freebsd/freebsd-src
1
0 Fork 0
mirror of https://github.com/freebsd/freebsd-src.git synced 2026-11-07 04:36:52 -10:00
Code Issues Packages Projects Releases Wiki Activity

OFED: Various changes from Linux 4.17

Browse Source 
This contains changes from the following Linux commits:

a3b641af72ba RDMA/CM: move rdma_id_private to cma_priv.h
6612b4983f7e IB/core: Fix comments of GID query functions
b26c4a1138df IB/{core, ipoib}: Simplify ib_find_gid() for unused ndev
5ac08a341303 IB/cma: Use rdma_protocol_roce() and remove cma_protocol_roce_dev_port()
a9c06aeba997 IB/core: Remove rdma_resolve_ip_route() as exported symbol
a22af59ea9a5 IB/cm: Add and use a helper function to add cm_id's to the port list
cb12a8e2fa5f IB/cm: Introduce and use helper function to get cm_port from path
0a5141593567 IB/core: Refactor ib_init_ah_attr_from_path() for RoCE
e41a7c41947d IB/core: Move rdma_addr_find_l2_eth_by_grh to core_priv.h
98f1f4e0ed26 IB/core: Refer to RoCE port property instead of GID table property
114cc9c4b182 IB/cma: Resolve route only while receiving CM requests
9137108cc3d6 RDMA/rdma_cm: Fix use after free race with process_one_req
190fb9c4d130 IB/core: Refer to RoCE port property to decide building cache
22d24f75a193 IB/core: Search GID only for IB link layer
4ab7cb4bf362 IB/core: Refer to RoCE port property instead of GID table property
3401857ea347 IB/core: Generate GID change event regardless of RoCE GID table property
218b9e3eb8b5 RDMA/cma: Move rdma_cm_state to cma_priv.h
8435168d50e6 RDMA/ucma: Don't allow setting RDMA_OPTION_IB_PATH without an RDMA device
72e1ff0fb7e0 RDMA/core: Update query_gid documentation for HCA drivers
0e1f9b924471 RDMA/providers: Simplify query_gid callback of RoCE providers
598ff6bae689 IB/core: Refactor GID modify code for RoCE
f35faa4ba956 IB/core: Simplify ib_query_gid to always refer to cache
14169e333e71 IB/providers: Avoid zero GID check for RoCE
414448d249d8 RDMA: Use ib_gid_attr during GID modification
ee6548d1d98d RDMA/rdma_cm: Delete rdma_addr_client
44e75052bc2a RDMA/rdma_cm: Make rdma_addr_cancel into a fence
e19c0d237873 RDMA/rdma_cm: Remove process_req and timer sorting
a66ed149b0da IB/core: Don't allow default GID addition at non reseved slots
dc5640f294e4 IB/core: Fix deleting default GIDs when changing mac adddress
22c01ee4b8a8 IB/core: Fix to avoid deleting IPv6 look alike default GIDs
9aa169213d11 RDMA/cma: Do not query GID during QP state transition to RTR
aec05afe641b IB/core: Remove redundant return

Tested by:	Wafa Hamzah <[email protected]> (mlx5_ib)
Tested by:	John Baldwin <[email protected]> (iw_cxgbe)
Sponsored by:	Chelsio Communications
This commit is contained in:
John Baldwin committed 2026-06-24 15:30:56 -04:00
1 parent 1d99feddd6
commit 75cde38e94
28 files changed
+709 -764
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

sys/dev/bnxt/bnxt_re/ib_verbs.c
+10 -12
View File
@@ -443,12 +443,11 @@ int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
return rc;
}
int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
unsigned int index, void **context)
int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context)
{
int rc = 0;
struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
struct bnxt_qplib_gid *gid_to_del;
u16 vlan_id = 0xFFFF;
@@ -482,14 +481,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
dev_dbg(rdev_to_dev(rdev),
"Trying to delete GID0 while QP1 is alive\n");
if (!ctx->refcnt) {
rdev->gid_map[index] = -1;
rdev->gid_map[attr->index] = -1;
ctx_tbl = sgid_tbl->ctx;
ctx_tbl[ctx->idx] = NULL;
kfree(ctx);
}
return 0;
}
rdev->gid_map[index] = -1;
rdev->gid_map[attr->index] = -1;
if (!ctx->refcnt) {
rc = bnxt_qplib_del_sgid(sgid_tbl, gid_to_del,
vlan_id, true);
@@ -510,15 +509,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
return rc;
}
int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
unsigned int index, const union ib_gid *gid,
int bnxt_re_add_gid(const union ib_gid *gid,
const struct ib_gid_attr *attr, void **context)
{
int rc;
u32 tbl_idx = 0;
u16 vlan_id = 0xFFFF;
struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
if ((attr->ndev) && is_vlan_dev(attr->ndev))
vlan_id = vlan_dev_vlan_id(attr->ndev);
@@ -540,8 +538,8 @@ int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
ctx_tbl[tbl_idx]->refcnt++;
}
*context = ctx_tbl[tbl_idx];
/* tbl_idx is the HW table index and index is the stack index */
rdev->gid_map[index] = tbl_idx;
/* tbl_idx is the HW table index and attr->index is the stack index */
rdev->gid_map[attr->index] = tbl_idx;
return 0;
} else if (rc < 0) {
dev_err(rdev_to_dev(rdev), "Add GID failed rc = 0x%x\n", rc);
@@ -556,8 +554,8 @@ int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
ctx->idx = tbl_idx;
ctx->refcnt = 1;
ctx_tbl[tbl_idx] = ctx;
/* tbl_idx is the HW table index and index is the stack index */
rdev->gid_map[index] = tbl_idx;
/* tbl_idx is the HW table index and attr->index is the stack index */
rdev->gid_map[attr->index] = tbl_idx;
*context = ctx;
}
return rc;
sys/dev/bnxt/bnxt_re/ib_verbs.h
+2 -4
View File
@@ -295,10 +295,8 @@ int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num,
void bnxt_re_compat_qfwstr(void);
int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num,
u16 index, u16 *pkey);
int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
unsigned int index, void **context);
int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
unsigned int index, const union ib_gid *gid,
int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context);
int bnxt_re_add_gid(const union ib_gid *gid,
const struct ib_gid_attr *attr, void **context);
int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
int index, union ib_gid *gid);
sys/dev/bnxt/bnxt_re/main.c
-1
View File
@@ -2179,7 +2179,6 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
ibdev->modify_port = bnxt_re_modify_port;
ibdev->get_port_immutable = bnxt_re_get_port_immutable;
ibdev->query_pkey = bnxt_re_query_pkey;
ibdev->query_gid = bnxt_re_query_gid;
ibdev->get_netdev = bnxt_re_get_netdev;
ibdev->add_gid = bnxt_re_add_gid;
ibdev->del_gid = bnxt_re_del_gid;
sys/dev/irdma/fbsd_kcompat.h
+2 -4
View File
@@ -156,8 +156,6 @@ int irdma_iw_port_immutable(struct ib_device *ibdev, u8 port_num,
struct ib_port_immutable *immutable);
int irdma_query_gid(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid);
int irdma_query_gid_roce(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid);
int irdma_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
u16 *pkey);
int irdma_query_port(struct ib_device *ibdev, u8 port,
@@ -255,9 +253,9 @@ int irdma_alloc_ucontext(struct ib_ucontext *uctx, struct ib_udata *udata);
void irdma_dealloc_ucontext(struct ib_ucontext *context);
int irdma_alloc_pd(struct ib_pd *pd, struct ib_udata *udata);
void irdma_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata);
int irdma_add_gid(struct ib_device *, u8, unsigned int, const union ib_gid *,
int irdma_add_gid(const union ib_gid *,
const struct ib_gid_attr *, void **);
int irdma_del_gid(struct ib_device *, u8, unsigned int, void **);
int irdma_del_gid(const struct ib_gid_attr *, void **);
struct ib_device *ib_device_get_by_netdev(struct ifnet *ndev, int driver_id);
void ib_device_put(struct ib_device *device);
void ib_unregister_device_put(struct ib_device *device);
sys/dev/irdma/irdma_kcompat.c
+2 -29
View File
@@ -77,10 +77,7 @@ irdma_get_dev_fw_str(struct ib_device *dev,
}
int
irdma_add_gid(struct ib_device *device,
u8 port_num,
unsigned int index,
const union ib_gid *gid,
irdma_add_gid(const union ib_gid *gid,
const struct ib_gid_attr *attr,
void **context)
{
@@ -88,9 +85,7 @@ irdma_add_gid(struct ib_device *device,
}
int
irdma_del_gid(struct ib_device *device,
u8 port_num,
unsigned int index,
irdma_del_gid(const struct ib_gid_attr *attr,
void **context)
{
return 0;
@@ -1516,28 +1511,6 @@ ib_unregister_device_put(struct ib_device *device)
ib_unregister_device(device);
}
/**
* irdma_query_gid_roce - Query port GID for Roce
* @ibdev: device pointer from stack
* @port: port number
* @index: Entry index
* @gid: Global ID
*/
int
irdma_query_gid_roce(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid)
{
int ret;
ret = ib_get_cached_gid(ibdev, port, index, gid, NULL);
if (ret == -EAGAIN) {
memcpy(gid, &zgid, sizeof(*gid));
return 0;
}
return ret;
}
/**
* irdma_modify_port - modify port attributes
* @ibdev: device pointer from stack
sys/dev/irdma/irdma_verbs.c
-1
View File
@@ -3328,7 +3328,6 @@ irdma_set_device_roce_ops(struct ib_device *ibdev)
dev_ops->get_port_immutable = irdma_roce_port_immutable;
dev_ops->modify_qp = irdma_modify_qp_roce;
dev_ops->query_ah = irdma_query_ah;
dev_ops->query_gid = irdma_query_gid_roce;
dev_ops->query_pkey = irdma_query_pkey;
ibdev->add_gid = irdma_add_gid;
ibdev->del_gid = irdma_del_gid;
sys/dev/mlx4/mlx4_ib/mlx4_ib_main.c
+13 -36
View File
@@ -252,14 +252,11 @@ static int mlx4_ib_update_gids(struct gid_entry *gids,
return mlx4_ib_update_gids_v1(gids, ibdev, port_num);
}
static int mlx4_ib_add_gid(struct ib_device *device,
u8 port_num,
unsigned int index,
const union ib_gid *gid,
static int mlx4_ib_add_gid(const union ib_gid *gid,
const struct ib_gid_attr *attr,
void **context)
{
struct mlx4_ib_dev *ibdev = to_mdev(device);
struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
struct mlx4_ib_iboe *iboe = &ibdev->iboe;
struct mlx4_port_gid_table *port_gid_table;
int free = -1, found = -1;
@@ -268,16 +265,16 @@ static int mlx4_ib_add_gid(struct ib_device *device,
int i;
struct gid_entry *gids = NULL;
if (!rdma_cap_roce_gid_table(device, port_num))
if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
return -EINVAL;
if (port_num > MLX4_MAX_PORTS)
if (attr->port_num > MLX4_MAX_PORTS)
return -EINVAL;
if (!context)
return -EINVAL;
port_gid_table = &iboe->gids[port_num - 1];
port_gid_table = &iboe->gids[attr->port_num - 1];
spin_lock_bh(&iboe->lock);
for (i = 0; i < MLX4_MAX_PORT_GIDS; ++i) {
if (!memcmp(&port_gid_table->gids[i].gid, gid, sizeof(*gid)) &&
@@ -324,33 +321,30 @@ static int mlx4_ib_add_gid(struct ib_device *device,
spin_unlock_bh(&iboe->lock);
if (!ret && hw_update) {
ret = mlx4_ib_update_gids(gids, ibdev, port_num);
ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
kfree(gids);
}
return ret;
}
static int mlx4_ib_del_gid(struct ib_device *device,
u8 port_num,
unsigned int index,
void **context)
static int mlx4_ib_del_gid(const struct ib_gid_attr *attr, void **context)
{
struct gid_cache_context *ctx = *context;
struct mlx4_ib_dev *ibdev = to_mdev(device);
struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
struct mlx4_ib_iboe *iboe = &ibdev->iboe;
struct mlx4_port_gid_table *port_gid_table;
int ret = 0;
int hw_update = 0;
struct gid_entry *gids = NULL;
if (!rdma_cap_roce_gid_table(device, port_num))
if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
return -EINVAL;
if (port_num > MLX4_MAX_PORTS)
if (attr->port_num > MLX4_MAX_PORTS)
return -EINVAL;
port_gid_table = &iboe->gids[port_num - 1];
port_gid_table = &iboe->gids[attr->port_num - 1];
spin_lock_bh(&iboe->lock);
if (ctx) {
ctx->refcount--;
@@ -382,7 +376,7 @@ static int mlx4_ib_del_gid(struct ib_device *device,
spin_unlock_bh(&iboe->lock);
if (!ret && hw_update) {
ret = mlx4_ib_update_gids(gids, ibdev, port_num);
ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
kfree(gids);
}
return ret;
@@ -417,9 +411,6 @@ int mlx4_ib_gid_index_to_real_index(struct mlx4_ib_dev *ibdev,
if (attr.ndev)
if_rele(attr.ndev);
if (!memcmp(&gid, &zgid, sizeof(gid)))
return -EINVAL;
spin_lock_irqsave(&iboe->lock, flags);
port_gid_table = &iboe->gids[port_num - 1];
@@ -825,24 +816,10 @@ out:
static int mlx4_ib_query_gid(struct ib_device *ibdev, u8 port, int index,
union ib_gid *gid)
{
int ret;
if (rdma_protocol_ib(ibdev, port))
return __mlx4_ib_query_gid(ibdev, port, index, gid, 0);
if (!rdma_protocol_roce(ibdev, port))
return -ENODEV;
if (!rdma_cap_roce_gid_table(ibdev, port))
return -ENODEV;
ret = ib_get_cached_gid(ibdev, port, index, gid, NULL);
if (ret == -EAGAIN) {
memcpy(gid, &zgid, sizeof(*gid));
return 0;
}
return ret;
return 0;
}
static int mlx4_ib_query_sl2vl(struct ib_device *ibdev, u8 port, u64 *sl2vl_tbl)
sys/dev/mlx4/mlx4_ib/mlx4_ib_qp.c
-2
View File
@@ -1787,8 +1787,6 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
status = ib_get_cached_gid(ibqp->device, port_num,
index, &gid, &gid_attr);
if (!status && !memcmp(&gid, &zgid, sizeof(gid)))
status = -ENOENT;
if (!status && gid_attr.ndev) {
vlan = rdma_vlan_dev_vlan_id(gid_attr.ndev);
memcpy(smac, if_getlladdr(gid_attr.ndev), ETH_ALEN);
sys/dev/mlx5/mlx5_ib/mlx5_ib_main.c
+7 -6
View File
@@ -428,18 +428,19 @@ static int set_roce_addr(struct ib_device *device, u8 port_num,
return mlx5_cmd_exec(dev->mdev, in, sizeof(in), out, sizeof(out));
}
static int mlx5_ib_add_gid(struct ib_device *device, u8 port_num,
unsigned int index, const union ib_gid *gid,
static int mlx5_ib_add_gid(const union ib_gid *gid,
const struct ib_gid_attr *attr,
__always_unused void **context)
{
return set_roce_addr(device, port_num, index, gid, attr);
return set_roce_addr(attr->device, attr->port_num, attr->index, gid,
attr);
}
static int mlx5_ib_del_gid(struct ib_device *device, u8 port_num,
unsigned int index, __always_unused void **context)
static int mlx5_ib_del_gid(const struct ib_gid_attr *attr,
__always_unused void **context)
{
return set_roce_addr(device, port_num, index, NULL, NULL);
return set_roce_addr(attr->device, attr->port_num, attr->index, NULL,
NULL);
}
__be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
sys/dev/qlnx/qlnxr/qlnxr_os.c
-1
View File
@@ -197,7 +197,6 @@ qlnxr_register_device(qlnxr_dev_t *dev)
ibdev->query_gid = qlnxr_iw_query_gid;
} else {
ibdev->node_type = RDMA_NODE_IB_CA;
ibdev->query_gid = qlnxr_query_gid;
ibdev->uverbs_cmd_mask |=
(1ull << IB_USER_VERBS_CMD_CREATE_SRQ) |
(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ) |
sys/dev/qlnx/qlnxr/qlnxr_verbs.c
-38
View File
@@ -118,44 +118,6 @@ qlnxr_iw_query_gid(struct ib_device *ibdev, u8 port, int index,
return 0;
}
int
qlnxr_query_gid(struct ib_device *ibdev, u8 port, int index,
union ib_gid *sgid)
{
struct qlnxr_dev *dev;
qlnx_host_t *ha;
dev = get_qlnxr_dev(ibdev);
ha = dev->ha;
QL_DPRINT12(ha, "enter index: %d\n", index);
#if 0
int ret = 0;
/* @@@: if DEFINE_ROCE_GID_TABLE to be used here */
//if (!rdma_cap_roce_gid_table(ibdev, port)) {
if (!(rdma_protocol_roce(ibdev, port) &&
ibdev->add_gid && ibdev->del_gid)) {
QL_DPRINT11(ha, "acquire gid failed\n");
return -ENODEV;
}
ret = ib_get_cached_gid(ibdev, port, index, sgid, NULL);
if (ret == -EAGAIN) {
memcpy(sgid, &zgid, sizeof(*sgid));
return 0;
}
#endif
if ((index >= QLNXR_MAX_SGID) || (index < 0)) {
QL_DPRINT12(ha, "invalid gid index %d\n", index);
memset(sgid, 0, sizeof(*sgid));
return -EINVAL;
}
memcpy(sgid, &dev->sgid_tbl[index], sizeof(*sgid));
QL_DPRINT12(ha, "exit : %p\n", sgid);
return 0;
}
int
qlnxr_create_srq(struct ib_srq *ibsrq,
struct ib_srq_init_attr *init_attr,
sys/dev/qlnx/qlnxr/qlnxr_verbs.h
-5
View File
@@ -34,11 +34,6 @@ extern int qlnxr_iw_query_gid(struct ib_device *,
int index,
union ib_gid *gid);
extern int qlnxr_query_gid(struct ib_device *,
u8 port,
int index,
union ib_gid *gid);
extern int qlnxr_create_srq(struct ib_srq *ibsrq,
struct ib_srq_init_attr *,
struct ib_udata *);
sys/ofed/drivers/infiniband/core/cma_priv.h
+95
View File
@@ -0,0 +1,95 @@
/*
* Copyright (c) 2005 Voltaire Inc. All rights reserved.
* Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved.
* Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2005-2006 Intel Corporation. All rights reserved.
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef _CMA_PRIV_H
#define _CMA_PRIV_H
enum rdma_cm_state {
RDMA_CM_IDLE,
RDMA_CM_ADDR_QUERY,
RDMA_CM_ADDR_RESOLVED,
RDMA_CM_ROUTE_QUERY,
RDMA_CM_ROUTE_RESOLVED,
RDMA_CM_CONNECT,
RDMA_CM_DISCONNECT,
RDMA_CM_ADDR_BOUND,
RDMA_CM_LISTEN,
RDMA_CM_DEVICE_REMOVAL,
RDMA_CM_DESTROYING
};
struct rdma_id_private {
struct rdma_cm_id id;
struct rdma_bind_list *bind_list;
struct hlist_node node;
struct list_head list; /* listen_any_list or cma_device.list */
struct list_head listen_list; /* per device listens */
struct cma_device *cma_dev;
struct list_head mc_list;
int internal_id;
enum rdma_cm_state state;
spinlock_t lock;
struct mutex qp_mutex;
struct completion comp;
atomic_t refcount;
struct mutex handler_mutex;
int backlog;
int timeout_ms;
struct ib_sa_query *query;
int query_id;
union {
struct ib_cm_id *ib;
struct iw_cm_id *iw;
} cm_id;
u32 seq_num;
u32 qkey;
u32 qp_num;
pid_t owner;
u32 options;
u8 srq;
u8 tos;
u8 tos_set:1;
u8 timeout_set:1;
u8 reuseaddr;
u8 afonly;
u8 timeout;
enum ib_gid_type gid_type;
};
#endif /* _CMA_PRIV_H */
sys/ofed/drivers/infiniband/core/core_priv.h
+10
View File
@@ -183,6 +183,16 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev,
return qp;
}
struct rdma_dev_addr;
int rdma_resolve_ip_route(struct sockaddr *src_addr,
const struct sockaddr *dst_addr,
struct rdma_dev_addr *addr);
int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
const union ib_gid *dgid,
u8 *dmac, if_t ndev,
int *hoplimit);
struct rdma_umap_priv {
struct vm_area_struct *vma;
struct list_head list;
sys/ofed/drivers/infiniband/core/ib_addr.c
+48 -100
View File
@@ -61,7 +61,6 @@ struct addr_req {
struct sockaddr_storage src_addr;
struct sockaddr_storage dst_addr;
struct rdma_dev_addr *addr;
struct rdma_addr_client *client;
void *context;
void (*callback)(int status, struct sockaddr *src_addr,
struct rdma_dev_addr *addr, void *context);
@@ -70,11 +69,8 @@ struct addr_req {
int status;
};
static void process_req(struct work_struct *work);
static DEFINE_MUTEX(lock);
static DEFINE_SPINLOCK(lock);
static LIST_HEAD(req_list);
static DECLARE_DELAYED_WORK(work, process_req);
static struct workqueue_struct *addr_wq;
int rdma_addr_size(struct sockaddr *addr)
@@ -108,28 +104,6 @@ int rdma_addr_size_kss(struct sockaddr_storage *addr)
}
EXPORT_SYMBOL(rdma_addr_size_kss);
static struct rdma_addr_client self;
void rdma_addr_register_client(struct rdma_addr_client *client)
{
atomic_set(&client->refcount, 1);
init_completion(&client->comp);
}
EXPORT_SYMBOL(rdma_addr_register_client);
static inline void put_client(struct rdma_addr_client *client)
{
if (atomic_dec_and_test(&client->refcount))
complete(&client->comp);
}
void rdma_addr_unregister_client(struct rdma_addr_client *client)
{
put_client(client);
wait_for_completion(&client->comp);
}
EXPORT_SYMBOL(rdma_addr_unregister_client);
static inline void
rdma_copy_addr_sub(u8 *dst, const u8 *src, unsigned min, unsigned max)
{
@@ -211,7 +185,7 @@ int rdma_translate_ip(const struct sockaddr *addr,
}
EXPORT_SYMBOL(rdma_translate_ip);
static void set_timeout(struct delayed_work *delayed_work, unsigned long time)
static void set_timeout(struct addr_req *req, unsigned long time)
{
unsigned long delay;
@@ -221,23 +195,15 @@ static void set_timeout(struct delayed_work *delayed_work, unsigned long time)
else if (delay > hz)
delay = hz;
mod_delayed_work(addr_wq, delayed_work, delay);
mod_delayed_work(addr_wq, &req->work, delay);
}
static void queue_req(struct addr_req *req)
{
struct addr_req *temp_req;
mutex_lock(&lock);
list_for_each_entry_reverse(temp_req, &req_list, list) {
if (time_after_eq(req->timeout, temp_req->timeout))
break;
}
list_add(&req->list, &temp_req->list);
set_timeout(&req->work, req->timeout);
mutex_unlock(&lock);
spin_lock_bh(&lock);
list_add_tail(&req->list, &req_list);
set_timeout(req, req->timeout);
spin_unlock_bh(&lock);
}
#if defined(INET) || defined(INET6)
@@ -714,7 +680,6 @@ static void process_one_req(struct work_struct *_work)
struct addr_req *req;
struct sockaddr *src_in, *dst_in;
mutex_lock(&lock);
req = container_of(_work, struct addr_req, work.work);
if (req->status == -ENODATA) {
@@ -725,62 +690,33 @@ static void process_one_req(struct work_struct *_work)
req->status = -ETIMEDOUT;
} else if (req->status == -ENODATA) {
/* requeue the work for retrying again */
set_timeout(&req->work, req->timeout);
mutex_unlock(&lock);
spin_lock_bh(&lock);
if (!list_empty(&req->list))
set_timeout(req, req->timeout);
spin_unlock_bh(&lock);
return;
}
}
list_del(&req->list);
mutex_unlock(&lock);
req->callback(req->status, (struct sockaddr *)&req->src_addr,
req->addr, req->context);
put_client(req->client);
kfree(req);
}
req->callback = NULL;
static void process_req(struct work_struct *work)
{
struct addr_req *req, *temp_req;
struct sockaddr *src_in, *dst_in;
struct list_head done_list;
INIT_LIST_HEAD(&done_list);
mutex_lock(&lock);
list_for_each_entry_safe(req, temp_req, &req_list, list) {
if (req->status == -ENODATA) {
src_in = (struct sockaddr *) &req->src_addr;
dst_in = (struct sockaddr *) &req->dst_addr;
req->status = addr_resolve(src_in, dst_in, req->addr);
if (req->status && time_after_eq(jiffies, req->timeout))
req->status = -ETIMEDOUT;
else if (req->status == -ENODATA) {
set_timeout(&req->work, req->timeout);
continue;
}
}
list_move_tail(&req->list, &done_list);
}
mutex_unlock(&lock);
list_for_each_entry_safe(req, temp_req, &done_list, list) {
list_del(&req->list);
/* It is safe to cancel other work items from this work item
* because at a time there can be only one work item running
* with this single threaded work queue.
spin_lock_bh(&lock);
if (!list_empty(&req->list)) {
/*
* Although the work will normally have been canceled by the
* workqueue, it can still be requeued as long as it is on the
* req_list.
*/
cancel_delayed_work(&req->work);
req->callback(req->status, (struct sockaddr *) &req->src_addr,
req->addr, req->context);
put_client(req->client);
list_del_init(&req->list);
kfree(req);
}
spin_unlock_bh(&lock);
}
int rdma_resolve_ip(struct rdma_addr_client *client,
struct sockaddr *src_addr, struct sockaddr *dst_addr,
int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr,
struct rdma_dev_addr *addr, int timeout_ms,
void (*callback)(int status, struct sockaddr *src_addr,
struct rdma_dev_addr *addr, void *context),
@@ -812,8 +748,6 @@ int rdma_resolve_ip(struct rdma_addr_client *client,
req->addr = addr;
req->callback = callback;
req->context = context;
req->client = client;
atomic_inc(&client->refcount);
INIT_DELAYED_WORK(&req->work, process_one_req);
req->status = addr_resolve(src_in, dst_in, addr);
@@ -828,7 +762,6 @@ int rdma_resolve_ip(struct rdma_addr_client *client,
break;
default:
ret = req->status;
atomic_dec(&client->refcount);
goto err;
}
return ret;
@@ -856,23 +789,40 @@ int rdma_resolve_ip_route(struct sockaddr *src_addr,
return addr_resolve(src_in, dst_addr, addr);
}
EXPORT_SYMBOL(rdma_resolve_ip_route);
void rdma_addr_cancel(struct rdma_dev_addr *addr)
{
struct addr_req *req, *temp_req;
struct addr_req *found = NULL;
mutex_lock(&lock);
spin_lock_bh(&lock);
list_for_each_entry_safe(req, temp_req, &req_list, list) {
if (req->addr == addr) {
req->status = -ECANCELED;
req->timeout = jiffies;
list_move(&req->list, &req_list);
set_timeout(&req->work, req->timeout);
/*
* Removing from the list means we take ownership of
* the req
*/
list_del_init(&req->list);
found = req;
break;
}
}
mutex_unlock(&lock);
spin_unlock_bh(&lock);
if (!found)
return;
/*
* sync canceling the work after removing it from the req_list
* guarentees no work is running and none will be started.
*/
cancel_delayed_work_sync(&found->work);
if (found->callback)
found->callback(-ECANCELED, (struct sockaddr *)&found->src_addr,
found->addr, found->context);
kfree(found);
}
EXPORT_SYMBOL(rdma_addr_cancel);
@@ -907,8 +857,8 @@ int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
dev_addr.net = dev_net(ndev);
init_completion(&ctx.comp);
ret = rdma_resolve_ip(&self, &sgid_addr._sockaddr, &dgid_addr._sockaddr,
&dev_addr, 1000, resolve_cb, &ctx);
ret = rdma_resolve_ip(&sgid_addr._sockaddr, &dgid_addr._sockaddr,
&dev_addr, 1000, resolve_cb, &ctx);
if (ret)
return ret;
@@ -929,13 +879,11 @@ int addr_init(void)
if (!addr_wq)
return -ENOMEM;
rdma_addr_register_client(&self);
return 0;
}
void addr_cleanup(void)
{
rdma_addr_unregister_client(&self);
destroy_workqueue(addr_wq);
WARN_ON(!list_empty(&req_list));
}
sys/ofed/drivers/infiniband/core/ib_cache.c
+307 -246
View File
@@ -62,8 +62,6 @@ struct ib_update_work {
union ib_gid zgid;
EXPORT_SYMBOL(zgid);
static const struct ib_gid_attr zattr;
enum gid_attr_find_mask {
GID_ATTR_FIND_MASK_GID = 1UL << 0,
GID_ATTR_FIND_MASK_NETDEV = 1UL << 1,
@@ -76,15 +74,6 @@ enum gid_table_entry_props {
GID_TABLE_ENTRY_DEFAULT = 1UL << 1,
};
enum gid_table_write_action {
GID_TABLE_WRITE_ACTION_ADD,
GID_TABLE_WRITE_ACTION_DEL,
/* MODIFY only updates the GID table. Currently only used by
* ib_cache_update.
*/
GID_TABLE_WRITE_ACTION_MODIFY
};
struct ib_gid_table_entry {
unsigned long props;
union ib_gid gid;
@@ -103,31 +92,26 @@ struct ib_gid_table {
* (a) Find the GID
* (b) Delete it.
*
* Add/delete should be carried out atomically.
* This is done by locking this mutex from multiple
* writers. We don't need this lock for IB, as the MAD
* layer replaces all entries. All data_vec entries
* are locked by this lock.
**/
struct mutex lock;
/* This lock protects the table entries from being
* read and written simultaneously.
/* Any writer to data_vec must hold this lock and the write side of
* rwlock. readers must hold only rwlock. All writers must be in a
* sleepable context.
*/
struct mutex lock;
/* rwlock protects data_vec[ix]->props. */
rwlock_t rwlock;
struct ib_gid_table_entry *data_vec;
};
static void dispatch_gid_change_event(struct ib_device *ib_dev, u8 port)
{
if (rdma_cap_roce_gid_table(ib_dev, port)) {
struct ib_event event;
struct ib_event event;
event.device = ib_dev;
event.element.port_num = port;
event.event = IB_EVENT_GID_CHANGE;
event.device = ib_dev;
event.element.port_num = port;
event.event = IB_EVENT_GID_CHANGE;
ib_dispatch_event(&event);
}
ib_dispatch_event(&event);
}
static const char * const gid_type_str[] = {
@@ -168,94 +152,127 @@ int ib_cache_gid_parse_type_str(const char *buf)
}
EXPORT_SYMBOL(ib_cache_gid_parse_type_str);
/* This function expects that rwlock will be write locked in all
* scenarios and that lock will be locked in sleep-able (RoCE)
* scenarios.
*/
static int write_gid(struct ib_device *ib_dev, u8 port,
struct ib_gid_table *table, int ix,
const union ib_gid *gid,
const struct ib_gid_attr *attr,
enum gid_table_write_action action,
bool default_gid)
__releases(&table->rwlock) __acquires(&table->rwlock)
static void del_roce_gid(struct ib_device *device, u8 port_num,
struct ib_gid_table *table, int ix)
{
pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
device->name, port_num, ix,
table->data_vec[ix].gid.raw);
if (rdma_cap_roce_gid_table(device, port_num))
device->del_gid(&table->data_vec[ix].attr,
&table->data_vec[ix].context);
dev_put(table->data_vec[ix].attr.ndev);
}
static int add_roce_gid(struct ib_gid_table *table,
const union ib_gid *gid,
const struct ib_gid_attr *attr)
{
struct ib_gid_table_entry *entry;
int ix = attr->index;
int ret = 0;
if_t old_net_dev;
enum ib_gid_type old_gid_type;
/* in rdma_cap_roce_gid_table, this funciton should be protected by a
* sleep-able lock.
*/
if (rdma_cap_roce_gid_table(ib_dev, port)) {
table->data_vec[ix].props |= GID_TABLE_ENTRY_INVALID;
write_unlock_irq(&table->rwlock);
/* GID_TABLE_WRITE_ACTION_MODIFY currently isn't supported by
* RoCE providers and thus only updates the cache.
*/
if (action == GID_TABLE_WRITE_ACTION_ADD)
ret = ib_dev->add_gid(ib_dev, port, ix, gid, attr,
&table->data_vec[ix].context);
else if (action == GID_TABLE_WRITE_ACTION_DEL)
ret = ib_dev->del_gid(ib_dev, port, ix,
&table->data_vec[ix].context);
write_lock_irq(&table->rwlock);
if (!attr->ndev) {
pr_err("%s NULL netdev device=%s port=%d index=%d\n",
__func__, attr->device->name, attr->port_num,
attr->index);
return -EINVAL;
}
old_net_dev = table->data_vec[ix].attr.ndev;
old_gid_type = table->data_vec[ix].attr.gid_type;
if (old_net_dev && old_net_dev != attr->ndev)
dev_put(old_net_dev);
/* if modify_gid failed, just delete the old gid */
if (ret || action == GID_TABLE_WRITE_ACTION_DEL) {
gid = &zgid;
attr = &zattr;
table->data_vec[ix].context = NULL;
entry = &table->data_vec[ix];
if ((entry->props & GID_TABLE_ENTRY_INVALID) == 0) {
WARN(1, "GID table corruption device=%s port=%d index=%d\n",
attr->device->name, attr->port_num,
attr->index);
return -EINVAL;
}
memcpy(&table->data_vec[ix].gid, gid, sizeof(*gid));
memcpy(&table->data_vec[ix].attr, attr, sizeof(*attr));
if (default_gid) {
table->data_vec[ix].props |= GID_TABLE_ENTRY_DEFAULT;
if (action == GID_TABLE_WRITE_ACTION_DEL)
table->data_vec[ix].attr.gid_type = old_gid_type;
if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) {
ret = attr->device->add_gid(gid, attr, &entry->context);
if (ret) {
pr_err("%s GID add failed device=%s port=%d index=%d\n",
__func__, attr->device->name, attr->port_num,
attr->index);
goto add_err;
}
}
if (table->data_vec[ix].attr.ndev &&
table->data_vec[ix].attr.ndev != old_net_dev)
dev_hold(table->data_vec[ix].attr.ndev);
table->data_vec[ix].props &= ~GID_TABLE_ENTRY_INVALID;
dev_hold(attr->ndev);
add_err:
if (!ret)
pr_debug("%s device=%s port=%d index=%d gid %pI6\n", __func__,
attr->device->name, attr->port_num, ix, gid->raw);
return ret;
}
static int add_gid(struct ib_device *ib_dev, u8 port,
struct ib_gid_table *table, int ix,
const union ib_gid *gid,
const struct ib_gid_attr *attr,
bool default_gid) {
return write_gid(ib_dev, port, table, ix, gid, attr,
GID_TABLE_WRITE_ACTION_ADD, default_gid);
/**
* add_modify_gid - Add or modify GID table entry
*
* @table: GID table in which GID to be added or modified
* @gid: GID content
* @attr: Attributes of the GID
*
* Returns 0 on success or appropriate error code. It accepts zero
* GID addition for non RoCE ports for HCA's who report them as valid
* GID. However such zero GIDs are not added to the cache.
*/
static int add_modify_gid(struct ib_gid_table *table,
const union ib_gid *gid,
const struct ib_gid_attr *attr)
{
int ret;
if (rdma_protocol_roce(attr->device, attr->port_num)) {
ret = add_roce_gid(table, gid, attr);
if (ret)
return ret;
} else {
/*
* Some HCA's report multiple GID entries with only one
* valid GID, but remaining as zero GID.
* So ignore such behavior for IB link layer and don't
* fail the call, but don't add such entry to GID cache.
*/
if (!memcmp(gid, &zgid, sizeof(*gid)))
return 0;
}
lockdep_assert_held(&table->lock);
memcpy(&table->data_vec[attr->index].gid, gid, sizeof(*gid));
memcpy(&table->data_vec[attr->index].attr, attr, sizeof(*attr));
write_lock_irq(&table->rwlock);
table->data_vec[attr->index].props &= ~GID_TABLE_ENTRY_INVALID;
write_unlock_irq(&table->rwlock);
return 0;
}
static int modify_gid(struct ib_device *ib_dev, u8 port,
struct ib_gid_table *table, int ix,
const union ib_gid *gid,
const struct ib_gid_attr *attr,
bool default_gid) {
return write_gid(ib_dev, port, table, ix, gid, attr,
GID_TABLE_WRITE_ACTION_MODIFY, default_gid);
/**
* del_gid - Delete GID table entry
*
* @ib_dev: IB device whose GID entry to be deleted
* @port: Port number of the IB device
* @table: GID table of the IB device for a port
* @ix: GID entry index to delete
*
*/
static void del_gid(struct ib_device *ib_dev, u8 port,
struct ib_gid_table *table, int ix)
{
lockdep_assert_held(&table->lock);
write_lock_irq(&table->rwlock);
table->data_vec[ix].props |= GID_TABLE_ENTRY_INVALID;
write_unlock_irq(&table->rwlock);
if (rdma_protocol_roce(ib_dev, port))
del_roce_gid(ib_dev, port, table, ix);
memcpy(&table->data_vec[ix].gid, &zgid, sizeof(zgid));
memset(&table->data_vec[ix].attr, 0, sizeof(table->data_vec[ix].attr));
table->data_vec[ix].context = NULL;
}
static int del_gid(struct ib_device *ib_dev, u8 port,
struct ib_gid_table *table, int ix,
bool default_gid) {
return write_gid(ib_dev, port, table, ix, &zgid, &zattr,
GID_TABLE_WRITE_ACTION_DEL, default_gid);
}
/* rwlock should be read locked */
/* rwlock should be read locked, or lock should be held */
static int find_gid(struct ib_gid_table *table, const union ib_gid *gid,
const struct ib_gid_attr *val, bool default_gid,
unsigned long mask, int *pempty)
@@ -271,15 +288,36 @@ static int find_gid(struct ib_gid_table *table, const union ib_gid *gid,
i++;
/* find_gid() is used during GID addition where it is expected
* to return a free entry slot which is not duplicate.
* Free entry slot is requested and returned if pempty is set,
* so lookup free slot only if requested.
*/
if (pempty && empty < 0) {
if (data->props & GID_TABLE_ENTRY_INVALID &&
(default_gid ==
!!(data->props & GID_TABLE_ENTRY_DEFAULT))) {
/*
* Found an invalid (free) entry; allocate it.
* If default GID is requested, then our
* found slot must be one of the DEFAULT
* reserved slots or we fail.
* This ensures that only DEFAULT reserved
* slots are used for default property GIDs.
*/
empty = curr_index;
}
}
/*
* Additionally find_gid() is used to find valid entry during
* lookup operation, where validity needs to be checked. So
* find the empty entry first to continue to search for a free
* slot and ignore its INVALID flag.
*/
if (data->props & GID_TABLE_ENTRY_INVALID)
continue;
if (empty < 0)
if (!memcmp(&data->gid, &zgid, sizeof(*gid)) &&
!memcmp(attr, &zattr, sizeof(*attr)) &&
!data->props)
empty = curr_index;
if (found >= 0)
continue;
@@ -329,25 +367,27 @@ static void make_default_gid(if_t dev, union ib_gid *gid)
addrconf_ifid_eui48(&gid->raw[8], dev);
}
int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
union ib_gid *gid, struct ib_gid_attr *attr)
static int __ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
union ib_gid *gid, struct ib_gid_attr *attr,
unsigned long mask, bool default_gid)
{
struct ib_gid_table *table;
int ix;
int ret = 0;
int empty;
int ix;
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
/* Do not allow adding zero GID in support of
* IB spec version 1.3 section 4.1.1 point (6) and
* section 12.7.10 and section 12.7.20
*/
if (!memcmp(gid, &zgid, sizeof(*gid)))
return -EINVAL;
mutex_lock(&table->lock);
write_lock_irq(&table->rwlock);
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
ix = find_gid(table, gid, attr, false, GID_ATTR_FIND_MASK_GID |
GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_NETDEV, &empty);
mutex_lock(&table->lock);
ix = find_gid(table, gid, attr, default_gid, mask, &empty);
if (ix >= 0)
goto out_unlock;
@@ -355,44 +395,74 @@ int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
ret = -ENOSPC;
goto out_unlock;
}
ret = add_gid(ib_dev, port, table, empty, gid, attr, false);
attr->device = ib_dev;
attr->index = empty;
attr->port_num = port;
ret = add_modify_gid(table, gid, attr);
if (!ret)
dispatch_gid_change_event(ib_dev, port);
out_unlock:
write_unlock_irq(&table->rwlock);
mutex_unlock(&table->lock);
if (ret)
pr_warn("%s: unable to add gid %pI6 error=%d\n",
__func__, gid->raw, ret);
return ret;
}
int ib_cache_gid_add(struct ib_device *ib_dev, u8 port,
union ib_gid *gid, struct ib_gid_attr *attr)
{
unsigned long mask;
int ret;
mask = GID_ATTR_FIND_MASK_GID |
GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_NETDEV;
ret = __ib_cache_gid_add(ib_dev, port, gid, attr, mask, false);
return ret;
}
static int
_ib_cache_gid_del(struct ib_device *ib_dev, u8 port,
union ib_gid *gid, struct ib_gid_attr *attr,
unsigned long mask, bool default_gid)
{
struct ib_gid_table *table;
int ret = 0;
int ix;
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
mutex_lock(&table->lock);
ix = find_gid(table, gid, attr, default_gid, mask, NULL);
if (ix < 0) {
ret = -EINVAL;
goto out_unlock;
}
del_gid(ib_dev, port, table, ix);
dispatch_gid_change_event(ib_dev, port);
out_unlock:
mutex_unlock(&table->lock);
if (ret)
pr_debug("%s: can't delete gid %pI6 error=%d\n",
__func__, gid->raw, ret);
return ret;
}
int ib_cache_gid_del(struct ib_device *ib_dev, u8 port,
union ib_gid *gid, struct ib_gid_attr *attr)
{
struct ib_gid_table *table;
int ix;
unsigned long mask = GID_ATTR_FIND_MASK_GID |
GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_DEFAULT |
GID_ATTR_FIND_MASK_NETDEV;
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
mutex_lock(&table->lock);
write_lock_irq(&table->rwlock);
ix = find_gid(table, gid, attr, false,
GID_ATTR_FIND_MASK_GID |
GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_NETDEV |
GID_ATTR_FIND_MASK_DEFAULT,
NULL);
if (ix < 0)
goto out_unlock;
if (!del_gid(ib_dev, port, table, ix, false))
dispatch_gid_change_event(ib_dev, port);
out_unlock:
write_unlock_irq(&table->rwlock);
mutex_unlock(&table->lock);
return 0;
return _ib_cache_gid_del(ib_dev, port, gid, attr, mask, false);
}
int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port,
@@ -405,16 +475,14 @@ int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port,
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
mutex_lock(&table->lock);
write_lock_irq(&table->rwlock);
for (ix = 0; ix < table->sz; ix++)
if (table->data_vec[ix].attr.ndev == ndev)
if (!del_gid(ib_dev, port, table, ix,
!!(table->data_vec[ix].props &
GID_TABLE_ENTRY_DEFAULT)))
deleted = true;
for (ix = 0; ix < table->sz; ix++) {
if (table->data_vec[ix].attr.ndev == ndev) {
del_gid(ib_dev, port, table, ix);
deleted = true;
}
}
write_unlock_irq(&table->rwlock);
mutex_unlock(&table->lock);
if (deleted)
@@ -497,6 +565,19 @@ static int ib_cache_gid_find(struct ib_device *ib_dev,
mask, port, index);
}
/**
* ib_find_cached_gid_by_port - Returns the GID table index where a specified
* GID value occurs. It searches for the specified GID value in the local
* software cache.
* @device: The device to query.
* @gid: The GID value to search for.
* @gid_type: The GID type to search for.
* @port_num: The port number of the device where the GID value should be
* searched.
* @ndev: In RoCE, the net device of the device. Null means ignore.
* @index: The index into the cached GID table where the GID was found. This
* parameter may be NULL.
*/
int ib_find_cached_gid_by_port(struct ib_device *ib_dev,
const union ib_gid *gid,
enum ib_gid_type gid_type,
@@ -533,7 +614,7 @@ int ib_find_cached_gid_by_port(struct ib_device *ib_dev,
EXPORT_SYMBOL(ib_find_cached_gid_by_port);
/**
* ib_find_gid_by_filter - Returns the GID table index where a specified
* ib_cache_gid_find_by_filter - Returns the GID table index where a specified
* GID value occurs
* @device: The device to query.
* @gid: The GID value to search for.
@@ -544,7 +625,7 @@ EXPORT_SYMBOL(ib_find_cached_gid_by_port);
* otherwise, we continue searching the GID table. It's guaranteed that
* while filter is executed, ndev field is valid and the structure won't
* change. filter is executed in an atomic context. filter must not be NULL.
* @index: The index into the cached GID table where the GID was found. This
* @index: The index into the cached GID table where the GID was found. This
* parameter may be NULL.
*
* ib_cache_gid_find_by_filter() searches for the specified GID value
@@ -604,6 +685,7 @@ static struct ib_gid_table *alloc_gid_table(int sz)
{
struct ib_gid_table *table =
kzalloc(sizeof(struct ib_gid_table), GFP_KERNEL);
int i;
if (!table)
return NULL;
@@ -617,6 +699,11 @@ static struct ib_gid_table *alloc_gid_table(int sz)
table->sz = sz;
rwlock_init(&table->rwlock);
/* Mark all entries as invalid so that allocator can allocate
* one of the invalid (free) entry.
*/
for (i = 0; i < sz; i++)
table->data_vec[i].props |= GID_TABLE_ENTRY_INVALID;
return table;
err_free_table:
@@ -641,16 +728,15 @@ static void cleanup_gid_table_port(struct ib_device *ib_dev, u8 port,
if (!table)
return;
write_lock_irq(&table->rwlock);
mutex_lock(&table->lock);
for (i = 0; i < table->sz; ++i) {
if (memcmp(&table->data_vec[i].gid, &zgid,
sizeof(table->data_vec[i].gid)))
if (!del_gid(ib_dev, port, table, i,
table->data_vec[i].props &
GID_ATTR_FIND_MASK_DEFAULT))
deleted = true;
sizeof(table->data_vec[i].gid))) {
del_gid(ib_dev, port, table, i);
deleted = true;
}
}
write_unlock_irq(&table->rwlock);
mutex_unlock(&table->lock);
if (deleted)
dispatch_gid_change_event(ib_dev, port);
@@ -661,15 +747,17 @@ void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port,
unsigned long gid_type_mask,
enum ib_cache_gid_default_mode mode)
{
union ib_gid gid;
union ib_gid gid = { };
struct ib_gid_attr gid_attr;
struct ib_gid_attr zattr_type = zattr;
struct ib_gid_table *table;
unsigned int gid_type;
unsigned long mask;
table = ib_dev->cache.ports[port - rdma_start_port(ib_dev)].gid;
make_default_gid(ndev, &gid);
mask = GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_DEFAULT |
GID_ATTR_FIND_MASK_NETDEV;
memset(&gid_attr, 0, sizeof(gid_attr));
gid_attr.ndev = ndev;
@@ -684,60 +772,19 @@ void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port,
#endif
for (gid_type = 0; gid_type < IB_GID_TYPE_SIZE; ++gid_type) {
int ix;
union ib_gid current_gid;
struct ib_gid_attr current_gid_attr = {};
if (1UL << gid_type & ~gid_type_mask)
continue;
gid_attr.gid_type = gid_type;
mutex_lock(&table->lock);
write_lock_irq(&table->rwlock);
ix = find_gid(table, NULL, &gid_attr, true,
GID_ATTR_FIND_MASK_GID_TYPE |
GID_ATTR_FIND_MASK_DEFAULT,
NULL);
/* Coudn't find default GID location */
if (WARN_ON(ix < 0))
goto release;
zattr_type.gid_type = gid_type;
if (!__ib_cache_gid_get(ib_dev, port, ix,
&current_gid, &current_gid_attr) &&
mode == IB_CACHE_GID_DEFAULT_MODE_SET &&
!memcmp(&gid, &current_gid, sizeof(gid)) &&
!memcmp(&gid_attr, &current_gid_attr, sizeof(gid_attr)))
goto release;
if (memcmp(&current_gid, &zgid, sizeof(current_gid)) ||
memcmp(&current_gid_attr, &zattr_type,
sizeof(current_gid_attr))) {
if (del_gid(ib_dev, port, table, ix, true)) {
pr_warn("ib_cache_gid: can't delete index %d for default gid %pI6\n",
ix, gid.raw);
goto release;
} else {
dispatch_gid_change_event(ib_dev, port);
}
}
if (mode == IB_CACHE_GID_DEFAULT_MODE_SET) {
if (add_gid(ib_dev, port, table, ix, &gid, &gid_attr, true))
pr_warn("ib_cache_gid: unable to add default gid %pI6\n",
gid.raw);
else
dispatch_gid_change_event(ib_dev, port);
make_default_gid(ndev, &gid);
__ib_cache_gid_add(ib_dev, port, &gid,
&gid_attr, mask, true);
} else if (mode == IB_CACHE_GID_DEFAULT_MODE_DELETE) {
_ib_cache_gid_del(ib_dev, port, &gid,
&gid_attr, mask, true);
}
release:
if (current_gid_attr.ndev)
dev_put(current_gid_attr.ndev);
write_unlock_irq(&table->rwlock);
mutex_unlock(&table->lock);
}
}
@@ -869,6 +916,20 @@ int ib_get_cached_gid(struct ib_device *device,
}
EXPORT_SYMBOL(ib_get_cached_gid);
/**
* ib_find_cached_gid - Returns the port number and GID table index where
* a specified GID value occurs.
* @device: The device to query.
* @gid: The GID value to search for.
* @gid_type: The GID type to search for.
* @ndev: In RoCE, the net device of the device. NULL means ignore.
* @port_num: The port number of the device where the GID value was found.
* @index: The index into the cached GID table where the GID was found. This
* parameter may be NULL.
*
* ib_find_cached_gid() searches for the specified GID value in
* the local software cache.
*/
int ib_find_cached_gid(struct ib_device *device,
const union ib_gid *gid,
enum ib_gid_type gid_type,
@@ -889,7 +950,7 @@ int ib_find_gid_by_filter(struct ib_device *device,
void *context, u16 *index)
{
/* Only RoCE GID table supports filter function */
if (!rdma_cap_roce_gid_table(device, port_num) && filter)
if (!rdma_protocol_roce(device, port_num) && filter)
return -EPROTONOSUPPORT;
return ib_cache_gid_find_by_filter(device, gid,
@@ -1034,20 +1095,45 @@ int ib_get_cached_port_state(struct ib_device *device,
}
EXPORT_SYMBOL(ib_get_cached_port_state);
static int config_non_roce_gid_cache(struct ib_device *device,
u8 port, int gid_tbl_len)
{
struct ib_gid_attr gid_attr = {};
struct ib_gid_table *table;
union ib_gid gid;
int ret = 0;
int i;
gid_attr.device = device;
gid_attr.port_num = port;
table = device->cache.ports[port - rdma_start_port(device)].gid;
mutex_lock(&table->lock);
for (i = 0; i < gid_tbl_len; ++i) {
if (!device->query_gid)
continue;
ret = device->query_gid(device, port, i, &gid);
if (ret) {
pr_warn("query_gid failed (%d) for %s (index %d)\n",
ret, device->name, i);
goto err;
}
gid_attr.index = i;
add_modify_gid(table, &gid, &gid_attr);
}
err:
mutex_unlock(&table->lock);
return ret;
}
static void ib_cache_update(struct ib_device *device,
u8 port)
{
struct ib_port_attr *tprops = NULL;
struct ib_pkey_cache *pkey_cache = NULL, *old_pkey_cache;
struct ib_gid_cache {
int table_len;
union ib_gid table[0];
} *gid_cache = NULL;
int i;
int ret;
struct ib_gid_table *table;
bool use_roce_gid_table =
rdma_cap_roce_gid_table(device, port);
if (!rdma_is_port_valid(device, port))
return;
@@ -1065,6 +1151,13 @@ static void ib_cache_update(struct ib_device *device,
goto err;
}
if (!rdma_protocol_roce(device, port)) {
ret = config_non_roce_gid_cache(device, port,
tprops->gid_tbl_len);
if (ret)
goto err;
}
pkey_cache = kmalloc(sizeof *pkey_cache + tprops->pkey_tbl_len *
sizeof *pkey_cache->table, GFP_KERNEL);
if (!pkey_cache)
@@ -1072,15 +1165,6 @@ static void ib_cache_update(struct ib_device *device,
pkey_cache->table_len = tprops->pkey_tbl_len;
if (!use_roce_gid_table) {
gid_cache = kmalloc(sizeof(*gid_cache) + tprops->gid_tbl_len *
sizeof(*gid_cache->table), GFP_KERNEL);
if (!gid_cache)
goto err;
gid_cache->table_len = tprops->gid_tbl_len;
}
for (i = 0; i < pkey_cache->table_len; ++i) {
ret = ib_query_pkey(device, port, i, pkey_cache->table + i);
if (ret) {
@@ -1090,47 +1174,24 @@ static void ib_cache_update(struct ib_device *device,
}
}
if (!use_roce_gid_table) {
for (i = 0; i < gid_cache->table_len; ++i) {
ret = ib_query_gid(device, port, i,
gid_cache->table + i, NULL);
if (ret) {
pr_warn("ib_query_gid failed (%d) for %s (index %d)\n",
ret, device->name, i);
goto err;
}
}
}
write_lock_irq(&device->cache.lock);
old_pkey_cache = device->cache.ports[port -
rdma_start_port(device)].pkey;
device->cache.ports[port - rdma_start_port(device)].pkey = pkey_cache;
if (!use_roce_gid_table) {
write_lock(&table->rwlock);
for (i = 0; i < gid_cache->table_len; i++) {
modify_gid(device, port, table, i, gid_cache->table + i,
&zattr, false);
}
write_unlock(&table->rwlock);
}
device->cache.ports[port - rdma_start_port(device)].lmc = tprops->lmc;
device->cache.ports[port - rdma_start_port(device)].port_state =
tprops->state;
write_unlock_irq(&device->cache.lock);
kfree(gid_cache);
kfree(old_pkey_cache);
kfree(tprops);
return;
err:
kfree(pkey_cache);
kfree(gid_cache);
kfree(tprops);
}
sys/ofed/drivers/infiniband/core/ib_cm.c
+39 -14
View File
@@ -494,13 +494,31 @@ static int cm_init_av_for_response(struct cm_port *port, struct ib_wc *wc,
grh, &av->ah_attr);
}
static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
struct cm_id_private *cm_id_priv)
static int add_cm_id_to_port_list(struct cm_id_private *cm_id_priv,
struct cm_av *av,
struct cm_port *port)
{
unsigned long flags;
int ret = 0;
spin_lock_irqsave(&cm.lock, flags);
if (&cm_id_priv->av == av)
list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list);
else if (&cm_id_priv->alt_av == av)
list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list);
else
ret = -EINVAL;
spin_unlock_irqrestore(&cm.lock, flags);
return ret;
}
static struct cm_port *get_cm_port_from_path(struct sa_path_rec *path)
{
struct cm_device *cm_dev;
struct cm_port *port = NULL;
unsigned long flags;
int ret;
u8 p;
if_t ndev = ib_get_ndev_from_path(path);
@@ -509,7 +527,7 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
if (!ib_find_cached_gid(cm_dev->ib_device, &path->sgid,
sa_conv_pathrec_to_gid_type(path),
ndev, &p, NULL)) {
port = cm_dev->port[p-1];
port = cm_dev->port[p - 1];
break;
}
}
@@ -517,9 +535,20 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
if (ndev)
dev_put(ndev);
return port;
}
static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
struct cm_id_private *cm_id_priv)
{
struct cm_device *cm_dev;
struct cm_port *port;
int ret;
port = get_cm_port_from_path(path);
if (!port)
return -EINVAL;
cm_dev = port->cm_dev;
ret = ib_find_cached_pkey(cm_dev->ib_device, port->port_num,
be16_to_cpu(path->pkey), &av->pkey_index);
@@ -534,16 +563,7 @@ static int cm_init_av_by_path(struct sa_path_rec *path, struct cm_av *av,
av->timeout = path->packet_life_time + 1;
spin_lock_irqsave(&cm.lock, flags);
if (&cm_id_priv->av == av)
list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list);
else if (&cm_id_priv->alt_av == av)
list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list);
else
ret = -EINVAL;
spin_unlock_irqrestore(&cm.lock, flags);
ret = add_cm_id_to_port_list(cm_id_priv, av, port);
return ret;
}
@@ -1548,6 +1568,8 @@ static void cm_format_paths_from_req(struct cm_req_msg *req_msg,
cm_req_get_primary_local_ack_timeout(req_msg);
primary_path->packet_life_time -= (primary_path->packet_life_time > 0);
primary_path->service_id = req_msg->service_id;
if (sa_path_is_roce(primary_path))
primary_path->roce.route_resolved = false;
if (cm_req_has_alt_path(req_msg)) {
alt_path->dgid = req_msg->alt_local_gid;
@@ -1567,6 +1589,9 @@ static void cm_format_paths_from_req(struct cm_req_msg *req_msg,
cm_req_get_alt_local_ack_timeout(req_msg);
alt_path->packet_life_time -= (alt_path->packet_life_time > 0);
alt_path->service_id = req_msg->service_id;
if (sa_path_is_roce(alt_path))
alt_path->roce.route_resolved = false;
}
cm_format_path_lid_from_req(req_msg, primary_path, alt_path);
}
sys/ofed/drivers/infiniband/core/ib_cma.c
+10 -67
View File
@@ -74,6 +74,7 @@
#include <sys/priv.h>
#include "core_priv.h"
#include "cma_priv.h"
MODULE_AUTHOR("Sean Hefty");
MODULE_DESCRIPTION("Generic RDMA CM Agent");
@@ -168,7 +169,6 @@ static struct ib_client cma_client = {
};
static struct ib_sa_client sa_client;
static struct rdma_addr_client addr_client;
static LIST_HEAD(dev_list);
static LIST_HEAD(listen_any_list);
static DEFINE_MUTEX(lock);
@@ -351,49 +351,6 @@ struct ib_device *cma_get_ib_dev(struct cma_device *cma_dev)
* We do this by disabling removal notification while a callback is in process,
* and reporting it after the callback completes.
*/
struct rdma_id_private {
struct rdma_cm_id id;
struct rdma_bind_list *bind_list;
struct hlist_node node;
struct list_head list; /* listen_any_list or cma_device.list */
struct list_head listen_list; /* per device listens */
struct cma_device *cma_dev;
struct list_head mc_list;
int internal_id;
enum rdma_cm_state state;
spinlock_t lock;
struct mutex qp_mutex;
struct completion comp;
atomic_t refcount;
struct mutex handler_mutex;
int backlog;
int timeout_ms;
struct ib_sa_query *query;
int query_id;
union {
struct ib_cm_id *ib;
struct iw_cm_id *iw;
} cm_id;
u32 seq_num;
u32 qkey;
u32 qp_num;
pid_t owner;
u32 options;
u8 srq;
u8 tos;
u8 tos_set:1;
u8 timeout_set:1;
u8 reuseaddr;
u8 afonly;
u8 timeout;
enum ib_gid_type gid_type;
};
struct cma_multicast {
struct rdma_id_private *id_priv;
union {
@@ -941,7 +898,6 @@ static int cma_modify_qp_rtr(struct rdma_id_private *id_priv,
{
struct ib_qp_attr qp_attr;
int qp_attr_mask, ret;
union ib_gid sgid;
mutex_lock(&id_priv->qp_mutex);
if (!id_priv->id.qp) {
@@ -964,12 +920,6 @@ static int cma_modify_qp_rtr(struct rdma_id_private *id_priv,
if (ret)
goto out;
ret = ib_query_gid(id_priv->id.device, id_priv->id.port_num,
rdma_ah_read_grh(&qp_attr.ah_attr)->sgid_index,
&sgid, NULL);
if (ret)
goto out;
BUG_ON(id_priv->cma_dev->device != id_priv->id.device);
if (conn_param)
@@ -1712,21 +1662,12 @@ static bool cma_match_private_data(struct rdma_id_private *id_priv,
return true;
}
static bool cma_protocol_roce_dev_port(struct ib_device *device, int port_num)
{
enum rdma_link_layer ll = rdma_port_get_link_layer(device, port_num);
enum rdma_transport_type transport =
rdma_node_get_transport(device->node_type);
return ll == IB_LINK_LAYER_ETHERNET && transport == RDMA_TRANSPORT_IB;
}
static bool cma_protocol_roce(const struct rdma_cm_id *id)
{
struct ib_device *device = id->device;
const int port_num = id->port_num ?: rdma_start_port(device);
return cma_protocol_roce_dev_port(device, port_num);
return rdma_protocol_roce(device, port_num);
}
static bool cma_match_net_dev(const struct rdma_cm_id *id,
@@ -1746,8 +1687,8 @@ static bool cma_match_net_dev(const struct rdma_cm_id *id,
return false;
}
/* This request is an AF_IB request or a RoCE request */
return addr->src_addr.ss_family == AF_IB ||
cma_protocol_roce_dev_port(id->device, port_num);
return (addr->src_addr.ss_family == AF_IB ||
rdma_protocol_roce(id->device, port_num));
}
return !addr->dev_addr.bound_dev_if ||
@@ -1808,6 +1749,10 @@ static struct rdma_id_private *cma_id_from_event(struct ib_cm_id *cm_id,
if (PTR_ERR(*net_dev) == -EAFNOSUPPORT) {
/* Assuming the protocol is AF_IB */
*net_dev = NULL;
} else if (rdma_protocol_roce(req.device, req.port)) {
/* TODO find the net dev matching the request parameters
* through the RoCE GID table */
*net_dev = NULL;
} else {
return ERR_CAST(*net_dev);
}
@@ -2840,6 +2785,7 @@ cma_iboe_set_path_rec_l2_fields(struct rdma_id_private *id_priv)
gid_type = ib_network_to_gid_type(addr->dev_addr.network);
route->path_rec->rec_type = sa_conv_gid_to_pathrec_type(gid_type);
route->path_rec->roce.route_resolved = true;
sa_path_set_ndev(route->path_rec, addr->dev_addr.net);
sa_path_set_ifindex(route->path_rec, if_getindex(ndev));
sa_path_set_dmac(route->path_rec, addr->dev_addr.dst_dev_addr);
@@ -3232,7 +3178,7 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr,
if (ret)
goto err;
ret = rdma_resolve_ip(&addr_client, cma_src_addr(id_priv),
ret = rdma_resolve_ip(cma_src_addr(id_priv),
dst_addr, &id->route.addr.dev_addr,
timeout_ms, addr_handler, id_priv);
}
@@ -4838,7 +4784,6 @@ static int __init cma_init(void)
return -ENOMEM;
ib_sa_register_client(&sa_client);
rdma_addr_register_client(&addr_client);
ret = ib_register_client(&cma_client);
if (ret)
@@ -4849,7 +4794,6 @@ static int __init cma_init(void)
return 0;
err:
rdma_addr_unregister_client(&addr_client);
ib_sa_unregister_client(&sa_client);
destroy_workqueue(cma_wq);
return ret;
@@ -4859,7 +4803,6 @@ static void __exit cma_cleanup(void)
{
cma_configfs_exit();
ib_unregister_client(&cma_client);
rdma_addr_unregister_client(&addr_client);
ib_sa_unregister_client(&sa_client);
destroy_workqueue(cma_wq);
}
sys/ofed/drivers/infiniband/core/ib_device.c
+5 -13
View File
@@ -94,7 +94,6 @@ static int ib_device_check_mandatory(struct ib_device *device)
IB_MANDATORY_FUNC(query_device),
IB_MANDATORY_FUNC(query_port),
IB_MANDATORY_FUNC(query_pkey),
IB_MANDATORY_FUNC(query_gid),
IB_MANDATORY_FUNC(alloc_pd),
IB_MANDATORY_FUNC(dealloc_pd),
IB_MANDATORY_FUNC(create_ah),
@@ -667,7 +666,7 @@ int ib_query_port(struct ib_device *device,
if (rdma_port_get_link_layer(device, port_num) != IB_LINK_LAYER_INFINIBAND)
return 0;
err = ib_query_gid(device, port_num, 0, &gid, NULL);
err = device->query_gid(device, port_num, 0, &gid);
if (err)
return err;
@@ -685,19 +684,13 @@ EXPORT_SYMBOL(ib_query_port);
* @attr: Returned GID attributes related to this GID index (only in RoCE).
* NULL means ignore.
*
* ib_query_gid() fetches the specified GID table entry.
* ib_query_gid() fetches the specified GID table entry from the cache.
*/
int ib_query_gid(struct ib_device *device,
u8 port_num, int index, union ib_gid *gid,
struct ib_gid_attr *attr)
{
if (rdma_cap_roce_gid_table(device, port_num))
return ib_get_cached_gid(device, port_num, index, gid, attr);
if (attr)
return -EINVAL;
return device->query_gid(device, port_num, index, gid);
return ib_get_cached_gid(device, port_num, index, gid, attr);
}
EXPORT_SYMBOL(ib_query_gid);
@@ -860,19 +853,18 @@ EXPORT_SYMBOL(ib_modify_port);
* a specified GID value occurs. Its searches only for IB link layer.
* @device: The device to query.
* @gid: The GID value to search for.
* @ndev: The ndev related to the GID to search for.
* @port_num: The port number of the device where the GID value was found.
* @index: The index into the GID table where the GID was found. This
* parameter may be NULL.
*/
int ib_find_gid(struct ib_device *device, union ib_gid *gid,
if_t ndev, u8 *port_num, u16 *index)
u8 *port_num, u16 *index)
{
union ib_gid tmp_gid;
int ret, port, i;
for (port = rdma_start_port(device); port <= rdma_end_port(device); ++port) {
if (rdma_cap_roce_gid_table(device, port))
if (!rdma_protocol_ib(device, port))
continue;
for (i = 0; i < device->port_immutable[port].gid_tbl_len; ++i) {
sys/ofed/drivers/infiniband/core/ib_sa_query.c
+104 -90
View File
@@ -790,112 +790,126 @@ static u8 get_src_path_mask(struct ib_device *device, u8 port_num)
return src_path_mask;
}
static int
roce_resolve_route_from_path(struct ib_device *device, u8 port_num,
struct sa_path_rec *rec)
{
if_t resolved_dev;
if_t ndev;
if_t idev;
struct rdma_dev_addr dev_addr = {
.bound_dev_if = ((sa_path_get_ifindex(rec) >= 0) ?
sa_path_get_ifindex(rec) : 0),
.net = sa_path_get_ndev(rec) ?
sa_path_get_ndev(rec) :
&init_net
};
union rdma_sockaddr sgid_addr, dgid_addr;
int ret;
if (rec->roce.route_resolved)
return 0;
if (!device->get_netdev)
return -EOPNOTSUPP;
rdma_gid2ip(&sgid_addr._sockaddr, &rec->sgid);
rdma_gid2ip(&dgid_addr._sockaddr, &rec->dgid);
/* validate the route */
ret = rdma_resolve_ip_route(&sgid_addr._sockaddr,
&dgid_addr._sockaddr, &dev_addr);
if (ret)
return ret;
if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
dev_addr.network == RDMA_NETWORK_IPV6) &&
rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2)
return -EINVAL;
idev = device->get_netdev(device, port_num);
if (!idev)
return -ENODEV;
resolved_dev = dev_get_by_index(dev_addr.net,
dev_addr.bound_dev_if);
if (!resolved_dev) {
ret = -ENODEV;
goto done;
}
ndev = ib_get_ndev_from_path(rec);
rcu_read_lock();
if ((ndev && ndev != resolved_dev) ||
(resolved_dev != idev &&
rdma_vlan_dev_real_dev(resolved_dev) != idev))
ret = -EHOSTUNREACH;
rcu_read_unlock();
dev_put(resolved_dev);
if (ndev)
dev_put(ndev);
done:
dev_put(idev);
if (!ret)
rec->roce.route_resolved = true;
return ret;
}
static int init_ah_attr_grh_fields(struct ib_device *device, u8 port_num,
struct sa_path_rec *rec,
struct rdma_ah_attr *ah_attr)
{
enum ib_gid_type type = sa_conv_pathrec_to_gid_type(rec);
if_t ndev;
u16 gid_index;
int ret;
ndev = ib_get_ndev_from_path(rec);
ret = ib_find_cached_gid_by_port(device, &rec->sgid, type,
port_num, ndev, &gid_index);
if (ndev)
dev_put(ndev);
if (ret)
return ret;
rdma_ah_set_grh(ah_attr, &rec->dgid,
be32_to_cpu(rec->flow_label),
gid_index, rec->hop_limit,
rec->traffic_class);
return 0;
}
int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num,
struct sa_path_rec *rec,
struct rdma_ah_attr *ah_attr)
{
int ret;
u16 gid_index;
int use_roce;
if_t ndev = NULL;
int ret = 0;
memset(ah_attr, 0, sizeof *ah_attr);
memset(ah_attr, 0, sizeof(*ah_attr));
ah_attr->type = rdma_ah_find_type(device, port_num);
rdma_ah_set_dlid(ah_attr, be32_to_cpu(sa_path_get_dlid(rec)));
if ((ah_attr->type == RDMA_AH_ATTR_TYPE_OPA) &&
(rdma_ah_get_dlid(ah_attr) == be16_to_cpu(IB_LID_PERMISSIVE)))
rdma_ah_set_make_grd(ah_attr, true);
rdma_ah_set_sl(ah_attr, rec->sl);
rdma_ah_set_path_bits(ah_attr, be32_to_cpu(sa_path_get_slid(rec)) &
get_src_path_mask(device, port_num));
rdma_ah_set_port_num(ah_attr, port_num);
rdma_ah_set_static_rate(ah_attr, rec->rate);
use_roce = rdma_cap_eth_ah(device, port_num);
if (use_roce) {
if_t idev;
if_t resolved_dev;
struct rdma_dev_addr dev_addr = {
.bound_dev_if = ((sa_path_get_ifindex(rec) >= 0) ?
sa_path_get_ifindex(rec) : 0),
.net = sa_path_get_ndev(rec) ?
sa_path_get_ndev(rec) :
&init_net
};
union rdma_sockaddr sgid_addr, dgid_addr;
if (!device->get_netdev)
return -EOPNOTSUPP;
rdma_gid2ip(&sgid_addr._sockaddr, &rec->sgid);
rdma_gid2ip(&dgid_addr._sockaddr, &rec->dgid);
/* validate the route */
ret = rdma_resolve_ip_route(&sgid_addr._sockaddr,
&dgid_addr._sockaddr, &dev_addr);
if (sa_path_is_roce(rec)) {
ret = roce_resolve_route_from_path(device, port_num, rec);
if (ret)
return ret;
if ((dev_addr.network == RDMA_NETWORK_IPV4 ||
dev_addr.network == RDMA_NETWORK_IPV6) &&
rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2)
return -EINVAL;
memcpy(ah_attr->roce.dmac, sa_path_get_dmac(rec), ETH_ALEN);
} else {
rdma_ah_set_dlid(ah_attr, be32_to_cpu(sa_path_get_dlid(rec)));
if (sa_path_is_opa(rec) &&
rdma_ah_get_dlid(ah_attr) == be16_to_cpu(IB_LID_PERMISSIVE))
rdma_ah_set_make_grd(ah_attr, true);
idev = device->get_netdev(device, port_num);
if (!idev)
return -ENODEV;
resolved_dev = dev_get_by_index(dev_addr.net,
dev_addr.bound_dev_if);
if (!resolved_dev) {
dev_put(idev);
return -ENODEV;
}
ndev = ib_get_ndev_from_path(rec);
if ((ndev && ndev != resolved_dev) ||
(resolved_dev != idev &&
rdma_vlan_dev_real_dev(resolved_dev) != idev))
ret = -EHOSTUNREACH;
dev_put(idev);
dev_put(resolved_dev);
if (ret) {
if (ndev)
dev_put(ndev);
return ret;
}
rdma_ah_set_path_bits(ah_attr,
be32_to_cpu(sa_path_get_slid(rec)) &
get_src_path_mask(device, port_num));
}
if (rec->hop_limit > 0 || use_roce) {
enum ib_gid_type type = sa_conv_pathrec_to_gid_type(rec);
ret = ib_find_cached_gid_by_port(device, &rec->sgid, type,
port_num, ndev, &gid_index);
if (ret) {
if (ndev)
dev_put(ndev);
return ret;
}
rdma_ah_set_grh(ah_attr, &rec->dgid,
be32_to_cpu(rec->flow_label),
gid_index, rec->hop_limit,
rec->traffic_class);
if (ndev)
dev_put(ndev);
}
if (use_roce) {
u8 *dmac = sa_path_get_dmac(rec);
if (!dmac)
return -EINVAL;
memcpy(ah_attr->roce.dmac, dmac, ETH_ALEN);
}
return 0;
if (rec->hop_limit > 0 || sa_path_is_roce(rec))
ret = init_ah_attr_grh_fields(device, port_num, rec, ah_attr);
return ret;
}
EXPORT_SYMBOL(ib_init_ah_attr_from_path);
sys/ofed/drivers/infiniband/core/ib_sysfs.c
+15 -1
View File
@@ -411,6 +411,7 @@ static ssize_t show_port_gid(struct ib_port *p, struct port_attribute *attr,
{
struct port_table_attribute *tab_attr =
container_of(attr, struct port_table_attribute, attr);
union ib_gid *pgid;
union ib_gid gid;
ssize_t ret;
@@ -418,7 +419,20 @@ static ssize_t show_port_gid(struct ib_port *p, struct port_attribute *attr,
if (ret)
return ret;
return sprintf(buf, GID_PRINT_FMT"\n", GID_PRINT_ARGS(gid.raw));
/* If reading GID fails, it is likely due to GID entry being empty
* (invalid) or reserved GID in the table.
* User space expects to read GID table entries as long as it given
* index is within GID table size.
* Administrative/debugging tool fails to query rest of the GID entries
* if it hits error while querying a GID of the given index.
* To avoid user space throwing such error on fail to read gid, return
* zero GID as before. This maintains backward compatibility.
*/
if (ret)
pgid = &zgid;
else
pgid = &gid;
return sprintf(buf, GID_PRINT_FMT"\n", GID_PRINT_ARGS(pgid->raw));
}
static ssize_t show_port_gid_attr_ndev(struct ib_port *p,
sys/ofed/drivers/infiniband/core/ib_ucma.c
+3
View File
@@ -1248,6 +1248,9 @@ static int ucma_set_ib_path(struct ucma_context *ctx,
if (!optlen)
return -EINVAL;
if (!ctx->cm_id->device)
return -EINVAL;
memset(&sa_path, 0, sizeof(sa_path));
sa_path.rec_type = SA_PATH_REC_TYPE_IB;
sys/ofed/drivers/infiniband/core/ib_umem.c
-2
View File
@@ -66,8 +66,6 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
}
sg_free_table(&umem->sg_head);
return;
}
/**
sys/ofed/include/rdma/ib_addr.h
+1 -28
View File
@@ -56,11 +56,6 @@
#define net_eq(a,b) ((a) == (b))
struct rdma_addr_client {
atomic_t refcount;
struct completion comp;
};
union rdma_sockaddr {
struct sockaddr _sockaddr;
struct sockaddr_in _sockaddr_in;
@@ -68,17 +63,6 @@ union rdma_sockaddr {
struct sockaddr_storage _sockaddr_ss;
};
/**
* rdma_addr_register_client - Register an address client.
*/
void rdma_addr_register_client(struct rdma_addr_client *client);
/**
* rdma_addr_unregister_client - Deregister an address client.
* @client: Client object to deregister.
*/
void rdma_addr_unregister_client(struct rdma_addr_client *client);
/**
* struct rdma_dev_addr - Contains resolved RDMA hardware addresses
* @src_dev_addr: Source MAC address.
@@ -114,7 +98,6 @@ int rdma_translate_ip(const struct sockaddr *addr,
/**
* rdma_resolve_ip - Resolve source and destination IP addresses to
* RDMA hardware addresses.
* @client: Address client associated with request.
* @src_addr: An optional source address to use in the resolution. If a
* source address is not provided, a usable address will be returned via
* the callback.
@@ -127,17 +110,12 @@ int rdma_translate_ip(const struct sockaddr *addr,
* or been canceled. A status of 0 indicates success.
* @context: User-specified context associated with the call.
*/
int rdma_resolve_ip(struct rdma_addr_client *client,
struct sockaddr *src_addr, struct sockaddr *dst_addr,
int rdma_resolve_ip(struct sockaddr *src_addr, struct sockaddr *dst_addr,
struct rdma_dev_addr *addr, int timeout_ms,
void (*callback)(int status, struct sockaddr *src_addr,
struct rdma_dev_addr *addr, void *context),
void *context);
int rdma_resolve_ip_route(struct sockaddr *src_addr,
const struct sockaddr *dst_addr,
struct rdma_dev_addr *addr);
void rdma_addr_cancel(struct rdma_dev_addr *addr);
void rdma_copy_addr(struct rdma_dev_addr *dev_addr,
@@ -148,11 +126,6 @@ int rdma_addr_size(struct sockaddr *addr);
int rdma_addr_size_in6(struct sockaddr_in6 *addr);
int rdma_addr_size_kss(struct sockaddr_storage *addr);
int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid,
const union ib_gid *dgid,
u8 *dmac, if_t ndev,
int *hoplimit);
static inline u16 ib_addr_get_pkey(struct rdma_dev_addr *dev_addr)
{
return ((u16)dev_addr->broadcast[8] << 8) | (u16)dev_addr->broadcast[9];
sys/ofed/include/rdma/ib_cache.h
-29
View File
@@ -57,20 +57,6 @@ int ib_get_cached_gid(struct ib_device *device,
union ib_gid *gid,
struct ib_gid_attr *attr);
/**
* ib_find_cached_gid - Returns the port number and GID table index where
* a specified GID value occurs.
* @device: The device to query.
* @gid: The GID value to search for.
* @gid_type: The GID type to search for.
* @ndev: In RoCE, the net device of the device. NULL means ignore.
* @port_num: The port number of the device where the GID value was found.
* @index: The index into the cached GID table where the GID was found. This
* parameter may be NULL.
*
* ib_find_cached_gid() searches for the specified GID value in
* the local software cache.
*/
int ib_find_cached_gid(struct ib_device *device,
const union ib_gid *gid,
enum ib_gid_type gid_type,
@@ -78,21 +64,6 @@ int ib_find_cached_gid(struct ib_device *device,
u8 *port_num,
u16 *index);
/**
* ib_find_cached_gid_by_port - Returns the GID table index where a specified
* GID value occurs
* @device: The device to query.
* @gid: The GID value to search for.
* @gid_type: The GID type to search for.
* @port_num: The port number of the device where the GID value sould be
* searched.
* @ndev: In RoCE, the net device of the device. Null means ignore.
* @index: The index into the cached GID table where the GID was found. This
* parameter may be NULL.
*
* ib_find_cached_gid() searches for the specified GID value in
* the local software cache.
*/
int ib_find_cached_gid_by_port(struct ib_device *device,
const union ib_gid *gid,
enum ib_gid_type gid_type,
sys/ofed/include/rdma/ib_sa.h
+13
View File
@@ -164,7 +164,15 @@ struct sa_path_rec_ib {
u8 raw_traffic;
};
/**
* struct sa_path_rec_roce - RoCE specific portion of the path record entry
* @route_resolved: When set, it indicates that this route is already
* resolved for this path record entry.
* @dmac: Destination mac address for the given DGID entry
* of the path record entry.
*/
struct sa_path_rec_roce {
bool route_resolved;
u8 dmac[ETH_ALEN];
/* ignored in IB */
int ifindex;
@@ -591,6 +599,11 @@ static inline bool sa_path_is_roce(struct sa_path_rec *rec)
(rec->rec_type == SA_PATH_REC_TYPE_ROCE_V2));
}
static inline bool sa_path_is_opa(struct sa_path_rec *rec)
{
return (rec->rec_type == SA_PATH_REC_TYPE_OPA);
}
static inline void sa_path_set_slid(struct sa_path_rec *rec, __be32 slid)
{
if (rec->rec_type == SA_PATH_REC_TYPE_IB)
sys/ofed/include/rdma/ib_verbs.h
+23 -21
View File
@@ -103,8 +103,11 @@ enum ib_gid_type {
#define ROCE_V2_UDP_DPORT 4791
struct ib_gid_attr {
enum ib_gid_type gid_type;
if_t ndev;
struct ib_device *device;
enum ib_gid_type gid_type;
u16 index;
u8 port_num;
};
enum rdma_node_type {
@@ -2182,37 +2185,36 @@ struct ib_device {
*/
if_t (*get_netdev)(struct ib_device *device,
u8 port_num);
/* query_gid should be return GID value for @device, when @port_num
* link layer is either IB or iWarp. It is no-op if @port_num port
* is RoCE link layer.
*/
int (*query_gid)(struct ib_device *device,
u8 port_num, int index,
union ib_gid *gid);
/* When calling add_gid, the HW vendor's driver should
* add the gid of device @device at gid index @index of
* port @port_num to be @gid. Meta-info of that gid (for example,
* the network device related to this gid is available
* at @attr. @context allows the HW vendor driver to store extra
* information together with a GID entry. The HW vendor may allocate
* memory to contain this information and store it in @context when a
* new GID entry is written to. Params are consistent until the next
* call of add_gid or delete_gid. The function should return 0 on
/* When calling add_gid, the HW vendor's driver should add the gid
* of device of port at gid index available at @attr. Meta-info of
* that gid (for example, the network device related to this gid) is
* available at @attr. @context allows the HW vendor driver to store
* extra information together with a GID entry. The HW vendor driver may
* allocate memory to contain this information and store it in @context
* when a new GID entry is written to. Params are consistent until the
* next call of add_gid or delete_gid. The function should return 0 on
* success or error otherwise. The function could be called
* concurrently for different ports. This function is only called
* when roce_gid_table is used.
* concurrently for different ports. This function is only called when
* roce_gid_table is used.
*/
int (*add_gid)(struct ib_device *device,
u8 port_num,
unsigned int index,
const union ib_gid *gid,
int (*add_gid)(const union ib_gid *gid,
const struct ib_gid_attr *attr,
void **context);
/* When calling del_gid, the HW vendor's driver should delete the
* gid of device @device at gid index @index of port @port_num.
* gid of device @device at gid index gid_index of port port_num
* available in @attr.
* Upon the deletion of a GID entry, the HW vendor must free any
* allocated memory. The caller will clear @context afterwards.
* This function is only called when roce_gid_table is used.
*/
int (*del_gid)(struct ib_device *device,
u8 port_num,
unsigned int index,
int (*del_gid)(const struct ib_gid_attr *attr,
void **context);
int (*query_pkey)(struct ib_device *device,
u8 port_num, u16 index, u16 *pkey);
@@ -2960,7 +2962,7 @@ int ib_modify_port(struct ib_device *device,
struct ib_port_modify *port_modify);
int ib_find_gid(struct ib_device *device, union ib_gid *gid,
if_t ndev, u8 *port_num, u16 *index);
u8 *port_num, u16 *index);
int ib_find_pkey(struct ib_device *device,
u8 port_num, u16 pkey, u16 *index);
sys/ofed/include/rdma/rdma_cm.h
-14
View File
@@ -122,20 +122,6 @@ struct rdma_cm_event {
} param;
};
enum rdma_cm_state {
RDMA_CM_IDLE,
RDMA_CM_ADDR_QUERY,
RDMA_CM_ADDR_RESOLVED,
RDMA_CM_ROUTE_QUERY,
RDMA_CM_ROUTE_RESOLVED,
RDMA_CM_CONNECT,
RDMA_CM_DISCONNECT,
RDMA_CM_ADDR_BOUND,
RDMA_CM_LISTEN,
RDMA_CM_DEVICE_REMOVAL,
RDMA_CM_DESTROYING
};
struct rdma_cm_id;
/**