1/* Set flags signalling availability of kernel features based on given
2 kernel version number.
3 Copyright (C) 1999-2023 Free Software Foundation, Inc.
4 This file is part of the GNU C Library.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <https://www.gnu.org/licenses/>. */
19
20/* This file must not contain any C code. At least it must be protected
21 to allow using the file also in assembler files. */
22
23#ifndef _LINUX_KERNEL_FEATURES_H
24#define _LINUX_KERNEL_FEATURES_H 1
25
26#include <bits/wordsize.h>
27
28#ifndef __LINUX_KERNEL_VERSION
29/* We assume the worst; all kernels should be supported. */
30# define __LINUX_KERNEL_VERSION 0
31#endif
32
33/* We assume for __LINUX_KERNEL_VERSION the same encoding used in
34 linux/version.h. I.e., the major, minor, and subminor all get a
35 byte with the major number being in the highest byte. This means
36 we can do numeric comparisons.
37
38 In the following we will define certain symbols depending on
39 whether the describes kernel feature is available in the kernel
40 version given by __LINUX_KERNEL_VERSION. We are not always exactly
41 recording the correct versions in which the features were
42 introduced. If somebody cares these values can afterwards be
43 corrected. */
44
45/* The statfs64 syscalls are available in 2.5.74 (but not for alpha). */
46#define __ASSUME_STATFS64 1
47
48/* pselect/ppoll were introduced just after 2.6.16-rc1. On x86_64 and
49 SH this appeared first in 2.6.19-rc1, on ia64 in 2.6.22-rc1. */
50#define __ASSUME_PSELECT 1
51
52/* Support for inter-process robust mutexes was added in 2.6.17 (but
53 some architectures lack futex_atomic_cmpxchg_inatomic in some
54 configurations). */
55#define __ASSUME_SET_ROBUST_LIST 1
56
57/* Support for various CLOEXEC and NONBLOCK flags was added in
58 2.6.27. */
59#define __ASSUME_IN_NONBLOCK 1
60
61/* Support for preadv and pwritev was added in 2.6.30. */
62#define __ASSUME_PREADV 1
63#define __ASSUME_PWRITEV 1
64
65/* Support for sendmmsg functionality was added in 3.0. */
66#define __ASSUME_SENDMMSG 1
67
68/* On most architectures, most socket syscalls are supported for all
69 supported kernel versions, but on some socketcall architectures
70 separate syscalls were only added later. */
71#define __ASSUME_SENDMSG_SYSCALL 1
72#define __ASSUME_RECVMSG_SYSCALL 1
73#define __ASSUME_ACCEPT_SYSCALL 1
74#define __ASSUME_CONNECT_SYSCALL 1
75#define __ASSUME_RECVFROM_SYSCALL 1
76#define __ASSUME_SENDTO_SYSCALL 1
77#define __ASSUME_ACCEPT4_SYSCALL 1
78#define __ASSUME_RECVMMSG_SYSCALL 1
79#define __ASSUME_SENDMMSG_SYSCALL 1
80#define __ASSUME_GETSOCKOPT_SYSCALL 1
81#define __ASSUME_SETSOCKOPT_SYSCALL 1
82#define __ASSUME_BIND_SYSCALL 1
83#define __ASSUME_SOCKET_SYSCALL 1
84#define __ASSUME_SOCKETPAIR_SYSCALL 1
85#define __ASSUME_LISTEN_SYSCALL 1
86#define __ASSUME_SHUTDOWN_SYSCALL 1
87
88/* Support for SysV IPC through wired syscalls. All supported architectures
89 either support ipc syscall and/or all the ipc correspondent syscalls. */
90#define __ASSUME_DIRECT_SYSVIPC_SYSCALLS 1
91/* The generic default __IPC_64 value is 0x0, however some architectures
92 require a different value of 0x100. */
93#define __ASSUME_SYSVIPC_DEFAULT_IPC_64 1
94
95/* All supported architectures reserve a 32-bit for MODE field in sysvipc
96 ipc_perm. However, some kernel ABI interfaces still expect a 16-bit
97 field. This is only an issue if arch-defined IPC_PERM padding is on a
98 wrong position regarding endianness. In this case, the IPC control
99 routines (msgctl, semctl, and semtctl) requires to shift the value to
100 correct place.
101 The ABIs that requires it define __ASSUME_SYSVIPC_BROKEN_MODE_T. */
102
103/* Support for p{read,write}v2 was added in 4.6. However Linux default
104 implementation does not assume the __ASSUME_* and instead use a fallback
105 implementation based on p{read,write}v and returning an error for
106 non supported flags. */
107
108/* Support for the renameat2 system call was added in kernel 3.15. */
109#if __LINUX_KERNEL_VERSION >= 0x030F00
110# define __ASSUME_RENAMEAT2
111#endif
112
113/* Support for the execveat syscall was added in 3.19. */
114#if __LINUX_KERNEL_VERSION >= 0x031300
115# define __ASSUME_EXECVEAT 1
116#endif
117
118#if __LINUX_KERNEL_VERSION >= 0x040400
119# define __ASSUME_MLOCK2 1
120#endif
121
122/* Support for statx was added in kernel 4.11. */
123#if __LINUX_KERNEL_VERSION >= 0x040B00
124# define __ASSUME_STATX 1
125#endif
126
127/* Support for clone call used on fork. The signature varies across the
128 architectures with current 4 different variants:
129
130 1. long int clone (unsigned long flags, unsigned long newsp,
131 int *parent_tidptr, unsigned long tls,
132 int *child_tidptr)
133
134 2. long int clone (unsigned long newsp, unsigned long clone_flags,
135 int *parent_tidptr, int * child_tidptr,
136 unsigned long tls)
137
138 3. long int clone (unsigned long flags, unsigned long newsp,
139 int stack_size, int *parent_tidptr,
140 int *child_tidptr, unsigned long tls)
141
142 4. long int clone (unsigned long flags, unsigned long newsp,
143 int *parent_tidptr, int *child_tidptr,
144 unsigned long tls)
145
146 The fourth variant is intended to be used as the default for newer ports,
147 Also IA64 uses the third variant but with __NR_clone2 instead of
148 __NR_clone.
149
150 The macros names to define the variant used for the architecture is
151 similar to kernel:
152
153 - __ASSUME_CLONE_BACKWARDS: for variant 1.
154 - __ASSUME_CLONE_BACKWARDS2: for variant 2 (s390).
155 - __ASSUME_CLONE_BACKWARDS3: for variant 3 (microblaze).
156 - __ASSUME_CLONE_DEFAULT: for variant 4.
157 - __ASSUME_CLONE2: for clone2 with variant 3 (ia64).
158 */
159
160#define __ASSUME_CLONE_DEFAULT 1
161
162/* Support for 64-bit time_t in the system call interface. When this
163 flag is set, the kernel provides a version of each of these system
164 calls that accepts 64-bit time_t:
165
166 clock_adjtime(64)
167 clock_gettime(64)
168 clock_settime(64)
169 clock_getres(_time64)
170 clock_nanosleep(_time64)
171 futex(_time64)
172 mq_timedreceive(_time64)
173 mq_timedsend(_time64)
174 ppoll(_time64)
175 pselect6(_time64)
176 rt_sigtimedwait(_time64)
177 sched_rr_get_interval(_time64)
178 timer_gettime(64)
179 timer_settime(64)
180 timerfd_gettime(64)
181 timerfd_settime(64)
182 utimensat(_time64)
183
184 On architectures where time_t has historically been 64 bits,
185 only the 64-bit version of each system call exists, and there
186 are no suffixes on the __NR_ constants.
187
188 On architectures where time_t has historically been 32 bits,
189 both 32-bit and 64-bit versions of each system call may exist,
190 depending on the kernel version. When the 64-bit version exists,
191 there is a '64' or '_time64' suffix on the name of its __NR_
192 constant, as shown above.
193
194 This flag is always set for Linux 5.1 and later. Prior to that
195 version, it is set only for some CPU architectures and ABIs:
196
197 - __WORDSIZE == 64 - all supported architectures where pointers
198 are 64 bits also have always had 64-bit time_t.
199
200 - __WORDSIZE == 32 && __SYSCALL_WORDSIZE == 64 - this describes
201 only one supported configuration, x86's 'x32' subarchitecture,
202 where pointers are 32 bits but time_t has always been 64 bits.
203
204 __ASSUME_TIME64_SYSCALLS being set does not mean __TIMESIZE is 64,
205 and __TIMESIZE equal to 64 does not mean __ASSUME_TIME64_SYSCALLS
206 is set. All four cases are possible. */
207
208#if __LINUX_KERNEL_VERSION >= 0x050100 \
209 || __WORDSIZE == 64 \
210 || (defined __SYSCALL_WORDSIZE && __SYSCALL_WORDSIZE == 64)
211# define __ASSUME_TIME64_SYSCALLS 1
212#endif
213
214/* Linux waitid prior kernel 5.4 does not support waiting for the current
215 process group. */
216#if __LINUX_KERNEL_VERSION >= 0x050400
217# define __ASSUME_WAITID_PID0_P_PGID
218#endif
219
220/* The faccessat2 system call was introduced across all architectures
221 in Linux 5.8. */
222#if __LINUX_KERNEL_VERSION >= 0x050800
223# define __ASSUME_FACCESSAT2 1
224#else
225# define __ASSUME_FACCESSAT2 0
226#endif
227
228/* The close_range system call was introduced across all architectures
229 in Linux 5.9. */
230#if __LINUX_KERNEL_VERSION >= 0x050900
231# define __ASSUME_CLOSE_RANGE 1
232#else
233# define __ASSUME_CLOSE_RANGE 0
234#endif
235
236/* The FUTEX_LOCK_PI2 operation was introduced across all architectures in Linux
237 5.14. */
238#if __LINUX_KERNEL_VERSION >= 0x050e00
239# define __ASSUME_FUTEX_LOCK_PI2 1
240#else
241# define __ASSUME_FUTEX_LOCK_PI2 0
242#endif
243
244#endif /* kernel-features.h */
245