1/* PLT trampolines. x86-64 version.
2 Copyright (C) 2004-2017 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
9
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
14
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, see
17 <http://www.gnu.org/licenses/>. */
18
19#include <config.h>
20#include <sysdep.h>
21#include <cpu-features.h>
22#include <link-defines.h>
23
24#ifndef DL_STACK_ALIGNMENT
25/* Due to GCC bug:
26
27 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58066
28
29 __tls_get_addr may be called with 8-byte stack alignment. Although
30 this bug has been fixed in GCC 4.9.4, 5.3 and 6, we can't assume
31 that stack will be always aligned at 16 bytes. We use unaligned
32 16-byte move to load and store SSE registers, which has no penalty
33 on modern processors if stack is 16-byte aligned. */
34# define DL_STACK_ALIGNMENT 8
35#endif
36
37#ifndef DL_RUNTIME_UNALIGNED_VEC_SIZE
38/* The maximum size in bytes of unaligned vector load and store in the
39 dynamic linker. Since SSE optimized memory/string functions with
40 aligned SSE register load and store are used in the dynamic linker,
41 we must set this to 8 so that _dl_runtime_resolve_sse will align the
42 stack before calling _dl_fixup. */
43# define DL_RUNTIME_UNALIGNED_VEC_SIZE 8
44#endif
45
46/* True if _dl_runtime_resolve should align stack to VEC_SIZE bytes. */
47#define DL_RUNTIME_RESOLVE_REALIGN_STACK \
48 (VEC_SIZE > DL_STACK_ALIGNMENT \
49 && VEC_SIZE > DL_RUNTIME_UNALIGNED_VEC_SIZE)
50
51/* Align vector register save area to 16 bytes. */
52#define REGISTER_SAVE_VEC_OFF 0
53
54/* Area on stack to save and restore registers used for parameter
55 passing when calling _dl_fixup. */
56#ifdef __ILP32__
57# define REGISTER_SAVE_RAX (REGISTER_SAVE_VEC_OFF + VEC_SIZE * 8)
58# define PRESERVE_BND_REGS_PREFIX
59#else
60/* Align bound register save area to 16 bytes. */
61# define REGISTER_SAVE_BND0 (REGISTER_SAVE_VEC_OFF + VEC_SIZE * 8)
62# define REGISTER_SAVE_BND1 (REGISTER_SAVE_BND0 + 16)
63# define REGISTER_SAVE_BND2 (REGISTER_SAVE_BND1 + 16)
64# define REGISTER_SAVE_BND3 (REGISTER_SAVE_BND2 + 16)
65# define REGISTER_SAVE_RAX (REGISTER_SAVE_BND3 + 16)
66# ifdef HAVE_MPX_SUPPORT
67# define PRESERVE_BND_REGS_PREFIX bnd
68# else
69# define PRESERVE_BND_REGS_PREFIX .byte 0xf2
70# endif
71#endif
72#define REGISTER_SAVE_RCX (REGISTER_SAVE_RAX + 8)
73#define REGISTER_SAVE_RDX (REGISTER_SAVE_RCX + 8)
74#define REGISTER_SAVE_RSI (REGISTER_SAVE_RDX + 8)
75#define REGISTER_SAVE_RDI (REGISTER_SAVE_RSI + 8)
76#define REGISTER_SAVE_R8 (REGISTER_SAVE_RDI + 8)
77#define REGISTER_SAVE_R9 (REGISTER_SAVE_R8 + 8)
78
79#define RESTORE_AVX
80
81#define VEC_SIZE 64
82#define VMOVA vmovdqa64
83#if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
84# define VMOV vmovdqa64
85#else
86# define VMOV vmovdqu64
87#endif
88#define VEC(i) zmm##i
89#define _dl_runtime_resolve _dl_runtime_resolve_avx512
90#define _dl_runtime_resolve_opt _dl_runtime_resolve_avx512_opt
91#define _dl_runtime_profile _dl_runtime_profile_avx512
92#include "dl-trampoline.h"
93#undef _dl_runtime_resolve
94#undef _dl_runtime_resolve_opt
95#undef _dl_runtime_profile
96#undef VEC
97#undef VMOV
98#undef VMOVA
99#undef VEC_SIZE
100
101#define VEC_SIZE 32
102#define VMOVA vmovdqa
103#if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
104# define VMOV vmovdqa
105#else
106# define VMOV vmovdqu
107#endif
108#define VEC(i) ymm##i
109#define _dl_runtime_resolve _dl_runtime_resolve_avx
110#define _dl_runtime_resolve_opt _dl_runtime_resolve_avx_opt
111#define _dl_runtime_profile _dl_runtime_profile_avx
112#include "dl-trampoline.h"
113#undef _dl_runtime_resolve
114#undef _dl_runtime_resolve_opt
115#undef _dl_runtime_profile
116#undef VEC
117#undef VMOV
118#undef VMOVA
119#undef VEC_SIZE
120
121/* movaps/movups is 1-byte shorter. */
122#define VEC_SIZE 16
123#define VMOVA movaps
124#if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
125# define VMOV movaps
126#else
127# define VMOV movups
128#endif
129#define VEC(i) xmm##i
130#define _dl_runtime_resolve _dl_runtime_resolve_sse
131#define _dl_runtime_profile _dl_runtime_profile_sse
132#undef RESTORE_AVX
133#include "dl-trampoline.h"
134#undef _dl_runtime_resolve
135#undef _dl_runtime_profile
136#undef VMOV
137#undef VMOVA
138
139/* Used by _dl_runtime_resolve_avx_opt/_dl_runtime_resolve_avx512_opt
140 to preserve the full vector registers with zero upper bits. */
141#define VMOVA vmovdqa
142#if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
143# define VMOV vmovdqa
144#else
145# define VMOV vmovdqu
146#endif
147#define _dl_runtime_resolve _dl_runtime_resolve_sse_vex
148#include "dl-trampoline.h"
149