Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/* Copyright (c) 2023 Isovalent */
3
4#include <linux/netdevice.h>
5#include <linux/ethtool.h>
6#include <linux/etherdevice.h>
7#include <linux/filter.h>
8#include <linux/netfilter_netdev.h>
9#include <linux/bpf_mprog.h>
10#include <linux/indirect_call_wrapper.h>
11
12#include <net/netkit.h>
13#include <net/dst.h>
14#include <net/tcx.h>
15
16#define DRV_NAME "netkit"
17
18struct netkit {
19 /* Needed in fast-path */
20 struct net_device __rcu *peer;
21 struct bpf_mprog_entry __rcu *active;
22 enum netkit_action policy;
23 struct bpf_mprog_bundle bundle;
24
25 /* Needed in slow-path */
26 enum netkit_mode mode;
27 bool primary;
28 u32 headroom;
29};
30
31struct netkit_link {
32 struct bpf_link link;
33 struct net_device *dev;
34 u32 location;
35};
36
37static __always_inline int
38netkit_run(const struct bpf_mprog_entry *entry, struct sk_buff *skb,
39 enum netkit_action ret)
40{
41 const struct bpf_mprog_fp *fp;
42 const struct bpf_prog *prog;
43
44 bpf_mprog_foreach_prog(entry, fp, prog) {
45 bpf_compute_data_pointers(skb);
46 ret = bpf_prog_run(prog, skb);
47 if (ret != NETKIT_NEXT)
48 break;
49 }
50 return ret;
51}
52
53static void netkit_prep_forward(struct sk_buff *skb, bool xnet)
54{
55 skb_scrub_packet(skb, xnet);
56 skb->priority = 0;
57 nf_skip_egress(skb, true);
58}
59
60static struct netkit *netkit_priv(const struct net_device *dev)
61{
62 return netdev_priv(dev);
63}
64
65static netdev_tx_t netkit_xmit(struct sk_buff *skb, struct net_device *dev)
66{
67 struct netkit *nk = netkit_priv(dev);
68 enum netkit_action ret = READ_ONCE(nk->policy);
69 netdev_tx_t ret_dev = NET_XMIT_SUCCESS;
70 const struct bpf_mprog_entry *entry;
71 struct net_device *peer;
72 int len = skb->len;
73
74 rcu_read_lock();
75 peer = rcu_dereference(nk->peer);
76 if (unlikely(!peer || !(peer->flags & IFF_UP) ||
77 !pskb_may_pull(skb, ETH_HLEN) ||
78 skb_orphan_frags(skb, GFP_ATOMIC)))
79 goto drop;
80 netkit_prep_forward(skb, !net_eq(dev_net(dev), dev_net(peer)));
81 skb->dev = peer;
82 entry = rcu_dereference(nk->active);
83 if (entry)
84 ret = netkit_run(entry, skb, ret);
85 switch (ret) {
86 case NETKIT_NEXT:
87 case NETKIT_PASS:
88 skb->protocol = eth_type_trans(skb, skb->dev);
89 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
90 if (likely(__netif_rx(skb) == NET_RX_SUCCESS)) {
91 dev_sw_netstats_tx_add(dev, 1, len);
92 dev_sw_netstats_rx_add(peer, len);
93 } else {
94 goto drop_stats;
95 }
96 break;
97 case NETKIT_REDIRECT:
98 dev_sw_netstats_tx_add(dev, 1, len);
99 skb_do_redirect(skb);
100 break;
101 case NETKIT_DROP:
102 default:
103drop:
104 kfree_skb(skb);
105drop_stats:
106 dev_core_stats_tx_dropped_inc(dev);
107 ret_dev = NET_XMIT_DROP;
108 break;
109 }
110 rcu_read_unlock();
111 return ret_dev;
112}
113
114static int netkit_open(struct net_device *dev)
115{
116 struct netkit *nk = netkit_priv(dev);
117 struct net_device *peer = rtnl_dereference(nk->peer);
118
119 if (!peer)
120 return -ENOTCONN;
121 if (peer->flags & IFF_UP) {
122 netif_carrier_on(dev);
123 netif_carrier_on(peer);
124 }
125 return 0;
126}
127
128static int netkit_close(struct net_device *dev)
129{
130 struct netkit *nk = netkit_priv(dev);
131 struct net_device *peer = rtnl_dereference(nk->peer);
132
133 netif_carrier_off(dev);
134 if (peer)
135 netif_carrier_off(peer);
136 return 0;
137}
138
139static int netkit_get_iflink(const struct net_device *dev)
140{
141 struct netkit *nk = netkit_priv(dev);
142 struct net_device *peer;
143 int iflink = 0;
144
145 rcu_read_lock();
146 peer = rcu_dereference(nk->peer);
147 if (peer)
148 iflink = peer->ifindex;
149 rcu_read_unlock();
150 return iflink;
151}
152
153static void netkit_set_multicast(struct net_device *dev)
154{
155 /* Nothing to do, we receive whatever gets pushed to us! */
156}
157
158static void netkit_set_headroom(struct net_device *dev, int headroom)
159{
160 struct netkit *nk = netkit_priv(dev), *nk2;
161 struct net_device *peer;
162
163 if (headroom < 0)
164 headroom = NET_SKB_PAD;
165
166 rcu_read_lock();
167 peer = rcu_dereference(nk->peer);
168 if (unlikely(!peer))
169 goto out;
170
171 nk2 = netkit_priv(peer);
172 nk->headroom = headroom;
173 headroom = max(nk->headroom, nk2->headroom);
174
175 peer->needed_headroom = headroom;
176 dev->needed_headroom = headroom;
177out:
178 rcu_read_unlock();
179}
180
181INDIRECT_CALLABLE_SCOPE struct net_device *netkit_peer_dev(struct net_device *dev)
182{
183 return rcu_dereference(netkit_priv(dev)->peer);
184}
185
186static void netkit_get_stats(struct net_device *dev,
187 struct rtnl_link_stats64 *stats)
188{
189 dev_fetch_sw_netstats(stats, dev->tstats);
190 stats->tx_dropped = DEV_STATS_READ(dev, tx_dropped);
191}
192
193static void netkit_uninit(struct net_device *dev);
194
195static const struct net_device_ops netkit_netdev_ops = {
196 .ndo_open = netkit_open,
197 .ndo_stop = netkit_close,
198 .ndo_start_xmit = netkit_xmit,
199 .ndo_set_rx_mode = netkit_set_multicast,
200 .ndo_set_rx_headroom = netkit_set_headroom,
201 .ndo_get_iflink = netkit_get_iflink,
202 .ndo_get_peer_dev = netkit_peer_dev,
203 .ndo_get_stats64 = netkit_get_stats,
204 .ndo_uninit = netkit_uninit,
205 .ndo_features_check = passthru_features_check,
206};
207
208static void netkit_get_drvinfo(struct net_device *dev,
209 struct ethtool_drvinfo *info)
210{
211 strscpy(info->driver, DRV_NAME, sizeof(info->driver));
212}
213
214static const struct ethtool_ops netkit_ethtool_ops = {
215 .get_drvinfo = netkit_get_drvinfo,
216};
217
218static void netkit_setup(struct net_device *dev)
219{
220 static const netdev_features_t netkit_features_hw_vlan =
221 NETIF_F_HW_VLAN_CTAG_TX |
222 NETIF_F_HW_VLAN_CTAG_RX |
223 NETIF_F_HW_VLAN_STAG_TX |
224 NETIF_F_HW_VLAN_STAG_RX;
225 static const netdev_features_t netkit_features =
226 netkit_features_hw_vlan |
227 NETIF_F_SG |
228 NETIF_F_FRAGLIST |
229 NETIF_F_HW_CSUM |
230 NETIF_F_RXCSUM |
231 NETIF_F_SCTP_CRC |
232 NETIF_F_HIGHDMA |
233 NETIF_F_GSO_SOFTWARE |
234 NETIF_F_GSO_ENCAP_ALL;
235
236 ether_setup(dev);
237 dev->max_mtu = ETH_MAX_MTU;
238 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
239
240 dev->flags |= IFF_NOARP;
241 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
242 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
243 dev->priv_flags |= IFF_PHONY_HEADROOM;
244 dev->priv_flags |= IFF_NO_QUEUE;
245
246 dev->ethtool_ops = &netkit_ethtool_ops;
247 dev->netdev_ops = &netkit_netdev_ops;
248
249 dev->features |= netkit_features | NETIF_F_LLTX;
250 dev->hw_features = netkit_features;
251 dev->hw_enc_features = netkit_features;
252 dev->mpls_features = NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
253 dev->vlan_features = dev->features & ~netkit_features_hw_vlan;
254
255 dev->needs_free_netdev = true;
256
257 netif_set_tso_max_size(dev, GSO_MAX_SIZE);
258}
259
260static struct net *netkit_get_link_net(const struct net_device *dev)
261{
262 struct netkit *nk = netkit_priv(dev);
263 struct net_device *peer = rtnl_dereference(nk->peer);
264
265 return peer ? dev_net(peer) : dev_net(dev);
266}
267
268static int netkit_check_policy(int policy, struct nlattr *tb,
269 struct netlink_ext_ack *extack)
270{
271 switch (policy) {
272 case NETKIT_PASS:
273 case NETKIT_DROP:
274 return 0;
275 default:
276 NL_SET_ERR_MSG_ATTR(extack, tb,
277 "Provided default xmit policy not supported");
278 return -EINVAL;
279 }
280}
281
282static int netkit_check_mode(int mode, struct nlattr *tb,
283 struct netlink_ext_ack *extack)
284{
285 switch (mode) {
286 case NETKIT_L2:
287 case NETKIT_L3:
288 return 0;
289 default:
290 NL_SET_ERR_MSG_ATTR(extack, tb,
291 "Provided device mode can only be L2 or L3");
292 return -EINVAL;
293 }
294}
295
296static int netkit_validate(struct nlattr *tb[], struct nlattr *data[],
297 struct netlink_ext_ack *extack)
298{
299 struct nlattr *attr = tb[IFLA_ADDRESS];
300
301 if (!attr)
302 return 0;
303 NL_SET_ERR_MSG_ATTR(extack, attr,
304 "Setting Ethernet address is not supported");
305 return -EOPNOTSUPP;
306}
307
308static struct rtnl_link_ops netkit_link_ops;
309
310static int netkit_new_link(struct net *src_net, struct net_device *dev,
311 struct nlattr *tb[], struct nlattr *data[],
312 struct netlink_ext_ack *extack)
313{
314 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp = tb, *attr;
315 enum netkit_action default_prim = NETKIT_PASS;
316 enum netkit_action default_peer = NETKIT_PASS;
317 enum netkit_mode mode = NETKIT_L3;
318 unsigned char ifname_assign_type;
319 struct ifinfomsg *ifmp = NULL;
320 struct net_device *peer;
321 char ifname[IFNAMSIZ];
322 struct netkit *nk;
323 struct net *net;
324 int err;
325
326 if (data) {
327 if (data[IFLA_NETKIT_MODE]) {
328 attr = data[IFLA_NETKIT_MODE];
329 mode = nla_get_u32(attr);
330 err = netkit_check_mode(mode, attr, extack);
331 if (err < 0)
332 return err;
333 }
334 if (data[IFLA_NETKIT_PEER_INFO]) {
335 attr = data[IFLA_NETKIT_PEER_INFO];
336 ifmp = nla_data(attr);
337 err = rtnl_nla_parse_ifinfomsg(peer_tb, attr, extack);
338 if (err < 0)
339 return err;
340 err = netkit_validate(peer_tb, NULL, extack);
341 if (err < 0)
342 return err;
343 tbp = peer_tb;
344 }
345 if (data[IFLA_NETKIT_POLICY]) {
346 attr = data[IFLA_NETKIT_POLICY];
347 default_prim = nla_get_u32(attr);
348 err = netkit_check_policy(default_prim, attr, extack);
349 if (err < 0)
350 return err;
351 }
352 if (data[IFLA_NETKIT_PEER_POLICY]) {
353 attr = data[IFLA_NETKIT_PEER_POLICY];
354 default_peer = nla_get_u32(attr);
355 err = netkit_check_policy(default_peer, attr, extack);
356 if (err < 0)
357 return err;
358 }
359 }
360
361 if (ifmp && tbp[IFLA_IFNAME]) {
362 nla_strscpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
363 ifname_assign_type = NET_NAME_USER;
364 } else {
365 strscpy(ifname, "nk%d", IFNAMSIZ);
366 ifname_assign_type = NET_NAME_ENUM;
367 }
368
369 net = rtnl_link_get_net(src_net, tbp);
370 if (IS_ERR(net))
371 return PTR_ERR(net);
372
373 peer = rtnl_create_link(net, ifname, ifname_assign_type,
374 &netkit_link_ops, tbp, extack);
375 if (IS_ERR(peer)) {
376 put_net(net);
377 return PTR_ERR(peer);
378 }
379
380 netif_inherit_tso_max(peer, dev);
381
382 if (mode == NETKIT_L2)
383 eth_hw_addr_random(peer);
384 if (ifmp && dev->ifindex)
385 peer->ifindex = ifmp->ifi_index;
386
387 nk = netkit_priv(peer);
388 nk->primary = false;
389 nk->policy = default_peer;
390 nk->mode = mode;
391 bpf_mprog_bundle_init(&nk->bundle);
392
393 err = register_netdevice(peer);
394 put_net(net);
395 if (err < 0)
396 goto err_register_peer;
397 netif_carrier_off(peer);
398 if (mode == NETKIT_L2)
399 dev_change_flags(peer, peer->flags & ~IFF_NOARP, NULL);
400
401 err = rtnl_configure_link(peer, NULL, 0, NULL);
402 if (err < 0)
403 goto err_configure_peer;
404
405 if (mode == NETKIT_L2)
406 eth_hw_addr_random(dev);
407 if (tb[IFLA_IFNAME])
408 nla_strscpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
409 else
410 strscpy(dev->name, "nk%d", IFNAMSIZ);
411
412 nk = netkit_priv(dev);
413 nk->primary = true;
414 nk->policy = default_prim;
415 nk->mode = mode;
416 bpf_mprog_bundle_init(&nk->bundle);
417
418 err = register_netdevice(dev);
419 if (err < 0)
420 goto err_configure_peer;
421 netif_carrier_off(dev);
422 if (mode == NETKIT_L2)
423 dev_change_flags(dev, dev->flags & ~IFF_NOARP, NULL);
424
425 rcu_assign_pointer(netkit_priv(dev)->peer, peer);
426 rcu_assign_pointer(netkit_priv(peer)->peer, dev);
427 return 0;
428err_configure_peer:
429 unregister_netdevice(peer);
430 return err;
431err_register_peer:
432 free_netdev(peer);
433 return err;
434}
435
436static struct bpf_mprog_entry *netkit_entry_fetch(struct net_device *dev,
437 bool bundle_fallback)
438{
439 struct netkit *nk = netkit_priv(dev);
440 struct bpf_mprog_entry *entry;
441
442 ASSERT_RTNL();
443 entry = rcu_dereference_rtnl(nk->active);
444 if (entry)
445 return entry;
446 if (bundle_fallback)
447 return &nk->bundle.a;
448 return NULL;
449}
450
451static void netkit_entry_update(struct net_device *dev,
452 struct bpf_mprog_entry *entry)
453{
454 struct netkit *nk = netkit_priv(dev);
455
456 ASSERT_RTNL();
457 rcu_assign_pointer(nk->active, entry);
458}
459
460static void netkit_entry_sync(void)
461{
462 synchronize_rcu();
463}
464
465static struct net_device *netkit_dev_fetch(struct net *net, u32 ifindex, u32 which)
466{
467 struct net_device *dev;
468 struct netkit *nk;
469
470 ASSERT_RTNL();
471
472 switch (which) {
473 case BPF_NETKIT_PRIMARY:
474 case BPF_NETKIT_PEER:
475 break;
476 default:
477 return ERR_PTR(-EINVAL);
478 }
479
480 dev = __dev_get_by_index(net, ifindex);
481 if (!dev)
482 return ERR_PTR(-ENODEV);
483 if (dev->netdev_ops != &netkit_netdev_ops)
484 return ERR_PTR(-ENXIO);
485
486 nk = netkit_priv(dev);
487 if (!nk->primary)
488 return ERR_PTR(-EACCES);
489 if (which == BPF_NETKIT_PEER) {
490 dev = rcu_dereference_rtnl(nk->peer);
491 if (!dev)
492 return ERR_PTR(-ENODEV);
493 }
494 return dev;
495}
496
497int netkit_prog_attach(const union bpf_attr *attr, struct bpf_prog *prog)
498{
499 struct bpf_mprog_entry *entry, *entry_new;
500 struct bpf_prog *replace_prog = NULL;
501 struct net_device *dev;
502 int ret;
503
504 rtnl_lock();
505 dev = netkit_dev_fetch(current->nsproxy->net_ns, attr->target_ifindex,
506 attr->attach_type);
507 if (IS_ERR(dev)) {
508 ret = PTR_ERR(dev);
509 goto out;
510 }
511 entry = netkit_entry_fetch(dev, true);
512 if (attr->attach_flags & BPF_F_REPLACE) {
513 replace_prog = bpf_prog_get_type(attr->replace_bpf_fd,
514 prog->type);
515 if (IS_ERR(replace_prog)) {
516 ret = PTR_ERR(replace_prog);
517 replace_prog = NULL;
518 goto out;
519 }
520 }
521 ret = bpf_mprog_attach(entry, &entry_new, prog, NULL, replace_prog,
522 attr->attach_flags, attr->relative_fd,
523 attr->expected_revision);
524 if (!ret) {
525 if (entry != entry_new) {
526 netkit_entry_update(dev, entry_new);
527 netkit_entry_sync();
528 }
529 bpf_mprog_commit(entry);
530 }
531out:
532 if (replace_prog)
533 bpf_prog_put(replace_prog);
534 rtnl_unlock();
535 return ret;
536}
537
538int netkit_prog_detach(const union bpf_attr *attr, struct bpf_prog *prog)
539{
540 struct bpf_mprog_entry *entry, *entry_new;
541 struct net_device *dev;
542 int ret;
543
544 rtnl_lock();
545 dev = netkit_dev_fetch(current->nsproxy->net_ns, attr->target_ifindex,
546 attr->attach_type);
547 if (IS_ERR(dev)) {
548 ret = PTR_ERR(dev);
549 goto out;
550 }
551 entry = netkit_entry_fetch(dev, false);
552 if (!entry) {
553 ret = -ENOENT;
554 goto out;
555 }
556 ret = bpf_mprog_detach(entry, &entry_new, prog, NULL, attr->attach_flags,
557 attr->relative_fd, attr->expected_revision);
558 if (!ret) {
559 if (!bpf_mprog_total(entry_new))
560 entry_new = NULL;
561 netkit_entry_update(dev, entry_new);
562 netkit_entry_sync();
563 bpf_mprog_commit(entry);
564 }
565out:
566 rtnl_unlock();
567 return ret;
568}
569
570int netkit_prog_query(const union bpf_attr *attr, union bpf_attr __user *uattr)
571{
572 struct net_device *dev;
573 int ret;
574
575 rtnl_lock();
576 dev = netkit_dev_fetch(current->nsproxy->net_ns,
577 attr->query.target_ifindex,
578 attr->query.attach_type);
579 if (IS_ERR(dev)) {
580 ret = PTR_ERR(dev);
581 goto out;
582 }
583 ret = bpf_mprog_query(attr, uattr, netkit_entry_fetch(dev, false));
584out:
585 rtnl_unlock();
586 return ret;
587}
588
589static struct netkit_link *netkit_link(const struct bpf_link *link)
590{
591 return container_of(link, struct netkit_link, link);
592}
593
594static int netkit_link_prog_attach(struct bpf_link *link, u32 flags,
595 u32 id_or_fd, u64 revision)
596{
597 struct netkit_link *nkl = netkit_link(link);
598 struct bpf_mprog_entry *entry, *entry_new;
599 struct net_device *dev = nkl->dev;
600 int ret;
601
602 ASSERT_RTNL();
603 entry = netkit_entry_fetch(dev, true);
604 ret = bpf_mprog_attach(entry, &entry_new, link->prog, link, NULL, flags,
605 id_or_fd, revision);
606 if (!ret) {
607 if (entry != entry_new) {
608 netkit_entry_update(dev, entry_new);
609 netkit_entry_sync();
610 }
611 bpf_mprog_commit(entry);
612 }
613 return ret;
614}
615
616static void netkit_link_release(struct bpf_link *link)
617{
618 struct netkit_link *nkl = netkit_link(link);
619 struct bpf_mprog_entry *entry, *entry_new;
620 struct net_device *dev;
621 int ret = 0;
622
623 rtnl_lock();
624 dev = nkl->dev;
625 if (!dev)
626 goto out;
627 entry = netkit_entry_fetch(dev, false);
628 if (!entry) {
629 ret = -ENOENT;
630 goto out;
631 }
632 ret = bpf_mprog_detach(entry, &entry_new, link->prog, link, 0, 0, 0);
633 if (!ret) {
634 if (!bpf_mprog_total(entry_new))
635 entry_new = NULL;
636 netkit_entry_update(dev, entry_new);
637 netkit_entry_sync();
638 bpf_mprog_commit(entry);
639 nkl->dev = NULL;
640 }
641out:
642 WARN_ON_ONCE(ret);
643 rtnl_unlock();
644}
645
646static int netkit_link_update(struct bpf_link *link, struct bpf_prog *nprog,
647 struct bpf_prog *oprog)
648{
649 struct netkit_link *nkl = netkit_link(link);
650 struct bpf_mprog_entry *entry, *entry_new;
651 struct net_device *dev;
652 int ret = 0;
653
654 rtnl_lock();
655 dev = nkl->dev;
656 if (!dev) {
657 ret = -ENOLINK;
658 goto out;
659 }
660 if (oprog && link->prog != oprog) {
661 ret = -EPERM;
662 goto out;
663 }
664 oprog = link->prog;
665 if (oprog == nprog) {
666 bpf_prog_put(nprog);
667 goto out;
668 }
669 entry = netkit_entry_fetch(dev, false);
670 if (!entry) {
671 ret = -ENOENT;
672 goto out;
673 }
674 ret = bpf_mprog_attach(entry, &entry_new, nprog, link, oprog,
675 BPF_F_REPLACE | BPF_F_ID,
676 link->prog->aux->id, 0);
677 if (!ret) {
678 WARN_ON_ONCE(entry != entry_new);
679 oprog = xchg(&link->prog, nprog);
680 bpf_prog_put(oprog);
681 bpf_mprog_commit(entry);
682 }
683out:
684 rtnl_unlock();
685 return ret;
686}
687
688static void netkit_link_dealloc(struct bpf_link *link)
689{
690 kfree(netkit_link(link));
691}
692
693static void netkit_link_fdinfo(const struct bpf_link *link, struct seq_file *seq)
694{
695 const struct netkit_link *nkl = netkit_link(link);
696 u32 ifindex = 0;
697
698 rtnl_lock();
699 if (nkl->dev)
700 ifindex = nkl->dev->ifindex;
701 rtnl_unlock();
702
703 seq_printf(seq, "ifindex:\t%u\n", ifindex);
704 seq_printf(seq, "attach_type:\t%u (%s)\n",
705 nkl->location,
706 nkl->location == BPF_NETKIT_PRIMARY ? "primary" : "peer");
707}
708
709static int netkit_link_fill_info(const struct bpf_link *link,
710 struct bpf_link_info *info)
711{
712 const struct netkit_link *nkl = netkit_link(link);
713 u32 ifindex = 0;
714
715 rtnl_lock();
716 if (nkl->dev)
717 ifindex = nkl->dev->ifindex;
718 rtnl_unlock();
719
720 info->netkit.ifindex = ifindex;
721 info->netkit.attach_type = nkl->location;
722 return 0;
723}
724
725static int netkit_link_detach(struct bpf_link *link)
726{
727 netkit_link_release(link);
728 return 0;
729}
730
731static const struct bpf_link_ops netkit_link_lops = {
732 .release = netkit_link_release,
733 .detach = netkit_link_detach,
734 .dealloc = netkit_link_dealloc,
735 .update_prog = netkit_link_update,
736 .show_fdinfo = netkit_link_fdinfo,
737 .fill_link_info = netkit_link_fill_info,
738};
739
740static int netkit_link_init(struct netkit_link *nkl,
741 struct bpf_link_primer *link_primer,
742 const union bpf_attr *attr,
743 struct net_device *dev,
744 struct bpf_prog *prog)
745{
746 bpf_link_init(&nkl->link, BPF_LINK_TYPE_NETKIT,
747 &netkit_link_lops, prog);
748 nkl->location = attr->link_create.attach_type;
749 nkl->dev = dev;
750 return bpf_link_prime(&nkl->link, link_primer);
751}
752
753int netkit_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
754{
755 struct bpf_link_primer link_primer;
756 struct netkit_link *nkl;
757 struct net_device *dev;
758 int ret;
759
760 rtnl_lock();
761 dev = netkit_dev_fetch(current->nsproxy->net_ns,
762 attr->link_create.target_ifindex,
763 attr->link_create.attach_type);
764 if (IS_ERR(dev)) {
765 ret = PTR_ERR(dev);
766 goto out;
767 }
768 nkl = kzalloc(sizeof(*nkl), GFP_KERNEL_ACCOUNT);
769 if (!nkl) {
770 ret = -ENOMEM;
771 goto out;
772 }
773 ret = netkit_link_init(nkl, &link_primer, attr, dev, prog);
774 if (ret) {
775 kfree(nkl);
776 goto out;
777 }
778 ret = netkit_link_prog_attach(&nkl->link,
779 attr->link_create.flags,
780 attr->link_create.netkit.relative_fd,
781 attr->link_create.netkit.expected_revision);
782 if (ret) {
783 nkl->dev = NULL;
784 bpf_link_cleanup(&link_primer);
785 goto out;
786 }
787 ret = bpf_link_settle(&link_primer);
788out:
789 rtnl_unlock();
790 return ret;
791}
792
793static void netkit_release_all(struct net_device *dev)
794{
795 struct bpf_mprog_entry *entry;
796 struct bpf_tuple tuple = {};
797 struct bpf_mprog_fp *fp;
798 struct bpf_mprog_cp *cp;
799
800 entry = netkit_entry_fetch(dev, false);
801 if (!entry)
802 return;
803 netkit_entry_update(dev, NULL);
804 netkit_entry_sync();
805 bpf_mprog_foreach_tuple(entry, fp, cp, tuple) {
806 if (tuple.link)
807 netkit_link(tuple.link)->dev = NULL;
808 else
809 bpf_prog_put(tuple.prog);
810 }
811}
812
813static void netkit_uninit(struct net_device *dev)
814{
815 netkit_release_all(dev);
816}
817
818static void netkit_del_link(struct net_device *dev, struct list_head *head)
819{
820 struct netkit *nk = netkit_priv(dev);
821 struct net_device *peer = rtnl_dereference(nk->peer);
822
823 RCU_INIT_POINTER(nk->peer, NULL);
824 unregister_netdevice_queue(dev, head);
825 if (peer) {
826 nk = netkit_priv(peer);
827 RCU_INIT_POINTER(nk->peer, NULL);
828 unregister_netdevice_queue(peer, head);
829 }
830}
831
832static int netkit_change_link(struct net_device *dev, struct nlattr *tb[],
833 struct nlattr *data[],
834 struct netlink_ext_ack *extack)
835{
836 struct netkit *nk = netkit_priv(dev);
837 struct net_device *peer = rtnl_dereference(nk->peer);
838 enum netkit_action policy;
839 struct nlattr *attr;
840 int err;
841
842 if (!nk->primary) {
843 NL_SET_ERR_MSG(extack,
844 "netkit link settings can be changed only through the primary device");
845 return -EACCES;
846 }
847
848 if (data[IFLA_NETKIT_MODE]) {
849 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_MODE],
850 "netkit link operating mode cannot be changed after device creation");
851 return -EACCES;
852 }
853
854 if (data[IFLA_NETKIT_PEER_INFO]) {
855 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_PEER_INFO],
856 "netkit peer info cannot be changed after device creation");
857 return -EINVAL;
858 }
859
860 if (data[IFLA_NETKIT_POLICY]) {
861 attr = data[IFLA_NETKIT_POLICY];
862 policy = nla_get_u32(attr);
863 err = netkit_check_policy(policy, attr, extack);
864 if (err)
865 return err;
866 WRITE_ONCE(nk->policy, policy);
867 }
868
869 if (data[IFLA_NETKIT_PEER_POLICY]) {
870 err = -EOPNOTSUPP;
871 attr = data[IFLA_NETKIT_PEER_POLICY];
872 policy = nla_get_u32(attr);
873 if (peer)
874 err = netkit_check_policy(policy, attr, extack);
875 if (err)
876 return err;
877 nk = netkit_priv(peer);
878 WRITE_ONCE(nk->policy, policy);
879 }
880
881 return 0;
882}
883
884static size_t netkit_get_size(const struct net_device *dev)
885{
886 return nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_POLICY */
887 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_PEER_POLICY */
888 nla_total_size(sizeof(u8)) + /* IFLA_NETKIT_PRIMARY */
889 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_MODE */
890 0;
891}
892
893static int netkit_fill_info(struct sk_buff *skb, const struct net_device *dev)
894{
895 struct netkit *nk = netkit_priv(dev);
896 struct net_device *peer = rtnl_dereference(nk->peer);
897
898 if (nla_put_u8(skb, IFLA_NETKIT_PRIMARY, nk->primary))
899 return -EMSGSIZE;
900 if (nla_put_u32(skb, IFLA_NETKIT_POLICY, nk->policy))
901 return -EMSGSIZE;
902 if (nla_put_u32(skb, IFLA_NETKIT_MODE, nk->mode))
903 return -EMSGSIZE;
904
905 if (peer) {
906 nk = netkit_priv(peer);
907 if (nla_put_u32(skb, IFLA_NETKIT_PEER_POLICY, nk->policy))
908 return -EMSGSIZE;
909 }
910
911 return 0;
912}
913
914static const struct nla_policy netkit_policy[IFLA_NETKIT_MAX + 1] = {
915 [IFLA_NETKIT_PEER_INFO] = { .len = sizeof(struct ifinfomsg) },
916 [IFLA_NETKIT_POLICY] = { .type = NLA_U32 },
917 [IFLA_NETKIT_MODE] = { .type = NLA_U32 },
918 [IFLA_NETKIT_PEER_POLICY] = { .type = NLA_U32 },
919 [IFLA_NETKIT_PRIMARY] = { .type = NLA_REJECT,
920 .reject_message = "Primary attribute is read-only" },
921};
922
923static struct rtnl_link_ops netkit_link_ops = {
924 .kind = DRV_NAME,
925 .priv_size = sizeof(struct netkit),
926 .setup = netkit_setup,
927 .newlink = netkit_new_link,
928 .dellink = netkit_del_link,
929 .changelink = netkit_change_link,
930 .get_link_net = netkit_get_link_net,
931 .get_size = netkit_get_size,
932 .fill_info = netkit_fill_info,
933 .policy = netkit_policy,
934 .validate = netkit_validate,
935 .maxtype = IFLA_NETKIT_MAX,
936};
937
938static __init int netkit_init(void)
939{
940 BUILD_BUG_ON((int)NETKIT_NEXT != (int)TCX_NEXT ||
941 (int)NETKIT_PASS != (int)TCX_PASS ||
942 (int)NETKIT_DROP != (int)TCX_DROP ||
943 (int)NETKIT_REDIRECT != (int)TCX_REDIRECT);
944
945 return rtnl_link_register(&netkit_link_ops);
946}
947
948static __exit void netkit_exit(void)
949{
950 rtnl_link_unregister(&netkit_link_ops);
951}
952
953module_init(netkit_init);
954module_exit(netkit_exit);
955
956MODULE_DESCRIPTION("BPF-programmable network device");
957MODULE_AUTHOR("Daniel Borkmann <daniel@iogearbox.net>");
958MODULE_AUTHOR("Nikolay Aleksandrov <razor@blackwall.org>");
959MODULE_LICENSE("GPL");
960MODULE_ALIAS_RTNL_LINK(DRV_NAME);
1// SPDX-License-Identifier: GPL-2.0-only
2/* Copyright (c) 2023 Isovalent */
3
4#include <linux/netdevice.h>
5#include <linux/ethtool.h>
6#include <linux/etherdevice.h>
7#include <linux/filter.h>
8#include <linux/netfilter_netdev.h>
9#include <linux/bpf_mprog.h>
10#include <linux/indirect_call_wrapper.h>
11
12#include <net/netkit.h>
13#include <net/dst.h>
14#include <net/tcx.h>
15
16#define DRV_NAME "netkit"
17
18struct netkit {
19 /* Needed in fast-path */
20 struct net_device __rcu *peer;
21 struct bpf_mprog_entry __rcu *active;
22 enum netkit_action policy;
23 enum netkit_scrub scrub;
24 struct bpf_mprog_bundle bundle;
25
26 /* Needed in slow-path */
27 enum netkit_mode mode;
28 bool primary;
29 u32 headroom;
30};
31
32struct netkit_link {
33 struct bpf_link link;
34 struct net_device *dev;
35 u32 location;
36};
37
38static __always_inline int
39netkit_run(const struct bpf_mprog_entry *entry, struct sk_buff *skb,
40 enum netkit_action ret)
41{
42 const struct bpf_mprog_fp *fp;
43 const struct bpf_prog *prog;
44
45 bpf_mprog_foreach_prog(entry, fp, prog) {
46 bpf_compute_data_pointers(skb);
47 ret = bpf_prog_run(prog, skb);
48 if (ret != NETKIT_NEXT)
49 break;
50 }
51 return ret;
52}
53
54static void netkit_xnet(struct sk_buff *skb)
55{
56 skb->priority = 0;
57 skb->mark = 0;
58}
59
60static void netkit_prep_forward(struct sk_buff *skb,
61 bool xnet, bool xnet_scrub)
62{
63 skb_scrub_packet(skb, false);
64 nf_skip_egress(skb, true);
65 skb_reset_mac_header(skb);
66 if (!xnet)
67 return;
68 ipvs_reset(skb);
69 skb_clear_tstamp(skb);
70 if (xnet_scrub)
71 netkit_xnet(skb);
72}
73
74static struct netkit *netkit_priv(const struct net_device *dev)
75{
76 return netdev_priv(dev);
77}
78
79static netdev_tx_t netkit_xmit(struct sk_buff *skb, struct net_device *dev)
80{
81 struct bpf_net_context __bpf_net_ctx, *bpf_net_ctx;
82 struct netkit *nk = netkit_priv(dev);
83 enum netkit_action ret = READ_ONCE(nk->policy);
84 netdev_tx_t ret_dev = NET_XMIT_SUCCESS;
85 const struct bpf_mprog_entry *entry;
86 struct net_device *peer;
87 int len = skb->len;
88
89 bpf_net_ctx = bpf_net_ctx_set(&__bpf_net_ctx);
90 rcu_read_lock();
91 peer = rcu_dereference(nk->peer);
92 if (unlikely(!peer || !(peer->flags & IFF_UP) ||
93 !pskb_may_pull(skb, ETH_HLEN) ||
94 skb_orphan_frags(skb, GFP_ATOMIC)))
95 goto drop;
96 netkit_prep_forward(skb, !net_eq(dev_net(dev), dev_net(peer)),
97 nk->scrub);
98 eth_skb_pkt_type(skb, peer);
99 skb->dev = peer;
100 entry = rcu_dereference(nk->active);
101 if (entry)
102 ret = netkit_run(entry, skb, ret);
103 switch (ret) {
104 case NETKIT_NEXT:
105 case NETKIT_PASS:
106 eth_skb_pull_mac(skb);
107 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
108 if (likely(__netif_rx(skb) == NET_RX_SUCCESS)) {
109 dev_sw_netstats_tx_add(dev, 1, len);
110 dev_sw_netstats_rx_add(peer, len);
111 } else {
112 goto drop_stats;
113 }
114 break;
115 case NETKIT_REDIRECT:
116 dev_sw_netstats_tx_add(dev, 1, len);
117 skb_do_redirect(skb);
118 break;
119 case NETKIT_DROP:
120 default:
121drop:
122 kfree_skb(skb);
123drop_stats:
124 dev_core_stats_tx_dropped_inc(dev);
125 ret_dev = NET_XMIT_DROP;
126 break;
127 }
128 rcu_read_unlock();
129 bpf_net_ctx_clear(bpf_net_ctx);
130 return ret_dev;
131}
132
133static int netkit_open(struct net_device *dev)
134{
135 struct netkit *nk = netkit_priv(dev);
136 struct net_device *peer = rtnl_dereference(nk->peer);
137
138 if (!peer)
139 return -ENOTCONN;
140 if (peer->flags & IFF_UP) {
141 netif_carrier_on(dev);
142 netif_carrier_on(peer);
143 }
144 return 0;
145}
146
147static int netkit_close(struct net_device *dev)
148{
149 struct netkit *nk = netkit_priv(dev);
150 struct net_device *peer = rtnl_dereference(nk->peer);
151
152 netif_carrier_off(dev);
153 if (peer)
154 netif_carrier_off(peer);
155 return 0;
156}
157
158static int netkit_get_iflink(const struct net_device *dev)
159{
160 struct netkit *nk = netkit_priv(dev);
161 struct net_device *peer;
162 int iflink = 0;
163
164 rcu_read_lock();
165 peer = rcu_dereference(nk->peer);
166 if (peer)
167 iflink = READ_ONCE(peer->ifindex);
168 rcu_read_unlock();
169 return iflink;
170}
171
172static void netkit_set_multicast(struct net_device *dev)
173{
174 /* Nothing to do, we receive whatever gets pushed to us! */
175}
176
177static int netkit_set_macaddr(struct net_device *dev, void *sa)
178{
179 struct netkit *nk = netkit_priv(dev);
180
181 if (nk->mode != NETKIT_L2)
182 return -EOPNOTSUPP;
183
184 return eth_mac_addr(dev, sa);
185}
186
187static void netkit_set_headroom(struct net_device *dev, int headroom)
188{
189 struct netkit *nk = netkit_priv(dev), *nk2;
190 struct net_device *peer;
191
192 if (headroom < 0)
193 headroom = NET_SKB_PAD;
194
195 rcu_read_lock();
196 peer = rcu_dereference(nk->peer);
197 if (unlikely(!peer))
198 goto out;
199
200 nk2 = netkit_priv(peer);
201 nk->headroom = headroom;
202 headroom = max(nk->headroom, nk2->headroom);
203
204 peer->needed_headroom = headroom;
205 dev->needed_headroom = headroom;
206out:
207 rcu_read_unlock();
208}
209
210INDIRECT_CALLABLE_SCOPE struct net_device *netkit_peer_dev(struct net_device *dev)
211{
212 return rcu_dereference(netkit_priv(dev)->peer);
213}
214
215static void netkit_get_stats(struct net_device *dev,
216 struct rtnl_link_stats64 *stats)
217{
218 dev_fetch_sw_netstats(stats, dev->tstats);
219 stats->tx_dropped = DEV_STATS_READ(dev, tx_dropped);
220}
221
222static void netkit_uninit(struct net_device *dev);
223
224static const struct net_device_ops netkit_netdev_ops = {
225 .ndo_open = netkit_open,
226 .ndo_stop = netkit_close,
227 .ndo_start_xmit = netkit_xmit,
228 .ndo_set_rx_mode = netkit_set_multicast,
229 .ndo_set_rx_headroom = netkit_set_headroom,
230 .ndo_set_mac_address = netkit_set_macaddr,
231 .ndo_get_iflink = netkit_get_iflink,
232 .ndo_get_peer_dev = netkit_peer_dev,
233 .ndo_get_stats64 = netkit_get_stats,
234 .ndo_uninit = netkit_uninit,
235 .ndo_features_check = passthru_features_check,
236};
237
238static void netkit_get_drvinfo(struct net_device *dev,
239 struct ethtool_drvinfo *info)
240{
241 strscpy(info->driver, DRV_NAME, sizeof(info->driver));
242}
243
244static const struct ethtool_ops netkit_ethtool_ops = {
245 .get_drvinfo = netkit_get_drvinfo,
246};
247
248static void netkit_setup(struct net_device *dev)
249{
250 static const netdev_features_t netkit_features_hw_vlan =
251 NETIF_F_HW_VLAN_CTAG_TX |
252 NETIF_F_HW_VLAN_CTAG_RX |
253 NETIF_F_HW_VLAN_STAG_TX |
254 NETIF_F_HW_VLAN_STAG_RX;
255 static const netdev_features_t netkit_features =
256 netkit_features_hw_vlan |
257 NETIF_F_SG |
258 NETIF_F_FRAGLIST |
259 NETIF_F_HW_CSUM |
260 NETIF_F_RXCSUM |
261 NETIF_F_SCTP_CRC |
262 NETIF_F_HIGHDMA |
263 NETIF_F_GSO_SOFTWARE |
264 NETIF_F_GSO_ENCAP_ALL;
265
266 ether_setup(dev);
267 dev->max_mtu = ETH_MAX_MTU;
268 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
269
270 dev->flags |= IFF_NOARP;
271 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
272 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
273 dev->priv_flags |= IFF_PHONY_HEADROOM;
274 dev->priv_flags |= IFF_NO_QUEUE;
275 dev->priv_flags |= IFF_DISABLE_NETPOLL;
276 dev->lltx = true;
277
278 dev->ethtool_ops = &netkit_ethtool_ops;
279 dev->netdev_ops = &netkit_netdev_ops;
280
281 dev->features |= netkit_features;
282 dev->hw_features = netkit_features;
283 dev->hw_enc_features = netkit_features;
284 dev->mpls_features = NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
285 dev->vlan_features = dev->features & ~netkit_features_hw_vlan;
286
287 dev->needs_free_netdev = true;
288
289 netif_set_tso_max_size(dev, GSO_MAX_SIZE);
290}
291
292static struct net *netkit_get_link_net(const struct net_device *dev)
293{
294 struct netkit *nk = netkit_priv(dev);
295 struct net_device *peer = rtnl_dereference(nk->peer);
296
297 return peer ? dev_net(peer) : dev_net(dev);
298}
299
300static int netkit_check_policy(int policy, struct nlattr *tb,
301 struct netlink_ext_ack *extack)
302{
303 switch (policy) {
304 case NETKIT_PASS:
305 case NETKIT_DROP:
306 return 0;
307 default:
308 NL_SET_ERR_MSG_ATTR(extack, tb,
309 "Provided default xmit policy not supported");
310 return -EINVAL;
311 }
312}
313
314static int netkit_validate(struct nlattr *tb[], struct nlattr *data[],
315 struct netlink_ext_ack *extack)
316{
317 struct nlattr *attr = tb[IFLA_ADDRESS];
318
319 if (!attr)
320 return 0;
321 if (nla_len(attr) != ETH_ALEN)
322 return -EINVAL;
323 if (!is_valid_ether_addr(nla_data(attr)))
324 return -EADDRNOTAVAIL;
325 return 0;
326}
327
328static struct rtnl_link_ops netkit_link_ops;
329
330static int netkit_new_link(struct net *peer_net, struct net_device *dev,
331 struct nlattr *tb[], struct nlattr *data[],
332 struct netlink_ext_ack *extack)
333{
334 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp = tb, *attr;
335 enum netkit_action policy_prim = NETKIT_PASS;
336 enum netkit_action policy_peer = NETKIT_PASS;
337 enum netkit_scrub scrub_prim = NETKIT_SCRUB_DEFAULT;
338 enum netkit_scrub scrub_peer = NETKIT_SCRUB_DEFAULT;
339 enum netkit_mode mode = NETKIT_L3;
340 unsigned char ifname_assign_type;
341 struct ifinfomsg *ifmp = NULL;
342 struct net_device *peer;
343 char ifname[IFNAMSIZ];
344 struct netkit *nk;
345 int err;
346
347 if (data) {
348 if (data[IFLA_NETKIT_MODE])
349 mode = nla_get_u32(data[IFLA_NETKIT_MODE]);
350 if (data[IFLA_NETKIT_PEER_INFO]) {
351 attr = data[IFLA_NETKIT_PEER_INFO];
352 ifmp = nla_data(attr);
353 rtnl_nla_parse_ifinfomsg(peer_tb, attr, extack);
354 tbp = peer_tb;
355 }
356 if (data[IFLA_NETKIT_SCRUB])
357 scrub_prim = nla_get_u32(data[IFLA_NETKIT_SCRUB]);
358 if (data[IFLA_NETKIT_PEER_SCRUB])
359 scrub_peer = nla_get_u32(data[IFLA_NETKIT_PEER_SCRUB]);
360 if (data[IFLA_NETKIT_POLICY]) {
361 attr = data[IFLA_NETKIT_POLICY];
362 policy_prim = nla_get_u32(attr);
363 err = netkit_check_policy(policy_prim, attr, extack);
364 if (err < 0)
365 return err;
366 }
367 if (data[IFLA_NETKIT_PEER_POLICY]) {
368 attr = data[IFLA_NETKIT_PEER_POLICY];
369 policy_peer = nla_get_u32(attr);
370 err = netkit_check_policy(policy_peer, attr, extack);
371 if (err < 0)
372 return err;
373 }
374 }
375
376 if (ifmp && tbp[IFLA_IFNAME]) {
377 nla_strscpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
378 ifname_assign_type = NET_NAME_USER;
379 } else {
380 strscpy(ifname, "nk%d", IFNAMSIZ);
381 ifname_assign_type = NET_NAME_ENUM;
382 }
383 if (mode != NETKIT_L2 &&
384 (tb[IFLA_ADDRESS] || tbp[IFLA_ADDRESS]))
385 return -EOPNOTSUPP;
386
387 peer = rtnl_create_link(peer_net, ifname, ifname_assign_type,
388 &netkit_link_ops, tbp, extack);
389 if (IS_ERR(peer))
390 return PTR_ERR(peer);
391
392 netif_inherit_tso_max(peer, dev);
393
394 if (mode == NETKIT_L2 && !(ifmp && tbp[IFLA_ADDRESS]))
395 eth_hw_addr_random(peer);
396 if (ifmp && dev->ifindex)
397 peer->ifindex = ifmp->ifi_index;
398
399 nk = netkit_priv(peer);
400 nk->primary = false;
401 nk->policy = policy_peer;
402 nk->scrub = scrub_peer;
403 nk->mode = mode;
404 bpf_mprog_bundle_init(&nk->bundle);
405
406 err = register_netdevice(peer);
407 if (err < 0)
408 goto err_register_peer;
409 netif_carrier_off(peer);
410 if (mode == NETKIT_L2)
411 dev_change_flags(peer, peer->flags & ~IFF_NOARP, NULL);
412
413 err = rtnl_configure_link(peer, NULL, 0, NULL);
414 if (err < 0)
415 goto err_configure_peer;
416
417 if (mode == NETKIT_L2 && !tb[IFLA_ADDRESS])
418 eth_hw_addr_random(dev);
419 if (tb[IFLA_IFNAME])
420 nla_strscpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
421 else
422 strscpy(dev->name, "nk%d", IFNAMSIZ);
423
424 nk = netkit_priv(dev);
425 nk->primary = true;
426 nk->policy = policy_prim;
427 nk->scrub = scrub_prim;
428 nk->mode = mode;
429 bpf_mprog_bundle_init(&nk->bundle);
430
431 err = register_netdevice(dev);
432 if (err < 0)
433 goto err_configure_peer;
434 netif_carrier_off(dev);
435 if (mode == NETKIT_L2)
436 dev_change_flags(dev, dev->flags & ~IFF_NOARP, NULL);
437
438 rcu_assign_pointer(netkit_priv(dev)->peer, peer);
439 rcu_assign_pointer(netkit_priv(peer)->peer, dev);
440 return 0;
441err_configure_peer:
442 unregister_netdevice(peer);
443 return err;
444err_register_peer:
445 free_netdev(peer);
446 return err;
447}
448
449static struct bpf_mprog_entry *netkit_entry_fetch(struct net_device *dev,
450 bool bundle_fallback)
451{
452 struct netkit *nk = netkit_priv(dev);
453 struct bpf_mprog_entry *entry;
454
455 ASSERT_RTNL();
456 entry = rcu_dereference_rtnl(nk->active);
457 if (entry)
458 return entry;
459 if (bundle_fallback)
460 return &nk->bundle.a;
461 return NULL;
462}
463
464static void netkit_entry_update(struct net_device *dev,
465 struct bpf_mprog_entry *entry)
466{
467 struct netkit *nk = netkit_priv(dev);
468
469 ASSERT_RTNL();
470 rcu_assign_pointer(nk->active, entry);
471}
472
473static void netkit_entry_sync(void)
474{
475 synchronize_rcu();
476}
477
478static struct net_device *netkit_dev_fetch(struct net *net, u32 ifindex, u32 which)
479{
480 struct net_device *dev;
481 struct netkit *nk;
482
483 ASSERT_RTNL();
484
485 switch (which) {
486 case BPF_NETKIT_PRIMARY:
487 case BPF_NETKIT_PEER:
488 break;
489 default:
490 return ERR_PTR(-EINVAL);
491 }
492
493 dev = __dev_get_by_index(net, ifindex);
494 if (!dev)
495 return ERR_PTR(-ENODEV);
496 if (dev->netdev_ops != &netkit_netdev_ops)
497 return ERR_PTR(-ENXIO);
498
499 nk = netkit_priv(dev);
500 if (!nk->primary)
501 return ERR_PTR(-EACCES);
502 if (which == BPF_NETKIT_PEER) {
503 dev = rcu_dereference_rtnl(nk->peer);
504 if (!dev)
505 return ERR_PTR(-ENODEV);
506 }
507 return dev;
508}
509
510int netkit_prog_attach(const union bpf_attr *attr, struct bpf_prog *prog)
511{
512 struct bpf_mprog_entry *entry, *entry_new;
513 struct bpf_prog *replace_prog = NULL;
514 struct net_device *dev;
515 int ret;
516
517 rtnl_lock();
518 dev = netkit_dev_fetch(current->nsproxy->net_ns, attr->target_ifindex,
519 attr->attach_type);
520 if (IS_ERR(dev)) {
521 ret = PTR_ERR(dev);
522 goto out;
523 }
524 entry = netkit_entry_fetch(dev, true);
525 if (attr->attach_flags & BPF_F_REPLACE) {
526 replace_prog = bpf_prog_get_type(attr->replace_bpf_fd,
527 prog->type);
528 if (IS_ERR(replace_prog)) {
529 ret = PTR_ERR(replace_prog);
530 replace_prog = NULL;
531 goto out;
532 }
533 }
534 ret = bpf_mprog_attach(entry, &entry_new, prog, NULL, replace_prog,
535 attr->attach_flags, attr->relative_fd,
536 attr->expected_revision);
537 if (!ret) {
538 if (entry != entry_new) {
539 netkit_entry_update(dev, entry_new);
540 netkit_entry_sync();
541 }
542 bpf_mprog_commit(entry);
543 }
544out:
545 if (replace_prog)
546 bpf_prog_put(replace_prog);
547 rtnl_unlock();
548 return ret;
549}
550
551int netkit_prog_detach(const union bpf_attr *attr, struct bpf_prog *prog)
552{
553 struct bpf_mprog_entry *entry, *entry_new;
554 struct net_device *dev;
555 int ret;
556
557 rtnl_lock();
558 dev = netkit_dev_fetch(current->nsproxy->net_ns, attr->target_ifindex,
559 attr->attach_type);
560 if (IS_ERR(dev)) {
561 ret = PTR_ERR(dev);
562 goto out;
563 }
564 entry = netkit_entry_fetch(dev, false);
565 if (!entry) {
566 ret = -ENOENT;
567 goto out;
568 }
569 ret = bpf_mprog_detach(entry, &entry_new, prog, NULL, attr->attach_flags,
570 attr->relative_fd, attr->expected_revision);
571 if (!ret) {
572 if (!bpf_mprog_total(entry_new))
573 entry_new = NULL;
574 netkit_entry_update(dev, entry_new);
575 netkit_entry_sync();
576 bpf_mprog_commit(entry);
577 }
578out:
579 rtnl_unlock();
580 return ret;
581}
582
583int netkit_prog_query(const union bpf_attr *attr, union bpf_attr __user *uattr)
584{
585 struct net_device *dev;
586 int ret;
587
588 rtnl_lock();
589 dev = netkit_dev_fetch(current->nsproxy->net_ns,
590 attr->query.target_ifindex,
591 attr->query.attach_type);
592 if (IS_ERR(dev)) {
593 ret = PTR_ERR(dev);
594 goto out;
595 }
596 ret = bpf_mprog_query(attr, uattr, netkit_entry_fetch(dev, false));
597out:
598 rtnl_unlock();
599 return ret;
600}
601
602static struct netkit_link *netkit_link(const struct bpf_link *link)
603{
604 return container_of(link, struct netkit_link, link);
605}
606
607static int netkit_link_prog_attach(struct bpf_link *link, u32 flags,
608 u32 id_or_fd, u64 revision)
609{
610 struct netkit_link *nkl = netkit_link(link);
611 struct bpf_mprog_entry *entry, *entry_new;
612 struct net_device *dev = nkl->dev;
613 int ret;
614
615 ASSERT_RTNL();
616 entry = netkit_entry_fetch(dev, true);
617 ret = bpf_mprog_attach(entry, &entry_new, link->prog, link, NULL, flags,
618 id_or_fd, revision);
619 if (!ret) {
620 if (entry != entry_new) {
621 netkit_entry_update(dev, entry_new);
622 netkit_entry_sync();
623 }
624 bpf_mprog_commit(entry);
625 }
626 return ret;
627}
628
629static void netkit_link_release(struct bpf_link *link)
630{
631 struct netkit_link *nkl = netkit_link(link);
632 struct bpf_mprog_entry *entry, *entry_new;
633 struct net_device *dev;
634 int ret = 0;
635
636 rtnl_lock();
637 dev = nkl->dev;
638 if (!dev)
639 goto out;
640 entry = netkit_entry_fetch(dev, false);
641 if (!entry) {
642 ret = -ENOENT;
643 goto out;
644 }
645 ret = bpf_mprog_detach(entry, &entry_new, link->prog, link, 0, 0, 0);
646 if (!ret) {
647 if (!bpf_mprog_total(entry_new))
648 entry_new = NULL;
649 netkit_entry_update(dev, entry_new);
650 netkit_entry_sync();
651 bpf_mprog_commit(entry);
652 nkl->dev = NULL;
653 }
654out:
655 WARN_ON_ONCE(ret);
656 rtnl_unlock();
657}
658
659static int netkit_link_update(struct bpf_link *link, struct bpf_prog *nprog,
660 struct bpf_prog *oprog)
661{
662 struct netkit_link *nkl = netkit_link(link);
663 struct bpf_mprog_entry *entry, *entry_new;
664 struct net_device *dev;
665 int ret = 0;
666
667 rtnl_lock();
668 dev = nkl->dev;
669 if (!dev) {
670 ret = -ENOLINK;
671 goto out;
672 }
673 if (oprog && link->prog != oprog) {
674 ret = -EPERM;
675 goto out;
676 }
677 oprog = link->prog;
678 if (oprog == nprog) {
679 bpf_prog_put(nprog);
680 goto out;
681 }
682 entry = netkit_entry_fetch(dev, false);
683 if (!entry) {
684 ret = -ENOENT;
685 goto out;
686 }
687 ret = bpf_mprog_attach(entry, &entry_new, nprog, link, oprog,
688 BPF_F_REPLACE | BPF_F_ID,
689 link->prog->aux->id, 0);
690 if (!ret) {
691 WARN_ON_ONCE(entry != entry_new);
692 oprog = xchg(&link->prog, nprog);
693 bpf_prog_put(oprog);
694 bpf_mprog_commit(entry);
695 }
696out:
697 rtnl_unlock();
698 return ret;
699}
700
701static void netkit_link_dealloc(struct bpf_link *link)
702{
703 kfree(netkit_link(link));
704}
705
706static void netkit_link_fdinfo(const struct bpf_link *link, struct seq_file *seq)
707{
708 const struct netkit_link *nkl = netkit_link(link);
709 u32 ifindex = 0;
710
711 rtnl_lock();
712 if (nkl->dev)
713 ifindex = nkl->dev->ifindex;
714 rtnl_unlock();
715
716 seq_printf(seq, "ifindex:\t%u\n", ifindex);
717 seq_printf(seq, "attach_type:\t%u (%s)\n",
718 nkl->location,
719 nkl->location == BPF_NETKIT_PRIMARY ? "primary" : "peer");
720}
721
722static int netkit_link_fill_info(const struct bpf_link *link,
723 struct bpf_link_info *info)
724{
725 const struct netkit_link *nkl = netkit_link(link);
726 u32 ifindex = 0;
727
728 rtnl_lock();
729 if (nkl->dev)
730 ifindex = nkl->dev->ifindex;
731 rtnl_unlock();
732
733 info->netkit.ifindex = ifindex;
734 info->netkit.attach_type = nkl->location;
735 return 0;
736}
737
738static int netkit_link_detach(struct bpf_link *link)
739{
740 netkit_link_release(link);
741 return 0;
742}
743
744static const struct bpf_link_ops netkit_link_lops = {
745 .release = netkit_link_release,
746 .detach = netkit_link_detach,
747 .dealloc = netkit_link_dealloc,
748 .update_prog = netkit_link_update,
749 .show_fdinfo = netkit_link_fdinfo,
750 .fill_link_info = netkit_link_fill_info,
751};
752
753static int netkit_link_init(struct netkit_link *nkl,
754 struct bpf_link_primer *link_primer,
755 const union bpf_attr *attr,
756 struct net_device *dev,
757 struct bpf_prog *prog)
758{
759 bpf_link_init(&nkl->link, BPF_LINK_TYPE_NETKIT,
760 &netkit_link_lops, prog);
761 nkl->location = attr->link_create.attach_type;
762 nkl->dev = dev;
763 return bpf_link_prime(&nkl->link, link_primer);
764}
765
766int netkit_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
767{
768 struct bpf_link_primer link_primer;
769 struct netkit_link *nkl;
770 struct net_device *dev;
771 int ret;
772
773 rtnl_lock();
774 dev = netkit_dev_fetch(current->nsproxy->net_ns,
775 attr->link_create.target_ifindex,
776 attr->link_create.attach_type);
777 if (IS_ERR(dev)) {
778 ret = PTR_ERR(dev);
779 goto out;
780 }
781 nkl = kzalloc(sizeof(*nkl), GFP_KERNEL_ACCOUNT);
782 if (!nkl) {
783 ret = -ENOMEM;
784 goto out;
785 }
786 ret = netkit_link_init(nkl, &link_primer, attr, dev, prog);
787 if (ret) {
788 kfree(nkl);
789 goto out;
790 }
791 ret = netkit_link_prog_attach(&nkl->link,
792 attr->link_create.flags,
793 attr->link_create.netkit.relative_fd,
794 attr->link_create.netkit.expected_revision);
795 if (ret) {
796 nkl->dev = NULL;
797 bpf_link_cleanup(&link_primer);
798 goto out;
799 }
800 ret = bpf_link_settle(&link_primer);
801out:
802 rtnl_unlock();
803 return ret;
804}
805
806static void netkit_release_all(struct net_device *dev)
807{
808 struct bpf_mprog_entry *entry;
809 struct bpf_tuple tuple = {};
810 struct bpf_mprog_fp *fp;
811 struct bpf_mprog_cp *cp;
812
813 entry = netkit_entry_fetch(dev, false);
814 if (!entry)
815 return;
816 netkit_entry_update(dev, NULL);
817 netkit_entry_sync();
818 bpf_mprog_foreach_tuple(entry, fp, cp, tuple) {
819 if (tuple.link)
820 netkit_link(tuple.link)->dev = NULL;
821 else
822 bpf_prog_put(tuple.prog);
823 }
824}
825
826static void netkit_uninit(struct net_device *dev)
827{
828 netkit_release_all(dev);
829}
830
831static void netkit_del_link(struct net_device *dev, struct list_head *head)
832{
833 struct netkit *nk = netkit_priv(dev);
834 struct net_device *peer = rtnl_dereference(nk->peer);
835
836 RCU_INIT_POINTER(nk->peer, NULL);
837 unregister_netdevice_queue(dev, head);
838 if (peer) {
839 nk = netkit_priv(peer);
840 RCU_INIT_POINTER(nk->peer, NULL);
841 unregister_netdevice_queue(peer, head);
842 }
843}
844
845static int netkit_change_link(struct net_device *dev, struct nlattr *tb[],
846 struct nlattr *data[],
847 struct netlink_ext_ack *extack)
848{
849 struct netkit *nk = netkit_priv(dev);
850 struct net_device *peer = rtnl_dereference(nk->peer);
851 enum netkit_action policy;
852 struct nlattr *attr;
853 int err;
854
855 if (!nk->primary) {
856 NL_SET_ERR_MSG(extack,
857 "netkit link settings can be changed only through the primary device");
858 return -EACCES;
859 }
860
861 if (data[IFLA_NETKIT_MODE]) {
862 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_MODE],
863 "netkit link operating mode cannot be changed after device creation");
864 return -EACCES;
865 }
866
867 if (data[IFLA_NETKIT_SCRUB]) {
868 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_SCRUB],
869 "netkit scrubbing cannot be changed after device creation");
870 return -EACCES;
871 }
872
873 if (data[IFLA_NETKIT_PEER_SCRUB]) {
874 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_PEER_SCRUB],
875 "netkit scrubbing cannot be changed after device creation");
876 return -EACCES;
877 }
878
879 if (data[IFLA_NETKIT_PEER_INFO]) {
880 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_NETKIT_PEER_INFO],
881 "netkit peer info cannot be changed after device creation");
882 return -EINVAL;
883 }
884
885 if (data[IFLA_NETKIT_POLICY]) {
886 attr = data[IFLA_NETKIT_POLICY];
887 policy = nla_get_u32(attr);
888 err = netkit_check_policy(policy, attr, extack);
889 if (err)
890 return err;
891 WRITE_ONCE(nk->policy, policy);
892 }
893
894 if (data[IFLA_NETKIT_PEER_POLICY]) {
895 err = -EOPNOTSUPP;
896 attr = data[IFLA_NETKIT_PEER_POLICY];
897 policy = nla_get_u32(attr);
898 if (peer)
899 err = netkit_check_policy(policy, attr, extack);
900 if (err)
901 return err;
902 nk = netkit_priv(peer);
903 WRITE_ONCE(nk->policy, policy);
904 }
905
906 return 0;
907}
908
909static size_t netkit_get_size(const struct net_device *dev)
910{
911 return nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_POLICY */
912 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_PEER_POLICY */
913 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_SCRUB */
914 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_PEER_SCRUB */
915 nla_total_size(sizeof(u32)) + /* IFLA_NETKIT_MODE */
916 nla_total_size(sizeof(u8)) + /* IFLA_NETKIT_PRIMARY */
917 0;
918}
919
920static int netkit_fill_info(struct sk_buff *skb, const struct net_device *dev)
921{
922 struct netkit *nk = netkit_priv(dev);
923 struct net_device *peer = rtnl_dereference(nk->peer);
924
925 if (nla_put_u8(skb, IFLA_NETKIT_PRIMARY, nk->primary))
926 return -EMSGSIZE;
927 if (nla_put_u32(skb, IFLA_NETKIT_POLICY, nk->policy))
928 return -EMSGSIZE;
929 if (nla_put_u32(skb, IFLA_NETKIT_MODE, nk->mode))
930 return -EMSGSIZE;
931 if (nla_put_u32(skb, IFLA_NETKIT_SCRUB, nk->scrub))
932 return -EMSGSIZE;
933
934 if (peer) {
935 nk = netkit_priv(peer);
936 if (nla_put_u32(skb, IFLA_NETKIT_PEER_POLICY, nk->policy))
937 return -EMSGSIZE;
938 if (nla_put_u32(skb, IFLA_NETKIT_PEER_SCRUB, nk->scrub))
939 return -EMSGSIZE;
940 }
941
942 return 0;
943}
944
945static const struct nla_policy netkit_policy[IFLA_NETKIT_MAX + 1] = {
946 [IFLA_NETKIT_PEER_INFO] = { .len = sizeof(struct ifinfomsg) },
947 [IFLA_NETKIT_MODE] = NLA_POLICY_MAX(NLA_U32, NETKIT_L3),
948 [IFLA_NETKIT_POLICY] = { .type = NLA_U32 },
949 [IFLA_NETKIT_PEER_POLICY] = { .type = NLA_U32 },
950 [IFLA_NETKIT_SCRUB] = NLA_POLICY_MAX(NLA_U32, NETKIT_SCRUB_DEFAULT),
951 [IFLA_NETKIT_PEER_SCRUB] = NLA_POLICY_MAX(NLA_U32, NETKIT_SCRUB_DEFAULT),
952 [IFLA_NETKIT_PRIMARY] = { .type = NLA_REJECT,
953 .reject_message = "Primary attribute is read-only" },
954};
955
956static struct rtnl_link_ops netkit_link_ops = {
957 .kind = DRV_NAME,
958 .priv_size = sizeof(struct netkit),
959 .setup = netkit_setup,
960 .newlink = netkit_new_link,
961 .dellink = netkit_del_link,
962 .changelink = netkit_change_link,
963 .get_link_net = netkit_get_link_net,
964 .get_size = netkit_get_size,
965 .fill_info = netkit_fill_info,
966 .policy = netkit_policy,
967 .validate = netkit_validate,
968 .peer_type = IFLA_NETKIT_PEER_INFO,
969 .maxtype = IFLA_NETKIT_MAX,
970};
971
972static __init int netkit_init(void)
973{
974 BUILD_BUG_ON((int)NETKIT_NEXT != (int)TCX_NEXT ||
975 (int)NETKIT_PASS != (int)TCX_PASS ||
976 (int)NETKIT_DROP != (int)TCX_DROP ||
977 (int)NETKIT_REDIRECT != (int)TCX_REDIRECT);
978
979 return rtnl_link_register(&netkit_link_ops);
980}
981
982static __exit void netkit_exit(void)
983{
984 rtnl_link_unregister(&netkit_link_ops);
985}
986
987module_init(netkit_init);
988module_exit(netkit_exit);
989
990MODULE_DESCRIPTION("BPF-programmable network device");
991MODULE_AUTHOR("Daniel Borkmann <daniel@iogearbox.net>");
992MODULE_AUTHOR("Nikolay Aleksandrov <razor@blackwall.org>");
993MODULE_LICENSE("GPL");
994MODULE_ALIAS_RTNL_LINK(DRV_NAME);