1 // SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
2 // Copyright (C) 2018 Facebook
3
4 #ifndef _GNU_SOURCE
5 #define _GNU_SOURCE
6 #endif
7 #include <errno.h>
8 #include <fcntl.h>
9 #include <stdlib.h>
10 #include <string.h>
11 #include <time.h>
12 #include <unistd.h>
13 #include <bpf/bpf.h>
14 #include <bpf/libbpf.h>
15 #include <net/if.h>
16 #include <linux/rtnetlink.h>
17 #include <linux/socket.h>
18 #include <linux/tc_act/tc_bpf.h>
19 #include <sys/socket.h>
20 #include <sys/stat.h>
21 #include <sys/types.h>
22
23 #include "bpf/nlattr.h"
24 #include "main.h"
25 #include "netlink_dumper.h"
26
27 #ifndef SOL_NETLINK
28 #define SOL_NETLINK 270
29 #endif
30
31 struct ip_devname_ifindex {
32 char devname[64];
33 int ifindex;
34 };
35
36 struct bpf_netdev_t {
37 struct ip_devname_ifindex *devices;
38 int used_len;
39 int array_len;
40 int filter_idx;
41 };
42
43 struct tc_kind_handle {
44 char kind[64];
45 int handle;
46 };
47
48 struct bpf_tcinfo_t {
49 struct tc_kind_handle *handle_array;
50 int used_len;
51 int array_len;
52 bool is_qdisc;
53 };
54
55 struct bpf_filter_t {
56 const char *kind;
57 const char *devname;
58 int ifindex;
59 };
60
61 struct bpf_attach_info {
62 __u32 flow_dissector_id;
63 };
64
65 enum net_attach_type {
66 NET_ATTACH_TYPE_XDP,
67 NET_ATTACH_TYPE_XDP_GENERIC,
68 NET_ATTACH_TYPE_XDP_DRIVER,
69 NET_ATTACH_TYPE_XDP_OFFLOAD,
70 NET_ATTACH_TYPE_TCX_INGRESS,
71 NET_ATTACH_TYPE_TCX_EGRESS,
72 };
73
74 static const char * const attach_type_strings[] = {
75 [NET_ATTACH_TYPE_XDP] = "xdp",
76 [NET_ATTACH_TYPE_XDP_GENERIC] = "xdpgeneric",
77 [NET_ATTACH_TYPE_XDP_DRIVER] = "xdpdrv",
78 [NET_ATTACH_TYPE_XDP_OFFLOAD] = "xdpoffload",
79 [NET_ATTACH_TYPE_TCX_INGRESS] = "tcx_ingress",
80 [NET_ATTACH_TYPE_TCX_EGRESS] = "tcx_egress",
81 };
82
83 static const char * const attach_loc_strings[] = {
84 [BPF_TCX_INGRESS] = "tcx/ingress",
85 [BPF_TCX_EGRESS] = "tcx/egress",
86 [BPF_NETKIT_PRIMARY] = "netkit/primary",
87 [BPF_NETKIT_PEER] = "netkit/peer",
88 };
89
90 const size_t net_attach_type_size = ARRAY_SIZE(attach_type_strings);
91
parse_attach_type(const char * str)92 static enum net_attach_type parse_attach_type(const char *str)
93 {
94 enum net_attach_type type;
95
96 for (type = 0; type < net_attach_type_size; type++) {
97 if (attach_type_strings[type] &&
98 is_prefix(str, attach_type_strings[type]))
99 return type;
100 }
101
102 return net_attach_type_size;
103 }
104
105 typedef int (*dump_nlmsg_t)(void *cookie, void *msg, struct nlattr **tb);
106
107 typedef int (*__dump_nlmsg_t)(struct nlmsghdr *nlmsg, dump_nlmsg_t, void *cookie);
108
netlink_open(__u32 * nl_pid)109 static int netlink_open(__u32 *nl_pid)
110 {
111 struct sockaddr_nl sa;
112 socklen_t addrlen;
113 int one = 1, ret;
114 int sock;
115
116 memset(&sa, 0, sizeof(sa));
117 sa.nl_family = AF_NETLINK;
118
119 sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
120 if (sock < 0)
121 return -errno;
122
123 if (setsockopt(sock, SOL_NETLINK, NETLINK_EXT_ACK,
124 &one, sizeof(one)) < 0) {
125 p_err("Netlink error reporting not supported");
126 }
127
128 if (bind(sock, (struct sockaddr *)&sa, sizeof(sa)) < 0) {
129 ret = -errno;
130 goto cleanup;
131 }
132
133 addrlen = sizeof(sa);
134 if (getsockname(sock, (struct sockaddr *)&sa, &addrlen) < 0) {
135 ret = -errno;
136 goto cleanup;
137 }
138
139 if (addrlen != sizeof(sa)) {
140 ret = -LIBBPF_ERRNO__INTERNAL;
141 goto cleanup;
142 }
143
144 *nl_pid = sa.nl_pid;
145 return sock;
146
147 cleanup:
148 close(sock);
149 return ret;
150 }
151
netlink_recv(int sock,__u32 nl_pid,__u32 seq,__dump_nlmsg_t _fn,dump_nlmsg_t fn,void * cookie)152 static int netlink_recv(int sock, __u32 nl_pid, __u32 seq,
153 __dump_nlmsg_t _fn, dump_nlmsg_t fn,
154 void *cookie)
155 {
156 bool multipart = true;
157 struct nlmsgerr *err;
158 struct nlmsghdr *nh;
159 char buf[4096];
160 int len, ret;
161
162 while (multipart) {
163 multipart = false;
164 len = recv(sock, buf, sizeof(buf), 0);
165 if (len < 0) {
166 ret = -errno;
167 goto done;
168 }
169
170 if (len == 0)
171 break;
172
173 for (nh = (struct nlmsghdr *)buf; NLMSG_OK(nh, (unsigned int)len);
174 nh = NLMSG_NEXT(nh, len)) {
175 if (nh->nlmsg_pid != nl_pid) {
176 ret = -LIBBPF_ERRNO__WRNGPID;
177 goto done;
178 }
179 if (nh->nlmsg_seq != seq) {
180 ret = -LIBBPF_ERRNO__INVSEQ;
181 goto done;
182 }
183 if (nh->nlmsg_flags & NLM_F_MULTI)
184 multipart = true;
185 switch (nh->nlmsg_type) {
186 case NLMSG_ERROR:
187 err = (struct nlmsgerr *)NLMSG_DATA(nh);
188 if (!err->error)
189 continue;
190 ret = err->error;
191 libbpf_nla_dump_errormsg(nh);
192 goto done;
193 case NLMSG_DONE:
194 return 0;
195 default:
196 break;
197 }
198 if (_fn) {
199 ret = _fn(nh, fn, cookie);
200 if (ret)
201 return ret;
202 }
203 }
204 }
205 ret = 0;
206 done:
207 return ret;
208 }
209
__dump_class_nlmsg(struct nlmsghdr * nlh,dump_nlmsg_t dump_class_nlmsg,void * cookie)210 static int __dump_class_nlmsg(struct nlmsghdr *nlh,
211 dump_nlmsg_t dump_class_nlmsg,
212 void *cookie)
213 {
214 struct nlattr *tb[TCA_MAX + 1], *attr;
215 struct tcmsg *t = NLMSG_DATA(nlh);
216 int len;
217
218 len = nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*t));
219 attr = (struct nlattr *) ((void *) t + NLMSG_ALIGN(sizeof(*t)));
220 if (libbpf_nla_parse(tb, TCA_MAX, attr, len, NULL) != 0)
221 return -LIBBPF_ERRNO__NLPARSE;
222
223 return dump_class_nlmsg(cookie, t, tb);
224 }
225
netlink_get_class(int sock,unsigned int nl_pid,int ifindex,dump_nlmsg_t dump_class_nlmsg,void * cookie)226 static int netlink_get_class(int sock, unsigned int nl_pid, int ifindex,
227 dump_nlmsg_t dump_class_nlmsg, void *cookie)
228 {
229 struct {
230 struct nlmsghdr nlh;
231 struct tcmsg t;
232 } req = {
233 .nlh.nlmsg_len = NLMSG_LENGTH(sizeof(struct tcmsg)),
234 .nlh.nlmsg_type = RTM_GETTCLASS,
235 .nlh.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST,
236 .t.tcm_family = AF_UNSPEC,
237 .t.tcm_ifindex = ifindex,
238 };
239 int seq = time(NULL);
240
241 req.nlh.nlmsg_seq = seq;
242 if (send(sock, &req, req.nlh.nlmsg_len, 0) < 0)
243 return -errno;
244
245 return netlink_recv(sock, nl_pid, seq, __dump_class_nlmsg,
246 dump_class_nlmsg, cookie);
247 }
248
__dump_qdisc_nlmsg(struct nlmsghdr * nlh,dump_nlmsg_t dump_qdisc_nlmsg,void * cookie)249 static int __dump_qdisc_nlmsg(struct nlmsghdr *nlh,
250 dump_nlmsg_t dump_qdisc_nlmsg,
251 void *cookie)
252 {
253 struct nlattr *tb[TCA_MAX + 1], *attr;
254 struct tcmsg *t = NLMSG_DATA(nlh);
255 int len;
256
257 len = nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*t));
258 attr = (struct nlattr *) ((void *) t + NLMSG_ALIGN(sizeof(*t)));
259 if (libbpf_nla_parse(tb, TCA_MAX, attr, len, NULL) != 0)
260 return -LIBBPF_ERRNO__NLPARSE;
261
262 return dump_qdisc_nlmsg(cookie, t, tb);
263 }
264
netlink_get_qdisc(int sock,unsigned int nl_pid,int ifindex,dump_nlmsg_t dump_qdisc_nlmsg,void * cookie)265 static int netlink_get_qdisc(int sock, unsigned int nl_pid, int ifindex,
266 dump_nlmsg_t dump_qdisc_nlmsg, void *cookie)
267 {
268 struct {
269 struct nlmsghdr nlh;
270 struct tcmsg t;
271 } req = {
272 .nlh.nlmsg_len = NLMSG_LENGTH(sizeof(struct tcmsg)),
273 .nlh.nlmsg_type = RTM_GETQDISC,
274 .nlh.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST,
275 .t.tcm_family = AF_UNSPEC,
276 .t.tcm_ifindex = ifindex,
277 };
278 int seq = time(NULL);
279
280 req.nlh.nlmsg_seq = seq;
281 if (send(sock, &req, req.nlh.nlmsg_len, 0) < 0)
282 return -errno;
283
284 return netlink_recv(sock, nl_pid, seq, __dump_qdisc_nlmsg,
285 dump_qdisc_nlmsg, cookie);
286 }
287
__dump_filter_nlmsg(struct nlmsghdr * nlh,dump_nlmsg_t dump_filter_nlmsg,void * cookie)288 static int __dump_filter_nlmsg(struct nlmsghdr *nlh,
289 dump_nlmsg_t dump_filter_nlmsg,
290 void *cookie)
291 {
292 struct nlattr *tb[TCA_MAX + 1], *attr;
293 struct tcmsg *t = NLMSG_DATA(nlh);
294 int len;
295
296 len = nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*t));
297 attr = (struct nlattr *) ((void *) t + NLMSG_ALIGN(sizeof(*t)));
298 if (libbpf_nla_parse(tb, TCA_MAX, attr, len, NULL) != 0)
299 return -LIBBPF_ERRNO__NLPARSE;
300
301 return dump_filter_nlmsg(cookie, t, tb);
302 }
303
netlink_get_filter(int sock,unsigned int nl_pid,int ifindex,int handle,dump_nlmsg_t dump_filter_nlmsg,void * cookie)304 static int netlink_get_filter(int sock, unsigned int nl_pid, int ifindex, int handle,
305 dump_nlmsg_t dump_filter_nlmsg, void *cookie)
306 {
307 struct {
308 struct nlmsghdr nlh;
309 struct tcmsg t;
310 } req = {
311 .nlh.nlmsg_len = NLMSG_LENGTH(sizeof(struct tcmsg)),
312 .nlh.nlmsg_type = RTM_GETTFILTER,
313 .nlh.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST,
314 .t.tcm_family = AF_UNSPEC,
315 .t.tcm_ifindex = ifindex,
316 .t.tcm_parent = handle,
317 };
318 int seq = time(NULL);
319
320 req.nlh.nlmsg_seq = seq;
321 if (send(sock, &req, req.nlh.nlmsg_len, 0) < 0)
322 return -errno;
323
324 return netlink_recv(sock, nl_pid, seq, __dump_filter_nlmsg,
325 dump_filter_nlmsg, cookie);
326 }
327
__dump_link_nlmsg(struct nlmsghdr * nlh,dump_nlmsg_t dump_link_nlmsg,void * cookie)328 static int __dump_link_nlmsg(struct nlmsghdr *nlh,
329 dump_nlmsg_t dump_link_nlmsg, void *cookie)
330 {
331 struct nlattr *tb[IFLA_MAX + 1], *attr;
332 struct ifinfomsg *ifi = NLMSG_DATA(nlh);
333 int len;
334
335 len = nlh->nlmsg_len - NLMSG_LENGTH(sizeof(*ifi));
336 attr = (struct nlattr *) ((void *) ifi + NLMSG_ALIGN(sizeof(*ifi)));
337 if (libbpf_nla_parse(tb, IFLA_MAX, attr, len, NULL) != 0)
338 return -LIBBPF_ERRNO__NLPARSE;
339
340 return dump_link_nlmsg(cookie, ifi, tb);
341 }
342
netlink_get_link(int sock,unsigned int nl_pid,dump_nlmsg_t dump_link_nlmsg,void * cookie)343 static int netlink_get_link(int sock, unsigned int nl_pid,
344 dump_nlmsg_t dump_link_nlmsg, void *cookie)
345 {
346 struct {
347 struct nlmsghdr nlh;
348 struct ifinfomsg ifm;
349 } req = {
350 .nlh.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
351 .nlh.nlmsg_type = RTM_GETLINK,
352 .nlh.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST,
353 .ifm.ifi_family = AF_PACKET,
354 };
355 int seq = time(NULL);
356
357 req.nlh.nlmsg_seq = seq;
358 if (send(sock, &req, req.nlh.nlmsg_len, 0) < 0)
359 return -errno;
360
361 return netlink_recv(sock, nl_pid, seq, __dump_link_nlmsg,
362 dump_link_nlmsg, cookie);
363 }
364
dump_link_nlmsg(void * cookie,void * msg,struct nlattr ** tb)365 static int dump_link_nlmsg(void *cookie, void *msg, struct nlattr **tb)
366 {
367 struct bpf_netdev_t *netinfo = cookie;
368 struct ifinfomsg *ifinfo = msg;
369 struct ip_devname_ifindex *tmp;
370
371 if (netinfo->filter_idx > 0 && netinfo->filter_idx != ifinfo->ifi_index)
372 return 0;
373
374 if (netinfo->used_len == netinfo->array_len) {
375 tmp = realloc(netinfo->devices,
376 (netinfo->array_len + 16) * sizeof(struct ip_devname_ifindex));
377 if (!tmp)
378 return -ENOMEM;
379
380 netinfo->devices = tmp;
381 netinfo->array_len += 16;
382 }
383 netinfo->devices[netinfo->used_len].ifindex = ifinfo->ifi_index;
384 snprintf(netinfo->devices[netinfo->used_len].devname,
385 sizeof(netinfo->devices[netinfo->used_len].devname),
386 "%s",
387 tb[IFLA_IFNAME]
388 ? libbpf_nla_getattr_str(tb[IFLA_IFNAME])
389 : "");
390 netinfo->used_len++;
391
392 return do_xdp_dump(ifinfo, tb);
393 }
394
dump_class_qdisc_nlmsg(void * cookie,void * msg,struct nlattr ** tb)395 static int dump_class_qdisc_nlmsg(void *cookie, void *msg, struct nlattr **tb)
396 {
397 struct bpf_tcinfo_t *tcinfo = cookie;
398 struct tcmsg *info = msg;
399 struct tc_kind_handle *tmp;
400
401 if (tcinfo->is_qdisc) {
402 /* skip clsact qdisc */
403 if (tb[TCA_KIND] &&
404 strcmp(libbpf_nla_data(tb[TCA_KIND]), "clsact") == 0)
405 return 0;
406 if (info->tcm_handle == 0)
407 return 0;
408 }
409
410 if (tcinfo->used_len == tcinfo->array_len) {
411 tmp = realloc(tcinfo->handle_array,
412 (tcinfo->array_len + 16) * sizeof(struct tc_kind_handle));
413 if (!tmp)
414 return -ENOMEM;
415
416 tcinfo->handle_array = tmp;
417 tcinfo->array_len += 16;
418 }
419 tcinfo->handle_array[tcinfo->used_len].handle = info->tcm_handle;
420 snprintf(tcinfo->handle_array[tcinfo->used_len].kind,
421 sizeof(tcinfo->handle_array[tcinfo->used_len].kind),
422 "%s",
423 tb[TCA_KIND]
424 ? libbpf_nla_getattr_str(tb[TCA_KIND])
425 : "unknown");
426 tcinfo->used_len++;
427
428 return 0;
429 }
430
dump_filter_nlmsg(void * cookie,void * msg,struct nlattr ** tb)431 static int dump_filter_nlmsg(void *cookie, void *msg, struct nlattr **tb)
432 {
433 const struct bpf_filter_t *filter_info = cookie;
434
435 return do_filter_dump((struct tcmsg *)msg, tb, filter_info->kind,
436 filter_info->devname, filter_info->ifindex);
437 }
438
__show_dev_tc_bpf_name(__u32 id,char * name,size_t len)439 static int __show_dev_tc_bpf_name(__u32 id, char *name, size_t len)
440 {
441 struct bpf_prog_info info = {};
442 __u32 ilen = sizeof(info);
443 int fd, ret;
444
445 fd = bpf_prog_get_fd_by_id(id);
446 if (fd < 0)
447 return fd;
448 ret = bpf_obj_get_info_by_fd(fd, &info, &ilen);
449 if (ret < 0)
450 goto out;
451 ret = -ENOENT;
452 if (info.name[0]) {
453 get_prog_full_name(&info, fd, name, len);
454 ret = 0;
455 }
456 out:
457 close(fd);
458 return ret;
459 }
460
__show_dev_tc_bpf(const struct ip_devname_ifindex * dev,const enum bpf_attach_type loc)461 static void __show_dev_tc_bpf(const struct ip_devname_ifindex *dev,
462 const enum bpf_attach_type loc)
463 {
464 __u32 prog_flags[64] = {}, link_flags[64] = {}, i, j;
465 __u32 prog_ids[64] = {}, link_ids[64] = {};
466 LIBBPF_OPTS(bpf_prog_query_opts, optq);
467 char prog_name[MAX_PROG_FULL_NAME];
468 int ret;
469
470 optq.prog_ids = prog_ids;
471 optq.prog_attach_flags = prog_flags;
472 optq.link_ids = link_ids;
473 optq.link_attach_flags = link_flags;
474 optq.count = ARRAY_SIZE(prog_ids);
475
476 ret = bpf_prog_query_opts(dev->ifindex, loc, &optq);
477 if (ret)
478 return;
479 for (i = 0; i < optq.count; i++) {
480 NET_START_OBJECT;
481 NET_DUMP_STR("devname", "%s", dev->devname);
482 NET_DUMP_UINT("ifindex", "(%u)", (unsigned int)dev->ifindex);
483 NET_DUMP_STR("kind", " %s", attach_loc_strings[loc]);
484 ret = __show_dev_tc_bpf_name(prog_ids[i], prog_name,
485 sizeof(prog_name));
486 if (!ret)
487 NET_DUMP_STR("name", " %s", prog_name);
488 NET_DUMP_UINT("prog_id", " prog_id %u ", prog_ids[i]);
489 if (prog_flags[i] || json_output) {
490 NET_START_ARRAY("prog_flags", "%s ");
491 for (j = 0; prog_flags[i] && j < 32; j++) {
492 if (!(prog_flags[i] & (1U << j)))
493 continue;
494 NET_DUMP_UINT_ONLY(1U << j);
495 }
496 NET_END_ARRAY("");
497 }
498 if (link_ids[i] || json_output) {
499 NET_DUMP_UINT("link_id", "link_id %u ", link_ids[i]);
500 if (link_flags[i] || json_output) {
501 NET_START_ARRAY("link_flags", "%s ");
502 for (j = 0; link_flags[i] && j < 32; j++) {
503 if (!(link_flags[i] & (1U << j)))
504 continue;
505 NET_DUMP_UINT_ONLY(1U << j);
506 }
507 NET_END_ARRAY("");
508 }
509 }
510 NET_END_OBJECT_FINAL;
511 }
512 }
513
show_dev_tc_bpf(struct ip_devname_ifindex * dev)514 static void show_dev_tc_bpf(struct ip_devname_ifindex *dev)
515 {
516 __show_dev_tc_bpf(dev, BPF_TCX_INGRESS);
517 __show_dev_tc_bpf(dev, BPF_TCX_EGRESS);
518
519 __show_dev_tc_bpf(dev, BPF_NETKIT_PRIMARY);
520 __show_dev_tc_bpf(dev, BPF_NETKIT_PEER);
521 }
522
show_dev_tc_bpf_classic(int sock,unsigned int nl_pid,struct ip_devname_ifindex * dev)523 static int show_dev_tc_bpf_classic(int sock, unsigned int nl_pid,
524 struct ip_devname_ifindex *dev)
525 {
526 struct bpf_filter_t filter_info;
527 struct bpf_tcinfo_t tcinfo;
528 int i, handle, ret = 0;
529
530 tcinfo.handle_array = NULL;
531 tcinfo.used_len = 0;
532 tcinfo.array_len = 0;
533
534 tcinfo.is_qdisc = false;
535 ret = netlink_get_class(sock, nl_pid, dev->ifindex,
536 dump_class_qdisc_nlmsg, &tcinfo);
537 if (ret)
538 goto out;
539
540 tcinfo.is_qdisc = true;
541 ret = netlink_get_qdisc(sock, nl_pid, dev->ifindex,
542 dump_class_qdisc_nlmsg, &tcinfo);
543 if (ret)
544 goto out;
545
546 filter_info.devname = dev->devname;
547 filter_info.ifindex = dev->ifindex;
548 for (i = 0; i < tcinfo.used_len; i++) {
549 filter_info.kind = tcinfo.handle_array[i].kind;
550 ret = netlink_get_filter(sock, nl_pid, dev->ifindex,
551 tcinfo.handle_array[i].handle,
552 dump_filter_nlmsg, &filter_info);
553 if (ret)
554 goto out;
555 }
556
557 /* root, ingress and egress handle */
558 handle = TC_H_ROOT;
559 filter_info.kind = "root";
560 ret = netlink_get_filter(sock, nl_pid, dev->ifindex, handle,
561 dump_filter_nlmsg, &filter_info);
562 if (ret)
563 goto out;
564
565 handle = TC_H_MAKE(TC_H_CLSACT, TC_H_MIN_INGRESS);
566 filter_info.kind = "clsact/ingress";
567 ret = netlink_get_filter(sock, nl_pid, dev->ifindex, handle,
568 dump_filter_nlmsg, &filter_info);
569 if (ret)
570 goto out;
571
572 handle = TC_H_MAKE(TC_H_CLSACT, TC_H_MIN_EGRESS);
573 filter_info.kind = "clsact/egress";
574 ret = netlink_get_filter(sock, nl_pid, dev->ifindex, handle,
575 dump_filter_nlmsg, &filter_info);
576 if (ret)
577 goto out;
578
579 out:
580 free(tcinfo.handle_array);
581 return 0;
582 }
583
query_flow_dissector(struct bpf_attach_info * attach_info)584 static int query_flow_dissector(struct bpf_attach_info *attach_info)
585 {
586 __u32 attach_flags;
587 __u32 prog_ids[1];
588 __u32 prog_cnt;
589 int err;
590 int fd;
591
592 fd = open("/proc/self/ns/net", O_RDONLY);
593 if (fd < 0) {
594 p_err("can't open /proc/self/ns/net: %s",
595 strerror(errno));
596 return -1;
597 }
598 prog_cnt = ARRAY_SIZE(prog_ids);
599 err = bpf_prog_query(fd, BPF_FLOW_DISSECTOR, 0,
600 &attach_flags, prog_ids, &prog_cnt);
601 close(fd);
602 if (err) {
603 if (errno == EINVAL) {
604 /* Older kernel's don't support querying
605 * flow dissector programs.
606 */
607 errno = 0;
608 return 0;
609 }
610 p_err("can't query prog: %s", strerror(errno));
611 return -1;
612 }
613
614 if (prog_cnt == 1)
615 attach_info->flow_dissector_id = prog_ids[0];
616
617 return 0;
618 }
619
net_parse_dev(int * argc,char *** argv)620 static int net_parse_dev(int *argc, char ***argv)
621 {
622 int ifindex;
623
624 if (is_prefix(**argv, "dev")) {
625 NEXT_ARGP();
626
627 ifindex = if_nametoindex(**argv);
628 if (!ifindex)
629 p_err("invalid devname %s", **argv);
630
631 NEXT_ARGP();
632 } else {
633 p_err("expected 'dev', got: '%s'?", **argv);
634 return -1;
635 }
636
637 return ifindex;
638 }
639
do_attach_detach_xdp(int progfd,enum net_attach_type attach_type,int ifindex,bool overwrite)640 static int do_attach_detach_xdp(int progfd, enum net_attach_type attach_type,
641 int ifindex, bool overwrite)
642 {
643 __u32 flags = 0;
644
645 if (!overwrite)
646 flags = XDP_FLAGS_UPDATE_IF_NOEXIST;
647 if (attach_type == NET_ATTACH_TYPE_XDP_GENERIC)
648 flags |= XDP_FLAGS_SKB_MODE;
649 if (attach_type == NET_ATTACH_TYPE_XDP_DRIVER)
650 flags |= XDP_FLAGS_DRV_MODE;
651 if (attach_type == NET_ATTACH_TYPE_XDP_OFFLOAD)
652 flags |= XDP_FLAGS_HW_MODE;
653
654 return bpf_xdp_attach(ifindex, progfd, flags, NULL);
655 }
656
get_tcx_type(enum net_attach_type attach_type)657 static int get_tcx_type(enum net_attach_type attach_type)
658 {
659 switch (attach_type) {
660 case NET_ATTACH_TYPE_TCX_INGRESS:
661 return BPF_TCX_INGRESS;
662 case NET_ATTACH_TYPE_TCX_EGRESS:
663 return BPF_TCX_EGRESS;
664 default:
665 return -1;
666 }
667 }
668
do_attach_tcx(int progfd,enum net_attach_type attach_type,int ifindex)669 static int do_attach_tcx(int progfd, enum net_attach_type attach_type, int ifindex)
670 {
671 int type = get_tcx_type(attach_type);
672
673 return bpf_prog_attach(progfd, ifindex, type, 0);
674 }
675
do_detach_tcx(int targetfd,enum net_attach_type attach_type)676 static int do_detach_tcx(int targetfd, enum net_attach_type attach_type)
677 {
678 int type = get_tcx_type(attach_type);
679
680 return bpf_prog_detach(targetfd, type);
681 }
682
do_attach(int argc,char ** argv)683 static int do_attach(int argc, char **argv)
684 {
685 enum net_attach_type attach_type;
686 int progfd, ifindex, err = 0;
687 bool overwrite = false;
688
689 /* parse attach args */
690 if (!REQ_ARGS(5))
691 return -EINVAL;
692
693 attach_type = parse_attach_type(*argv);
694 if (attach_type == net_attach_type_size) {
695 p_err("invalid net attach/detach type: %s", *argv);
696 return -EINVAL;
697 }
698 NEXT_ARG();
699
700 progfd = prog_parse_fd(&argc, &argv);
701 if (progfd < 0)
702 return -EINVAL;
703
704 ifindex = net_parse_dev(&argc, &argv);
705 if (ifindex < 1) {
706 err = -EINVAL;
707 goto cleanup;
708 }
709
710 if (argc) {
711 if (is_prefix(*argv, "overwrite")) {
712 overwrite = true;
713 } else {
714 p_err("expected 'overwrite', got: '%s'?", *argv);
715 err = -EINVAL;
716 goto cleanup;
717 }
718 }
719
720 switch (attach_type) {
721 /* attach xdp prog */
722 case NET_ATTACH_TYPE_XDP:
723 case NET_ATTACH_TYPE_XDP_GENERIC:
724 case NET_ATTACH_TYPE_XDP_DRIVER:
725 case NET_ATTACH_TYPE_XDP_OFFLOAD:
726 err = do_attach_detach_xdp(progfd, attach_type, ifindex, overwrite);
727 break;
728 /* attach tcx prog */
729 case NET_ATTACH_TYPE_TCX_INGRESS:
730 case NET_ATTACH_TYPE_TCX_EGRESS:
731 err = do_attach_tcx(progfd, attach_type, ifindex);
732 break;
733 default:
734 break;
735 }
736
737 if (err) {
738 p_err("interface %s attach failed: %s",
739 attach_type_strings[attach_type], strerror(-err));
740 goto cleanup;
741 }
742
743 if (json_output)
744 jsonw_null(json_wtr);
745 cleanup:
746 close(progfd);
747 return err;
748 }
749
do_detach(int argc,char ** argv)750 static int do_detach(int argc, char **argv)
751 {
752 enum net_attach_type attach_type;
753 int progfd, ifindex, err = 0;
754
755 /* parse detach args */
756 if (!REQ_ARGS(3))
757 return -EINVAL;
758
759 attach_type = parse_attach_type(*argv);
760 if (attach_type == net_attach_type_size) {
761 p_err("invalid net attach/detach type: %s", *argv);
762 return -EINVAL;
763 }
764 NEXT_ARG();
765
766 ifindex = net_parse_dev(&argc, &argv);
767 if (ifindex < 1)
768 return -EINVAL;
769
770 switch (attach_type) {
771 /* detach xdp prog */
772 case NET_ATTACH_TYPE_XDP:
773 case NET_ATTACH_TYPE_XDP_GENERIC:
774 case NET_ATTACH_TYPE_XDP_DRIVER:
775 case NET_ATTACH_TYPE_XDP_OFFLOAD:
776 progfd = -1;
777 err = do_attach_detach_xdp(progfd, attach_type, ifindex, NULL);
778 break;
779 /* detach tcx prog */
780 case NET_ATTACH_TYPE_TCX_INGRESS:
781 case NET_ATTACH_TYPE_TCX_EGRESS:
782 err = do_detach_tcx(ifindex, attach_type);
783 break;
784 default:
785 break;
786 }
787
788 if (err < 0) {
789 p_err("interface %s detach failed: %s",
790 attach_type_strings[attach_type], strerror(-err));
791 return err;
792 }
793
794 if (json_output)
795 jsonw_null(json_wtr);
796
797 return 0;
798 }
799
netfilter_link_compar(const void * a,const void * b)800 static int netfilter_link_compar(const void *a, const void *b)
801 {
802 const struct bpf_link_info *nfa = a;
803 const struct bpf_link_info *nfb = b;
804 int delta;
805
806 delta = nfa->netfilter.pf - nfb->netfilter.pf;
807 if (delta)
808 return delta;
809
810 delta = nfa->netfilter.hooknum - nfb->netfilter.hooknum;
811 if (delta)
812 return delta;
813
814 if (nfa->netfilter.priority < nfb->netfilter.priority)
815 return -1;
816 if (nfa->netfilter.priority > nfb->netfilter.priority)
817 return 1;
818
819 return nfa->netfilter.flags - nfb->netfilter.flags;
820 }
821
show_link_netfilter(void)822 static void show_link_netfilter(void)
823 {
824 unsigned int nf_link_len = 0, nf_link_count = 0;
825 struct bpf_link_info *nf_link_info = NULL;
826 __u32 id = 0;
827
828 while (true) {
829 struct bpf_link_info info;
830 int fd, err;
831 __u32 len;
832
833 err = bpf_link_get_next_id(id, &id);
834 if (err) {
835 if (errno == ENOENT)
836 break;
837 p_err("can't get next link: %s (id %u)", strerror(errno), id);
838 break;
839 }
840
841 fd = bpf_link_get_fd_by_id(id);
842 if (fd < 0) {
843 p_err("can't get link by id (%u): %s", id, strerror(errno));
844 continue;
845 }
846
847 memset(&info, 0, sizeof(info));
848 len = sizeof(info);
849
850 err = bpf_link_get_info_by_fd(fd, &info, &len);
851
852 close(fd);
853
854 if (err) {
855 p_err("can't get link info for fd %d: %s", fd, strerror(errno));
856 continue;
857 }
858
859 if (info.type != BPF_LINK_TYPE_NETFILTER)
860 continue;
861
862 if (nf_link_count >= nf_link_len) {
863 static const unsigned int max_link_count = INT_MAX / sizeof(info);
864 struct bpf_link_info *expand;
865
866 if (nf_link_count > max_link_count) {
867 p_err("cannot handle more than %u links\n", max_link_count);
868 break;
869 }
870
871 nf_link_len += 16;
872
873 expand = realloc(nf_link_info, nf_link_len * sizeof(info));
874 if (!expand) {
875 p_err("realloc: %s", strerror(errno));
876 break;
877 }
878
879 nf_link_info = expand;
880 }
881
882 nf_link_info[nf_link_count] = info;
883 nf_link_count++;
884 }
885
886 if (!nf_link_info)
887 return;
888
889 qsort(nf_link_info, nf_link_count, sizeof(*nf_link_info), netfilter_link_compar);
890
891 for (id = 0; id < nf_link_count; id++) {
892 NET_START_OBJECT;
893 if (json_output)
894 netfilter_dump_json(&nf_link_info[id], json_wtr);
895 else
896 netfilter_dump_plain(&nf_link_info[id]);
897
898 NET_DUMP_UINT("id", " prog_id %u", nf_link_info[id].prog_id);
899 NET_END_OBJECT;
900 }
901
902 free(nf_link_info);
903 }
904
do_show(int argc,char ** argv)905 static int do_show(int argc, char **argv)
906 {
907 struct bpf_attach_info attach_info = {};
908 int i, sock, ret, filter_idx = -1;
909 struct bpf_netdev_t dev_array;
910 unsigned int nl_pid = 0;
911 char err_buf[256];
912
913 if (argc == 2) {
914 filter_idx = net_parse_dev(&argc, &argv);
915 if (filter_idx < 1)
916 return -1;
917 } else if (argc != 0) {
918 usage();
919 }
920
921 ret = query_flow_dissector(&attach_info);
922 if (ret)
923 return -1;
924
925 sock = netlink_open(&nl_pid);
926 if (sock < 0) {
927 fprintf(stderr, "failed to open netlink sock\n");
928 return -1;
929 }
930
931 dev_array.devices = NULL;
932 dev_array.used_len = 0;
933 dev_array.array_len = 0;
934 dev_array.filter_idx = filter_idx;
935
936 if (json_output)
937 jsonw_start_array(json_wtr);
938 NET_START_OBJECT;
939 NET_START_ARRAY("xdp", "%s:\n");
940 ret = netlink_get_link(sock, nl_pid, dump_link_nlmsg, &dev_array);
941 NET_END_ARRAY("\n");
942
943 if (!ret) {
944 NET_START_ARRAY("tc", "%s:\n");
945 for (i = 0; i < dev_array.used_len; i++) {
946 show_dev_tc_bpf(&dev_array.devices[i]);
947 ret = show_dev_tc_bpf_classic(sock, nl_pid,
948 &dev_array.devices[i]);
949 if (ret)
950 break;
951 }
952 NET_END_ARRAY("\n");
953 }
954
955 NET_START_ARRAY("flow_dissector", "%s:\n");
956 if (attach_info.flow_dissector_id > 0)
957 NET_DUMP_UINT("id", "id %u", attach_info.flow_dissector_id);
958 NET_END_ARRAY("\n");
959
960 NET_START_ARRAY("netfilter", "%s:\n");
961 show_link_netfilter();
962 NET_END_ARRAY("\n");
963
964 NET_END_OBJECT;
965 if (json_output)
966 jsonw_end_array(json_wtr);
967
968 if (ret) {
969 if (json_output)
970 jsonw_null(json_wtr);
971 libbpf_strerror(ret, err_buf, sizeof(err_buf));
972 fprintf(stderr, "Error: %s\n", err_buf);
973 }
974 free(dev_array.devices);
975 close(sock);
976 return ret;
977 }
978
do_help(int argc,char ** argv)979 static int do_help(int argc, char **argv)
980 {
981 if (json_output) {
982 jsonw_null(json_wtr);
983 return 0;
984 }
985
986 fprintf(stderr,
987 "Usage: %1$s %2$s { show | list } [dev <devname>]\n"
988 " %1$s %2$s attach ATTACH_TYPE PROG dev <devname> [ overwrite ]\n"
989 " %1$s %2$s detach ATTACH_TYPE dev <devname>\n"
990 " %1$s %2$s help\n"
991 "\n"
992 " " HELP_SPEC_PROGRAM "\n"
993 " ATTACH_TYPE := { xdp | xdpgeneric | xdpdrv | xdpoffload | tcx_ingress\n"
994 " | tcx_egress }\n"
995 " " HELP_SPEC_OPTIONS " }\n"
996 "\n"
997 "Note: Only xdp, tcx, tc, netkit, flow_dissector and netfilter attachments\n"
998 " are currently supported.\n"
999 " For progs attached to cgroups, use \"bpftool cgroup\"\n"
1000 " to dump program attachments. For program types\n"
1001 " sk_{filter,skb,msg,reuseport} and lwt/seg6, please\n"
1002 " consult iproute2.\n"
1003 "",
1004 bin_name, argv[-2]);
1005
1006 return 0;
1007 }
1008
1009 static const struct cmd cmds[] = {
1010 { "show", do_show },
1011 { "list", do_show },
1012 { "attach", do_attach },
1013 { "detach", do_detach },
1014 { "help", do_help },
1015 { 0 }
1016 };
1017
do_net(int argc,char ** argv)1018 int do_net(int argc, char **argv)
1019 {
1020 return cmd_select(cmds, argc, argv, do_help);
1021 }
1022