1 /*
2 * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
3 * Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
4 *
5 * This software is available to you under a choice of one of two
6 * licenses. You may choose to be licensed under the terms of the GNU
7 * General Public License (GPL) Version 2, available from the file
8 * COPYING in the main directory of this source tree, or the
9 * OpenIB.org BSD license below:
10 *
11 * Redistribution and use in source and binary forms, with or
12 * without modification, are permitted provided that the following
13 * conditions are met:
14 *
15 * - Redistributions of source code must retain the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer.
18 *
19 * - Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials
22 * provided with the distribution.
23 *
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
25 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
26 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
27 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
28 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
29 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
30 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
31 * SOFTWARE.
32 */
33
34 #include <net/addrconf.h>
35 #include "rxe.h"
36 #include "rxe_loc.h"
37
38 MODULE_AUTHOR("Bob Pearson, Frank Zago, John Groves, Kamal Heib");
39 MODULE_DESCRIPTION("Soft RDMA transport");
40 MODULE_LICENSE("Dual BSD/GPL");
41
42 /* free resources for all ports on a device */
rxe_cleanup_ports(struct rxe_dev * rxe)43 static void rxe_cleanup_ports(struct rxe_dev *rxe)
44 {
45 kfree(rxe->port.pkey_tbl);
46 rxe->port.pkey_tbl = NULL;
47
48 }
49
50 /* free resources for a rxe device all objects created for this device must
51 * have been destroyed
52 */
rxe_cleanup(struct rxe_dev * rxe)53 static void rxe_cleanup(struct rxe_dev *rxe)
54 {
55 rxe_pool_cleanup(&rxe->uc_pool);
56 rxe_pool_cleanup(&rxe->pd_pool);
57 rxe_pool_cleanup(&rxe->ah_pool);
58 rxe_pool_cleanup(&rxe->srq_pool);
59 rxe_pool_cleanup(&rxe->qp_pool);
60 rxe_pool_cleanup(&rxe->cq_pool);
61 rxe_pool_cleanup(&rxe->mr_pool);
62 rxe_pool_cleanup(&rxe->mw_pool);
63 rxe_pool_cleanup(&rxe->mc_grp_pool);
64 rxe_pool_cleanup(&rxe->mc_elem_pool);
65
66 rxe_cleanup_ports(rxe);
67
68 crypto_free_shash(rxe->tfm);
69 }
70
71 /* called when all references have been dropped */
rxe_release(struct kref * kref)72 void rxe_release(struct kref *kref)
73 {
74 struct rxe_dev *rxe = container_of(kref, struct rxe_dev, ref_cnt);
75
76 rxe_cleanup(rxe);
77 ib_dealloc_device(&rxe->ib_dev);
78 }
79
80 /* initialize rxe device parameters */
rxe_init_device_param(struct rxe_dev * rxe)81 static void rxe_init_device_param(struct rxe_dev *rxe)
82 {
83 rxe->max_inline_data = RXE_MAX_INLINE_DATA;
84
85 rxe->attr.fw_ver = RXE_FW_VER;
86 rxe->attr.max_mr_size = RXE_MAX_MR_SIZE;
87 rxe->attr.page_size_cap = RXE_PAGE_SIZE_CAP;
88 rxe->attr.vendor_id = RXE_VENDOR_ID;
89 rxe->attr.vendor_part_id = RXE_VENDOR_PART_ID;
90 rxe->attr.hw_ver = RXE_HW_VER;
91 rxe->attr.max_qp = RXE_MAX_QP;
92 rxe->attr.max_qp_wr = RXE_MAX_QP_WR;
93 rxe->attr.device_cap_flags = RXE_DEVICE_CAP_FLAGS;
94 rxe->attr.max_send_sge = RXE_MAX_SGE;
95 rxe->attr.max_recv_sge = RXE_MAX_SGE;
96 rxe->attr.max_sge_rd = RXE_MAX_SGE_RD;
97 rxe->attr.max_cq = RXE_MAX_CQ;
98 rxe->attr.max_cqe = (1 << RXE_MAX_LOG_CQE) - 1;
99 rxe->attr.max_mr = RXE_MAX_MR;
100 rxe->attr.max_pd = RXE_MAX_PD;
101 rxe->attr.max_qp_rd_atom = RXE_MAX_QP_RD_ATOM;
102 rxe->attr.max_ee_rd_atom = RXE_MAX_EE_RD_ATOM;
103 rxe->attr.max_res_rd_atom = RXE_MAX_RES_RD_ATOM;
104 rxe->attr.max_qp_init_rd_atom = RXE_MAX_QP_INIT_RD_ATOM;
105 rxe->attr.max_ee_init_rd_atom = RXE_MAX_EE_INIT_RD_ATOM;
106 rxe->attr.atomic_cap = RXE_ATOMIC_CAP;
107 rxe->attr.max_ee = RXE_MAX_EE;
108 rxe->attr.max_rdd = RXE_MAX_RDD;
109 rxe->attr.max_mw = RXE_MAX_MW;
110 rxe->attr.max_raw_ipv6_qp = RXE_MAX_RAW_IPV6_QP;
111 rxe->attr.max_raw_ethy_qp = RXE_MAX_RAW_ETHY_QP;
112 rxe->attr.max_mcast_grp = RXE_MAX_MCAST_GRP;
113 rxe->attr.max_mcast_qp_attach = RXE_MAX_MCAST_QP_ATTACH;
114 rxe->attr.max_total_mcast_qp_attach = RXE_MAX_TOT_MCAST_QP_ATTACH;
115 rxe->attr.max_ah = RXE_MAX_AH;
116 rxe->attr.max_fmr = RXE_MAX_FMR;
117 rxe->attr.max_map_per_fmr = RXE_MAX_MAP_PER_FMR;
118 rxe->attr.max_srq = RXE_MAX_SRQ;
119 rxe->attr.max_srq_wr = RXE_MAX_SRQ_WR;
120 rxe->attr.max_srq_sge = RXE_MAX_SRQ_SGE;
121 rxe->attr.max_fast_reg_page_list_len = RXE_MAX_FMR_PAGE_LIST_LEN;
122 rxe->attr.max_pkeys = RXE_MAX_PKEYS;
123 rxe->attr.local_ca_ack_delay = RXE_LOCAL_CA_ACK_DELAY;
124
125 rxe->max_ucontext = RXE_MAX_UCONTEXT;
126 }
127
128 /* initialize port attributes */
rxe_init_port_param(struct rxe_port * port)129 static int rxe_init_port_param(struct rxe_port *port)
130 {
131 port->attr.state = RXE_PORT_STATE;
132 port->attr.max_mtu = RXE_PORT_MAX_MTU;
133 port->attr.active_mtu = RXE_PORT_ACTIVE_MTU;
134 port->attr.gid_tbl_len = RXE_PORT_GID_TBL_LEN;
135 port->attr.port_cap_flags = RXE_PORT_PORT_CAP_FLAGS;
136 port->attr.max_msg_sz = RXE_PORT_MAX_MSG_SZ;
137 port->attr.bad_pkey_cntr = RXE_PORT_BAD_PKEY_CNTR;
138 port->attr.qkey_viol_cntr = RXE_PORT_QKEY_VIOL_CNTR;
139 port->attr.pkey_tbl_len = RXE_PORT_PKEY_TBL_LEN;
140 port->attr.lid = RXE_PORT_LID;
141 port->attr.sm_lid = RXE_PORT_SM_LID;
142 port->attr.lmc = RXE_PORT_LMC;
143 port->attr.max_vl_num = RXE_PORT_MAX_VL_NUM;
144 port->attr.sm_sl = RXE_PORT_SM_SL;
145 port->attr.subnet_timeout = RXE_PORT_SUBNET_TIMEOUT;
146 port->attr.init_type_reply = RXE_PORT_INIT_TYPE_REPLY;
147 port->attr.active_width = RXE_PORT_ACTIVE_WIDTH;
148 port->attr.active_speed = RXE_PORT_ACTIVE_SPEED;
149 port->attr.phys_state = RXE_PORT_PHYS_STATE;
150 port->mtu_cap =
151 ib_mtu_enum_to_int(RXE_PORT_ACTIVE_MTU);
152 port->subnet_prefix = cpu_to_be64(RXE_PORT_SUBNET_PREFIX);
153
154 return 0;
155 }
156
157 /* initialize port state, note IB convention that HCA ports are always
158 * numbered from 1
159 */
rxe_init_ports(struct rxe_dev * rxe)160 static int rxe_init_ports(struct rxe_dev *rxe)
161 {
162 struct rxe_port *port = &rxe->port;
163
164 rxe_init_port_param(port);
165
166 if (!port->attr.pkey_tbl_len || !port->attr.gid_tbl_len)
167 return -EINVAL;
168
169 port->pkey_tbl = kcalloc(port->attr.pkey_tbl_len,
170 sizeof(*port->pkey_tbl), GFP_KERNEL);
171
172 if (!port->pkey_tbl)
173 return -ENOMEM;
174
175 port->pkey_tbl[0] = 0xffff;
176 addrconf_addr_eui48((unsigned char *)&port->port_guid,
177 rxe->ndev->dev_addr);
178
179 spin_lock_init(&port->port_lock);
180
181 return 0;
182 }
183
184 /* init pools of managed objects */
rxe_init_pools(struct rxe_dev * rxe)185 static int rxe_init_pools(struct rxe_dev *rxe)
186 {
187 int err;
188
189 err = rxe_pool_init(rxe, &rxe->uc_pool, RXE_TYPE_UC,
190 rxe->max_ucontext);
191 if (err)
192 goto err1;
193
194 err = rxe_pool_init(rxe, &rxe->pd_pool, RXE_TYPE_PD,
195 rxe->attr.max_pd);
196 if (err)
197 goto err2;
198
199 err = rxe_pool_init(rxe, &rxe->ah_pool, RXE_TYPE_AH,
200 rxe->attr.max_ah);
201 if (err)
202 goto err3;
203
204 err = rxe_pool_init(rxe, &rxe->srq_pool, RXE_TYPE_SRQ,
205 rxe->attr.max_srq);
206 if (err)
207 goto err4;
208
209 err = rxe_pool_init(rxe, &rxe->qp_pool, RXE_TYPE_QP,
210 rxe->attr.max_qp);
211 if (err)
212 goto err5;
213
214 err = rxe_pool_init(rxe, &rxe->cq_pool, RXE_TYPE_CQ,
215 rxe->attr.max_cq);
216 if (err)
217 goto err6;
218
219 err = rxe_pool_init(rxe, &rxe->mr_pool, RXE_TYPE_MR,
220 rxe->attr.max_mr);
221 if (err)
222 goto err7;
223
224 err = rxe_pool_init(rxe, &rxe->mw_pool, RXE_TYPE_MW,
225 rxe->attr.max_mw);
226 if (err)
227 goto err8;
228
229 err = rxe_pool_init(rxe, &rxe->mc_grp_pool, RXE_TYPE_MC_GRP,
230 rxe->attr.max_mcast_grp);
231 if (err)
232 goto err9;
233
234 err = rxe_pool_init(rxe, &rxe->mc_elem_pool, RXE_TYPE_MC_ELEM,
235 rxe->attr.max_total_mcast_qp_attach);
236 if (err)
237 goto err10;
238
239 return 0;
240
241 err10:
242 rxe_pool_cleanup(&rxe->mc_grp_pool);
243 err9:
244 rxe_pool_cleanup(&rxe->mw_pool);
245 err8:
246 rxe_pool_cleanup(&rxe->mr_pool);
247 err7:
248 rxe_pool_cleanup(&rxe->cq_pool);
249 err6:
250 rxe_pool_cleanup(&rxe->qp_pool);
251 err5:
252 rxe_pool_cleanup(&rxe->srq_pool);
253 err4:
254 rxe_pool_cleanup(&rxe->ah_pool);
255 err3:
256 rxe_pool_cleanup(&rxe->pd_pool);
257 err2:
258 rxe_pool_cleanup(&rxe->uc_pool);
259 err1:
260 return err;
261 }
262
263 /* initialize rxe device state */
rxe_init(struct rxe_dev * rxe)264 static int rxe_init(struct rxe_dev *rxe)
265 {
266 int err;
267
268 /* init default device parameters */
269 rxe_init_device_param(rxe);
270
271 err = rxe_init_ports(rxe);
272 if (err)
273 goto err1;
274
275 err = rxe_init_pools(rxe);
276 if (err)
277 goto err2;
278
279 /* init pending mmap list */
280 spin_lock_init(&rxe->mmap_offset_lock);
281 spin_lock_init(&rxe->pending_lock);
282 INIT_LIST_HEAD(&rxe->pending_mmaps);
283 INIT_LIST_HEAD(&rxe->list);
284
285 mutex_init(&rxe->usdev_lock);
286
287 return 0;
288
289 err2:
290 rxe_cleanup_ports(rxe);
291 err1:
292 return err;
293 }
294
rxe_set_mtu(struct rxe_dev * rxe,unsigned int ndev_mtu)295 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int ndev_mtu)
296 {
297 struct rxe_port *port = &rxe->port;
298 enum ib_mtu mtu;
299
300 mtu = eth_mtu_int_to_enum(ndev_mtu);
301
302 /* Make sure that new MTU in range */
303 mtu = mtu ? min_t(enum ib_mtu, mtu, RXE_PORT_MAX_MTU) : IB_MTU_256;
304
305 port->attr.active_mtu = mtu;
306 port->mtu_cap = ib_mtu_enum_to_int(mtu);
307 }
308
309 /* called by ifc layer to create new rxe device.
310 * The caller should allocate memory for rxe by calling ib_alloc_device.
311 */
rxe_add(struct rxe_dev * rxe,unsigned int mtu)312 int rxe_add(struct rxe_dev *rxe, unsigned int mtu)
313 {
314 int err;
315
316 kref_init(&rxe->ref_cnt);
317
318 err = rxe_init(rxe);
319 if (err)
320 goto err1;
321
322 rxe_set_mtu(rxe, mtu);
323
324 err = rxe_register_device(rxe);
325 if (err)
326 goto err1;
327
328 return 0;
329
330 err1:
331 rxe_dev_put(rxe);
332 return err;
333 }
334
335 /* called by the ifc layer to remove a device */
rxe_remove(struct rxe_dev * rxe)336 void rxe_remove(struct rxe_dev *rxe)
337 {
338 rxe_unregister_device(rxe);
339
340 rxe_dev_put(rxe);
341 }
342
rxe_module_init(void)343 static int __init rxe_module_init(void)
344 {
345 int err;
346
347 /* initialize slab caches for managed objects */
348 err = rxe_cache_init();
349 if (err) {
350 pr_err("unable to init object pools\n");
351 return err;
352 }
353
354 err = rxe_net_init();
355 if (err)
356 return err;
357
358 pr_info("loaded\n");
359 return 0;
360 }
361
rxe_module_exit(void)362 static void __exit rxe_module_exit(void)
363 {
364 rxe_remove_all();
365 rxe_net_exit();
366 rxe_cache_exit();
367
368 pr_info("unloaded\n");
369 }
370
371 late_initcall(rxe_module_init);
372 module_exit(rxe_module_exit);
373