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, bool quiet) 134 { 135 int fd; 136 137 fd = bpf_obj_get(path); 138 if (fd < 0) { 139 if (!quiet) 140 p_err("bpf obj get (%s): %s", path, 141 errno == EACCES && !is_bpffs(dirname(path)) ? 142 "directory not in bpf file system (bpffs)" : 143 strerror(errno)); 144 return -1; 145 } 146 147 return fd; 148 } 149 150 int open_obj_pinned_any(char *path, enum bpf_obj_type exp_type) 151 { 152 enum bpf_obj_type type; 153 int fd; 154 155 fd = open_obj_pinned(path, false); 156 if (fd < 0) 157 return -1; 158 159 type = get_fd_type(fd); 160 if (type < 0) { 161 close(fd); 162 return type; 163 } 164 if (type != exp_type) { 165 p_err("incorrect object type: %s", get_fd_type_name(type)); 166 close(fd); 167 return -1; 168 } 169 170 return fd; 171 } 172 173 int do_pin_fd(int fd, const char *name) 174 { 175 char err_str[ERR_MAX_LEN]; 176 char *file; 177 char *dir; 178 int err = 0; 179 180 err = bpf_obj_pin(fd, name); 181 if (!err) 182 goto out; 183 184 file = malloc(strlen(name) + 1); 185 strcpy(file, name); 186 dir = dirname(file); 187 188 if (errno != EPERM || is_bpffs(dir)) { 189 p_err("can't pin the object (%s): %s", name, strerror(errno)); 190 goto out_free; 191 } 192 193 /* Attempt to mount bpffs, then retry pinning. */ 194 err = mnt_bpffs(dir, err_str, ERR_MAX_LEN); 195 if (!err) { 196 err = bpf_obj_pin(fd, name); 197 if (err) 198 p_err("can't pin the object (%s): %s", name, 199 strerror(errno)); 200 } else { 201 err_str[ERR_MAX_LEN - 1] = '\0'; 202 p_err("can't mount BPF file system to pin the object (%s): %s", 203 name, err_str); 204 } 205 206 out_free: 207 free(file); 208 out: 209 return err; 210 } 211 212 int do_pin_any(int argc, char **argv, int (*get_fd_by_id)(__u32)) 213 { 214 unsigned int id; 215 char *endptr; 216 int err; 217 int fd; 218 219 if (argc < 3) { 220 p_err("too few arguments, id ID and FILE path is required"); 221 return -1; 222 } else if (argc > 3) { 223 p_err("too many arguments"); 224 return -1; 225 } 226 227 if (!is_prefix(*argv, "id")) { 228 p_err("expected 'id' got %s", *argv); 229 return -1; 230 } 231 NEXT_ARG(); 232 233 id = strtoul(*argv, &endptr, 0); 234 if (*endptr) { 235 p_err("can't parse %s as ID", *argv); 236 return -1; 237 } 238 NEXT_ARG(); 239 240 fd = get_fd_by_id(id); 241 if (fd < 0) { 242 p_err("can't get prog by id (%u): %s", id, strerror(errno)); 243 return -1; 244 } 245 246 err = do_pin_fd(fd, *argv); 247 248 close(fd); 249 return err; 250 } 251 252 const char *get_fd_type_name(enum bpf_obj_type type) 253 { 254 static const char * const names[] = { 255 [BPF_OBJ_UNKNOWN] = "unknown", 256 [BPF_OBJ_PROG] = "prog", 257 [BPF_OBJ_MAP] = "map", 258 }; 259 260 if (type < 0 || type >= ARRAY_SIZE(names) || !names[type]) 261 return names[BPF_OBJ_UNKNOWN]; 262 263 return names[type]; 264 } 265 266 int get_fd_type(int fd) 267 { 268 char path[PATH_MAX]; 269 char buf[512]; 270 ssize_t n; 271 272 snprintf(path, sizeof(path), "/proc/%d/fd/%d", getpid(), fd); 273 274 n = readlink(path, buf, sizeof(buf)); 275 if (n < 0) { 276 p_err("can't read link type: %s", strerror(errno)); 277 return -1; 278 } 279 if (n == sizeof(path)) { 280 p_err("can't read link type: path too long!"); 281 return -1; 282 } 283 284 if (strstr(buf, "bpf-map")) 285 return BPF_OBJ_MAP; 286 else if (strstr(buf, "bpf-prog")) 287 return BPF_OBJ_PROG; 288 289 return BPF_OBJ_UNKNOWN; 290 } 291 292 char *get_fdinfo(int fd, const char *key) 293 { 294 char path[PATH_MAX]; 295 char *line = NULL; 296 size_t line_n = 0; 297 ssize_t n; 298 FILE *fdi; 299 300 snprintf(path, sizeof(path), "/proc/%d/fdinfo/%d", getpid(), fd); 301 302 fdi = fopen(path, "r"); 303 if (!fdi) { 304 p_err("can't open fdinfo: %s", strerror(errno)); 305 return NULL; 306 } 307 308 while ((n = getline(&line, &line_n, fdi)) > 0) { 309 char *value; 310 int len; 311 312 if (!strstr(line, key)) 313 continue; 314 315 fclose(fdi); 316 317 value = strchr(line, '\t'); 318 if (!value || !value[1]) { 319 p_err("malformed fdinfo!?"); 320 free(line); 321 return NULL; 322 } 323 value++; 324 325 len = strlen(value); 326 memmove(line, value, len); 327 line[len - 1] = '\0'; 328 329 return line; 330 } 331 332 p_err("key '%s' not found in fdinfo", key); 333 free(line); 334 fclose(fdi); 335 return NULL; 336 } 337 338 void print_data_json(uint8_t *data, size_t len) 339 { 340 unsigned int i; 341 342 jsonw_start_array(json_wtr); 343 for (i = 0; i < len; i++) 344 jsonw_printf(json_wtr, "%d", data[i]); 345 jsonw_end_array(json_wtr); 346 } 347 348 void print_hex_data_json(uint8_t *data, size_t len) 349 { 350 unsigned int i; 351 352 jsonw_start_array(json_wtr); 353 for (i = 0; i < len; i++) 354 jsonw_printf(json_wtr, "\"0x%02hhx\"", data[i]); 355 jsonw_end_array(json_wtr); 356 } 357 358 int build_pinned_obj_table(struct pinned_obj_table *tab, 359 enum bpf_obj_type type) 360 { 361 struct bpf_prog_info pinned_info = {}; 362 struct pinned_obj *obj_node = NULL; 363 __u32 len = sizeof(pinned_info); 364 struct mntent *mntent = NULL; 365 enum bpf_obj_type objtype; 366 FILE *mntfile = NULL; 367 FTSENT *ftse = NULL; 368 FTS *fts = NULL; 369 int fd, err; 370 371 mntfile = setmntent("/proc/mounts", "r"); 372 if (!mntfile) 373 return -1; 374 375 while ((mntent = getmntent(mntfile))) { 376 char *path[] = { mntent->mnt_dir, NULL }; 377 378 if (strncmp(mntent->mnt_type, "bpf", 3) != 0) 379 continue; 380 381 fts = fts_open(path, 0, NULL); 382 if (!fts) 383 continue; 384 385 while ((ftse = fts_read(fts))) { 386 if (!(ftse->fts_info & FTS_F)) 387 continue; 388 fd = open_obj_pinned(ftse->fts_path, true); 389 if (fd < 0) 390 continue; 391 392 objtype = get_fd_type(fd); 393 if (objtype != type) { 394 close(fd); 395 continue; 396 } 397 memset(&pinned_info, 0, sizeof(pinned_info)); 398 err = bpf_obj_get_info_by_fd(fd, &pinned_info, &len); 399 if (err) { 400 close(fd); 401 continue; 402 } 403 404 obj_node = malloc(sizeof(*obj_node)); 405 if (!obj_node) { 406 close(fd); 407 fts_close(fts); 408 fclose(mntfile); 409 return -1; 410 } 411 412 memset(obj_node, 0, sizeof(*obj_node)); 413 obj_node->id = pinned_info.id; 414 obj_node->path = strdup(ftse->fts_path); 415 hash_add(tab->table, &obj_node->hash, obj_node->id); 416 417 close(fd); 418 } 419 fts_close(fts); 420 } 421 fclose(mntfile); 422 return 0; 423 } 424 425 void delete_pinned_obj_table(struct pinned_obj_table *tab) 426 { 427 struct pinned_obj *obj; 428 struct hlist_node *tmp; 429 unsigned int bkt; 430 431 hash_for_each_safe(tab->table, bkt, tmp, obj, hash) { 432 hash_del(&obj->hash); 433 free(obj->path); 434 free(obj); 435 } 436 } 437 438 unsigned int get_page_size(void) 439 { 440 static int result; 441 442 if (!result) 443 result = getpagesize(); 444 return result; 445 } 446 447 unsigned int get_possible_cpus(void) 448 { 449 static unsigned int result; 450 char buf[128]; 451 long int n; 452 char *ptr; 453 int fd; 454 455 if (result) 456 return result; 457 458 fd = open("/sys/devices/system/cpu/possible", O_RDONLY); 459 if (fd < 0) { 460 p_err("can't open sysfs possible cpus"); 461 exit(-1); 462 } 463 464 n = read(fd, buf, sizeof(buf)); 465 if (n < 2) { 466 p_err("can't read sysfs possible cpus"); 467 exit(-1); 468 } 469 close(fd); 470 471 if (n == sizeof(buf)) { 472 p_err("read sysfs possible cpus overflow"); 473 exit(-1); 474 } 475 476 ptr = buf; 477 n = 0; 478 while (*ptr && *ptr != '\n') { 479 unsigned int a, b; 480 481 if (sscanf(ptr, "%u-%u", &a, &b) == 2) { 482 n += b - a + 1; 483 484 ptr = strchr(ptr, '-') + 1; 485 } else if (sscanf(ptr, "%u", &a) == 1) { 486 n++; 487 } else { 488 assert(0); 489 } 490 491 while (isdigit(*ptr)) 492 ptr++; 493 if (*ptr == ',') 494 ptr++; 495 } 496 497 result = n; 498 499 return result; 500 } 501 502 static char * 503 ifindex_to_name_ns(__u32 ifindex, __u32 ns_dev, __u32 ns_ino, char *buf) 504 { 505 struct stat st; 506 int err; 507 508 err = stat("/proc/self/ns/net", &st); 509 if (err) { 510 p_err("Can't stat /proc/self: %s", strerror(errno)); 511 return NULL; 512 } 513 514 if (st.st_dev != ns_dev || st.st_ino != ns_ino) 515 return NULL; 516 517 return if_indextoname(ifindex, buf); 518 } 519 520 static int read_sysfs_hex_int(char *path) 521 { 522 char vendor_id_buf[8]; 523 int len; 524 int fd; 525 526 fd = open(path, O_RDONLY); 527 if (fd < 0) { 528 p_err("Can't open %s: %s", path, strerror(errno)); 529 return -1; 530 } 531 532 len = read(fd, vendor_id_buf, sizeof(vendor_id_buf)); 533 close(fd); 534 if (len < 0) { 535 p_err("Can't read %s: %s", path, strerror(errno)); 536 return -1; 537 } 538 if (len >= (int)sizeof(vendor_id_buf)) { 539 p_err("Value in %s too long", path); 540 return -1; 541 } 542 543 vendor_id_buf[len] = 0; 544 545 return strtol(vendor_id_buf, NULL, 0); 546 } 547 548 static int read_sysfs_netdev_hex_int(char *devname, const char *entry_name) 549 { 550 char full_path[64]; 551 552 snprintf(full_path, sizeof(full_path), "/sys/class/net/%s/device/%s", 553 devname, entry_name); 554 555 return read_sysfs_hex_int(full_path); 556 } 557 558 const char * 559 ifindex_to_bfd_params(__u32 ifindex, __u64 ns_dev, __u64 ns_ino, 560 const char **opt) 561 { 562 char devname[IF_NAMESIZE]; 563 int vendor_id; 564 int device_id; 565 566 if (!ifindex_to_name_ns(ifindex, ns_dev, ns_ino, devname)) { 567 p_err("Can't get net device name for ifindex %d: %s", ifindex, 568 strerror(errno)); 569 return NULL; 570 } 571 572 vendor_id = read_sysfs_netdev_hex_int(devname, "vendor"); 573 if (vendor_id < 0) { 574 p_err("Can't get device vendor id for %s", devname); 575 return NULL; 576 } 577 578 switch (vendor_id) { 579 case 0x19ee: 580 device_id = read_sysfs_netdev_hex_int(devname, "device"); 581 if (device_id != 0x4000 && 582 device_id != 0x6000 && 583 device_id != 0x6003) 584 p_info("Unknown NFP device ID, assuming it is NFP-6xxx arch"); 585 *opt = "ctx4"; 586 return "NFP-6xxx"; 587 default: 588 p_err("Can't get bfd arch name for device vendor id 0x%04x", 589 vendor_id); 590 return NULL; 591 } 592 } 593 594 void print_dev_plain(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 595 { 596 char name[IF_NAMESIZE]; 597 598 if (!ifindex) 599 return; 600 601 printf(" dev "); 602 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 603 printf("%s", name); 604 else 605 printf("ifindex %u ns_dev %llu ns_ino %llu", 606 ifindex, ns_dev, ns_inode); 607 } 608 609 void print_dev_json(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 610 { 611 char name[IF_NAMESIZE]; 612 613 if (!ifindex) 614 return; 615 616 jsonw_name(json_wtr, "dev"); 617 jsonw_start_object(json_wtr); 618 jsonw_uint_field(json_wtr, "ifindex", ifindex); 619 jsonw_uint_field(json_wtr, "ns_dev", ns_dev); 620 jsonw_uint_field(json_wtr, "ns_inode", ns_inode); 621 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 622 jsonw_string_field(json_wtr, "ifname", name); 623 jsonw_end_object(json_wtr); 624 } 625 626 int parse_u32_arg(int *argc, char ***argv, __u32 *val, const char *what) 627 { 628 char *endptr; 629 630 NEXT_ARGP(); 631 632 if (*val) { 633 p_err("%s already specified", what); 634 return -1; 635 } 636 637 *val = strtoul(**argv, &endptr, 0); 638 if (*endptr) { 639 p_err("can't parse %s as %s", **argv, what); 640 return -1; 641 } 642 NEXT_ARGP(); 643 644 return 0; 645 } 646