1 /* 2 * Copyright (C) 2017-2018 Netronome Systems, Inc. 3 * 4 * This software is dual licensed under the GNU General License Version 2, 5 * June 1991 as shown in the file COPYING in the top-level directory of this 6 * source tree or the BSD 2-Clause License provided below. You have the 7 * option to license this software under the complete terms of either license. 8 * 9 * The BSD 2-Clause License: 10 * 11 * Redistribution and use in source and binary forms, with or 12 * without modification, are permitted provided that the following 13 * conditions are met: 14 * 15 * 1. Redistributions of source code must retain the above 16 * copyright notice, this list of conditions and the following 17 * disclaimer. 18 * 19 * 2. Redistributions in binary form must reproduce the above 20 * copyright notice, this list of conditions and the following 21 * disclaimer in the documentation and/or other materials 22 * provided with the distribution. 23 * 24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 25 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 26 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 27 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 28 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 29 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 30 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 31 * SOFTWARE. 32 */ 33 34 #include <ctype.h> 35 #include <errno.h> 36 #include <fcntl.h> 37 #include <fts.h> 38 #include <libgen.h> 39 #include <mntent.h> 40 #include <stdbool.h> 41 #include <stdio.h> 42 #include <stdlib.h> 43 #include <string.h> 44 #include <unistd.h> 45 #include <linux/limits.h> 46 #include <linux/magic.h> 47 #include <net/if.h> 48 #include <sys/mount.h> 49 #include <sys/stat.h> 50 #include <sys/types.h> 51 #include <sys/vfs.h> 52 53 #include <bpf.h> 54 55 #include "main.h" 56 57 #ifndef BPF_FS_MAGIC 58 #define BPF_FS_MAGIC 0xcafe4a11 59 #endif 60 61 void p_err(const char *fmt, ...) 62 { 63 va_list ap; 64 65 va_start(ap, fmt); 66 if (json_output) { 67 jsonw_start_object(json_wtr); 68 jsonw_name(json_wtr, "error"); 69 jsonw_vprintf_enquote(json_wtr, fmt, ap); 70 jsonw_end_object(json_wtr); 71 } else { 72 fprintf(stderr, "Error: "); 73 vfprintf(stderr, fmt, ap); 74 fprintf(stderr, "\n"); 75 } 76 va_end(ap); 77 } 78 79 void p_info(const char *fmt, ...) 80 { 81 va_list ap; 82 83 if (json_output) 84 return; 85 86 va_start(ap, fmt); 87 vfprintf(stderr, fmt, ap); 88 fprintf(stderr, "\n"); 89 va_end(ap); 90 } 91 92 static bool is_bpffs(char *path) 93 { 94 struct statfs st_fs; 95 96 if (statfs(path, &st_fs) < 0) 97 return false; 98 99 return (unsigned long)st_fs.f_type == BPF_FS_MAGIC; 100 } 101 102 static int mnt_bpffs(const char *target, char *buff, size_t bufflen) 103 { 104 bool bind_done = false; 105 106 while (mount("", target, "none", MS_PRIVATE | MS_REC, NULL)) { 107 if (errno != EINVAL || bind_done) { 108 snprintf(buff, bufflen, 109 "mount --make-private %s failed: %s", 110 target, strerror(errno)); 111 return -1; 112 } 113 114 if (mount(target, target, "none", MS_BIND, NULL)) { 115 snprintf(buff, bufflen, 116 "mount --bind %s %s failed: %s", 117 target, target, strerror(errno)); 118 return -1; 119 } 120 121 bind_done = true; 122 } 123 124 if (mount("bpf", target, "bpf", 0, "mode=0700")) { 125 snprintf(buff, bufflen, "mount -t bpf bpf %s failed: %s", 126 target, strerror(errno)); 127 return -1; 128 } 129 130 return 0; 131 } 132 133 int open_obj_pinned(char *path) 134 { 135 int fd; 136 137 fd = bpf_obj_get(path); 138 if (fd < 0) { 139 p_err("bpf obj get (%s): %s", path, 140 errno == EACCES && !is_bpffs(dirname(path)) ? 141 "directory not in bpf file system (bpffs)" : 142 strerror(errno)); 143 return -1; 144 } 145 146 return fd; 147 } 148 149 int open_obj_pinned_any(char *path, enum bpf_obj_type exp_type) 150 { 151 enum bpf_obj_type type; 152 int fd; 153 154 fd = open_obj_pinned(path); 155 if (fd < 0) 156 return -1; 157 158 type = get_fd_type(fd); 159 if (type < 0) { 160 close(fd); 161 return type; 162 } 163 if (type != exp_type) { 164 p_err("incorrect object type: %s", get_fd_type_name(type)); 165 close(fd); 166 return -1; 167 } 168 169 return fd; 170 } 171 172 int do_pin_fd(int fd, const char *name) 173 { 174 char err_str[ERR_MAX_LEN]; 175 char *file; 176 char *dir; 177 int err = 0; 178 179 err = bpf_obj_pin(fd, name); 180 if (!err) 181 goto out; 182 183 file = malloc(strlen(name) + 1); 184 strcpy(file, name); 185 dir = dirname(file); 186 187 if (errno != EPERM || is_bpffs(dir)) { 188 p_err("can't pin the object (%s): %s", name, strerror(errno)); 189 goto out_free; 190 } 191 192 /* Attempt to mount bpffs, then retry pinning. */ 193 err = mnt_bpffs(dir, err_str, ERR_MAX_LEN); 194 if (!err) { 195 err = bpf_obj_pin(fd, name); 196 if (err) 197 p_err("can't pin the object (%s): %s", name, 198 strerror(errno)); 199 } else { 200 err_str[ERR_MAX_LEN - 1] = '\0'; 201 p_err("can't mount BPF file system to pin the object (%s): %s", 202 name, err_str); 203 } 204 205 out_free: 206 free(file); 207 out: 208 return err; 209 } 210 211 int do_pin_any(int argc, char **argv, int (*get_fd_by_id)(__u32)) 212 { 213 unsigned int id; 214 char *endptr; 215 int err; 216 int fd; 217 218 if (!is_prefix(*argv, "id")) { 219 p_err("expected 'id' got %s", *argv); 220 return -1; 221 } 222 NEXT_ARG(); 223 224 id = strtoul(*argv, &endptr, 0); 225 if (*endptr) { 226 p_err("can't parse %s as ID", *argv); 227 return -1; 228 } 229 NEXT_ARG(); 230 231 if (argc != 1) 232 usage(); 233 234 fd = get_fd_by_id(id); 235 if (fd < 0) { 236 p_err("can't get prog by id (%u): %s", id, strerror(errno)); 237 return -1; 238 } 239 240 err = do_pin_fd(fd, *argv); 241 242 close(fd); 243 return err; 244 } 245 246 const char *get_fd_type_name(enum bpf_obj_type type) 247 { 248 static const char * const names[] = { 249 [BPF_OBJ_UNKNOWN] = "unknown", 250 [BPF_OBJ_PROG] = "prog", 251 [BPF_OBJ_MAP] = "map", 252 }; 253 254 if (type < 0 || type >= ARRAY_SIZE(names) || !names[type]) 255 return names[BPF_OBJ_UNKNOWN]; 256 257 return names[type]; 258 } 259 260 int get_fd_type(int fd) 261 { 262 char path[PATH_MAX]; 263 char buf[512]; 264 ssize_t n; 265 266 snprintf(path, sizeof(path), "/proc/%d/fd/%d", getpid(), fd); 267 268 n = readlink(path, buf, sizeof(buf)); 269 if (n < 0) { 270 p_err("can't read link type: %s", strerror(errno)); 271 return -1; 272 } 273 if (n == sizeof(path)) { 274 p_err("can't read link type: path too long!"); 275 return -1; 276 } 277 278 if (strstr(buf, "bpf-map")) 279 return BPF_OBJ_MAP; 280 else if (strstr(buf, "bpf-prog")) 281 return BPF_OBJ_PROG; 282 283 return BPF_OBJ_UNKNOWN; 284 } 285 286 char *get_fdinfo(int fd, const char *key) 287 { 288 char path[PATH_MAX]; 289 char *line = NULL; 290 size_t line_n = 0; 291 ssize_t n; 292 FILE *fdi; 293 294 snprintf(path, sizeof(path), "/proc/%d/fdinfo/%d", getpid(), fd); 295 296 fdi = fopen(path, "r"); 297 if (!fdi) { 298 p_err("can't open fdinfo: %s", strerror(errno)); 299 return NULL; 300 } 301 302 while ((n = getline(&line, &line_n, fdi))) { 303 char *value; 304 int len; 305 306 if (!strstr(line, key)) 307 continue; 308 309 fclose(fdi); 310 311 value = strchr(line, '\t'); 312 if (!value || !value[1]) { 313 p_err("malformed fdinfo!?"); 314 free(line); 315 return NULL; 316 } 317 value++; 318 319 len = strlen(value); 320 memmove(line, value, len); 321 line[len - 1] = '\0'; 322 323 return line; 324 } 325 326 p_err("key '%s' not found in fdinfo", key); 327 free(line); 328 fclose(fdi); 329 return NULL; 330 } 331 332 void print_data_json(uint8_t *data, size_t len) 333 { 334 unsigned int i; 335 336 jsonw_start_array(json_wtr); 337 for (i = 0; i < len; i++) 338 jsonw_printf(json_wtr, "%d", data[i]); 339 jsonw_end_array(json_wtr); 340 } 341 342 void print_hex_data_json(uint8_t *data, size_t len) 343 { 344 unsigned int i; 345 346 jsonw_start_array(json_wtr); 347 for (i = 0; i < len; i++) 348 jsonw_printf(json_wtr, "\"0x%02hhx\"", data[i]); 349 jsonw_end_array(json_wtr); 350 } 351 352 int build_pinned_obj_table(struct pinned_obj_table *tab, 353 enum bpf_obj_type type) 354 { 355 struct bpf_prog_info pinned_info = {}; 356 struct pinned_obj *obj_node = NULL; 357 __u32 len = sizeof(pinned_info); 358 struct mntent *mntent = NULL; 359 enum bpf_obj_type objtype; 360 FILE *mntfile = NULL; 361 FTSENT *ftse = NULL; 362 FTS *fts = NULL; 363 int fd, err; 364 365 mntfile = setmntent("/proc/mounts", "r"); 366 if (!mntfile) 367 return -1; 368 369 while ((mntent = getmntent(mntfile))) { 370 char *path[] = { mntent->mnt_dir, NULL }; 371 372 if (strncmp(mntent->mnt_type, "bpf", 3) != 0) 373 continue; 374 375 fts = fts_open(path, 0, NULL); 376 if (!fts) 377 continue; 378 379 while ((ftse = fts_read(fts))) { 380 if (!(ftse->fts_info & FTS_F)) 381 continue; 382 fd = open_obj_pinned(ftse->fts_path); 383 if (fd < 0) 384 continue; 385 386 objtype = get_fd_type(fd); 387 if (objtype != type) { 388 close(fd); 389 continue; 390 } 391 memset(&pinned_info, 0, sizeof(pinned_info)); 392 err = bpf_obj_get_info_by_fd(fd, &pinned_info, &len); 393 if (err) { 394 close(fd); 395 continue; 396 } 397 398 obj_node = malloc(sizeof(*obj_node)); 399 if (!obj_node) { 400 close(fd); 401 fts_close(fts); 402 fclose(mntfile); 403 return -1; 404 } 405 406 memset(obj_node, 0, sizeof(*obj_node)); 407 obj_node->id = pinned_info.id; 408 obj_node->path = strdup(ftse->fts_path); 409 hash_add(tab->table, &obj_node->hash, obj_node->id); 410 411 close(fd); 412 } 413 fts_close(fts); 414 } 415 fclose(mntfile); 416 return 0; 417 } 418 419 void delete_pinned_obj_table(struct pinned_obj_table *tab) 420 { 421 struct pinned_obj *obj; 422 struct hlist_node *tmp; 423 unsigned int bkt; 424 425 hash_for_each_safe(tab->table, bkt, tmp, obj, hash) { 426 hash_del(&obj->hash); 427 free(obj->path); 428 free(obj); 429 } 430 } 431 432 unsigned int get_page_size(void) 433 { 434 static int result; 435 436 if (!result) 437 result = getpagesize(); 438 return result; 439 } 440 441 unsigned int get_possible_cpus(void) 442 { 443 static unsigned int result; 444 char buf[128]; 445 long int n; 446 char *ptr; 447 int fd; 448 449 if (result) 450 return result; 451 452 fd = open("/sys/devices/system/cpu/possible", O_RDONLY); 453 if (fd < 0) { 454 p_err("can't open sysfs possible cpus"); 455 exit(-1); 456 } 457 458 n = read(fd, buf, sizeof(buf)); 459 if (n < 2) { 460 p_err("can't read sysfs possible cpus"); 461 exit(-1); 462 } 463 close(fd); 464 465 if (n == sizeof(buf)) { 466 p_err("read sysfs possible cpus overflow"); 467 exit(-1); 468 } 469 470 ptr = buf; 471 n = 0; 472 while (*ptr && *ptr != '\n') { 473 unsigned int a, b; 474 475 if (sscanf(ptr, "%u-%u", &a, &b) == 2) { 476 n += b - a + 1; 477 478 ptr = strchr(ptr, '-') + 1; 479 } else if (sscanf(ptr, "%u", &a) == 1) { 480 n++; 481 } else { 482 assert(0); 483 } 484 485 while (isdigit(*ptr)) 486 ptr++; 487 if (*ptr == ',') 488 ptr++; 489 } 490 491 result = n; 492 493 return result; 494 } 495 496 static char * 497 ifindex_to_name_ns(__u32 ifindex, __u32 ns_dev, __u32 ns_ino, char *buf) 498 { 499 struct stat st; 500 int err; 501 502 err = stat("/proc/self/ns/net", &st); 503 if (err) { 504 p_err("Can't stat /proc/self: %s", strerror(errno)); 505 return NULL; 506 } 507 508 if (st.st_dev != ns_dev || st.st_ino != ns_ino) 509 return NULL; 510 511 return if_indextoname(ifindex, buf); 512 } 513 514 static int read_sysfs_hex_int(char *path) 515 { 516 char vendor_id_buf[8]; 517 int len; 518 int fd; 519 520 fd = open(path, O_RDONLY); 521 if (fd < 0) { 522 p_err("Can't open %s: %s", path, strerror(errno)); 523 return -1; 524 } 525 526 len = read(fd, vendor_id_buf, sizeof(vendor_id_buf)); 527 close(fd); 528 if (len < 0) { 529 p_err("Can't read %s: %s", path, strerror(errno)); 530 return -1; 531 } 532 if (len >= (int)sizeof(vendor_id_buf)) { 533 p_err("Value in %s too long", path); 534 return -1; 535 } 536 537 vendor_id_buf[len] = 0; 538 539 return strtol(vendor_id_buf, NULL, 0); 540 } 541 542 static int read_sysfs_netdev_hex_int(char *devname, const char *entry_name) 543 { 544 char full_path[64]; 545 546 snprintf(full_path, sizeof(full_path), "/sys/class/net/%s/device/%s", 547 devname, entry_name); 548 549 return read_sysfs_hex_int(full_path); 550 } 551 552 const char *ifindex_to_bfd_name_ns(__u32 ifindex, __u64 ns_dev, __u64 ns_ino) 553 { 554 char devname[IF_NAMESIZE]; 555 int vendor_id; 556 int device_id; 557 558 if (!ifindex_to_name_ns(ifindex, ns_dev, ns_ino, devname)) { 559 p_err("Can't get net device name for ifindex %d: %s", ifindex, 560 strerror(errno)); 561 return NULL; 562 } 563 564 vendor_id = read_sysfs_netdev_hex_int(devname, "vendor"); 565 if (vendor_id < 0) { 566 p_err("Can't get device vendor id for %s", devname); 567 return NULL; 568 } 569 570 switch (vendor_id) { 571 case 0x19ee: 572 device_id = read_sysfs_netdev_hex_int(devname, "device"); 573 if (device_id != 0x4000 && 574 device_id != 0x6000 && 575 device_id != 0x6003) 576 p_info("Unknown NFP device ID, assuming it is NFP-6xxx arch"); 577 return "NFP-6xxx"; 578 default: 579 p_err("Can't get bfd arch name for device vendor id 0x%04x", 580 vendor_id); 581 return NULL; 582 } 583 } 584 585 void print_dev_plain(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 586 { 587 char name[IF_NAMESIZE]; 588 589 if (!ifindex) 590 return; 591 592 printf(" dev "); 593 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 594 printf("%s", name); 595 else 596 printf("ifindex %u ns_dev %llu ns_ino %llu", 597 ifindex, ns_dev, ns_inode); 598 } 599 600 void print_dev_json(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 601 { 602 char name[IF_NAMESIZE]; 603 604 if (!ifindex) 605 return; 606 607 jsonw_name(json_wtr, "dev"); 608 jsonw_start_object(json_wtr); 609 jsonw_uint_field(json_wtr, "ifindex", ifindex); 610 jsonw_uint_field(json_wtr, "ns_dev", ns_dev); 611 jsonw_uint_field(json_wtr, "ns_inode", ns_inode); 612 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 613 jsonw_string_field(json_wtr, "ifname", name); 614 jsonw_end_object(json_wtr); 615 } 616