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/resource.h> 50 #include <sys/stat.h> 51 #include <sys/types.h> 52 #include <sys/vfs.h> 53 54 #include <bpf.h> 55 56 #include "main.h" 57 58 #ifndef BPF_FS_MAGIC 59 #define BPF_FS_MAGIC 0xcafe4a11 60 #endif 61 62 void p_err(const char *fmt, ...) 63 { 64 va_list ap; 65 66 va_start(ap, fmt); 67 if (json_output) { 68 jsonw_start_object(json_wtr); 69 jsonw_name(json_wtr, "error"); 70 jsonw_vprintf_enquote(json_wtr, fmt, ap); 71 jsonw_end_object(json_wtr); 72 } else { 73 fprintf(stderr, "Error: "); 74 vfprintf(stderr, fmt, ap); 75 fprintf(stderr, "\n"); 76 } 77 va_end(ap); 78 } 79 80 void p_info(const char *fmt, ...) 81 { 82 va_list ap; 83 84 if (json_output) 85 return; 86 87 va_start(ap, fmt); 88 vfprintf(stderr, fmt, ap); 89 fprintf(stderr, "\n"); 90 va_end(ap); 91 } 92 93 static bool is_bpffs(char *path) 94 { 95 struct statfs st_fs; 96 97 if (statfs(path, &st_fs) < 0) 98 return false; 99 100 return (unsigned long)st_fs.f_type == BPF_FS_MAGIC; 101 } 102 103 void set_max_rlimit(void) 104 { 105 struct rlimit rinf = { RLIM_INFINITY, RLIM_INFINITY }; 106 107 setrlimit(RLIMIT_MEMLOCK, &rinf); 108 } 109 110 static int mnt_bpffs(const char *target, char *buff, size_t bufflen) 111 { 112 bool bind_done = false; 113 114 while (mount("", target, "none", MS_PRIVATE | MS_REC, NULL)) { 115 if (errno != EINVAL || bind_done) { 116 snprintf(buff, bufflen, 117 "mount --make-private %s failed: %s", 118 target, strerror(errno)); 119 return -1; 120 } 121 122 if (mount(target, target, "none", MS_BIND, NULL)) { 123 snprintf(buff, bufflen, 124 "mount --bind %s %s failed: %s", 125 target, target, strerror(errno)); 126 return -1; 127 } 128 129 bind_done = true; 130 } 131 132 if (mount("bpf", target, "bpf", 0, "mode=0700")) { 133 snprintf(buff, bufflen, "mount -t bpf bpf %s failed: %s", 134 target, strerror(errno)); 135 return -1; 136 } 137 138 return 0; 139 } 140 141 int open_obj_pinned(char *path, bool quiet) 142 { 143 int fd; 144 145 fd = bpf_obj_get(path); 146 if (fd < 0) { 147 if (!quiet) 148 p_err("bpf obj get (%s): %s", path, 149 errno == EACCES && !is_bpffs(dirname(path)) ? 150 "directory not in bpf file system (bpffs)" : 151 strerror(errno)); 152 return -1; 153 } 154 155 return fd; 156 } 157 158 int open_obj_pinned_any(char *path, enum bpf_obj_type exp_type) 159 { 160 enum bpf_obj_type type; 161 int fd; 162 163 fd = open_obj_pinned(path, false); 164 if (fd < 0) 165 return -1; 166 167 type = get_fd_type(fd); 168 if (type < 0) { 169 close(fd); 170 return type; 171 } 172 if (type != exp_type) { 173 p_err("incorrect object type: %s", get_fd_type_name(type)); 174 close(fd); 175 return -1; 176 } 177 178 return fd; 179 } 180 181 int mount_bpffs_for_pin(const char *name) 182 { 183 char err_str[ERR_MAX_LEN]; 184 char *file; 185 char *dir; 186 int err = 0; 187 188 file = malloc(strlen(name) + 1); 189 strcpy(file, name); 190 dir = dirname(file); 191 192 if (is_bpffs(dir)) 193 /* nothing to do if already mounted */ 194 goto out_free; 195 196 err = mnt_bpffs(dir, err_str, ERR_MAX_LEN); 197 if (err) { 198 err_str[ERR_MAX_LEN - 1] = '\0'; 199 p_err("can't mount BPF file system to pin the object (%s): %s", 200 name, err_str); 201 } 202 203 out_free: 204 free(file); 205 return err; 206 } 207 208 int do_pin_fd(int fd, const char *name) 209 { 210 int err; 211 212 err = mount_bpffs_for_pin(name); 213 if (err) 214 return err; 215 216 return bpf_obj_pin(fd, name); 217 } 218 219 int do_pin_any(int argc, char **argv, int (*get_fd_by_id)(__u32)) 220 { 221 unsigned int id; 222 char *endptr; 223 int err; 224 int fd; 225 226 if (argc < 3) { 227 p_err("too few arguments, id ID and FILE path is required"); 228 return -1; 229 } else if (argc > 3) { 230 p_err("too many arguments"); 231 return -1; 232 } 233 234 if (!is_prefix(*argv, "id")) { 235 p_err("expected 'id' got %s", *argv); 236 return -1; 237 } 238 NEXT_ARG(); 239 240 id = strtoul(*argv, &endptr, 0); 241 if (*endptr) { 242 p_err("can't parse %s as ID", *argv); 243 return -1; 244 } 245 NEXT_ARG(); 246 247 fd = get_fd_by_id(id); 248 if (fd < 0) { 249 p_err("can't get prog by id (%u): %s", id, strerror(errno)); 250 return -1; 251 } 252 253 err = do_pin_fd(fd, *argv); 254 255 close(fd); 256 return err; 257 } 258 259 const char *get_fd_type_name(enum bpf_obj_type type) 260 { 261 static const char * const names[] = { 262 [BPF_OBJ_UNKNOWN] = "unknown", 263 [BPF_OBJ_PROG] = "prog", 264 [BPF_OBJ_MAP] = "map", 265 }; 266 267 if (type < 0 || type >= ARRAY_SIZE(names) || !names[type]) 268 return names[BPF_OBJ_UNKNOWN]; 269 270 return names[type]; 271 } 272 273 int get_fd_type(int fd) 274 { 275 char path[PATH_MAX]; 276 char buf[512]; 277 ssize_t n; 278 279 snprintf(path, sizeof(path), "/proc/%d/fd/%d", getpid(), fd); 280 281 n = readlink(path, buf, sizeof(buf)); 282 if (n < 0) { 283 p_err("can't read link type: %s", strerror(errno)); 284 return -1; 285 } 286 if (n == sizeof(path)) { 287 p_err("can't read link type: path too long!"); 288 return -1; 289 } 290 291 if (strstr(buf, "bpf-map")) 292 return BPF_OBJ_MAP; 293 else if (strstr(buf, "bpf-prog")) 294 return BPF_OBJ_PROG; 295 296 return BPF_OBJ_UNKNOWN; 297 } 298 299 char *get_fdinfo(int fd, const char *key) 300 { 301 char path[PATH_MAX]; 302 char *line = NULL; 303 size_t line_n = 0; 304 ssize_t n; 305 FILE *fdi; 306 307 snprintf(path, sizeof(path), "/proc/%d/fdinfo/%d", getpid(), fd); 308 309 fdi = fopen(path, "r"); 310 if (!fdi) { 311 p_err("can't open fdinfo: %s", strerror(errno)); 312 return NULL; 313 } 314 315 while ((n = getline(&line, &line_n, fdi)) > 0) { 316 char *value; 317 int len; 318 319 if (!strstr(line, key)) 320 continue; 321 322 fclose(fdi); 323 324 value = strchr(line, '\t'); 325 if (!value || !value[1]) { 326 p_err("malformed fdinfo!?"); 327 free(line); 328 return NULL; 329 } 330 value++; 331 332 len = strlen(value); 333 memmove(line, value, len); 334 line[len - 1] = '\0'; 335 336 return line; 337 } 338 339 p_err("key '%s' not found in fdinfo", key); 340 free(line); 341 fclose(fdi); 342 return NULL; 343 } 344 345 void print_data_json(uint8_t *data, size_t len) 346 { 347 unsigned int i; 348 349 jsonw_start_array(json_wtr); 350 for (i = 0; i < len; i++) 351 jsonw_printf(json_wtr, "%d", data[i]); 352 jsonw_end_array(json_wtr); 353 } 354 355 void print_hex_data_json(uint8_t *data, size_t len) 356 { 357 unsigned int i; 358 359 jsonw_start_array(json_wtr); 360 for (i = 0; i < len; i++) 361 jsonw_printf(json_wtr, "\"0x%02hhx\"", data[i]); 362 jsonw_end_array(json_wtr); 363 } 364 365 int build_pinned_obj_table(struct pinned_obj_table *tab, 366 enum bpf_obj_type type) 367 { 368 struct bpf_prog_info pinned_info = {}; 369 struct pinned_obj *obj_node = NULL; 370 __u32 len = sizeof(pinned_info); 371 struct mntent *mntent = NULL; 372 enum bpf_obj_type objtype; 373 FILE *mntfile = NULL; 374 FTSENT *ftse = NULL; 375 FTS *fts = NULL; 376 int fd, err; 377 378 mntfile = setmntent("/proc/mounts", "r"); 379 if (!mntfile) 380 return -1; 381 382 while ((mntent = getmntent(mntfile))) { 383 char *path[] = { mntent->mnt_dir, NULL }; 384 385 if (strncmp(mntent->mnt_type, "bpf", 3) != 0) 386 continue; 387 388 fts = fts_open(path, 0, NULL); 389 if (!fts) 390 continue; 391 392 while ((ftse = fts_read(fts))) { 393 if (!(ftse->fts_info & FTS_F)) 394 continue; 395 fd = open_obj_pinned(ftse->fts_path, true); 396 if (fd < 0) 397 continue; 398 399 objtype = get_fd_type(fd); 400 if (objtype != type) { 401 close(fd); 402 continue; 403 } 404 memset(&pinned_info, 0, sizeof(pinned_info)); 405 err = bpf_obj_get_info_by_fd(fd, &pinned_info, &len); 406 if (err) { 407 close(fd); 408 continue; 409 } 410 411 obj_node = malloc(sizeof(*obj_node)); 412 if (!obj_node) { 413 close(fd); 414 fts_close(fts); 415 fclose(mntfile); 416 return -1; 417 } 418 419 memset(obj_node, 0, sizeof(*obj_node)); 420 obj_node->id = pinned_info.id; 421 obj_node->path = strdup(ftse->fts_path); 422 hash_add(tab->table, &obj_node->hash, obj_node->id); 423 424 close(fd); 425 } 426 fts_close(fts); 427 } 428 fclose(mntfile); 429 return 0; 430 } 431 432 void delete_pinned_obj_table(struct pinned_obj_table *tab) 433 { 434 struct pinned_obj *obj; 435 struct hlist_node *tmp; 436 unsigned int bkt; 437 438 hash_for_each_safe(tab->table, bkt, tmp, obj, hash) { 439 hash_del(&obj->hash); 440 free(obj->path); 441 free(obj); 442 } 443 } 444 445 unsigned int get_page_size(void) 446 { 447 static int result; 448 449 if (!result) 450 result = getpagesize(); 451 return result; 452 } 453 454 unsigned int get_possible_cpus(void) 455 { 456 static unsigned int result; 457 char buf[128]; 458 long int n; 459 char *ptr; 460 int fd; 461 462 if (result) 463 return result; 464 465 fd = open("/sys/devices/system/cpu/possible", O_RDONLY); 466 if (fd < 0) { 467 p_err("can't open sysfs possible cpus"); 468 exit(-1); 469 } 470 471 n = read(fd, buf, sizeof(buf)); 472 if (n < 2) { 473 p_err("can't read sysfs possible cpus"); 474 exit(-1); 475 } 476 close(fd); 477 478 if (n == sizeof(buf)) { 479 p_err("read sysfs possible cpus overflow"); 480 exit(-1); 481 } 482 483 ptr = buf; 484 n = 0; 485 while (*ptr && *ptr != '\n') { 486 unsigned int a, b; 487 488 if (sscanf(ptr, "%u-%u", &a, &b) == 2) { 489 n += b - a + 1; 490 491 ptr = strchr(ptr, '-') + 1; 492 } else if (sscanf(ptr, "%u", &a) == 1) { 493 n++; 494 } else { 495 assert(0); 496 } 497 498 while (isdigit(*ptr)) 499 ptr++; 500 if (*ptr == ',') 501 ptr++; 502 } 503 504 result = n; 505 506 return result; 507 } 508 509 static char * 510 ifindex_to_name_ns(__u32 ifindex, __u32 ns_dev, __u32 ns_ino, char *buf) 511 { 512 struct stat st; 513 int err; 514 515 err = stat("/proc/self/ns/net", &st); 516 if (err) { 517 p_err("Can't stat /proc/self: %s", strerror(errno)); 518 return NULL; 519 } 520 521 if (st.st_dev != ns_dev || st.st_ino != ns_ino) 522 return NULL; 523 524 return if_indextoname(ifindex, buf); 525 } 526 527 static int read_sysfs_hex_int(char *path) 528 { 529 char vendor_id_buf[8]; 530 int len; 531 int fd; 532 533 fd = open(path, O_RDONLY); 534 if (fd < 0) { 535 p_err("Can't open %s: %s", path, strerror(errno)); 536 return -1; 537 } 538 539 len = read(fd, vendor_id_buf, sizeof(vendor_id_buf)); 540 close(fd); 541 if (len < 0) { 542 p_err("Can't read %s: %s", path, strerror(errno)); 543 return -1; 544 } 545 if (len >= (int)sizeof(vendor_id_buf)) { 546 p_err("Value in %s too long", path); 547 return -1; 548 } 549 550 vendor_id_buf[len] = 0; 551 552 return strtol(vendor_id_buf, NULL, 0); 553 } 554 555 static int read_sysfs_netdev_hex_int(char *devname, const char *entry_name) 556 { 557 char full_path[64]; 558 559 snprintf(full_path, sizeof(full_path), "/sys/class/net/%s/device/%s", 560 devname, entry_name); 561 562 return read_sysfs_hex_int(full_path); 563 } 564 565 const char * 566 ifindex_to_bfd_params(__u32 ifindex, __u64 ns_dev, __u64 ns_ino, 567 const char **opt) 568 { 569 char devname[IF_NAMESIZE]; 570 int vendor_id; 571 int device_id; 572 573 if (!ifindex_to_name_ns(ifindex, ns_dev, ns_ino, devname)) { 574 p_err("Can't get net device name for ifindex %d: %s", ifindex, 575 strerror(errno)); 576 return NULL; 577 } 578 579 vendor_id = read_sysfs_netdev_hex_int(devname, "vendor"); 580 if (vendor_id < 0) { 581 p_err("Can't get device vendor id for %s", devname); 582 return NULL; 583 } 584 585 switch (vendor_id) { 586 case 0x19ee: 587 device_id = read_sysfs_netdev_hex_int(devname, "device"); 588 if (device_id != 0x4000 && 589 device_id != 0x6000 && 590 device_id != 0x6003) 591 p_info("Unknown NFP device ID, assuming it is NFP-6xxx arch"); 592 *opt = "ctx4"; 593 return "NFP-6xxx"; 594 default: 595 p_err("Can't get bfd arch name for device vendor id 0x%04x", 596 vendor_id); 597 return NULL; 598 } 599 } 600 601 void print_dev_plain(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 602 { 603 char name[IF_NAMESIZE]; 604 605 if (!ifindex) 606 return; 607 608 printf(" dev "); 609 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 610 printf("%s", name); 611 else 612 printf("ifindex %u ns_dev %llu ns_ino %llu", 613 ifindex, ns_dev, ns_inode); 614 } 615 616 void print_dev_json(__u32 ifindex, __u64 ns_dev, __u64 ns_inode) 617 { 618 char name[IF_NAMESIZE]; 619 620 if (!ifindex) 621 return; 622 623 jsonw_name(json_wtr, "dev"); 624 jsonw_start_object(json_wtr); 625 jsonw_uint_field(json_wtr, "ifindex", ifindex); 626 jsonw_uint_field(json_wtr, "ns_dev", ns_dev); 627 jsonw_uint_field(json_wtr, "ns_inode", ns_inode); 628 if (ifindex_to_name_ns(ifindex, ns_dev, ns_inode, name)) 629 jsonw_string_field(json_wtr, "ifname", name); 630 jsonw_end_object(json_wtr); 631 } 632 633 int parse_u32_arg(int *argc, char ***argv, __u32 *val, const char *what) 634 { 635 char *endptr; 636 637 NEXT_ARGP(); 638 639 if (*val) { 640 p_err("%s already specified", what); 641 return -1; 642 } 643 644 *val = strtoul(**argv, &endptr, 0); 645 if (*endptr) { 646 p_err("can't parse %s as %s", **argv, what); 647 return -1; 648 } 649 NEXT_ARGP(); 650 651 return 0; 652 } 653