提交 0a9a0188 编写于 作者: J Jack Morgenstein 提交者: Roland Dreier

mlx4: MAD_IFC paravirtualization

The MAD_IFC firmware command fulfills two functions.

First, it is used in the QP0/QP1 MAD-handling flow to obtain
information from the FW (for answering queries), and for setting
variables in the HCA (MAD SET packets).

For this, MAD_IFC should provide the FW (physical) view of the data.
This is the view that OpenSM needs.  We call this the "network view".

In the second case, MAD_IFC is used by various verbs to obtain data
regarding the local HCA (e.g., ib_query_device()).  We call this the
"host view".

This data needs to be paravirtualized.

MAD_IFC therefore needs a wrapper function, and also needs another
flag indicating whether it should provide the network view (when it is
called by ib_process_mad in special-qp packet handling), or the host
view (when it is called while implementing a verb).

There are currently 2 flag parameters in mlx4_MAD_IFC already:
ignore_bkey and ignore_mkey.  These two parameters are replaced by a
single "mad_ifc_flags" parameter, with different bits set for each
flag.  A third flag is added: "network-view/host-view".
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NRoland Dreier <roland@purestorage.com>
上级 37bfc7c1
......@@ -75,7 +75,7 @@ struct mlx4_rcv_tunnel_mad {
struct ib_mad mad;
} __packed;
int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int ignore_mkey, int ignore_bkey,
int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int mad_ifc_flags,
int port, struct ib_wc *in_wc, struct ib_grh *in_grh,
void *in_mad, void *response_mad)
{
......@@ -102,10 +102,13 @@ int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int ignore_mkey, int ignore_bkey,
* Key check traps can't be generated unless we have in_wc to
* tell us where to send the trap.
*/
if (ignore_mkey || !in_wc)
if ((mad_ifc_flags & MLX4_MAD_IFC_IGNORE_MKEY) || !in_wc)
op_modifier |= 0x1;
if (ignore_bkey || !in_wc)
if ((mad_ifc_flags & MLX4_MAD_IFC_IGNORE_BKEY) || !in_wc)
op_modifier |= 0x2;
if (mlx4_is_mfunc(dev->dev) &&
(mad_ifc_flags & MLX4_MAD_IFC_NET_VIEW || in_wc))
op_modifier |= 0x8;
if (in_wc) {
struct {
......@@ -138,10 +141,10 @@ int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int ignore_mkey, int ignore_bkey,
in_modifier |= in_wc->slid << 16;
}
err = mlx4_cmd_box(dev->dev, inmailbox->dma, outmailbox->dma,
in_modifier, op_modifier,
err = mlx4_cmd_box(dev->dev, inmailbox->dma, outmailbox->dma, in_modifier,
mlx4_is_master(dev->dev) ? (op_modifier & ~0x8) : op_modifier,
MLX4_CMD_MAD_IFC, MLX4_CMD_TIME_CLASS_C,
MLX4_CMD_NATIVE);
(op_modifier & 0x8) ? MLX4_CMD_NATIVE : MLX4_CMD_WRAPPED);
if (!err)
memcpy(response_mad, outmailbox->buf, 256);
......@@ -614,8 +617,9 @@ static int ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
prev_lid = pattr.lid;
err = mlx4_MAD_IFC(to_mdev(ibdev),
mad_flags & IB_MAD_IGNORE_MKEY,
mad_flags & IB_MAD_IGNORE_BKEY,
(mad_flags & IB_MAD_IGNORE_MKEY ? MLX4_MAD_IFC_IGNORE_MKEY : 0) |
(mad_flags & IB_MAD_IGNORE_BKEY ? MLX4_MAD_IFC_IGNORE_BKEY : 0) |
MLX4_MAD_IFC_NET_VIEW,
port_num, in_wc, in_grh, in_mad, out_mad);
if (err)
return IB_MAD_RESULT_FAILURE;
......
......@@ -98,7 +98,8 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
init_query_mad(in_mad);
in_mad->attr_id = IB_SMP_ATTR_NODE_INFO;
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, 1, NULL, NULL, in_mad, out_mad);
err = mlx4_MAD_IFC(to_mdev(ibdev), MLX4_MAD_IFC_IGNORE_KEYS,
1, NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......@@ -182,11 +183,12 @@ mlx4_ib_port_link_layer(struct ib_device *device, u8 port_num)
}
static int ib_link_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props)
struct ib_port_attr *props, int netw_view)
{
struct ib_smp *in_mad = NULL;
struct ib_smp *out_mad = NULL;
int ext_active_speed;
int mad_ifc_flags = MLX4_MAD_IFC_IGNORE_KEYS;
int err = -ENOMEM;
in_mad = kzalloc(sizeof *in_mad, GFP_KERNEL);
......@@ -198,7 +200,10 @@ static int ib_link_query_port(struct ib_device *ibdev, u8 port,
in_mad->attr_id = IB_SMP_ATTR_PORT_INFO;
in_mad->attr_mod = cpu_to_be32(port);
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL,
if (mlx4_is_mfunc(to_mdev(ibdev)->dev) && netw_view)
mad_ifc_flags |= MLX4_MAD_IFC_NET_VIEW;
err = mlx4_MAD_IFC(to_mdev(ibdev), mad_ifc_flags, port, NULL, NULL,
in_mad, out_mad);
if (err)
goto out;
......@@ -211,7 +216,10 @@ static int ib_link_query_port(struct ib_device *ibdev, u8 port,
props->state = out_mad->data[32] & 0xf;
props->phys_state = out_mad->data[33] >> 4;
props->port_cap_flags = be32_to_cpup((__be32 *) (out_mad->data + 20));
props->gid_tbl_len = to_mdev(ibdev)->dev->caps.gid_table_len[port];
if (netw_view)
props->gid_tbl_len = out_mad->data[50];
else
props->gid_tbl_len = to_mdev(ibdev)->dev->caps.gid_table_len[port];
props->max_msg_sz = to_mdev(ibdev)->dev->caps.max_msg_sz;
props->pkey_tbl_len = to_mdev(ibdev)->dev->caps.pkey_table_len[port];
props->bad_pkey_cntr = be16_to_cpup((__be16 *) (out_mad->data + 46));
......@@ -244,7 +252,7 @@ static int ib_link_query_port(struct ib_device *ibdev, u8 port,
in_mad->attr_id = MLX4_ATTR_EXTENDED_PORT_INFO;
in_mad->attr_mod = cpu_to_be32(port);
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port,
err = mlx4_MAD_IFC(to_mdev(ibdev), mad_ifc_flags, port,
NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......@@ -270,7 +278,7 @@ static u8 state_to_phys_state(enum ib_port_state state)
}
static int eth_link_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props)
struct ib_port_attr *props, int netw_view)
{
struct mlx4_ib_dev *mdev = to_mdev(ibdev);
......@@ -320,20 +328,27 @@ static int eth_link_query_port(struct ib_device *ibdev, u8 port,
return err;
}
static int mlx4_ib_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props)
int __mlx4_ib_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props, int netw_view)
{
int err;
memset(props, 0, sizeof *props);
err = mlx4_ib_port_link_layer(ibdev, port) == IB_LINK_LAYER_INFINIBAND ?
ib_link_query_port(ibdev, port, props) :
eth_link_query_port(ibdev, port, props);
ib_link_query_port(ibdev, port, props, netw_view) :
eth_link_query_port(ibdev, port, props, netw_view);
return err;
}
static int mlx4_ib_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props)
{
/* returns host view */
return __mlx4_ib_query_port(ibdev, port, props, 0);
}
static int __mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid)
{
......@@ -350,7 +365,8 @@ static int __mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
in_mad->attr_id = IB_SMP_ATTR_PORT_INFO;
in_mad->attr_mod = cpu_to_be32(port);
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad, out_mad);
err = mlx4_MAD_IFC(to_mdev(ibdev), MLX4_MAD_IFC_IGNORE_KEYS, port,
NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......@@ -360,7 +376,8 @@ static int __mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
in_mad->attr_id = IB_SMP_ATTR_GUID_INFO;
in_mad->attr_mod = cpu_to_be32(index / 8);
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad, out_mad);
err = mlx4_MAD_IFC(to_mdev(ibdev), MLX4_MAD_IFC_IGNORE_KEYS, port,
NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......@@ -391,11 +408,12 @@ static int mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
return iboe_query_gid(ibdev, port, index, gid);
}
static int mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey)
int __mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey, int netw_view)
{
struct ib_smp *in_mad = NULL;
struct ib_smp *out_mad = NULL;
int mad_ifc_flags = MLX4_MAD_IFC_IGNORE_KEYS;
int err = -ENOMEM;
in_mad = kzalloc(sizeof *in_mad, GFP_KERNEL);
......@@ -407,7 +425,11 @@ static int mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
in_mad->attr_id = IB_SMP_ATTR_PKEY_TABLE;
in_mad->attr_mod = cpu_to_be32(index / 32);
err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port, NULL, NULL, in_mad, out_mad);
if (mlx4_is_mfunc(to_mdev(ibdev)->dev) && netw_view)
mad_ifc_flags |= MLX4_MAD_IFC_NET_VIEW;
err = mlx4_MAD_IFC(to_mdev(ibdev), mad_ifc_flags, port, NULL, NULL,
in_mad, out_mad);
if (err)
goto out;
......@@ -419,6 +441,11 @@ static int mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
return err;
}
static int mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index, u16 *pkey)
{
return __mlx4_ib_query_pkey(ibdev, port, index, pkey, 0);
}
static int mlx4_ib_modify_device(struct ib_device *ibdev, int mask,
struct ib_device_modify *props)
{
......@@ -849,6 +876,7 @@ static int init_node_data(struct mlx4_ib_dev *dev)
{
struct ib_smp *in_mad = NULL;
struct ib_smp *out_mad = NULL;
int mad_ifc_flags = MLX4_MAD_IFC_IGNORE_KEYS;
int err = -ENOMEM;
in_mad = kzalloc(sizeof *in_mad, GFP_KERNEL);
......@@ -858,8 +886,10 @@ static int init_node_data(struct mlx4_ib_dev *dev)
init_query_mad(in_mad);
in_mad->attr_id = IB_SMP_ATTR_NODE_DESC;
if (mlx4_is_master(dev->dev))
mad_ifc_flags |= MLX4_MAD_IFC_NET_VIEW;
err = mlx4_MAD_IFC(dev, 1, 1, 1, NULL, NULL, in_mad, out_mad);
err = mlx4_MAD_IFC(dev, mad_ifc_flags, 1, NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......@@ -867,7 +897,7 @@ static int init_node_data(struct mlx4_ib_dev *dev)
in_mad->attr_id = IB_SMP_ATTR_NODE_INFO;
err = mlx4_MAD_IFC(dev, 1, 1, 1, NULL, NULL, in_mad, out_mad);
err = mlx4_MAD_IFC(dev, mad_ifc_flags, 1, NULL, NULL, in_mad, out_mad);
if (err)
goto out;
......
......@@ -176,6 +176,14 @@ enum mlx4_ib_qp_type {
MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER | \
MLX4_IB_QPT_TUN_SMI | MLX4_IB_QPT_TUN_GSI)
enum mlx4_ib_mad_ifc_flags {
MLX4_MAD_IFC_IGNORE_MKEY = 1,
MLX4_MAD_IFC_IGNORE_BKEY = 2,
MLX4_MAD_IFC_IGNORE_KEYS = (MLX4_MAD_IFC_IGNORE_MKEY |
MLX4_MAD_IFC_IGNORE_BKEY),
MLX4_MAD_IFC_NET_VIEW = 4,
};
enum {
MLX4_NUM_TUNNEL_BUFS = 256,
};
......@@ -512,7 +520,7 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
int mlx4_ib_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
struct ib_recv_wr **bad_wr);
int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int ignore_mkey, int ignore_bkey,
int mlx4_MAD_IFC(struct mlx4_ib_dev *dev, int mad_ifc_flags,
int port, struct ib_wc *in_wc, struct ib_grh *in_grh,
void *in_mad, void *response_mad);
int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
......@@ -527,6 +535,10 @@ int mlx4_ib_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list, int npages,
u64 iova);
int mlx4_ib_unmap_fmr(struct list_head *fmr_list);
int mlx4_ib_fmr_dealloc(struct ib_fmr *fmr);
int __mlx4_ib_query_port(struct ib_device *ibdev, u8 port,
struct ib_port_attr *props, int netw_view);
int __mlx4_ib_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey, int netw_view);
int mlx4_ib_resolve_grh(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah_attr,
u8 *mac, int *is_mcast, u8 port);
......
......@@ -40,6 +40,7 @@
#include <linux/mlx4/cmd.h>
#include <linux/semaphore.h>
#include <rdma/ib_smi.h>
#include <asm/io.h>
......@@ -627,6 +628,149 @@ static int mlx4_ACCESS_MEM(struct mlx4_dev *dev, u64 master_addr,
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
}
static int query_pkey_block(struct mlx4_dev *dev, u8 port, u16 index, u16 *pkey,
struct mlx4_cmd_mailbox *inbox,
struct mlx4_cmd_mailbox *outbox)
{
struct ib_smp *in_mad = (struct ib_smp *)(inbox->buf);
struct ib_smp *out_mad = (struct ib_smp *)(outbox->buf);
int err;
int i;
if (index & 0x1f)
return -EINVAL;
in_mad->attr_mod = cpu_to_be32(index / 32);
err = mlx4_cmd_box(dev, inbox->dma, outbox->dma, port, 3,
MLX4_CMD_MAD_IFC, MLX4_CMD_TIME_CLASS_C,
MLX4_CMD_NATIVE);
if (err)
return err;
for (i = 0; i < 32; ++i)
pkey[i] = be16_to_cpu(((__be16 *) out_mad->data)[i]);
return err;
}
static int get_full_pkey_table(struct mlx4_dev *dev, u8 port, u16 *table,
struct mlx4_cmd_mailbox *inbox,
struct mlx4_cmd_mailbox *outbox)
{
int i;
int err;
for (i = 0; i < dev->caps.pkey_table_len[port]; i += 32) {
err = query_pkey_block(dev, port, i, table + i, inbox, outbox);
if (err)
return err;
}
return 0;
}
#define PORT_CAPABILITY_LOCATION_IN_SMP 20
#define PORT_STATE_OFFSET 32
static enum ib_port_state vf_port_state(struct mlx4_dev *dev, int port, int vf)
{
/* will be modified when add alias_guid feature */
return IB_PORT_DOWN;
}
static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
struct mlx4_vhcr *vhcr,
struct mlx4_cmd_mailbox *inbox,
struct mlx4_cmd_mailbox *outbox,
struct mlx4_cmd_info *cmd)
{
struct ib_smp *smp = inbox->buf;
u32 index;
u8 port;
u16 *table;
int err;
int vidx, pidx;
struct mlx4_priv *priv = mlx4_priv(dev);
struct ib_smp *outsmp = outbox->buf;
__be16 *outtab = (__be16 *)(outsmp->data);
__be32 slave_cap_mask;
port = vhcr->in_modifier;
if (smp->base_version == 1 &&
smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
smp->class_version == 1) {
if (smp->method == IB_MGMT_METHOD_GET) {
if (smp->attr_id == IB_SMP_ATTR_PKEY_TABLE) {
index = be32_to_cpu(smp->attr_mod);
if (port < 1 || port > dev->caps.num_ports)
return -EINVAL;
table = kcalloc(dev->caps.pkey_table_len[port], sizeof *table, GFP_KERNEL);
if (!table)
return -ENOMEM;
/* need to get the full pkey table because the paravirtualized
* pkeys may be scattered among several pkey blocks.
*/
err = get_full_pkey_table(dev, port, table, inbox, outbox);
if (!err) {
for (vidx = index * 32; vidx < (index + 1) * 32; ++vidx) {
pidx = priv->virt2phys_pkey[slave][port - 1][vidx];
outtab[vidx % 32] = cpu_to_be16(table[pidx]);
}
}
kfree(table);
return err;
}
if (smp->attr_id == IB_SMP_ATTR_PORT_INFO) {
/*get the slave specific caps:*/
/*do the command */
err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
vhcr->in_modifier, vhcr->op_modifier,
vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
/* modify the response for slaves */
if (!err && slave != mlx4_master_func_num(dev)) {
u8 *state = outsmp->data + PORT_STATE_OFFSET;
*state = (*state & 0xf0) | vf_port_state(dev, port, slave);
slave_cap_mask = priv->mfunc.master.slave_state[slave].ib_cap_mask[port];
memcpy(outsmp->data + PORT_CAPABILITY_LOCATION_IN_SMP, &slave_cap_mask, 4);
}
return err;
}
if (smp->attr_id == IB_SMP_ATTR_GUID_INFO) {
/* compute slave's gid block */
smp->attr_mod = cpu_to_be32(slave / 8);
/* execute cmd */
err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
vhcr->in_modifier, vhcr->op_modifier,
vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
if (!err) {
/* if needed, move slave gid to index 0 */
if (slave % 8)
memcpy(outsmp->data,
outsmp->data + (slave % 8) * 8, 8);
/* delete all other gids */
memset(outsmp->data + 8, 0, 56);
}
return err;
}
}
}
if (slave != mlx4_master_func_num(dev) &&
((smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) ||
(smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
smp->method == IB_MGMT_METHOD_SET))) {
mlx4_err(dev, "slave %d is trying to execute a Subnet MGMT MAD, "
"class 0x%x, method 0x%x for attr 0x%x. Rejecting\n",
slave, smp->method, smp->mgmt_class,
be16_to_cpu(smp->attr_id));
return -EPERM;
}
/*default:*/
return mlx4_cmd_box(dev, inbox->dma, outbox->dma,
vhcr->in_modifier, vhcr->op_modifier,
vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
}
int mlx4_DMA_wrapper(struct mlx4_dev *dev, int slave,
struct mlx4_vhcr *vhcr,
struct mlx4_cmd_mailbox *inbox,
......@@ -1060,6 +1204,24 @@ static struct mlx4_cmd_info cmd_info[] = {
.verify = NULL,
.wrapper = mlx4_GEN_QP_wrapper
},
{
.opcode = MLX4_CMD_CONF_SPECIAL_QP,
.has_inbox = false,
.has_outbox = false,
.out_is_imm = false,
.encode_slave_id = false,
.verify = NULL, /* XXX verify: only demux can do this */
.wrapper = NULL
},
{
.opcode = MLX4_CMD_MAD_IFC,
.has_inbox = true,
.has_outbox = true,
.out_is_imm = false,
.encode_slave_id = false,
.verify = NULL,
.wrapper = mlx4_MAD_IFC_wrapper
},
{
.opcode = MLX4_CMD_QUERY_IF_STAT,
.has_inbox = false,
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册