startprog64.S revision 1.2.2.3
1 1.2.2.3 skrll /* $NetBSD: startprog64.S,v 1.2.2.3 2017/08/28 17:51:41 skrll Exp $ */ 2 1.2.2.2 skrll /* NetBSD: startprog.S,v 1.3 2003/02/01 14:48:18 dsl Exp */ 3 1.2.2.2 skrll 4 1.2.2.2 skrll /* starts program in protected mode / flat space 5 1.2.2.2 skrll with given stackframe 6 1.2.2.2 skrll needs global variables flatcodeseg and flatdataseg 7 1.2.2.2 skrll (gdt offsets) 8 1.2.2.2 skrll derived from: NetBSD:sys/arch/i386/boot/asm.S 9 1.2.2.2 skrll */ 10 1.2.2.2 skrll 11 1.2.2.2 skrll /* 12 1.2.2.2 skrll * Ported to boot 386BSD by Julian Elischer (julian (at) tfs.com) Sept 1992 13 1.2.2.2 skrll * 14 1.2.2.2 skrll * Mach Operating System 15 1.2.2.2 skrll * Copyright (c) 1992, 1991 Carnegie Mellon University 16 1.2.2.2 skrll * All Rights Reserved. 17 1.2.2.2 skrll * 18 1.2.2.2 skrll * Permission to use, copy, modify and distribute this software and its 19 1.2.2.2 skrll * documentation is hereby granted, provided that both the copyright 20 1.2.2.2 skrll * notice and this permission notice appear in all copies of the 21 1.2.2.2 skrll * software, derivative works or modified versions, and any portions 22 1.2.2.2 skrll * thereof, and that both notices appear in supporting documentation. 23 1.2.2.2 skrll * 24 1.2.2.2 skrll * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 25 1.2.2.2 skrll * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR 26 1.2.2.2 skrll * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 27 1.2.2.2 skrll * 28 1.2.2.2 skrll * Carnegie Mellon requests users of this software to return to 29 1.2.2.2 skrll * 30 1.2.2.2 skrll * Software Distribution Coordinator or Software.Distribution (at) CS.CMU.EDU 31 1.2.2.2 skrll * School of Computer Science 32 1.2.2.2 skrll * Carnegie Mellon University 33 1.2.2.2 skrll * Pittsburgh PA 15213-3890 34 1.2.2.2 skrll * 35 1.2.2.2 skrll * any improvements or extensions that they make and grant Carnegie Mellon 36 1.2.2.2 skrll * the rights to redistribute these changes. 37 1.2.2.2 skrll */ 38 1.2.2.2 skrll 39 1.2.2.2 skrll /* 40 1.2.2.2 skrll Copyright 1988, 1989, 1990, 1991, 1992 41 1.2.2.2 skrll by Intel Corporation, Santa Clara, California. 42 1.2.2.2 skrll 43 1.2.2.2 skrll All Rights Reserved 44 1.2.2.2 skrll 45 1.2.2.2 skrll Permission to use, copy, modify, and distribute this software and 46 1.2.2.2 skrll its documentation for any purpose and without fee is hereby 47 1.2.2.2 skrll granted, provided that the above copyright notice appears in all 48 1.2.2.2 skrll copies and that both the copyright notice and this permission notice 49 1.2.2.2 skrll appear in supporting documentation, and that the name of Intel 50 1.2.2.2 skrll not be used in advertising or publicity pertaining to distribution 51 1.2.2.2 skrll of the software without specific, written prior permission. 52 1.2.2.2 skrll 53 1.2.2.2 skrll INTEL DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE 54 1.2.2.2 skrll INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, 55 1.2.2.2 skrll IN NO EVENT SHALL INTEL BE LIABLE FOR ANY SPECIAL, INDIRECT, OR 56 1.2.2.2 skrll CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM 57 1.2.2.2 skrll LOSS OF USE, DATA OR PROFITS, WHETHER IN ACTION OF CONTRACT, 58 1.2.2.2 skrll NEGLIGENCE, OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION 59 1.2.2.2 skrll WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 60 1.2.2.2 skrll */ 61 1.2.2.2 skrll 62 1.2.2.2 skrll #include <machine/asm.h> 63 1.2.2.2 skrll #include <machine/specialreg.h> 64 1.2.2.2 skrll 65 1.2.2.2 skrll #define CODE_SEGMENT 0x08 66 1.2.2.2 skrll #define DATA_SEGMENT 0x10 67 1.2.2.2 skrll 68 1.2.2.2 skrll .align 16 69 1.2.2.2 skrll .globl _C_LABEL(startprog64) 70 1.2.2.2 skrll _C_LABEL(startprog64): 71 1.2.2.2 skrll .quad 0 72 1.2.2.2 skrll 73 1.2.2.2 skrll .globl _C_LABEL(startprog64_size) 74 1.2.2.2 skrll _C_LABEL(startprog64_size): 75 1.2.2.2 skrll .long startprog64_end - _C_LABEL(startprog64_start) 76 1.2.2.2 skrll 77 1.2.2.2 skrll .text 78 1.2.2.2 skrll .p2align 4,,15 79 1.2.2.2 skrll 80 1.2.2.2 skrll /* 81 1.2.2.3 skrll * startprog64(loadddr,entry,stack,kern_load,kern_start,kern_size) 82 1.2.2.2 skrll */ 83 1.2.2.2 skrll ENTRY(startprog64_start) 84 1.2.2.2 skrll start: 85 1.2.2.2 skrll /* 86 1.2.2.2 skrll * This function is to call the loaded kernel's start() with 87 1.2.2.2 skrll * 32bit segment mode from x64 mode. 88 1.2.2.3 skrll * %rdi: kernel start address 89 1.2.2.3 skrll * %rsi: loaded kernel address 90 1.2.2.2 skrll * %rdx: stack address 91 1.2.2.3 skrll * %rcx: loaded kernel size 92 1.2.2.3 skrll * %r8 : loaded start address 93 1.2.2.3 skrll * %r9 : kernel entry address 94 1.2.2.2 skrll */ 95 1.2.2.2 skrll 96 1.2.2.3 skrll cld /* LynxOS depends on it */ 97 1.2.2.3 skrll 98 1.2.2.3 skrll cli 99 1.2.2.3 skrll 100 1.2.2.3 skrll /* Copy kernel */ 101 1.2.2.3 skrll mov %rcx, %r12 /* original kernel size */ 102 1.2.2.3 skrll movq %rdi, %r11 /* for misaligned check */ 103 1.2.2.3 skrll 104 1.2.2.3 skrll #if !defined(NO_OVERLAP) 105 1.2.2.3 skrll movq %rdi, %r13 106 1.2.2.3 skrll subq %rsi, %r13 107 1.2.2.3 skrll #endif 108 1.2.2.3 skrll 109 1.2.2.3 skrll shrq $3, %rcx /* count for copy by words */ 110 1.2.2.3 skrll jz 8f /* j if less than 8 bytes */ 111 1.2.2.3 skrll 112 1.2.2.3 skrll lea -8(%rdi, %r12), %r14 /* target address of last 8 */ 113 1.2.2.3 skrll mov -8(%rsi, %r12), %r15 /* get last word */ 114 1.2.2.3 skrll #if !defined(NO_OVERLAP) 115 1.2.2.3 skrll cmpq %r12, %r13 /* overlapping? */ 116 1.2.2.3 skrll jb 10f 117 1.2.2.3 skrll #endif 118 1.2.2.3 skrll 119 1.2.2.3 skrll /* 120 1.2.2.3 skrll * Non-overlaping, copy forwards. 121 1.2.2.3 skrll * Newer Intel cpus (Nehalem) will do 16byte read/write transfers 122 1.2.2.3 skrll * if %ecx is more than 76. 123 1.2.2.3 skrll * AMD might do something similar some day. 124 1.2.2.3 skrll */ 125 1.2.2.3 skrll and $7, %r11 /* destination misaligned ? */ 126 1.2.2.3 skrll jnz 2f 127 1.2.2.3 skrll rep 128 1.2.2.3 skrll movsq 129 1.2.2.3 skrll mov %r15, (%r14) /* write last word */ 130 1.2.2.3 skrll jmp .Lcopy_done 131 1.2.2.3 skrll 132 1.2.2.3 skrll /* 133 1.2.2.3 skrll * Destination misaligned 134 1.2.2.3 skrll * AMD say it is better to align the destination (not the source). 135 1.2.2.3 skrll * This will also re-align copies if the source and dest are both 136 1.2.2.3 skrll * misaligned by the same amount) 137 1.2.2.3 skrll * (I think Nehalem will use its accelerated copy if the source 138 1.2.2.3 skrll * and destination have the same alignment.) 139 1.2.2.3 skrll */ 140 1.2.2.3 skrll 2: 141 1.2.2.3 skrll lea -9(%r11, %r12), %rcx /* post re-alignment count */ 142 1.2.2.3 skrll neg %r11 /* now -1 .. -7 */ 143 1.2.2.3 skrll mov (%rsi), %r12 /* get first word */ 144 1.2.2.3 skrll mov %rdi, %r13 /* target for first word */ 145 1.2.2.3 skrll lea 8(%rsi, %r11), %rsi 146 1.2.2.3 skrll lea 8(%rdi, %r11), %rdi 147 1.2.2.3 skrll shr $3, %rcx 148 1.2.2.3 skrll rep 149 1.2.2.3 skrll movsq 150 1.2.2.3 skrll mov %r12, (%r13) /* write first word */ 151 1.2.2.3 skrll mov %r15, (%r14) /* write last word */ 152 1.2.2.3 skrll jmp .Lcopy_done 153 1.2.2.3 skrll 154 1.2.2.3 skrll #if !defined(NO_OVERLAP) 155 1.2.2.3 skrll /* Must copy backwards. 156 1.2.2.3 skrll * Reverse copy is probably easy to code faster than 'rep movds' 157 1.2.2.3 skrll * since that requires (IIRC) an extra clock every 3 iterations (AMD). 158 1.2.2.3 skrll * However I don't suppose anything cares that much! 159 1.2.2.3 skrll * The big cost is the std/cld pair - reputedly 50+ cycles on Netburst P4. 160 1.2.2.3 skrll * The copy is aligned with the buffer start (more likely to 161 1.2.2.3 skrll * be a multiple of 8 than the end). 162 1.2.2.3 skrll */ 163 1.2.2.3 skrll 10: 164 1.2.2.3 skrll lea -8(%rsi, %rcx, 8), %rsi 165 1.2.2.3 skrll lea -8(%rdi, %rcx, 8), %rdi 166 1.2.2.3 skrll std 167 1.2.2.3 skrll rep 168 1.2.2.3 skrll movsq 169 1.2.2.3 skrll cld 170 1.2.2.3 skrll mov %r15, (%r14) /* write last bytes */ 171 1.2.2.3 skrll jmp .Lcopy_done 172 1.2.2.3 skrll #endif 173 1.2.2.3 skrll 174 1.2.2.3 skrll /* Less than 8 bytes to copy, copy by bytes */ 175 1.2.2.3 skrll /* Intel Nehalem optimise 'rep movsb' for <= 7 bytes (9-15 clocks). 176 1.2.2.3 skrll * For longer transfers it is 50+ ! 177 1.2.2.3 skrll */ 178 1.2.2.3 skrll 8: mov %r12, %rcx 179 1.2.2.3 skrll 180 1.2.2.3 skrll #if !defined(NO_OVERLAP) 181 1.2.2.3 skrll cmpq %r12, %r13 /* overlapping? */ 182 1.2.2.3 skrll jb 81f 183 1.2.2.3 skrll #endif 184 1.2.2.3 skrll 185 1.2.2.3 skrll /* nope, copy forwards. */ 186 1.2.2.3 skrll rep 187 1.2.2.3 skrll movsb 188 1.2.2.3 skrll jmp .Lcopy_done 189 1.2.2.3 skrll 190 1.2.2.3 skrll #if !defined(NO_OVERLAP) 191 1.2.2.3 skrll /* Must copy backwards */ 192 1.2.2.3 skrll 81: 193 1.2.2.3 skrll lea -1(%rsi, %rcx), %rsi 194 1.2.2.3 skrll lea -1(%rdi, %rcx), %rdi 195 1.2.2.3 skrll std 196 1.2.2.3 skrll rep 197 1.2.2.3 skrll movsb 198 1.2.2.3 skrll cld 199 1.2.2.3 skrll #endif 200 1.2.2.3 skrll /* End of copy kernel */ 201 1.2.2.3 skrll .Lcopy_done: 202 1.2.2.3 skrll 203 1.2.2.3 skrll mov %r8, %rdi /* %rdi: loaded start address */ 204 1.2.2.3 skrll mov %r9, %rsi /* %rsi: kernel entry address */ 205 1.2.2.2 skrll 206 1.2.2.2 skrll /* Prepare jump address */ 207 1.2.2.2 skrll lea (start32a - start)(%rdi), %rax 208 1.2.2.2 skrll movl %eax, (start32r - start)(%rdi) 209 1.2.2.2 skrll 210 1.2.2.2 skrll /* Setup GDT */ 211 1.2.2.2 skrll lea (gdt - start)(%rdi), %rax 212 1.2.2.2 skrll mov %rax, (gdtrr - start)(%rdi) 213 1.2.2.2 skrll lgdt (gdtr - start)(%rdi) 214 1.2.2.2 skrll 215 1.2.2.2 skrll /* Jump to set %cs */ 216 1.2.2.2 skrll ljmp *(start32r - start)(%rdi) 217 1.2.2.2 skrll 218 1.2.2.2 skrll .align 4 219 1.2.2.2 skrll .code32 220 1.2.2.2 skrll start32a: 221 1.2.2.2 skrll movl $DATA_SEGMENT, %eax 222 1.2.2.3 skrll movw %ax, %ds 223 1.2.2.3 skrll movw %ax, %es 224 1.2.2.3 skrll movw %ax, %fs 225 1.2.2.3 skrll movw %ax, %gs 226 1.2.2.3 skrll movw %ax, %ss 227 1.2.2.2 skrll 228 1.2.2.2 skrll movl %edx, %esp 229 1.2.2.2 skrll 230 1.2.2.2 skrll /* Disable Paging in CR0 */ 231 1.2.2.2 skrll movl %cr0, %eax 232 1.2.2.2 skrll andl $(~CR0_PG), %eax 233 1.2.2.2 skrll movl %eax, %cr0 234 1.2.2.2 skrll 235 1.2.2.2 skrll /* Disable PAE in CR4 */ 236 1.2.2.2 skrll movl %cr4, %eax 237 1.2.2.2 skrll andl $(~CR4_PAE), %eax 238 1.2.2.2 skrll movl %eax, %cr4 239 1.2.2.2 skrll 240 1.2.2.2 skrll jmp start32b 241 1.2.2.2 skrll 242 1.2.2.2 skrll .align 4 243 1.2.2.2 skrll start32b: 244 1.2.2.2 skrll xor %eax, %eax 245 1.2.2.2 skrll call *%esi 246 1.2.2.2 skrll 247 1.2.2.2 skrll .align 16 248 1.2.2.2 skrll start32r: 249 1.2.2.2 skrll .long 0 250 1.2.2.2 skrll .long CODE_SEGMENT 251 1.2.2.2 skrll .align 16 252 1.2.2.2 skrll gdt: 253 1.2.2.2 skrll .long 0, 0 254 1.2.2.2 skrll .byte 0xff, 0xff, 0x00, 0x00, 0x00, 0x9f, 0xcf, 0x00 255 1.2.2.2 skrll .byte 0xff, 0xff, 0x00, 0x00, 0x00, 0x93, 0xcf, 0x00 256 1.2.2.2 skrll gdtr: 257 1.2.2.2 skrll .word gdtr - gdt 258 1.2.2.2 skrll gdtrr: 259 1.2.2.2 skrll .quad 260 1.2.2.2 skrll start32end: 261 1.2.2.2 skrll /* Space for the stack */ 262 1.2.2.2 skrll .align 16 263 1.2.2.2 skrll .space 8192 264 1.2.2.2 skrll startprog64_end: 265
Indexes created Sat Oct 25 01:09:55 GMT 2025