dev->dev, qp->port, u64_mac);
+ if (smac_index >= 0) {
+ qp->pri.candidate_smac_index = smac_index;
+ qp->pri.candidate_smac = u64_mac;
+ qp->pri.candidate_smac_port = qp->port;
+ context->pri_path.grh_mylmc = 0x80 | (u8) smac_index;
+ } else {
+ return -ENOENT;
+ }
+ }
+ return 0;
+}
+
static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
const struct ib_qp_attr *attr, int attr_mask,
enum ib_qp_state cur_state, enum ib_qp_state new_state)
@@ -1403,7 +1529,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
}
if (attr_mask & IB_QP_AV) {
- if (mlx4_set_path(dev, attr, attr_mask, &context->pri_path,
+ if (mlx4_set_path(dev, attr, attr_mask, qp, &context->pri_path,
attr_mask & IB_QP_PORT ?
attr->port_num : qp->port))
goto out;
@@ -1426,7 +1552,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
dev->dev->caps.pkey_table_len[attr->alt_port_num])
goto out;
- if (mlx4_set_alt_path(dev, attr, attr_mask, &context->alt_path,
+ if (mlx4_set_alt_path(dev, attr, attr_mask, qp,
+ &context->alt_path,
attr->alt_port_num))
goto out;
@@ -1532,6 +1659,22 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
context->pri_path.fl = 0x80;
context->pri_path.sched_queue |= MLX4_IB_DEFAULT_SCHED_QUEUE;
}
+ if (rdma_port_get_link_layer(&dev->ib_dev, qp->port) ==
+ IB_LINK_LAYER_ETHERNET) {
+ if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_TUN_GSI ||
+ qp->mlx4_ib_qp_type == MLX4_IB_QPT_GSI)
+ context->pri_path.feup = 1 << 7; /* don't fsm */
+ /* handle smac_index */
+ if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_UD ||
+ qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI ||
+ qp->mlx4_ib_qp_type == MLX4_IB_QPT_TUN_GSI) {
+ err = handle_eth_ud_smac_index(dev, qp, (u8 *)attr->smac, context);
+ if (err)
+ return -EINVAL;
+ if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI)
+ dev->qp1_proxy[qp->port - 1] = qp;
+ }
+ }
}
if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET)
@@ -1619,28 +1762,113 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
* If we moved a kernel QP to RESET, clean up all old CQ
* entries and reinitialize the QP.
*/
- if (new_state == IB_QPS_RESET && !ibqp->uobject) {
- mlx4_ib_cq_clean(recv_cq, qp->mqp.qpn,
- ibqp->srq ? to_msrq(ibqp->srq): NULL);
- if (send_cq != recv_cq)
- mlx4_ib_cq_clean(send_cq, qp->mqp.qpn, NULL);
+ if (new_state == IB_QPS_RESET) {
+ if (!ibqp->uobject) {
+ mlx4_ib_cq_clean(recv_cq, qp->mqp.qpn,
+ ibqp->srq ? to_msrq(ibqp->srq) : NULL);
+ if (send_cq != recv_cq)
+ mlx4_ib_cq_clean(send_cq, qp->mqp.qpn, NULL);
+
+ qp->rq.head = 0;
+ qp->rq.tail = 0;
+ qp->sq.head = 0;
+ qp->sq.tail = 0;
+ qp->sq_next_wqe = 0;
+ if (qp->rq.wqe_cnt)
+ *qp->db.db = 0;
- qp->rq.head = 0;
- qp->rq.tail = 0;
- qp->sq.head = 0;
- qp->sq.tail = 0;
- qp->sq_next_wqe = 0;
- if (qp->rq.wqe_cnt)
- *qp->db.db = 0;
+ if (qp->flags & MLX4_IB_QP_NETIF)
+ mlx4_ib_steer_qp_reg(dev, qp, 0);
+ }
+ 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->flags & MLX4_IB_QP_NETIF)
- mlx4_ib_steer_qp_reg(dev, qp, 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;
+ }
}
-
out:
if (err && steer_qp)
mlx4_ib_steer_qp_reg(dev, qp, 0);
kfree(context);
+ if (qp->pri.candidate_smac) {
+ if (err) {
+ mlx4_unregister_mac(dev->dev, qp->pri.candidate_smac_port, qp->pri.candidate_smac);
+ } else {
+ if (qp->pri.smac)
+ mlx4_unregister_mac(dev->dev, qp->pri.smac_port, qp->pri.smac);
+ qp->pri.smac = qp->pri.candidate_smac;
+ qp->pri.smac_index = qp->pri.candidate_smac_index;
+ qp->pri.smac_port = qp->pri.candidate_smac_port;
+ }
+ qp->pri.candidate_smac = 0;
+ qp->pri.candidate_smac_index = 0;
+ qp->pri.candidate_smac_port = 0;
+ }
+ if (qp->alt.candidate_smac) {
+ if (err) {
+ mlx4_unregister_mac(dev->dev, qp->alt.candidate_smac_port, qp->alt.candidate_smac);
+ } else {
+ if (qp->alt.smac)
+ mlx4_unregister_mac(dev->dev, qp->alt.smac_port, qp->alt.smac);
+ qp->alt.smac = qp->alt.candidate_smac;
+ qp->alt.smac_index = qp->alt.candidate_smac_index;
+ qp->alt.smac_port = qp->alt.candidate_smac_port;
+ }
+ qp->alt.candidate_smac = 0;
+ qp->alt.candidate_smac_index = 0;
+ qp->alt.candidate_smac_port = 0;
+ }
+
+ if (qp->pri.update_vid) {
+ if (err) {
+ if (qp->pri.candidate_vid < 0x1000)
+ mlx4_unregister_vlan(dev->dev, qp->pri.candidate_vlan_port,
+ qp->pri.candidate_vid);
+ } else {
+ if (qp->pri.vid < 0x1000)
+ mlx4_unregister_vlan(dev->dev, qp->pri.vlan_port,
+ qp->pri.vid);
+ qp->pri.vid = qp->pri.candidate_vid;
+ qp->pri.vlan_port = qp->pri.candidate_vlan_port;
+ qp->pri.vlan_index = qp->pri.candidate_vlan_index;
+ }
+ qp->pri.candidate_vid = 0xFFFF;
+ qp->pri.update_vid = 0;
+ }
+
+ if (qp->alt.update_vid) {
+ if (err) {
+ if (qp->alt.candidate_vid < 0x1000)
+ mlx4_unregister_vlan(dev->dev, qp->alt.candidate_vlan_port,
+ qp->alt.candidate_vid);
+ } else {
+ if (qp->alt.vid < 0x1000)
+ mlx4_unregister_vlan(dev->dev, qp->alt.vlan_port,
+ qp->alt.vid);
+ qp->alt.vid = qp->alt.candidate_vid;
+ qp->alt.vlan_port = qp->alt.candidate_vlan_port;
+ qp->alt.vlan_index = qp->alt.candidate_vlan_index;
+ }
+ qp->alt.candidate_vid = 0xFFFF;
+ qp->alt.update_vid = 0;
+ }
+
return err;
}
@@ -1729,6 +1957,19 @@ out:
return err;
}
+static int vf_get_qp0_qkey(struct mlx4_dev *dev, int qpn, u32 *qkey)
+{
+ int i;
+ for (i = 0; i < dev->caps.num_ports; i++) {
+ if (qpn == dev->caps.qp0_proxy[i] ||
+ qpn == dev->caps.qp0_tunnel[i]) {
+ *qkey = dev->caps.qp0_qkey[i];
+ return 0;
+ }
+ }
+ return -EINVAL;
+}
+
static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
struct ib_send_wr *wr,
void *wqe, unsigned *mlx_seg_len)
@@ -1786,8 +2027,13 @@ static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
cpu_to_be32(mdev->dev->caps.qp0_tunnel[sqp->qp.port - 1]);
sqp->ud_header.bth.psn = cpu_to_be32((sqp->send_psn++) & ((1 << 24) - 1));
- if (mlx4_get_parav_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
- return -EINVAL;
+ if (mlx4_is_master(mdev->dev)) {
+ if (mlx4_get_parav_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+ return -EINVAL;
+ } else {
+ if (vf_get_qp0_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+ return -EINVAL;
+ }
sqp->ud_header.deth.qkey = cpu_to_be32(qkey);
sqp->ud_header.deth.source_qpn = cpu_to_be32(sqp->qp.mqp.qpn);
@@ -1842,9 +2088,9 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
{
struct ib_device *ib_dev = sqp->qp.ibqp.device;
struct mlx4_wqe_mlx_seg *mlx = wqe;
+ struct mlx4_wqe_ctrl_seg *ctrl = wqe;
struct mlx4_wqe_inline_seg *inl = wqe + sizeof *mlx;
struct mlx4_ib_ah *ah = to_mah(wr->wr.ud.ah);
- struct net_device *ndev;
union ib_gid sgid;
u16 pkey;
int send_size;
@@ -1868,12 +2114,11 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
/* When multi-function is enabled, the ib_core gid
* indexes don't necessarily match the hw ones, so
* we must use our own cache */
- sgid.global.subnet_prefix =
- to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1].
- subnet_prefix;
- sgid.global.interface_id =
- to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1].
- guid_cache[ah->av.ib.gid_index];
+ err = mlx4_get_roce_gid_from_slave(to_mdev(ib_dev)->dev,
+ be32_to_cpu(ah->av.ib.port_pd) >> 24,
+ ah->av.ib.gid_index, &sgid.raw[0]);
+ if (err)
+ return err;
} else {
err = ib_get_cached_gid(ib_dev,
be32_to_cpu(ah->av.ib.port_pd) >> 24,
@@ -1882,7 +2127,7 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
return err;
}
- if (ah->av.eth.vlan != 0xffff) {
+ if (ah->av.eth.vlan != cpu_to_be16(0xffff)) {
vlan = be16_to_cpu(ah->av.eth.vlan) & 0x0fff;
is_vlan = 1;
}
@@ -1902,6 +2147,9 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
sqp->ud_header.grh.flow_label =
ah->av.ib.sl_tclass_flowlabel & cpu_to_be32(0xfffff);
sqp->ud_header.grh.hop_limit = ah->av.ib.hop_limit;
+ if (is_eth)
+ memcpy(sqp->ud_header.grh.source_gid.raw, sgid.raw, 16);
+ else {
if (mlx4_is_mfunc(to_mdev(ib_dev)->dev)) {
/* When multi-function is enabled, the ib_core gid
* indexes don't necessarily match the hw ones, so
@@ -1917,6 +2165,7 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
be32_to_cpu(ah->av.ib.port_pd) >> 24,
ah->av.ib.gid_index,
&sqp->ud_header.grh.source_gid);
+ }
memcpy(sqp->ud_header.grh.destination_gid.raw,
ah->av.ib.dgid, 16);
}
@@ -1949,16 +2198,23 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr,
if (is_eth) {
u8 *smac;
+ struct in6_addr in6;
+
u16 pcp = (be32_to_cpu(ah->av.ib.sl_tclass_flowlabel) >> 29) << 13;
mlx->sched_prio = cpu_to_be16(pcp);
memcpy(sqp->ud_header.eth.dmac_h, ah->av.eth.mac, 6);
/* FIXME: cache smac value? */
- ndev = to_mdev(sqp->qp.ibqp.device)->iboe.netdevs[sqp->qp.port - 1];
- if (!ndev)
- return -ENODEV;
- smac = ndev->dev_addr;
+ memcpy(&ctrl->srcrb_flags16[0], ah->av.eth.mac, 2);
+ memcpy(&ctrl->imm, ah->av.eth.mac + 2, 4);
+ memcpy(&in6, sgid.raw, sizeof(in6));
+
+ if (!mlx4_is_mfunc(to_mdev(ib_dev)->dev))
+ smac = to_mdev(sqp->qp.ibqp.device)->
+ iboe.netdevs[sqp->qp.port - 1]->dev_addr;
+ else /* use the src mac of the tunnel */
+ smac = ah->av.eth.s_mac;
memcpy(sqp->ud_header.eth.smac_h, smac, 6);
if (!memcmp(sqp->ud_header.eth.smac_h, sqp->ud_header.eth.dmac_h, 6))
mlx->flags |= cpu_to_be32(MLX4_WQE_CTRL_FORCE_LOOPBACK);
@@ -2159,7 +2415,8 @@ static void set_datagram_seg(struct mlx4_wqe_datagram_seg *dseg,
static void set_tunnel_datagram_seg(struct mlx4_ib_dev *dev,
struct mlx4_wqe_datagram_seg *dseg,
- struct ib_send_wr *wr, enum ib_qp_type qpt)
+ struct ib_send_wr *wr,
+ enum mlx4_ib_qp_type qpt)
{
union mlx4_ext_av *av = &to_mah(wr->wr.ud.ah)->av;
struct mlx4_av sqp_av = {0};
@@ -2172,8 +2429,10 @@ static void set_tunnel_datagram_seg(struct mlx4_ib_dev *dev,
cpu_to_be32(0xf0000000);
memcpy(dseg->av, &sqp_av, sizeof (struct mlx4_av));
- /* This function used only for sending on QP1 proxies */
- dseg->dqpn = cpu_to_be32(dev->dev->caps.qp1_tunnel[port - 1]);
+ if (qpt == MLX4_IB_QPT_PROXY_GSI)
+ dseg->dqpn = cpu_to_be32(dev->dev->caps.qp1_tunnel[port - 1]);
+ else
+ dseg->dqpn = cpu_to_be32(dev->dev->caps.qp0_tunnel[port - 1]);
/* Use QKEY from the QP context, which is set by master */
dseg->qkey = cpu_to_be32(IB_QP_SET_QKEY);
}
@@ -2190,6 +2449,8 @@ static void build_tunnel_header(struct ib_send_wr *wr, void *wqe, unsigned *mlx_
hdr.remote_qpn = cpu_to_be32(wr->wr.ud.remote_qpn);
hdr.pkey_index = cpu_to_be16(wr->wr.ud.pkey_index);
hdr.qkey = cpu_to_be32(wr->wr.ud.remote_qkey);
+ memcpy(hdr.mac, ah->av.eth.mac, 6);
+ hdr.vlan = ah->av.eth.vlan;
spc = MLX4_INLINE_ALIGN -
((unsigned long) (inl + 1) & (MLX4_INLINE_ALIGN - 1));
@@ -2466,11 +2727,6 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
break;
case MLX4_IB_QPT_PROXY_SMI_OWNER:
- if (unlikely(!mlx4_is_master(to_mdev(ibqp->device)->dev))) {
- err = -ENOSYS;
- *bad_wr = wr;
- goto out;
- }
err = build_sriov_qp0_header(to_msqp(qp), wr, ctrl, &seglen);
if (unlikely(err)) {
*bad_wr = wr;
@@ -2487,16 +2743,13 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
size += seglen / 16;
break;
case MLX4_IB_QPT_PROXY_SMI:
- /* don't allow QP0 sends on guests */
- err = -ENOSYS;
- *bad_wr = wr;
- goto out;
case MLX4_IB_QPT_PROXY_GSI:
/* If we are tunneling special qps, this is a UD qp.
* In this case we first add a UD segment targeting
* the tunnel qp, and then add a header with address
* information */
- set_tunnel_datagram_seg(to_mdev(ibqp->device), wqe, wr, ibqp->qp_type);
+ set_tunnel_datagram_seg(to_mdev(ibqp->device), wqe, wr,
+ qp->mlx4_ib_qp_type);
wqe += sizeof (struct mlx4_wqe_datagram_seg);
size += sizeof (struct mlx4_wqe_datagram_seg) / 16;
build_tunnel_header(wr, wqe, &seglen);
diff --git a/drivers/infiniband/hw/mlx4/srq.c b/drivers/infiniband/hw/mlx4/srq.c
index 60c5fb025fc..62d9285300a 100644
--- a/drivers/infiniband/hw/mlx4/srq.c
+++ b/drivers/infiniband/hw/mlx4/srq.c
@@ -134,13 +134,14 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
if (err)
goto err_mtt;
} else {
- err = mlx4_db_alloc(dev->dev, &srq->db, 0);
+ err = mlx4_db_alloc(dev->dev, &srq->db, 0, GFP_KERNEL);
if (err)
goto err_srq;
*srq->db.db = 0;
- if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &srq->buf)) {
+ if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &srq->buf,
+ GFP_KERNEL)) {
err = -ENOMEM;
goto err_db;
}
@@ -165,7 +166,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
if (err)
goto err_buf;
- err = mlx4_buf_write_mtt(dev->dev, &srq->mtt, &srq->buf);
+ err = mlx4_buf_write_mtt(dev->dev, &srq->mtt, &srq->buf, GFP_KERNEL);
if (err)
goto err_mtt;
diff --git a/drivers/infiniband/hw/mlx4/sysfs.c b/drivers/infiniband/hw/mlx4/sysfs.c
index db2ea31df83..cb4c66e723b 100644
--- a/drivers/infiniband/hw/mlx4/sysfs.c
+++ b/drivers/infiniband/hw/mlx4/sysfs.c
@@ -389,8 +389,10 @@ struct mlx4_port {
struct mlx4_ib_dev *dev;
struct attribute_group pkey_group;
struct attribute_group gid_group;
- u8 port_num;
+ struct device_attribute enable_smi_admin;
+ struct device_attribute smi_enabled;
int slave;
+ u8 port_num;
};
@@ -558,6 +560,101 @@ err:
return NULL;
}
+static ssize_t sysfs_show_smi_enabled(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct mlx4_port *p =
+ container_of(attr, struct mlx4_port, smi_enabled);
+ ssize_t len = 0;
+
+ if (mlx4_vf_smi_enabled(p->dev->dev, p->slave, p->port_num))
+ len = sprintf(buf, "%d\n", 1);
+ else
+ len = sprintf(buf, "%d\n", 0);
+
+ return len;
+}
+
+static ssize_t sysfs_show_enable_smi_admin(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct mlx4_port *p =
+ container_of(attr, struct mlx4_port, enable_smi_admin);
+ ssize_t len = 0;
+
+ if (mlx4_vf_get_enable_smi_admin(p->dev->dev, p->slave, p->port_num))
+ len = sprintf(buf, "%d\n", 1);
+ else
+ len = sprintf(buf, "%d\n", 0);
+
+ return len;
+}
+
+static ssize_t sysfs_store_enable_smi_admin(struct device *dev,
+ struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct mlx4_port *p =
+ container_of(attr, struct mlx4_port, enable_smi_admin);
+ int enable;
+
+ if (sscanf(buf, "%i", &enable) != 1 ||
+ enable < 0 || enable > 1)
+ return -EINVAL;
+
+ if (mlx4_vf_set_enable_smi_admin(p->dev->dev, p->slave, p->port_num, enable))
+ return -EINVAL;
+ return count;
+}
+
+static int add_vf_smi_entries(struct mlx4_port *p)
+{
+ int is_eth = rdma_port_get_link_layer(&p->dev->ib_dev, p->port_num) ==
+ IB_LINK_LAYER_ETHERNET;
+ int ret;
+
+ /* do not display entries if eth transport, or if master */
+ if (is_eth || p->slave == mlx4_master_func_num(p->dev->dev))
+ return 0;
+
+ sysfs_attr_init(&p->smi_enabled.attr);
+ p->smi_enabled.show = sysfs_show_smi_enabled;
+ p->smi_enabled.store = NULL;
+ p->smi_enabled.attr.name = "smi_enabled";
+ p->smi_enabled.attr.mode = 0444;
+ ret = sysfs_create_file(&p->kobj, &p->smi_enabled.attr);
+ if (ret) {
+ pr_err("failed to create smi_enabled\n");
+ return ret;
+ }
+
+ sysfs_attr_init(&p->enable_smi_admin.attr);
+ p->enable_smi_admin.show = sysfs_show_enable_smi_admin;
+ p->enable_smi_admin.store = sysfs_store_enable_smi_admin;
+ p->enable_smi_admin.attr.name = "enable_smi_admin";
+ p->enable_smi_admin.attr.mode = 0644;
+ ret = sysfs_create_file(&p->kobj, &p->enable_smi_admin.attr);
+ if (ret) {
+ pr_err("failed to create enable_smi_admin\n");
+ sysfs_remove_file(&p->kobj, &p->smi_enabled.attr);
+ return ret;
+ }
+ return 0;
+}
+
+static void remove_vf_smi_entries(struct mlx4_port *p)
+{
+ int is_eth = rdma_port_get_link_layer(&p->dev->ib_dev, p->port_num) ==
+ IB_LINK_LAYER_ETHERNET;
+
+ if (is_eth || p->slave == mlx4_master_func_num(p->dev->dev))
+ return;
+
+ sysfs_remove_file(&p->kobj, &p->smi_enabled.attr);
+ sysfs_remove_file(&p->kobj, &p->enable_smi_admin.attr);
+}
+
static int add_port(struct mlx4_ib_dev *dev, int port_num, int slave)
{
struct mlx4_port *p;
@@ -602,6 +699,10 @@ static int add_port(struct mlx4_ib_dev *dev, int port_num, int slave)
if (ret)
goto err_free_gid;
+ ret = add_vf_smi_entries(p);
+ if (ret)
+ goto err_free_gid;
+
list_add_tail(&p->kobj.entry, &dev->pkeys.pkey_port_list[slave]);
return 0;
@@ -627,6 +728,7 @@ static int register_one_pkey_tree(struct mlx4_ib_dev *dev, int slave)
int port;
struct kobject *p, *t;
struct mlx4_port *mport;
+ struct mlx4_active_ports actv_ports;
get_name(dev, name, slave, sizeof name);
@@ -649,7 +751,11 @@ static int register_one_pkey_tree(struct mlx4_ib_dev *dev, int slave)
goto err_ports;
}
+ actv_ports = mlx4_get_active_ports(dev->dev, slave);
+
for (port = 1; port <= dev->dev->caps.num_ports; ++port) {
+ if (!test_bit(port - 1, actv_ports.ports))
+ continue;
err = add_port(dev, port, slave);
if (err)
goto err_add;
@@ -664,6 +770,7 @@ err_add:
mport = container_of(p, struct mlx4_port, kobj);
sysfs_remove_group(p, &mport->pkey_group);
sysfs_remove_group(p, &mport->gid_group);
+ remove_vf_smi_entries(mport);
kobject_put(p);
}
kobject_put(dev->dev_ports_parent[slave]);
@@ -708,6 +815,7 @@ static void unregister_pkey_tree(struct mlx4_ib_dev *device)
port = container_of(p, struct mlx4_port, kobj);
sysfs_remove_group(p, &port->pkey_group);
sysfs_remove_group(p, &port->gid_group);
+ remove_vf_smi_entries(port);
kobject_put(p);
kobject_put(device->dev_ports_parent[slave]);
}
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2025-11-12 17:37:03 +0000