1/* PLT trampolines. x86-64 version.
2 Copyright (C) 2004-2021 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 <https://www.gnu.org/licenses/>. */
18
19#include <config.h>
20#include <sysdep.h>
21#include <cpu-features-offsets.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/* True if _dl_runtime_resolve should align stack for STATE_SAVE or align
38 stack to 16 bytes before calling _dl_fixup. */
39#define DL_RUNTIME_RESOLVE_REALIGN_STACK \
40 (STATE_SAVE_ALIGNMENT > DL_STACK_ALIGNMENT \
41 || 16 > DL_STACK_ALIGNMENT)
42
43/* Area on stack to save and restore registers used for parameter
44 passing when calling _dl_fixup. */
45#ifdef __ILP32__
46# define PRESERVE_BND_REGS_PREFIX
47#else
48# ifdef HAVE_MPX_SUPPORT
49# define PRESERVE_BND_REGS_PREFIX bnd
50# else
51# define PRESERVE_BND_REGS_PREFIX .byte 0xf2
52# endif
53#endif
54#define REGISTER_SAVE_RAX 0
55#define REGISTER_SAVE_RCX (REGISTER_SAVE_RAX + 8)
56#define REGISTER_SAVE_RDX (REGISTER_SAVE_RCX + 8)
57#define REGISTER_SAVE_RSI (REGISTER_SAVE_RDX + 8)
58#define REGISTER_SAVE_RDI (REGISTER_SAVE_RSI + 8)
59#define REGISTER_SAVE_R8 (REGISTER_SAVE_RDI + 8)
60#define REGISTER_SAVE_R9 (REGISTER_SAVE_R8 + 8)
61
62#define RESTORE_AVX
63
64#define VEC_SIZE 64
65#define VMOVA vmovdqa64
66#define VEC(i) zmm##i
67#define _dl_runtime_profile _dl_runtime_profile_avx512
68#include "dl-trampoline.h"
69#undef _dl_runtime_profile
70#undef VEC
71#undef VMOVA
72#undef VEC_SIZE
73
74#define VEC_SIZE 32
75#define VMOVA vmovdqa
76#define VEC(i) ymm##i
77#define _dl_runtime_profile _dl_runtime_profile_avx
78#include "dl-trampoline.h"
79#undef _dl_runtime_profile
80#undef VEC
81#undef VMOVA
82#undef VEC_SIZE
83
84/* movaps/movups is 1-byte shorter. */
85#define VEC_SIZE 16
86#define VMOVA movaps
87#define VEC(i) xmm##i
88#define _dl_runtime_profile _dl_runtime_profile_sse
89#undef RESTORE_AVX
90#include "dl-trampoline.h"
91#undef _dl_runtime_profile
92#undef VEC
93#undef VMOVA
94#undef VEC_SIZE
95
96#define USE_FXSAVE
97#define STATE_SAVE_ALIGNMENT 16
98#define _dl_runtime_resolve _dl_runtime_resolve_fxsave
99#include "dl-trampoline.h"
100#undef _dl_runtime_resolve
101#undef USE_FXSAVE
102#undef STATE_SAVE_ALIGNMENT
103
104#define USE_XSAVE
105#define STATE_SAVE_ALIGNMENT 64
106#define _dl_runtime_resolve _dl_runtime_resolve_xsave
107#include "dl-trampoline.h"
108#undef _dl_runtime_resolve
109#undef USE_XSAVE
110#undef STATE_SAVE_ALIGNMENT
111
112#define USE_XSAVEC
113#define STATE_SAVE_ALIGNMENT 64
114#define _dl_runtime_resolve _dl_runtime_resolve_xsavec
115#include "dl-trampoline.h"
116#undef _dl_runtime_resolve
117#undef USE_XSAVEC
118#undef STATE_SAVE_ALIGNMENT
119