1/* 2 * linux/boot/head.S 3 * 4 * Copyright (C) 1991, 1992, 1993 Linus Torvalds 5 */ 6 7/* 8 * head.S contains the 32-bit startup code. 9 * 10 * NOTE!!! Startup happens at absolute address 0x00001000, which is also where 11 * the page directory will exist. The startup code will be overwritten by 12 * the page directory. [According to comments etc elsewhere on a compressed 13 * kernel it will end up at 0x1000 + 1Mb I hope so as I assume this. - AC] 14 * 15 * Page 0 is deliberately kept safe, since System Management Mode code in 16 * laptops may need to access the BIOS data stored there. This is also 17 * useful for future device drivers that either access the BIOS via VM86 18 * mode. 19 */ 20 21/* 22 * High loaded stuff by Hans Lermen & Werner Almesberger, Feb. 1996 23 */ 24 .code32 25 .text 26 27#include <linux/init.h> 28#include <linux/linkage.h> 29#include <asm/segment.h> 30#include <asm/pgtable_types.h> 31#include <asm/page_types.h> 32#include <asm/boot.h> 33#include <asm/msr.h> 34#include <asm/processor-flags.h> 35#include <asm/asm-offsets.h> 36 37 __HEAD 38 .code32 39ENTRY(startup_32) 40 cld 41 /* 42 * Test KEEP_SEGMENTS flag to see if the bootloader is asking 43 * us to not reload segments 44 */ 45 testb $(1<<6), BP_loadflags(%esi) 46 jnz 1f 47 48 cli 49 movl $(__KERNEL_DS), %eax 50 movl %eax, %ds 51 movl %eax, %es 52 movl %eax, %ss 531: 54 55/* 56 * Calculate the delta between where we were compiled to run 57 * at and where we were actually loaded at. This can only be done 58 * with a short local call on x86. Nothing else will tell us what 59 * address we are running at. The reserved chunk of the real-mode 60 * data at 0x1e4 (defined as a scratch field) are used as the stack 61 * for this calculation. Only 4 bytes are needed. 62 */ 63 leal (BP_scratch+4)(%esi), %esp 64 call 1f 651: popl %ebp 66 subl $1b, %ebp 67 68/* setup a stack and make sure cpu supports long mode. */ 69 movl $boot_stack_end, %eax 70 addl %ebp, %eax 71 movl %eax, %esp 72 73 call verify_cpu 74 testl %eax, %eax 75 jnz no_longmode 76 77/* 78 * Compute the delta between where we were compiled to run at 79 * and where the code will actually run at. 80 * 81 * %ebp contains the address we are loaded at by the boot loader and %ebx 82 * contains the address where we should move the kernel image temporarily 83 * for safe in-place decompression. 84 */ 85 86#ifdef CONFIG_RELOCATABLE 87 movl %ebp, %ebx 88 movl BP_kernel_alignment(%esi), %eax 89 decl %eax 90 addl %eax, %ebx 91 notl %eax 92 andl %eax, %ebx 93#else 94 movl $LOAD_PHYSICAL_ADDR, %ebx 95#endif 96 97 /* Target address to relocate to for decompression */ 98 addl $z_extract_offset, %ebx 99 100/* 101 * Prepare for entering 64 bit mode 102 */ 103 104 /* Load new GDT with the 64bit segments using 32bit descriptor */ 105 leal gdt(%ebp), %eax 106 movl %eax, gdt+2(%ebp) 107 lgdt gdt(%ebp) 108 109 /* Enable PAE mode */ 110 movl $(X86_CR4_PAE), %eax 111 movl %eax, %cr4 112 113 /* 114 * Build early 4G boot pagetable 115 */ 116 /* Initialize Page tables to 0 */ 117 leal pgtable(%ebx), %edi 118 xorl %eax, %eax 119 movl $((4096*6)/4), %ecx 120 rep stosl 121 122 /* Build Level 4 */ 123 leal pgtable + 0(%ebx), %edi 124 leal 0x1007 (%edi), %eax 125 movl %eax, 0(%edi) 126 127 /* Build Level 3 */ 128 leal pgtable + 0x1000(%ebx), %edi 129 leal 0x1007(%edi), %eax 130 movl $4, %ecx 1311: movl %eax, 0x00(%edi) 132 addl $0x00001000, %eax 133 addl $8, %edi 134 decl %ecx 135 jnz 1b 136 137 /* Build Level 2 */ 138 leal pgtable + 0x2000(%ebx), %edi 139 movl $0x00000183, %eax 140 movl $2048, %ecx 1411: movl %eax, 0(%edi) 142 addl $0x00200000, %eax 143 addl $8, %edi 144 decl %ecx 145 jnz 1b 146 147 /* Enable the boot page tables */ 148 leal pgtable(%ebx), %eax 149 movl %eax, %cr3 150 151 /* Enable Long mode in EFER (Extended Feature Enable Register) */ 152 movl $MSR_EFER, %ecx 153 rdmsr 154 btsl $_EFER_LME, %eax 155 wrmsr 156 157 /* 158 * Setup for the jump to 64bit mode 159 * 160 * When the jump is performend we will be in long mode but 161 * in 32bit compatibility mode with EFER.LME = 1, CS.L = 0, CS.D = 1 162 * (and in turn EFER.LMA = 1). To jump into 64bit mode we use 163 * the new gdt/idt that has __KERNEL_CS with CS.L = 1. 164 * We place all of the values on our mini stack so lret can 165 * used to perform that far jump. 166 */ 167 pushl $__KERNEL_CS 168 leal startup_64(%ebp), %eax 169 pushl %eax 170 171 /* Enter paged protected Mode, activating Long Mode */ 172 movl $(X86_CR0_PG | X86_CR0_PE), %eax /* Enable Paging and Protected mode */ 173 movl %eax, %cr0 174 175 /* Jump from 32bit compatibility mode into 64bit mode. */ 176 lret 177ENDPROC(startup_32) 178 179no_longmode: 180 /* This isn't an x86-64 CPU so hang */ 1811: 182 hlt 183 jmp 1b 184 185#include "../../kernel/verify_cpu_64.S" 186 187 /* 188 * Be careful here startup_64 needs to be at a predictable 189 * address so I can export it in an ELF header. Bootloaders 190 * should look at the ELF header to find this address, as 191 * it may change in the future. 192 */ 193 .code64 194 .org 0x200 195ENTRY(startup_64) 196 /* 197 * We come here either from startup_32 or directly from a 198 * 64bit bootloader. If we come here from a bootloader we depend on 199 * an identity mapped page table being provied that maps our 200 * entire text+data+bss and hopefully all of memory. 201 */ 202 203 /* Setup data segments. */ 204 xorl %eax, %eax 205 movl %eax, %ds 206 movl %eax, %es 207 movl %eax, %ss 208 movl %eax, %fs 209 movl %eax, %gs 210 lldt %ax 211 movl $0x20, %eax 212 ltr %ax 213 214 /* 215 * Compute the decompressed kernel start address. It is where 216 * we were loaded at aligned to a 2M boundary. %rbp contains the 217 * decompressed kernel start address. 218 * 219 * If it is a relocatable kernel then decompress and run the kernel 220 * from load address aligned to 2MB addr, otherwise decompress and 221 * run the kernel from LOAD_PHYSICAL_ADDR 222 * 223 * We cannot rely on the calculation done in 32-bit mode, since we 224 * may have been invoked via the 64-bit entry point. 225 */ 226 227 /* Start with the delta to where the kernel will run at. */ 228#ifdef CONFIG_RELOCATABLE 229 leaq startup_32(%rip) /* - $startup_32 */, %rbp 230 movl BP_kernel_alignment(%rsi), %eax 231 decl %eax 232 addq %rax, %rbp 233 notq %rax 234 andq %rax, %rbp 235#else 236 movq $LOAD_PHYSICAL_ADDR, %rbp 237#endif 238 239 /* Target address to relocate to for decompression */ 240 leaq z_extract_offset(%rbp), %rbx 241 242 /* Set up the stack */ 243 leaq boot_stack_end(%rbx), %rsp 244 245 /* Zero EFLAGS */ 246 pushq $0 247 popfq 248 249/* 250 * Copy the compressed kernel to the end of our buffer 251 * where decompression in place becomes safe. 252 */ 253 pushq %rsi 254 leaq (_bss-8)(%rip), %rsi 255 leaq (_bss-8)(%rbx), %rdi 256 movq $_bss /* - $startup_32 */, %rcx 257 shrq $3, %rcx 258 std 259 rep movsq 260 cld 261 popq %rsi 262 263/* 264 * Jump to the relocated address. 265 */ 266 leaq relocated(%rbx), %rax 267 jmp *%rax 268 269 .text 270relocated: 271 272/* 273 * Clear BSS (stack is currently empty) 274 */ 275 xorl %eax, %eax 276 leaq _bss(%rip), %rdi 277 leaq _ebss(%rip), %rcx 278 subq %rdi, %rcx 279 shrq $3, %rcx 280 rep stosq 281 282/* 283 * Adjust our own GOT 284 */ 285 leaq _got(%rip), %rdx 286 leaq _egot(%rip), %rcx 2871: 288 cmpq %rcx, %rdx 289 jae 2f 290 addq %rbx, (%rdx) 291 addq $8, %rdx 292 jmp 1b 2932: 294 295/* 296 * Do the decompression, and jump to the new kernel.. 297 */ 298 pushq %rsi /* Save the real mode argument */ 299 movq %rsi, %rdi /* real mode address */ 300 leaq boot_heap(%rip), %rsi /* malloc area for uncompression */ 301 leaq input_data(%rip), %rdx /* input_data */ 302 movl $z_input_len, %ecx /* input_len */ 303 movq %rbp, %r8 /* output target address */ 304 call decompress_kernel 305 popq %rsi 306 307/* 308 * Jump to the decompressed kernel. 309 */ 310 jmp *%rbp 311 312 .data 313gdt: 314 .word gdt_end - gdt 315 .long gdt 316 .word 0 317 .quad 0x0000000000000000 /* NULL descriptor */ 318 .quad 0x00af9a000000ffff /* __KERNEL_CS */ 319 .quad 0x00cf92000000ffff /* __KERNEL_DS */ 320 .quad 0x0080890000000000 /* TS descriptor */ 321 .quad 0x0000000000000000 /* TS continued */ 322gdt_end: 323 324/* 325 * Stack and heap for uncompression 326 */ 327 .bss 328 .balign 4 329boot_heap: 330 .fill BOOT_HEAP_SIZE, 1, 0 331boot_stack: 332 .fill BOOT_STACK_SIZE, 1, 0 333boot_stack_end: 334 335/* 336 * Space for page tables (not in .bss so not zeroed) 337 */ 338 .section ".pgtable","a",@nobits 339 .balign 4096 340pgtable: 341 .fill 6*4096, 1, 0 342