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 /* Author: Jakub Kicinski <kubakici@wp.pl> */ 35 36 #include <ctype.h> 37 #include <errno.h> 38 #include <fcntl.h> 39 #include <fts.h> 40 #include <libgen.h> 41 #include <mntent.h> 42 #include <stdbool.h> 43 #include <stdio.h> 44 #include <stdlib.h> 45 #include <string.h> 46 #include <unistd.h> 47 #include <linux/limits.h> 48 #include <linux/magic.h> 49 #include <net/if.h> 50 #include <sys/mount.h> 51 #include <sys/stat.h> 52 #include <sys/types.h> 53 #include <sys/vfs.h> 54 55 #include <bpf.h> 56 57 #include "main.h" 58 59 #ifndef BPF_FS_MAGIC 60 #define BPF_FS_MAGIC 0xcafe4a11 61 #endif 62 63 void p_err(const char *fmt, ...) 64 { 65 va_list ap; 66 67 va_start(ap, fmt); 68 if (json_output) { 69 jsonw_start_object(json_wtr); 70 jsonw_name(json_wtr, "error"); 71 jsonw_vprintf_enquote(json_wtr, fmt, ap); 72 jsonw_end_object(json_wtr); 73 } else { 74 fprintf(stderr, "Error: "); 75 vfprintf(stderr, fmt, ap); 76 fprintf(stderr, "\n"); 77 } 78 va_end(ap); 79 } 80 81 void p_info(const char *fmt, ...) 82 { 83 va_list ap; 84 85 if (json_output) 86 return; 87 88 va_start(ap, fmt); 89 vfprintf(stderr, fmt, ap); 90 fprintf(stderr, "\n"); 91 va_end(ap); 92 } 93 94 static bool is_bpffs(char *path) 95 { 96 struct statfs st_fs; 97 98 if (statfs(path, &st_fs) < 0) 99 return false; 100 101 return (unsigned long)st_fs.f_type == BPF_FS_MAGIC; 102 } 103 104 static int mnt_bpffs(const char *target, char *buff, size_t bufflen) 105 { 106 bool bind_done = false; 107 108 while (mount("", target, "none", MS_PRIVATE | MS_REC, NULL)) { 109 if (errno != EINVAL || bind_done) { 110 snprintf(buff, bufflen, 111 "mount --make-private %s failed: %s", 112 target, strerror(errno)); 113 return -1; 114 } 115 116 if (mount(target, target, "none", MS_BIND, NULL)) { 117 snprintf(buff, bufflen, 118 "mount --bind %s %s failed: %s", 119 target, target, strerror(errno)); 120 return -1; 121 } 122 123 bind_done = true; 124 } 125 126 if (mount("bpf", target, "bpf", 0, "mode=0700")) { 127 snprintf(buff, bufflen, "mount -t bpf bpf %s failed: %s", 128 target, strerror(errno)); 129 return -1; 130 } 131 132 return 0; 133 } 134 135 int open_obj_pinned(char *path) 136 { 137 int fd; 138 139 fd = bpf_obj_get(path); 140 if (fd < 0) { 141 p_err("bpf obj get (%s): %s", path, 142 errno == EACCES && !is_bpffs(dirname(path)) ? 143 "directory not in bpf file system (bpffs)" : 144 strerror(errno)); 145 return -1; 146 } 147 148 return fd; 149 } 150 151 int open_obj_pinned_any(char *path, enum bpf_obj_type exp_type) 152 { 153 enum bpf_obj_type type; 154 int fd; 155 156 fd = open_obj_pinned(path); 157 if (fd < 0) 158 return -1; 159 160 type = get_fd_type(fd); 161 if (type < 0) { 162 close(fd); 163 return type; 164 } 165 if (type != exp_type) { 166 p_err("incorrect object type: %s", get_fd_type_name(type)); 167 close(fd); 168 return -1; 169 } 170 171 return fd; 172 } 173 174 int do_pin_fd(int fd, const char *name) 175 { 176 char err_str[ERR_MAX_LEN]; 177 char *file; 178 char *dir; 179 int err = 0; 180 181 err = bpf_obj_pin(fd, name); 182 if (!err) 183 goto out; 184 185 file = malloc(strlen(name) + 1); 186 strcpy(file, name); 187 dir = dirname(file); 188 189 if (errno != EPERM || is_bpffs(dir)) { 190 p_err("can't pin the object (%s): %s", name, strerror(errno)); 191 goto out_free; 192 } 193 194 /* Attempt to mount bpffs, then retry pinning. */ 195 err = mnt_bpffs(dir, err_str, ERR_MAX_LEN); 196 if (!err) { 197 err = bpf_obj_pin(fd, name); 198 if (err) 199 p_err("can't pin the object (%s): %s", name, 200 strerror(errno)); 201 } else { 202 err_str[ERR_MAX_LEN - 1] = '\0'; 203 p_err("can't mount BPF file system to pin the object (%s): %s", 204 name, err_str); 205 } 206 207 out_free: 208 free(file); 209 out: 210 return err; 211 } 212 213 int do_pin_any(int argc, char **argv, int (*get_fd_by_id)(__u32)) 214 { 215 unsigned int id; 216 char *endptr; 217 int err; 218 int fd; 219 220 if (argc < 3) { 221 p_err("too few arguments, id ID and FILE path is required"); 222 return -1; 223 } else if (argc > 3) { 224 p_err("too many arguments"); 225 return -1; 226 } 227 228 if (!is_prefix(*argv, "id")) { 229 p_err("expected 'id' got %s", *argv); 230 return -1; 231 } 232 NEXT_ARG(); 233 234 id = strtoul(*argv, &endptr, 0); 235 if (*endptr) { 236 p_err("can't parse %s as ID", *argv); 237 return -1; 238 } 239 NEXT_ARG(); 240 241 fd = get_fd_by_id(id); 242 if (fd < 0) { 243 p_err("can't get prog by id (%u): %s", id, strerror(errno)); 244 return -1; 245 } 246 247 err = do_pin_fd(fd, *argv); 248 249 close(fd); 250 return err; 251 } 252 253 const char *get_fd_type_name(enum bpf_obj_type type) 254 { 255 static const char * const names[] = { 256 [BPF_OBJ_UNKNOWN] = "unknown", 257 [BPF_OBJ_PROG] = "prog", 258 [BPF_OBJ_MAP] = "map", 259 }; 260 261 if (type < 0 || type >= ARRAY_SIZE(names) || !names[type]) 262 return names[BPF_OBJ_UNKNOWN]; 263 264 return names[type]; 265 } 266 267 int get_fd_type(int fd) 268 { 269 char path[PATH_MAX]; 270 char buf[512]; 271 ssize_t n; 272 273 snprintf(path, sizeof(path), "/proc/%d/fd/%d", getpid(), fd); 274 275 n = readlink(path, buf, sizeof(buf)); 276 if (n < 0) { 277 p_err("can't read link type: %s", strerror(errno)); 278 return -1; 279 } 280 if (n == sizeof(path)) { 281 p_err("can't read link type: path too long!"); 282 return -1; 283 } 284 285 if (strstr(buf, "bpf-map")) 286 return BPF_OBJ_MAP; 287 else if (strstr(buf, "bpf-prog")) 288 return BPF_OBJ_PROG; 289 290 return BPF_OBJ_UNKNOWN; 291 } 292 293 char *get_fdinfo(int fd, const char *key) 294 { 295 char path[PATH_MAX]; 296 char *line = NULL; 297 size_t line_n = 0; 298 ssize_t n; 299 FILE *fdi; 300 301 snprintf(path, sizeof(path), "/proc/%d/fdinfo/%d", getpid(), fd); 302 303 fdi = fopen(path, "r"); 304 if (!fdi) { 305 p_err("can't open fdinfo: %s", strerror(errno)); 306 return NULL; 307 } 308 309 while ((n = getline(&line, &line_n, fdi))) { 310 char *value; 311 int len; 312 313 if (!strstr(line, key)) 314 continue; 315 316 fclose(fdi); 317 318 value = strchr(line, '\t'); 319 if (!value || !value[1]) { 320 p_err("malformed fdinfo!?"); 321 free(line); 322 return NULL; 323 } 324 value++; 325 326 len = strlen(value); 327 memmove(line, value, len); 328 line[len - 1] = '\0'; 329 330 return line; 331 } 332 333 p_err("key '%s' not found in fdinfo", key); 334 free(line); 335 fclose(fdi); 336 return NULL; 337 } 338 339 void print_data_json(uint8_t *data, size_t len) 340 { 341 unsigned int i; 342 343 jsonw_start_array(json_wtr); 344 for (i = 0; i < len; i++) 345 jsonw_printf(json_wtr, "%d", data[i]); 346 jsonw_end_array(json_wtr); 347 } 348 349 void print_hex_data_json(uint8_t *data, size_t len) 350 { 351 unsigned int i; 352 353 jsonw_start_array(json_wtr); 354 for (i = 0; i < len; i++) 355 jsonw_printf(json_wtr, "\"0x%02hhx\"", data[i]); 356 jsonw_end_array(json_wtr); 357 } 358 359 int build_pinned_obj_table(struct pinned_obj_table *tab, 360 enum bpf_obj_type type) 361 { 362 struct bpf_prog_info pinned_info = {}; 363 struct pinned_obj *obj_node = NULL; 364 __u32 len = sizeof(pinned_info); 365 struct mntent *mntent = NULL; 366 enum bpf_obj_type objtype; 367 FILE *mntfile = NULL; 368 FTSENT *ftse = NULL; 369 FTS *fts = NULL; 370 int fd, err; 371 372 mntfile = setmntent("/proc/mounts", "r"); 373 if (!mntfile) 374 return -1; 375 376 while ((mntent = getmntent(mntfile))) { 377 char *path[] = { mntent->mnt_dir, NULL }; 378 379 if (strncmp(mntent->mnt_type, "bpf", 3) != 0) 380 continue; 381 382 fts = fts_open(path, 0, NULL); 383 if (!fts) 384 continue; 385 386 while ((ftse = fts_read(fts))) { 387 if (!(ftse->fts_info & FTS_F)) 388 continue; 389 fd = open_obj_pinned(ftse->fts_path); 390 if (fd < 0) 391 continue; 392 393 objtype = get_fd_type(fd); 394 if (objtype != type) { 395 close(fd); 396 continue; 397 } 398 memset(&pinned_info, 0, sizeof(pinned_info)); 399 err = bpf_obj_get_info_by_fd(fd, &pinned_info, &len); 400 if (err) { 401 close(fd); 402 continue; 403 } 404 405 obj_node = malloc(sizeof(*obj_node)); 406 if (!obj_node) { 407 close(fd); 408 fts_close(fts); 409 fclose(mntfile); 410 return -1; 411 } 412 413 memset(obj_node, 0, sizeof(*obj_node)); 414 obj_node->id = pinned_info.id; 415 obj_node->path = strdup(ftse->fts_path); 416 hash_add(tab->table, &obj_node->hash, obj_node->id); 417 418 close(fd); 419 } 420 fts_close(fts); 421 } 422 fclose(mntfile); 423 return 0; 424 } 425 426 void delete_pinned_obj_table(struct pinned_obj_table *tab) 427 { 428 struct pinned_obj *obj; 429 struct hlist_node *tmp; 430 unsigned int bkt; 431 432 hash_for_each_safe(tab->table, bkt, tmp, obj, hash) { 433 hash_del(&obj->hash); 434 free(obj->path); 435 free(obj); 436 } 437 } 438 439 unsigned int get_page_size(void) 440 { 441 static int result; 442 443 if (!result) 444 result = getpagesize(); 445 return result; 446 } 447 448 unsigned int get_possible_cpus(void) 449 { 450 static unsigned int result; 451 char buf[128]; 452 long int n; 453 char *ptr; 454 int fd; 455 456 if (result) 457 return result; 458 459 fd = open("/sys/devices/system/cpu/possible", O_RDONLY); 460 if (fd < 0) { 461 p_err("can't open sysfs possible cpus"); 462 exit(-1); 463 } 464 465 n = read(fd, buf, sizeof(buf)); 466 if (n < 2) { 467 p_err("can't read sysfs possible cpus"); 468 exit(-1); 469 } 470 close(fd); 471 472 if (n == sizeof(buf)) { 473 p_err("read sysfs possible cpus overflow"); 474 exit(-1); 475 } 476 477 ptr = buf; 478 n = 0; 479 while (*ptr && *ptr != '\n') { 480 unsigned int a, b; 481 482 if (sscanf(ptr, "%u-%u", &a, &b) == 2) { 483 n += b - a + 1; 484 485 ptr = strchr(ptr, '-') + 1; 486 } else if (sscanf(ptr, "%u", &a) == 1) { 487 n++; 488 } else { 489 assert(0); 490 } 491 492 while (isdigit(*ptr)) 493 ptr++; 494 if (*ptr == ',') 495 ptr++; 496 } 497 498 result = n; 499 500 return result; 501 } 502 503 static char * 504 ifindex_to_name_ns(__u32 ifindex, __u32 ns_dev, __u32 ns_ino, char *buf) 505 { 506 struct stat st; 507 int err; 508 509 err = stat("/proc/self/ns/net", &st); 510 if (err) { 511 p_err("Can't stat /proc/self: %s", strerror(errno)); 512 return NULL; 513 } 514 515 if (st.st_dev != ns_dev || st.st_ino != ns_ino) 516 return NULL; 517 518 return if_indextoname(ifindex, buf); 519 } 520 521 static int read_sysfs_hex_int(char *path) 522 { 523 char vendor_id_buf[8]; 524 int len; 525 int fd; 526 527 fd = open(path, O_RDONLY); 528 if (fd < 0) { 529 p_err("Can't open %s: %s", path, strerror(errno)); 530 return -1; 531 } 532 533 len = read(fd, vendor_id_buf, sizeof(vendor_id_buf)); 534 close(fd); 535 if (len < 0) { 536 p_err("Can't read %s: %s", path, strerror(errno)); 537 return -1; 538 } 539 if (len >= (int)sizeof(vendor_id_buf)) { 540 p_err("Value in %s too long", path); 541 return -1; 542 } 543 544 vendor_id_buf[len] = 0; 545 546 return strtol(vendor_id_buf, NULL, 0); 547 } 548 549 static int read_sysfs_netdev_hex_int(char *devname, const char *entry_name) 550 { 551 char full_path[64]; 552 553 snprintf(full_path, sizeof(full_path), "/sys/class/net/%s/device/%s", 554 devname, entry_name); 555 556 return read_sysfs_hex_int(full_path); 557 } 558 559 const char *ifindex_to_bfd_name_ns(__u32 ifindex, __u64 ns_dev, __u64 ns_ino) 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 return "NFP-6xxx"; 585 default: 586 p_err("Can't get bfd arch name for device vendor id 0x%04x", 587 vendor_id); 588 return NULL; 589 } 590 } 591 592 void print_dev_plain(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 593 { 594 char name[IF_NAMESIZE]; 595 596 if (!ifindex) 597 return; 598 599 printf(" dev "); 600 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 601 printf("%s", name); 602 else 603 printf("ifindex %u ns_dev %llu ns_ino %llu", 604 ifindex, ns_dev, ns_inode); 605 } 606 607 void print_dev_json(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 608 { 609 char name[IF_NAMESIZE]; 610 611 if (!ifindex) 612 return; 613 614 jsonw_name(json_wtr, "dev"); 615 jsonw_start_object(json_wtr); 616 jsonw_uint_field(json_wtr, "ifindex", ifindex); 617 jsonw_uint_field(json_wtr, "ns_dev", ns_dev); 618 jsonw_uint_field(json_wtr, "ns_inode", ns_inode); 619 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 620 jsonw_string_field(json_wtr, "ifname", name); 621 jsonw_end_object(json_wtr); 622 } 623