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