src/basic/missing_syscall.h

   1 /* SPDX-License-Identifier: LGPL-2.1-or-later */
   2 #pragma once
   3
   4 /* Missing glibc definitions to access certain kernel APIs */
   5
   6 #include <errno.h>
   7 #include <fcntl.h>
   8 #if HAVE_LINUX_TIME_TYPES_H
   9 /* This header defines __kernel_timespec for us, but is only available since Linux 5.1, hence conditionally
  10  * include this. */
  11 #include <linux/time_types.h>
  12 #endif
  13 #include <signal.h>
  14 #include <sys/syscall.h>
  15 #include <sys/types.h>
  16 #include <sys/wait.h>
  17 #include <unistd.h>
  18
  19 #ifdef ARCH_MIPS
  20 #include <asm/sgidefs.h>
  21 #endif
  22
  23 #include "macro.h"
  24 #include "missing_keyctl.h"
  25 #include "missing_stat.h"
  26 #include "missing_syscall_def.h"
  27
  28 /* linux/kcmp.h */
  29 #ifndef KCMP_FILE /* 3f4994cfc15f38a3159c6e3a4b3ab2e1481a6b02 (3.19) */
  30 #define KCMP_FILE 0
  31 #endif
  32
  33 /* ======================================================================= */
  34
  35 #if !HAVE_PIVOT_ROOT
  36 static inline int missing_pivot_root(const char *new_root, const char *put_old) {
  37         return syscall(__NR_pivot_root, new_root, put_old);
  38 }
  39
  40 #  define pivot_root missing_pivot_root
  41 #endif
  42
  43 /* ======================================================================= */
  44
  45 #if !HAVE_IOPRIO_GET
  46 static inline int missing_ioprio_get(int which, int who) {
  47         return syscall(__NR_ioprio_get, which, who);
  48 }
  49
  50 #  define ioprio_get missing_ioprio_get
  51 #endif
  52
  53 /* ======================================================================= */
  54
  55 #if !HAVE_IOPRIO_SET
  56 static inline int missing_ioprio_set(int which, int who, int ioprio) {
  57         return syscall(__NR_ioprio_set, which, who, ioprio);
  58 }
  59
  60 #  define ioprio_set missing_ioprio_set
  61 #endif
  62
  63 /* ======================================================================= */
  64
  65 #if !HAVE_MEMFD_CREATE
  66 static inline int missing_memfd_create(const char *name, unsigned int flags) {
  67 #  ifdef __NR_memfd_create
  68         return syscall(__NR_memfd_create, name, flags);
  69 #  else
  70         errno = ENOSYS;
  71         return -1;
  72 #  endif
  73 }
  74
  75 #  define memfd_create missing_memfd_create
  76 #endif
  77
  78 /* ======================================================================= */
  79
  80 #if !HAVE_GETRANDOM
  81 /* glibc says getrandom() returns ssize_t */
  82 static inline ssize_t missing_getrandom(void *buffer, size_t count, unsigned flags) {
  83 #  ifdef __NR_getrandom
  84         return syscall(__NR_getrandom, buffer, count, flags);
  85 #  else
  86         errno = ENOSYS;
  87         return -1;
  88 #  endif
  89 }
  90
  91 #  define getrandom missing_getrandom
  92 #endif
  93
  94 /* ======================================================================= */
  95
  96 /* The syscall has been defined since forever, but the glibc wrapper was missing. */
  97 #if !HAVE_GETTID
  98 static inline pid_t missing_gettid(void) {
  99 #  if defined __NR_gettid && __NR_gettid >= 0
 100         return (pid_t) syscall(__NR_gettid);
 101 #  else
 102 #    error "__NR_gettid not defined"
 103 #  endif
 104 }
 105
 106 #  define gettid missing_gettid
 107 #endif
 108
 109 /* ======================================================================= */
 110
 111 #if !HAVE_NAME_TO_HANDLE_AT
 112 struct file_handle {
 113         unsigned int handle_bytes;
 114         int handle_type;
 115         unsigned char f_handle[0];
 116 };
 117
 118 static inline int missing_name_to_handle_at(int fd, const char *name, struct file_handle *handle, int *mnt_id, int flags) {
 119 #  ifdef __NR_name_to_handle_at
 120         return syscall(__NR_name_to_handle_at, fd, name, handle, mnt_id, flags);
 121 #  else
 122         errno = ENOSYS;
 123         return -1;
 124 #  endif
 125 }
 126
 127 #  define name_to_handle_at missing_name_to_handle_at
 128 #endif
 129
 130 /* ======================================================================= */
 131
 132 #if !HAVE_SETNS
 133 static inline int missing_setns(int fd, int nstype) {
 134 #  ifdef __NR_setns
 135         return syscall(__NR_setns, fd, nstype);
 136 #  else
 137         errno = ENOSYS;
 138         return -1;
 139 #  endif
 140 }
 141
 142 #  define setns missing_setns
 143 #endif
 144
 145 /* ======================================================================= */
 146
 147 static inline pid_t raw_getpid(void) {
 148 #if defined(__alpha__)
 149         return (pid_t) syscall(__NR_getxpid);
 150 #else
 151         return (pid_t) syscall(__NR_getpid);
 152 #endif
 153 }
 154
 155 /* ======================================================================= */
 156
 157 #if !HAVE_RENAMEAT2
 158 static inline int missing_renameat2(int oldfd, const char *oldname, int newfd, const char *newname, unsigned flags) {
 159 #  ifdef __NR_renameat2
 160         return syscall(__NR_renameat2, oldfd, oldname, newfd, newname, flags);
 161 #  else
 162         errno = ENOSYS;
 163         return -1;
 164 #  endif
 165 }
 166
 167 #  define renameat2 missing_renameat2
 168 #endif
 169
 170 /* ======================================================================= */
 171
 172 #if !HAVE_KCMP
 173 static inline int missing_kcmp(pid_t pid1, pid_t pid2, int type, unsigned long idx1, unsigned long idx2) {
 174 #  if defined __NR_kcmp && __NR_kcmp >= 0
 175         return syscall(__NR_kcmp, pid1, pid2, type, idx1, idx2);
 176 #  else
 177         errno = ENOSYS;
 178         return -1;
 179 #  endif
 180 }
 181
 182 #  define kcmp missing_kcmp
 183 #endif
 184
 185 /* ======================================================================= */
 186
 187 #if !HAVE_KEYCTL
 188 static inline long missing_keyctl(int cmd, unsigned long arg2, unsigned long arg3, unsigned long arg4, unsigned long arg5) {
 189 #  if defined __NR_keyctl && __NR_keyctl >= 0
 190         return syscall(__NR_keyctl, cmd, arg2, arg3, arg4, arg5);
 191 #  else
 192         errno = ENOSYS;
 193         return -1;
 194 #  endif
 195
 196 #  define keyctl missing_keyctl
 197 }
 198
 199 static inline key_serial_t missing_add_key(const char *type, const char *description, const void *payload, size_t plen, key_serial_t ringid) {
 200 #  if defined __NR_add_key && __NR_add_key >= 0
 201         return syscall(__NR_add_key, type, description, payload, plen, ringid);
 202 #  else
 203         errno = ENOSYS;
 204         return -1;
 205 #  endif
 206
 207 #  define add_key missing_add_key
 208 }
 209
 210 static inline key_serial_t missing_request_key(const char *type, const char *description, const char * callout_info, key_serial_t destringid) {
 211 #  if defined __NR_request_key && __NR_request_key >= 0
 212         return syscall(__NR_request_key, type, description, callout_info, destringid);
 213 #  else
 214         errno = ENOSYS;
 215         return -1;
 216 #  endif
 217
 218 #  define request_key missing_request_key
 219 }
 220 #endif
 221
 222 /* ======================================================================= */
 223
 224 #if !HAVE_COPY_FILE_RANGE
 225 static inline ssize_t missing_copy_file_range(int fd_in, loff_t *off_in,
 226                                               int fd_out, loff_t *off_out,
 227                                               size_t len,
 228                                               unsigned int flags) {
 229 #  ifdef __NR_copy_file_range
 230         return syscall(__NR_copy_file_range, fd_in, off_in, fd_out, off_out, len, flags);
 231 #  else
 232         errno = ENOSYS;
 233         return -1;
 234 #  endif
 235 }
 236
 237 #  define copy_file_range missing_copy_file_range
 238 #endif
 239
 240 /* ======================================================================= */
 241
 242 #if !HAVE_BPF
 243 union bpf_attr;
 244
 245 static inline int missing_bpf(int cmd, union bpf_attr *attr, size_t size) {
 246 #ifdef __NR_bpf
 247         return (int) syscall(__NR_bpf, cmd, attr, size);
 248 #else
 249         errno = ENOSYS;
 250         return -1;
 251 #endif
 252 }
 253
 254 #  define bpf missing_bpf
 255 #endif
 256
 257 /* ======================================================================= */
 258
 259 #if !HAVE_STATX
 260 struct statx;
 261
 262 static inline ssize_t missing_statx(int dfd, const char *filename, unsigned flags, unsigned int mask, struct statx *buffer) {
 263 #  ifdef __NR_statx
 264         return syscall(__NR_statx, dfd, filename, flags, mask, buffer);
 265 #  else
 266         errno = ENOSYS;
 267         return -1;
 268 #  endif
 269 }
 270 #endif
 271
 272 /* This typedef is supposed to be always defined. */
 273 typedef struct statx struct_statx;
 274
 275 #if !HAVE_STATX
 276 #  define statx(dfd, filename, flags, mask, buffer) missing_statx(dfd, filename, flags, mask, buffer)
 277 #endif
 278
 279 /* ======================================================================= */
 280
 281 #if !HAVE_SET_MEMPOLICY
 282 enum {
 283         MPOL_DEFAULT,
 284         MPOL_PREFERRED,
 285         MPOL_BIND,
 286         MPOL_INTERLEAVE,
 287         MPOL_LOCAL,
 288 };
 289
 290 static inline long missing_set_mempolicy(int mode, const unsigned long *nodemask,
 291                            unsigned long maxnode) {
 292         long i;
 293 #  if defined __NR_set_mempolicy && __NR_set_mempolicy >= 0
 294         i = syscall(__NR_set_mempolicy, mode, nodemask, maxnode);
 295 #  else
 296         errno = ENOSYS;
 297         i = -1;
 298 #  endif
 299         return i;
 300 }
 301
 302 #  define set_mempolicy missing_set_mempolicy
 303 #endif
 304
 305 #if !HAVE_GET_MEMPOLICY
 306 static inline long missing_get_mempolicy(int *mode, unsigned long *nodemask,
 307                            unsigned long maxnode, void *addr,
 308                            unsigned long flags) {
 309         long i;
 310 #  if defined __NR_get_mempolicy && __NR_get_mempolicy >= 0
 311         i = syscall(__NR_get_mempolicy, mode, nodemask, maxnode, addr, flags);
 312 #  else
 313         errno = ENOSYS;
 314         i = -1;
 315 #  endif
 316         return i;
 317 }
 318
 319 #  define get_mempolicy missing_get_mempolicy
 320 #endif
 321
 322 /* ======================================================================= */
 323
 324 #if !HAVE_PIDFD_SEND_SIGNAL
 325 static inline int missing_pidfd_send_signal(int fd, int sig, siginfo_t *info, unsigned flags) {
 326 #  ifdef __NR_pidfd_send_signal
 327         return syscall(__NR_pidfd_send_signal, fd, sig, info, flags);
 328 #  else
 329         errno = ENOSYS;
 330         return -1;
 331 #  endif
 332 }
 333
 334 #  define pidfd_send_signal missing_pidfd_send_signal
 335 #endif
 336
 337 #if !HAVE_PIDFD_OPEN
 338 static inline int missing_pidfd_open(pid_t pid, unsigned flags) {
 339 #  ifdef __NR_pidfd_open
 340         return syscall(__NR_pidfd_open, pid, flags);
 341 #  else
 342         errno = ENOSYS;
 343         return -1;
 344 #  endif
 345 }
 346
 347 #  define pidfd_open missing_pidfd_open
 348 #endif
 349
 350 /* ======================================================================= */
 351
 352 #if !HAVE_RT_SIGQUEUEINFO
 353 static inline int missing_rt_sigqueueinfo(pid_t tgid, int sig, siginfo_t *info) {
 354 #  if defined __NR_rt_sigqueueinfo && __NR_rt_sigqueueinfo >= 0
 355         return syscall(__NR_rt_sigqueueinfo, tgid, sig, info);
 356 #  else
 357 #    error "__NR_rt_sigqueueinfo not defined"
 358 #  endif
 359 }
 360
 361 #  define rt_sigqueueinfo missing_rt_sigqueueinfo
 362 #endif
 363
 364 /* ======================================================================= */
 365
 366 #if !HAVE_RT_TGSIGQUEUEINFO
 367 static inline int missing_rt_tgsigqueueinfo(pid_t tgid, pid_t tid, int sig, siginfo_t *info) {
 368 #  if defined __NR_rt_tgsigqueueinfo && __NR_rt_tgsigqueueinfo >= 0
 369         return syscall(__NR_rt_tgsigqueueinfo, tgid, tid, sig, info);
 370 #  else
 371 #    error "__NR_rt_tgsigqueueinfo not defined"
 372 #  endif
 373 }
 374
 375 #  define rt_tgsigqueueinfo missing_rt_tgsigqueueinfo
 376 #endif
 377
 378 /* ======================================================================= */
 379
 380 #if !HAVE_EXECVEAT
 381 static inline int missing_execveat(int dirfd, const char *pathname,
 382                                    char *const argv[], char *const envp[],
 383                                    int flags) {
 384 #  if defined __NR_execveat && __NR_execveat >= 0
 385         return syscall(__NR_execveat, dirfd, pathname, argv, envp, flags);
 386 #  else
 387         errno = ENOSYS;
 388         return -1;
 389 #  endif
 390 }
 391
 392 #  undef AT_EMPTY_PATH
 393 #  define AT_EMPTY_PATH 0x1000
 394 #  define execveat missing_execveat
 395 #endif
 396
 397 /* ======================================================================= */
 398
 399 #if !HAVE_CLOSE_RANGE
 400 static inline int missing_close_range(int first_fd, int end_fd, unsigned flags) {
 401 #  ifdef __NR_close_range
 402         /* Kernel-side the syscall expects fds as unsigned integers (just like close() actually), while
 403          * userspace exclusively uses signed integers for fds. We don't know just yet how glibc is going to
 404          * wrap this syscall, but let's assume it's going to be similar to what they do for close(),
 405          * i.e. make the same unsigned → signed type change from the raw kernel syscall compared to the
 406          * userspace wrapper. There's only one caveat for this: unlike for close() there's the special
 407          * UINT_MAX fd value for the 'end_fd' argument. Let's safely map that to -1 here. And let's refuse
 408          * any other negative values. */
 409         if ((first_fd < 0) || (end_fd < 0 && end_fd != -1)) {
 410                 errno = -EBADF;
 411                 return -1;
 412         }
 413
 414         return syscall(__NR_close_range,
 415                        (unsigned) first_fd,
 416                        end_fd == -1 ? UINT_MAX : (unsigned) end_fd, /* Of course, the compiler should figure out that this is the identity mapping IRL */
 417                        flags);
 418 #  else
 419         errno = ENOSYS;
 420         return -1;
 421 #  endif
 422 }
 423
 424 #  define close_range missing_close_range
 425 #endif
 426
 427 /* ======================================================================= */
 428
 429 #if !HAVE_MOUNT_SETATTR
 430
 431 #if !HAVE_STRUCT_MOUNT_ATTR
 432 struct mount_attr {
 433         uint64_t attr_set;
 434         uint64_t attr_clr;
 435         uint64_t propagation;
 436         uint64_t userns_fd;
 437 };
 438 #else
 439 struct mount_attr;
 440 #endif
 441
 442 #ifndef MOUNT_ATTR_RDONLY
 443 #define MOUNT_ATTR_RDONLY       0x00000001 /* Mount read-only */
 444 #endif
 445
 446 #ifndef MOUNT_ATTR_NOSUID
 447 #define MOUNT_ATTR_NOSUID       0x00000002 /* Ignore suid and sgid bits */
 448 #endif
 449
 450 #ifndef MOUNT_ATTR_NODEV
 451 #define MOUNT_ATTR_NODEV        0x00000004 /* Disallow access to device special files */
 452 #endif
 453
 454 #ifndef MOUNT_ATTR_NOEXEC
 455 #define MOUNT_ATTR_NOEXEC       0x00000008 /* Disallow program execution */
 456 #endif
 457
 458 #ifndef MOUNT_ATTR__ATIME
 459 #define MOUNT_ATTR__ATIME       0x00000070 /* Setting on how atime should be updated */
 460 #endif
 461
 462 #ifndef MOUNT_ATTR_RELATIME
 463 #define MOUNT_ATTR_RELATIME     0x00000000 /* - Update atime relative to mtime/ctime. */
 464 #endif
 465
 466 #ifndef MOUNT_ATTR_NOATIME
 467 #define MOUNT_ATTR_NOATIME      0x00000010 /* - Do not update access times. */
 468 #endif
 469
 470 #ifndef MOUNT_ATTR_STRICTATIME
 471 #define MOUNT_ATTR_STRICTATIME  0x00000020 /* - Always perform atime updates */
 472 #endif
 473
 474 #ifndef MOUNT_ATTR_NODIRATIME
 475 #define MOUNT_ATTR_NODIRATIME   0x00000080 /* Do not update directory access times */
 476 #endif
 477
 478 #ifndef MOUNT_ATTR_IDMAP
 479 #define MOUNT_ATTR_IDMAP        0x00100000 /* Idmap mount to @userns_fd in struct mount_attr. */
 480 #endif
 481
 482 #ifndef MOUNT_ATTR_NOSYMFOLLOW
 483 #define MOUNT_ATTR_NOSYMFOLLOW  0x00200000 /* Do not follow symlinks */
 484 #endif
 485
 486 #ifndef MOUNT_ATTR_SIZE_VER0
 487 #define MOUNT_ATTR_SIZE_VER0    32 /* sizeof first published struct */
 488 #endif
 489
 490 #ifndef AT_RECURSIVE
 491 #define AT_RECURSIVE 0x8000
 492 #endif
 493
 494 static inline int missing_mount_setattr(
 495                 int dfd,
 496                 const char *path,
 497                 unsigned flags,
 498                 struct mount_attr *attr,
 499                 size_t size) {
 500
 501 #  if defined __NR_mount_setattr && __NR_mount_setattr >= 0
 502         return syscall(__NR_mount_setattr, dfd, path, flags, attr, size);
 503 #  else
 504         errno = ENOSYS;
 505         return -1;
 506 #  endif
 507 }
 508
 509 #  define mount_setattr missing_mount_setattr
 510 #endif
 511
 512 /* ======================================================================= */
 513
 514 #if !HAVE_OPEN_TREE
 515
 516 #ifndef OPEN_TREE_CLONE
 517 #define OPEN_TREE_CLONE 1
 518 #endif
 519
 520 #ifndef OPEN_TREE_CLOEXEC
 521 #define OPEN_TREE_CLOEXEC O_CLOEXEC
 522 #endif
 523
 524 static inline int missing_open_tree(
 525                 int dfd,
 526                 const char *filename,
 527                 unsigned flags) {
 528
 529 #  if defined __NR_open_tree && __NR_open_tree >= 0
 530         return syscall(__NR_open_tree, dfd, filename, flags);
 531 #  else
 532         errno = ENOSYS;
 533         return -1;
 534 #  endif
 535 }
 536
 537 #  define open_tree missing_open_tree
 538 #endif
 539
 540 /* ======================================================================= */
 541
 542 #ifndef MOVE_MOUNT_BENEATH
 543 #define MOVE_MOUNT_BENEATH 0x00000200
 544 #endif
 545
 546 #if !HAVE_MOVE_MOUNT
 547
 548 #ifndef MOVE_MOUNT_F_EMPTY_PATH
 549 #define MOVE_MOUNT_F_EMPTY_PATH 0x00000004 /* Empty from path permitted */
 550 #endif
 551
 552 #ifndef MOVE_MOUNT_T_EMPTY_PATH
 553 #define MOVE_MOUNT_T_EMPTY_PATH 0x00000040 /* Empty to path permitted */
 554 #endif
 555
 556 static inline int missing_move_mount(
 557                 int from_dfd,
 558                 const char *from_pathname,
 559                 int to_dfd,
 560                 const char *to_pathname,
 561                 unsigned flags) {
 562
 563 #  if defined __NR_move_mount && __NR_move_mount >= 0
 564         return syscall(__NR_move_mount, from_dfd, from_pathname, to_dfd, to_pathname, flags);
 565 #  else
 566         errno = ENOSYS;
 567         return -1;
 568 #  endif
 569 }
 570
 571 #  define move_mount missing_move_mount
 572 #endif
 573
 574 /* ======================================================================= */
 575
 576 #if !HAVE_FSOPEN
 577
 578 #ifndef FSOPEN_CLOEXEC
 579 #define FSOPEN_CLOEXEC 0x00000001
 580 #endif
 581
 582 static inline int missing_fsopen(const char *fsname, unsigned flags) {
 583 #  if defined __NR_fsopen && __NR_fsopen >= 0
 584         return syscall(__NR_fsopen, fsname, flags);
 585 #  else
 586         errno = ENOSYS;
 587         return -1;
 588 #  endif
 589 }
 590
 591 #  define fsopen missing_fsopen
 592 #endif
 593
 594 /* ======================================================================= */
 595
 596 #if !HAVE_FSCONFIG
 597
 598 #ifndef FSCONFIG_SET_FLAG
 599 #define FSCONFIG_SET_FLAG 0 /* Set parameter, supplying no value */
 600 #endif
 601
 602 #ifndef FSCONFIG_SET_STRING
 603 #define FSCONFIG_SET_STRING 1 /* Set parameter, supplying a string value */
 604 #endif
 605
 606 #ifndef FSCONFIG_SET_FD
 607 #define FSCONFIG_SET_FD 5 /* Set parameter, supplying an object by fd */
 608 #endif
 609
 610 #ifndef FSCONFIG_CMD_CREATE
 611 #define FSCONFIG_CMD_CREATE 6 /* Invoke superblock creation */
 612 #endif
 613
 614 static inline int missing_fsconfig(int fd, unsigned cmd, const char *key, const void *value, int aux) {
 615 #  if defined __NR_fsconfig && __NR_fsconfig >= 0
 616         return syscall(__NR_fsconfig, fd, cmd, key, value, aux);
 617 #  else
 618         errno = ENOSYS;
 619         return -1;
 620 #  endif
 621 }
 622
 623 #  define fsconfig missing_fsconfig
 624 #endif
 625
 626 /* ======================================================================= */
 627
 628 #if !HAVE_FSMOUNT
 629
 630 #ifndef FSMOUNT_CLOEXEC
 631 #define FSMOUNT_CLOEXEC 0x00000001
 632 #endif
 633
 634 static inline int missing_fsmount(int fd, unsigned flags, unsigned ms_flags) {
 635 #  if defined __NR_fsmount && __NR_fsmount >= 0
 636         return syscall(__NR_fsmount, fd, flags, ms_flags);
 637 #  else
 638         errno = ENOSYS;
 639         return -1;
 640 #  endif
 641 }
 642
 643 #  define fsmount missing_fsmount
 644 #endif
 645
 646 /* ======================================================================= */
 647
 648 #if !HAVE_GETDENTS64
 649
 650 static inline ssize_t missing_getdents64(int fd, void *buffer, size_t length) {
 651 #  if defined __NR_getdents64 && __NR_getdents64 >= 0
 652         return syscall(__NR_getdents64, fd, buffer, length);
 653 #  else
 654         errno = ENOSYS;
 655         return -1;
 656 #  endif
 657 }
 658
 659 #  define getdents64 missing_getdents64
 660 #endif
 661
 662 /* ======================================================================= */
 663
 664 /* glibc does not provide clone() on ia64, only clone2(). Not only that, but it also doesn't provide a
 665  * prototype, only the symbol in the shared library (it provides a prototype for clone(), but not the
 666  * symbol in the shared library). */
 667 #if defined(__ia64__)
 668 int __clone2(int (*fn)(void *), void *stack_base, size_t stack_size, int flags, void *arg);
 669 #define HAVE_CLONE 0
 670 #else
 671 /* We know that everywhere else clone() is available, so we don't bother with a meson check (that takes time
 672  * at build time) and just define it. Once the kernel drops ia64 support, we can drop this too. */
 673 #define HAVE_CLONE 1
 674 #endif