• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  *
32  * Author: Upinder Malhi <umalhi@cisco.com>
33  * Author: Anant Deepak <anadeepa@cisco.com>
34  * Author: Cesare Cantu' <cantuc@cisco.com>
35  * Author: Jeff Squyres <jsquyres@cisco.com>
36  * Author: Kiran Thirumalai <kithirum@cisco.com>
37  * Author: Xuyang Wang <xuywang@cisco.com>
38  * Author: Reese Faucette <rfaucett@cisco.com>
39  *
40  */
41 
42 #include <linux/module.h>
43 #include <linux/inetdevice.h>
44 #include <linux/init.h>
45 #include <linux/slab.h>
46 #include <linux/errno.h>
47 #include <linux/pci.h>
48 #include <linux/netdevice.h>
49 
50 #include <rdma/ib_user_verbs.h>
51 #include <rdma/ib_addr.h>
52 
53 #include "usnic_abi.h"
54 #include "usnic_common_util.h"
55 #include "usnic_ib.h"
56 #include "usnic_ib_qp_grp.h"
57 #include "usnic_log.h"
58 #include "usnic_fwd.h"
59 #include "usnic_debugfs.h"
60 #include "usnic_ib_verbs.h"
61 #include "usnic_transport.h"
62 #include "usnic_uiom.h"
63 #include "usnic_ib_sysfs.h"
64 
65 unsigned int usnic_log_lvl = USNIC_LOG_LVL_ERR;
66 unsigned int usnic_ib_share_vf = 1;
67 
68 static const char usnic_version[] =
69 	DRV_NAME ": Cisco VIC (USNIC) Verbs Driver v"
70 	DRV_VERSION " (" DRV_RELDATE ")\n";
71 
72 static DEFINE_MUTEX(usnic_ib_ibdev_list_lock);
73 static LIST_HEAD(usnic_ib_ibdev_list);
74 
75 /* Callback dump funcs */
usnic_ib_dump_vf_hdr(void * obj,char * buf,int buf_sz)76 static int usnic_ib_dump_vf_hdr(void *obj, char *buf, int buf_sz)
77 {
78 	struct usnic_ib_vf *vf = obj;
79 	return scnprintf(buf, buf_sz, "PF: %s ", dev_name(&vf->pf->ib_dev.dev));
80 }
81 /* End callback dump funcs */
82 
usnic_ib_dump_vf(struct usnic_ib_vf * vf,char * buf,int buf_sz)83 static void usnic_ib_dump_vf(struct usnic_ib_vf *vf, char *buf, int buf_sz)
84 {
85 	usnic_vnic_dump(vf->vnic, buf, buf_sz, vf,
86 			usnic_ib_dump_vf_hdr,
87 			usnic_ib_qp_grp_dump_hdr, usnic_ib_qp_grp_dump_rows);
88 }
89 
usnic_ib_log_vf(struct usnic_ib_vf * vf)90 void usnic_ib_log_vf(struct usnic_ib_vf *vf)
91 {
92 	char *buf = kzalloc(1000, GFP_KERNEL);
93 
94 	if (!buf)
95 		return;
96 
97 	usnic_ib_dump_vf(vf, buf, 1000);
98 	usnic_dbg("%s\n", buf);
99 
100 	kfree(buf);
101 }
102 
103 /* Start of netdev section */
usnic_ib_qp_grp_modify_active_to_err(struct usnic_ib_dev * us_ibdev)104 static void usnic_ib_qp_grp_modify_active_to_err(struct usnic_ib_dev *us_ibdev)
105 {
106 	struct usnic_ib_ucontext *ctx;
107 	struct usnic_ib_qp_grp *qp_grp;
108 	enum ib_qp_state cur_state;
109 	int status;
110 
111 	BUG_ON(!mutex_is_locked(&us_ibdev->usdev_lock));
112 
113 	list_for_each_entry(ctx, &us_ibdev->ctx_list, link) {
114 		list_for_each_entry(qp_grp, &ctx->qp_grp_list, link) {
115 			cur_state = qp_grp->state;
116 			if (cur_state == IB_QPS_INIT ||
117 				cur_state == IB_QPS_RTR ||
118 				cur_state == IB_QPS_RTS) {
119 				status = usnic_ib_qp_grp_modify(qp_grp,
120 								IB_QPS_ERR,
121 								NULL);
122 				if (status) {
123 					usnic_err("Failed to transition qp grp %u from %s to %s\n",
124 						qp_grp->grp_id,
125 						usnic_ib_qp_grp_state_to_string
126 						(cur_state),
127 						usnic_ib_qp_grp_state_to_string
128 						(IB_QPS_ERR));
129 				}
130 			}
131 		}
132 	}
133 }
134 
usnic_ib_handle_usdev_event(struct usnic_ib_dev * us_ibdev,unsigned long event)135 static void usnic_ib_handle_usdev_event(struct usnic_ib_dev *us_ibdev,
136 					unsigned long event)
137 {
138 	struct net_device *netdev;
139 	struct ib_event ib_event;
140 
141 	memset(&ib_event, 0, sizeof(ib_event));
142 
143 	mutex_lock(&us_ibdev->usdev_lock);
144 	netdev = us_ibdev->netdev;
145 	switch (event) {
146 	case NETDEV_REBOOT:
147 		usnic_info("PF Reset on %s\n", dev_name(&us_ibdev->ib_dev.dev));
148 		usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
149 		ib_event.event = IB_EVENT_PORT_ERR;
150 		ib_event.device = &us_ibdev->ib_dev;
151 		ib_event.element.port_num = 1;
152 		ib_dispatch_event(&ib_event);
153 		break;
154 	case NETDEV_UP:
155 	case NETDEV_DOWN:
156 	case NETDEV_CHANGE:
157 		if (!us_ibdev->ufdev->link_up &&
158 				netif_carrier_ok(netdev)) {
159 			usnic_fwd_carrier_up(us_ibdev->ufdev);
160 			usnic_info("Link UP on %s\n",
161 				   dev_name(&us_ibdev->ib_dev.dev));
162 			ib_event.event = IB_EVENT_PORT_ACTIVE;
163 			ib_event.device = &us_ibdev->ib_dev;
164 			ib_event.element.port_num = 1;
165 			ib_dispatch_event(&ib_event);
166 		} else if (us_ibdev->ufdev->link_up &&
167 				!netif_carrier_ok(netdev)) {
168 			usnic_fwd_carrier_down(us_ibdev->ufdev);
169 			usnic_info("Link DOWN on %s\n",
170 				   dev_name(&us_ibdev->ib_dev.dev));
171 			usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
172 			ib_event.event = IB_EVENT_PORT_ERR;
173 			ib_event.device = &us_ibdev->ib_dev;
174 			ib_event.element.port_num = 1;
175 			ib_dispatch_event(&ib_event);
176 		} else {
177 			usnic_dbg("Ignoring %s on %s\n",
178 					netdev_cmd_to_name(event),
179 					dev_name(&us_ibdev->ib_dev.dev));
180 		}
181 		break;
182 	case NETDEV_CHANGEADDR:
183 		if (!memcmp(us_ibdev->ufdev->mac, netdev->dev_addr,
184 				sizeof(us_ibdev->ufdev->mac))) {
185 			usnic_dbg("Ignoring addr change on %s\n",
186 				  dev_name(&us_ibdev->ib_dev.dev));
187 		} else {
188 			usnic_info(" %s old mac: %pM new mac: %pM\n",
189 					dev_name(&us_ibdev->ib_dev.dev),
190 					us_ibdev->ufdev->mac,
191 					netdev->dev_addr);
192 			usnic_fwd_set_mac(us_ibdev->ufdev, netdev->dev_addr);
193 			usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
194 			ib_event.event = IB_EVENT_GID_CHANGE;
195 			ib_event.device = &us_ibdev->ib_dev;
196 			ib_event.element.port_num = 1;
197 			ib_dispatch_event(&ib_event);
198 		}
199 
200 		break;
201 	case NETDEV_CHANGEMTU:
202 		if (us_ibdev->ufdev->mtu != netdev->mtu) {
203 			usnic_info("MTU Change on %s old: %u new: %u\n",
204 					dev_name(&us_ibdev->ib_dev.dev),
205 					us_ibdev->ufdev->mtu, netdev->mtu);
206 			usnic_fwd_set_mtu(us_ibdev->ufdev, netdev->mtu);
207 			usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
208 		} else {
209 			usnic_dbg("Ignoring MTU change on %s\n",
210 				  dev_name(&us_ibdev->ib_dev.dev));
211 		}
212 		break;
213 	default:
214 		usnic_dbg("Ignoring event %s on %s",
215 				netdev_cmd_to_name(event),
216 				dev_name(&us_ibdev->ib_dev.dev));
217 	}
218 	mutex_unlock(&us_ibdev->usdev_lock);
219 }
220 
usnic_ib_netdevice_event(struct notifier_block * notifier,unsigned long event,void * ptr)221 static int usnic_ib_netdevice_event(struct notifier_block *notifier,
222 					unsigned long event, void *ptr)
223 {
224 	struct usnic_ib_dev *us_ibdev;
225 	struct ib_device *ibdev;
226 
227 	struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
228 
229 	ibdev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_USNIC);
230 	if (!ibdev)
231 		return NOTIFY_DONE;
232 
233 	us_ibdev = container_of(ibdev, struct usnic_ib_dev, ib_dev);
234 	usnic_ib_handle_usdev_event(us_ibdev, event);
235 	ib_device_put(ibdev);
236 	return NOTIFY_DONE;
237 }
238 
239 static struct notifier_block usnic_ib_netdevice_notifier = {
240 	.notifier_call = usnic_ib_netdevice_event
241 };
242 /* End of netdev section */
243 
244 /* Start of inet section */
usnic_ib_handle_inet_event(struct usnic_ib_dev * us_ibdev,unsigned long event,void * ptr)245 static int usnic_ib_handle_inet_event(struct usnic_ib_dev *us_ibdev,
246 					unsigned long event, void *ptr)
247 {
248 	struct in_ifaddr *ifa = ptr;
249 	struct ib_event ib_event;
250 
251 	mutex_lock(&us_ibdev->usdev_lock);
252 
253 	switch (event) {
254 	case NETDEV_DOWN:
255 		usnic_info("%s via ip notifiers",
256 				netdev_cmd_to_name(event));
257 		usnic_fwd_del_ipaddr(us_ibdev->ufdev);
258 		usnic_ib_qp_grp_modify_active_to_err(us_ibdev);
259 		ib_event.event = IB_EVENT_GID_CHANGE;
260 		ib_event.device = &us_ibdev->ib_dev;
261 		ib_event.element.port_num = 1;
262 		ib_dispatch_event(&ib_event);
263 		break;
264 	case NETDEV_UP:
265 		usnic_fwd_add_ipaddr(us_ibdev->ufdev, ifa->ifa_address);
266 		usnic_info("%s via ip notifiers: ip %pI4",
267 				netdev_cmd_to_name(event),
268 				&us_ibdev->ufdev->inaddr);
269 		ib_event.event = IB_EVENT_GID_CHANGE;
270 		ib_event.device = &us_ibdev->ib_dev;
271 		ib_event.element.port_num = 1;
272 		ib_dispatch_event(&ib_event);
273 		break;
274 	default:
275 		usnic_info("Ignoring event %s on %s",
276 				netdev_cmd_to_name(event),
277 				dev_name(&us_ibdev->ib_dev.dev));
278 	}
279 	mutex_unlock(&us_ibdev->usdev_lock);
280 
281 	return NOTIFY_DONE;
282 }
283 
usnic_ib_inetaddr_event(struct notifier_block * notifier,unsigned long event,void * ptr)284 static int usnic_ib_inetaddr_event(struct notifier_block *notifier,
285 					unsigned long event, void *ptr)
286 {
287 	struct usnic_ib_dev *us_ibdev;
288 	struct in_ifaddr *ifa = ptr;
289 	struct net_device *netdev = ifa->ifa_dev->dev;
290 	struct ib_device *ibdev;
291 
292 	ibdev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_USNIC);
293 	if (!ibdev)
294 		return NOTIFY_DONE;
295 
296 	us_ibdev = container_of(ibdev, struct usnic_ib_dev, ib_dev);
297 	usnic_ib_handle_inet_event(us_ibdev, event, ptr);
298 	ib_device_put(ibdev);
299 	return NOTIFY_DONE;
300 }
301 static struct notifier_block usnic_ib_inetaddr_notifier = {
302 	.notifier_call = usnic_ib_inetaddr_event
303 };
304 /* End of inet section*/
305 
usnic_port_immutable(struct ib_device * ibdev,u8 port_num,struct ib_port_immutable * immutable)306 static int usnic_port_immutable(struct ib_device *ibdev, u8 port_num,
307 			        struct ib_port_immutable *immutable)
308 {
309 	struct ib_port_attr attr;
310 	int err;
311 
312 	immutable->core_cap_flags = RDMA_CORE_PORT_USNIC;
313 
314 	err = ib_query_port(ibdev, port_num, &attr);
315 	if (err)
316 		return err;
317 
318 	immutable->gid_tbl_len = attr.gid_tbl_len;
319 
320 	return 0;
321 }
322 
usnic_get_dev_fw_str(struct ib_device * device,char * str)323 static void usnic_get_dev_fw_str(struct ib_device *device, char *str)
324 {
325 	struct usnic_ib_dev *us_ibdev =
326 		container_of(device, struct usnic_ib_dev, ib_dev);
327 	struct ethtool_drvinfo info;
328 
329 	mutex_lock(&us_ibdev->usdev_lock);
330 	us_ibdev->netdev->ethtool_ops->get_drvinfo(us_ibdev->netdev, &info);
331 	mutex_unlock(&us_ibdev->usdev_lock);
332 
333 	snprintf(str, IB_FW_VERSION_NAME_MAX, "%s", info.fw_version);
334 }
335 
336 static const struct ib_device_ops usnic_dev_ops = {
337 	.owner = THIS_MODULE,
338 	.driver_id = RDMA_DRIVER_USNIC,
339 	.uverbs_abi_ver = USNIC_UVERBS_ABI_VERSION,
340 
341 	.alloc_pd = usnic_ib_alloc_pd,
342 	.alloc_ucontext = usnic_ib_alloc_ucontext,
343 	.create_cq = usnic_ib_create_cq,
344 	.create_qp = usnic_ib_create_qp,
345 	.dealloc_pd = usnic_ib_dealloc_pd,
346 	.dealloc_ucontext = usnic_ib_dealloc_ucontext,
347 	.dereg_mr = usnic_ib_dereg_mr,
348 	.destroy_cq = usnic_ib_destroy_cq,
349 	.destroy_qp = usnic_ib_destroy_qp,
350 	.get_dev_fw_str = usnic_get_dev_fw_str,
351 	.get_link_layer = usnic_ib_port_link_layer,
352 	.get_port_immutable = usnic_port_immutable,
353 	.mmap = usnic_ib_mmap,
354 	.modify_qp = usnic_ib_modify_qp,
355 	.query_device = usnic_ib_query_device,
356 	.query_gid = usnic_ib_query_gid,
357 	.query_port = usnic_ib_query_port,
358 	.query_qp = usnic_ib_query_qp,
359 	.reg_user_mr = usnic_ib_reg_mr,
360 	INIT_RDMA_OBJ_SIZE(ib_pd, usnic_ib_pd, ibpd),
361 	INIT_RDMA_OBJ_SIZE(ib_cq, usnic_ib_cq, ibcq),
362 	INIT_RDMA_OBJ_SIZE(ib_ucontext, usnic_ib_ucontext, ibucontext),
363 };
364 
365 /* Start of PF discovery section */
usnic_ib_device_add(struct pci_dev * dev)366 static void *usnic_ib_device_add(struct pci_dev *dev)
367 {
368 	struct usnic_ib_dev *us_ibdev;
369 	union ib_gid gid;
370 	struct in_device *ind;
371 	struct net_device *netdev;
372 	int ret;
373 
374 	usnic_dbg("\n");
375 	netdev = pci_get_drvdata(dev);
376 
377 	us_ibdev = ib_alloc_device(usnic_ib_dev, ib_dev);
378 	if (!us_ibdev) {
379 		usnic_err("Device %s context alloc failed\n",
380 				netdev_name(pci_get_drvdata(dev)));
381 		return ERR_PTR(-EFAULT);
382 	}
383 
384 	us_ibdev->ufdev = usnic_fwd_dev_alloc(dev);
385 	if (!us_ibdev->ufdev) {
386 		usnic_err("Failed to alloc ufdev for %s\n", pci_name(dev));
387 		goto err_dealloc;
388 	}
389 
390 	mutex_init(&us_ibdev->usdev_lock);
391 	INIT_LIST_HEAD(&us_ibdev->vf_dev_list);
392 	INIT_LIST_HEAD(&us_ibdev->ctx_list);
393 
394 	us_ibdev->pdev = dev;
395 	us_ibdev->netdev = pci_get_drvdata(dev);
396 	us_ibdev->ib_dev.node_type = RDMA_NODE_USNIC_UDP;
397 	us_ibdev->ib_dev.phys_port_cnt = USNIC_IB_PORT_CNT;
398 	us_ibdev->ib_dev.num_comp_vectors = USNIC_IB_NUM_COMP_VECTORS;
399 	us_ibdev->ib_dev.dev.parent = &dev->dev;
400 
401 	us_ibdev->ib_dev.uverbs_cmd_mask =
402 		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
403 		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) |
404 		(1ull << IB_USER_VERBS_CMD_QUERY_PORT) |
405 		(1ull << IB_USER_VERBS_CMD_ALLOC_PD) |
406 		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD) |
407 		(1ull << IB_USER_VERBS_CMD_REG_MR) |
408 		(1ull << IB_USER_VERBS_CMD_DEREG_MR) |
409 		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
410 		(1ull << IB_USER_VERBS_CMD_CREATE_CQ) |
411 		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ) |
412 		(1ull << IB_USER_VERBS_CMD_CREATE_QP) |
413 		(1ull << IB_USER_VERBS_CMD_MODIFY_QP) |
414 		(1ull << IB_USER_VERBS_CMD_QUERY_QP) |
415 		(1ull << IB_USER_VERBS_CMD_DESTROY_QP) |
416 		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST) |
417 		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST) |
418 		(1ull << IB_USER_VERBS_CMD_OPEN_QP);
419 
420 	ib_set_device_ops(&us_ibdev->ib_dev, &usnic_dev_ops);
421 
422 	rdma_set_device_sysfs_group(&us_ibdev->ib_dev, &usnic_attr_group);
423 
424 	ret = ib_device_set_netdev(&us_ibdev->ib_dev, us_ibdev->netdev, 1);
425 	if (ret)
426 		goto err_fwd_dealloc;
427 
428 	dma_set_max_seg_size(&dev->dev, SZ_2G);
429 	if (ib_register_device(&us_ibdev->ib_dev, "usnic_%d", &dev->dev))
430 		goto err_fwd_dealloc;
431 
432 	usnic_fwd_set_mtu(us_ibdev->ufdev, us_ibdev->netdev->mtu);
433 	usnic_fwd_set_mac(us_ibdev->ufdev, us_ibdev->netdev->dev_addr);
434 	if (netif_carrier_ok(us_ibdev->netdev))
435 		usnic_fwd_carrier_up(us_ibdev->ufdev);
436 
437 	rcu_read_lock();
438 	ind = __in_dev_get_rcu(netdev);
439 	if (ind) {
440 		const struct in_ifaddr *ifa;
441 
442 		ifa = rcu_dereference(ind->ifa_list);
443 		if (ifa)
444 			usnic_fwd_add_ipaddr(us_ibdev->ufdev, ifa->ifa_address);
445 	}
446 	rcu_read_unlock();
447 
448 	usnic_mac_ip_to_gid(us_ibdev->netdev->perm_addr,
449 				us_ibdev->ufdev->inaddr, &gid.raw[0]);
450 	memcpy(&us_ibdev->ib_dev.node_guid, &gid.global.interface_id,
451 		sizeof(gid.global.interface_id));
452 	kref_init(&us_ibdev->vf_cnt);
453 
454 	usnic_info("Added ibdev: %s netdev: %s with mac %pM Link: %u MTU: %u\n",
455 		   dev_name(&us_ibdev->ib_dev.dev),
456 		   netdev_name(us_ibdev->netdev), us_ibdev->ufdev->mac,
457 		   us_ibdev->ufdev->link_up, us_ibdev->ufdev->mtu);
458 	return us_ibdev;
459 
460 err_fwd_dealloc:
461 	usnic_fwd_dev_free(us_ibdev->ufdev);
462 err_dealloc:
463 	usnic_err("failed -- deallocing device\n");
464 	ib_dealloc_device(&us_ibdev->ib_dev);
465 	return NULL;
466 }
467 
usnic_ib_device_remove(struct usnic_ib_dev * us_ibdev)468 static void usnic_ib_device_remove(struct usnic_ib_dev *us_ibdev)
469 {
470 	usnic_info("Unregistering %s\n", dev_name(&us_ibdev->ib_dev.dev));
471 	usnic_ib_sysfs_unregister_usdev(us_ibdev);
472 	usnic_fwd_dev_free(us_ibdev->ufdev);
473 	ib_unregister_device(&us_ibdev->ib_dev);
474 	ib_dealloc_device(&us_ibdev->ib_dev);
475 }
476 
usnic_ib_undiscover_pf(struct kref * kref)477 static void usnic_ib_undiscover_pf(struct kref *kref)
478 {
479 	struct usnic_ib_dev *us_ibdev, *tmp;
480 	struct pci_dev *dev;
481 	bool found = false;
482 
483 	dev = container_of(kref, struct usnic_ib_dev, vf_cnt)->pdev;
484 	mutex_lock(&usnic_ib_ibdev_list_lock);
485 	list_for_each_entry_safe(us_ibdev, tmp,
486 				&usnic_ib_ibdev_list, ib_dev_link) {
487 		if (us_ibdev->pdev == dev) {
488 			list_del(&us_ibdev->ib_dev_link);
489 			found = true;
490 			break;
491 		}
492 	}
493 
494 
495 	mutex_unlock(&usnic_ib_ibdev_list_lock);
496 	if (found)
497 		usnic_ib_device_remove(us_ibdev);
498 	else
499 		WARN(1, "Failed to remove PF %s\n", pci_name(dev));
500 }
501 
usnic_ib_discover_pf(struct usnic_vnic * vnic)502 static struct usnic_ib_dev *usnic_ib_discover_pf(struct usnic_vnic *vnic)
503 {
504 	struct usnic_ib_dev *us_ibdev;
505 	struct pci_dev *parent_pci, *vf_pci;
506 	int err;
507 
508 	vf_pci = usnic_vnic_get_pdev(vnic);
509 	parent_pci = pci_physfn(vf_pci);
510 
511 	BUG_ON(!parent_pci);
512 
513 	mutex_lock(&usnic_ib_ibdev_list_lock);
514 	list_for_each_entry(us_ibdev, &usnic_ib_ibdev_list, ib_dev_link) {
515 		if (us_ibdev->pdev == parent_pci) {
516 			kref_get(&us_ibdev->vf_cnt);
517 			goto out;
518 		}
519 	}
520 
521 	us_ibdev = usnic_ib_device_add(parent_pci);
522 	if (IS_ERR_OR_NULL(us_ibdev)) {
523 		us_ibdev = us_ibdev ? us_ibdev : ERR_PTR(-EFAULT);
524 		goto out;
525 	}
526 
527 	err = usnic_ib_sysfs_register_usdev(us_ibdev);
528 	if (err) {
529 		usnic_ib_device_remove(us_ibdev);
530 		us_ibdev = ERR_PTR(err);
531 		goto out;
532 	}
533 
534 	list_add(&us_ibdev->ib_dev_link, &usnic_ib_ibdev_list);
535 out:
536 	mutex_unlock(&usnic_ib_ibdev_list_lock);
537 	return us_ibdev;
538 }
539 /* End of PF discovery section */
540 
541 /* Start of PCI section */
542 
543 static const struct pci_device_id usnic_ib_pci_ids[] = {
544 	{PCI_DEVICE(PCI_VENDOR_ID_CISCO, PCI_DEVICE_ID_CISCO_VIC_USPACE_NIC)},
545 	{0,}
546 };
547 
usnic_ib_pci_probe(struct pci_dev * pdev,const struct pci_device_id * id)548 static int usnic_ib_pci_probe(struct pci_dev *pdev,
549 				const struct pci_device_id *id)
550 {
551 	int err;
552 	struct usnic_ib_dev *pf;
553 	struct usnic_ib_vf *vf;
554 	enum usnic_vnic_res_type res_type;
555 
556 	vf = kzalloc(sizeof(*vf), GFP_KERNEL);
557 	if (!vf)
558 		return -ENOMEM;
559 
560 	err = pci_enable_device(pdev);
561 	if (err) {
562 		usnic_err("Failed to enable %s with err %d\n",
563 				pci_name(pdev), err);
564 		goto out_clean_vf;
565 	}
566 
567 	err = pci_request_regions(pdev, DRV_NAME);
568 	if (err) {
569 		usnic_err("Failed to request region for %s with err %d\n",
570 				pci_name(pdev), err);
571 		goto out_disable_device;
572 	}
573 
574 	pci_set_master(pdev);
575 	pci_set_drvdata(pdev, vf);
576 
577 	vf->vnic = usnic_vnic_alloc(pdev);
578 	if (IS_ERR_OR_NULL(vf->vnic)) {
579 		err = vf->vnic ? PTR_ERR(vf->vnic) : -ENOMEM;
580 		usnic_err("Failed to alloc vnic for %s with err %d\n",
581 				pci_name(pdev), err);
582 		goto out_release_regions;
583 	}
584 
585 	pf = usnic_ib_discover_pf(vf->vnic);
586 	if (IS_ERR_OR_NULL(pf)) {
587 		usnic_err("Failed to discover pf of vnic %s with err%ld\n",
588 				pci_name(pdev), PTR_ERR(pf));
589 		err = pf ? PTR_ERR(pf) : -EFAULT;
590 		goto out_clean_vnic;
591 	}
592 
593 	vf->pf = pf;
594 	spin_lock_init(&vf->lock);
595 	mutex_lock(&pf->usdev_lock);
596 	list_add_tail(&vf->link, &pf->vf_dev_list);
597 	/*
598 	 * Save max settings (will be same for each VF, easier to re-write than
599 	 * to say "if (!set) { set_values(); set=1; }
600 	 */
601 	for (res_type = USNIC_VNIC_RES_TYPE_EOL+1;
602 			res_type < USNIC_VNIC_RES_TYPE_MAX;
603 			res_type++) {
604 		pf->vf_res_cnt[res_type] = usnic_vnic_res_cnt(vf->vnic,
605 								res_type);
606 	}
607 
608 	mutex_unlock(&pf->usdev_lock);
609 
610 	usnic_info("Registering usnic VF %s into PF %s\n", pci_name(pdev),
611 		   dev_name(&pf->ib_dev.dev));
612 	usnic_ib_log_vf(vf);
613 	return 0;
614 
615 out_clean_vnic:
616 	usnic_vnic_free(vf->vnic);
617 out_release_regions:
618 	pci_set_drvdata(pdev, NULL);
619 	pci_clear_master(pdev);
620 	pci_release_regions(pdev);
621 out_disable_device:
622 	pci_disable_device(pdev);
623 out_clean_vf:
624 	kfree(vf);
625 	return err;
626 }
627 
usnic_ib_pci_remove(struct pci_dev * pdev)628 static void usnic_ib_pci_remove(struct pci_dev *pdev)
629 {
630 	struct usnic_ib_vf *vf = pci_get_drvdata(pdev);
631 	struct usnic_ib_dev *pf = vf->pf;
632 
633 	mutex_lock(&pf->usdev_lock);
634 	list_del(&vf->link);
635 	mutex_unlock(&pf->usdev_lock);
636 
637 	kref_put(&pf->vf_cnt, usnic_ib_undiscover_pf);
638 	usnic_vnic_free(vf->vnic);
639 	pci_set_drvdata(pdev, NULL);
640 	pci_clear_master(pdev);
641 	pci_release_regions(pdev);
642 	pci_disable_device(pdev);
643 	kfree(vf);
644 
645 	usnic_info("Removed VF %s\n", pci_name(pdev));
646 }
647 
648 /* PCI driver entry points */
649 static struct pci_driver usnic_ib_pci_driver = {
650 	.name = DRV_NAME,
651 	.id_table = usnic_ib_pci_ids,
652 	.probe = usnic_ib_pci_probe,
653 	.remove = usnic_ib_pci_remove,
654 };
655 /* End of PCI section */
656 
657 /* Start of module section */
usnic_ib_init(void)658 static int __init usnic_ib_init(void)
659 {
660 	int err;
661 
662 	printk_once(KERN_INFO "%s", usnic_version);
663 
664 	err = usnic_uiom_init(DRV_NAME);
665 	if (err) {
666 		usnic_err("Unable to initialize umem with err %d\n", err);
667 		return err;
668 	}
669 
670 	err = pci_register_driver(&usnic_ib_pci_driver);
671 	if (err) {
672 		usnic_err("Unable to register with PCI\n");
673 		goto out_umem_fini;
674 	}
675 
676 	err = register_netdevice_notifier(&usnic_ib_netdevice_notifier);
677 	if (err) {
678 		usnic_err("Failed to register netdev notifier\n");
679 		goto out_pci_unreg;
680 	}
681 
682 	err = register_inetaddr_notifier(&usnic_ib_inetaddr_notifier);
683 	if (err) {
684 		usnic_err("Failed to register inet addr notifier\n");
685 		goto out_unreg_netdev_notifier;
686 	}
687 
688 	err = usnic_transport_init();
689 	if (err) {
690 		usnic_err("Failed to initialize transport\n");
691 		goto out_unreg_inetaddr_notifier;
692 	}
693 
694 	usnic_debugfs_init();
695 
696 	return 0;
697 
698 out_unreg_inetaddr_notifier:
699 	unregister_inetaddr_notifier(&usnic_ib_inetaddr_notifier);
700 out_unreg_netdev_notifier:
701 	unregister_netdevice_notifier(&usnic_ib_netdevice_notifier);
702 out_pci_unreg:
703 	pci_unregister_driver(&usnic_ib_pci_driver);
704 out_umem_fini:
705 
706 	return err;
707 }
708 
usnic_ib_destroy(void)709 static void __exit usnic_ib_destroy(void)
710 {
711 	usnic_dbg("\n");
712 	usnic_debugfs_exit();
713 	usnic_transport_fini();
714 	unregister_inetaddr_notifier(&usnic_ib_inetaddr_notifier);
715 	unregister_netdevice_notifier(&usnic_ib_netdevice_notifier);
716 	pci_unregister_driver(&usnic_ib_pci_driver);
717 }
718 
719 MODULE_DESCRIPTION("Cisco VIC (usNIC) Verbs Driver");
720 MODULE_AUTHOR("Upinder Malhi <umalhi@cisco.com>");
721 MODULE_LICENSE("Dual BSD/GPL");
722 module_param(usnic_log_lvl, uint, S_IRUGO | S_IWUSR);
723 module_param(usnic_ib_share_vf, uint, S_IRUGO | S_IWUSR);
724 MODULE_PARM_DESC(usnic_log_lvl, " Off=0, Err=1, Info=2, Debug=3");
725 MODULE_PARM_DESC(usnic_ib_share_vf, "Off=0, On=1 VF sharing amongst QPs");
726 MODULE_DEVICE_TABLE(pci, usnic_ib_pci_ids);
727 
728 module_init(usnic_ib_init);
729 module_exit(usnic_ib_destroy);
730 /* End of module section */
731