xref: /linux/drivers/net/netdevsim/netdev.c (revision 7f71507851fc7764b36a3221839607d3a45c2025)
1 /*
2  * Copyright (C) 2017 Netronome Systems, Inc.
3  *
4  * This software is licensed under the GNU General License Version 2,
5  * June 1991 as shown in the file COPYING in the top-level directory of this
6  * source tree.
7  *
8  * THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS"
9  * WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING,
10  * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
11  * FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE
12  * OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME
13  * THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
14  */
15 
16 #include <linux/debugfs.h>
17 #include <linux/etherdevice.h>
18 #include <linux/kernel.h>
19 #include <linux/module.h>
20 #include <linux/netdevice.h>
21 #include <linux/slab.h>
22 #include <net/netdev_queues.h>
23 #include <net/page_pool/helpers.h>
24 #include <net/netlink.h>
25 #include <net/net_shaper.h>
26 #include <net/pkt_cls.h>
27 #include <net/rtnetlink.h>
28 #include <net/udp_tunnel.h>
29 
30 #include "netdevsim.h"
31 
32 #define NSIM_RING_SIZE		256
33 
34 static int nsim_napi_rx(struct nsim_rq *rq, struct sk_buff *skb)
35 {
36 	if (skb_queue_len(&rq->skb_queue) > NSIM_RING_SIZE) {
37 		dev_kfree_skb_any(skb);
38 		return NET_RX_DROP;
39 	}
40 
41 	skb_queue_tail(&rq->skb_queue, skb);
42 	return NET_RX_SUCCESS;
43 }
44 
45 static int nsim_forward_skb(struct net_device *dev, struct sk_buff *skb,
46 			    struct nsim_rq *rq)
47 {
48 	return __dev_forward_skb(dev, skb) ?: nsim_napi_rx(rq, skb);
49 }
50 
51 static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
52 {
53 	struct netdevsim *ns = netdev_priv(dev);
54 	struct net_device *peer_dev;
55 	unsigned int len = skb->len;
56 	struct netdevsim *peer_ns;
57 	struct nsim_rq *rq;
58 	int rxq;
59 
60 	rcu_read_lock();
61 	if (!nsim_ipsec_tx(ns, skb))
62 		goto out_drop_free;
63 
64 	peer_ns = rcu_dereference(ns->peer);
65 	if (!peer_ns)
66 		goto out_drop_free;
67 
68 	peer_dev = peer_ns->netdev;
69 	rxq = skb_get_queue_mapping(skb);
70 	if (rxq >= peer_dev->num_rx_queues)
71 		rxq = rxq % peer_dev->num_rx_queues;
72 	rq = &peer_ns->rq[rxq];
73 
74 	skb_tx_timestamp(skb);
75 	if (unlikely(nsim_forward_skb(peer_dev, skb, rq) == NET_RX_DROP))
76 		goto out_drop_cnt;
77 
78 	napi_schedule(&rq->napi);
79 
80 	rcu_read_unlock();
81 	u64_stats_update_begin(&ns->syncp);
82 	ns->tx_packets++;
83 	ns->tx_bytes += len;
84 	u64_stats_update_end(&ns->syncp);
85 	return NETDEV_TX_OK;
86 
87 out_drop_free:
88 	dev_kfree_skb(skb);
89 out_drop_cnt:
90 	rcu_read_unlock();
91 	u64_stats_update_begin(&ns->syncp);
92 	ns->tx_dropped++;
93 	u64_stats_update_end(&ns->syncp);
94 	return NETDEV_TX_OK;
95 }
96 
97 static void nsim_set_rx_mode(struct net_device *dev)
98 {
99 }
100 
101 static int nsim_change_mtu(struct net_device *dev, int new_mtu)
102 {
103 	struct netdevsim *ns = netdev_priv(dev);
104 
105 	if (ns->xdp.prog && new_mtu > NSIM_XDP_MAX_MTU)
106 		return -EBUSY;
107 
108 	WRITE_ONCE(dev->mtu, new_mtu);
109 
110 	return 0;
111 }
112 
113 static void
114 nsim_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
115 {
116 	struct netdevsim *ns = netdev_priv(dev);
117 	unsigned int start;
118 
119 	do {
120 		start = u64_stats_fetch_begin(&ns->syncp);
121 		stats->tx_bytes = ns->tx_bytes;
122 		stats->tx_packets = ns->tx_packets;
123 		stats->tx_dropped = ns->tx_dropped;
124 	} while (u64_stats_fetch_retry(&ns->syncp, start));
125 }
126 
127 static int
128 nsim_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
129 {
130 	return nsim_bpf_setup_tc_block_cb(type, type_data, cb_priv);
131 }
132 
133 static int nsim_set_vf_mac(struct net_device *dev, int vf, u8 *mac)
134 {
135 	struct netdevsim *ns = netdev_priv(dev);
136 	struct nsim_dev *nsim_dev = ns->nsim_dev;
137 
138 	/* Only refuse multicast addresses, zero address can mean unset/any. */
139 	if (vf >= nsim_dev_get_vfs(nsim_dev) || is_multicast_ether_addr(mac))
140 		return -EINVAL;
141 	memcpy(nsim_dev->vfconfigs[vf].vf_mac, mac, ETH_ALEN);
142 
143 	return 0;
144 }
145 
146 static int nsim_set_vf_vlan(struct net_device *dev, int vf,
147 			    u16 vlan, u8 qos, __be16 vlan_proto)
148 {
149 	struct netdevsim *ns = netdev_priv(dev);
150 	struct nsim_dev *nsim_dev = ns->nsim_dev;
151 
152 	if (vf >= nsim_dev_get_vfs(nsim_dev) || vlan > 4095 || qos > 7)
153 		return -EINVAL;
154 
155 	nsim_dev->vfconfigs[vf].vlan = vlan;
156 	nsim_dev->vfconfigs[vf].qos = qos;
157 	nsim_dev->vfconfigs[vf].vlan_proto = vlan_proto;
158 
159 	return 0;
160 }
161 
162 static int nsim_set_vf_rate(struct net_device *dev, int vf, int min, int max)
163 {
164 	struct netdevsim *ns = netdev_priv(dev);
165 	struct nsim_dev *nsim_dev = ns->nsim_dev;
166 
167 	if (nsim_esw_mode_is_switchdev(ns->nsim_dev)) {
168 		pr_err("Not supported in switchdev mode. Please use devlink API.\n");
169 		return -EOPNOTSUPP;
170 	}
171 
172 	if (vf >= nsim_dev_get_vfs(nsim_dev))
173 		return -EINVAL;
174 
175 	nsim_dev->vfconfigs[vf].min_tx_rate = min;
176 	nsim_dev->vfconfigs[vf].max_tx_rate = max;
177 
178 	return 0;
179 }
180 
181 static int nsim_set_vf_spoofchk(struct net_device *dev, int vf, bool val)
182 {
183 	struct netdevsim *ns = netdev_priv(dev);
184 	struct nsim_dev *nsim_dev = ns->nsim_dev;
185 
186 	if (vf >= nsim_dev_get_vfs(nsim_dev))
187 		return -EINVAL;
188 	nsim_dev->vfconfigs[vf].spoofchk_enabled = val;
189 
190 	return 0;
191 }
192 
193 static int nsim_set_vf_rss_query_en(struct net_device *dev, int vf, bool val)
194 {
195 	struct netdevsim *ns = netdev_priv(dev);
196 	struct nsim_dev *nsim_dev = ns->nsim_dev;
197 
198 	if (vf >= nsim_dev_get_vfs(nsim_dev))
199 		return -EINVAL;
200 	nsim_dev->vfconfigs[vf].rss_query_enabled = val;
201 
202 	return 0;
203 }
204 
205 static int nsim_set_vf_trust(struct net_device *dev, int vf, bool val)
206 {
207 	struct netdevsim *ns = netdev_priv(dev);
208 	struct nsim_dev *nsim_dev = ns->nsim_dev;
209 
210 	if (vf >= nsim_dev_get_vfs(nsim_dev))
211 		return -EINVAL;
212 	nsim_dev->vfconfigs[vf].trusted = val;
213 
214 	return 0;
215 }
216 
217 static int
218 nsim_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivi)
219 {
220 	struct netdevsim *ns = netdev_priv(dev);
221 	struct nsim_dev *nsim_dev = ns->nsim_dev;
222 
223 	if (vf >= nsim_dev_get_vfs(nsim_dev))
224 		return -EINVAL;
225 
226 	ivi->vf = vf;
227 	ivi->linkstate = nsim_dev->vfconfigs[vf].link_state;
228 	ivi->min_tx_rate = nsim_dev->vfconfigs[vf].min_tx_rate;
229 	ivi->max_tx_rate = nsim_dev->vfconfigs[vf].max_tx_rate;
230 	ivi->vlan = nsim_dev->vfconfigs[vf].vlan;
231 	ivi->vlan_proto = nsim_dev->vfconfigs[vf].vlan_proto;
232 	ivi->qos = nsim_dev->vfconfigs[vf].qos;
233 	memcpy(&ivi->mac, nsim_dev->vfconfigs[vf].vf_mac, ETH_ALEN);
234 	ivi->spoofchk = nsim_dev->vfconfigs[vf].spoofchk_enabled;
235 	ivi->trusted = nsim_dev->vfconfigs[vf].trusted;
236 	ivi->rss_query_en = nsim_dev->vfconfigs[vf].rss_query_enabled;
237 
238 	return 0;
239 }
240 
241 static int nsim_set_vf_link_state(struct net_device *dev, int vf, int state)
242 {
243 	struct netdevsim *ns = netdev_priv(dev);
244 	struct nsim_dev *nsim_dev = ns->nsim_dev;
245 
246 	if (vf >= nsim_dev_get_vfs(nsim_dev))
247 		return -EINVAL;
248 
249 	switch (state) {
250 	case IFLA_VF_LINK_STATE_AUTO:
251 	case IFLA_VF_LINK_STATE_ENABLE:
252 	case IFLA_VF_LINK_STATE_DISABLE:
253 		break;
254 	default:
255 		return -EINVAL;
256 	}
257 
258 	nsim_dev->vfconfigs[vf].link_state = state;
259 
260 	return 0;
261 }
262 
263 static void nsim_taprio_stats(struct tc_taprio_qopt_stats *stats)
264 {
265 	stats->window_drops = 0;
266 	stats->tx_overruns = 0;
267 }
268 
269 static int nsim_setup_tc_taprio(struct net_device *dev,
270 				struct tc_taprio_qopt_offload *offload)
271 {
272 	int err = 0;
273 
274 	switch (offload->cmd) {
275 	case TAPRIO_CMD_REPLACE:
276 	case TAPRIO_CMD_DESTROY:
277 		break;
278 	case TAPRIO_CMD_STATS:
279 		nsim_taprio_stats(&offload->stats);
280 		break;
281 	default:
282 		err = -EOPNOTSUPP;
283 	}
284 
285 	return err;
286 }
287 
288 static LIST_HEAD(nsim_block_cb_list);
289 
290 static int
291 nsim_setup_tc(struct net_device *dev, enum tc_setup_type type, void *type_data)
292 {
293 	struct netdevsim *ns = netdev_priv(dev);
294 
295 	switch (type) {
296 	case TC_SETUP_QDISC_TAPRIO:
297 		return nsim_setup_tc_taprio(dev, type_data);
298 	case TC_SETUP_BLOCK:
299 		return flow_block_cb_setup_simple(type_data,
300 						  &nsim_block_cb_list,
301 						  nsim_setup_tc_block_cb,
302 						  ns, ns, true);
303 	default:
304 		return -EOPNOTSUPP;
305 	}
306 }
307 
308 static int
309 nsim_set_features(struct net_device *dev, netdev_features_t features)
310 {
311 	struct netdevsim *ns = netdev_priv(dev);
312 
313 	if ((dev->features & NETIF_F_HW_TC) > (features & NETIF_F_HW_TC))
314 		return nsim_bpf_disable_tc(ns);
315 
316 	return 0;
317 }
318 
319 static int nsim_get_iflink(const struct net_device *dev)
320 {
321 	struct netdevsim *nsim, *peer;
322 	int iflink;
323 
324 	nsim = netdev_priv(dev);
325 
326 	rcu_read_lock();
327 	peer = rcu_dereference(nsim->peer);
328 	iflink = peer ? READ_ONCE(peer->netdev->ifindex) :
329 			READ_ONCE(dev->ifindex);
330 	rcu_read_unlock();
331 
332 	return iflink;
333 }
334 
335 static int nsim_rcv(struct nsim_rq *rq, int budget)
336 {
337 	struct sk_buff *skb;
338 	int i;
339 
340 	for (i = 0; i < budget; i++) {
341 		if (skb_queue_empty(&rq->skb_queue))
342 			break;
343 
344 		skb = skb_dequeue(&rq->skb_queue);
345 		netif_receive_skb(skb);
346 	}
347 
348 	return i;
349 }
350 
351 static int nsim_poll(struct napi_struct *napi, int budget)
352 {
353 	struct nsim_rq *rq = container_of(napi, struct nsim_rq, napi);
354 	int done;
355 
356 	done = nsim_rcv(rq, budget);
357 	napi_complete(napi);
358 
359 	return done;
360 }
361 
362 static int nsim_create_page_pool(struct nsim_rq *rq)
363 {
364 	struct page_pool_params p = {
365 		.order = 0,
366 		.pool_size = NSIM_RING_SIZE,
367 		.nid = NUMA_NO_NODE,
368 		.dev = &rq->napi.dev->dev,
369 		.napi = &rq->napi,
370 		.dma_dir = DMA_BIDIRECTIONAL,
371 		.netdev = rq->napi.dev,
372 	};
373 
374 	rq->page_pool = page_pool_create(&p);
375 	if (IS_ERR(rq->page_pool)) {
376 		int err = PTR_ERR(rq->page_pool);
377 
378 		rq->page_pool = NULL;
379 		return err;
380 	}
381 	return 0;
382 }
383 
384 static int nsim_init_napi(struct netdevsim *ns)
385 {
386 	struct net_device *dev = ns->netdev;
387 	struct nsim_rq *rq;
388 	int err, i;
389 
390 	for (i = 0; i < dev->num_rx_queues; i++) {
391 		rq = &ns->rq[i];
392 
393 		netif_napi_add(dev, &rq->napi, nsim_poll);
394 	}
395 
396 	for (i = 0; i < dev->num_rx_queues; i++) {
397 		rq = &ns->rq[i];
398 
399 		err = nsim_create_page_pool(rq);
400 		if (err)
401 			goto err_pp_destroy;
402 	}
403 
404 	return 0;
405 
406 err_pp_destroy:
407 	while (i--) {
408 		page_pool_destroy(ns->rq[i].page_pool);
409 		ns->rq[i].page_pool = NULL;
410 	}
411 
412 	for (i = 0; i < dev->num_rx_queues; i++)
413 		__netif_napi_del(&ns->rq[i].napi);
414 
415 	return err;
416 }
417 
418 static void nsim_enable_napi(struct netdevsim *ns)
419 {
420 	struct net_device *dev = ns->netdev;
421 	int i;
422 
423 	for (i = 0; i < dev->num_rx_queues; i++) {
424 		struct nsim_rq *rq = &ns->rq[i];
425 
426 		netif_queue_set_napi(dev, i, NETDEV_QUEUE_TYPE_RX, &rq->napi);
427 		napi_enable(&rq->napi);
428 	}
429 }
430 
431 static int nsim_open(struct net_device *dev)
432 {
433 	struct netdevsim *ns = netdev_priv(dev);
434 	int err;
435 
436 	err = nsim_init_napi(ns);
437 	if (err)
438 		return err;
439 
440 	nsim_enable_napi(ns);
441 
442 	return 0;
443 }
444 
445 static void nsim_del_napi(struct netdevsim *ns)
446 {
447 	struct net_device *dev = ns->netdev;
448 	int i;
449 
450 	for (i = 0; i < dev->num_rx_queues; i++) {
451 		struct nsim_rq *rq = &ns->rq[i];
452 
453 		napi_disable(&rq->napi);
454 		__netif_napi_del(&rq->napi);
455 	}
456 	synchronize_net();
457 
458 	for (i = 0; i < dev->num_rx_queues; i++) {
459 		page_pool_destroy(ns->rq[i].page_pool);
460 		ns->rq[i].page_pool = NULL;
461 	}
462 }
463 
464 static int nsim_stop(struct net_device *dev)
465 {
466 	struct netdevsim *ns = netdev_priv(dev);
467 	struct netdevsim *peer;
468 
469 	netif_carrier_off(dev);
470 	peer = rtnl_dereference(ns->peer);
471 	if (peer)
472 		netif_carrier_off(peer->netdev);
473 
474 	nsim_del_napi(ns);
475 
476 	return 0;
477 }
478 
479 static int nsim_shaper_set(struct net_shaper_binding *binding,
480 			   const struct net_shaper *shaper,
481 			   struct netlink_ext_ack *extack)
482 {
483 	return 0;
484 }
485 
486 static int nsim_shaper_del(struct net_shaper_binding *binding,
487 			   const struct net_shaper_handle *handle,
488 			   struct netlink_ext_ack *extack)
489 {
490 	return 0;
491 }
492 
493 static int nsim_shaper_group(struct net_shaper_binding *binding,
494 			     int leaves_count,
495 			     const struct net_shaper *leaves,
496 			     const struct net_shaper *root,
497 			     struct netlink_ext_ack *extack)
498 {
499 	return 0;
500 }
501 
502 static void nsim_shaper_cap(struct net_shaper_binding *binding,
503 			    enum net_shaper_scope scope,
504 			    unsigned long *flags)
505 {
506 	*flags = ULONG_MAX;
507 }
508 
509 static const struct net_shaper_ops nsim_shaper_ops = {
510 	.set			= nsim_shaper_set,
511 	.delete			= nsim_shaper_del,
512 	.group			= nsim_shaper_group,
513 	.capabilities		= nsim_shaper_cap,
514 };
515 
516 static const struct net_device_ops nsim_netdev_ops = {
517 	.ndo_start_xmit		= nsim_start_xmit,
518 	.ndo_set_rx_mode	= nsim_set_rx_mode,
519 	.ndo_set_mac_address	= eth_mac_addr,
520 	.ndo_validate_addr	= eth_validate_addr,
521 	.ndo_change_mtu		= nsim_change_mtu,
522 	.ndo_get_stats64	= nsim_get_stats64,
523 	.ndo_set_vf_mac		= nsim_set_vf_mac,
524 	.ndo_set_vf_vlan	= nsim_set_vf_vlan,
525 	.ndo_set_vf_rate	= nsim_set_vf_rate,
526 	.ndo_set_vf_spoofchk	= nsim_set_vf_spoofchk,
527 	.ndo_set_vf_trust	= nsim_set_vf_trust,
528 	.ndo_get_vf_config	= nsim_get_vf_config,
529 	.ndo_set_vf_link_state	= nsim_set_vf_link_state,
530 	.ndo_set_vf_rss_query_en = nsim_set_vf_rss_query_en,
531 	.ndo_setup_tc		= nsim_setup_tc,
532 	.ndo_set_features	= nsim_set_features,
533 	.ndo_get_iflink		= nsim_get_iflink,
534 	.ndo_bpf		= nsim_bpf,
535 	.ndo_open		= nsim_open,
536 	.ndo_stop		= nsim_stop,
537 	.net_shaper_ops		= &nsim_shaper_ops,
538 };
539 
540 static const struct net_device_ops nsim_vf_netdev_ops = {
541 	.ndo_start_xmit		= nsim_start_xmit,
542 	.ndo_set_rx_mode	= nsim_set_rx_mode,
543 	.ndo_set_mac_address	= eth_mac_addr,
544 	.ndo_validate_addr	= eth_validate_addr,
545 	.ndo_change_mtu		= nsim_change_mtu,
546 	.ndo_get_stats64	= nsim_get_stats64,
547 	.ndo_setup_tc		= nsim_setup_tc,
548 	.ndo_set_features	= nsim_set_features,
549 };
550 
551 /* We don't have true per-queue stats, yet, so do some random fakery here.
552  * Only report stuff for queue 0.
553  */
554 static void nsim_get_queue_stats_rx(struct net_device *dev, int idx,
555 				    struct netdev_queue_stats_rx *stats)
556 {
557 	struct rtnl_link_stats64 rtstats = {};
558 
559 	if (!idx)
560 		nsim_get_stats64(dev, &rtstats);
561 
562 	stats->packets = rtstats.rx_packets - !!rtstats.rx_packets;
563 	stats->bytes = rtstats.rx_bytes;
564 }
565 
566 static void nsim_get_queue_stats_tx(struct net_device *dev, int idx,
567 				    struct netdev_queue_stats_tx *stats)
568 {
569 	struct rtnl_link_stats64 rtstats = {};
570 
571 	if (!idx)
572 		nsim_get_stats64(dev, &rtstats);
573 
574 	stats->packets = rtstats.tx_packets - !!rtstats.tx_packets;
575 	stats->bytes = rtstats.tx_bytes;
576 }
577 
578 static void nsim_get_base_stats(struct net_device *dev,
579 				struct netdev_queue_stats_rx *rx,
580 				struct netdev_queue_stats_tx *tx)
581 {
582 	struct rtnl_link_stats64 rtstats = {};
583 
584 	nsim_get_stats64(dev, &rtstats);
585 
586 	rx->packets = !!rtstats.rx_packets;
587 	rx->bytes = 0;
588 	tx->packets = !!rtstats.tx_packets;
589 	tx->bytes = 0;
590 }
591 
592 static const struct netdev_stat_ops nsim_stat_ops = {
593 	.get_queue_stats_tx	= nsim_get_queue_stats_tx,
594 	.get_queue_stats_rx	= nsim_get_queue_stats_rx,
595 	.get_base_stats		= nsim_get_base_stats,
596 };
597 
598 static ssize_t
599 nsim_pp_hold_read(struct file *file, char __user *data,
600 		  size_t count, loff_t *ppos)
601 {
602 	struct netdevsim *ns = file->private_data;
603 	char buf[3] = "n\n";
604 
605 	if (ns->page)
606 		buf[0] = 'y';
607 
608 	return simple_read_from_buffer(data, count, ppos, buf, 2);
609 }
610 
611 static ssize_t
612 nsim_pp_hold_write(struct file *file, const char __user *data,
613 		   size_t count, loff_t *ppos)
614 {
615 	struct netdevsim *ns = file->private_data;
616 	ssize_t ret;
617 	bool val;
618 
619 	ret = kstrtobool_from_user(data, count, &val);
620 	if (ret)
621 		return ret;
622 
623 	rtnl_lock();
624 	ret = count;
625 	if (val == !!ns->page)
626 		goto exit;
627 
628 	if (!netif_running(ns->netdev) && val) {
629 		ret = -ENETDOWN;
630 	} else if (val) {
631 		ns->page = page_pool_dev_alloc_pages(ns->rq[0].page_pool);
632 		if (!ns->page)
633 			ret = -ENOMEM;
634 	} else {
635 		page_pool_put_full_page(ns->page->pp, ns->page, false);
636 		ns->page = NULL;
637 	}
638 	rtnl_unlock();
639 
640 exit:
641 	return count;
642 }
643 
644 static const struct file_operations nsim_pp_hold_fops = {
645 	.open = simple_open,
646 	.read = nsim_pp_hold_read,
647 	.write = nsim_pp_hold_write,
648 	.llseek = generic_file_llseek,
649 	.owner = THIS_MODULE,
650 };
651 
652 static void nsim_setup(struct net_device *dev)
653 {
654 	ether_setup(dev);
655 	eth_hw_addr_random(dev);
656 
657 	dev->tx_queue_len = 0;
658 	dev->flags &= ~IFF_MULTICAST;
659 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE |
660 			   IFF_NO_QUEUE;
661 	dev->features |= NETIF_F_HIGHDMA |
662 			 NETIF_F_SG |
663 			 NETIF_F_FRAGLIST |
664 			 NETIF_F_HW_CSUM |
665 			 NETIF_F_TSO;
666 	dev->hw_features |= NETIF_F_HW_TC |
667 			    NETIF_F_SG |
668 			    NETIF_F_FRAGLIST |
669 			    NETIF_F_HW_CSUM |
670 			    NETIF_F_TSO;
671 	dev->max_mtu = ETH_MAX_MTU;
672 	dev->xdp_features = NETDEV_XDP_ACT_HW_OFFLOAD;
673 }
674 
675 static int nsim_queue_init(struct netdevsim *ns)
676 {
677 	struct net_device *dev = ns->netdev;
678 	int i;
679 
680 	ns->rq = kvcalloc(dev->num_rx_queues, sizeof(*ns->rq),
681 			  GFP_KERNEL_ACCOUNT | __GFP_RETRY_MAYFAIL);
682 	if (!ns->rq)
683 		return -ENOMEM;
684 
685 	for (i = 0; i < dev->num_rx_queues; i++)
686 		skb_queue_head_init(&ns->rq[i].skb_queue);
687 
688 	return 0;
689 }
690 
691 static void nsim_queue_free(struct netdevsim *ns)
692 {
693 	struct net_device *dev = ns->netdev;
694 	int i;
695 
696 	for (i = 0; i < dev->num_rx_queues; i++)
697 		skb_queue_purge_reason(&ns->rq[i].skb_queue,
698 				       SKB_DROP_REASON_QUEUE_PURGE);
699 
700 	kvfree(ns->rq);
701 	ns->rq = NULL;
702 }
703 
704 static int nsim_init_netdevsim(struct netdevsim *ns)
705 {
706 	struct mock_phc *phc;
707 	int err;
708 
709 	phc = mock_phc_create(&ns->nsim_bus_dev->dev);
710 	if (IS_ERR(phc))
711 		return PTR_ERR(phc);
712 
713 	ns->phc = phc;
714 	ns->netdev->netdev_ops = &nsim_netdev_ops;
715 	ns->netdev->stat_ops = &nsim_stat_ops;
716 
717 	err = nsim_udp_tunnels_info_create(ns->nsim_dev, ns->netdev);
718 	if (err)
719 		goto err_phc_destroy;
720 
721 	rtnl_lock();
722 	err = nsim_queue_init(ns);
723 	if (err)
724 		goto err_utn_destroy;
725 
726 	err = nsim_bpf_init(ns);
727 	if (err)
728 		goto err_rq_destroy;
729 
730 	nsim_macsec_init(ns);
731 	nsim_ipsec_init(ns);
732 
733 	err = register_netdevice(ns->netdev);
734 	if (err)
735 		goto err_ipsec_teardown;
736 	rtnl_unlock();
737 	return 0;
738 
739 err_ipsec_teardown:
740 	nsim_ipsec_teardown(ns);
741 	nsim_macsec_teardown(ns);
742 	nsim_bpf_uninit(ns);
743 err_rq_destroy:
744 	nsim_queue_free(ns);
745 err_utn_destroy:
746 	rtnl_unlock();
747 	nsim_udp_tunnels_info_destroy(ns->netdev);
748 err_phc_destroy:
749 	mock_phc_destroy(ns->phc);
750 	return err;
751 }
752 
753 static int nsim_init_netdevsim_vf(struct netdevsim *ns)
754 {
755 	int err;
756 
757 	ns->netdev->netdev_ops = &nsim_vf_netdev_ops;
758 	rtnl_lock();
759 	err = register_netdevice(ns->netdev);
760 	rtnl_unlock();
761 	return err;
762 }
763 
764 static void nsim_exit_netdevsim(struct netdevsim *ns)
765 {
766 	nsim_udp_tunnels_info_destroy(ns->netdev);
767 	mock_phc_destroy(ns->phc);
768 }
769 
770 struct netdevsim *
771 nsim_create(struct nsim_dev *nsim_dev, struct nsim_dev_port *nsim_dev_port)
772 {
773 	struct net_device *dev;
774 	struct netdevsim *ns;
775 	int err;
776 
777 	dev = alloc_netdev_mq(sizeof(*ns), "eth%d", NET_NAME_UNKNOWN, nsim_setup,
778 			      nsim_dev->nsim_bus_dev->num_queues);
779 	if (!dev)
780 		return ERR_PTR(-ENOMEM);
781 
782 	dev_net_set(dev, nsim_dev_net(nsim_dev));
783 	ns = netdev_priv(dev);
784 	ns->netdev = dev;
785 	u64_stats_init(&ns->syncp);
786 	ns->nsim_dev = nsim_dev;
787 	ns->nsim_dev_port = nsim_dev_port;
788 	ns->nsim_bus_dev = nsim_dev->nsim_bus_dev;
789 	SET_NETDEV_DEV(dev, &ns->nsim_bus_dev->dev);
790 	SET_NETDEV_DEVLINK_PORT(dev, &nsim_dev_port->devlink_port);
791 	nsim_ethtool_init(ns);
792 	if (nsim_dev_port_is_pf(nsim_dev_port))
793 		err = nsim_init_netdevsim(ns);
794 	else
795 		err = nsim_init_netdevsim_vf(ns);
796 	if (err)
797 		goto err_free_netdev;
798 
799 	ns->pp_dfs = debugfs_create_file("pp_hold", 0600, nsim_dev_port->ddir,
800 					 ns, &nsim_pp_hold_fops);
801 
802 	return ns;
803 
804 err_free_netdev:
805 	free_netdev(dev);
806 	return ERR_PTR(err);
807 }
808 
809 void nsim_destroy(struct netdevsim *ns)
810 {
811 	struct net_device *dev = ns->netdev;
812 	struct netdevsim *peer;
813 
814 	debugfs_remove(ns->pp_dfs);
815 
816 	rtnl_lock();
817 	peer = rtnl_dereference(ns->peer);
818 	if (peer)
819 		RCU_INIT_POINTER(peer->peer, NULL);
820 	RCU_INIT_POINTER(ns->peer, NULL);
821 	unregister_netdevice(dev);
822 	if (nsim_dev_port_is_pf(ns->nsim_dev_port)) {
823 		nsim_macsec_teardown(ns);
824 		nsim_ipsec_teardown(ns);
825 		nsim_bpf_uninit(ns);
826 		nsim_queue_free(ns);
827 	}
828 	rtnl_unlock();
829 	if (nsim_dev_port_is_pf(ns->nsim_dev_port))
830 		nsim_exit_netdevsim(ns);
831 
832 	/* Put this intentionally late to exercise the orphaning path */
833 	if (ns->page) {
834 		page_pool_put_full_page(ns->page->pp, ns->page, false);
835 		ns->page = NULL;
836 	}
837 
838 	free_netdev(dev);
839 }
840 
841 bool netdev_is_nsim(struct net_device *dev)
842 {
843 	return dev->netdev_ops == &nsim_netdev_ops;
844 }
845 
846 static int nsim_validate(struct nlattr *tb[], struct nlattr *data[],
847 			 struct netlink_ext_ack *extack)
848 {
849 	NL_SET_ERR_MSG_MOD(extack,
850 			   "Please use: echo \"[ID] [PORT_COUNT] [NUM_QUEUES]\" > /sys/bus/netdevsim/new_device");
851 	return -EOPNOTSUPP;
852 }
853 
854 static struct rtnl_link_ops nsim_link_ops __read_mostly = {
855 	.kind		= DRV_NAME,
856 	.validate	= nsim_validate,
857 };
858 
859 static int __init nsim_module_init(void)
860 {
861 	int err;
862 
863 	err = nsim_dev_init();
864 	if (err)
865 		return err;
866 
867 	err = nsim_bus_init();
868 	if (err)
869 		goto err_dev_exit;
870 
871 	err = rtnl_link_register(&nsim_link_ops);
872 	if (err)
873 		goto err_bus_exit;
874 
875 	return 0;
876 
877 err_bus_exit:
878 	nsim_bus_exit();
879 err_dev_exit:
880 	nsim_dev_exit();
881 	return err;
882 }
883 
884 static void __exit nsim_module_exit(void)
885 {
886 	rtnl_link_unregister(&nsim_link_ops);
887 	nsim_bus_exit();
888 	nsim_dev_exit();
889 }
890 
891 module_init(nsim_module_init);
892 module_exit(nsim_module_exit);
893 MODULE_LICENSE("GPL");
894 MODULE_DESCRIPTION("Simulated networking device for testing");
895 MODULE_ALIAS_RTNL_LINK(DRV_NAME);
896