lib/netns_linux.c

   1 /*
   2  * NS functions.
   3  * Copyright (C) 2014 6WIND S.A.
   4  *
   5  * This file is part of GNU Zebra.
   6  *
   7  * GNU Zebra is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published
   9  * by the Free Software Foundation; either version 2, or (at your
  10  * option) any later version.
  11  *
  12  * GNU Zebra is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with this program; see the file COPYING; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include <zebra.h>
  23
  24 #ifdef HAVE_NETNS
  25 #undef _GNU_SOURCE
  26 #define _GNU_SOURCE
  27
  28 #include <sched.h>
  29 #endif
  30
  31 /* for basename */
  32 #include <libgen.h>
  33
  34 #include "if.h"
  35 #include "ns.h"
  36 #include "log.h"
  37 #include "memory.h"
  38 #include "command.h"
  39 #include "vty.h"
  40 #include "vrf.h"
  41 #include "lib_errors.h"
  42
  43 DEFINE_MTYPE_STATIC(LIB, NS, "NetNS Context")
  44 DEFINE_MTYPE_STATIC(LIB, NS_NAME, "NetNS Name")
  45
  46 /* default NS ID value used when VRF backend is not NETNS */
  47 #define NS_DEFAULT_INTERNAL 0
  48
  49 static inline int ns_compare(const struct ns *ns, const struct ns *ns2);
  50 static struct ns *ns_lookup_name_internal(const char *name);
  51
  52 RB_GENERATE(ns_head, ns, entry, ns_compare)
  53
  54 struct ns_head ns_tree = RB_INITIALIZER(&ns_tree);
  55
  56 static struct ns *default_ns;
  57 static int ns_current_ns_fd;
  58 static int ns_default_ns_fd;
  59
  60 static int ns_debug;
  61
  62 struct ns_map_nsid {
  63         RB_ENTRY(ns_map_nsid) id_entry;
  64         ns_id_t ns_id_external;
  65         ns_id_t ns_id;
  66 };
  67
  68 static inline int ns_map_compare(const struct ns_map_nsid *a,
  69                                    const struct ns_map_nsid *b)
  70 {
  71         return (a->ns_id - b->ns_id);
  72 }
  73
  74 RB_HEAD(ns_map_nsid_head, ns_map_nsid);
  75 RB_PROTOTYPE(ns_map_nsid_head, ns_map_nsid, id_entry, ns_map_compare);
  76 RB_GENERATE(ns_map_nsid_head, ns_map_nsid, id_entry, ns_map_compare);
  77 struct ns_map_nsid_head ns_map_nsid_list = RB_INITIALIZER(&ns_map_nsid_list);
  78
  79 static ns_id_t ns_id_external_numbering;
  80
  81
  82 #ifndef CLONE_NEWNET
  83 #define CLONE_NEWNET 0x40000000
  84 /* New network namespace (lo, device, names sockets, etc) */
  85 #endif
  86
  87 #ifndef HAVE_SETNS
  88 static inline int setns(int fd, int nstype)
  89 {
  90 #ifdef __NR_setns
  91         return syscall(__NR_setns, fd, nstype);
  92 #else
  93         errno = EINVAL;
  94         return -1;
  95 #endif
  96 }
  97 #endif /* !HAVE_SETNS */
  98
  99 #ifdef HAVE_NETNS
 100 static int have_netns_enabled = -1;
 101 #endif /* HAVE_NETNS */
 102
 103 /* default NS ID value used when VRF backend is not NETNS */
 104 #define NS_DEFAULT_INTERNAL 0
 105
 106 static int have_netns(void)
 107 {
 108 #ifdef HAVE_NETNS
 109         if (have_netns_enabled < 0) {
 110                 int fd = open(NS_DEFAULT_NAME, O_RDONLY);
 111
 112                 if (fd < 0)
 113                         have_netns_enabled = 0;
 114                 else {
 115                         have_netns_enabled = 1;
 116                         close(fd);
 117                 }
 118         }
 119         return have_netns_enabled;
 120 #else
 121         return 0;
 122 #endif
 123 }
 124
 125 /* Holding NS hooks  */
 126 struct ns_master {
 127         int (*ns_new_hook)(struct ns *ns);
 128         int (*ns_delete_hook)(struct ns *ns);
 129         int (*ns_enable_hook)(struct ns *ns);
 130         int (*ns_disable_hook)(struct ns *ns);
 131 } ns_master = {
 132         0,
 133 };
 134
 135 static int ns_is_enabled(struct ns *ns);
 136
 137 static inline int ns_compare(const struct ns *a, const struct ns *b)
 138 {
 139         return (a->ns_id - b->ns_id);
 140 }
 141
 142 /* Look up a NS by identifier. */
 143 static struct ns *ns_lookup_internal(ns_id_t ns_id)
 144 {
 145         struct ns ns;
 146
 147         ns.ns_id = ns_id;
 148         return RB_FIND(ns_head, &ns_tree, &ns);
 149 }
 150
 151 /* Look up a NS by name */
 152 static struct ns *ns_lookup_name_internal(const char *name)
 153 {
 154         struct ns *ns = NULL;
 155
 156         RB_FOREACH (ns, ns_head, &ns_tree) {
 157                 if (ns->name != NULL) {
 158                         if (strcmp(name, ns->name) == 0)
 159                                 return ns;
 160                 }
 161         }
 162         return NULL;
 163 }
 164
 165 static struct ns *ns_get_created_internal(struct ns *ns, char *name,
 166                                           ns_id_t ns_id)
 167 {
 168         int created = 0;
 169         /*
 170          * Initialize interfaces.
 171          */
 172         if (!ns && !name && ns_id != NS_UNKNOWN)
 173                 ns = ns_lookup_internal(ns_id);
 174         if (!ns && name)
 175                 ns = ns_lookup_name_internal(name);
 176         if (!ns) {
 177                 ns = XCALLOC(MTYPE_NS, sizeof(struct ns));
 178                 ns->ns_id = ns_id;
 179                 if (name)
 180                         ns->name = XSTRDUP(MTYPE_NS_NAME, name);
 181                 ns->fd = -1;
 182                 RB_INSERT(ns_head, &ns_tree, ns);
 183                 created = 1;
 184         }
 185         if (ns_id != ns->ns_id) {
 186                 RB_REMOVE(ns_head, &ns_tree, ns);
 187                 ns->ns_id = ns_id;
 188                 RB_INSERT(ns_head, &ns_tree, ns);
 189         }
 190         if (!created)
 191                 return ns;
 192         if (ns_debug) {
 193                 if (ns->ns_id != NS_UNKNOWN)
 194                         zlog_info("NS %u is created.", ns->ns_id);
 195                 else
 196                         zlog_info("NS %s is created.", ns->name);
 197         }
 198         if (ns_master.ns_new_hook)
 199                 (*ns_master.ns_new_hook)(ns);
 200         return ns;
 201 }
 202
 203 /*
 204  * Enable a NS - that is, let the NS be ready to use.
 205  * The NS_ENABLE_HOOK callback will be called to inform
 206  * that they can allocate resources in this NS.
 207  *
 208  * RETURN: 1 - enabled successfully; otherwise, 0.
 209  */
 210 static int ns_enable_internal(struct ns *ns, void (*func)(ns_id_t, void *))
 211 {
 212         if (!ns_is_enabled(ns)) {
 213                 if (have_netns()) {
 214                         ns->fd = open(ns->name, O_RDONLY);
 215                 } else {
 216                         ns->fd = -2;
 217                         /* Remember ns_enable_hook has been called */
 218                         errno = -ENOTSUP;
 219                 }
 220
 221                 if (!ns_is_enabled(ns)) {
 222                         flog_err_sys(EC_LIB_SYSTEM_CALL,
 223                                      "Can not enable NS %u: %s!", ns->ns_id,
 224                                      safe_strerror(errno));
 225                         return 0;
 226                 }
 227
 228                 /* Non default NS. leave */
 229                 if (ns->ns_id == NS_UNKNOWN) {
 230                         flog_err(EC_LIB_NS,
 231                                  "Can not enable NS %s %u: Invalid NSID",
 232                                  ns->name, ns->ns_id);
 233                         return 0;
 234                 }
 235                 if (func)
 236                         func(ns->ns_id, (void *)ns->vrf_ctxt);
 237                 if (ns_debug) {
 238                         if (have_netns())
 239                                 zlog_info("NS %u is associated with NETNS %s.",
 240                                           ns->ns_id, ns->name);
 241                         zlog_info("NS %u is enabled.", ns->ns_id);
 242                 }
 243                 /* zebra first receives NS enable event,
 244                  * then VRF enable event
 245                  */
 246                 if (ns_master.ns_enable_hook)
 247                         (*ns_master.ns_enable_hook)(ns);
 248         }
 249
 250         return 1;
 251 }
 252
 253 /*
 254  * Check whether the NS is enabled - that is, whether the NS
 255  * is ready to allocate resources. Currently there's only one
 256  * type of resource: socket.
 257  */
 258 static int ns_is_enabled(struct ns *ns)
 259 {
 260         if (have_netns())
 261                 return ns && ns->fd >= 0;
 262         else
 263                 return ns && ns->fd == -2 && ns->ns_id == NS_DEFAULT;
 264 }
 265
 266 /*
 267  * Disable a NS - that is, let the NS be unusable.
 268  * The NS_DELETE_HOOK callback will be called to inform
 269  * that they must release the resources in the NS.
 270  */
 271 static void ns_disable_internal(struct ns *ns)
 272 {
 273         if (ns_is_enabled(ns)) {
 274                 if (ns_debug)
 275                         zlog_info("NS %u is to be disabled.", ns->ns_id);
 276
 277                 if (ns_master.ns_disable_hook)
 278                         (*ns_master.ns_disable_hook)(ns);
 279
 280                 if (have_netns())
 281                         close(ns->fd);
 282
 283                 ns->fd = -1;
 284         }
 285 }
 286
 287 /* VRF list existance check by name. */
 288 static struct ns_map_nsid *ns_map_nsid_lookup_by_nsid(ns_id_t ns_id)
 289 {
 290         struct ns_map_nsid ns_map;
 291
 292         ns_map.ns_id = ns_id;
 293         return RB_FIND(ns_map_nsid_head, &ns_map_nsid_list, &ns_map);
 294 }
 295
 296 ns_id_t ns_map_nsid_with_external(ns_id_t ns_id, bool map)
 297 {
 298         struct ns_map_nsid *ns_map;
 299         vrf_id_t ns_id_external;
 300
 301         ns_map = ns_map_nsid_lookup_by_nsid(ns_id);
 302         if (ns_map && !map) {
 303                 ns_id_external = ns_map->ns_id_external;
 304                 RB_REMOVE(ns_map_nsid_head, &ns_map_nsid_list, ns_map);
 305                 return ns_id_external;
 306         }
 307         if (ns_map)
 308                 return ns_map->ns_id_external;
 309         ns_map = XCALLOC(MTYPE_NS, sizeof(struct ns_map_nsid));
 310         /* increase vrf_id
 311          * default vrf is the first one : 0
 312          */
 313         ns_map->ns_id_external = ns_id_external_numbering++;
 314         ns_map->ns_id = ns_id;
 315         RB_INSERT(ns_map_nsid_head, &ns_map_nsid_list, ns_map);
 316         return ns_map->ns_id_external;
 317 }
 318
 319 struct ns *ns_get_created(struct ns *ns, char *name, ns_id_t ns_id)
 320 {
 321         return ns_get_created_internal(ns, name, ns_id);
 322 }
 323
 324 int ns_have_netns(void)
 325 {
 326         return have_netns();
 327 }
 328
 329 /* Delete a NS. This is called in ns_terminate(). */
 330 void ns_delete(struct ns *ns)
 331 {
 332         if (ns_debug)
 333                 zlog_info("NS %u is to be deleted.", ns->ns_id);
 334
 335         ns_disable(ns);
 336
 337         if (ns_master.ns_delete_hook)
 338                 (*ns_master.ns_delete_hook)(ns);
 339
 340         /*
 341          * I'm not entirely sure if the vrf->iflist
 342          * needs to be moved into here or not.
 343          */
 344         // if_terminate (&ns->iflist);
 345
 346         RB_REMOVE(ns_head, &ns_tree, ns);
 347         XFREE(MTYPE_NS_NAME, ns->name);
 348
 349         XFREE(MTYPE_NS, ns);
 350 }
 351
 352 /* Look up the data pointer of the specified VRF. */
 353 void *ns_info_lookup(ns_id_t ns_id)
 354 {
 355         struct ns *ns = ns_lookup_internal(ns_id);
 356
 357         return ns ? ns->info : NULL;
 358 }
 359
 360 /* Look up a NS by name */
 361 struct ns *ns_lookup_name(const char *name)
 362 {
 363         return ns_lookup_name_internal(name);
 364 }
 365
 366 int ns_enable(struct ns *ns, void (*func)(ns_id_t, void *))
 367 {
 368         return ns_enable_internal(ns, func);
 369 }
 370
 371 void ns_disable(struct ns *ns)
 372 {
 373         return ns_disable_internal(ns);
 374 }
 375
 376 struct ns *ns_lookup(ns_id_t ns_id)
 377 {
 378         return ns_lookup_internal(ns_id);
 379 }
 380
 381 void ns_walk_func(int (*func)(struct ns *))
 382 {
 383         struct ns *ns = NULL;
 384
 385         RB_FOREACH (ns, ns_head, &ns_tree)
 386                 func(ns);
 387 }
 388
 389 const char *ns_get_name(struct ns *ns)
 390 {
 391         if (!ns)
 392                 return NULL;
 393         return ns->name;
 394 }
 395
 396 /* Add a NS hook. Please add hooks before calling ns_init(). */
 397 void ns_add_hook(int type, int (*func)(struct ns *))
 398 {
 399         switch (type) {
 400         case NS_NEW_HOOK:
 401                 ns_master.ns_new_hook = func;
 402                 break;
 403         case NS_DELETE_HOOK:
 404                 ns_master.ns_delete_hook = func;
 405                 break;
 406         case NS_ENABLE_HOOK:
 407                 ns_master.ns_enable_hook = func;
 408                 break;
 409         case NS_DISABLE_HOOK:
 410                 ns_master.ns_disable_hook = func;
 411                 break;
 412         default:
 413                 break;
 414         }
 415 }
 416
 417 /*
 418  * NS realization with NETNS
 419  */
 420
 421 char *ns_netns_pathname(struct vty *vty, const char *name)
 422 {
 423         static char pathname[PATH_MAX];
 424         char *result;
 425         char *check_base;
 426
 427         if (name[0] == '/') /* absolute pathname */
 428                 result = realpath(name, pathname);
 429         else {
 430                 /* relevant pathname */
 431                 char tmp_name[PATH_MAX];
 432
 433                 snprintf(tmp_name, PATH_MAX, "%s/%s", NS_RUN_DIR, name);
 434                 result = realpath(tmp_name, pathname);
 435         }
 436
 437         if (!result) {
 438                 if (vty)
 439                         vty_out(vty, "Invalid pathname for %s: %s\n",
 440                                 pathname,
 441                                 safe_strerror(errno));
 442                 else
 443                         flog_warn(EC_LIB_LINUX_NS,
 444                                   "Invalid pathname for %s: %s", pathname,
 445                                   safe_strerror(errno));
 446                 return NULL;
 447         }
 448         check_base = basename(pathname);
 449         if (check_base != NULL && strlen(check_base) + 1 > NS_NAMSIZ) {
 450                 if (vty)
 451                         vty_out(vty, "NS name (%s) invalid: too long (>%d)\n",
 452                                 check_base, NS_NAMSIZ - 1);
 453                 else
 454                         flog_warn(EC_LIB_LINUX_NS,
 455                                   "NS name (%s) invalid: too long (>%d)",
 456                                   check_base, NS_NAMSIZ - 1);
 457                 return NULL;
 458         }
 459         return pathname;
 460 }
 461
 462 void ns_init(void)
 463 {
 464         static int ns_initialised;
 465
 466         ns_debug = 0;
 467         /* silently return as initialisation done */
 468         if (ns_initialised == 1)
 469                 return;
 470         errno = 0;
 471         if (have_netns())
 472                 ns_default_ns_fd = open(NS_DEFAULT_NAME, O_RDONLY);
 473         else {
 474                 ns_default_ns_fd = -1;
 475                 default_ns = NULL;
 476         }
 477         ns_current_ns_fd = -1;
 478         ns_initialised = 1;
 479 }
 480
 481 /* Initialize NS module. */
 482 void ns_init_management(ns_id_t default_ns_id, ns_id_t internal_ns)
 483 {
 484         int fd;
 485
 486         ns_init();
 487         default_ns = ns_get_created_internal(NULL, NULL, default_ns_id);
 488         if (!default_ns) {
 489                 flog_err(EC_LIB_NS, "%s: failed to create the default NS!",
 490                          __func__);
 491                 exit(1);
 492         }
 493         if (have_netns()) {
 494                 fd = open(NS_DEFAULT_NAME, O_RDONLY);
 495                 default_ns->fd = fd;
 496         }
 497         default_ns->internal_ns_id = internal_ns;
 498
 499         /* Set the default NS name. */
 500         default_ns->name = XSTRDUP(MTYPE_NS_NAME, NS_DEFAULT_NAME);
 501         if (ns_debug)
 502                 zlog_info("%s: default NSID is %u", __func__,
 503                           default_ns->ns_id);
 504
 505         /* Enable the default NS. */
 506         if (!ns_enable(default_ns, NULL)) {
 507                 flog_err(EC_LIB_NS, "%s: failed to enable the default NS!",
 508                          __func__);
 509                 exit(1);
 510         }
 511 }
 512
 513 /* Terminate NS module. */
 514 void ns_terminate(void)
 515 {
 516         struct ns *ns;
 517
 518         while (!RB_EMPTY(ns_head, &ns_tree)) {
 519                 ns = RB_ROOT(ns_head, &ns_tree);
 520
 521                 ns_delete(ns);
 522         }
 523 }
 524
 525 int ns_switch_to_netns(const char *name)
 526 {
 527         int ret;
 528         int fd;
 529
 530         if (name == NULL)
 531                 return -1;
 532         if (ns_default_ns_fd == -1)
 533                 return -1;
 534         fd = open(name, O_RDONLY);
 535         if (fd == -1) {
 536                 errno = EINVAL;
 537                 return -1;
 538         }
 539         ret = setns(fd, CLONE_NEWNET);
 540         ns_current_ns_fd = fd;
 541         close(fd);
 542         return ret;
 543 }
 544
 545 /* returns 1 if switch() was not called before
 546  * return status of setns() otherwise
 547  */
 548 int ns_switchback_to_initial(void)
 549 {
 550         if (ns_current_ns_fd != -1 && ns_default_ns_fd != -1) {
 551                 int ret;
 552
 553                 ret = setns(ns_default_ns_fd, CLONE_NEWNET);
 554                 ns_current_ns_fd = -1;
 555                 return ret;
 556         }
 557         /* silently ignore if setns() is not called */
 558         return 1;
 559 }
 560
 561 /* Create a socket for the NS. */
 562 int ns_socket(int domain, int type, int protocol, ns_id_t ns_id)
 563 {
 564         struct ns *ns = ns_lookup(ns_id);
 565         int ret;
 566
 567         if (!ns || !ns_is_enabled(ns)) {
 568                 errno = EINVAL;
 569                 return -1;
 570         }
 571         if (have_netns()) {
 572                 ret = (ns_id != NS_DEFAULT) ? setns(ns->fd, CLONE_NEWNET) : 0;
 573                 if (ret >= 0) {
 574                         ret = socket(domain, type, protocol);
 575                         if (ns_id != NS_DEFAULT) {
 576                                 setns(ns_lookup(NS_DEFAULT)->fd, CLONE_NEWNET);
 577                                 ns_current_ns_fd = ns_id;
 578                         }
 579                 }
 580         } else
 581                 ret = socket(domain, type, protocol);
 582
 583         return ret;
 584 }
 585
 586 ns_id_t ns_get_default_id(void)
 587 {
 588         if (default_ns)
 589                 return default_ns->ns_id;
 590         return NS_DEFAULT_INTERNAL;
 591 }