| blob | abdf1c7027072a08a04d258211ca613a9bbc01a3 |
1 .\" Copyright (c) 1980, 1991, 1993
2 .\" The Regents of the University of California. All rights reserved.
3 .\"
4 .\" %%%LICENSE_START(BSD_4_CLAUSE_UCB)
5 .\" Redistribution and use in source and binary forms, with or without
6 .\" modification, are permitted provided that the following conditions
7 .\" are met:
8 .\" 1. Redistributions of source code must retain the above copyright
9 .\" notice, this list of conditions and the following disclaimer.
10 .\" 2. Redistributions in binary form must reproduce the above copyright
11 .\" notice, this list of conditions and the following disclaimer in the
12 .\" documentation and/or other materials provided with the distribution.
13 .\" 3. All advertising materials mentioning features or use of this software
14 .\" must display the following acknowledgement:
15 .\" This product includes software developed by the University of
16 .\" California, Berkeley and its contributors.
17 .\" 4. Neither the name of the University nor the names of its contributors
18 .\" may be used to endorse or promote products derived from this software
19 .\" without specific prior written permission.
20 .\"
21 .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 .\" SUCH DAMAGE.
32 .\" %%%LICENSE_END
33 .\"
34 .\" @(#)syscall.2 8.1 (Berkeley) 6/16/93
35 .\"
36 .\"
37 .\" 2002-03-20 Christoph Hellwig <hch@infradead.org>
38 .\" - adopted for Linux
39 .\" 2015-01-17, Kees Cook <keescook@chromium.org>
40 .\" Added mips and arm64.
41 .\"
42 .TH SYSCALL 2 2021-03-22 "Linux" "Linux Programmer's Manual"
43 .SH NAME
44 syscall \- indirect system call
45 .SH SYNOPSIS
46 .nf
47 .BR "#include <sys/syscall.h>" " /* Definition of " SYS_* " constants */"
48 .B #include <unistd.h>
49 .PP
50 .BI "long syscall(long " number ", ...);"
51 .fi
52 .PP
53 .RS -4
54 Feature Test Macro Requirements for glibc (see
55 .BR feature_test_macros (7)):
56 .RE
57 .PP
58 .BR syscall ():
59 .nf
60 Since glibc 2.19:
61 _DEFAULT_SOURCE
62 Before glibc 2.19:
63 _BSD_SOURCE || _SVID_SOURCE
64 .fi
65 .SH DESCRIPTION
66 .BR syscall ()
67 is a small library function that invokes
68 the system call whose assembly language
69 interface has the specified
70 .I number
71 with the specified arguments.
72 Employing
73 .BR syscall ()
74 is useful, for example,
75 when invoking a system call that has no wrapper function in the C library.
76 .PP
77 .BR syscall ()
78 saves CPU registers before making the system call,
79 restores the registers upon return from the system call,
80 and stores any error returned by the system call in
81 .BR errno (3).
82 .PP
83 Symbolic constants for system call numbers can be found in the header file
84 .IR <sys/syscall.h> .
85 .SH RETURN VALUE
86 The return value is defined by the system call being invoked.
87 In general, a 0 return value indicates success.
88 A \-1 return value indicates an error,
89 and an error number is stored in
90 .IR errno .
91 .SH NOTES
92 .BR syscall ()
93 first appeared in
94 4BSD.
95 .SS Architecture-specific requirements
96 Each architecture ABI has its own requirements on how
97 system call arguments are passed to the kernel.
98 For system calls that have a glibc wrapper (e.g., most system calls),
99 glibc handles the details of copying arguments to the right registers
100 in a manner suitable for the architecture.
101 However, when using
102 .BR syscall ()
103 to make a system call,
104 the caller might need to handle architecture-dependent details;
105 this requirement is most commonly encountered on certain 32-bit architectures.
106 .PP
107 For example, on the ARM architecture Embedded ABI (EABI), a
108 64-bit value (e.g.,
109 .IR "long long" )
110 must be aligned to an even register pair.
111 Thus, using
112 .BR syscall ()
113 instead of the wrapper provided by glibc,
114 the
115 .BR readahead (2)
116 system call would be invoked as follows on the ARM architecture with the EABI
117 in little endian mode:
118 .PP
119 .in +4n
120 .EX
121 syscall(SYS_readahead, fd, 0,
122 (unsigned int) (offset & 0xFFFFFFFF),
123 (unsigned int) (offset >> 32),
124 count);
125 .EE
126 .in
127 .PP
128 Since the offset argument is 64 bits, and the first argument
129 .RI ( fd )
130 is passed in
131 .IR r0 ,
132 the caller must manually split and align the 64-bit value
133 so that it is passed in the
134 .IR r2 / r3
135 register pair.
136 That means inserting a dummy value into
137 .I r1
138 (the second argument of 0).
139 Care also must be taken so that the split follows endian conventions
140 (according to the C ABI for the platform).
141 .PP
142 Similar issues can occur on MIPS with the O32 ABI,
143 on PowerPC and parisc with the 32-bit ABI, and on Xtensa.
144 .\" Mike Frysinger: this issue ends up forcing MIPS
145 .\" O32 to take 7 arguments to syscall()
146 .PP
147 .\" See arch/parisc/kernel/sys_parisc.c.
148 Note that while the parisc C ABI also uses aligned register pairs,
149 it uses a shim layer to hide the issue from user space.
150 .PP
151 The affected system calls are
152 .BR fadvise64_64 (2),
153 .BR ftruncate64 (2),
154 .BR posix_fadvise (2),
155 .BR pread64 (2),
156 .BR pwrite64 (2),
157 .BR readahead (2),
158 .BR sync_file_range (2),
159 and
160 .BR truncate64 (2).
161 .PP
162 .\" You need to look up the syscalls directly in the kernel source to see if
163 .\" they should be in this list. For example, look at fs/read_write.c and
164 .\" the function signatures that do:
165 .\" ..., unsigned long, pos_l, unsigned long, pos_h, ...
166 .\" If they use off_t, then they most likely do not belong in this list.
167 This does not affect syscalls that manually split and assemble 64-bit values
168 such as
169 .BR _llseek (2),
170 .BR preadv (2),
171 .BR preadv2 (2),
172 .BR pwritev (2),
173 and
174 .BR pwritev2 (2).
175 Welcome to the wonderful world of historical baggage.
176 .SS Architecture calling conventions
177 Every architecture has its own way of invoking and passing arguments to the
178 kernel.
179 The details for various architectures are listed in the two tables below.
180 .PP
181 The first table lists the instruction used to transition to kernel mode
182 (which might not be the fastest or best way to transition to the kernel,
183 so you might have to refer to
184 .BR vdso (7)),
185 the register used to indicate the system call number,
186 the register(s) used to return the system call result,
187 and the register used to signal an error.
188 .if t \{\
189 .ft CW
190 \}
191 .TS
192 l2 l2 l2 l2 l1 l2 l.
193 Arch/ABI Instruction System Ret Ret Error Notes
194 call # val val2
195 _
196 alpha callsys v0 v0 a4 a3 1, 6
197 arc trap0 r8 r0 - -
198 arm/OABI swi NR - r0 - - 2
199 arm/EABI swi 0x0 r7 r0 r1 -
200 arm64 svc #0 w8 x0 x1 -
201 blackfin excpt 0x0 P0 R0 - -
202 i386 int $0x80 eax eax edx -
203 ia64 break 0x100000 r15 r8 r9 r10 1, 6
204 m68k trap #0 d0 d0 - -
205 microblaze brki r14,8 r12 r3 - -
206 mips syscall v0 v0 v1 a3 1, 6
207 nios2 trap r2 r2 - r7
208 parisc ble 0x100(%sr2, %r0) r20 r28 - -
209 powerpc sc r0 r3 - r0 1
210 powerpc64 sc r0 r3 - cr0.SO 1
211 riscv ecall a7 a0 a1 -
212 s390 svc 0 r1 r2 r3 - 3
213 s390x svc 0 r1 r2 r3 - 3
214 superh trapa #31 r3 r0 r1 - 4, 6
215 sparc/32 t 0x10 g1 o0 o1 psr/csr 1, 6
216 sparc/64 t 0x6d g1 o0 o1 psr/csr 1, 6
217 tile swint1 R10 R00 - R01 1
218 x86-64 syscall rax rax rdx - 5
219 x32 syscall rax rax rdx - 5
220 xtensa syscall a2 a2 - -
221 .TE
222 .PP
223 Notes:
224 .IP [1] 4
225 On a few architectures,
226 a register is used as a boolean
227 (0 indicating no error, and \-1 indicating an error) to signal that the
228 system call failed.
229 The actual error value is still contained in the return register.
230 On sparc, the carry bit
231 .RI ( csr )
232 in the processor status register
233 .RI ( psr )
234 is used instead of a full register.
235 On powerpc64, the summary overflow bit
236 .RI ( SO )
237 in field 0 of the condition register
238 .RI ( cr0 )
239 is used.
240 .IP [2]
241 .I NR
242 is the system call number.
243 .IP [3]
244 For s390 and s390x,
245 .I NR
246 (the system call number) may be passed directly with
247 .I "svc\ NR"
248 if it is less than 256.
249 .IP [4]
250 On SuperH additional trap numbers are supported for historic reasons, but
251 .BR trapa #31
252 is the recommended "unified" ABI.
253 .IP [5]
254 The x32 ABI shares syscall table with x86-64 ABI, but there are some
255 nuances:
256 .RS
257 .IP \(bu 3
258 In order to indicate that a system call is called under the x32 ABI,
259 an additional bit,
260 .BR __X32_SYSCALL_BIT ,
261 is bitwise-ORed with the system call number.
262 The ABI used by a process affects some process behaviors,
263 including signal handling or system call restarting.
264 .IP \(bu
265 Since x32 has different sizes for
266 .I long
267 and pointer types, layouts of some (but not all;
268 .I struct timeval
269 or
270 .I struct rlimit
271 are 64-bit, for example) structures are different.
272 In order to handle this,
273 additional system calls are added to the system call table,
274 starting from number 512
275 (without the
276 .BR __X32_SYSCALL_BIT ).
277 For example,
278 .B __NR_readv
279 is defined as 19 for the x86-64 ABI and as
280 .IR __X32_SYSCALL_BIT " | " \fB515\fP
281 for the x32 ABI.
282 Most of these additional system calls are actually identical
283 to the system calls used for providing i386 compat.
284 There are some notable exceptions, however, such as
285 .BR preadv2 (2),
286 which uses
287 .I struct iovec
288 entities with 4-byte pointers and sizes ("compat_iovec" in kernel terms),
289 but passes an 8-byte
290 .I pos
291 argument in a single register and not two, as is done in every other ABI.
292 .RE
293 .IP [6]
294 Some architectures
295 (namely, Alpha, IA-64, MIPS, SuperH, sparc/32, and sparc/64)
296 use an additional register ("Retval2" in the above table)
297 to pass back a second return value from the
298 .BR pipe (2)
299 system call;
300 Alpha uses this technique in the architecture-specific
301 .BR getxpid (2),
302 .BR getxuid (2),
303 and
304 .BR getxgid (2)
305 system calls as well.
306 Other architectures do not use the second return value register
307 in the system call interface, even if it is defined in the System V ABI.
308 .if t \{\
309 .in
310 .ft P
311 \}
312 .PP
313 The second table shows the registers used to pass the system call arguments.
314 .if t \{\
315 .ft CW
316 \}
317 .TS
318 l l2 l2 l2 l2 l2 l2 l2 l.
319 Arch/ABI arg1 arg2 arg3 arg4 arg5 arg6 arg7 Notes
320 _
321 alpha a0 a1 a2 a3 a4 a5 -
322 arc r0 r1 r2 r3 r4 r5 -
323 arm/OABI r0 r1 r2 r3 r4 r5 r6
324 arm/EABI r0 r1 r2 r3 r4 r5 r6
325 arm64 x0 x1 x2 x3 x4 x5 -
326 blackfin R0 R1 R2 R3 R4 R5 -
327 i386 ebx ecx edx esi edi ebp -
328 ia64 out0 out1 out2 out3 out4 out5 -
329 m68k d1 d2 d3 d4 d5 a0 -
330 microblaze r5 r6 r7 r8 r9 r10 -
331 mips/o32 a0 a1 a2 a3 - - - 1
332 mips/n32,64 a0 a1 a2 a3 a4 a5 -
333 nios2 r4 r5 r6 r7 r8 r9 -
334 parisc r26 r25 r24 r23 r22 r21 -
335 powerpc r3 r4 r5 r6 r7 r8 r9
336 powerpc64 r3 r4 r5 r6 r7 r8 -
337 riscv a0 a1 a2 a3 a4 a5 -
338 s390 r2 r3 r4 r5 r6 r7 -
339 s390x r2 r3 r4 r5 r6 r7 -
340 superh r4 r5 r6 r7 r0 r1 r2
341 sparc/32 o0 o1 o2 o3 o4 o5 -
342 sparc/64 o0 o1 o2 o3 o4 o5 -
343 tile R00 R01 R02 R03 R04 R05 -
344 x86-64 rdi rsi rdx r10 r8 r9 -
345 x32 rdi rsi rdx r10 r8 r9 -
346 xtensa a6 a3 a4 a5 a8 a9 -
347 .TE
348 .PP
349 Notes:
350 .IP [1] 4
351 The mips/o32 system call convention passes
352 arguments 5 through 8 on the user stack.
353 .if t \{\
354 .in
355 .ft P
356 \}
357 .PP
358 Note that these tables don't cover the entire calling convention\(emsome
359 architectures may indiscriminately clobber other registers not listed here.
360 .SH EXAMPLES
361 .EX
362 #define _GNU_SOURCE
363 #include <unistd.h>
364 #include <sys/syscall.h>
365 #include <sys/types.h>
366 #include <signal.h>
368 int
369 main(int argc, char *argv[])
370 {
371 pid_t tid;
373 tid = syscall(SYS_gettid);
374 syscall(SYS_tgkill, getpid(), tid, SIGHUP);
375 }
376 .EE
377 .SH SEE ALSO
378 .BR _syscall (2),
379 .BR intro (2),
380 .BR syscalls (2),
381 .BR errno (3),
382 .BR vdso (7)