1SYSCALL(2) Linux Programmer's Manual SYSCALL(2)
2
3
4
6 syscall - indirect system call
7
9 #include <sys/syscall.h> /* Definition of SYS_* constants */
10 #include <unistd.h>
11
12 long syscall(long number, ...);
13
14 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
15
16 syscall():
17 Since glibc 2.19:
18 _DEFAULT_SOURCE
19 Before glibc 2.19:
20 _BSD_SOURCE || _SVID_SOURCE
21
23 syscall() is a small library function that invokes the system call
24 whose assembly language interface has the specified number with the
25 specified arguments. Employing syscall() is useful, for example, when
26 invoking a system call that has no wrapper function in the C library.
27
28 syscall() saves CPU registers before making the system call, restores
29 the registers upon return from the system call, and stores any error
30 returned by the system call in errno(3).
31
32 Symbolic constants for system call numbers can be found in the header
33 file <sys/syscall.h>.
34
36 The return value is defined by the system call being invoked. In gen‐
37 eral, a 0 return value indicates success. A -1 return value indicates
38 an error, and an error number is stored in errno.
39
41 syscall() first appeared in 4BSD.
42
43 Architecture-specific requirements
44 Each architecture ABI has its own requirements on how system call argu‐
45 ments are passed to the kernel. For system calls that have a glibc
46 wrapper (e.g., most system calls), glibc handles the details of copying
47 arguments to the right registers in a manner suitable for the architec‐
48 ture. However, when using syscall() to make a system call, the caller
49 might need to handle architecture-dependent details; this requirement
50 is most commonly encountered on certain 32-bit architectures.
51
52 For example, on the ARM architecture Embedded ABI (EABI), a 64-bit
53 value (e.g., long long) must be aligned to an even register pair.
54 Thus, using syscall() instead of the wrapper provided by glibc, the
55 readahead(2) system call would be invoked as follows on the ARM archi‐
56 tecture with the EABI in little endian mode:
57
58 syscall(SYS_readahead, fd, 0,
59 (unsigned int) (offset & 0xFFFFFFFF),
60 (unsigned int) (offset >> 32),
61 count);
62
63 Since the offset argument is 64 bits, and the first argument (fd) is
64 passed in r0, the caller must manually split and align the 64-bit value
65 so that it is passed in the r2/r3 register pair. That means inserting
66 a dummy value into r1 (the second argument of 0). Care also must be
67 taken so that the split follows endian conventions (according to the C
68 ABI for the platform).
69
70 Similar issues can occur on MIPS with the O32 ABI, on PowerPC and
71 parisc with the 32-bit ABI, and on Xtensa.
72
73 Note that while the parisc C ABI also uses aligned register pairs, it
74 uses a shim layer to hide the issue from user space.
75
76 The affected system calls are fadvise64_64(2), ftruncate64(2),
77 posix_fadvise(2), pread64(2), pwrite64(2), readahead(2),
78 sync_file_range(2), and truncate64(2).
79
80 This does not affect syscalls that manually split and assemble 64-bit
81 values such as _llseek(2), preadv(2), preadv2(2), pwritev(2), and
82 pwritev2(2). Welcome to the wonderful world of historical baggage.
83
84 Architecture calling conventions
85 Every architecture has its own way of invoking and passing arguments to
86 the kernel. The details for various architectures are listed in the
87 two tables below.
88
89 The first table lists the instruction used to transition to kernel mode
90 (which might not be the fastest or best way to transition to the ker‐
91 nel, so you might have to refer to vdso(7)), the register used to indi‐
92 cate the system call number, the register(s) used to return the system
93 call result, and the register used to signal an error.
94
95 Arch/ABI Instruction System Ret Ret Error Notes
96 call # val val2
97 ───────────────────────────────────────────────────────────────────
98 alpha callsys v0 v0 a4 a3 1, 6
99 arc trap0 r8 r0 - -
100 arm/OABI swi NR - r0 - - 2
101 arm/EABI swi 0x0 r7 r0 r1 -
102 arm64 svc #0 w8 x0 x1 -
103 blackfin excpt 0x0 P0 R0 - -
104 i386 int $0x80 eax eax edx -
105 ia64 break 0x100000 r15 r8 r9 r10 1, 6
106 m68k trap #0 d0 d0 - -
107 microblaze brki r14,8 r12 r3 - -
108 mips syscall v0 v0 v1 a3 1, 6
109 nios2 trap r2 r2 - r7
110 parisc ble 0x100(%sr2, %r0) r20 r28 - -
111 powerpc sc r0 r3 - r0 1
112 powerpc64 sc r0 r3 - cr0.SO 1
113 riscv ecall a7 a0 a1 -
114 s390 svc 0 r1 r2 r3 - 3
115 s390x svc 0 r1 r2 r3 - 3
116 superh trapa #31 r3 r0 r1 - 4, 6
117 sparc/32 t 0x10 g1 o0 o1 psr/csr 1, 6
118 sparc/64 t 0x6d g1 o0 o1 psr/csr 1, 6
119 tile swint1 R10 R00 - R01 1
120 x86-64 syscall rax rax rdx - 5
121 x32 syscall rax rax rdx - 5
122 xtensa syscall a2 a2 - -
123
124 Notes:
125
126 [1] On a few architectures, a register is used as a boolean (0 indicat‐
127 ing no error, and -1 indicating an error) to signal that the system
128 call failed. The actual error value is still contained in the re‐
129 turn register. On sparc, the carry bit (csr) in the processor sta‐
130 tus register (psr) is used instead of a full register. On pow‐
131 erpc64, the summary overflow bit (SO) in field 0 of the condition
132 register (cr0) is used.
133
134 [2] NR is the system call number.
135
136 [3] For s390 and s390x, NR (the system call number) may be passed di‐
137 rectly with svc NR if it is less than 256.
138
139 [4] On SuperH additional trap numbers are supported for historic rea‐
140 sons, but trapa#31 is the recommended "unified" ABI.
141
142 [5] The x32 ABI shares syscall table with x86-64 ABI, but there are
143 some nuances:
144
145 • In order to indicate that a system call is called under the x32
146 ABI, an additional bit, __X32_SYSCALL_BIT, is bitwise-ORed with
147 the system call number. The ABI used by a process affects some
148 process behaviors, including signal handling or system call
149 restarting.
150
151 • Since x32 has different sizes for long and pointer types, lay‐
152 outs of some (but not all; struct timeval or struct rlimit are
153 64-bit, for example) structures are different. In order to han‐
154 dle this, additional system calls are added to the system call
155 table, starting from number 512 (without the __X32_SYSCALL_BIT).
156 For example, __NR_readv is defined as 19 for the x86-64 ABI and
157 as __X32_SYSCALL_BIT | 515 for the x32 ABI. Most of these addi‐
158 tional system calls are actually identical to the system calls
159 used for providing i386 compat. There are some notable excep‐
160 tions, however, such as preadv2(2), which uses struct iovec en‐
161 tities with 4-byte pointers and sizes ("compat_iovec" in kernel
162 terms), but passes an 8-byte pos argument in a single register
163 and not two, as is done in every other ABI.
164
165 [6] Some architectures (namely, Alpha, IA-64, MIPS, SuperH, sparc/32,
166 and sparc/64) use an additional register ("Retval2" in the above
167 table) to pass back a second return value from the pipe(2) system
168 call; Alpha uses this technique in the architecture-specific getx‐
169 pid(2), getxuid(2), and getxgid(2) system calls as well. Other ar‐
170 chitectures do not use the second return value register in the sys‐
171 tem call interface, even if it is defined in the System V ABI.
172
173 The second table shows the registers used to pass the system call argu‐
174 ments.
175
176 Arch/ABI arg1 arg2 arg3 arg4 arg5 arg6 arg7 Notes
177 ──────────────────────────────────────────────────────────────
178 alpha a0 a1 a2 a3 a4 a5 -
179 arc r0 r1 r2 r3 r4 r5 -
180 arm/OABI r0 r1 r2 r3 r4 r5 r6
181 arm/EABI r0 r1 r2 r3 r4 r5 r6
182 arm64 x0 x1 x2 x3 x4 x5 -
183 blackfin R0 R1 R2 R3 R4 R5 -
184 i386 ebx ecx edx esi edi ebp -
185 ia64 out0 out1 out2 out3 out4 out5 -
186 m68k d1 d2 d3 d4 d5 a0 -
187 microblaze r5 r6 r7 r8 r9 r10 -
188 mips/o32 a0 a1 a2 a3 - - - 1
189 mips/n32,64 a0 a1 a2 a3 a4 a5 -
190 nios2 r4 r5 r6 r7 r8 r9 -
191 parisc r26 r25 r24 r23 r22 r21 -
192 powerpc r3 r4 r5 r6 r7 r8 r9
193 powerpc64 r3 r4 r5 r6 r7 r8 -
194 riscv a0 a1 a2 a3 a4 a5 -
195 s390 r2 r3 r4 r5 r6 r7 -
196 s390x r2 r3 r4 r5 r6 r7 -
197 superh r4 r5 r6 r7 r0 r1 r2
198
199 sparc/32 o0 o1 o2 o3 o4 o5 -
200 sparc/64 o0 o1 o2 o3 o4 o5 -
201 tile R00 R01 R02 R03 R04 R05 -
202 x86-64 rdi rsi rdx r10 r8 r9 -
203 x32 rdi rsi rdx r10 r8 r9 -
204 xtensa a6 a3 a4 a5 a8 a9 -
205
206 Notes:
207
208 [1] The mips/o32 system call convention passes arguments 5 through 8 on
209 the user stack.
210
211 Note that these tables don't cover the entire calling convention—some
212 architectures may indiscriminately clobber other registers not listed
213 here.
214
216 #define _GNU_SOURCE
217 #include <unistd.h>
218 #include <sys/syscall.h>
219 #include <sys/types.h>
220 #include <signal.h>
221
222 int
223 main(int argc, char *argv[])
224 {
225 pid_t tid;
226
227 tid = syscall(SYS_gettid);
228 syscall(SYS_tgkill, getpid(), tid, SIGHUP);
229 }
230
232 _syscall(2), intro(2), syscalls(2), errno(3), vdso(7)
233
235 This page is part of release 5.12 of the Linux man-pages project. A
236 description of the project, information about reporting bugs, and the
237 latest version of this page, can be found at
238 https://www.kernel.org/doc/man-pages/.
239
240
241
242Linux 2021-03-22 SYSCALL(2)