1 | /* Copyright (C) 2001-2020 Free Software Foundation, Inc. |
2 | This file is part of the GNU C Library. |
3 | |
4 | The GNU C Library is free software; you can redistribute it and/or |
5 | modify it under the terms of the GNU Lesser General Public |
6 | License as published by the Free Software Foundation; either |
7 | version 2.1 of the License, or (at your option) any later version. |
8 | |
9 | The GNU C Library is distributed in the hope that it will be useful, |
10 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
11 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
12 | Lesser General Public License for more details. |
13 | |
14 | You should have received a copy of the GNU Lesser General Public |
15 | License along with the GNU C Library; if not, see |
16 | <https://www.gnu.org/licenses/>. */ |
17 | |
18 | #ifndef _LINUX_X86_64_SYSDEP_H |
19 | #define _LINUX_X86_64_SYSDEP_H 1 |
20 | |
21 | /* There is some commonality. */ |
22 | #include <sysdeps/unix/sysv/linux/sysdep.h> |
23 | #include <sysdeps/unix/x86_64/sysdep.h> |
24 | #include <tls.h> |
25 | |
26 | /* Defines RTLD_PRIVATE_ERRNO. */ |
27 | #include <dl-sysdep.h> |
28 | |
29 | /* For Linux we can use the system call table in the header file |
30 | /usr/include/asm/unistd.h |
31 | of the kernel. But these symbols do not follow the SYS_* syntax |
32 | so we have to redefine the `SYS_ify' macro here. */ |
33 | #undef SYS_ify |
34 | #define SYS_ify(syscall_name) __NR_##syscall_name |
35 | |
36 | /* This is to help the old kernel headers where __NR_semtimedop is not |
37 | available. */ |
38 | #ifndef __NR_semtimedop |
39 | # define __NR_semtimedop 220 |
40 | #endif |
41 | |
42 | |
43 | #ifdef __ASSEMBLER__ |
44 | |
45 | /* Linux uses a negative return value to indicate syscall errors, |
46 | unlike most Unices, which use the condition codes' carry flag. |
47 | |
48 | Since version 2.1 the return value of a system call might be |
49 | negative even if the call succeeded. E.g., the `lseek' system call |
50 | might return a large offset. Therefore we must not anymore test |
51 | for < 0, but test for a real error by making sure the value in %eax |
52 | is a real error number. Linus said he will make sure the no syscall |
53 | returns a value in -1 .. -4095 as a valid result so we can savely |
54 | test with -4095. */ |
55 | |
56 | /* We don't want the label for the error handle to be global when we define |
57 | it here. */ |
58 | # ifdef PIC |
59 | # define SYSCALL_ERROR_LABEL 0f |
60 | # else |
61 | # define SYSCALL_ERROR_LABEL syscall_error |
62 | # endif |
63 | |
64 | # undef PSEUDO |
65 | # define PSEUDO(name, syscall_name, args) \ |
66 | .text; \ |
67 | ENTRY (name) \ |
68 | DO_CALL (syscall_name, args); \ |
69 | cmpq $-4095, %rax; \ |
70 | jae SYSCALL_ERROR_LABEL |
71 | |
72 | # undef PSEUDO_END |
73 | # define PSEUDO_END(name) \ |
74 | SYSCALL_ERROR_HANDLER \ |
75 | END (name) |
76 | |
77 | # undef PSEUDO_NOERRNO |
78 | # define PSEUDO_NOERRNO(name, syscall_name, args) \ |
79 | .text; \ |
80 | ENTRY (name) \ |
81 | DO_CALL (syscall_name, args) |
82 | |
83 | # undef PSEUDO_END_NOERRNO |
84 | # define PSEUDO_END_NOERRNO(name) \ |
85 | END (name) |
86 | |
87 | # define ret_NOERRNO ret |
88 | |
89 | # undef PSEUDO_ERRVAL |
90 | # define PSEUDO_ERRVAL(name, syscall_name, args) \ |
91 | .text; \ |
92 | ENTRY (name) \ |
93 | DO_CALL (syscall_name, args); \ |
94 | negq %rax |
95 | |
96 | # undef PSEUDO_END_ERRVAL |
97 | # define PSEUDO_END_ERRVAL(name) \ |
98 | END (name) |
99 | |
100 | # define ret_ERRVAL ret |
101 | |
102 | # if defined PIC && RTLD_PRIVATE_ERRNO |
103 | # define SYSCALL_SET_ERRNO \ |
104 | lea rtld_errno(%rip), %RCX_LP; \ |
105 | neg %eax; \ |
106 | movl %eax, (%rcx) |
107 | # else |
108 | # if IS_IN (libc) |
109 | # define SYSCALL_ERROR_ERRNO __libc_errno |
110 | # else |
111 | # define SYSCALL_ERROR_ERRNO errno |
112 | # endif |
113 | # define SYSCALL_SET_ERRNO \ |
114 | movq SYSCALL_ERROR_ERRNO@GOTTPOFF(%rip), %rcx;\ |
115 | neg %eax; \ |
116 | movl %eax, %fs:(%rcx); |
117 | # endif |
118 | |
119 | # ifndef PIC |
120 | # define SYSCALL_ERROR_HANDLER /* Nothing here; code in sysdep.S is used. */ |
121 | # else |
122 | # define SYSCALL_ERROR_HANDLER \ |
123 | 0: \ |
124 | SYSCALL_SET_ERRNO; \ |
125 | or $-1, %RAX_LP; \ |
126 | ret; |
127 | # endif /* PIC */ |
128 | |
129 | /* The Linux/x86-64 kernel expects the system call parameters in |
130 | registers according to the following table: |
131 | |
132 | syscall number rax |
133 | arg 1 rdi |
134 | arg 2 rsi |
135 | arg 3 rdx |
136 | arg 4 r10 |
137 | arg 5 r8 |
138 | arg 6 r9 |
139 | |
140 | The Linux kernel uses and destroys internally these registers: |
141 | return address from |
142 | syscall rcx |
143 | eflags from syscall r11 |
144 | |
145 | Normal function call, including calls to the system call stub |
146 | functions in the libc, get the first six parameters passed in |
147 | registers and the seventh parameter and later on the stack. The |
148 | register use is as follows: |
149 | |
150 | system call number in the DO_CALL macro |
151 | arg 1 rdi |
152 | arg 2 rsi |
153 | arg 3 rdx |
154 | arg 4 rcx |
155 | arg 5 r8 |
156 | arg 6 r9 |
157 | |
158 | We have to take care that the stack is aligned to 16 bytes. When |
159 | called the stack is not aligned since the return address has just |
160 | been pushed. |
161 | |
162 | |
163 | Syscalls of more than 6 arguments are not supported. */ |
164 | |
165 | # undef DO_CALL |
166 | # define DO_CALL(syscall_name, args) \ |
167 | DOARGS_##args \ |
168 | movl $SYS_ify (syscall_name), %eax; \ |
169 | syscall; |
170 | |
171 | # define DOARGS_0 /* nothing */ |
172 | # define DOARGS_1 /* nothing */ |
173 | # define DOARGS_2 /* nothing */ |
174 | # define DOARGS_3 /* nothing */ |
175 | # define DOARGS_4 movq %rcx, %r10; |
176 | # define DOARGS_5 DOARGS_4 |
177 | # define DOARGS_6 DOARGS_5 |
178 | |
179 | #else /* !__ASSEMBLER__ */ |
180 | /* Define a macro which expands inline into the wrapper code for a system |
181 | call. */ |
182 | # undef INLINE_SYSCALL |
183 | # define INLINE_SYSCALL(name, nr, args...) \ |
184 | ({ \ |
185 | unsigned long int resultvar = INTERNAL_SYSCALL (name, , nr, args); \ |
186 | if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (resultvar, ))) \ |
187 | { \ |
188 | __set_errno (INTERNAL_SYSCALL_ERRNO (resultvar, )); \ |
189 | resultvar = (unsigned long int) -1; \ |
190 | } \ |
191 | (long int) resultvar; }) |
192 | |
193 | /* Define a macro with explicit types for arguments, which expands inline |
194 | into the wrapper code for a system call. It should be used when size |
195 | of any argument > size of long int. */ |
196 | # undef INLINE_SYSCALL_TYPES |
197 | # define INLINE_SYSCALL_TYPES(name, nr, args...) \ |
198 | ({ \ |
199 | unsigned long int resultvar = INTERNAL_SYSCALL_TYPES (name, , nr, args); \ |
200 | if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (resultvar, ))) \ |
201 | { \ |
202 | __set_errno (INTERNAL_SYSCALL_ERRNO (resultvar, )); \ |
203 | resultvar = (unsigned long int) -1; \ |
204 | } \ |
205 | (long int) resultvar; }) |
206 | |
207 | # undef INTERNAL_SYSCALL_DECL |
208 | # define INTERNAL_SYSCALL_DECL(err) do { } while (0) |
209 | |
210 | /* Registers clobbered by syscall. */ |
211 | # define REGISTERS_CLOBBERED_BY_SYSCALL "cc", "r11", "cx" |
212 | |
213 | /* Create a variable 'name' based on type 'X' to avoid explicit types. |
214 | This is mainly used set use 64-bits arguments in x32. */ |
215 | #define TYPEFY(X, name) __typeof__ ((X) - (X)) name |
216 | /* Explicit cast the argument to avoid integer from pointer warning on |
217 | x32. */ |
218 | #define ARGIFY(X) ((__typeof__ ((X) - (X))) (X)) |
219 | |
220 | #undef INTERNAL_SYSCALL |
221 | #define INTERNAL_SYSCALL(name, err, nr, args...) \ |
222 | internal_syscall##nr (SYS_ify (name), err, args) |
223 | |
224 | #undef INTERNAL_SYSCALL_NCS |
225 | #define INTERNAL_SYSCALL_NCS(number, err, nr, args...) \ |
226 | internal_syscall##nr (number, err, args) |
227 | |
228 | #undef internal_syscall0 |
229 | #define internal_syscall0(number, err, dummy...) \ |
230 | ({ \ |
231 | unsigned long int resultvar; \ |
232 | asm volatile ( \ |
233 | "syscall\n\t" \ |
234 | : "=a" (resultvar) \ |
235 | : "0" (number) \ |
236 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
237 | (long int) resultvar; \ |
238 | }) |
239 | |
240 | #undef internal_syscall1 |
241 | #define internal_syscall1(number, err, arg1) \ |
242 | ({ \ |
243 | unsigned long int resultvar; \ |
244 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
245 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
246 | asm volatile ( \ |
247 | "syscall\n\t" \ |
248 | : "=a" (resultvar) \ |
249 | : "0" (number), "r" (_a1) \ |
250 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
251 | (long int) resultvar; \ |
252 | }) |
253 | |
254 | #undef internal_syscall2 |
255 | #define internal_syscall2(number, err, arg1, arg2) \ |
256 | ({ \ |
257 | unsigned long int resultvar; \ |
258 | TYPEFY (arg2, __arg2) = ARGIFY (arg2); \ |
259 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
260 | register TYPEFY (arg2, _a2) asm ("rsi") = __arg2; \ |
261 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
262 | asm volatile ( \ |
263 | "syscall\n\t" \ |
264 | : "=a" (resultvar) \ |
265 | : "0" (number), "r" (_a1), "r" (_a2) \ |
266 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
267 | (long int) resultvar; \ |
268 | }) |
269 | |
270 | #undef internal_syscall3 |
271 | #define internal_syscall3(number, err, arg1, arg2, arg3) \ |
272 | ({ \ |
273 | unsigned long int resultvar; \ |
274 | TYPEFY (arg3, __arg3) = ARGIFY (arg3); \ |
275 | TYPEFY (arg2, __arg2) = ARGIFY (arg2); \ |
276 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
277 | register TYPEFY (arg3, _a3) asm ("rdx") = __arg3; \ |
278 | register TYPEFY (arg2, _a2) asm ("rsi") = __arg2; \ |
279 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
280 | asm volatile ( \ |
281 | "syscall\n\t" \ |
282 | : "=a" (resultvar) \ |
283 | : "0" (number), "r" (_a1), "r" (_a2), "r" (_a3) \ |
284 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
285 | (long int) resultvar; \ |
286 | }) |
287 | |
288 | #undef internal_syscall4 |
289 | #define internal_syscall4(number, err, arg1, arg2, arg3, arg4) \ |
290 | ({ \ |
291 | unsigned long int resultvar; \ |
292 | TYPEFY (arg4, __arg4) = ARGIFY (arg4); \ |
293 | TYPEFY (arg3, __arg3) = ARGIFY (arg3); \ |
294 | TYPEFY (arg2, __arg2) = ARGIFY (arg2); \ |
295 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
296 | register TYPEFY (arg4, _a4) asm ("r10") = __arg4; \ |
297 | register TYPEFY (arg3, _a3) asm ("rdx") = __arg3; \ |
298 | register TYPEFY (arg2, _a2) asm ("rsi") = __arg2; \ |
299 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
300 | asm volatile ( \ |
301 | "syscall\n\t" \ |
302 | : "=a" (resultvar) \ |
303 | : "0" (number), "r" (_a1), "r" (_a2), "r" (_a3), "r" (_a4) \ |
304 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
305 | (long int) resultvar; \ |
306 | }) |
307 | |
308 | #undef internal_syscall5 |
309 | #define internal_syscall5(number, err, arg1, arg2, arg3, arg4, arg5) \ |
310 | ({ \ |
311 | unsigned long int resultvar; \ |
312 | TYPEFY (arg5, __arg5) = ARGIFY (arg5); \ |
313 | TYPEFY (arg4, __arg4) = ARGIFY (arg4); \ |
314 | TYPEFY (arg3, __arg3) = ARGIFY (arg3); \ |
315 | TYPEFY (arg2, __arg2) = ARGIFY (arg2); \ |
316 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
317 | register TYPEFY (arg5, _a5) asm ("r8") = __arg5; \ |
318 | register TYPEFY (arg4, _a4) asm ("r10") = __arg4; \ |
319 | register TYPEFY (arg3, _a3) asm ("rdx") = __arg3; \ |
320 | register TYPEFY (arg2, _a2) asm ("rsi") = __arg2; \ |
321 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
322 | asm volatile ( \ |
323 | "syscall\n\t" \ |
324 | : "=a" (resultvar) \ |
325 | : "0" (number), "r" (_a1), "r" (_a2), "r" (_a3), "r" (_a4), \ |
326 | "r" (_a5) \ |
327 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
328 | (long int) resultvar; \ |
329 | }) |
330 | |
331 | #undef internal_syscall6 |
332 | #define internal_syscall6(number, err, arg1, arg2, arg3, arg4, arg5, arg6) \ |
333 | ({ \ |
334 | unsigned long int resultvar; \ |
335 | TYPEFY (arg6, __arg6) = ARGIFY (arg6); \ |
336 | TYPEFY (arg5, __arg5) = ARGIFY (arg5); \ |
337 | TYPEFY (arg4, __arg4) = ARGIFY (arg4); \ |
338 | TYPEFY (arg3, __arg3) = ARGIFY (arg3); \ |
339 | TYPEFY (arg2, __arg2) = ARGIFY (arg2); \ |
340 | TYPEFY (arg1, __arg1) = ARGIFY (arg1); \ |
341 | register TYPEFY (arg6, _a6) asm ("r9") = __arg6; \ |
342 | register TYPEFY (arg5, _a5) asm ("r8") = __arg5; \ |
343 | register TYPEFY (arg4, _a4) asm ("r10") = __arg4; \ |
344 | register TYPEFY (arg3, _a3) asm ("rdx") = __arg3; \ |
345 | register TYPEFY (arg2, _a2) asm ("rsi") = __arg2; \ |
346 | register TYPEFY (arg1, _a1) asm ("rdi") = __arg1; \ |
347 | asm volatile ( \ |
348 | "syscall\n\t" \ |
349 | : "=a" (resultvar) \ |
350 | : "0" (number), "r" (_a1), "r" (_a2), "r" (_a3), "r" (_a4), \ |
351 | "r" (_a5), "r" (_a6) \ |
352 | : "memory", REGISTERS_CLOBBERED_BY_SYSCALL); \ |
353 | (long int) resultvar; \ |
354 | }) |
355 | |
356 | # undef INTERNAL_SYSCALL_ERROR_P |
357 | # define INTERNAL_SYSCALL_ERROR_P(val, err) \ |
358 | ((unsigned long int) (long int) (val) >= -4095L) |
359 | |
360 | # undef INTERNAL_SYSCALL_ERRNO |
361 | # define INTERNAL_SYSCALL_ERRNO(val, err) (-(val)) |
362 | |
363 | # define VDSO_NAME "LINUX_2.6" |
364 | # define VDSO_HASH 61765110 |
365 | |
366 | /* List of system calls which are supported as vsyscalls. */ |
367 | # define HAVE_CLOCK_GETTIME64_VSYSCALL "__vdso_clock_gettime" |
368 | # define HAVE_GETTIMEOFDAY_VSYSCALL "__vdso_gettimeofday" |
369 | # define HAVE_TIME_VSYSCALL "__vdso_time" |
370 | # define HAVE_GETCPU_VSYSCALL "__vdso_getcpu" |
371 | # define HAVE_CLOCK_GETRES64_VSYSCALL "__vdso_clock_getres" |
372 | |
373 | # define SINGLE_THREAD_BY_GLOBAL 1 |
374 | |
375 | #endif /* __ASSEMBLER__ */ |
376 | |
377 | |
378 | /* Pointer mangling support. */ |
379 | #if IS_IN (rtld) |
380 | /* We cannot use the thread descriptor because in ld.so we use setjmp |
381 | earlier than the descriptor is initialized. */ |
382 | # ifdef __ASSEMBLER__ |
383 | # define PTR_MANGLE(reg) xor __pointer_chk_guard_local(%rip), reg; \ |
384 | rol $2*LP_SIZE+1, reg |
385 | # define PTR_DEMANGLE(reg) ror $2*LP_SIZE+1, reg; \ |
386 | xor __pointer_chk_guard_local(%rip), reg |
387 | # else |
388 | # define PTR_MANGLE(reg) asm ("xor __pointer_chk_guard_local(%%rip), %0\n" \ |
389 | "rol $2*" LP_SIZE "+1, %0" \ |
390 | : "=r" (reg) : "0" (reg)) |
391 | # define PTR_DEMANGLE(reg) asm ("ror $2*" LP_SIZE "+1, %0\n" \ |
392 | "xor __pointer_chk_guard_local(%%rip), %0" \ |
393 | : "=r" (reg) : "0" (reg)) |
394 | # endif |
395 | #else |
396 | # ifdef __ASSEMBLER__ |
397 | # define PTR_MANGLE(reg) xor %fs:POINTER_GUARD, reg; \ |
398 | rol $2*LP_SIZE+1, reg |
399 | # define PTR_DEMANGLE(reg) ror $2*LP_SIZE+1, reg; \ |
400 | xor %fs:POINTER_GUARD, reg |
401 | # else |
402 | # define PTR_MANGLE(var) asm ("xor %%fs:%c2, %0\n" \ |
403 | "rol $2*" LP_SIZE "+1, %0" \ |
404 | : "=r" (var) \ |
405 | : "0" (var), \ |
406 | "i" (offsetof (tcbhead_t, \ |
407 | pointer_guard))) |
408 | # define PTR_DEMANGLE(var) asm ("ror $2*" LP_SIZE "+1, %0\n" \ |
409 | "xor %%fs:%c2, %0" \ |
410 | : "=r" (var) \ |
411 | : "0" (var), \ |
412 | "i" (offsetof (tcbhead_t, \ |
413 | pointer_guard))) |
414 | # endif |
415 | #endif |
416 | |
417 | /* How to pass the off{64}_t argument on p{readv,writev}{64}. */ |
418 | #undef LO_HI_LONG |
419 | #define LO_HI_LONG(val) (val), 0 |
420 | |
421 | /* Each shadow stack slot takes 8 bytes. Assuming that each stack |
422 | frame takes 256 bytes, this is used to compute shadow stack size |
423 | from stack size. */ |
424 | #define STACK_SIZE_TO_SHADOW_STACK_SIZE_SHIFT 5 |
425 | |
426 | #endif /* linux/x86_64/sysdep.h */ |
427 | |