src/lxc/utils.h

   1 /*
   2  * lxc: linux Container library
   3  *
   4  * (C) Copyright IBM Corp. 2007, 2008
   5  *
   6  * Authors:
   7  * Daniel Lezcano <daniel.lezcano at free.fr>
   8  *
   9  * This library 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 library 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 library; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  22  */
  23 #ifndef __LXC_UTILS_H
  24 #define __LXC_UTILS_H
  25
  26 /* Properly support loop devices on 32bit systems. */
  27 #define _FILE_OFFSET_BITS 64
  28
  29 #include "config.h"
  30
  31 #include <errno.h>
  32 #include <stdarg.h>
  33 #include <stdio.h>
  34 #include <stdbool.h>
  35 #include <unistd.h>
  36 #include <linux/loop.h>
  37 #include <linux/magic.h>
  38 #include <linux/types.h>
  39 #include <sys/syscall.h>
  40 #include <sys/types.h>
  41 #include <sys/vfs.h>
  42
  43 #ifdef HAVE_LINUX_MEMFD_H
  44 #include <linux/memfd.h>
  45 #endif
  46
  47 #include "initutils.h"
  48
  49 /* Define __S_ISTYPE if missing from the C library. */
  50 #ifndef __S_ISTYPE
  51 #define __S_ISTYPE(mode, mask) (((mode)&S_IFMT) == (mask))
  52 #endif
  53
  54 #if HAVE_LIBCAP
  55 #ifndef CAP_SETFCAP
  56 #define CAP_SETFCAP 31
  57 #endif
  58
  59 #ifndef CAP_MAC_OVERRIDE
  60 #define CAP_MAC_OVERRIDE 32
  61 #endif
  62
  63 #ifndef CAP_MAC_ADMIN
  64 #define CAP_MAC_ADMIN 33
  65 #endif
  66 #endif
  67
  68 #ifndef PR_CAPBSET_DROP
  69 #define PR_CAPBSET_DROP 24
  70 #endif
  71
  72 #ifndef LO_FLAGS_AUTOCLEAR
  73 #define LO_FLAGS_AUTOCLEAR 4
  74 #endif
  75
  76 #ifndef CAP_SETUID
  77 #define CAP_SETUID 7
  78 #endif
  79
  80 #ifndef CAP_SETGID
  81 #define CAP_SETGID 6
  82 #endif
  83
  84 /* needed for cgroup automount checks, regardless of whether we
  85  * have included linux/capability.h or not */
  86 #ifndef CAP_SYS_ADMIN
  87 #define CAP_SYS_ADMIN 21
  88 #endif
  89
  90 #ifndef CGROUP_SUPER_MAGIC
  91 #define CGROUP_SUPER_MAGIC 0x27e0eb
  92 #endif
  93
  94 #ifndef CGROUP2_SUPER_MAGIC
  95 #define CGROUP2_SUPER_MAGIC 0x63677270
  96 #endif
  97
  98 /* Useful macros */
  99 /* Maximum number for 64 bit integer is a string with 21 digits: 2^64 - 1 = 21 */
 100 #define LXC_NUMSTRLEN64 21
 101 #define LXC_LINELEN 4096
 102 #define LXC_IDMAPLEN 4096
 103 #define LXC_MAX_BUFFER 4096
 104
 105 /* returns 1 on success, 0 if there were any failures */
 106 extern int lxc_rmdir_onedev(const char *path, const char *exclude);
 107 extern int get_u16(unsigned short *val, const char *arg, int base);
 108 extern int mkdir_p(const char *dir, mode_t mode);
 109 extern char *get_rundir(void);
 110
 111 /* Define getline() if missing from the C library */
 112 #ifndef HAVE_GETLINE
 113 #ifdef HAVE_FGETLN
 114 #include <../include/getline.h>
 115 #endif
 116 #endif
 117
 118 #if !defined(__NR_setns) && !defined(__NR_set_ns)
 119         #if defined(__x86_64__)
 120                 #define __NR_setns 308
 121         #elif defined(__i386__)
 122                 #define __NR_setns 346
 123         #elif defined(__arm__)
 124                 #define __NR_setns 375
 125         #elif defined(__aarch64__)
 126                 #define __NR_setns 375
 127         #elif defined(__powerpc__)
 128                 #define __NR_setns 350
 129         #elif defined(__s390__)
 130                 #define __NR_setns 339
 131         #endif
 132 #endif
 133
 134 /* Define setns() if missing from the C library */
 135 #ifndef HAVE_SETNS
 136 static inline int setns(int fd, int nstype)
 137 {
 138 #ifdef __NR_setns
 139         return syscall(__NR_setns, fd, nstype);
 140 #elif defined(__NR_set_ns)
 141         return syscall(__NR_set_ns, fd, nstype);
 142 #else
 143         errno = ENOSYS;
 144         return -1;
 145 #endif
 146 }
 147 #endif
 148
 149 /* Define sethostname() if missing from the C library */
 150 #ifndef HAVE_SETHOSTNAME
 151 static inline int sethostname(const char * name, size_t len)
 152 {
 153 #ifdef __NR_sethostname
 154 return syscall(__NR_sethostname, name, len);
 155 #else
 156 errno = ENOSYS;
 157 return -1;
 158 #endif
 159 }
 160 #endif
 161
 162 /* Define unshare() if missing from the C library */
 163 #ifndef HAVE_UNSHARE
 164 static inline int unshare(int flags)
 165 {
 166 #ifdef __NR_unshare
 167         return syscall(__NR_unshare, flags);
 168 #else
 169         errno = ENOSYS;
 170         return -1;
 171 #endif
 172 }
 173 #else
 174 extern int unshare(int);
 175 #endif
 176
 177 /* Define signalfd() if missing from the C library */
 178 #ifdef HAVE_SYS_SIGNALFD_H
 179 #  include <sys/signalfd.h>
 180 #else
 181 /* assume kernel headers are too old */
 182 #include <stdint.h>
 183 struct signalfd_siginfo
 184 {
 185         uint32_t ssi_signo;
 186         int32_t ssi_errno;
 187         int32_t ssi_code;
 188         uint32_t ssi_pid;
 189         uint32_t ssi_uid;
 190         int32_t ssi_fd;
 191         uint32_t ssi_tid;
 192         uint32_t ssi_band;
 193         uint32_t ssi_overrun;
 194         uint32_t ssi_trapno;
 195         int32_t ssi_status;
 196         int32_t ssi_int;
 197         uint64_t ssi_ptr;
 198         uint64_t ssi_utime;
 199         uint64_t ssi_stime;
 200         uint64_t ssi_addr;
 201         uint8_t __pad[48];
 202 };
 203
 204 #  ifndef __NR_signalfd4
 205 /* assume kernel headers are too old */
 206 #    if __i386__
 207 #      define __NR_signalfd4 327
 208 #    elif __x86_64__
 209 #      define __NR_signalfd4 289
 210 #    elif __powerpc__
 211 #      define __NR_signalfd4 313
 212 #    elif __s390x__
 213 #      define __NR_signalfd4 322
 214 #    elif __arm__
 215 #      define __NR_signalfd4 355
 216 #    elif __mips__ && _MIPS_SIM == _ABIO32
 217 #      define __NR_signalfd4 4324
 218 #    elif __mips__ && _MIPS_SIM == _ABI64
 219 #      define __NR_signalfd4 5283
 220 #    elif __mips__ && _MIPS_SIM == _ABIN32
 221 #      define __NR_signalfd4 6287
 222 #    endif
 223 #endif
 224
 225 #  ifndef __NR_signalfd
 226 /* assume kernel headers are too old */
 227 #    if __i386__
 228 #      define __NR_signalfd 321
 229 #    elif __x86_64__
 230 #      define __NR_signalfd 282
 231 #    elif __powerpc__
 232 #      define __NR_signalfd 305
 233 #    elif __s390x__
 234 #      define __NR_signalfd 316
 235 #    elif __arm__
 236 #      define __NR_signalfd 349
 237 #    elif __mips__ && _MIPS_SIM == _ABIO32
 238 #      define __NR_signalfd 4317
 239 #    elif __mips__ && _MIPS_SIM == _ABI64
 240 #      define __NR_signalfd 5276
 241 #    elif __mips__ && _MIPS_SIM == _ABIN32
 242 #      define __NR_signalfd 6280
 243 #    endif
 244 #endif
 245
 246 static inline int signalfd(int fd, const sigset_t *mask, int flags)
 247 {
 248         int retval;
 249
 250         retval = syscall (__NR_signalfd4, fd, mask, _NSIG / 8, flags);
 251         if (errno == ENOSYS && flags == 0)
 252                 retval = syscall (__NR_signalfd, fd, mask, _NSIG / 8);
 253         return retval;
 254 }
 255 #endif
 256
 257 /* loop devices */
 258 #ifndef LO_FLAGS_AUTOCLEAR
 259 #define LO_FLAGS_AUTOCLEAR 4
 260 #endif
 261
 262 #ifndef LOOP_CTL_GET_FREE
 263 #define LOOP_CTL_GET_FREE 0x4C82
 264 #endif
 265
 266 /* memfd_create() */
 267 #ifndef MFD_CLOEXEC
 268 #define MFD_CLOEXEC 0x0001U
 269 #endif
 270
 271 #ifndef MFD_ALLOW_SEALING
 272 #define MFD_ALLOW_SEALING 0x0002U
 273 #endif
 274
 275 #ifndef HAVE_MEMFD_CREATE
 276 static inline int memfd_create(const char *name, unsigned int flags) {
 277         #ifndef __NR_memfd_create
 278                 #if defined __i386__
 279                         #define __NR_memfd_create 356
 280                 #elif defined __x86_64__
 281                         #define __NR_memfd_create 319
 282                 #elif defined __arm__
 283                         #define __NR_memfd_create 385
 284                 #elif defined __aarch64__
 285                         #define __NR_memfd_create 279
 286                 #elif defined __s390__
 287                         #define __NR_memfd_create 350
 288                 #elif defined __powerpc__
 289                         #define __NR_memfd_create 360
 290                 #elif defined __sparc__
 291                         #define __NR_memfd_create 348
 292                 #elif defined __blackfin__
 293                         #define __NR_memfd_create 390
 294                 #elif defined __ia64__
 295                         #define __NR_memfd_create 1340
 296                 #elif defined _MIPS_SIM
 297                         #if _MIPS_SIM == _MIPS_SIM_ABI32
 298                                 #define __NR_memfd_create 4354
 299                         #endif
 300                         #if _MIPS_SIM == _MIPS_SIM_NABI32
 301                                 #define __NR_memfd_create 6318
 302                         #endif
 303                         #if _MIPS_SIM == _MIPS_SIM_ABI64
 304                                 #define __NR_memfd_create 5314
 305                         #endif
 306                 #endif
 307         #endif
 308         #ifdef __NR_memfd_create
 309         return syscall(__NR_memfd_create, name, flags);
 310         #else
 311         errno = ENOSYS;
 312         return -1;
 313         #endif
 314 }
 315 #else
 316 extern int memfd_create(const char *name, unsigned int flags);
 317 #endif
 318
 319 static inline int lxc_set_cloexec(int fd)
 320 {
 321         return fcntl(fd, F_SETFD, FD_CLOEXEC);
 322 }
 323
 324 /* Struct to carry child pid from lxc_popen() to lxc_pclose().
 325  * Not an opaque struct to allow direct access to the underlying FILE *
 326  * (i.e., struct lxc_popen_FILE *file; fgets(buf, sizeof(buf), file->f))
 327  * without additional wrappers.
 328  */
 329 struct lxc_popen_FILE {
 330         int pipe;
 331         FILE *f;
 332         pid_t child_pid;
 333 };
 334
 335 /* popen(command, "re") replacement that restores default signal mask
 336  * via sigprocmask(2) (unblocks all signals) after fork(2) but prior to calling exec(3).
 337  * In short, popen(command, "re") does pipe() + fork()                 + exec()
 338  * while lxc_popen(command)       does pipe() + fork() + sigprocmask() + exec().
 339  * Returns pointer to struct lxc_popen_FILE, that should be freed with lxc_pclose().
 340  * On error returns NULL.
 341  */
 342 extern struct lxc_popen_FILE *lxc_popen(const char *command);
 343
 344 /* pclose() replacement to be used on struct lxc_popen_FILE *,
 345  * returned by lxc_popen().
 346  * Waits for associated process to terminate, returns its exit status and
 347  * frees resources, pointed to by struct lxc_popen_FILE *.
 348  */
 349 extern int lxc_pclose(struct lxc_popen_FILE *fp);
 350
 351 /**
 352  * BUILD_BUG_ON - break compile if a condition is true.
 353  * @condition: the condition which the compiler should know is false.
 354  *
 355  * If you have some code which relies on certain constants being equal, or
 356  * other compile-time-evaluated condition, you should use BUILD_BUG_ON to
 357  * detect if someone changes it.
 358  *
 359  * The implementation uses gcc's reluctance to create a negative array, but
 360  * gcc (as of 4.4) only emits that error for obvious cases (eg. not arguments
 361  * to inline functions).  So as a fallback we use the optimizer; if it can't
 362  * prove the condition is false, it will cause a link error on the undefined
 363  * "__build_bug_on_failed".  This error message can be harder to track down
 364  * though, hence the two different methods.
 365  */
 366 #ifndef __OPTIMIZE__
 367 #define BUILD_BUG_ON(condition) ((void)sizeof(char[1 - 2*!!(condition)]))
 368 #else
 369 extern int __build_bug_on_failed;
 370 #define BUILD_BUG_ON(condition)                                 \
 371         do {                                                    \
 372                 ((void)sizeof(char[1 - 2*!!(condition)]));      \
 373                 if (condition) __build_bug_on_failed = 1;       \
 374         } while(0)
 375 #endif
 376
 377 /*
 378  * wait on a child we forked
 379  */
 380 extern int wait_for_pid(pid_t pid);
 381 extern int lxc_wait_for_pid_status(pid_t pid);
 382
 383 /* send and receive buffers completely */
 384 extern ssize_t lxc_write_nointr(int fd, const void* buf, size_t count);
 385 extern ssize_t lxc_read_nointr(int fd, void* buf, size_t count);
 386 extern ssize_t lxc_read_nointr_expect(int fd, void *buf, size_t count,
 387                                       const void *expected_buf);
 388 #if HAVE_LIBGNUTLS
 389 #define SHA_DIGEST_LENGTH 20
 390 extern int sha1sum_file(char *fnam, unsigned char *md_value);
 391 #endif
 392
 393 /* read and write whole files */
 394 extern int lxc_write_to_file(const char *filename, const void *buf,
 395                              size_t count, bool add_newline);
 396 extern int lxc_read_from_file(const char *filename, void* buf, size_t count);
 397
 398 /* convert variadic argument lists to arrays (for execl type argument lists) */
 399 extern char** lxc_va_arg_list_to_argv(va_list ap, size_t skip, int do_strdup);
 400 extern const char** lxc_va_arg_list_to_argv_const(va_list ap, size_t skip);
 401
 402 /* Some simple string functions; if they return pointers, they are allocated
 403  * buffers.
 404  */
 405 extern char *lxc_string_replace(const char *needle, const char *replacement,
 406                                 const char *haystack);
 407 extern bool lxc_string_in_array(const char *needle, const char **haystack);
 408 extern char *lxc_string_join(const char *sep, const char **parts,
 409                              bool use_as_prefix);
 410 /* Normalize and split path: Leading and trailing / are removed, multiple
 411  * / are compactified, .. and . are resolved (.. on the top level is considered
 412  * identical to .).
 413  * Examples:
 414  *     /            ->   { NULL }
 415  *     foo/../bar   ->   { bar, NULL }
 416  *     ../../       ->   { NULL }
 417  *     ./bar/baz/.. ->   { bar, NULL }
 418  *     foo//bar     ->   { foo, bar, NULL }
 419  */
 420 extern char **lxc_normalize_path(const char *path);
 421 /* remove multiple slashes from the path, e.g. ///foo//bar -> /foo/bar */
 422 extern char *lxc_deslashify(const char *path);
 423 extern char *lxc_append_paths(const char *first, const char *second);
 424 /* Note: the following two functions use strtok(), so they will never
 425  *       consider an empty element, even if two delimiters are next to
 426  *       each other.
 427  */
 428 extern bool lxc_string_in_list(const char *needle, const char *haystack,
 429                                char sep);
 430 extern char **lxc_string_split(const char *string, char sep);
 431 extern char **lxc_string_split_and_trim(const char *string, char sep);
 432 extern char **lxc_string_split_quoted(char *string);
 433 /* Append string to NULL-terminated string array. */
 434 extern int lxc_append_string(char ***list, char *entry);
 435
 436 /* some simple array manipulation utilities */
 437 typedef void (*lxc_free_fn)(void *);
 438 typedef void *(*lxc_dup_fn)(void *);
 439 extern int lxc_grow_array(void ***array, size_t *capacity, size_t new_size,
 440                           size_t capacity_increment);
 441 extern void lxc_free_array(void **array, lxc_free_fn element_free_fn);
 442 extern size_t lxc_array_len(void **array);
 443
 444 extern void **lxc_append_null_to_array(void **array, size_t count);
 445
 446 /* initialize rand with urandom */
 447 extern int randseed(bool);
 448
 449 /* are we unprivileged with respect to our namespaces */
 450 inline static bool am_guest_unpriv(void) {
 451         return geteuid() != 0;
 452 }
 453
 454 /* are we unprivileged with respect to init_user_ns */
 455 inline static bool am_host_unpriv(void)
 456 {
 457         FILE *f;
 458         uid_t user, host, count;
 459         int ret;
 460
 461         if (geteuid() != 0)
 462                 return true;
 463
 464         /* Now: are we in a user namespace? Because then we're also
 465          * unprivileged.
 466          */
 467         f = fopen("/proc/self/uid_map", "r");
 468         if (!f) {
 469                 return false;
 470         }
 471
 472         ret = fscanf(f, "%u %u %u", &user, &host, &count);
 473         fclose(f);
 474         if (ret != 3) {
 475                 return false;
 476         }
 477
 478         if (user != 0 || host != 0 || count != UINT32_MAX)
 479                 return true;
 480         return false;
 481 }
 482
 483 /*
 484  * parse /proc/self/uid_map to find what @orig maps to
 485  */
 486 extern uid_t get_ns_uid(uid_t orig);
 487
 488 extern bool dir_exists(const char *path);
 489
 490 #define FNV1A_64_INIT ((uint64_t)0xcbf29ce484222325ULL)
 491 extern uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval);
 492
 493 extern int detect_shared_rootfs(void);
 494 extern bool detect_ramfs_rootfs(void);
 495 extern char *on_path(const char *cmd, const char *rootfs);
 496 extern bool file_exists(const char *f);
 497 extern bool cgns_supported(void);
 498 extern char *choose_init(const char *rootfs);
 499 extern int print_to_file(const char *file, const char *content);
 500 extern bool switch_to_ns(pid_t pid, const char *ns);
 501 extern int is_dir(const char *path);
 502 extern char *get_template_path(const char *t);
 503 extern int safe_mount(const char *src, const char *dest, const char *fstype,
 504                       unsigned long flags, const void *data,
 505                       const char *rootfs);
 506 extern int lxc_mount_proc_if_needed(const char *rootfs);
 507 extern int open_devnull(void);
 508 extern int set_stdfds(int fd);
 509 extern int null_stdfds(void);
 510 extern int lxc_count_file_lines(const char *fn);
 511 extern int lxc_preserve_ns(const int pid, const char *ns);
 512
 513 /* Check whether a signal is blocked by a process. */
 514 extern bool task_blocking_signal(pid_t pid, int signal);
 515
 516 /* Helper functions to parse numbers. */
 517 extern int lxc_safe_uint(const char *numstr, unsigned int *converted);
 518 extern int lxc_safe_int(const char *numstr, int *converted);
 519 extern int lxc_safe_long(const char *numstr, long int *converted);
 520 extern int lxc_safe_long_long(const char *numstr, long long int *converted);
 521 extern int lxc_safe_ulong(const char *numstr, unsigned long *converted);
 522 /* Handles B, kb, MB, GB. Detects overflows and reports -ERANGE. */
 523 extern int parse_byte_size_string(const char *s, int64_t *converted);
 524
 525 /* Switch to a new uid and gid. */
 526 extern int lxc_switch_uid_gid(uid_t uid, gid_t gid);
 527 extern int lxc_setgroups(int size, gid_t list[]);
 528
 529 /* Find an unused loop device and associate it with source. */
 530 extern int lxc_prepare_loop_dev(const char *source, char *loop_dev, int flags);
 531
 532 /* Clear all mounts on a given node.
 533  * >= 0 successfully cleared. The number returned is the number of umounts
 534  *      performed.
 535  * < 0  error umounting. Return -errno.
 536  */
 537 extern int lxc_unstack_mountpoint(const char *path, bool lazy);
 538
 539 /*
 540  * run_command runs a command and collect it's std{err,out} output in buf.
 541  *
 542  * @param[out] buf     The buffer where the commands std{err,out] output will be
 543  *                     read into. If no output was produced, buf will be memset
 544  *                     to 0.
 545  * @param[in] buf_size The size of buf. This function will reserve one byte for
 546  *                     \0-termination.
 547  * @param[in] child_fn The function to be run in the child process. This
 548  *                     function must exec.
 549  * @param[in] args     Arguments to be passed to child_fn.
 550  */
 551 extern int run_command(char *buf, size_t buf_size, int (*child_fn)(void *),
 552                        void *args);
 553
 554 /* Concatenate all passed-in strings into one path. Do not fail. If any piece
 555  * is not prefixed with '/', add a '/'.
 556  */
 557 __attribute__((sentinel)) extern char *must_make_path(const char *first, ...);
 558 __attribute__((sentinel)) extern char *must_append_path(char *first, ...);
 559
 560 /* return copy of string @entry;  do not fail. */
 561 extern char *must_copy_string(const char *entry);
 562
 563 /* Re-alllocate a pointer, do not fail */
 564 extern void *must_realloc(void *orig, size_t sz);
 565
 566 /* __typeof__ should be safe to use with all compilers. */
 567 typedef __typeof__(((struct statfs *)NULL)->f_type) fs_type_magic;
 568 extern bool has_fs_type(const char *path, fs_type_magic magic_val);
 569 extern bool is_fs_type(const struct statfs *fs, fs_type_magic magic_val);
 570 extern bool lxc_nic_exists(char *nic);
 571 extern int lxc_make_tmpfile(char *template, bool rm);
 572
 573 static inline uint64_t lxc_getpagesize(void)
 574 {
 575         int64_t pgsz;
 576
 577         pgsz = sysconf(_SC_PAGESIZE);
 578         if (pgsz <= 0)
 579                 pgsz = 1 << 12;
 580
 581         return pgsz;
 582 }
 583
 584 /* If n is not a power of 2 this function will return the next power of 2
 585  * greater than that number. Note that this function always returns the *next*
 586  * power of 2 *greater* that number not the *nearest*. For example, passing 1025
 587  * as argument this function will return 2048 although the closest power of 2
 588  * would be 1024.
 589  * If the caller passes in 0 they will receive 0 in return since this is invalid
 590  * input and 0 is not a power of 2.
 591  */
 592 extern uint64_t lxc_find_next_power2(uint64_t n);
 593
 594 static inline pid_t lxc_raw_gettid(void)
 595 {
 596 #ifdef SYS_gettid
 597         return syscall(SYS_gettid);
 598 #else
 599         return lxc_raw_getpid();
 600 #endif
 601 }
 602
 603 /* Set a signal the child process will receive after the parent has died. */
 604 extern int lxc_set_death_signal(int signal);
 605
 606 #endif /* __LXC_UTILS_H */