1 // SPDX-License-Identifier: GPL-2.0
2 // Copyright (c) 2020 Cloudflare
3 #include <error.h>
4 #include <linux/tcp.h>
5 #include <linux/socket.h>
6 #include <sys/epoll.h>
7
8 #include "test_progs.h"
9 #include "test_skmsg_load_helpers.skel.h"
10 #include "test_sockmap_update.skel.h"
11 #include "test_sockmap_invalid_update.skel.h"
12 #include "test_sockmap_skb_verdict_attach.skel.h"
13 #include "test_sockmap_progs_query.skel.h"
14 #include "test_sockmap_pass_prog.skel.h"
15 #include "test_sockmap_drop_prog.skel.h"
16 #include "test_sockmap_change_tail.skel.h"
17 #include "bpf_iter_sockmap.skel.h"
18
19 #include "sockmap_helpers.h"
20
21 #define TCP_REPAIR 19 /* TCP sock is under repair right now */
22
23 #define TCP_REPAIR_ON 1
24 #define TCP_REPAIR_OFF_NO_WP -1 /* Turn off without window probes */
25
26 /**
27 * SOL_TCP is defined in <netinet/tcp.h> (glibc), but the copybuf_address
28 * field of tcp_zerocopy_receive is not yet included in older versions.
29 * This workaround remains necessary until the glibc update propagates.
30 */
31 #ifndef SOL_TCP
32 #define SOL_TCP 6
33 #endif
34
connected_socket_v4(void)35 static int connected_socket_v4(void)
36 {
37 struct sockaddr_in addr = {
38 .sin_family = AF_INET,
39 .sin_port = htons(80),
40 .sin_addr = { inet_addr("127.0.0.1") },
41 };
42 socklen_t len = sizeof(addr);
43 int s, repair, err;
44
45 s = socket(AF_INET, SOCK_STREAM, 0);
46 if (!ASSERT_GE(s, 0, "socket"))
47 goto error;
48
49 repair = TCP_REPAIR_ON;
50 err = setsockopt(s, SOL_TCP, TCP_REPAIR, &repair, sizeof(repair));
51 if (!ASSERT_OK(err, "setsockopt(TCP_REPAIR)"))
52 goto error;
53
54 err = connect(s, (struct sockaddr *)&addr, len);
55 if (!ASSERT_OK(err, "connect"))
56 goto error;
57
58 repair = TCP_REPAIR_OFF_NO_WP;
59 err = setsockopt(s, SOL_TCP, TCP_REPAIR, &repair, sizeof(repair));
60 if (!ASSERT_OK(err, "setsockopt(TCP_REPAIR)"))
61 goto error;
62
63 return s;
64 error:
65 perror(__func__);
66 close(s);
67 return -1;
68 }
69
compare_cookies(struct bpf_map * src,struct bpf_map * dst)70 static void compare_cookies(struct bpf_map *src, struct bpf_map *dst)
71 {
72 __u32 i, max_entries = bpf_map__max_entries(src);
73 int err, src_fd, dst_fd;
74
75 src_fd = bpf_map__fd(src);
76 dst_fd = bpf_map__fd(dst);
77
78 for (i = 0; i < max_entries; i++) {
79 __u64 src_cookie, dst_cookie;
80
81 err = bpf_map_lookup_elem(src_fd, &i, &src_cookie);
82 if (err && errno == ENOENT) {
83 err = bpf_map_lookup_elem(dst_fd, &i, &dst_cookie);
84 ASSERT_ERR(err, "map_lookup_elem(dst)");
85 ASSERT_EQ(errno, ENOENT, "map_lookup_elem(dst)");
86 continue;
87 }
88 if (!ASSERT_OK(err, "lookup_elem(src)"))
89 continue;
90
91 err = bpf_map_lookup_elem(dst_fd, &i, &dst_cookie);
92 if (!ASSERT_OK(err, "lookup_elem(dst)"))
93 continue;
94
95 ASSERT_EQ(dst_cookie, src_cookie, "cookie mismatch");
96 }
97 }
98
99 /* Create a map, populate it with one socket, and free the map. */
test_sockmap_create_update_free(enum bpf_map_type map_type)100 static void test_sockmap_create_update_free(enum bpf_map_type map_type)
101 {
102 const int zero = 0;
103 int s, map, err;
104
105 s = connected_socket_v4();
106 if (!ASSERT_GE(s, 0, "connected_socket_v4"))
107 return;
108
109 map = bpf_map_create(map_type, NULL, sizeof(int), sizeof(int), 1, NULL);
110 if (!ASSERT_GE(map, 0, "bpf_map_create"))
111 goto out;
112
113 err = bpf_map_update_elem(map, &zero, &s, BPF_NOEXIST);
114 if (!ASSERT_OK(err, "bpf_map_update"))
115 goto out;
116
117 out:
118 close(map);
119 close(s);
120 }
121
test_sockmap_vsock_delete_on_close(void)122 static void test_sockmap_vsock_delete_on_close(void)
123 {
124 int map, c, p, err, zero = 0;
125
126 map = bpf_map_create(BPF_MAP_TYPE_SOCKMAP, NULL, sizeof(int),
127 sizeof(int), 1, NULL);
128 if (!ASSERT_OK_FD(map, "bpf_map_create"))
129 return;
130
131 err = create_pair(AF_VSOCK, SOCK_STREAM, &c, &p);
132 if (!ASSERT_OK(err, "create_pair"))
133 goto close_map;
134
135 if (xbpf_map_update_elem(map, &zero, &c, BPF_NOEXIST))
136 goto close_socks;
137
138 xclose(c);
139 xclose(p);
140
141 err = create_pair(AF_VSOCK, SOCK_STREAM, &c, &p);
142 if (!ASSERT_OK(err, "create_pair"))
143 goto close_map;
144
145 err = bpf_map_update_elem(map, &zero, &c, BPF_NOEXIST);
146 ASSERT_OK(err, "after close(), bpf_map_update");
147
148 close_socks:
149 xclose(c);
150 xclose(p);
151 close_map:
152 xclose(map);
153 }
154
test_skmsg_helpers(enum bpf_map_type map_type)155 static void test_skmsg_helpers(enum bpf_map_type map_type)
156 {
157 struct test_skmsg_load_helpers *skel;
158 int err, map, verdict;
159
160 skel = test_skmsg_load_helpers__open_and_load();
161 if (!ASSERT_OK_PTR(skel, "test_skmsg_load_helpers__open_and_load"))
162 return;
163
164 verdict = bpf_program__fd(skel->progs.prog_msg_verdict);
165 map = bpf_map__fd(skel->maps.sock_map);
166
167 err = bpf_prog_attach(verdict, map, BPF_SK_MSG_VERDICT, 0);
168 if (!ASSERT_OK(err, "bpf_prog_attach"))
169 goto out;
170
171 err = bpf_prog_detach2(verdict, map, BPF_SK_MSG_VERDICT);
172 if (!ASSERT_OK(err, "bpf_prog_detach2"))
173 goto out;
174 out:
175 test_skmsg_load_helpers__destroy(skel);
176 }
177
test_skmsg_helpers_with_link(enum bpf_map_type map_type)178 static void test_skmsg_helpers_with_link(enum bpf_map_type map_type)
179 {
180 struct bpf_program *prog, *prog_clone, *prog_clone2;
181 DECLARE_LIBBPF_OPTS(bpf_link_update_opts, opts);
182 struct test_skmsg_load_helpers *skel;
183 struct bpf_link *link, *link2;
184 int err, map;
185
186 skel = test_skmsg_load_helpers__open_and_load();
187 if (!ASSERT_OK_PTR(skel, "test_skmsg_load_helpers__open_and_load"))
188 return;
189
190 prog = skel->progs.prog_msg_verdict;
191 prog_clone = skel->progs.prog_msg_verdict_clone;
192 prog_clone2 = skel->progs.prog_msg_verdict_clone2;
193 map = bpf_map__fd(skel->maps.sock_map);
194
195 link = bpf_program__attach_sockmap(prog, map);
196 if (!ASSERT_OK_PTR(link, "bpf_program__attach_sockmap"))
197 goto out;
198
199 /* Fail since bpf_link for the same prog has been created. */
200 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_MSG_VERDICT, 0);
201 if (!ASSERT_ERR(err, "bpf_prog_attach"))
202 goto out;
203
204 /* Fail since bpf_link for the same prog type has been created. */
205 link2 = bpf_program__attach_sockmap(prog_clone, map);
206 if (!ASSERT_ERR_PTR(link2, "bpf_program__attach_sockmap")) {
207 bpf_link__destroy(link2);
208 goto out;
209 }
210
211 err = bpf_link__update_program(link, prog_clone);
212 if (!ASSERT_OK(err, "bpf_link__update_program"))
213 goto out;
214
215 /* Fail since a prog with different type attempts to do update. */
216 err = bpf_link__update_program(link, skel->progs.prog_skb_verdict);
217 if (!ASSERT_ERR(err, "bpf_link__update_program"))
218 goto out;
219
220 /* Fail since the old prog does not match the one in the kernel. */
221 opts.old_prog_fd = bpf_program__fd(prog_clone2);
222 opts.flags = BPF_F_REPLACE;
223 err = bpf_link_update(bpf_link__fd(link), bpf_program__fd(prog), &opts);
224 if (!ASSERT_ERR(err, "bpf_link_update"))
225 goto out;
226
227 opts.old_prog_fd = bpf_program__fd(prog_clone);
228 opts.flags = BPF_F_REPLACE;
229 err = bpf_link_update(bpf_link__fd(link), bpf_program__fd(prog), &opts);
230 if (!ASSERT_OK(err, "bpf_link_update"))
231 goto out;
232 out:
233 bpf_link__destroy(link);
234 test_skmsg_load_helpers__destroy(skel);
235 }
236
test_sockmap_update(enum bpf_map_type map_type)237 static void test_sockmap_update(enum bpf_map_type map_type)
238 {
239 int err, prog, src;
240 struct test_sockmap_update *skel;
241 struct bpf_map *dst_map;
242 const __u32 zero = 0;
243 char dummy[14] = {0};
244 LIBBPF_OPTS(bpf_test_run_opts, topts,
245 .data_in = dummy,
246 .data_size_in = sizeof(dummy),
247 .repeat = 1,
248 );
249 __s64 sk;
250
251 sk = connected_socket_v4();
252 if (!ASSERT_NEQ(sk, -1, "connected_socket_v4"))
253 return;
254
255 skel = test_sockmap_update__open_and_load();
256 if (!ASSERT_OK_PTR(skel, "open_and_load"))
257 goto close_sk;
258
259 prog = bpf_program__fd(skel->progs.copy_sock_map);
260 src = bpf_map__fd(skel->maps.src);
261 if (map_type == BPF_MAP_TYPE_SOCKMAP)
262 dst_map = skel->maps.dst_sock_map;
263 else
264 dst_map = skel->maps.dst_sock_hash;
265
266 err = bpf_map_update_elem(src, &zero, &sk, BPF_NOEXIST);
267 if (!ASSERT_OK(err, "update_elem(src)"))
268 goto out;
269
270 err = bpf_prog_test_run_opts(prog, &topts);
271 if (!ASSERT_OK(err, "test_run"))
272 goto out;
273 if (!ASSERT_NEQ(topts.retval, 0, "test_run retval"))
274 goto out;
275
276 compare_cookies(skel->maps.src, dst_map);
277
278 out:
279 test_sockmap_update__destroy(skel);
280 close_sk:
281 close(sk);
282 }
283
test_sockmap_invalid_update(void)284 static void test_sockmap_invalid_update(void)
285 {
286 struct test_sockmap_invalid_update *skel;
287
288 skel = test_sockmap_invalid_update__open_and_load();
289 if (!ASSERT_NULL(skel, "open_and_load"))
290 test_sockmap_invalid_update__destroy(skel);
291 }
292
test_sockmap_copy(enum bpf_map_type map_type)293 static void test_sockmap_copy(enum bpf_map_type map_type)
294 {
295 DECLARE_LIBBPF_OPTS(bpf_iter_attach_opts, opts);
296 int err, len, src_fd, iter_fd;
297 union bpf_iter_link_info linfo = {};
298 __u32 i, num_sockets, num_elems;
299 struct bpf_iter_sockmap *skel;
300 __s64 *sock_fd = NULL;
301 struct bpf_link *link;
302 struct bpf_map *src;
303 char buf[64];
304
305 skel = bpf_iter_sockmap__open_and_load();
306 if (!ASSERT_OK_PTR(skel, "bpf_iter_sockmap__open_and_load"))
307 return;
308
309 if (map_type == BPF_MAP_TYPE_SOCKMAP) {
310 src = skel->maps.sockmap;
311 num_elems = bpf_map__max_entries(src);
312 num_sockets = num_elems - 1;
313 } else {
314 src = skel->maps.sockhash;
315 num_elems = bpf_map__max_entries(src) - 1;
316 num_sockets = num_elems;
317 }
318
319 sock_fd = calloc(num_sockets, sizeof(*sock_fd));
320 if (!ASSERT_OK_PTR(sock_fd, "calloc(sock_fd)"))
321 goto out;
322
323 for (i = 0; i < num_sockets; i++)
324 sock_fd[i] = -1;
325
326 src_fd = bpf_map__fd(src);
327
328 for (i = 0; i < num_sockets; i++) {
329 sock_fd[i] = connected_socket_v4();
330 if (!ASSERT_NEQ(sock_fd[i], -1, "connected_socket_v4"))
331 goto out;
332
333 err = bpf_map_update_elem(src_fd, &i, &sock_fd[i], BPF_NOEXIST);
334 if (!ASSERT_OK(err, "map_update"))
335 goto out;
336 }
337
338 linfo.map.map_fd = src_fd;
339 opts.link_info = &linfo;
340 opts.link_info_len = sizeof(linfo);
341 link = bpf_program__attach_iter(skel->progs.copy, &opts);
342 if (!ASSERT_OK_PTR(link, "attach_iter"))
343 goto out;
344
345 iter_fd = bpf_iter_create(bpf_link__fd(link));
346 if (!ASSERT_GE(iter_fd, 0, "create_iter"))
347 goto free_link;
348
349 /* do some tests */
350 while ((len = read(iter_fd, buf, sizeof(buf))) > 0)
351 ;
352 if (!ASSERT_GE(len, 0, "read"))
353 goto close_iter;
354
355 /* test results */
356 if (!ASSERT_EQ(skel->bss->elems, num_elems, "elems"))
357 goto close_iter;
358
359 if (!ASSERT_EQ(skel->bss->socks, num_sockets, "socks"))
360 goto close_iter;
361
362 compare_cookies(src, skel->maps.dst);
363
364 close_iter:
365 close(iter_fd);
366 free_link:
367 bpf_link__destroy(link);
368 out:
369 for (i = 0; sock_fd && i < num_sockets; i++)
370 if (sock_fd[i] >= 0)
371 close(sock_fd[i]);
372 if (sock_fd)
373 free(sock_fd);
374 bpf_iter_sockmap__destroy(skel);
375 }
376
test_sockmap_skb_verdict_attach(enum bpf_attach_type first,enum bpf_attach_type second)377 static void test_sockmap_skb_verdict_attach(enum bpf_attach_type first,
378 enum bpf_attach_type second)
379 {
380 struct test_sockmap_skb_verdict_attach *skel;
381 int err, map, verdict;
382
383 skel = test_sockmap_skb_verdict_attach__open_and_load();
384 if (!ASSERT_OK_PTR(skel, "open_and_load"))
385 return;
386
387 verdict = bpf_program__fd(skel->progs.prog_skb_verdict);
388 map = bpf_map__fd(skel->maps.sock_map);
389
390 err = bpf_prog_attach(verdict, map, first, 0);
391 if (!ASSERT_OK(err, "bpf_prog_attach"))
392 goto out;
393
394 err = bpf_prog_attach(verdict, map, second, 0);
395 ASSERT_EQ(err, -EBUSY, "prog_attach_fail");
396
397 err = bpf_prog_detach2(verdict, map, first);
398 if (!ASSERT_OK(err, "bpf_prog_detach2"))
399 goto out;
400 out:
401 test_sockmap_skb_verdict_attach__destroy(skel);
402 }
403
test_sockmap_skb_verdict_attach_with_link(void)404 static void test_sockmap_skb_verdict_attach_with_link(void)
405 {
406 struct test_sockmap_skb_verdict_attach *skel;
407 struct bpf_program *prog;
408 struct bpf_link *link;
409 int err, map;
410
411 skel = test_sockmap_skb_verdict_attach__open_and_load();
412 if (!ASSERT_OK_PTR(skel, "open_and_load"))
413 return;
414 prog = skel->progs.prog_skb_verdict;
415 map = bpf_map__fd(skel->maps.sock_map);
416 link = bpf_program__attach_sockmap(prog, map);
417 if (!ASSERT_OK_PTR(link, "bpf_program__attach_sockmap"))
418 goto out;
419
420 bpf_link__destroy(link);
421
422 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_VERDICT, 0);
423 if (!ASSERT_OK(err, "bpf_prog_attach"))
424 goto out;
425
426 /* Fail since attaching with the same prog/map has been done. */
427 link = bpf_program__attach_sockmap(prog, map);
428 if (!ASSERT_ERR_PTR(link, "bpf_program__attach_sockmap"))
429 bpf_link__destroy(link);
430
431 err = bpf_prog_detach2(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_VERDICT);
432 if (!ASSERT_OK(err, "bpf_prog_detach2"))
433 goto out;
434 out:
435 test_sockmap_skb_verdict_attach__destroy(skel);
436 }
437
query_prog_id(int prog_fd)438 static __u32 query_prog_id(int prog_fd)
439 {
440 struct bpf_prog_info info = {};
441 __u32 info_len = sizeof(info);
442 int err;
443
444 err = bpf_prog_get_info_by_fd(prog_fd, &info, &info_len);
445 if (!ASSERT_OK(err, "bpf_prog_get_info_by_fd") ||
446 !ASSERT_EQ(info_len, sizeof(info), "bpf_prog_get_info_by_fd"))
447 return 0;
448
449 return info.id;
450 }
451
test_sockmap_progs_query(enum bpf_attach_type attach_type)452 static void test_sockmap_progs_query(enum bpf_attach_type attach_type)
453 {
454 struct test_sockmap_progs_query *skel;
455 int err, map_fd, verdict_fd;
456 __u32 attach_flags = 0;
457 __u32 prog_ids[3] = {};
458 __u32 prog_cnt = 3;
459
460 skel = test_sockmap_progs_query__open_and_load();
461 if (!ASSERT_OK_PTR(skel, "test_sockmap_progs_query__open_and_load"))
462 return;
463
464 map_fd = bpf_map__fd(skel->maps.sock_map);
465
466 if (attach_type == BPF_SK_MSG_VERDICT)
467 verdict_fd = bpf_program__fd(skel->progs.prog_skmsg_verdict);
468 else
469 verdict_fd = bpf_program__fd(skel->progs.prog_skb_verdict);
470
471 err = bpf_prog_query(map_fd, attach_type, 0 /* query flags */,
472 &attach_flags, prog_ids, &prog_cnt);
473 ASSERT_OK(err, "bpf_prog_query failed");
474 ASSERT_EQ(attach_flags, 0, "wrong attach_flags on query");
475 ASSERT_EQ(prog_cnt, 0, "wrong program count on query");
476
477 err = bpf_prog_attach(verdict_fd, map_fd, attach_type, 0);
478 if (!ASSERT_OK(err, "bpf_prog_attach failed"))
479 goto out;
480
481 prog_cnt = 1;
482 err = bpf_prog_query(map_fd, attach_type, 0 /* query flags */,
483 &attach_flags, prog_ids, &prog_cnt);
484 ASSERT_OK(err, "bpf_prog_query failed");
485 ASSERT_EQ(attach_flags, 0, "wrong attach_flags on query");
486 ASSERT_EQ(prog_cnt, 1, "wrong program count on query");
487 ASSERT_EQ(prog_ids[0], query_prog_id(verdict_fd),
488 "wrong prog_ids on query");
489
490 bpf_prog_detach2(verdict_fd, map_fd, attach_type);
491 out:
492 test_sockmap_progs_query__destroy(skel);
493 }
494
495 #define MAX_EVENTS 10
test_sockmap_skb_verdict_shutdown(void)496 static void test_sockmap_skb_verdict_shutdown(void)
497 {
498 int n, err, map, verdict, c1 = -1, p1 = -1;
499 struct epoll_event ev, events[MAX_EVENTS];
500 struct test_sockmap_pass_prog *skel;
501 int zero = 0;
502 int epollfd;
503 char b;
504
505 skel = test_sockmap_pass_prog__open_and_load();
506 if (!ASSERT_OK_PTR(skel, "open_and_load"))
507 return;
508
509 verdict = bpf_program__fd(skel->progs.prog_skb_verdict);
510 map = bpf_map__fd(skel->maps.sock_map_rx);
511
512 err = bpf_prog_attach(verdict, map, BPF_SK_SKB_STREAM_VERDICT, 0);
513 if (!ASSERT_OK(err, "bpf_prog_attach"))
514 goto out;
515
516 err = create_pair(AF_INET, SOCK_STREAM, &c1, &p1);
517 if (err < 0)
518 goto out;
519
520 err = bpf_map_update_elem(map, &zero, &c1, BPF_NOEXIST);
521 if (err < 0)
522 goto out_close;
523
524 shutdown(p1, SHUT_WR);
525
526 ev.events = EPOLLIN;
527 ev.data.fd = c1;
528
529 epollfd = epoll_create1(0);
530 if (!ASSERT_GT(epollfd, -1, "epoll_create(0)"))
531 goto out_close;
532 err = epoll_ctl(epollfd, EPOLL_CTL_ADD, c1, &ev);
533 if (!ASSERT_OK(err, "epoll_ctl(EPOLL_CTL_ADD)"))
534 goto out_close;
535 err = epoll_wait(epollfd, events, MAX_EVENTS, -1);
536 if (!ASSERT_EQ(err, 1, "epoll_wait(fd)"))
537 goto out_close;
538
539 n = recv(c1, &b, 1, MSG_DONTWAIT);
540 ASSERT_EQ(n, 0, "recv(fin)");
541 out_close:
542 close(c1);
543 close(p1);
544 out:
545 test_sockmap_pass_prog__destroy(skel);
546 }
547
548
do_test_sockmap_skb_verdict_fionread(int sotype,bool pass_prog)549 static void do_test_sockmap_skb_verdict_fionread(int sotype, bool pass_prog)
550 {
551 int err, map, verdict, c0 = -1, c1 = -1, p0 = -1, p1 = -1;
552 int expected, zero = 0, sent, recvd, avail;
553 struct test_sockmap_pass_prog *pass = NULL;
554 struct test_sockmap_drop_prog *drop = NULL;
555 char buf[256] = "0123456789";
556 int split_len = sizeof(buf) / 2;
557
558 if (pass_prog) {
559 pass = test_sockmap_pass_prog__open_and_load();
560 if (!ASSERT_OK_PTR(pass, "open_and_load"))
561 return;
562 verdict = bpf_program__fd(pass->progs.prog_skb_verdict);
563 map = bpf_map__fd(pass->maps.sock_map_rx);
564 if (sotype == SOCK_DGRAM)
565 expected = split_len; /* FIONREAD for UDP is different from TCP */
566 else
567 expected = sizeof(buf);
568 } else {
569 drop = test_sockmap_drop_prog__open_and_load();
570 if (!ASSERT_OK_PTR(drop, "open_and_load"))
571 return;
572 verdict = bpf_program__fd(drop->progs.prog_skb_verdict);
573 map = bpf_map__fd(drop->maps.sock_map_rx);
574 /* On drop data is consumed immediately and copied_seq inc'd */
575 expected = 0;
576 }
577
578
579 err = bpf_prog_attach(verdict, map, BPF_SK_SKB_STREAM_VERDICT, 0);
580 if (!ASSERT_OK(err, "bpf_prog_attach"))
581 goto out;
582
583 err = create_socket_pairs(AF_INET, sotype, &c0, &c1, &p0, &p1);
584 if (!ASSERT_OK(err, "create_socket_pairs()"))
585 goto out;
586
587 err = bpf_map_update_elem(map, &zero, &c1, BPF_NOEXIST);
588 if (!ASSERT_OK(err, "bpf_map_update_elem(c1)"))
589 goto out_close;
590
591 sent = xsend(p1, &buf, split_len, 0);
592 sent += xsend(p1, &buf, sizeof(buf) - split_len, 0);
593 ASSERT_EQ(sent, sizeof(buf), "xsend(p1)");
594 err = ioctl(c1, FIONREAD, &avail);
595 ASSERT_OK(err, "ioctl(FIONREAD) error");
596 ASSERT_EQ(avail, expected, "ioctl(FIONREAD)");
597 /* On DROP test there will be no data to read */
598 if (pass_prog) {
599 recvd = recv_timeout(c1, &buf, sizeof(buf), MSG_DONTWAIT, IO_TIMEOUT_SEC);
600 ASSERT_EQ(recvd, sizeof(buf), "recv_timeout(c0)");
601 }
602
603 out_close:
604 close(c0);
605 close(p0);
606 close(c1);
607 close(p1);
608 out:
609 if (pass_prog)
610 test_sockmap_pass_prog__destroy(pass);
611 else
612 test_sockmap_drop_prog__destroy(drop);
613 }
614
test_sockmap_skb_verdict_fionread(bool pass_prog)615 static void test_sockmap_skb_verdict_fionread(bool pass_prog)
616 {
617 do_test_sockmap_skb_verdict_fionread(SOCK_STREAM, pass_prog);
618 do_test_sockmap_skb_verdict_fionread(SOCK_DGRAM, pass_prog);
619 }
620
test_sockmap_skb_verdict_change_tail(void)621 static void test_sockmap_skb_verdict_change_tail(void)
622 {
623 struct test_sockmap_change_tail *skel;
624 int err, map, verdict;
625 int c1, p1, sent, recvd;
626 int zero = 0;
627 char buf[2];
628
629 skel = test_sockmap_change_tail__open_and_load();
630 if (!ASSERT_OK_PTR(skel, "open_and_load"))
631 return;
632 verdict = bpf_program__fd(skel->progs.prog_skb_verdict);
633 map = bpf_map__fd(skel->maps.sock_map_rx);
634
635 err = bpf_prog_attach(verdict, map, BPF_SK_SKB_STREAM_VERDICT, 0);
636 if (!ASSERT_OK(err, "bpf_prog_attach"))
637 goto out;
638 err = create_pair(AF_INET, SOCK_STREAM, &c1, &p1);
639 if (!ASSERT_OK(err, "create_pair()"))
640 goto out;
641 err = bpf_map_update_elem(map, &zero, &c1, BPF_NOEXIST);
642 if (!ASSERT_OK(err, "bpf_map_update_elem(c1)"))
643 goto out_close;
644 sent = xsend(p1, "Tr", 2, 0);
645 ASSERT_EQ(sent, 2, "xsend(p1)");
646 recvd = recv(c1, buf, 2, 0);
647 ASSERT_EQ(recvd, 1, "recv(c1)");
648 ASSERT_EQ(skel->data->change_tail_ret, 0, "change_tail_ret");
649
650 sent = xsend(p1, "G", 1, 0);
651 ASSERT_EQ(sent, 1, "xsend(p1)");
652 recvd = recv(c1, buf, 2, 0);
653 ASSERT_EQ(recvd, 2, "recv(c1)");
654 ASSERT_EQ(skel->data->change_tail_ret, 0, "change_tail_ret");
655
656 sent = xsend(p1, "E", 1, 0);
657 ASSERT_EQ(sent, 1, "xsend(p1)");
658 recvd = recv(c1, buf, 1, 0);
659 ASSERT_EQ(recvd, 1, "recv(c1)");
660 ASSERT_EQ(skel->data->change_tail_ret, -EINVAL, "change_tail_ret");
661
662 out_close:
663 close(c1);
664 close(p1);
665 out:
666 test_sockmap_change_tail__destroy(skel);
667 }
668
test_sockmap_skb_verdict_peek_helper(int map)669 static void test_sockmap_skb_verdict_peek_helper(int map)
670 {
671 int err, c1, p1, zero = 0, sent, recvd, avail;
672 char snd[256] = "0123456789";
673 char rcv[256] = "0";
674
675 err = create_pair(AF_INET, SOCK_STREAM, &c1, &p1);
676 if (!ASSERT_OK(err, "create_pair()"))
677 return;
678
679 err = bpf_map_update_elem(map, &zero, &c1, BPF_NOEXIST);
680 if (!ASSERT_OK(err, "bpf_map_update_elem(c1)"))
681 goto out_close;
682
683 sent = xsend(p1, snd, sizeof(snd), 0);
684 ASSERT_EQ(sent, sizeof(snd), "xsend(p1)");
685 recvd = recv(c1, rcv, sizeof(rcv), MSG_PEEK);
686 ASSERT_EQ(recvd, sizeof(rcv), "recv(c1)");
687 err = ioctl(c1, FIONREAD, &avail);
688 ASSERT_OK(err, "ioctl(FIONREAD) error");
689 ASSERT_EQ(avail, sizeof(snd), "after peek ioctl(FIONREAD)");
690 recvd = recv(c1, rcv, sizeof(rcv), 0);
691 ASSERT_EQ(recvd, sizeof(rcv), "recv(p0)");
692 err = ioctl(c1, FIONREAD, &avail);
693 ASSERT_OK(err, "ioctl(FIONREAD) error");
694 ASSERT_EQ(avail, 0, "after read ioctl(FIONREAD)");
695
696 out_close:
697 close(c1);
698 close(p1);
699 }
700
test_sockmap_skb_verdict_peek(void)701 static void test_sockmap_skb_verdict_peek(void)
702 {
703 struct test_sockmap_pass_prog *pass;
704 int err, map, verdict;
705
706 pass = test_sockmap_pass_prog__open_and_load();
707 if (!ASSERT_OK_PTR(pass, "open_and_load"))
708 return;
709 verdict = bpf_program__fd(pass->progs.prog_skb_verdict);
710 map = bpf_map__fd(pass->maps.sock_map_rx);
711
712 err = bpf_prog_attach(verdict, map, BPF_SK_SKB_STREAM_VERDICT, 0);
713 if (!ASSERT_OK(err, "bpf_prog_attach"))
714 goto out;
715
716 test_sockmap_skb_verdict_peek_helper(map);
717
718 out:
719 test_sockmap_pass_prog__destroy(pass);
720 }
721
test_sockmap_skb_verdict_peek_with_link(void)722 static void test_sockmap_skb_verdict_peek_with_link(void)
723 {
724 struct test_sockmap_pass_prog *pass;
725 struct bpf_program *prog;
726 struct bpf_link *link;
727 int err, map;
728
729 pass = test_sockmap_pass_prog__open_and_load();
730 if (!ASSERT_OK_PTR(pass, "open_and_load"))
731 return;
732 prog = pass->progs.prog_skb_verdict;
733 map = bpf_map__fd(pass->maps.sock_map_rx);
734 link = bpf_program__attach_sockmap(prog, map);
735 if (!ASSERT_OK_PTR(link, "bpf_program__attach_sockmap"))
736 goto out;
737
738 err = bpf_link__update_program(link, pass->progs.prog_skb_verdict_clone);
739 if (!ASSERT_OK(err, "bpf_link__update_program"))
740 goto out;
741
742 /* Fail since a prog with different attach type attempts to do update. */
743 err = bpf_link__update_program(link, pass->progs.prog_skb_parser);
744 if (!ASSERT_ERR(err, "bpf_link__update_program"))
745 goto out;
746
747 test_sockmap_skb_verdict_peek_helper(map);
748 ASSERT_EQ(pass->bss->clone_called, 1, "clone_called");
749 out:
750 bpf_link__destroy(link);
751 test_sockmap_pass_prog__destroy(pass);
752 }
753
test_sockmap_unconnected_unix(void)754 static void test_sockmap_unconnected_unix(void)
755 {
756 int err, map, stream = -1, dgram = -1, zero = 0;
757 struct test_sockmap_pass_prog *skel;
758
759 skel = test_sockmap_pass_prog__open_and_load();
760 if (!ASSERT_OK_PTR(skel, "open_and_load"))
761 return;
762
763 map = bpf_map__fd(skel->maps.sock_map_rx);
764
765 stream = xsocket(AF_UNIX, SOCK_STREAM, 0);
766 if (stream < 0)
767 goto out;
768
769 dgram = xsocket(AF_UNIX, SOCK_DGRAM, 0);
770 if (dgram < 0)
771 goto out;
772
773 err = bpf_map_update_elem(map, &zero, &stream, BPF_ANY);
774 if (!ASSERT_ERR(err, "bpf_map_update_elem(stream)"))
775 goto out;
776
777 err = bpf_map_update_elem(map, &zero, &dgram, BPF_ANY);
778 ASSERT_OK(err, "bpf_map_update_elem(dgram)");
779 out:
780 close(stream);
781 close(dgram);
782 test_sockmap_pass_prog__destroy(skel);
783 }
784
test_sockmap_many_socket(void)785 static void test_sockmap_many_socket(void)
786 {
787 struct test_sockmap_pass_prog *skel;
788 int stream[2], dgram, udp, tcp;
789 int i, err, map, entry = 0;
790
791 skel = test_sockmap_pass_prog__open_and_load();
792 if (!ASSERT_OK_PTR(skel, "open_and_load"))
793 return;
794
795 map = bpf_map__fd(skel->maps.sock_map_rx);
796
797 dgram = xsocket(AF_UNIX, SOCK_DGRAM, 0);
798 if (dgram < 0) {
799 test_sockmap_pass_prog__destroy(skel);
800 return;
801 }
802
803 tcp = connected_socket_v4();
804 if (!ASSERT_GE(tcp, 0, "connected_socket_v4")) {
805 close(dgram);
806 test_sockmap_pass_prog__destroy(skel);
807 return;
808 }
809
810 udp = xsocket(AF_INET, SOCK_DGRAM | SOCK_NONBLOCK, 0);
811 if (udp < 0) {
812 close(dgram);
813 close(tcp);
814 test_sockmap_pass_prog__destroy(skel);
815 return;
816 }
817
818 err = socketpair(AF_UNIX, SOCK_STREAM, 0, stream);
819 ASSERT_OK(err, "socketpair(af_unix, sock_stream)");
820 if (err)
821 goto out;
822
823 for (i = 0; i < 2; i++, entry++) {
824 err = bpf_map_update_elem(map, &entry, &stream[0], BPF_ANY);
825 ASSERT_OK(err, "bpf_map_update_elem(stream)");
826 }
827 for (i = 0; i < 2; i++, entry++) {
828 err = bpf_map_update_elem(map, &entry, &dgram, BPF_ANY);
829 ASSERT_OK(err, "bpf_map_update_elem(dgram)");
830 }
831 for (i = 0; i < 2; i++, entry++) {
832 err = bpf_map_update_elem(map, &entry, &udp, BPF_ANY);
833 ASSERT_OK(err, "bpf_map_update_elem(udp)");
834 }
835 for (i = 0; i < 2; i++, entry++) {
836 err = bpf_map_update_elem(map, &entry, &tcp, BPF_ANY);
837 ASSERT_OK(err, "bpf_map_update_elem(tcp)");
838 }
839 for (entry--; entry >= 0; entry--) {
840 err = bpf_map_delete_elem(map, &entry);
841 ASSERT_OK(err, "bpf_map_delete_elem(entry)");
842 }
843
844 close(stream[0]);
845 close(stream[1]);
846 out:
847 close(dgram);
848 close(tcp);
849 close(udp);
850 test_sockmap_pass_prog__destroy(skel);
851 }
852
test_sockmap_many_maps(void)853 static void test_sockmap_many_maps(void)
854 {
855 struct test_sockmap_pass_prog *skel;
856 int stream[2], dgram, udp, tcp;
857 int i, err, map[2], entry = 0;
858
859 skel = test_sockmap_pass_prog__open_and_load();
860 if (!ASSERT_OK_PTR(skel, "open_and_load"))
861 return;
862
863 map[0] = bpf_map__fd(skel->maps.sock_map_rx);
864 map[1] = bpf_map__fd(skel->maps.sock_map_tx);
865
866 dgram = xsocket(AF_UNIX, SOCK_DGRAM, 0);
867 if (dgram < 0) {
868 test_sockmap_pass_prog__destroy(skel);
869 return;
870 }
871
872 tcp = connected_socket_v4();
873 if (!ASSERT_GE(tcp, 0, "connected_socket_v4")) {
874 close(dgram);
875 test_sockmap_pass_prog__destroy(skel);
876 return;
877 }
878
879 udp = xsocket(AF_INET, SOCK_DGRAM | SOCK_NONBLOCK, 0);
880 if (udp < 0) {
881 close(dgram);
882 close(tcp);
883 test_sockmap_pass_prog__destroy(skel);
884 return;
885 }
886
887 err = socketpair(AF_UNIX, SOCK_STREAM, 0, stream);
888 ASSERT_OK(err, "socketpair(af_unix, sock_stream)");
889 if (err)
890 goto out;
891
892 for (i = 0; i < 2; i++, entry++) {
893 err = bpf_map_update_elem(map[i], &entry, &stream[0], BPF_ANY);
894 ASSERT_OK(err, "bpf_map_update_elem(stream)");
895 }
896 for (i = 0; i < 2; i++, entry++) {
897 err = bpf_map_update_elem(map[i], &entry, &dgram, BPF_ANY);
898 ASSERT_OK(err, "bpf_map_update_elem(dgram)");
899 }
900 for (i = 0; i < 2; i++, entry++) {
901 err = bpf_map_update_elem(map[i], &entry, &udp, BPF_ANY);
902 ASSERT_OK(err, "bpf_map_update_elem(udp)");
903 }
904 for (i = 0; i < 2; i++, entry++) {
905 err = bpf_map_update_elem(map[i], &entry, &tcp, BPF_ANY);
906 ASSERT_OK(err, "bpf_map_update_elem(tcp)");
907 }
908 for (entry--; entry >= 0; entry--) {
909 err = bpf_map_delete_elem(map[1], &entry);
910 entry--;
911 ASSERT_OK(err, "bpf_map_delete_elem(entry)");
912 err = bpf_map_delete_elem(map[0], &entry);
913 ASSERT_OK(err, "bpf_map_delete_elem(entry)");
914 }
915
916 close(stream[0]);
917 close(stream[1]);
918 out:
919 close(dgram);
920 close(tcp);
921 close(udp);
922 test_sockmap_pass_prog__destroy(skel);
923 }
924
test_sockmap_same_sock(void)925 static void test_sockmap_same_sock(void)
926 {
927 struct test_sockmap_pass_prog *skel;
928 int stream[2], dgram, udp, tcp;
929 int i, err, map, zero = 0;
930
931 skel = test_sockmap_pass_prog__open_and_load();
932 if (!ASSERT_OK_PTR(skel, "open_and_load"))
933 return;
934
935 map = bpf_map__fd(skel->maps.sock_map_rx);
936
937 dgram = xsocket(AF_UNIX, SOCK_DGRAM, 0);
938 if (dgram < 0) {
939 test_sockmap_pass_prog__destroy(skel);
940 return;
941 }
942
943 tcp = connected_socket_v4();
944 if (!ASSERT_GE(tcp, 0, "connected_socket_v4")) {
945 close(dgram);
946 test_sockmap_pass_prog__destroy(skel);
947 return;
948 }
949
950 udp = xsocket(AF_INET, SOCK_DGRAM | SOCK_NONBLOCK, 0);
951 if (udp < 0) {
952 close(dgram);
953 close(tcp);
954 test_sockmap_pass_prog__destroy(skel);
955 return;
956 }
957
958 err = socketpair(AF_UNIX, SOCK_STREAM, 0, stream);
959 ASSERT_OK(err, "socketpair(af_unix, sock_stream)");
960 if (err) {
961 close(tcp);
962 goto out;
963 }
964
965 for (i = 0; i < 2; i++) {
966 err = bpf_map_update_elem(map, &zero, &stream[0], BPF_ANY);
967 ASSERT_OK(err, "bpf_map_update_elem(stream)");
968 }
969 for (i = 0; i < 2; i++) {
970 err = bpf_map_update_elem(map, &zero, &dgram, BPF_ANY);
971 ASSERT_OK(err, "bpf_map_update_elem(dgram)");
972 }
973 for (i = 0; i < 2; i++) {
974 err = bpf_map_update_elem(map, &zero, &udp, BPF_ANY);
975 ASSERT_OK(err, "bpf_map_update_elem(udp)");
976 }
977 for (i = 0; i < 2; i++) {
978 err = bpf_map_update_elem(map, &zero, &tcp, BPF_ANY);
979 ASSERT_OK(err, "bpf_map_update_elem(tcp)");
980 }
981
982 close(tcp);
983 err = bpf_map_delete_elem(map, &zero);
984 ASSERT_ERR(err, "bpf_map_delete_elem(entry)");
985
986 close(stream[0]);
987 close(stream[1]);
988 out:
989 close(dgram);
990 close(udp);
991 test_sockmap_pass_prog__destroy(skel);
992 }
993
test_sockmap_skb_verdict_vsock_poll(void)994 static void test_sockmap_skb_verdict_vsock_poll(void)
995 {
996 struct test_sockmap_pass_prog *skel;
997 int err, map, conn, peer;
998 struct bpf_program *prog;
999 struct bpf_link *link;
1000 char buf = 'x';
1001 int zero = 0;
1002
1003 skel = test_sockmap_pass_prog__open_and_load();
1004 if (!ASSERT_OK_PTR(skel, "open_and_load"))
1005 return;
1006
1007 if (create_pair(AF_VSOCK, SOCK_STREAM, &conn, &peer))
1008 goto destroy;
1009
1010 prog = skel->progs.prog_skb_verdict;
1011 map = bpf_map__fd(skel->maps.sock_map_rx);
1012 link = bpf_program__attach_sockmap(prog, map);
1013 if (!ASSERT_OK_PTR(link, "bpf_program__attach_sockmap"))
1014 goto close;
1015
1016 err = bpf_map_update_elem(map, &zero, &conn, BPF_ANY);
1017 if (!ASSERT_OK(err, "bpf_map_update_elem"))
1018 goto detach;
1019
1020 if (xsend(peer, &buf, 1, 0) != 1)
1021 goto detach;
1022
1023 err = poll_read(conn, IO_TIMEOUT_SEC);
1024 if (!ASSERT_OK(err, "poll"))
1025 goto detach;
1026
1027 if (xrecv_nonblock(conn, &buf, 1, 0) != 1)
1028 FAIL("xrecv_nonblock");
1029 detach:
1030 bpf_link__destroy(link);
1031 close:
1032 xclose(conn);
1033 xclose(peer);
1034 destroy:
1035 test_sockmap_pass_prog__destroy(skel);
1036 }
1037
test_sockmap_vsock_unconnected(void)1038 static void test_sockmap_vsock_unconnected(void)
1039 {
1040 struct sockaddr_storage addr;
1041 int map, s, zero = 0;
1042 socklen_t alen;
1043
1044 map = bpf_map_create(BPF_MAP_TYPE_SOCKMAP, NULL, sizeof(int),
1045 sizeof(int), 1, NULL);
1046 if (!ASSERT_OK_FD(map, "bpf_map_create"))
1047 return;
1048
1049 s = xsocket(AF_VSOCK, SOCK_STREAM, 0);
1050 if (s < 0)
1051 goto close_map;
1052
1053 /* Fail connect(), but trigger transport assignment. */
1054 init_addr_loopback(AF_VSOCK, &addr, &alen);
1055 if (!ASSERT_ERR(connect(s, sockaddr(&addr), alen), "connect"))
1056 goto close_sock;
1057
1058 ASSERT_ERR(bpf_map_update_elem(map, &zero, &s, BPF_ANY), "map_update");
1059
1060 close_sock:
1061 xclose(s);
1062 close_map:
1063 xclose(map);
1064 }
1065
1066 /* it is used to reproduce WARNING */
test_sockmap_zc(void)1067 static void test_sockmap_zc(void)
1068 {
1069 int map, err, sent, recvd, zero = 0, one = 1, on = 1;
1070 char buf[10] = "0123456789", rcv[11], addr[100];
1071 struct test_sockmap_pass_prog *skel = NULL;
1072 int c0 = -1, p0 = -1, c1 = -1, p1 = -1;
1073 struct tcp_zerocopy_receive zc;
1074 socklen_t zc_len = sizeof(zc);
1075 struct bpf_program *prog;
1076
1077 skel = test_sockmap_pass_prog__open_and_load();
1078 if (!ASSERT_OK_PTR(skel, "open_and_load"))
1079 return;
1080
1081 if (create_socket_pairs(AF_INET, SOCK_STREAM, &c0, &c1, &p0, &p1))
1082 goto end;
1083
1084 prog = skel->progs.prog_skb_verdict_ingress;
1085 map = bpf_map__fd(skel->maps.sock_map_rx);
1086
1087 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_VERDICT, 0);
1088 if (!ASSERT_OK(err, "bpf_prog_attach"))
1089 goto end;
1090
1091 err = bpf_map_update_elem(map, &zero, &p0, BPF_ANY);
1092 if (!ASSERT_OK(err, "bpf_map_update_elem"))
1093 goto end;
1094
1095 err = bpf_map_update_elem(map, &one, &p1, BPF_ANY);
1096 if (!ASSERT_OK(err, "bpf_map_update_elem"))
1097 goto end;
1098
1099 sent = xsend(c0, buf, sizeof(buf), 0);
1100 if (!ASSERT_EQ(sent, sizeof(buf), "xsend"))
1101 goto end;
1102
1103 /* trigger tcp_bpf_recvmsg_parser and inc copied_seq of p1 */
1104 recvd = recv_timeout(p1, rcv, sizeof(rcv), MSG_DONTWAIT, 1);
1105 if (!ASSERT_EQ(recvd, sent, "recv_timeout(p1)"))
1106 goto end;
1107
1108 /* uninstall sockmap of p1 */
1109 bpf_map_delete_elem(map, &one);
1110
1111 /* trigger tcp stack and the rcv_nxt of p1 is less than copied_seq */
1112 sent = xsend(c1, buf, sizeof(buf) - 1, 0);
1113 if (!ASSERT_EQ(sent, sizeof(buf) - 1, "xsend"))
1114 goto end;
1115
1116 err = setsockopt(p1, SOL_SOCKET, SO_ZEROCOPY, &on, sizeof(on));
1117 if (!ASSERT_OK(err, "setsockopt"))
1118 goto end;
1119
1120 memset(&zc, 0, sizeof(zc));
1121 zc.copybuf_address = (__u64)((unsigned long)addr);
1122 zc.copybuf_len = sizeof(addr);
1123
1124 err = getsockopt(p1, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, &zc, &zc_len);
1125 if (!ASSERT_OK(err, "getsockopt"))
1126 goto end;
1127
1128 end:
1129 if (c0 >= 0)
1130 close(c0);
1131 if (p0 >= 0)
1132 close(p0);
1133 if (c1 >= 0)
1134 close(c1);
1135 if (p1 >= 0)
1136 close(p1);
1137 test_sockmap_pass_prog__destroy(skel);
1138 }
1139
1140 /* it is used to check whether copied_seq of sk is correct */
test_sockmap_copied_seq(bool strp)1141 static void test_sockmap_copied_seq(bool strp)
1142 {
1143 int i, map, err, sent, recvd, zero = 0, one = 1;
1144 struct test_sockmap_pass_prog *skel = NULL;
1145 int c0 = -1, p0 = -1, c1 = -1, p1 = -1;
1146 char buf[10] = "0123456789", rcv[11];
1147 struct bpf_program *prog;
1148
1149 skel = test_sockmap_pass_prog__open_and_load();
1150 if (!ASSERT_OK_PTR(skel, "open_and_load"))
1151 return;
1152
1153 if (create_socket_pairs(AF_INET, SOCK_STREAM, &c0, &c1, &p0, &p1))
1154 goto end;
1155
1156 prog = skel->progs.prog_skb_verdict_ingress;
1157 map = bpf_map__fd(skel->maps.sock_map_rx);
1158
1159 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_VERDICT, 0);
1160 if (!ASSERT_OK(err, "bpf_prog_attach verdict"))
1161 goto end;
1162
1163 if (strp) {
1164 prog = skel->progs.prog_skb_verdict_ingress_strp;
1165 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_PARSER, 0);
1166 if (!ASSERT_OK(err, "bpf_prog_attach parser"))
1167 goto end;
1168 }
1169
1170 err = bpf_map_update_elem(map, &zero, &p0, BPF_ANY);
1171 if (!ASSERT_OK(err, "bpf_map_update_elem(p0)"))
1172 goto end;
1173
1174 err = bpf_map_update_elem(map, &one, &p1, BPF_ANY);
1175 if (!ASSERT_OK(err, "bpf_map_update_elem(p1)"))
1176 goto end;
1177
1178 /* just trigger sockamp: data sent by c0 will be received by p1 */
1179 sent = xsend(c0, buf, sizeof(buf), 0);
1180 if (!ASSERT_EQ(sent, sizeof(buf), "xsend(c0), bpf"))
1181 goto end;
1182
1183 /* do partial read */
1184 recvd = recv_timeout(p1, rcv, 1, MSG_DONTWAIT, 1);
1185 recvd += recv_timeout(p1, rcv + 1, sizeof(rcv) - 1, MSG_DONTWAIT, 1);
1186 if (!ASSERT_EQ(recvd, sent, "recv_timeout(p1), bpf") ||
1187 !ASSERT_OK(memcmp(buf, rcv, recvd), "data mismatch"))
1188 goto end;
1189
1190 /* uninstall sockmap of p1 and p0 */
1191 err = bpf_map_delete_elem(map, &one);
1192 if (!ASSERT_OK(err, "bpf_map_delete_elem(1)"))
1193 goto end;
1194
1195 err = bpf_map_delete_elem(map, &zero);
1196 if (!ASSERT_OK(err, "bpf_map_delete_elem(0)"))
1197 goto end;
1198
1199 /* now all sockets become plain socket, they should still work */
1200 for (i = 0; i < 5; i++) {
1201 /* test copied_seq of p1 by running tcp native stack */
1202 sent = xsend(c1, buf, sizeof(buf), 0);
1203 if (!ASSERT_EQ(sent, sizeof(buf), "xsend(c1), native"))
1204 goto end;
1205
1206 recvd = recv(p1, rcv, sizeof(rcv), MSG_DONTWAIT);
1207 if (!ASSERT_EQ(recvd, sent, "recv_timeout(p1), native"))
1208 goto end;
1209
1210 /* p0 previously redirected skb to p1, we also check copied_seq of p0 */
1211 sent = xsend(c0, buf, sizeof(buf), 0);
1212 if (!ASSERT_EQ(sent, sizeof(buf), "xsend(c0), native"))
1213 goto end;
1214
1215 recvd = recv(p0, rcv, sizeof(rcv), MSG_DONTWAIT);
1216 if (!ASSERT_EQ(recvd, sent, "recv_timeout(p0), native"))
1217 goto end;
1218 }
1219
1220 end:
1221 if (c0 >= 0)
1222 close(c0);
1223 if (p0 >= 0)
1224 close(p0);
1225 if (c1 >= 0)
1226 close(c1);
1227 if (p1 >= 0)
1228 close(p1);
1229 test_sockmap_pass_prog__destroy(skel);
1230 }
1231
1232 /* Wait until FIONREAD returns the expected value or timeout */
wait_for_fionread(int fd,int expected,unsigned int timeout_ms)1233 static int wait_for_fionread(int fd, int expected, unsigned int timeout_ms)
1234 {
1235 unsigned int elapsed = 0;
1236 int avail = 0;
1237
1238 while (elapsed < timeout_ms) {
1239 if (ioctl(fd, FIONREAD, &avail) < 0)
1240 return -errno;
1241 if (avail >= expected)
1242 return avail;
1243 usleep(1000);
1244 elapsed++;
1245 }
1246 return avail;
1247 }
1248
1249 /* it is used to send data to via native stack and BPF redirecting */
test_sockmap_multi_channels(int sotype)1250 static void test_sockmap_multi_channels(int sotype)
1251 {
1252 int map, err, sent, recvd, zero = 0, one = 1, avail = 0, expected;
1253 struct test_sockmap_pass_prog *skel = NULL;
1254 int c0 = -1, p0 = -1, c1 = -1, p1 = -1;
1255 char buf[10] = "0123456789", rcv[11];
1256 struct bpf_program *prog;
1257
1258 skel = test_sockmap_pass_prog__open_and_load();
1259 if (!ASSERT_OK_PTR(skel, "open_and_load"))
1260 return;
1261
1262 err = create_socket_pairs(AF_INET, sotype, &c0, &c1, &p0, &p1);
1263 if (err)
1264 goto end;
1265
1266 prog = skel->progs.prog_skb_verdict_ingress;
1267 map = bpf_map__fd(skel->maps.sock_map_rx);
1268
1269 err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_SKB_STREAM_VERDICT, 0);
1270 if (!ASSERT_OK(err, "bpf_prog_attach verdict"))
1271 goto end;
1272
1273 err = bpf_map_update_elem(map, &zero, &p0, BPF_ANY);
1274 if (!ASSERT_OK(err, "bpf_map_update_elem(p0)"))
1275 goto end;
1276
1277 err = bpf_map_update_elem(map, &one, &p1, BPF_ANY);
1278 if (!ASSERT_OK(err, "bpf_map_update_elem"))
1279 goto end;
1280
1281 /* send data to p1 via native stack */
1282 sent = xsend(c1, buf, 2, 0);
1283 if (!ASSERT_EQ(sent, 2, "xsend(2)"))
1284 goto end;
1285
1286 avail = wait_for_fionread(p1, 2, IO_TIMEOUT_SEC);
1287 ASSERT_EQ(avail, 2, "ioctl(FIONREAD) partial return");
1288
1289 /* send data to p1 via bpf redirecting */
1290 sent = xsend(c0, buf + 2, sizeof(buf) - 2, 0);
1291 if (!ASSERT_EQ(sent, sizeof(buf) - 2, "xsend(remain-data)"))
1292 goto end;
1293
1294 /* Poll FIONREAD until expected bytes arrive, poll_read() is unreliable
1295 * here since it may return immediately if prior data is already queued.
1296 */
1297 expected = sotype == SOCK_DGRAM ? 2 : sizeof(buf);
1298 avail = wait_for_fionread(p1, expected, IO_TIMEOUT_SEC);
1299 ASSERT_EQ(avail, expected, "ioctl(FIONREAD) full return");
1300
1301 recvd = recv_timeout(p1, rcv, sizeof(rcv), MSG_DONTWAIT, 1);
1302 if (!ASSERT_EQ(recvd, sizeof(buf), "recv_timeout(p1)") ||
1303 !ASSERT_OK(memcmp(buf, rcv, recvd), "data mismatch"))
1304 goto end;
1305 end:
1306 if (c0 >= 0)
1307 close(c0);
1308 if (p0 >= 0)
1309 close(p0);
1310 if (c1 >= 0)
1311 close(c1);
1312 if (p1 >= 0)
1313 close(p1);
1314 test_sockmap_pass_prog__destroy(skel);
1315 }
1316
test_sockmap_basic(void)1317 void test_sockmap_basic(void)
1318 {
1319 if (test__start_subtest("sockmap create_update_free"))
1320 test_sockmap_create_update_free(BPF_MAP_TYPE_SOCKMAP);
1321 if (test__start_subtest("sockhash create_update_free"))
1322 test_sockmap_create_update_free(BPF_MAP_TYPE_SOCKHASH);
1323 if (test__start_subtest("sockmap vsock delete on close"))
1324 test_sockmap_vsock_delete_on_close();
1325 if (test__start_subtest("sockmap sk_msg load helpers"))
1326 test_skmsg_helpers(BPF_MAP_TYPE_SOCKMAP);
1327 if (test__start_subtest("sockhash sk_msg load helpers"))
1328 test_skmsg_helpers(BPF_MAP_TYPE_SOCKHASH);
1329 if (test__start_subtest("sockmap update"))
1330 test_sockmap_update(BPF_MAP_TYPE_SOCKMAP);
1331 if (test__start_subtest("sockhash update"))
1332 test_sockmap_update(BPF_MAP_TYPE_SOCKHASH);
1333 if (test__start_subtest("sockmap update in unsafe context"))
1334 test_sockmap_invalid_update();
1335 if (test__start_subtest("sockmap copy"))
1336 test_sockmap_copy(BPF_MAP_TYPE_SOCKMAP);
1337 if (test__start_subtest("sockhash copy"))
1338 test_sockmap_copy(BPF_MAP_TYPE_SOCKHASH);
1339 if (test__start_subtest("sockmap skb_verdict attach")) {
1340 test_sockmap_skb_verdict_attach(BPF_SK_SKB_VERDICT,
1341 BPF_SK_SKB_STREAM_VERDICT);
1342 test_sockmap_skb_verdict_attach(BPF_SK_SKB_STREAM_VERDICT,
1343 BPF_SK_SKB_VERDICT);
1344 }
1345 if (test__start_subtest("sockmap skb_verdict attach_with_link"))
1346 test_sockmap_skb_verdict_attach_with_link();
1347 if (test__start_subtest("sockmap msg_verdict progs query"))
1348 test_sockmap_progs_query(BPF_SK_MSG_VERDICT);
1349 if (test__start_subtest("sockmap stream_parser progs query"))
1350 test_sockmap_progs_query(BPF_SK_SKB_STREAM_PARSER);
1351 if (test__start_subtest("sockmap stream_verdict progs query"))
1352 test_sockmap_progs_query(BPF_SK_SKB_STREAM_VERDICT);
1353 if (test__start_subtest("sockmap skb_verdict progs query"))
1354 test_sockmap_progs_query(BPF_SK_SKB_VERDICT);
1355 if (test__start_subtest("sockmap skb_verdict shutdown"))
1356 test_sockmap_skb_verdict_shutdown();
1357 if (test__start_subtest("sockmap skb_verdict fionread"))
1358 test_sockmap_skb_verdict_fionread(true);
1359 if (test__start_subtest("sockmap skb_verdict fionread on drop"))
1360 test_sockmap_skb_verdict_fionread(false);
1361 if (test__start_subtest("sockmap skb_verdict change tail"))
1362 test_sockmap_skb_verdict_change_tail();
1363 if (test__start_subtest("sockmap skb_verdict msg_f_peek"))
1364 test_sockmap_skb_verdict_peek();
1365 if (test__start_subtest("sockmap skb_verdict msg_f_peek with link"))
1366 test_sockmap_skb_verdict_peek_with_link();
1367 if (test__start_subtest("sockmap unconnected af_unix"))
1368 test_sockmap_unconnected_unix();
1369 if (test__start_subtest("sockmap one socket to many map entries"))
1370 test_sockmap_many_socket();
1371 if (test__start_subtest("sockmap one socket to many maps"))
1372 test_sockmap_many_maps();
1373 if (test__start_subtest("sockmap same socket replace"))
1374 test_sockmap_same_sock();
1375 if (test__start_subtest("sockmap sk_msg attach sockmap helpers with link"))
1376 test_skmsg_helpers_with_link(BPF_MAP_TYPE_SOCKMAP);
1377 if (test__start_subtest("sockhash sk_msg attach sockhash helpers with link"))
1378 test_skmsg_helpers_with_link(BPF_MAP_TYPE_SOCKHASH);
1379 if (test__start_subtest("sockmap skb_verdict vsock poll"))
1380 test_sockmap_skb_verdict_vsock_poll();
1381 if (test__start_subtest("sockmap vsock unconnected"))
1382 test_sockmap_vsock_unconnected();
1383 if (test__start_subtest("sockmap with zc"))
1384 test_sockmap_zc();
1385 if (test__start_subtest("sockmap recover"))
1386 test_sockmap_copied_seq(false);
1387 if (test__start_subtest("sockmap recover with strp"))
1388 test_sockmap_copied_seq(true);
1389 if (test__start_subtest("sockmap tcp multi channels"))
1390 test_sockmap_multi_channels(SOCK_STREAM);
1391 if (test__start_subtest("sockmap udp multi channels"))
1392 test_sockmap_multi_channels(SOCK_DGRAM);
1393 }
1394