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#define REGISTER_SAVE_RAX 0 46#define REGISTER_SAVE_RCX (REGISTER_SAVE_RAX + 8) 47#define REGISTER_SAVE_RDX (REGISTER_SAVE_RCX + 8) 48#define REGISTER_SAVE_RSI (REGISTER_SAVE_RDX + 8) 49#define REGISTER_SAVE_RDI (REGISTER_SAVE_RSI + 8) 50#define REGISTER_SAVE_R8 (REGISTER_SAVE_RDI + 8) 51#define REGISTER_SAVE_R9 (REGISTER_SAVE_R8 + 8) 52 53#define RESTORE_AVX 54 55#define VEC_SIZE 64 56#define VMOVA vmovdqa64 57#define VEC(i) zmm##i 58#define _dl_runtime_profile _dl_runtime_profile_avx512 59#include "dl-trampoline.h" 60#undef _dl_runtime_profile 61#undef VEC 62#undef VMOVA 63#undef VEC_SIZE 64 65#define VEC_SIZE 32 66#define VMOVA vmovdqa 67#define VEC(i) ymm##i 68#define _dl_runtime_profile _dl_runtime_profile_avx 69#include "dl-trampoline.h" 70#undef _dl_runtime_profile 71#undef VEC 72#undef VMOVA 73#undef VEC_SIZE 74 75/* movaps/movups is 1-byte shorter. */ 76#define VEC_SIZE 16 77#define VMOVA movaps 78#define VEC(i) xmm##i 79#define _dl_runtime_profile _dl_runtime_profile_sse 80#undef RESTORE_AVX 81#include "dl-trampoline.h" 82#undef _dl_runtime_profile 83#undef VEC 84#undef VMOVA 85#undef VEC_SIZE 86 87#define USE_FXSAVE 88#define STATE_SAVE_ALIGNMENT 16 89#define _dl_runtime_resolve _dl_runtime_resolve_fxsave 90#include "dl-trampoline.h" 91#undef _dl_runtime_resolve 92#undef USE_FXSAVE 93#undef STATE_SAVE_ALIGNMENT 94 95#define USE_XSAVE 96#define STATE_SAVE_ALIGNMENT 64 97#define _dl_runtime_resolve _dl_runtime_resolve_xsave 98#include "dl-trampoline.h" 99#undef _dl_runtime_resolve 100#undef USE_XSAVE 101#undef STATE_SAVE_ALIGNMENT 102 103#define USE_XSAVEC 104#define STATE_SAVE_ALIGNMENT 64 105#define _dl_runtime_resolve _dl_runtime_resolve_xsavec 106#include "dl-trampoline.h" 107#undef _dl_runtime_resolve 108#undef USE_XSAVEC 109#undef STATE_SAVE_ALIGNMENT 110