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.
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.
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.
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/>. */
21 #include <cpu-features.h>
22 #include <link-defines.h>
24 #ifndef DL_STACK_ALIGNMENT
27 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58066
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
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
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)
51 /* Align vector register save area to 16 bytes. */
52 #define REGISTER_SAVE_VEC_OFF 0
54 /* Area on stack to save and restore registers used for parameter
55 passing when calling _dl_fixup. */
57 # define REGISTER_SAVE_RAX (REGISTER_SAVE_VEC_OFF + VEC_SIZE * 8)
58 # define PRESERVE_BND_REGS_PREFIX
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
69 # define PRESERVE_BND_REGS_PREFIX .byte 0xf2
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)
82 #define VMOVA vmovdqa64
83 #if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
84 # define VMOV vmovdqa64
86 # define VMOV vmovdqu64
89 #define _dl_runtime_resolve _dl_runtime_resolve_avx512
90 #define _dl_runtime_profile _dl_runtime_profile_avx512
91 #include "dl-trampoline.h"
92 #undef _dl_runtime_resolve
93 #undef _dl_runtime_profile
100 #define VMOVA vmovdqa
101 #if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
102 # define VMOV vmovdqa
104 # define VMOV vmovdqu
106 #define VEC(i) ymm##i
107 #define _dl_runtime_resolve _dl_runtime_resolve_avx
108 #define _dl_runtime_resolve_opt _dl_runtime_resolve_avx_opt
109 #define _dl_runtime_profile _dl_runtime_profile_avx
110 #include "dl-trampoline.h"
111 #undef _dl_runtime_resolve
112 #undef _dl_runtime_resolve_opt
113 #undef _dl_runtime_profile
119 /* movaps/movups is 1-byte shorter. */
122 #if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
127 #define VEC(i) xmm##i
128 #define _dl_runtime_resolve _dl_runtime_resolve_sse
129 #define _dl_runtime_profile _dl_runtime_profile_sse
131 #include "dl-trampoline.h"
132 #undef _dl_runtime_resolve
133 #undef _dl_runtime_profile
137 /* Used by _dl_runtime_resolve_avx_opt/_dl_runtime_resolve_avx512_opt
138 to preserve the full vector registers with zero upper bits. */
139 #define VMOVA vmovdqa
140 #if DL_RUNTIME_RESOLVE_REALIGN_STACK || VEC_SIZE <= DL_STACK_ALIGNMENT
141 # define VMOV vmovdqa
143 # define VMOV vmovdqu
145 #define _dl_runtime_resolve _dl_runtime_resolve_sse_vex
146 #define _dl_runtime_resolve_opt _dl_runtime_resolve_avx512_opt
147 #include "dl-trampoline.h"