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