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