1SYSCALL(2) Linux Programmer's Manual SYSCALL(2)
2
6 syscall - indirect system call
7
9 #include <unistd.h>
10 #include <sys/syscall.h> /* For SYS_xxx definitions */
11 long syscall(long number, ...);
13 Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
15 syscall():
16 Since glibc 2.19:
17 _DEFAULT_SOURCE
18 Before glibc 2.19:
19 _BSD_SOURCE || _SVID_SOURCE
20
22 syscall() is a small library function that invokes the system call
23 whose assembly language interface has the specified number with the
24 specified arguments. Employing syscall() is useful, for example, when
25 invoking a system call that has no wrapper function in the C library.
26 syscall() saves CPU registers before making the system call, restores
28 the registers upon return from the system call, and stores any error
29 returned by the system call in errno(3).
30 Symbolic constants for system call numbers can be found in the header
32 file <sys/syscall.h>.
33
35 The return value is defined by the system call being invoked. In gen‐
36 eral, a 0 return value indicates success. A -1 return value indicates
37 an error, and an error number is stored in errno.
38
40 syscall() first appeared in 4BSD.
41 Architecture-specific requirements
43 Each architecture ABI has its own requirements on how system call argu‐
44 ments are passed to the kernel. For system calls that have a glibc
45 wrapper (e.g., most system calls), glibc handles the details of copying
46 arguments to the right registers in a manner suitable for the architec‐
47 ture. However, when using syscall() to make a system call, the caller
48 might need to handle architecture-dependent details; this requirement
49 is most commonly encountered on certain 32-bit architectures.
50 For example, on the ARM architecture Embedded ABI (EABI), a 64-bit
52 value (e.g., long long) must be aligned to an even register pair.
53 Thus, using syscall() instead of the wrapper provided by glibc, the
54 readahead(2) system call would be invoked as follows on the ARM archi‐
55 tecture with the EABI in little endian mode:
56 syscall(SYS_readahead, fd, 0,
58 (unsigned int) (offset & 0xFFFFFFFF),
59 (unsigned int) (offset >> 32),
60 count);
61 Since the offset argument is 64 bits, and the first argument (fd) is
63 passed in r0, the caller must manually split and align the 64-bit value
64 so that it is passed in the r2/r3 register pair. That means inserting
65 a dummy value into r1 (the second argument of 0). Care also must be
66 taken so that the split follows endian conventions (according to the C
67 ABI for the platform).
68 Similar issues can occur on MIPS with the O32 ABI, on PowerPC and
70 parisc with the 32-bit ABI, and on Xtensa.
71 Note that while the parisc C ABI also uses aligned register pairs, it
73 uses a shim layer to hide the issue from user space.
74 The affected system calls are fadvise64_64(2), ftruncate64(2),
76 posix_fadvise(2), pread64(2), pwrite64(2), readahead(2),
77 sync_file_range(2), and truncate64(2).
78 This does not affect syscalls that manually split and assemble 64-bit
80 values such as _llseek(2), preadv(2), preadv2(2), pwritev(2), and
81 pwritev2(2). Welcome to the wonderful world of historical baggage.
82 Architecture calling conventions
84 Every architecture has its own way of invoking and passing arguments to
85 the kernel. The details for various architectures are listed in the
86 two tables below.
87 The first table lists the instruction used to transition to kernel mode
89 (which might not be the fastest or best way to transition to the ker‐
90 nel, so you might have to refer to vdso(7)), the register used to indi‐
91 cate the system call number, the register(s) used to return the system
92 call result, and the register used to signal an error.
93 Arch/ABI Instruction System Ret Ret Error Notes
95 call # val val2
96 ───────────────────────────────────────────────────────────────────
97 alpha callsys v0 v0 a4 a3 1, 6
98 arc trap0 r8 r0 - -
99 arm/OABI swi NR - r0 - - 2
100 arm/EABI swi 0x0 r7 r0 r1 -
101 arm64 svc #0 w8 x0 x1 -
102 blackfin excpt 0x0 P0 R0 - -
103 i386 int $0x80 eax eax edx -
104 ia64 break 0x100000 r15 r8 r9 r10 1, 6
105 m68k trap #0 d0 d0 - -
106 microblaze brki r14,8 r12 r3 - -
107 mips syscall v0 v0 v1 a3 1, 6
108 nios2 trap r2 r2 - r7
109 parisc ble 0x100(%sr2, %r0) r20 r28 - -
110 powerpc sc r0 r3 - r0 1
111 powerpc64 sc r0 r3 - cr0.SO 1
112 riscv ecall a7 a0 a1 -
113 s390 svc 0 r1 r2 r3 - 3
114 s390x svc 0 r1 r2 r3 - 3
115 superh trap #0x17 r3 r0 r1 - 4, 6
116 sparc/32 t 0x10 g1 o0 o1 psr/csr 1, 6
117 sparc/64 t 0x6d g1 o0 o1 psr/csr 1, 6
118 tile swint1 R10 R00 - R01 1
119 x86-64 syscall rax rax rdx - 5
120 x32 syscall rax rax rdx - 5
121 xtensa syscall a2 a2 - -
122 Notes:
124 [1] On a few architectures, a register is used as a boolean (0 indicat‐
126 ing no error, and -1 indicating an error) to signal that the system
127 call failed. The actual error value is still contained in the
128 return register. On sparc, the carry bit (csr) in the processor
129 status register (psr) is used instead of a full register. On pow‐
130 erpc64, the summary overflow bit (SO) in field 0 of the condition
131 register (cr0) is used.
132 [2] NR is the system call number.
134 [3] For s390 and s390x, NR (the system call number) may be passed
136 directly with svc NR if it is less than 256.
137 [4] On SuperH, the trap number controls the maximum number of arguments
139 passed. A trap #0x10 can be used with only 0-argument system
140 calls, a trap #0x11 can be used with 0- or 1-argument system calls,
141 and so on up to trap #0x17 for 7-argument system calls.
142 [5] The x32 ABI shares syscall table with x86-64 ABI, but there are
144 some nuances:
145 · In order to indicate that a system call is called under the x32
147 ABI, an additional bit, __X32_SYSCALL_BIT, is bitwise-ORed with
148 the system call number. The ABI used by a process affects some
149 process behaviors, including signal handling or system call
150 restarting.
151 · Since x32 has different sizes for long and pointer types, lay‐
153 outs of some (but not all; struct timeval or struct rlimit are
154 64-bit, for example) structures are different. In order to han‐
155 dle this, additional system calls are added to the system call
156 table, starting from number 512 (without the __X32_SYSCALL_BIT).
157 For example, __NR_readv is defined as 19 for the x86-64 ABI and
158 as __X32_SYSCALL_BIT | 515 for the x32 ABI. Most of these addi‐
159 tional system calls are actually identical to the system calls
160 used for providing i386 compat. There are some notable excep‐
161 tions, however, such as preadv2(2), which uses struct iovec
162 entities with 4-byte pointers and sizes ("compat_iovec" in ker‐
163 nel terms), but passes an 8-byte pos argument in a single regis‐
164 ter and not two, as is done in every other ABI.
165 [6] Some architectures (namely, Alpha, IA-64, MIPS, SuperH, sparc/32,
167 and sparc/64) use an additional register ("Retval2" in the above
168 table) to pass back a second return value from the pipe(2) system
169 call; Alpha uses this technique in the architecture-specific getx‐
170 pid(2), getxuid(2), and getxgid(2) system calls as well. Other
171 architectures do not use the second return value register in the
172 system call interface, even if it is defined in the System V ABI.
173 The second table shows the registers used to pass the system call argu‐
175 ments.
176 Arch/ABI arg1 arg2 arg3 arg4 arg5 arg6 arg7 Notes
178 ──────────────────────────────────────────────────────────────
179 alpha a0 a1 a2 a3 a4 a5 -
180 arc r0 r1 r2 r3 r4 r5 -
181 arm/OABI r0 r1 r2 r3 r4 r5 r6
182 arm/EABI r0 r1 r2 r3 r4 r5 r6
183 arm64 x0 x1 x2 x3 x4 x5 -
184 blackfin R0 R1 R2 R3 R4 R5 -
185 i386 ebx ecx edx esi edi ebp -
186 ia64 out0 out1 out2 out3 out4 out5 -
187 m68k d1 d2 d3 d4 d5 a0 -
188 microblaze r5 r6 r7 r8 r9 r10 -
189 mips/o32 a0 a1 a2 a3 - - - 1
190 mips/n32,64 a0 a1 a2 a3 a4 a5 -
191 nios2 r4 r5 r6 r7 r8 r9 -
192 parisc r26 r25 r24 r23 r22 r21 -
193 powerpc r3 r4 r5 r6 r7 r8 r9
194 powerpc64 r3 r4 r5 r6 r7 r8 -
195 riscv a0 a1 a2 a3 a4 a5 -
196 s390 r2 r3 r4 r5 r6 r7 -
197 s390x r2 r3 r4 r5 r6 r7 -
198 superh r4 r5 r6 r7 r0 r1 r2
200 sparc/32 o0 o1 o2 o3 o4 o5 -
201 sparc/64 o0 o1 o2 o3 o4 o5 -
202 tile R00 R01 R02 R03 R04 R05 -
203 x86-64 rdi rsi rdx r10 r8 r9 -
204 x32 rdi rsi rdx r10 r8 r9 -
205 xtensa a6 a3 a4 a5 a8 a9 -
206 Notes:
208 [1] The mips/o32 system call convention passes arguments 5 through 8 on
210 the user stack.
211 Note that these tables don't cover the entire calling convention—some
213 architectures may indiscriminately clobber other registers not listed
214 here.
215
217 #define _GNU_SOURCE
218 #include <unistd.h>
219 #include <sys/syscall.h>
220 #include <sys/types.h>
221 #include <signal.h>
222 int
224 main(int argc, char *argv[])
225 {
226 pid_t tid;
227 tid = syscall(SYS_gettid);
229 syscall(SYS_tgkill, getpid(), tid, SIGHUP);
230 }
231
233 _syscall(2), intro(2), syscalls(2), errno(3), vdso(7)
234
236 This page is part of release 5.07 of the Linux man-pages project. A
237 description of the project, information about reporting bugs, and the
238 latest version of this page, can be found at
239 https://www.kernel.org/doc/man-pages/.
240Linux 2020-06-09 SYSCALL(2)