1 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
2 /*
3 * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
4 * Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
5 */
6
7 #include <rdma/rdma_netlink.h>
8 #include <net/addrconf.h>
9 #include "rxe.h"
10 #include "rxe_loc.h"
11
12 MODULE_AUTHOR("Bob Pearson, Frank Zago, John Groves, Kamal Heib");
13 MODULE_DESCRIPTION("Soft RDMA transport");
14 MODULE_LICENSE("Dual BSD/GPL");
15
16 /* free resources for a rxe device all objects created for this device must
17 * have been destroyed
18 */
rxe_dealloc(struct ib_device * ib_dev)19 void rxe_dealloc(struct ib_device *ib_dev)
20 {
21 struct rxe_dev *rxe = container_of(ib_dev, struct rxe_dev, ib_dev);
22
23 rxe_pool_cleanup(&rxe->uc_pool);
24 rxe_pool_cleanup(&rxe->pd_pool);
25 rxe_pool_cleanup(&rxe->ah_pool);
26 rxe_pool_cleanup(&rxe->srq_pool);
27 rxe_pool_cleanup(&rxe->qp_pool);
28 rxe_pool_cleanup(&rxe->cq_pool);
29 rxe_pool_cleanup(&rxe->mr_pool);
30 rxe_pool_cleanup(&rxe->mw_pool);
31
32 WARN_ON(!RB_EMPTY_ROOT(&rxe->mcg_tree));
33
34 if (rxe->tfm)
35 crypto_free_shash(rxe->tfm);
36 }
37
38 /* initialize rxe device parameters */
rxe_init_device_param(struct rxe_dev * rxe)39 static void rxe_init_device_param(struct rxe_dev *rxe)
40 {
41 rxe->max_inline_data = RXE_MAX_INLINE_DATA;
42
43 rxe->attr.vendor_id = RXE_VENDOR_ID;
44 rxe->attr.max_mr_size = RXE_MAX_MR_SIZE;
45 rxe->attr.page_size_cap = RXE_PAGE_SIZE_CAP;
46 rxe->attr.max_qp = RXE_MAX_QP;
47 rxe->attr.max_qp_wr = RXE_MAX_QP_WR;
48 rxe->attr.device_cap_flags = RXE_DEVICE_CAP_FLAGS;
49 rxe->attr.kernel_cap_flags = IBK_ALLOW_USER_UNREG;
50 rxe->attr.max_send_sge = RXE_MAX_SGE;
51 rxe->attr.max_recv_sge = RXE_MAX_SGE;
52 rxe->attr.max_sge_rd = RXE_MAX_SGE_RD;
53 rxe->attr.max_cq = RXE_MAX_CQ;
54 rxe->attr.max_cqe = (1 << RXE_MAX_LOG_CQE) - 1;
55 rxe->attr.max_mr = RXE_MAX_MR;
56 rxe->attr.max_mw = RXE_MAX_MW;
57 rxe->attr.max_pd = RXE_MAX_PD;
58 rxe->attr.max_qp_rd_atom = RXE_MAX_QP_RD_ATOM;
59 rxe->attr.max_res_rd_atom = RXE_MAX_RES_RD_ATOM;
60 rxe->attr.max_qp_init_rd_atom = RXE_MAX_QP_INIT_RD_ATOM;
61 rxe->attr.atomic_cap = IB_ATOMIC_HCA;
62 rxe->attr.max_mcast_grp = RXE_MAX_MCAST_GRP;
63 rxe->attr.max_mcast_qp_attach = RXE_MAX_MCAST_QP_ATTACH;
64 rxe->attr.max_total_mcast_qp_attach = RXE_MAX_TOT_MCAST_QP_ATTACH;
65 rxe->attr.max_ah = RXE_MAX_AH;
66 rxe->attr.max_srq = RXE_MAX_SRQ;
67 rxe->attr.max_srq_wr = RXE_MAX_SRQ_WR;
68 rxe->attr.max_srq_sge = RXE_MAX_SRQ_SGE;
69 rxe->attr.max_fast_reg_page_list_len = RXE_MAX_FMR_PAGE_LIST_LEN;
70 rxe->attr.max_pkeys = RXE_MAX_PKEYS;
71 rxe->attr.local_ca_ack_delay = RXE_LOCAL_CA_ACK_DELAY;
72 addrconf_addr_eui48((unsigned char *)&rxe->attr.sys_image_guid,
73 rxe->ndev->dev_addr);
74
75 rxe->max_ucontext = RXE_MAX_UCONTEXT;
76 }
77
78 /* initialize port attributes */
rxe_init_port_param(struct rxe_port * port)79 static void rxe_init_port_param(struct rxe_port *port)
80 {
81 port->attr.state = IB_PORT_DOWN;
82 port->attr.max_mtu = IB_MTU_4096;
83 port->attr.active_mtu = IB_MTU_256;
84 port->attr.gid_tbl_len = RXE_PORT_GID_TBL_LEN;
85 port->attr.port_cap_flags = RXE_PORT_PORT_CAP_FLAGS;
86 port->attr.max_msg_sz = RXE_PORT_MAX_MSG_SZ;
87 port->attr.bad_pkey_cntr = RXE_PORT_BAD_PKEY_CNTR;
88 port->attr.qkey_viol_cntr = RXE_PORT_QKEY_VIOL_CNTR;
89 port->attr.pkey_tbl_len = RXE_PORT_PKEY_TBL_LEN;
90 port->attr.lid = RXE_PORT_LID;
91 port->attr.sm_lid = RXE_PORT_SM_LID;
92 port->attr.lmc = RXE_PORT_LMC;
93 port->attr.max_vl_num = RXE_PORT_MAX_VL_NUM;
94 port->attr.sm_sl = RXE_PORT_SM_SL;
95 port->attr.subnet_timeout = RXE_PORT_SUBNET_TIMEOUT;
96 port->attr.init_type_reply = RXE_PORT_INIT_TYPE_REPLY;
97 port->attr.active_width = RXE_PORT_ACTIVE_WIDTH;
98 port->attr.active_speed = RXE_PORT_ACTIVE_SPEED;
99 port->attr.phys_state = RXE_PORT_PHYS_STATE;
100 port->mtu_cap = ib_mtu_enum_to_int(IB_MTU_256);
101 port->subnet_prefix = cpu_to_be64(RXE_PORT_SUBNET_PREFIX);
102 }
103
104 /* initialize port state, note IB convention that HCA ports are always
105 * numbered from 1
106 */
rxe_init_ports(struct rxe_dev * rxe)107 static void rxe_init_ports(struct rxe_dev *rxe)
108 {
109 struct rxe_port *port = &rxe->port;
110
111 rxe_init_port_param(port);
112 addrconf_addr_eui48((unsigned char *)&port->port_guid,
113 rxe->ndev->dev_addr);
114 spin_lock_init(&port->port_lock);
115 }
116
117 /* init pools of managed objects */
rxe_init_pools(struct rxe_dev * rxe)118 static void rxe_init_pools(struct rxe_dev *rxe)
119 {
120 rxe_pool_init(rxe, &rxe->uc_pool, RXE_TYPE_UC);
121 rxe_pool_init(rxe, &rxe->pd_pool, RXE_TYPE_PD);
122 rxe_pool_init(rxe, &rxe->ah_pool, RXE_TYPE_AH);
123 rxe_pool_init(rxe, &rxe->srq_pool, RXE_TYPE_SRQ);
124 rxe_pool_init(rxe, &rxe->qp_pool, RXE_TYPE_QP);
125 rxe_pool_init(rxe, &rxe->cq_pool, RXE_TYPE_CQ);
126 rxe_pool_init(rxe, &rxe->mr_pool, RXE_TYPE_MR);
127 rxe_pool_init(rxe, &rxe->mw_pool, RXE_TYPE_MW);
128 }
129
130 /* initialize rxe device state */
rxe_init(struct rxe_dev * rxe)131 static void rxe_init(struct rxe_dev *rxe)
132 {
133 /* init default device parameters */
134 rxe_init_device_param(rxe);
135
136 rxe_init_ports(rxe);
137 rxe_init_pools(rxe);
138
139 /* init pending mmap list */
140 spin_lock_init(&rxe->mmap_offset_lock);
141 spin_lock_init(&rxe->pending_lock);
142 INIT_LIST_HEAD(&rxe->pending_mmaps);
143
144 /* init multicast support */
145 spin_lock_init(&rxe->mcg_lock);
146 rxe->mcg_tree = RB_ROOT;
147
148 mutex_init(&rxe->usdev_lock);
149 }
150
rxe_set_mtu(struct rxe_dev * rxe,unsigned int ndev_mtu)151 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)
152 {
153 struct rxe_port *port = &rxe->port;
154 enum ib_mtu mtu;
155
156 mtu = eth_mtu_int_to_enum(ndev_mtu);
157
158 /* Make sure that new MTU in range */
159 mtu = mtu ? min_t(enum ib_mtu, mtu, IB_MTU_4096) : IB_MTU_256;
160
161 port->attr.active_mtu = mtu;
162 port->mtu_cap = ib_mtu_enum_to_int(mtu);
163 }
164
165 /* called by ifc layer to create new rxe device.
166 * The caller should allocate memory for rxe by calling ib_alloc_device.
167 */
rxe_add(struct rxe_dev * rxe,unsigned int mtu,const char * ibdev_name)168 int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name)
169 {
170 rxe_init(rxe);
171 rxe_set_mtu(rxe, mtu);
172
173 return rxe_register_device(rxe, ibdev_name);
174 }
175
rxe_newlink(const char * ibdev_name,struct net_device * ndev)176 static int rxe_newlink(const char *ibdev_name, struct net_device *ndev)
177 {
178 struct rxe_dev *exists;
179 int err = 0;
180
181 if (is_vlan_dev(ndev)) {
182 pr_err("rxe creation allowed on top of a real device only\n");
183 err = -EPERM;
184 goto err;
185 }
186
187 exists = rxe_get_dev_from_net(ndev);
188 if (exists) {
189 ib_device_put(&exists->ib_dev);
190 rxe_dbg(exists, "already configured on %s\n", ndev->name);
191 err = -EEXIST;
192 goto err;
193 }
194
195 err = rxe_net_add(ibdev_name, ndev);
196 if (err) {
197 rxe_dbg(exists, "failed to add %s\n", ndev->name);
198 goto err;
199 }
200 err:
201 return err;
202 }
203
204 static struct rdma_link_ops rxe_link_ops = {
205 .type = "rxe",
206 .newlink = rxe_newlink,
207 };
208
rxe_module_init(void)209 static int __init rxe_module_init(void)
210 {
211 int err;
212
213 err = rxe_net_init();
214 if (err)
215 return err;
216
217 rdma_link_register(&rxe_link_ops);
218 pr_info("loaded\n");
219 return 0;
220 }
221
rxe_module_exit(void)222 static void __exit rxe_module_exit(void)
223 {
224 rdma_link_unregister(&rxe_link_ops);
225 ib_unregister_driver(RDMA_DRIVER_RXE);
226 rxe_net_exit();
227
228 pr_info("unloaded\n");
229 }
230
231 late_initcall(rxe_module_init);
232 module_exit(rxe_module_exit);
233
234 MODULE_ALIAS_RDMA_LINK("rxe");
235