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
missing_pivot_root(const char * new_root,const char * put_old)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
missing_ioprio_get(int which,int who)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
missing_ioprio_set(int which,int who,int ioprio)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
missing_memfd_create(const char * name,unsigned int flags)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 */
missing_getrandom(void * buffer,size_t count,unsigned flags)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
missing_gettid(void)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 
missing_name_to_handle_at(int fd,const char * name,struct file_handle * handle,int * mnt_id,int flags)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
missing_setns(int fd,int nstype)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 
raw_getpid(void)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
missing_renameat2(int oldfd,const char * oldname,int newfd,const char * newname,unsigned flags)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
missing_kcmp(pid_t pid1,pid_t pid2,int type,unsigned long idx1,unsigned long idx2)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
missing_keyctl(int cmd,unsigned long arg2,unsigned long arg3,unsigned long arg4,unsigned long arg5)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 
missing_add_key(const char * type,const char * description,const void * payload,size_t plen,key_serial_t ringid)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 
missing_request_key(const char * type,const char * description,const char * callout_info,key_serial_t destringid)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
missing_copy_file_range(int fd_in,loff_t * off_in,int fd_out,loff_t * off_out,size_t len,unsigned int flags)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 
missing_bpf(int cmd,union bpf_attr * attr,size_t size)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 
missing_statx(int dfd,const char * filename,unsigned flags,unsigned int mask,struct statx * buffer)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 
missing_set_mempolicy(int mode,const unsigned long * nodemask,unsigned long maxnode)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
missing_get_mempolicy(int * mode,unsigned long * nodemask,unsigned long maxnode,void * addr,unsigned long flags)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
missing_pidfd_send_signal(int fd,int sig,siginfo_t * info,unsigned flags)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
missing_pidfd_open(pid_t pid,unsigned flags)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
missing_rt_sigqueueinfo(pid_t tgid,int sig,siginfo_t * info)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_EXECVEAT
missing_execveat(int dirfd,const char * pathname,char * const argv[],char * const envp[],int flags)367 static inline int missing_execveat(int dirfd, const char *pathname,
368                                    char *const argv[], char *const envp[],
369                                    int flags) {
370 #  if defined __NR_execveat && __NR_execveat >= 0
371         return syscall(__NR_execveat, dirfd, pathname, argv, envp, flags);
372 #  else
373         errno = ENOSYS;
374         return -1;
375 #  endif
376 }
377 
378 #  undef AT_EMPTY_PATH
379 #  define AT_EMPTY_PATH 0x1000
380 #  define execveat missing_execveat
381 #endif
382 
383 /* ======================================================================= */
384 
385 #if !HAVE_CLOSE_RANGE
missing_close_range(int first_fd,int end_fd,unsigned flags)386 static inline int missing_close_range(int first_fd, int end_fd, unsigned flags) {
387 #  ifdef __NR_close_range
388         /* Kernel-side the syscall expects fds as unsigned integers (just like close() actually), while
389          * userspace exclusively uses signed integers for fds. We don't know just yet how glibc is going to
390          * wrap this syscall, but let's assume it's going to be similar to what they do for close(),
391          * i.e. make the same unsigned → signed type change from the raw kernel syscall compared to the
392          * userspace wrapper. There's only one caveat for this: unlike for close() there's the special
393          * UINT_MAX fd value for the 'end_fd' argument. Let's safely map that to -1 here. And let's refuse
394          * any other negative values. */
395         if ((first_fd < 0) || (end_fd < 0 && end_fd != -1)) {
396                 errno = -EBADF;
397                 return -1;
398         }
399 
400         return syscall(__NR_close_range,
401                        (unsigned) first_fd,
402                        end_fd == -1 ? UINT_MAX : (unsigned) end_fd, /* Of course, the compiler should figure out that this is the identity mapping IRL */
403                        flags);
404 #  else
405         errno = ENOSYS;
406         return -1;
407 #  endif
408 }
409 
410 #  define close_range missing_close_range
411 #endif
412 
413 /* ======================================================================= */
414 
415 #if !HAVE_EPOLL_PWAIT2
416 
417 /* Defined to be equivalent to the kernel's _NSIG_WORDS, i.e. the size of the array of longs that is
418  * encapsulated by sigset_t. */
419 #define KERNEL_NSIG_WORDS (64 / (sizeof(long) * 8))
420 #define KERNEL_NSIG_BYTES (KERNEL_NSIG_WORDS * sizeof(long))
421 
422 struct epoll_event;
423 
missing_epoll_pwait2(int fd,struct epoll_event * events,int maxevents,const struct timespec * timeout,const sigset_t * sigset)424 static inline int missing_epoll_pwait2(
425                 int fd,
426                 struct epoll_event *events,
427                 int maxevents,
428                 const struct timespec *timeout,
429                 const sigset_t *sigset) {
430 
431 #  if defined(__NR_epoll_pwait2) && HAVE_LINUX_TIME_TYPES_H
432         if (timeout) {
433                 /* Convert from userspace timespec to kernel timespec */
434                 struct __kernel_timespec ts = {
435                         .tv_sec = timeout->tv_sec,
436                         .tv_nsec = timeout->tv_nsec,
437                 };
438 
439                 return syscall(__NR_epoll_pwait2, fd, events, maxevents, &ts, sigset, sigset ? KERNEL_NSIG_BYTES : 0);
440         } else
441                 return syscall(__NR_epoll_pwait2, fd, events, maxevents, NULL, sigset, sigset ? KERNEL_NSIG_BYTES : 0);
442 #  else
443         errno = ENOSYS;
444         return -1;
445 #  endif
446 }
447 
448 #  define epoll_pwait2 missing_epoll_pwait2
449 #endif
450 
451 /* ======================================================================= */
452 
453 #if !HAVE_MOUNT_SETATTR
454 
455 #if !HAVE_STRUCT_MOUNT_ATTR
456 struct mount_attr {
457         uint64_t attr_set;
458         uint64_t attr_clr;
459         uint64_t propagation;
460         uint64_t userns_fd;
461 };
462 #else
463 struct mount_attr;
464 #endif
465 
466 #ifndef MOUNT_ATTR_RDONLY
467 #define MOUNT_ATTR_RDONLY       0x00000001 /* Mount read-only */
468 #endif
469 
470 #ifndef MOUNT_ATTR_NOSUID
471 #define MOUNT_ATTR_NOSUID       0x00000002 /* Ignore suid and sgid bits */
472 #endif
473 
474 #ifndef MOUNT_ATTR_NODEV
475 #define MOUNT_ATTR_NODEV        0x00000004 /* Disallow access to device special files */
476 #endif
477 
478 #ifndef MOUNT_ATTR_NOEXEC
479 #define MOUNT_ATTR_NOEXEC       0x00000008 /* Disallow program execution */
480 #endif
481 
482 #ifndef MOUNT_ATTR__ATIME
483 #define MOUNT_ATTR__ATIME       0x00000070 /* Setting on how atime should be updated */
484 #endif
485 
486 #ifndef MOUNT_ATTR_RELATIME
487 #define MOUNT_ATTR_RELATIME     0x00000000 /* - Update atime relative to mtime/ctime. */
488 #endif
489 
490 #ifndef MOUNT_ATTR_NOATIME
491 #define MOUNT_ATTR_NOATIME      0x00000010 /* - Do not update access times. */
492 #endif
493 
494 #ifndef MOUNT_ATTR_STRICTATIME
495 #define MOUNT_ATTR_STRICTATIME  0x00000020 /* - Always perform atime updates */
496 #endif
497 
498 #ifndef MOUNT_ATTR_NODIRATIME
499 #define MOUNT_ATTR_NODIRATIME   0x00000080 /* Do not update directory access times */
500 #endif
501 
502 #ifndef MOUNT_ATTR_IDMAP
503 #define MOUNT_ATTR_IDMAP        0x00100000 /* Idmap mount to @userns_fd in struct mount_attr. */
504 #endif
505 
506 #ifndef MOUNT_ATTR_NOSYMFOLLOW
507 #define MOUNT_ATTR_NOSYMFOLLOW  0x00200000 /* Do not follow symlinks */
508 #endif
509 
510 #ifndef MOUNT_ATTR_SIZE_VER0
511 #define MOUNT_ATTR_SIZE_VER0    32 /* sizeof first published struct */
512 #endif
513 
514 #ifndef AT_RECURSIVE
515 #define AT_RECURSIVE 0x8000
516 #endif
517 
missing_mount_setattr(int dfd,const char * path,unsigned flags,struct mount_attr * attr,size_t size)518 static inline int missing_mount_setattr(
519                 int dfd,
520                 const char *path,
521                 unsigned flags,
522                 struct mount_attr *attr,
523                 size_t size) {
524 
525 #  if defined __NR_mount_setattr && __NR_mount_setattr >= 0
526         return syscall(__NR_mount_setattr, dfd, path, flags, attr, size);
527 #  else
528         errno = ENOSYS;
529         return -1;
530 #  endif
531 }
532 
533 #  define mount_setattr missing_mount_setattr
534 #endif
535 
536 /* ======================================================================= */
537 
538 #if !HAVE_OPEN_TREE
539 
540 #ifndef OPEN_TREE_CLONE
541 #define OPEN_TREE_CLONE 1
542 #endif
543 
544 #ifndef OPEN_TREE_CLOEXEC
545 #define OPEN_TREE_CLOEXEC O_CLOEXEC
546 #endif
547 
missing_open_tree(int dfd,const char * filename,unsigned flags)548 static inline int missing_open_tree(
549                 int dfd,
550                 const char *filename,
551                 unsigned flags) {
552 
553 #  if defined __NR_open_tree && __NR_open_tree >= 0
554         return syscall(__NR_open_tree, dfd, filename, flags);
555 #  else
556         errno = ENOSYS;
557         return -1;
558 #  endif
559 }
560 
561 #  define open_tree missing_open_tree
562 #endif
563 
564 /* ======================================================================= */
565 
566 #if !HAVE_MOVE_MOUNT
567 
568 #ifndef MOVE_MOUNT_F_EMPTY_PATH
569 #define MOVE_MOUNT_F_EMPTY_PATH 0x00000004 /* Empty from path permitted */
570 #endif
571 
572 #ifndef MOVE_MOUNT_T_EMPTY_PATH
573 #define MOVE_MOUNT_T_EMPTY_PATH 0x00000040 /* Empty to path permitted */
574 #endif
575 
missing_move_mount(int from_dfd,const char * from_pathname,int to_dfd,const char * to_pathname,unsigned flags)576 static inline int missing_move_mount(
577                 int from_dfd,
578                 const char *from_pathname,
579                 int to_dfd,
580                 const char *to_pathname,
581                 unsigned flags) {
582 
583 #  if defined __NR_move_mount && __NR_move_mount >= 0
584         return syscall(__NR_move_mount, from_dfd, from_pathname, to_dfd, to_pathname, flags);
585 #  else
586         errno = ENOSYS;
587         return -1;
588 #  endif
589 }
590 
591 #  define move_mount missing_move_mount
592 #endif
593 
594 /* ======================================================================= */
595 
596 #if !HAVE_GETDENTS64
597 
missing_getdents64(int fd,void * buffer,size_t length)598 static inline ssize_t missing_getdents64(int fd, void *buffer, size_t length) {
599 #  if defined __NR_getdents64 && __NR_getdents64 >= 0
600         return syscall(__NR_getdents64, fd, buffer, length);
601 #  else
602         errno = ENOSYS;
603         return -1;
604 #  endif
605 }
606 
607 #  define getdents64 missing_getdents64
608 #endif
609