1 // SPDX-License-Identifier: GPL-2.0
2 #include <test_progs.h>
3 #include <network_helpers.h>
4 #include <linux/ipv6.h>
5 #include <arpa/inet.h>
6 #include "test_xdp_context_test_run.skel.h"
7 #include "test_xdp_meta.skel.h"
8
9 #define RX_NAME "veth0"
10 #define TX_NAME "veth1"
11 #define TX_NETNS "xdp_context_tx"
12 #define RX_NETNS "xdp_context_rx"
13 #define RX_MAC "02:00:00:00:00:01"
14 #define TX_MAC "02:00:00:00:00:02"
15 #define TAP_NAME "tap0"
16 #define DUMMY_NAME "dum0"
17 #define TAP_NETNS "xdp_context_tuntap"
18 #define LWT_NETNS "xdp_context_lwt"
19
20 #define TEST_PAYLOAD_LEN 32
21 static const __u8 test_payload[TEST_PAYLOAD_LEN] = {
22 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
23 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18,
24 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28,
25 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38,
26 };
27
test_xdp_context_error(int prog_fd,struct bpf_test_run_opts opts,__u32 data_meta,__u32 data,__u32 data_end,__u32 ingress_ifindex,__u32 rx_queue_index,__u32 egress_ifindex)28 void test_xdp_context_error(int prog_fd, struct bpf_test_run_opts opts,
29 __u32 data_meta, __u32 data, __u32 data_end,
30 __u32 ingress_ifindex, __u32 rx_queue_index,
31 __u32 egress_ifindex)
32 {
33 struct xdp_md ctx = {
34 .data = data,
35 .data_end = data_end,
36 .data_meta = data_meta,
37 .ingress_ifindex = ingress_ifindex,
38 .rx_queue_index = rx_queue_index,
39 .egress_ifindex = egress_ifindex,
40 };
41 int err;
42
43 opts.ctx_in = &ctx;
44 opts.ctx_size_in = sizeof(ctx);
45 err = bpf_prog_test_run_opts(prog_fd, &opts);
46 ASSERT_EQ(errno, EINVAL, "errno-EINVAL");
47 ASSERT_ERR(err, "bpf_prog_test_run");
48 }
49
test_xdp_context_test_run(void)50 void test_xdp_context_test_run(void)
51 {
52 struct test_xdp_context_test_run *skel = NULL;
53 char data[sizeof(pkt_v4) + sizeof(__u32)];
54 char bad_ctx[sizeof(struct xdp_md) + 1];
55 char large_data[256];
56 struct xdp_md ctx_in, ctx_out;
57 DECLARE_LIBBPF_OPTS(bpf_test_run_opts, opts,
58 .data_in = &data,
59 .data_size_in = sizeof(data),
60 .ctx_out = &ctx_out,
61 .ctx_size_out = sizeof(ctx_out),
62 .repeat = 1,
63 );
64 int err, prog_fd;
65
66 skel = test_xdp_context_test_run__open_and_load();
67 if (!ASSERT_OK_PTR(skel, "skel"))
68 return;
69 prog_fd = bpf_program__fd(skel->progs.xdp_context);
70
71 /* Data past the end of the kernel's struct xdp_md must be 0 */
72 bad_ctx[sizeof(bad_ctx) - 1] = 1;
73 opts.ctx_in = bad_ctx;
74 opts.ctx_size_in = sizeof(bad_ctx);
75 err = bpf_prog_test_run_opts(prog_fd, &opts);
76 ASSERT_EQ(errno, E2BIG, "extradata-errno");
77 ASSERT_ERR(err, "bpf_prog_test_run(extradata)");
78
79 *(__u32 *)data = XDP_PASS;
80 *(struct ipv4_packet *)(data + sizeof(__u32)) = pkt_v4;
81 opts.ctx_in = &ctx_in;
82 opts.ctx_size_in = sizeof(ctx_in);
83 memset(&ctx_in, 0, sizeof(ctx_in));
84 ctx_in.data_meta = 0;
85 ctx_in.data = sizeof(__u32);
86 ctx_in.data_end = ctx_in.data + sizeof(pkt_v4);
87 err = bpf_prog_test_run_opts(prog_fd, &opts);
88 ASSERT_OK(err, "bpf_prog_test_run(valid)");
89 ASSERT_EQ(opts.retval, XDP_PASS, "valid-retval");
90 ASSERT_EQ(opts.data_size_out, sizeof(pkt_v4), "valid-datasize");
91 ASSERT_EQ(opts.ctx_size_out, opts.ctx_size_in, "valid-ctxsize");
92 ASSERT_EQ(ctx_out.data_meta, 0, "valid-datameta");
93 ASSERT_EQ(ctx_out.data, 0, "valid-data");
94 ASSERT_EQ(ctx_out.data_end, sizeof(pkt_v4), "valid-dataend");
95
96 /* Meta data's size must be a multiple of 4 */
97 test_xdp_context_error(prog_fd, opts, 0, 1, sizeof(data), 0, 0, 0);
98
99 /* data_meta must reference the start of data */
100 test_xdp_context_error(prog_fd, opts, 4, sizeof(__u32), sizeof(data),
101 0, 0, 0);
102
103 /* Total size of data must be data_end - data_meta or larger */
104 test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32),
105 sizeof(data) + 1, 0, 0, 0);
106
107 /* RX queue cannot be specified without specifying an ingress */
108 test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
109 0, 1, 0);
110
111 /* Interface 1 is always the loopback interface which always has only
112 * one RX queue (index 0). This makes index 1 an invalid rx queue index
113 * for interface 1.
114 */
115 test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
116 1, 1, 0);
117
118 /* The egress cannot be specified */
119 test_xdp_context_error(prog_fd, opts, 0, sizeof(__u32), sizeof(data),
120 0, 0, 1);
121
122 /* Meta data must be 216 bytes or smaller (256 - sizeof(struct
123 * xdp_frame)). Test both nearest invalid size and nearest invalid
124 * 4-byte-aligned size, and make sure data_in is large enough that we
125 * actually hit the check on metadata length
126 */
127 opts.data_in = large_data;
128 opts.data_size_in = sizeof(large_data);
129 test_xdp_context_error(prog_fd, opts, 0, 217, sizeof(large_data), 0, 0, 0);
130 test_xdp_context_error(prog_fd, opts, 0, 220, sizeof(large_data), 0, 0, 0);
131
132 test_xdp_context_test_run__destroy(skel);
133 }
134
send_test_packet(int ifindex)135 static int send_test_packet(int ifindex)
136 {
137 int n, sock = -1;
138 __u8 packet[sizeof(struct ethhdr) + TEST_PAYLOAD_LEN];
139
140 /* We use the Ethernet header only to identify the test packet */
141 struct ethhdr eth = {
142 .h_source = { 0x12, 0x34, 0xDE, 0xAD, 0xBE, 0xEF },
143 };
144
145 memcpy(packet, ð, sizeof(eth));
146 memcpy(packet + sizeof(eth), test_payload, TEST_PAYLOAD_LEN);
147
148 sock = socket(AF_PACKET, SOCK_RAW, IPPROTO_RAW);
149 if (!ASSERT_GE(sock, 0, "socket"))
150 goto err;
151
152 struct sockaddr_ll saddr = {
153 .sll_family = PF_PACKET,
154 .sll_ifindex = ifindex,
155 .sll_halen = ETH_ALEN
156 };
157 n = sendto(sock, packet, sizeof(packet), 0, (struct sockaddr *)&saddr,
158 sizeof(saddr));
159 if (!ASSERT_EQ(n, sizeof(packet), "sendto"))
160 goto err;
161
162 close(sock);
163 return 0;
164
165 err:
166 if (sock >= 0)
167 close(sock);
168 return -1;
169 }
170
write_test_packet(int tap_fd)171 static int write_test_packet(int tap_fd)
172 {
173 __u8 packet[sizeof(struct ethhdr) + TEST_PAYLOAD_LEN];
174 int n;
175
176 /* The Ethernet header is mostly not relevant. We use it to identify the
177 * test packet and some BPF helpers we exercise expect to operate on
178 * Ethernet frames carrying IP packets. Pretend that's the case.
179 */
180 struct ethhdr eth = {
181 .h_source = { 0x12, 0x34, 0xDE, 0xAD, 0xBE, 0xEF },
182 .h_proto = htons(ETH_P_IP),
183 };
184
185 memcpy(packet, ð, sizeof(eth));
186 memcpy(packet + sizeof(struct ethhdr), test_payload, TEST_PAYLOAD_LEN);
187
188 n = write(tap_fd, packet, sizeof(packet));
189 if (!ASSERT_EQ(n, sizeof(packet), "write packet"))
190 return -1;
191
192 return 0;
193 }
194
195 /* Inject Ethernet+IPv6+UDP frame into TAP */
write_test_packet_udp(int tap_fd)196 static int write_test_packet_udp(int tap_fd)
197 {
198 __u8 pkt[sizeof(struct ethhdr) + sizeof(struct ipv6hdr) +
199 sizeof(struct udphdr) + TEST_PAYLOAD_LEN] = {};
200 struct ethhdr *eth = (void *)pkt;
201 struct ipv6hdr *ip6 = (void *)(eth + 1);
202 struct udphdr *udp = (void *)(ip6 + 1);
203 __u8 *payload = (void *)(udp + 1);
204 const __u8 tap_mac[ETH_ALEN] = { 0x02, 0, 0, 0, 0, 0x01 };
205 int n;
206
207 memcpy(eth->h_dest, tap_mac, ETH_ALEN);
208 eth->h_proto = htons(ETH_P_IPV6);
209
210 ip6->version = 6;
211 ip6->hop_limit = 64;
212 ip6->nexthdr = IPPROTO_UDP;
213 ip6->payload_len = htons(sizeof(*udp) + TEST_PAYLOAD_LEN);
214 inet_pton(AF_INET6, "fd00::2", &ip6->saddr);
215 inet_pton(AF_INET6, "fd00:1::1", &ip6->daddr);
216
217 udp->source = htons(42);
218 udp->dest = htons(42);
219 udp->len = htons(sizeof(*udp) + TEST_PAYLOAD_LEN);
220 /* UDP checksum is not validated on the forwarding path. */
221
222 memcpy(payload, test_payload, TEST_PAYLOAD_LEN);
223
224 n = write(tap_fd, pkt, sizeof(pkt));
225 if (!ASSERT_EQ(n, sizeof(pkt), "write frame"))
226 return -1;
227
228 return 0;
229 }
230
dump_err_stream(const struct bpf_program * prog)231 static void dump_err_stream(const struct bpf_program *prog)
232 {
233 char buf[512];
234 int ret;
235
236 ret = 0;
237 do {
238 ret = bpf_prog_stream_read(bpf_program__fd(prog),
239 BPF_STREAM_STDERR, buf, sizeof(buf),
240 NULL);
241 if (ret > 0)
242 fwrite(buf, sizeof(buf[0]), ret, stderr);
243 } while (ret > 0);
244 }
245
test_xdp_context_veth(void)246 void test_xdp_context_veth(void)
247 {
248 LIBBPF_OPTS(bpf_tc_hook, tc_hook, .attach_point = BPF_TC_INGRESS);
249 LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 1);
250 struct netns_obj *rx_ns = NULL, *tx_ns = NULL;
251 struct bpf_program *tc_prog, *xdp_prog;
252 struct test_xdp_meta *skel = NULL;
253 struct nstoken *nstoken = NULL;
254 int rx_ifindex, tx_ifindex;
255 int ret;
256
257 tx_ns = netns_new(TX_NETNS, false);
258 if (!ASSERT_OK_PTR(tx_ns, "create tx_ns"))
259 return;
260
261 rx_ns = netns_new(RX_NETNS, false);
262 if (!ASSERT_OK_PTR(rx_ns, "create rx_ns"))
263 goto close;
264
265 SYS(close, "ip link add " RX_NAME " netns " RX_NETNS
266 " type veth peer name " TX_NAME " netns " TX_NETNS);
267
268 nstoken = open_netns(RX_NETNS);
269 if (!ASSERT_OK_PTR(nstoken, "setns rx_ns"))
270 goto close;
271
272 SYS(close, "ip link set dev " RX_NAME " up");
273
274 skel = test_xdp_meta__open_and_load();
275 if (!ASSERT_OK_PTR(skel, "open and load skeleton"))
276 goto close;
277
278 rx_ifindex = if_nametoindex(RX_NAME);
279 if (!ASSERT_GE(rx_ifindex, 0, "if_nametoindex rx"))
280 goto close;
281
282 tc_hook.ifindex = rx_ifindex;
283 ret = bpf_tc_hook_create(&tc_hook);
284 if (!ASSERT_OK(ret, "bpf_tc_hook_create"))
285 goto close;
286
287 tc_prog = bpf_object__find_program_by_name(skel->obj, "ing_cls");
288 if (!ASSERT_OK_PTR(tc_prog, "open ing_cls prog"))
289 goto close;
290
291 tc_opts.prog_fd = bpf_program__fd(tc_prog);
292 ret = bpf_tc_attach(&tc_hook, &tc_opts);
293 if (!ASSERT_OK(ret, "bpf_tc_attach"))
294 goto close;
295
296 xdp_prog = bpf_object__find_program_by_name(skel->obj, "ing_xdp");
297 if (!ASSERT_OK_PTR(xdp_prog, "open ing_xdp prog"))
298 goto close;
299
300 ret = bpf_xdp_attach(rx_ifindex,
301 bpf_program__fd(xdp_prog),
302 0, NULL);
303 if (!ASSERT_GE(ret, 0, "bpf_xdp_attach"))
304 goto close;
305
306 close_netns(nstoken);
307
308 nstoken = open_netns(TX_NETNS);
309 if (!ASSERT_OK_PTR(nstoken, "setns tx_ns"))
310 goto close;
311
312 SYS(close, "ip link set dev " TX_NAME " up");
313
314 tx_ifindex = if_nametoindex(TX_NAME);
315 if (!ASSERT_GE(tx_ifindex, 0, "if_nametoindex tx"))
316 goto close;
317
318 skel->bss->test_pass = false;
319
320 ret = send_test_packet(tx_ifindex);
321 if (!ASSERT_OK(ret, "send_test_packet"))
322 goto close;
323
324 if (!ASSERT_TRUE(skel->bss->test_pass, "test_pass"))
325 dump_err_stream(tc_prog);
326
327 close:
328 close_netns(nstoken);
329 test_xdp_meta__destroy(skel);
330 netns_free(rx_ns);
331 netns_free(tx_ns);
332 }
333
test_tuntap(struct bpf_program * xdp_prog,struct bpf_program * tc_prio_1_prog,struct bpf_program * tc_prio_2_prog,bool * test_pass)334 static void test_tuntap(struct bpf_program *xdp_prog,
335 struct bpf_program *tc_prio_1_prog,
336 struct bpf_program *tc_prio_2_prog,
337 bool *test_pass)
338 {
339 LIBBPF_OPTS(bpf_tc_hook, tc_hook, .attach_point = BPF_TC_INGRESS);
340 LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 1);
341 struct netns_obj *ns = NULL;
342 int tap_fd = -1;
343 int tap_ifindex;
344 int ret;
345
346 *test_pass = false;
347
348 ns = netns_new(TAP_NETNS, true);
349 if (!ASSERT_OK_PTR(ns, "create and open ns"))
350 return;
351
352 tap_fd = open_tuntap(TAP_NAME, true);
353 if (!ASSERT_GE(tap_fd, 0, "open_tuntap"))
354 goto close;
355
356 SYS(close, "ip link set dev " TAP_NAME " up");
357
358 tap_ifindex = if_nametoindex(TAP_NAME);
359 if (!ASSERT_GE(tap_ifindex, 0, "if_nametoindex"))
360 goto close;
361
362 tc_hook.ifindex = tap_ifindex;
363 ret = bpf_tc_hook_create(&tc_hook);
364 if (!ASSERT_OK(ret, "bpf_tc_hook_create"))
365 goto close;
366
367 tc_opts.prog_fd = bpf_program__fd(tc_prio_1_prog);
368 ret = bpf_tc_attach(&tc_hook, &tc_opts);
369 if (!ASSERT_OK(ret, "bpf_tc_attach"))
370 goto close;
371
372 if (tc_prio_2_prog) {
373 LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 2,
374 .prog_fd = bpf_program__fd(tc_prio_2_prog));
375
376 ret = bpf_tc_attach(&tc_hook, &tc_opts);
377 if (!ASSERT_OK(ret, "bpf_tc_attach"))
378 goto close;
379 }
380
381 ret = bpf_xdp_attach(tap_ifindex, bpf_program__fd(xdp_prog),
382 0, NULL);
383 if (!ASSERT_GE(ret, 0, "bpf_xdp_attach"))
384 goto close;
385
386 ret = write_test_packet(tap_fd);
387 if (!ASSERT_OK(ret, "write_test_packet"))
388 goto close;
389
390 if (!ASSERT_TRUE(*test_pass, "test_pass"))
391 dump_err_stream(tc_prio_2_prog ? : tc_prio_1_prog);
392
393 close:
394 if (tap_fd >= 0)
395 close(tap_fd);
396 netns_free(ns);
397 }
398
399 /* Write a packet to a tap dev and copy it to ingress of a dummy dev */
test_tuntap_mirred(struct bpf_program * xdp_prog,struct bpf_program * tc_prog,bool * test_pass)400 static void test_tuntap_mirred(struct bpf_program *xdp_prog,
401 struct bpf_program *tc_prog,
402 bool *test_pass)
403 {
404 LIBBPF_OPTS(bpf_tc_hook, tc_hook, .attach_point = BPF_TC_INGRESS);
405 LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 1);
406 struct netns_obj *ns = NULL;
407 int dummy_ifindex;
408 int tap_fd = -1;
409 int tap_ifindex;
410 int ret;
411
412 *test_pass = false;
413
414 ns = netns_new(TAP_NETNS, true);
415 if (!ASSERT_OK_PTR(ns, "netns_new"))
416 return;
417
418 /* Setup dummy interface */
419 SYS(close, "ip link add name " DUMMY_NAME " type dummy");
420 SYS(close, "ip link set dev " DUMMY_NAME " up");
421
422 dummy_ifindex = if_nametoindex(DUMMY_NAME);
423 if (!ASSERT_GE(dummy_ifindex, 0, "if_nametoindex"))
424 goto close;
425
426 tc_hook.ifindex = dummy_ifindex;
427 ret = bpf_tc_hook_create(&tc_hook);
428 if (!ASSERT_OK(ret, "bpf_tc_hook_create"))
429 goto close;
430
431 tc_opts.prog_fd = bpf_program__fd(tc_prog);
432 ret = bpf_tc_attach(&tc_hook, &tc_opts);
433 if (!ASSERT_OK(ret, "bpf_tc_attach"))
434 goto close;
435
436 /* Setup TAP interface */
437 tap_fd = open_tuntap(TAP_NAME, true);
438 if (!ASSERT_GE(tap_fd, 0, "open_tuntap"))
439 goto close;
440
441 SYS(close, "ip link set dev " TAP_NAME " up");
442
443 tap_ifindex = if_nametoindex(TAP_NAME);
444 if (!ASSERT_GE(tap_ifindex, 0, "if_nametoindex"))
445 goto close;
446
447 ret = bpf_xdp_attach(tap_ifindex, bpf_program__fd(xdp_prog), 0, NULL);
448 if (!ASSERT_GE(ret, 0, "bpf_xdp_attach"))
449 goto close;
450
451 /* Copy all packets received from TAP to dummy ingress */
452 SYS(close, "tc qdisc add dev " TAP_NAME " clsact");
453 SYS(close, "tc filter add dev " TAP_NAME " ingress "
454 "protocol all matchall "
455 "action mirred ingress mirror dev " DUMMY_NAME);
456
457 /* Receive a packet on TAP */
458 ret = write_test_packet(tap_fd);
459 if (!ASSERT_OK(ret, "write_test_packet"))
460 goto close;
461
462 if (!ASSERT_TRUE(*test_pass, "test_pass"))
463 dump_err_stream(tc_prog);
464
465 close:
466 if (tap_fd >= 0)
467 close(tap_fd);
468 netns_free(ns);
469 }
470
test_xdp_context_tuntap(void)471 void test_xdp_context_tuntap(void)
472 {
473 struct test_xdp_meta *skel = NULL;
474
475 skel = test_xdp_meta__open_and_load();
476 if (!ASSERT_OK_PTR(skel, "open and load skeleton"))
477 return;
478
479 if (test__start_subtest("data_meta"))
480 test_tuntap(skel->progs.ing_xdp,
481 skel->progs.ing_cls,
482 NULL, /* tc prio 2 */
483 &skel->bss->test_pass);
484 if (test__start_subtest("dynptr_read"))
485 test_tuntap(skel->progs.ing_xdp,
486 skel->progs.ing_cls_dynptr_read,
487 NULL, /* tc prio 2 */
488 &skel->bss->test_pass);
489 if (test__start_subtest("dynptr_slice"))
490 test_tuntap(skel->progs.ing_xdp,
491 skel->progs.ing_cls_dynptr_slice,
492 NULL, /* tc prio 2 */
493 &skel->bss->test_pass);
494 if (test__start_subtest("dynptr_write"))
495 test_tuntap(skel->progs.ing_xdp_zalloc_meta,
496 skel->progs.ing_cls_dynptr_write,
497 skel->progs.ing_cls_dynptr_read,
498 &skel->bss->test_pass);
499 if (test__start_subtest("dynptr_slice_rdwr"))
500 test_tuntap(skel->progs.ing_xdp_zalloc_meta,
501 skel->progs.ing_cls_dynptr_slice_rdwr,
502 skel->progs.ing_cls_dynptr_slice,
503 &skel->bss->test_pass);
504 if (test__start_subtest("dynptr_offset"))
505 test_tuntap(skel->progs.ing_xdp_zalloc_meta,
506 skel->progs.ing_cls_dynptr_offset_wr,
507 skel->progs.ing_cls_dynptr_offset_rd,
508 &skel->bss->test_pass);
509 if (test__start_subtest("dynptr_offset_oob"))
510 test_tuntap(skel->progs.ing_xdp,
511 skel->progs.ing_cls_dynptr_offset_oob,
512 skel->progs.ing_cls,
513 &skel->bss->test_pass);
514 if (test__start_subtest("clone_data_meta_survives_data_write"))
515 test_tuntap_mirred(skel->progs.ing_xdp,
516 skel->progs.clone_data_meta_survives_data_write,
517 &skel->bss->test_pass);
518 if (test__start_subtest("clone_data_meta_survives_meta_write"))
519 test_tuntap_mirred(skel->progs.ing_xdp,
520 skel->progs.clone_data_meta_survives_meta_write,
521 &skel->bss->test_pass);
522 if (test__start_subtest("clone_meta_dynptr_survives_data_slice_write"))
523 test_tuntap_mirred(skel->progs.ing_xdp,
524 skel->progs.clone_meta_dynptr_survives_data_slice_write,
525 &skel->bss->test_pass);
526 if (test__start_subtest("clone_meta_dynptr_survives_meta_slice_write"))
527 test_tuntap_mirred(skel->progs.ing_xdp,
528 skel->progs.clone_meta_dynptr_survives_meta_slice_write,
529 &skel->bss->test_pass);
530 if (test__start_subtest("clone_meta_dynptr_rw_before_data_dynptr_write"))
531 test_tuntap_mirred(skel->progs.ing_xdp,
532 skel->progs.clone_meta_dynptr_rw_before_data_dynptr_write,
533 &skel->bss->test_pass);
534 if (test__start_subtest("clone_meta_dynptr_rw_before_meta_dynptr_write"))
535 test_tuntap_mirred(skel->progs.ing_xdp,
536 skel->progs.clone_meta_dynptr_rw_before_meta_dynptr_write,
537 &skel->bss->test_pass);
538 /* Tests for BPF helpers which touch headroom */
539 if (test__start_subtest("helper_skb_vlan_push_pop"))
540 test_tuntap(skel->progs.ing_xdp,
541 skel->progs.helper_skb_vlan_push_pop,
542 NULL, /* tc prio 2 */
543 &skel->bss->test_pass);
544 if (test__start_subtest("helper_skb_adjust_room"))
545 test_tuntap(skel->progs.ing_xdp,
546 skel->progs.helper_skb_adjust_room,
547 NULL, /* tc prio 2 */
548 &skel->bss->test_pass);
549 if (test__start_subtest("helper_skb_change_head_tail"))
550 test_tuntap(skel->progs.ing_xdp,
551 skel->progs.helper_skb_change_head_tail,
552 NULL, /* tc prio 2 */
553 &skel->bss->test_pass);
554 if (test__start_subtest("helper_skb_change_proto"))
555 test_tuntap(skel->progs.ing_xdp,
556 skel->progs.helper_skb_change_proto,
557 NULL, /* tc prio 2 */
558 &skel->bss->test_pass);
559
560 test_xdp_meta__destroy(skel);
561 }
562
563 /*
564 * Test topology:
565 *
566 * tap0 fd00::1
567 * RX: injected IPv6 UDP frame, XDP ingress sets metadata
568 * fwd: encap route prepends outer header(s)
569 * TX: TC egress validates metadata
570 *
571 * A routable IPv6 UDP frame is written into the tap fd, so it enters the RX
572 * path where XDP stores metadata. Routing then forwards it back out the same
573 * tap through an encapsulating route that prepends outer header(s). The TC
574 * egress program checks that the pushed header did not silently corrupt
575 * metadata.
576 */
577 #define LWT_PIN_PATH "/sys/fs/bpf/xdp_context_lwt_xmit"
578
579 enum lwt_encap_type {
580 LWT_ENCAP_BPF,
581 LWT_ENCAP_MPLS,
582 LWT_ENCAP_SEG6,
583 LWT_ENCAP_IOAM6,
584 };
585
test_lwt_encap(struct test_xdp_meta * skel,enum lwt_encap_type type)586 static void test_lwt_encap(struct test_xdp_meta *skel,
587 enum lwt_encap_type type)
588 {
589 LIBBPF_OPTS(bpf_tc_hook, tc_hook, .attach_point = BPF_TC_EGRESS);
590 LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 1);
591 struct bpf_program *lwt_prog = NULL;
592 struct netns_obj *ns = NULL;
593 const char *encap;
594 bool pinned = false;
595 int tap_ifindex;
596 int tap_fd = -1;
597 int ret;
598
599 skel->bss->test_pass = false;
600
601 switch (type) {
602 case LWT_ENCAP_BPF:
603 encap = "encap bpf xmit pinned " LWT_PIN_PATH " via fd00::2";
604 lwt_prog = skel->progs.dummy_lwt_xmit;
605 break;
606 case LWT_ENCAP_MPLS:
607 encap = "encap mpls 100 via inet6 fd00::2";
608 break;
609 case LWT_ENCAP_SEG6:
610 encap = "encap seg6 mode encap segs fd00::2";
611 break;
612 case LWT_ENCAP_IOAM6:
613 encap = "encap ioam6 mode encap tundst fd00::2 "
614 "trace prealloc type 0x800000 ns 0 size 4 via fd00::2";
615 break;
616 default:
617 return;
618 }
619
620 if (lwt_prog) {
621 unlink(LWT_PIN_PATH);
622 ret = bpf_program__pin(lwt_prog, LWT_PIN_PATH);
623 if (!ASSERT_OK(ret, "pin lwt prog"))
624 return;
625 pinned = true;
626 }
627
628 ns = netns_new(LWT_NETNS, true);
629 if (!ASSERT_OK_PTR(ns, "netns_new"))
630 goto close;
631
632 tap_fd = open_tuntap(TAP_NAME, true);
633 if (!ASSERT_GE(tap_fd, 0, "open_tuntap"))
634 goto close;
635
636 SYS(close, "ip link set dev " TAP_NAME " address " RX_MAC);
637 SYS(close, "sysctl -wq net.ipv6.conf.all.forwarding=1");
638 SYS(close, "ip addr add fd00::1/64 dev " TAP_NAME " nodad");
639 SYS(close, "ip link set dev " TAP_NAME " up");
640 SYS(close, "ip neigh add fd00::2 lladdr " TX_MAC " nud permanent dev " TAP_NAME);
641 SYS(close, "ip -6 route add fd00:1::/64 %s dev %s", encap, TAP_NAME);
642
643 tap_ifindex = if_nametoindex(TAP_NAME);
644 if (!ASSERT_GE(tap_ifindex, 0, "if_nametoindex"))
645 goto close;
646
647 ret = bpf_xdp_attach(tap_ifindex, bpf_program__fd(skel->progs.ing_xdp),
648 0, NULL);
649 if (!ASSERT_GE(ret, 0, "bpf_xdp_attach"))
650 goto close;
651
652 tc_hook.ifindex = tap_ifindex;
653 ret = bpf_tc_hook_create(&tc_hook);
654 if (!ASSERT_OK(ret, "bpf_tc_hook_create"))
655 goto close;
656
657 tc_opts.prog_fd = bpf_program__fd(skel->progs.tc_is_meta_empty);
658 ret = bpf_tc_attach(&tc_hook, &tc_opts);
659 if (!ASSERT_OK(ret, "bpf_tc_attach"))
660 goto close;
661
662 ret = write_test_packet_udp(tap_fd);
663 if (!ASSERT_OK(ret, "write_test_packet_udp"))
664 goto close;
665
666 if (!ASSERT_TRUE(skel->bss->test_pass, "test_pass"))
667 dump_err_stream(skel->progs.tc_is_meta_empty);
668
669 close:
670 if (tap_fd >= 0)
671 close(tap_fd);
672 netns_free(ns);
673 if (pinned)
674 unlink(LWT_PIN_PATH);
675 }
676
test_xdp_context_lwt_encap(void)677 void test_xdp_context_lwt_encap(void)
678 {
679 struct test_xdp_meta *skel;
680
681 skel = test_xdp_meta__open_and_load();
682 if (!ASSERT_OK_PTR(skel, "open and load skeleton"))
683 return;
684
685 if (test__start_subtest("bpf_encap"))
686 test_lwt_encap(skel, LWT_ENCAP_BPF);
687 if (test__start_subtest("mpls_encap"))
688 test_lwt_encap(skel, LWT_ENCAP_MPLS);
689 if (test__start_subtest("seg6_encap"))
690 test_lwt_encap(skel, LWT_ENCAP_SEG6);
691 if (test__start_subtest("ioam6_encap"))
692 test_lwt_encap(skel, LWT_ENCAP_IOAM6);
693
694 test_xdp_meta__destroy(skel);
695 }
696