src/lxc/storage/storage_utils.c

   1 /*
   2  * lxc: linux Container library
   3  *
   4  * Copyright © 2017 Canonical Ltd.
   5  *
   6  * Authors:
   7  * Christian Brauner <christian.brauner@ubuntu.com>
   8  *
   9  * This program is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public
  11  * License as published by the Free Software Foundation; either
  12  * version 2.1 of the License, or (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  22  */
  23
  24 #define _GNU_SOURCE
  25 #include <dirent.h>
  26 #include <errno.h>
  27 #include <fcntl.h>
  28 #include <grp.h>
  29 #include <inttypes.h>
  30 #include <libgen.h>
  31 #include <sched.h>
  32 #include <stdint.h>
  33 #include <stdio.h>
  34 #include <string.h>
  35 #include <unistd.h>
  36 #include <sys/mount.h>
  37 #include <sys/prctl.h>
  38 #include <sys/types.h>
  39 #include <sys/stat.h>
  40 #include <sys/wait.h>
  41
  42 #include "log.h"
  43 #include "nbd.h"
  44 #include "parse.h"
  45 #include "storage.h"
  46 #include "storage_utils.h"
  47 #include "utils.h"
  48
  49 #ifndef BLKGETSIZE64
  50 #define BLKGETSIZE64 _IOR(0x12, 114, size_t)
  51 #endif
  52
  53 lxc_log_define(storage_utils, lxc);
  54
  55 /* the bulk of this needs to become a common helper */
  56 char *dir_new_path(char *src, const char *oldname, const char *name,
  57                    const char *oldpath, const char *lxcpath)
  58 {
  59         char *ret, *p, *p2;
  60         int l1, l2, nlen;
  61
  62         nlen = strlen(src) + 1;
  63         l1 = strlen(oldpath);
  64         p = src;
  65         /* if src starts with oldpath, look for oldname only after
  66          * that path */
  67         if (strncmp(src, oldpath, l1) == 0) {
  68                 p += l1;
  69                 nlen += (strlen(lxcpath) - l1);
  70         }
  71         l2 = strlen(oldname);
  72         while ((p = strstr(p, oldname)) != NULL) {
  73                 p += l2;
  74                 nlen += strlen(name) - l2;
  75         }
  76
  77         ret = malloc(nlen);
  78         if (!ret)
  79                 return NULL;
  80
  81         p = ret;
  82         if (strncmp(src, oldpath, l1) == 0) {
  83                 p += sprintf(p, "%s", lxcpath);
  84                 src += l1;
  85         }
  86
  87         while ((p2 = strstr(src, oldname)) != NULL) {
  88                 /* copy text up to oldname */
  89                 strncpy(p, src, p2 - src);
  90                 /* move target pointer (p) */
  91                 p += p2 - src;
  92                 /* print new name in place of oldname */
  93                 p += sprintf(p, "%s", name);
  94                 /* move src to end of oldname */
  95                 src = p2 + l2;
  96         }
  97         /* copy the rest of src */
  98         sprintf(p, "%s", src);
  99         return ret;
 100 }
 101
 102 /*
 103  * attach_block_device returns true if all went well,
 104  * meaning either a block device was attached or was not
 105  * needed.  It returns false if something went wrong and
 106  * container startup should be stopped.
 107  */
 108 bool attach_block_device(struct lxc_conf *conf)
 109 {
 110         char *path;
 111
 112         if (!conf->rootfs.path)
 113                 return true;
 114
 115         path = conf->rootfs.path;
 116         if (!requires_nbd(path))
 117                 return true;
 118
 119         path = strchr(path, ':');
 120         if (!path)
 121                 return false;
 122
 123         path++;
 124         if (!attach_nbd(path, conf))
 125                 return false;
 126
 127         return true;
 128 }
 129
 130 /*
 131  * return block size of dev->src in units of bytes
 132  */
 133 int blk_getsize(struct lxc_storage *bdev, uint64_t *size)
 134 {
 135         int fd, ret;
 136         char *src;
 137
 138         src = lxc_storage_get_path(bdev->src, bdev->type);
 139         fd = open(src, O_RDONLY);
 140         if (fd < 0)
 141                 return -1;
 142
 143         /* size of device in bytes */
 144         ret = ioctl(fd, BLKGETSIZE64, size);
 145         close(fd);
 146         return ret;
 147 }
 148
 149 void detach_block_device(struct lxc_conf *conf)
 150 {
 151         if (conf->nbd_idx != -1)
 152                 detach_nbd_idx(conf->nbd_idx);
 153 }
 154
 155 /*
 156  * Given a lxc_storage (presumably blockdev-based), detect the fstype
 157  * by trying mounting (in a private mntns) it.
 158  * @lxc_storage: bdev to investigate
 159  * @type: preallocated char* in which to write the fstype
 160  * @len: length of passed in char*
 161  * Returns length of fstype, of -1 on error
 162  */
 163 int detect_fs(struct lxc_storage *bdev, char *type, int len)
 164 {
 165         int p[2], ret;
 166         size_t linelen;
 167         pid_t pid;
 168         FILE *f;
 169         char *sp1, *sp2, *sp3, *srcdev, *line = NULL;
 170
 171         if (!bdev || !bdev->src || !bdev->dest)
 172                 return -1;
 173
 174         srcdev = lxc_storage_get_path(bdev->src, bdev->type);
 175
 176         ret = pipe(p);
 177         if (ret < 0)
 178                 return -1;
 179
 180         if ((pid = fork()) < 0)
 181                 return -1;
 182
 183         if (pid > 0) {
 184                 int status;
 185                 close(p[1]);
 186                 memset(type, 0, len);
 187                 ret = read(p[0], type, len - 1);
 188                 close(p[0]);
 189                 if (ret < 0) {
 190                         SYSERROR("error reading from pipe");
 191                         wait(&status);
 192                         return -1;
 193                 } else if (ret == 0) {
 194                         ERROR("child exited early - fstype not found");
 195                         wait(&status);
 196                         return -1;
 197                 }
 198                 wait(&status);
 199                 type[len - 1] = '\0';
 200                 INFO("detected fstype %s for %s", type, srcdev);
 201                 return ret;
 202         }
 203
 204         if (unshare(CLONE_NEWNS) < 0)
 205                 exit(1);
 206
 207         if (detect_shared_rootfs()) {
 208                 if (mount(NULL, "/", NULL, MS_SLAVE | MS_REC, NULL)) {
 209                         SYSERROR("Failed to make / rslave");
 210                         ERROR("Continuing...");
 211                 }
 212         }
 213
 214         ret = mount_unknown_fs(srcdev, bdev->dest, bdev->mntopts);
 215         if (ret < 0) {
 216                 ERROR("failed mounting %s onto %s to detect fstype", srcdev,
 217                       bdev->dest);
 218                 exit(1);
 219         }
 220
 221         /* if symlink, get the real dev name */
 222         char devpath[MAXPATHLEN];
 223         char *l = linkderef(srcdev, devpath);
 224         if (!l)
 225                 exit(1);
 226         f = fopen("/proc/self/mounts", "r");
 227         if (!f)
 228                 exit(1);
 229
 230         while (getline(&line, &linelen, f) != -1) {
 231                 sp1 = strchr(line, ' ');
 232                 if (!sp1)
 233                         exit(1);
 234                 *sp1 = '\0';
 235                 if (strcmp(line, l))
 236                         continue;
 237                 sp2 = strchr(sp1 + 1, ' ');
 238                 if (!sp2)
 239                         exit(1);
 240                 *sp2 = '\0';
 241                 sp3 = strchr(sp2 + 1, ' ');
 242                 if (!sp3)
 243                         exit(1);
 244                 *sp3 = '\0';
 245                 sp2++;
 246                 if (write(p[1], sp2, strlen(sp2)) != strlen(sp2))
 247                         exit(1);
 248
 249                 exit(0);
 250         }
 251
 252         exit(1);
 253 }
 254
 255 int do_mkfs_exec_wrapper(void *args)
 256 {
 257         int ret;
 258         char *mkfs;
 259         char **data = args;
 260         /* strlen("mkfs.")
 261          * +
 262          * strlen(data[0])
 263          * +
 264          * \0
 265          */
 266         size_t len = 5 + strlen(data[0]) + 1;
 267
 268         mkfs = malloc(len);
 269         if (!mkfs)
 270                 return -1;
 271
 272         ret = snprintf(mkfs, len, "mkfs.%s", data[0]);
 273         if (ret < 0 || (size_t)ret >= len) {
 274                 free(mkfs);
 275                 return -1;
 276         }
 277
 278         TRACE("executing \"%s %s\"", mkfs, data[1]);
 279         execlp(mkfs, mkfs, data[1], (char *)NULL);
 280         SYSERROR("failed to run \"%s %s \"", mkfs, data[1]);
 281         return -1;
 282 }
 283
 284 /*
 285  * This will return 1 for physical disks, qemu-nbd, loop, etc right now only lvm
 286  * is a block device.
 287  */
 288 int is_blktype(struct lxc_storage *b)
 289 {
 290         if (strcmp(b->type, "lvm") == 0)
 291                 return 1;
 292
 293         return 0;
 294 }
 295
 296 int mount_unknown_fs(const char *rootfs, const char *target,
 297                      const char *options)
 298 {
 299         size_t i;
 300         int ret;
 301         struct cbarg {
 302                 const char *rootfs;
 303                 const char *target;
 304                 const char *options;
 305         } cbarg = {
 306             .rootfs = rootfs,
 307             .target = target,
 308             .options = options,
 309         };
 310
 311         /*
 312          * find the filesystem type with brute force:
 313          * first we check with /etc/filesystems, in case the modules
 314          * are auto-loaded and fall back to the supported kernel fs
 315          */
 316         char *fsfile[] = {
 317             "/etc/filesystems",
 318             "/proc/filesystems",
 319         };
 320
 321         for (i = 0; i < sizeof(fsfile) / sizeof(fsfile[0]); i++) {
 322                 if (access(fsfile[i], F_OK))
 323                         continue;
 324
 325                 ret = lxc_file_for_each_line(fsfile[i], find_fstype_cb, &cbarg);
 326                 if (ret < 0) {
 327                         ERROR("failed to parse '%s'", fsfile[i]);
 328                         return -1;
 329                 }
 330
 331                 if (ret)
 332                         return 0;
 333         }
 334
 335         ERROR("failed to determine fs type for '%s'", rootfs);
 336         return -1;
 337 }
 338
 339 /*
 340  * These are copied from conf.c.  However as conf.c will be moved to using
 341  * the callback system, they can be pulled from there eventually, so we
 342  * don't need to pollute utils.c with these low level functions
 343  */
 344 int find_fstype_cb(char *buffer, void *data)
 345 {
 346         struct cbarg {
 347                 const char *rootfs;
 348                 const char *target;
 349                 const char *options;
 350         } *cbarg = data;
 351
 352         unsigned long mntflags;
 353         char *mntdata;
 354         char *fstype;
 355
 356         /* we don't try 'nodev' entries */
 357         if (strstr(buffer, "nodev"))
 358                 return 0;
 359
 360         fstype = buffer;
 361         fstype += lxc_char_left_gc(fstype, strlen(fstype));
 362         fstype[lxc_char_right_gc(fstype, strlen(fstype))] = '\0';
 363
 364         DEBUG("trying to mount '%s'->'%s' with fstype '%s'", cbarg->rootfs,
 365               cbarg->target, fstype);
 366
 367         if (parse_mntopts(cbarg->options, &mntflags, &mntdata) < 0) {
 368                 free(mntdata);
 369                 return 0;
 370         }
 371
 372         if (mount(cbarg->rootfs, cbarg->target, fstype, mntflags, mntdata)) {
 373                 DEBUG("mount failed with error: %s", strerror(errno));
 374                 free(mntdata);
 375                 return 0;
 376         }
 377
 378         free(mntdata);
 379
 380         INFO("mounted '%s' on '%s', with fstype '%s'", cbarg->rootfs,
 381              cbarg->target, fstype);
 382
 383         return 1;
 384 }
 385
 386 char *linkderef(char *path, char *dest)
 387 {
 388         struct stat sbuf;
 389         ssize_t ret;
 390
 391         ret = stat(path, &sbuf);
 392         if (ret < 0)
 393                 return NULL;
 394
 395         if (!S_ISLNK(sbuf.st_mode))
 396                 return path;
 397
 398         ret = readlink(path, dest, MAXPATHLEN);
 399         if (ret < 0) {
 400                 SYSERROR("error reading link %s", path);
 401                 return NULL;
 402         } else if (ret >= MAXPATHLEN) {
 403                 ERROR("link in %s too long", path);
 404                 return NULL;
 405         }
 406         dest[ret] = '\0';
 407
 408         return dest;
 409 }
 410
 411 /*
 412  * is an unprivileged user allowed to make this kind of snapshot
 413  */
 414 bool unpriv_snap_allowed(struct lxc_storage *b, const char *t, bool snap,
 415                          bool maybesnap)
 416 {
 417         if (!t) {
 418                 /* New type will be same as original (unless snap && b->type ==
 419                  * dir, in which case it will be overlayfs -- which is also
 420                  * allowed).
 421                  */
 422                 if (strcmp(b->type, "dir") == 0 ||
 423                     strcmp(b->type, "aufs") == 0 ||
 424                     strcmp(b->type, "overlay") == 0 ||
 425                     strcmp(b->type, "overlayfs") == 0 ||
 426                     strcmp(b->type, "btrfs") == 0 ||
 427                     strcmp(b->type, "loop") == 0)
 428                         return true;
 429
 430                 return false;
 431         }
 432
 433         /* Unprivileged users can copy and snapshot dir, overlayfs, and loop.
 434          * In particular, not zfs, btrfs, or lvm.
 435          */
 436         if (strcmp(t, "dir") == 0 ||
 437             strcmp(t, "aufs") == 0 ||
 438             strcmp(t, "overlay") == 0 ||
 439             strcmp(t, "overlayfs") == 0 ||
 440             strcmp(t, "btrfs") == 0 ||
 441             strcmp(t, "loop") == 0)
 442                 return true;
 443
 444         return false;
 445 }
 446
 447 bool is_valid_storage_type(const char *type)
 448 {
 449         if (strcmp(type, "dir") == 0 ||
 450             strcmp(type, "btrfs") == 0 ||
 451             strcmp(type, "aufs") == 0 ||
 452             strcmp(type, "loop") == 0 ||
 453             strcmp(type, "lvm") == 0 ||
 454             strcmp(type, "nbd") == 0 ||
 455             strcmp(type, "overlay") == 0 ||
 456             strcmp(type, "overlayfs") == 0 ||
 457             strcmp(type, "rbd") == 0 ||
 458             strcmp(type, "zfs") == 0)
 459                 return true;
 460
 461         return false;
 462 }
 463
 464 int storage_destroy_wrapper(void *data)
 465 {
 466         struct lxc_conf *conf = data;
 467
 468         if (setgid(0) < 0) {
 469                 ERROR("Failed to setgid to 0");
 470                 return -1;
 471         }
 472
 473         if (setgroups(0, NULL) < 0)
 474                 WARN("Failed to clear groups");
 475
 476         if (setuid(0) < 0) {
 477                 ERROR("Failed to setuid to 0");
 478                 return -1;
 479         }
 480
 481         if (!storage_destroy(conf))
 482                 return -1;
 483
 484         return 0;
 485 }