1 /*
2  * Copyright(c) 2017 Intel Corporation.
3  *
4  * This file is provided under a dual BSD/GPLv2 license.  When using or
5  * redistributing this file, you may do so under either license.
6  *
7  * GPL LICENSE SUMMARY
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of version 2 of the GNU General Public License as
11  * published by the Free Software Foundation.
12  *
13  * This program is distributed in the hope that it will be useful, but
14  * WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16  * General Public License for more details.
17  *
18  * BSD LICENSE
19  *
20  * Redistribution and use in source and binary forms, with or without
21  * modification, are permitted provided that the following conditions
22  * are met:
23  *
24  *  - Redistributions of source code must retain the above copyright
25  *    notice, this list of conditions and the following disclaimer.
26  *  - Redistributions in binary form must reproduce the above copyright
27  *    notice, this list of conditions and the following disclaimer in
28  *    the documentation and/or other materials provided with the
29  *    distribution.
30  *  - Neither the name of Intel Corporation nor the names of its
31  *    contributors may be used to endorse or promote products derived
32  *    from this software without specific prior written permission.
33  *
34  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
35  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
36  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
37  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
38  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
39  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
40  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
41  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
42  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
43  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
44  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
45  *
46  */
47 
48 /*
49  * This file contains OPA Virtual Network Interface Controller (VNIC)
50  * Ethernet Management Agent (EMA) driver
51  */
52 
53 #include <linux/module.h>
54 #include <linux/xarray.h>
55 #include <rdma/ib_addr.h>
56 #include <rdma/ib_verbs.h>
57 #include <rdma/opa_smi.h>
58 #include <rdma/opa_port_info.h>
59 
60 #include "opa_vnic_internal.h"
61 
62 #define DRV_VERSION "1.0"
63 char opa_vnic_driver_name[] = "opa_vnic";
64 const char opa_vnic_driver_version[] = DRV_VERSION;
65 
66 /*
67  * The trap service level is kept in bits 3 to 7 in the trap_sl_rsvd
68  * field in the class port info MAD.
69  */
70 #define GET_TRAP_SL_FROM_CLASS_PORT_INFO(x)  (((x) >> 3) & 0x1f)
71 
72 /* Cap trap bursts to a reasonable limit good for normal cases */
73 #define OPA_VNIC_TRAP_BURST_LIMIT 4
74 
75 /*
76  * VNIC trap limit timeout.
77  * Inverse of cap2_mask response time out (1.0737 secs) = 0.9
78  * secs approx IB spec 13.4.6.2.1 PortInfoSubnetTimeout and
79  * 13.4.9 Traps.
80  */
81 #define OPA_VNIC_TRAP_TIMEOUT  ((4096 * (1UL << 18)) / 1000)
82 
83 #define OPA_VNIC_UNSUP_ATTR  \
84 		cpu_to_be16(IB_MGMT_MAD_STATUS_UNSUPPORTED_METHOD_ATTRIB)
85 
86 #define OPA_VNIC_INVAL_ATTR  \
87 		cpu_to_be16(IB_MGMT_MAD_STATUS_INVALID_ATTRIB_VALUE)
88 
89 #define OPA_VNIC_CLASS_CAP_TRAP   0x1
90 
91 /* Maximum number of VNIC ports supported */
92 #define OPA_VNIC_MAX_NUM_VPORT    255
93 
94 /**
95  * struct opa_vnic_vema_port -- VNIC VEMA port details
96  * @cport: pointer to port
97  * @mad_agent: pointer to mad agent for port
98  * @class_port_info: Class port info information.
99  * @tid: Transaction id
100  * @port_num: OPA port number
101  * @vports: vnic ports
102  * @event_handler: ib event handler
103  * @lock: adapter interface lock
104  */
105 struct opa_vnic_vema_port {
106 	struct opa_vnic_ctrl_port      *cport;
107 	struct ib_mad_agent            *mad_agent;
108 	struct opa_class_port_info      class_port_info;
109 	u64                             tid;
110 	u8                              port_num;
111 	struct xarray                   vports;
112 	struct ib_event_handler         event_handler;
113 
114 	/* Lock to query/update network adapter */
115 	struct mutex                    lock;
116 };
117 
118 static void opa_vnic_vema_add_one(struct ib_device *device);
119 static void opa_vnic_vema_rem_one(struct ib_device *device,
120 				  void *client_data);
121 
122 static struct ib_client opa_vnic_client = {
123 	.name   = opa_vnic_driver_name,
124 	.add    = opa_vnic_vema_add_one,
125 	.remove = opa_vnic_vema_rem_one,
126 };
127 
128 /**
129  * vema_get_vport_num -- Get the vnic from the mad
130  * @recvd_mad:  Received mad
131  *
132  * Return: returns value of the vnic port number
133  */
134 static inline u8 vema_get_vport_num(struct opa_vnic_vema_mad *recvd_mad)
135 {
136 	return be32_to_cpu(recvd_mad->mad_hdr.attr_mod) & 0xff;
137 }
138 
139 /**
140  * vema_get_vport_adapter -- Get vnic port adapter from recvd mad
141  * @recvd_mad: received mad
142  * @port: ptr to port struct on which MAD was recvd
143  *
144  * Return: vnic adapter
145  */
146 static inline struct opa_vnic_adapter *
147 vema_get_vport_adapter(struct opa_vnic_vema_mad *recvd_mad,
148 		       struct opa_vnic_vema_port *port)
149 {
150 	u8 vport_num = vema_get_vport_num(recvd_mad);
151 
152 	return xa_load(&port->vports, vport_num);
153 }
154 
155 /**
156  * vema_mac_tbl_req_ok -- Check if mac request has correct values
157  * @mac_tbl: mac table
158  *
159  * This function checks for the validity of the offset and number of
160  * entries required.
161  *
162  * Return: true if offset and num_entries are valid
163  */
164 static inline bool vema_mac_tbl_req_ok(struct opa_veswport_mactable *mac_tbl)
165 {
166 	u16 offset, num_entries;
167 	u16 req_entries = ((OPA_VNIC_EMA_DATA - sizeof(*mac_tbl)) /
168 			   sizeof(mac_tbl->tbl_entries[0]));
169 
170 	offset = be16_to_cpu(mac_tbl->offset);
171 	num_entries = be16_to_cpu(mac_tbl->num_entries);
172 
173 	return ((num_entries <= req_entries) &&
174 		(offset + num_entries <= OPA_VNIC_MAC_TBL_MAX_ENTRIES));
175 }
176 
177 /*
178  * Return the power on default values in the port info structure
179  * in big endian format as required by MAD.
180  */
181 static inline void vema_get_pod_values(struct opa_veswport_info *port_info)
182 {
183 	memset(port_info, 0, sizeof(*port_info));
184 	port_info->vport.max_mac_tbl_ent =
185 		cpu_to_be16(OPA_VNIC_MAC_TBL_MAX_ENTRIES);
186 	port_info->vport.max_smac_ent =
187 		cpu_to_be16(OPA_VNIC_MAX_SMAC_LIMIT);
188 	port_info->vport.oper_state = OPA_VNIC_STATE_DROP_ALL;
189 	port_info->vport.config_state = OPA_VNIC_STATE_DROP_ALL;
190 	port_info->vesw.eth_mtu = cpu_to_be16(ETH_DATA_LEN);
191 }
192 
193 /**
194  * vema_add_vport -- Add a new vnic port
195  * @port: ptr to opa_vnic_vema_port struct
196  * @vport_num: vnic port number (to be added)
197  *
198  * Return a pointer to the vnic adapter structure
199  */
200 static struct opa_vnic_adapter *vema_add_vport(struct opa_vnic_vema_port *port,
201 					       u8 vport_num)
202 {
203 	struct opa_vnic_ctrl_port *cport = port->cport;
204 	struct opa_vnic_adapter *adapter;
205 
206 	adapter = opa_vnic_add_netdev(cport->ibdev, port->port_num, vport_num);
207 	if (!IS_ERR(adapter)) {
208 		int rc;
209 
210 		adapter->cport = cport;
211 		rc = xa_insert(&port->vports, vport_num, adapter, GFP_KERNEL);
212 		if (rc < 0) {
213 			opa_vnic_rem_netdev(adapter);
214 			adapter = ERR_PTR(rc);
215 		}
216 	}
217 
218 	return adapter;
219 }
220 
221 /**
222  * vema_get_class_port_info -- Get class info for port
223  * @port:  Port on whic MAD was received
224  * @recvd_mad: pointer to the received mad
225  * @rsp_mad:   pointer to respose mad
226  *
227  * This function copies the latest class port info value set for the
228  * port and stores it for generating traps
229  */
230 static void vema_get_class_port_info(struct opa_vnic_vema_port *port,
231 				     struct opa_vnic_vema_mad *recvd_mad,
232 				     struct opa_vnic_vema_mad *rsp_mad)
233 {
234 	struct opa_class_port_info *port_info;
235 
236 	port_info = (struct opa_class_port_info *)rsp_mad->data;
237 	memcpy(port_info, &port->class_port_info, sizeof(*port_info));
238 	port_info->base_version = OPA_MGMT_BASE_VERSION,
239 	port_info->class_version = OPA_EMA_CLASS_VERSION;
240 
241 	/*
242 	 * Set capability mask bit indicating agent generates traps,
243 	 * and set the maximum number of VNIC ports supported.
244 	 */
245 	port_info->cap_mask = cpu_to_be16((OPA_VNIC_CLASS_CAP_TRAP |
246 					   (OPA_VNIC_MAX_NUM_VPORT << 8)));
247 
248 	/*
249 	 * Since a get routine is always sent by the EM first we
250 	 * set the expected response time to
251 	 * 4.096 usec * 2^18 == 1.0737 sec here.
252 	 */
253 	port_info->cap_mask2_resp_time = cpu_to_be32(18);
254 }
255 
256 /**
257  * vema_set_class_port_info -- Get class info for port
258  * @port:  Port on whic MAD was received
259  * @recvd_mad: pointer to the received mad
260  * @rsp_mad:   pointer to respose mad
261  *
262  * This function updates the port class info for the specific vnic
263  * and sets up the response mad data
264  */
265 static void vema_set_class_port_info(struct opa_vnic_vema_port *port,
266 				     struct opa_vnic_vema_mad *recvd_mad,
267 				     struct opa_vnic_vema_mad *rsp_mad)
268 {
269 	memcpy(&port->class_port_info, recvd_mad->data,
270 	       sizeof(port->class_port_info));
271 
272 	vema_get_class_port_info(port, recvd_mad, rsp_mad);
273 }
274 
275 /**
276  * vema_get_veswport_info -- Get veswport info
277  * @port:      source port on which MAD was received
278  * @recvd_mad: pointer to the received mad
279  * @rsp_mad:   pointer to respose mad
280  */
281 static void vema_get_veswport_info(struct opa_vnic_vema_port *port,
282 				   struct opa_vnic_vema_mad *recvd_mad,
283 				   struct opa_vnic_vema_mad *rsp_mad)
284 {
285 	struct opa_veswport_info *port_info =
286 				  (struct opa_veswport_info *)rsp_mad->data;
287 	struct opa_vnic_adapter *adapter;
288 
289 	adapter = vema_get_vport_adapter(recvd_mad, port);
290 	if (adapter) {
291 		memset(port_info, 0, sizeof(*port_info));
292 		opa_vnic_get_vesw_info(adapter, &port_info->vesw);
293 		opa_vnic_get_per_veswport_info(adapter,
294 					       &port_info->vport);
295 	} else {
296 		vema_get_pod_values(port_info);
297 	}
298 }
299 
300 /**
301  * vema_set_veswport_info -- Set veswport info
302  * @port:      source port on which MAD was received
303  * @recvd_mad: pointer to the received mad
304  * @rsp_mad:   pointer to respose mad
305  *
306  * This function gets the port class infor for vnic
307  */
308 static void vema_set_veswport_info(struct opa_vnic_vema_port *port,
309 				   struct opa_vnic_vema_mad *recvd_mad,
310 				   struct opa_vnic_vema_mad *rsp_mad)
311 {
312 	struct opa_vnic_ctrl_port *cport = port->cport;
313 	struct opa_veswport_info *port_info;
314 	struct opa_vnic_adapter *adapter;
315 	u8 vport_num;
316 
317 	vport_num = vema_get_vport_num(recvd_mad);
318 
319 	adapter = vema_get_vport_adapter(recvd_mad, port);
320 	if (!adapter) {
321 		adapter = vema_add_vport(port, vport_num);
322 		if (IS_ERR(adapter)) {
323 			c_err("failed to add vport %d: %ld\n",
324 			      vport_num, PTR_ERR(adapter));
325 			goto err_exit;
326 		}
327 	}
328 
329 	port_info = (struct opa_veswport_info *)recvd_mad->data;
330 	opa_vnic_set_vesw_info(adapter, &port_info->vesw);
331 	opa_vnic_set_per_veswport_info(adapter, &port_info->vport);
332 
333 	/* Process the new config settings */
334 	opa_vnic_process_vema_config(adapter);
335 
336 	vema_get_veswport_info(port, recvd_mad, rsp_mad);
337 	return;
338 
339 err_exit:
340 	rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
341 }
342 
343 /**
344  * vema_get_mac_entries -- Get MAC entries in VNIC MAC table
345  * @port:      source port on which MAD was received
346  * @recvd_mad: pointer to the received mad
347  * @rsp_mad:   pointer to respose mad
348  *
349  * This function gets the MAC entries that are programmed into
350  * the VNIC MAC forwarding table. It checks for the validity of
351  * the index into the MAC table and the number of entries that
352  * are to be retrieved.
353  */
354 static void vema_get_mac_entries(struct opa_vnic_vema_port *port,
355 				 struct opa_vnic_vema_mad *recvd_mad,
356 				 struct opa_vnic_vema_mad *rsp_mad)
357 {
358 	struct opa_veswport_mactable *mac_tbl_in, *mac_tbl_out;
359 	struct opa_vnic_adapter *adapter;
360 
361 	adapter = vema_get_vport_adapter(recvd_mad, port);
362 	if (!adapter) {
363 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
364 		return;
365 	}
366 
367 	mac_tbl_in = (struct opa_veswport_mactable *)recvd_mad->data;
368 	mac_tbl_out = (struct opa_veswport_mactable *)rsp_mad->data;
369 
370 	if (vema_mac_tbl_req_ok(mac_tbl_in)) {
371 		mac_tbl_out->offset = mac_tbl_in->offset;
372 		mac_tbl_out->num_entries = mac_tbl_in->num_entries;
373 		opa_vnic_query_mac_tbl(adapter, mac_tbl_out);
374 	} else {
375 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
376 	}
377 }
378 
379 /**
380  * vema_set_mac_entries -- Set MAC entries in VNIC MAC table
381  * @port:      source port on which MAD was received
382  * @recvd_mad: pointer to the received mad
383  * @rsp_mad:   pointer to respose mad
384  *
385  * This function sets the MAC entries in the VNIC forwarding table
386  * It checks for the validity of the index and the number of forwarding
387  * table entries to be programmed.
388  */
389 static void vema_set_mac_entries(struct opa_vnic_vema_port *port,
390 				 struct opa_vnic_vema_mad *recvd_mad,
391 				 struct opa_vnic_vema_mad *rsp_mad)
392 {
393 	struct opa_veswport_mactable *mac_tbl;
394 	struct opa_vnic_adapter *adapter;
395 
396 	adapter = vema_get_vport_adapter(recvd_mad, port);
397 	if (!adapter) {
398 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
399 		return;
400 	}
401 
402 	mac_tbl = (struct opa_veswport_mactable *)recvd_mad->data;
403 	if (vema_mac_tbl_req_ok(mac_tbl)) {
404 		if (opa_vnic_update_mac_tbl(adapter, mac_tbl))
405 			rsp_mad->mad_hdr.status = OPA_VNIC_UNSUP_ATTR;
406 	} else {
407 		rsp_mad->mad_hdr.status = OPA_VNIC_UNSUP_ATTR;
408 	}
409 	vema_get_mac_entries(port, recvd_mad, rsp_mad);
410 }
411 
412 /**
413  * vema_set_delete_vesw -- Reset VESW info to POD values
414  * @port:      source port on which MAD was received
415  * @recvd_mad: pointer to the received mad
416  * @rsp_mad:   pointer to respose mad
417  *
418  * This function clears all the fields of veswport info for the requested vesw
419  * and sets them back to the power-on default values. It does not delete the
420  * vesw.
421  */
422 static void vema_set_delete_vesw(struct opa_vnic_vema_port *port,
423 				 struct opa_vnic_vema_mad *recvd_mad,
424 				 struct opa_vnic_vema_mad *rsp_mad)
425 {
426 	struct opa_veswport_info *port_info =
427 				  (struct opa_veswport_info *)rsp_mad->data;
428 	struct opa_vnic_adapter *adapter;
429 
430 	adapter = vema_get_vport_adapter(recvd_mad, port);
431 	if (!adapter) {
432 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
433 		return;
434 	}
435 
436 	vema_get_pod_values(port_info);
437 	opa_vnic_set_vesw_info(adapter, &port_info->vesw);
438 	opa_vnic_set_per_veswport_info(adapter, &port_info->vport);
439 
440 	/* Process the new config settings */
441 	opa_vnic_process_vema_config(adapter);
442 
443 	opa_vnic_release_mac_tbl(adapter);
444 
445 	vema_get_veswport_info(port, recvd_mad, rsp_mad);
446 }
447 
448 /**
449  * vema_get_mac_list -- Get the unicast/multicast macs.
450  * @port:      source port on which MAD was received
451  * @recvd_mad: Received mad contains fields to set vnic parameters
452  * @rsp_mad:   Response mad to be built
453  * @attr_id:   Attribute ID indicating multicast or unicast mac list
454  */
455 static void vema_get_mac_list(struct opa_vnic_vema_port *port,
456 			      struct opa_vnic_vema_mad *recvd_mad,
457 			      struct opa_vnic_vema_mad *rsp_mad,
458 			      u16 attr_id)
459 {
460 	struct opa_veswport_iface_macs *macs_in, *macs_out;
461 	int max_entries = (OPA_VNIC_EMA_DATA - sizeof(*macs_out)) / ETH_ALEN;
462 	struct opa_vnic_adapter *adapter;
463 
464 	adapter = vema_get_vport_adapter(recvd_mad, port);
465 	if (!adapter) {
466 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
467 		return;
468 	}
469 
470 	macs_in = (struct opa_veswport_iface_macs *)recvd_mad->data;
471 	macs_out = (struct opa_veswport_iface_macs *)rsp_mad->data;
472 
473 	macs_out->start_idx = macs_in->start_idx;
474 	if (macs_in->num_macs_in_msg)
475 		macs_out->num_macs_in_msg = macs_in->num_macs_in_msg;
476 	else
477 		macs_out->num_macs_in_msg = cpu_to_be16(max_entries);
478 
479 	if (attr_id == OPA_EM_ATTR_IFACE_MCAST_MACS)
480 		opa_vnic_query_mcast_macs(adapter, macs_out);
481 	else
482 		opa_vnic_query_ucast_macs(adapter, macs_out);
483 }
484 
485 /**
486  * vema_get_summary_counters -- Gets summary counters.
487  * @port:      source port on which MAD was received
488  * @recvd_mad: Received mad contains fields to set vnic parameters
489  * @rsp_mad:   Response mad to be built
490  */
491 static void vema_get_summary_counters(struct opa_vnic_vema_port *port,
492 				      struct opa_vnic_vema_mad *recvd_mad,
493 				      struct opa_vnic_vema_mad *rsp_mad)
494 {
495 	struct opa_veswport_summary_counters *cntrs;
496 	struct opa_vnic_adapter *adapter;
497 
498 	adapter = vema_get_vport_adapter(recvd_mad, port);
499 	if (adapter) {
500 		cntrs = (struct opa_veswport_summary_counters *)rsp_mad->data;
501 		opa_vnic_get_summary_counters(adapter, cntrs);
502 	} else {
503 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
504 	}
505 }
506 
507 /**
508  * vema_get_error_counters -- Gets summary counters.
509  * @port:      source port on which MAD was received
510  * @recvd_mad: Received mad contains fields to set vnic parameters
511  * @rsp_mad:   Response mad to be built
512  */
513 static void vema_get_error_counters(struct opa_vnic_vema_port *port,
514 				    struct opa_vnic_vema_mad *recvd_mad,
515 				    struct opa_vnic_vema_mad *rsp_mad)
516 {
517 	struct opa_veswport_error_counters *cntrs;
518 	struct opa_vnic_adapter *adapter;
519 
520 	adapter = vema_get_vport_adapter(recvd_mad, port);
521 	if (adapter) {
522 		cntrs = (struct opa_veswport_error_counters *)rsp_mad->data;
523 		opa_vnic_get_error_counters(adapter, cntrs);
524 	} else {
525 		rsp_mad->mad_hdr.status = OPA_VNIC_INVAL_ATTR;
526 	}
527 }
528 
529 /**
530  * vema_get -- Process received get MAD
531  * @port:      source port on which MAD was received
532  * @recvd_mad: Received mad
533  * @rsp_mad:   Response mad to be built
534  */
535 static void vema_get(struct opa_vnic_vema_port *port,
536 		     struct opa_vnic_vema_mad *recvd_mad,
537 		     struct opa_vnic_vema_mad *rsp_mad)
538 {
539 	u16 attr_id = be16_to_cpu(recvd_mad->mad_hdr.attr_id);
540 
541 	switch (attr_id) {
542 	case OPA_EM_ATTR_CLASS_PORT_INFO:
543 		vema_get_class_port_info(port, recvd_mad, rsp_mad);
544 		break;
545 	case OPA_EM_ATTR_VESWPORT_INFO:
546 		vema_get_veswport_info(port, recvd_mad, rsp_mad);
547 		break;
548 	case OPA_EM_ATTR_VESWPORT_MAC_ENTRIES:
549 		vema_get_mac_entries(port, recvd_mad, rsp_mad);
550 		break;
551 	case OPA_EM_ATTR_IFACE_UCAST_MACS:
552 		/* fall through */
553 	case OPA_EM_ATTR_IFACE_MCAST_MACS:
554 		vema_get_mac_list(port, recvd_mad, rsp_mad, attr_id);
555 		break;
556 	case OPA_EM_ATTR_VESWPORT_SUMMARY_COUNTERS:
557 		vema_get_summary_counters(port, recvd_mad, rsp_mad);
558 		break;
559 	case OPA_EM_ATTR_VESWPORT_ERROR_COUNTERS:
560 		vema_get_error_counters(port, recvd_mad, rsp_mad);
561 		break;
562 	default:
563 		rsp_mad->mad_hdr.status = OPA_VNIC_UNSUP_ATTR;
564 		break;
565 	}
566 }
567 
568 /**
569  * vema_set -- Process received set MAD
570  * @port:      source port on which MAD was received
571  * @recvd_mad: Received mad contains fields to set vnic parameters
572  * @rsp_mad:   Response mad to be built
573  */
574 static void vema_set(struct opa_vnic_vema_port *port,
575 		     struct opa_vnic_vema_mad *recvd_mad,
576 		     struct opa_vnic_vema_mad *rsp_mad)
577 {
578 	u16 attr_id = be16_to_cpu(recvd_mad->mad_hdr.attr_id);
579 
580 	switch (attr_id) {
581 	case OPA_EM_ATTR_CLASS_PORT_INFO:
582 		vema_set_class_port_info(port, recvd_mad, rsp_mad);
583 		break;
584 	case OPA_EM_ATTR_VESWPORT_INFO:
585 		vema_set_veswport_info(port, recvd_mad, rsp_mad);
586 		break;
587 	case OPA_EM_ATTR_VESWPORT_MAC_ENTRIES:
588 		vema_set_mac_entries(port, recvd_mad, rsp_mad);
589 		break;
590 	case OPA_EM_ATTR_DELETE_VESW:
591 		vema_set_delete_vesw(port, recvd_mad, rsp_mad);
592 		break;
593 	default:
594 		rsp_mad->mad_hdr.status = OPA_VNIC_UNSUP_ATTR;
595 		break;
596 	}
597 }
598 
599 /**
600  * vema_send -- Send handler for VEMA MAD agent
601  * @mad_agent: pointer to the mad agent
602  * @mad_wc:    pointer to mad send work completion information
603  *
604  * Free all the data structures associated with the sent MAD
605  */
606 static void vema_send(struct ib_mad_agent *mad_agent,
607 		      struct ib_mad_send_wc *mad_wc)
608 {
609 	rdma_destroy_ah(mad_wc->send_buf->ah, RDMA_DESTROY_AH_SLEEPABLE);
610 	ib_free_send_mad(mad_wc->send_buf);
611 }
612 
613 /**
614  * vema_recv -- Recv handler for VEMA MAD agent
615  * @mad_agent: pointer to the mad agent
616  * @send_buf: Send buffer if found, else NULL
617  * @mad_wc:    pointer to mad send work completion information
618  *
619  * Handle only set and get methods and respond to other methods
620  * as unsupported. Allocate response buffer and address handle
621  * for the response MAD.
622  */
623 static void vema_recv(struct ib_mad_agent *mad_agent,
624 		      struct ib_mad_send_buf *send_buf,
625 		      struct ib_mad_recv_wc *mad_wc)
626 {
627 	struct opa_vnic_vema_port *port;
628 	struct ib_ah              *ah;
629 	struct ib_mad_send_buf    *rsp;
630 	struct opa_vnic_vema_mad  *vema_mad;
631 
632 	if (!mad_wc || !mad_wc->recv_buf.mad)
633 		return;
634 
635 	port = mad_agent->context;
636 	ah = ib_create_ah_from_wc(mad_agent->qp->pd, mad_wc->wc,
637 				  mad_wc->recv_buf.grh, mad_agent->port_num);
638 	if (IS_ERR(ah))
639 		goto free_recv_mad;
640 
641 	rsp = ib_create_send_mad(mad_agent, mad_wc->wc->src_qp,
642 				 mad_wc->wc->pkey_index, 0,
643 				 IB_MGMT_VENDOR_HDR, OPA_VNIC_EMA_DATA,
644 				 GFP_KERNEL, OPA_MGMT_BASE_VERSION);
645 	if (IS_ERR(rsp))
646 		goto err_rsp;
647 
648 	rsp->ah = ah;
649 	vema_mad = rsp->mad;
650 	memcpy(vema_mad, mad_wc->recv_buf.mad, IB_MGMT_VENDOR_HDR);
651 	vema_mad->mad_hdr.method = IB_MGMT_METHOD_GET_RESP;
652 	vema_mad->mad_hdr.status = 0;
653 
654 	/* Lock ensures network adapter is not removed */
655 	mutex_lock(&port->lock);
656 
657 	switch (mad_wc->recv_buf.mad->mad_hdr.method) {
658 	case IB_MGMT_METHOD_GET:
659 		vema_get(port, (struct opa_vnic_vema_mad *)mad_wc->recv_buf.mad,
660 			 vema_mad);
661 		break;
662 	case IB_MGMT_METHOD_SET:
663 		vema_set(port, (struct opa_vnic_vema_mad *)mad_wc->recv_buf.mad,
664 			 vema_mad);
665 		break;
666 	default:
667 		vema_mad->mad_hdr.status = OPA_VNIC_UNSUP_ATTR;
668 		break;
669 	}
670 	mutex_unlock(&port->lock);
671 
672 	if (!ib_post_send_mad(rsp, NULL)) {
673 		/*
674 		 * with post send successful ah and send mad
675 		 * will be destroyed in send handler
676 		 */
677 		goto free_recv_mad;
678 	}
679 
680 	ib_free_send_mad(rsp);
681 
682 err_rsp:
683 	rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE);
684 free_recv_mad:
685 	ib_free_recv_mad(mad_wc);
686 }
687 
688 /**
689  * vema_get_port -- Gets the opa_vnic_vema_port
690  * @cport: pointer to control dev
691  * @port_num: Port number
692  *
693  * This function loops through the ports and returns
694  * the opa_vnic_vema port structure that is associated
695  * with the OPA port number
696  *
697  * Return: ptr to requested opa_vnic_vema_port strucure
698  *         if success, NULL if not
699  */
700 static struct opa_vnic_vema_port *
701 vema_get_port(struct opa_vnic_ctrl_port *cport, u8 port_num)
702 {
703 	struct opa_vnic_vema_port *port = (void *)cport + sizeof(*cport);
704 
705 	if (port_num > cport->num_ports)
706 		return NULL;
707 
708 	return port + (port_num - 1);
709 }
710 
711 /**
712  * opa_vnic_vema_send_trap -- This function sends a trap to the EM
713  * @adapter: pointer to vnic adapter
714  * @data: pointer to trap data filled by calling function
715  * @lid:  issuers lid (encap_slid from vesw_port_info)
716  *
717  * This function is called from the VNIC driver to send a trap if there
718  * is somethng the EM should be notified about. These events currently
719  * are
720  * 1) UNICAST INTERFACE MACADDRESS changes
721  * 2) MULTICAST INTERFACE MACADDRESS changes
722  * 3) ETHERNET LINK STATUS changes
723  * While allocating the send mad the remote site qpn used is 1
724  * as this is the well known QP.
725  *
726  */
727 void opa_vnic_vema_send_trap(struct opa_vnic_adapter *adapter,
728 			     struct __opa_veswport_trap *data, u32 lid)
729 {
730 	struct opa_vnic_ctrl_port *cport = adapter->cport;
731 	struct ib_mad_send_buf *send_buf;
732 	struct opa_vnic_vema_port *port;
733 	struct ib_device *ibp;
734 	struct opa_vnic_vema_mad_trap *trap_mad;
735 	struct opa_class_port_info *class;
736 	struct rdma_ah_attr ah_attr;
737 	struct ib_ah *ah;
738 	struct opa_veswport_trap *trap;
739 	u32 trap_lid;
740 	u16 pkey_idx;
741 
742 	if (!cport)
743 		goto err_exit;
744 	ibp = cport->ibdev;
745 	port = vema_get_port(cport, data->opaportnum);
746 	if (!port || !port->mad_agent)
747 		goto err_exit;
748 
749 	if (time_before(jiffies, adapter->trap_timeout)) {
750 		if (adapter->trap_count == OPA_VNIC_TRAP_BURST_LIMIT) {
751 			v_warn("Trap rate exceeded\n");
752 			goto err_exit;
753 		} else {
754 			adapter->trap_count++;
755 		}
756 	} else {
757 		adapter->trap_count = 0;
758 	}
759 
760 	class = &port->class_port_info;
761 	/* Set up address handle */
762 	memset(&ah_attr, 0, sizeof(ah_attr));
763 	ah_attr.type = rdma_ah_find_type(ibp, port->port_num);
764 	rdma_ah_set_sl(&ah_attr,
765 		       GET_TRAP_SL_FROM_CLASS_PORT_INFO(class->trap_sl_rsvd));
766 	rdma_ah_set_port_num(&ah_attr, port->port_num);
767 	trap_lid = be32_to_cpu(class->trap_lid);
768 	/*
769 	 * check for trap lid validity, must not be zero
770 	 * The trap sink could change after we fashion the MAD but since traps
771 	 * are not guaranteed we won't use a lock as anyway the change will take
772 	 * place even with locking.
773 	 */
774 	if (!trap_lid) {
775 		c_err("%s: Invalid dlid\n", __func__);
776 		goto err_exit;
777 	}
778 
779 	rdma_ah_set_dlid(&ah_attr, trap_lid);
780 	ah = rdma_create_ah(port->mad_agent->qp->pd, &ah_attr, 0);
781 	if (IS_ERR(ah)) {
782 		c_err("%s:Couldn't create new AH = %p\n", __func__, ah);
783 		c_err("%s:dlid = %d, sl = %d, port = %d\n", __func__,
784 		      rdma_ah_get_dlid(&ah_attr), rdma_ah_get_sl(&ah_attr),
785 		      rdma_ah_get_port_num(&ah_attr));
786 		goto err_exit;
787 	}
788 
789 	if (ib_find_pkey(ibp, data->opaportnum, IB_DEFAULT_PKEY_FULL,
790 			 &pkey_idx) < 0) {
791 		c_err("%s:full key not found, defaulting to partial\n",
792 		      __func__);
793 		if (ib_find_pkey(ibp, data->opaportnum, IB_DEFAULT_PKEY_PARTIAL,
794 				 &pkey_idx) < 0)
795 			pkey_idx = 1;
796 	}
797 
798 	send_buf = ib_create_send_mad(port->mad_agent, 1, pkey_idx, 0,
799 				      IB_MGMT_VENDOR_HDR, IB_MGMT_MAD_DATA,
800 				      GFP_ATOMIC, OPA_MGMT_BASE_VERSION);
801 	if (IS_ERR(send_buf)) {
802 		c_err("%s:Couldn't allocate send buf\n", __func__);
803 		goto err_sndbuf;
804 	}
805 
806 	send_buf->ah = ah;
807 
808 	/* Set up common MAD hdr */
809 	trap_mad = send_buf->mad;
810 	trap_mad->mad_hdr.base_version = OPA_MGMT_BASE_VERSION;
811 	trap_mad->mad_hdr.mgmt_class = OPA_MGMT_CLASS_INTEL_EMA;
812 	trap_mad->mad_hdr.class_version = OPA_EMA_CLASS_VERSION;
813 	trap_mad->mad_hdr.method = IB_MGMT_METHOD_TRAP;
814 	port->tid++;
815 	trap_mad->mad_hdr.tid = cpu_to_be64(port->tid);
816 	trap_mad->mad_hdr.attr_id = IB_SMP_ATTR_NOTICE;
817 
818 	/* Set up vendor OUI */
819 	trap_mad->oui[0] = INTEL_OUI_1;
820 	trap_mad->oui[1] = INTEL_OUI_2;
821 	trap_mad->oui[2] = INTEL_OUI_3;
822 
823 	/* Setup notice attribute portion */
824 	trap_mad->notice.gen_type = OPA_INTEL_EMA_NOTICE_TYPE_INFO << 1;
825 	trap_mad->notice.oui_1 = INTEL_OUI_1;
826 	trap_mad->notice.oui_2 = INTEL_OUI_2;
827 	trap_mad->notice.oui_3 = INTEL_OUI_3;
828 	trap_mad->notice.issuer_lid = cpu_to_be32(lid);
829 
830 	/* copy the actual trap data */
831 	trap = (struct opa_veswport_trap *)trap_mad->notice.raw_data;
832 	trap->fabric_id = cpu_to_be16(data->fabric_id);
833 	trap->veswid = cpu_to_be16(data->veswid);
834 	trap->veswportnum = cpu_to_be32(data->veswportnum);
835 	trap->opaportnum = cpu_to_be16(data->opaportnum);
836 	trap->veswportindex = data->veswportindex;
837 	trap->opcode = data->opcode;
838 
839 	/* If successful send set up rate limit timeout else bail */
840 	if (ib_post_send_mad(send_buf, NULL)) {
841 		ib_free_send_mad(send_buf);
842 	} else {
843 		if (adapter->trap_count)
844 			return;
845 		adapter->trap_timeout = jiffies +
846 					usecs_to_jiffies(OPA_VNIC_TRAP_TIMEOUT);
847 		return;
848 	}
849 
850 err_sndbuf:
851 	rdma_destroy_ah(ah, 0);
852 err_exit:
853 	v_err("Aborting trap\n");
854 }
855 
856 static void opa_vnic_event(struct ib_event_handler *handler,
857 			   struct ib_event *record)
858 {
859 	struct opa_vnic_vema_port *port =
860 		container_of(handler, struct opa_vnic_vema_port, event_handler);
861 	struct opa_vnic_ctrl_port *cport = port->cport;
862 	struct opa_vnic_adapter *adapter;
863 	unsigned long index;
864 
865 	if (record->element.port_num != port->port_num)
866 		return;
867 
868 	c_dbg("OPA_VNIC received event %d on device %s port %d\n",
869 	      record->event, dev_name(&record->device->dev),
870 	      record->element.port_num);
871 
872 	if (record->event != IB_EVENT_PORT_ERR &&
873 	    record->event != IB_EVENT_PORT_ACTIVE)
874 		return;
875 
876 	xa_for_each(&port->vports, index, adapter) {
877 		if (record->event == IB_EVENT_PORT_ACTIVE)
878 			netif_carrier_on(adapter->netdev);
879 		else
880 			netif_carrier_off(adapter->netdev);
881 	}
882 }
883 
884 /**
885  * vema_unregister -- Unregisters agent
886  * @cport: pointer to control port
887  *
888  * This deletes the registration by VEMA for MADs
889  */
890 static void vema_unregister(struct opa_vnic_ctrl_port *cport)
891 {
892 	struct opa_vnic_adapter *adapter;
893 	unsigned long index;
894 	int i;
895 
896 	for (i = 1; i <= cport->num_ports; i++) {
897 		struct opa_vnic_vema_port *port = vema_get_port(cport, i);
898 
899 		if (!port->mad_agent)
900 			continue;
901 
902 		/* Lock ensures no MAD is being processed */
903 		mutex_lock(&port->lock);
904 		xa_for_each(&port->vports, index, adapter)
905 			opa_vnic_rem_netdev(adapter);
906 		mutex_unlock(&port->lock);
907 
908 		ib_unregister_mad_agent(port->mad_agent);
909 		port->mad_agent = NULL;
910 		mutex_destroy(&port->lock);
911 		xa_destroy(&port->vports);
912 		ib_unregister_event_handler(&port->event_handler);
913 	}
914 }
915 
916 /**
917  * vema_register -- Registers agent
918  * @cport: pointer to control port
919  *
920  * This function registers the handlers for the VEMA MADs
921  *
922  * Return: returns 0 on success. non zero otherwise
923  */
924 static int vema_register(struct opa_vnic_ctrl_port *cport)
925 {
926 	struct ib_mad_reg_req reg_req = {
927 		.mgmt_class = OPA_MGMT_CLASS_INTEL_EMA,
928 		.mgmt_class_version = OPA_MGMT_BASE_VERSION,
929 		.oui = { INTEL_OUI_1, INTEL_OUI_2, INTEL_OUI_3 }
930 	};
931 	int i;
932 
933 	set_bit(IB_MGMT_METHOD_GET, reg_req.method_mask);
934 	set_bit(IB_MGMT_METHOD_SET, reg_req.method_mask);
935 
936 	/* register ib event handler and mad agent for each port on dev */
937 	for (i = 1; i <= cport->num_ports; i++) {
938 		struct opa_vnic_vema_port *port = vema_get_port(cport, i);
939 		int ret;
940 
941 		port->cport = cport;
942 		port->port_num = i;
943 
944 		INIT_IB_EVENT_HANDLER(&port->event_handler,
945 				      cport->ibdev, opa_vnic_event);
946 		ib_register_event_handler(&port->event_handler);
947 
948 		xa_init(&port->vports);
949 		mutex_init(&port->lock);
950 		port->mad_agent = ib_register_mad_agent(cport->ibdev, i,
951 							IB_QPT_GSI, &reg_req,
952 							IB_MGMT_RMPP_VERSION,
953 							vema_send, vema_recv,
954 							port, 0);
955 		if (IS_ERR(port->mad_agent)) {
956 			ret = PTR_ERR(port->mad_agent);
957 			port->mad_agent = NULL;
958 			mutex_destroy(&port->lock);
959 			vema_unregister(cport);
960 			return ret;
961 		}
962 	}
963 
964 	return 0;
965 }
966 
967 /**
968  * opa_vnic_ctrl_config_dev -- This function sends a trap to the EM
969  * by way of ib_modify_port to indicate support for ethernet on the
970  * fabric.
971  * @cport: pointer to control port
972  * @en: enable or disable ethernet on fabric support
973  */
974 static void opa_vnic_ctrl_config_dev(struct opa_vnic_ctrl_port *cport, bool en)
975 {
976 	struct ib_port_modify pm = { 0 };
977 	int i;
978 
979 	if (en)
980 		pm.set_port_cap_mask = OPA_CAP_MASK3_IsEthOnFabricSupported;
981 	else
982 		pm.clr_port_cap_mask = OPA_CAP_MASK3_IsEthOnFabricSupported;
983 
984 	for (i = 1; i <= cport->num_ports; i++)
985 		ib_modify_port(cport->ibdev, i, IB_PORT_OPA_MASK_CHG, &pm);
986 }
987 
988 /**
989  * opa_vnic_vema_add_one -- Handle new ib device
990  * @device: ib device pointer
991  *
992  * Allocate the vnic control port and initialize it.
993  */
994 static void opa_vnic_vema_add_one(struct ib_device *device)
995 {
996 	struct opa_vnic_ctrl_port *cport;
997 	int rc, size = sizeof(*cport);
998 
999 	if (!rdma_cap_opa_vnic(device))
1000 		return;
1001 
1002 	size += device->phys_port_cnt * sizeof(struct opa_vnic_vema_port);
1003 	cport = kzalloc(size, GFP_KERNEL);
1004 	if (!cport)
1005 		return;
1006 
1007 	cport->num_ports = device->phys_port_cnt;
1008 	cport->ibdev = device;
1009 
1010 	/* Initialize opa vnic management agent (vema) */
1011 	rc = vema_register(cport);
1012 	if (!rc)
1013 		c_info("VNIC client initialized\n");
1014 
1015 	ib_set_client_data(device, &opa_vnic_client, cport);
1016 	opa_vnic_ctrl_config_dev(cport, true);
1017 }
1018 
1019 /**
1020  * opa_vnic_vema_rem_one -- Handle ib device removal
1021  * @device: ib device pointer
1022  * @client_data: ib client data
1023  *
1024  * Uninitialize and free the vnic control port.
1025  */
1026 static void opa_vnic_vema_rem_one(struct ib_device *device,
1027 				  void *client_data)
1028 {
1029 	struct opa_vnic_ctrl_port *cport = client_data;
1030 
1031 	if (!cport)
1032 		return;
1033 
1034 	c_info("removing VNIC client\n");
1035 	opa_vnic_ctrl_config_dev(cport, false);
1036 	vema_unregister(cport);
1037 	kfree(cport);
1038 }
1039 
1040 static int __init opa_vnic_init(void)
1041 {
1042 	int rc;
1043 
1044 	pr_info("OPA Virtual Network Driver - v%s\n",
1045 		opa_vnic_driver_version);
1046 
1047 	rc = ib_register_client(&opa_vnic_client);
1048 	if (rc)
1049 		pr_err("VNIC driver register failed %d\n", rc);
1050 
1051 	return rc;
1052 }
1053 module_init(opa_vnic_init);
1054 
1055 static void opa_vnic_deinit(void)
1056 {
1057 	ib_unregister_client(&opa_vnic_client);
1058 }
1059 module_exit(opa_vnic_deinit);
1060 
1061 MODULE_LICENSE("Dual BSD/GPL");
1062 MODULE_AUTHOR("Intel Corporation");
1063 MODULE_DESCRIPTION("Intel OPA Virtual Network driver");
1064