Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/boot/head.S |
| 3 | * |
| 4 | * Copyright (C) 1991, 1992, 1993 Linus Torvalds |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | /* |
Domen Puncer | f454944 | 2005-06-25 14:58:59 -0700 | [diff] [blame] | 22 | * High loaded stuff by Hans Lermen & Werner Almesberger, Feb. 1996 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 23 | */ |
| 24 | .code32 |
| 25 | .text |
| 26 | |
| 27 | #include <linux/linkage.h> |
| 28 | #include <asm/segment.h> |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 29 | #include <asm/pgtable.h> |
Eric W. Biederman | d053750 | 2005-06-25 14:57:52 -0700 | [diff] [blame] | 30 | #include <asm/page.h> |
Alexander van Heukelum | 7c53976 | 2008-04-08 12:54:30 +0200 | [diff] [blame] | 31 | #include <asm/boot.h> |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 32 | #include <asm/msr.h> |
Cyrill Gorcunov | e83e31f4 | 2008-05-12 15:43:39 +0200 | [diff] [blame] | 33 | #include <asm/processor-flags.h> |
Eric W. Biederman | bd53147 | 2007-10-26 11:29:04 -0600 | [diff] [blame] | 34 | #include <asm/asm-offsets.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 36 | .section ".text.head" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | .code32 |
| 38 | .globl startup_32 |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 39 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | startup_32: |
| 41 | cld |
Eric W. Biederman | bd53147 | 2007-10-26 11:29:04 -0600 | [diff] [blame] | 42 | /* test KEEP_SEGMENTS flag to see if the bootloader is asking |
| 43 | * us to not reload segments */ |
| 44 | testb $(1<<6), BP_loadflags(%esi) |
| 45 | jnz 1f |
| 46 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | cli |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 48 | movl $(__KERNEL_DS), %eax |
| 49 | movl %eax, %ds |
| 50 | movl %eax, %es |
| 51 | movl %eax, %ss |
Eric W. Biederman | bd53147 | 2007-10-26 11:29:04 -0600 | [diff] [blame] | 52 | 1: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 54 | /* Calculate the delta between where we were compiled to run |
| 55 | * at and where we were actually loaded at. This can only be done |
| 56 | * with a short local call on x86. Nothing else will tell us what |
| 57 | * address we are running at. The reserved chunk of the real-mode |
H. Peter Anvin | 85414b6 | 2007-07-11 12:18:33 -0700 | [diff] [blame] | 58 | * data at 0x1e4 (defined as a scratch field) are used as the stack |
| 59 | * for this calculation. Only 4 bytes are needed. |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 60 | */ |
H. Peter Anvin | 85414b6 | 2007-07-11 12:18:33 -0700 | [diff] [blame] | 61 | leal (0x1e4+4)(%esi), %esp |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 62 | call 1f |
| 63 | 1: popl %ebp |
| 64 | subl $1b, %ebp |
| 65 | |
Vivek Goyal | a4831e0 | 2007-05-02 19:27:08 +0200 | [diff] [blame] | 66 | /* setup a stack and make sure cpu supports long mode. */ |
Alexander van Heukelum | 7c53976 | 2008-04-08 12:54:30 +0200 | [diff] [blame] | 67 | movl $boot_stack_end, %eax |
Vivek Goyal | a4831e0 | 2007-05-02 19:27:08 +0200 | [diff] [blame] | 68 | addl %ebp, %eax |
| 69 | movl %eax, %esp |
| 70 | |
| 71 | call verify_cpu |
| 72 | testl %eax, %eax |
| 73 | jnz no_longmode |
| 74 | |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 75 | /* Compute the delta between where we were compiled to run at |
| 76 | * and where the code will actually run at. |
| 77 | */ |
| 78 | /* %ebp contains the address we are loaded at by the boot loader and %ebx |
| 79 | * contains the address where we should move the kernel image temporarily |
| 80 | * for safe in-place decompression. |
| 81 | */ |
| 82 | |
| 83 | #ifdef CONFIG_RELOCATABLE |
| 84 | movl %ebp, %ebx |
Andi Kleen | 31422c5 | 2008-02-04 16:48:08 +0100 | [diff] [blame] | 85 | addl $(PMD_PAGE_SIZE -1), %ebx |
| 86 | andl $PMD_PAGE_MASK, %ebx |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 87 | #else |
| 88 | movl $CONFIG_PHYSICAL_START, %ebx |
| 89 | #endif |
| 90 | |
| 91 | /* Replace the compressed data size with the uncompressed size */ |
| 92 | subl input_len(%ebp), %ebx |
| 93 | movl output_len(%ebp), %eax |
| 94 | addl %eax, %ebx |
| 95 | /* Add 8 bytes for every 32K input block */ |
| 96 | shrl $12, %eax |
| 97 | addl %eax, %ebx |
| 98 | /* Add 32K + 18 bytes of extra slack and align on a 4K boundary */ |
| 99 | addl $(32768 + 18 + 4095), %ebx |
| 100 | andl $~4095, %ebx |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 101 | |
| 102 | /* |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 103 | * Prepare for entering 64 bit mode |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 104 | */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 105 | |
| 106 | /* Load new GDT with the 64bit segments using 32bit descriptor */ |
| 107 | leal gdt(%ebp), %eax |
| 108 | movl %eax, gdt+2(%ebp) |
| 109 | lgdt gdt(%ebp) |
| 110 | |
| 111 | /* Enable PAE mode */ |
| 112 | xorl %eax, %eax |
Cyrill Gorcunov | e83e31f4 | 2008-05-12 15:43:39 +0200 | [diff] [blame] | 113 | orl $(X86_CR4_PAE), %eax |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 114 | movl %eax, %cr4 |
| 115 | |
| 116 | /* |
| 117 | * Build early 4G boot pagetable |
| 118 | */ |
| 119 | /* Initialize Page tables to 0*/ |
| 120 | leal pgtable(%ebx), %edi |
| 121 | xorl %eax, %eax |
| 122 | movl $((4096*6)/4), %ecx |
| 123 | rep stosl |
| 124 | |
| 125 | /* Build Level 4 */ |
| 126 | leal pgtable + 0(%ebx), %edi |
| 127 | leal 0x1007 (%edi), %eax |
| 128 | movl %eax, 0(%edi) |
| 129 | |
| 130 | /* Build Level 3 */ |
| 131 | leal pgtable + 0x1000(%ebx), %edi |
| 132 | leal 0x1007(%edi), %eax |
| 133 | movl $4, %ecx |
| 134 | 1: movl %eax, 0x00(%edi) |
| 135 | addl $0x00001000, %eax |
| 136 | addl $8, %edi |
| 137 | decl %ecx |
| 138 | jnz 1b |
| 139 | |
| 140 | /* Build Level 2 */ |
| 141 | leal pgtable + 0x2000(%ebx), %edi |
| 142 | movl $0x00000183, %eax |
| 143 | movl $2048, %ecx |
| 144 | 1: movl %eax, 0(%edi) |
| 145 | addl $0x00200000, %eax |
| 146 | addl $8, %edi |
| 147 | decl %ecx |
| 148 | jnz 1b |
| 149 | |
| 150 | /* Enable the boot page tables */ |
| 151 | leal pgtable(%ebx), %eax |
| 152 | movl %eax, %cr3 |
| 153 | |
| 154 | /* Enable Long mode in EFER (Extended Feature Enable Register) */ |
| 155 | movl $MSR_EFER, %ecx |
| 156 | rdmsr |
| 157 | btsl $_EFER_LME, %eax |
| 158 | wrmsr |
| 159 | |
| 160 | /* Setup for the jump to 64bit mode |
| 161 | * |
| 162 | * When the jump is performend we will be in long mode but |
| 163 | * in 32bit compatibility mode with EFER.LME = 1, CS.L = 0, CS.D = 1 |
| 164 | * (and in turn EFER.LMA = 1). To jump into 64bit mode we use |
| 165 | * the new gdt/idt that has __KERNEL_CS with CS.L = 1. |
| 166 | * We place all of the values on our mini stack so lret can |
| 167 | * used to perform that far jump. |
| 168 | */ |
| 169 | pushl $__KERNEL_CS |
| 170 | leal startup_64(%ebp), %eax |
| 171 | pushl %eax |
| 172 | |
| 173 | /* Enter paged protected Mode, activating Long Mode */ |
Cyrill Gorcunov | e83e31f4 | 2008-05-12 15:43:39 +0200 | [diff] [blame] | 174 | movl $(X86_CR0_PG | X86_CR0_PE), %eax /* Enable Paging and Protected mode */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 175 | movl %eax, %cr0 |
| 176 | |
| 177 | /* Jump from 32bit compatibility mode into 64bit mode. */ |
| 178 | lret |
| 179 | |
Vivek Goyal | a4831e0 | 2007-05-02 19:27:08 +0200 | [diff] [blame] | 180 | no_longmode: |
| 181 | /* This isn't an x86-64 CPU so hang */ |
| 182 | 1: |
| 183 | hlt |
| 184 | jmp 1b |
| 185 | |
Thomas Gleixner | 250c227 | 2007-10-11 11:17:24 +0200 | [diff] [blame] | 186 | #include "../../kernel/verify_cpu_64.S" |
Vivek Goyal | a4831e0 | 2007-05-02 19:27:08 +0200 | [diff] [blame] | 187 | |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 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 |
Vivek Goyal | a4831e0 | 2007-05-02 19:27:08 +0200 | [diff] [blame] | 194 | .org 0x200 |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 195 | ENTRY(startup_64) |
| 196 | /* We come here either from startup_32 or directly from a |
| 197 | * 64bit bootloader. If we come here from a bootloader we depend on |
| 198 | * an identity mapped page table being provied that maps our |
| 199 | * entire text+data+bss and hopefully all of memory. |
| 200 | */ |
| 201 | |
| 202 | /* Setup data segments. */ |
| 203 | xorl %eax, %eax |
| 204 | movl %eax, %ds |
| 205 | movl %eax, %es |
| 206 | movl %eax, %ss |
Zachary Amsden | 08da5a2 | 2007-08-10 22:31:05 +0200 | [diff] [blame] | 207 | movl %eax, %fs |
| 208 | movl %eax, %gs |
| 209 | lldt %ax |
| 210 | movl $0x20, %eax |
| 211 | ltr %ax |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 212 | |
| 213 | /* Compute the decompressed kernel start address. It is where |
| 214 | * we were loaded at aligned to a 2M boundary. %rbp contains the |
| 215 | * decompressed kernel start address. |
| 216 | * |
| 217 | * If it is a relocatable kernel then decompress and run the kernel |
| 218 | * from load address aligned to 2MB addr, otherwise decompress and |
| 219 | * run the kernel from CONFIG_PHYSICAL_START |
| 220 | */ |
| 221 | |
| 222 | /* Start with the delta to where the kernel will run at. */ |
| 223 | #ifdef CONFIG_RELOCATABLE |
| 224 | leaq startup_32(%rip) /* - $startup_32 */, %rbp |
Andi Kleen | 31422c5 | 2008-02-04 16:48:08 +0100 | [diff] [blame] | 225 | addq $(PMD_PAGE_SIZE - 1), %rbp |
| 226 | andq $PMD_PAGE_MASK, %rbp |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 227 | movq %rbp, %rbx |
| 228 | #else |
| 229 | movq $CONFIG_PHYSICAL_START, %rbp |
| 230 | movq %rbp, %rbx |
| 231 | #endif |
| 232 | |
| 233 | /* Replace the compressed data size with the uncompressed size */ |
| 234 | movl input_len(%rip), %eax |
| 235 | subq %rax, %rbx |
| 236 | movl output_len(%rip), %eax |
| 237 | addq %rax, %rbx |
| 238 | /* Add 8 bytes for every 32K input block */ |
| 239 | shrq $12, %rax |
| 240 | addq %rax, %rbx |
| 241 | /* Add 32K + 18 bytes of extra slack and align on a 4K boundary */ |
| 242 | addq $(32768 + 18 + 4095), %rbx |
| 243 | andq $~4095, %rbx |
| 244 | |
| 245 | /* Copy the compressed kernel to the end of our buffer |
| 246 | * where decompression in place becomes safe. |
| 247 | */ |
Yinghai Lu | 4a9f54c | 2008-04-10 15:06:38 -0700 | [diff] [blame] | 248 | leaq _end_before_pgt(%rip), %r8 |
| 249 | leaq _end_before_pgt(%rbx), %r9 |
| 250 | movq $_end_before_pgt /* - $startup_32 */, %rcx |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 251 | 1: subq $8, %r8 |
| 252 | subq $8, %r9 |
| 253 | movq 0(%r8), %rax |
| 254 | movq %rax, 0(%r9) |
| 255 | subq $8, %rcx |
| 256 | jnz 1b |
| 257 | |
| 258 | /* |
| 259 | * Jump to the relocated address. |
| 260 | */ |
| 261 | leaq relocated(%rbx), %rax |
| 262 | jmp *%rax |
| 263 | |
| 264 | .section ".text" |
| 265 | relocated: |
| 266 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 267 | /* |
| 268 | * Clear BSS |
| 269 | */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 270 | xorq %rax, %rax |
| 271 | leaq _edata(%rbx), %rdi |
Yinghai Lu | 4a9f54c | 2008-04-10 15:06:38 -0700 | [diff] [blame] | 272 | leaq _end_before_pgt(%rbx), %rcx |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 273 | subq %rdi, %rcx |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 274 | cld |
| 275 | rep |
| 276 | stosb |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 277 | |
| 278 | /* Setup the stack */ |
Alexander van Heukelum | 7c53976 | 2008-04-08 12:54:30 +0200 | [diff] [blame] | 279 | leaq boot_stack_end(%rip), %rsp |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 280 | |
| 281 | /* zero EFLAGS after setting rsp */ |
| 282 | pushq $0 |
| 283 | popfq |
| 284 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 285 | /* |
| 286 | * Do the decompression, and jump to the new kernel.. |
| 287 | */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 288 | pushq %rsi # Save the real mode argument |
| 289 | movq %rsi, %rdi # real mode address |
Alexander van Heukelum | 7c53976 | 2008-04-08 12:54:30 +0200 | [diff] [blame] | 290 | leaq boot_heap(%rip), %rsi # malloc area for uncompression |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 291 | leaq input_data(%rip), %rdx # input_data |
| 292 | movl input_len(%rip), %eax |
| 293 | movq %rax, %rcx # input_len |
| 294 | movq %rbp, %r8 # output |
| 295 | call decompress_kernel |
| 296 | popq %rsi |
| 297 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 298 | |
| 299 | /* |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 300 | * Jump to the decompressed kernel. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 301 | */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 302 | jmp *%rbp |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 303 | |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 304 | .data |
| 305 | gdt: |
| 306 | .word gdt_end - gdt |
| 307 | .long gdt |
| 308 | .word 0 |
| 309 | .quad 0x0000000000000000 /* NULL descriptor */ |
| 310 | .quad 0x00af9a000000ffff /* __KERNEL_CS */ |
| 311 | .quad 0x00cf92000000ffff /* __KERNEL_DS */ |
Zachary Amsden | 08da5a2 | 2007-08-10 22:31:05 +0200 | [diff] [blame] | 312 | .quad 0x0080890000000000 /* TS descriptor */ |
| 313 | .quad 0x0000000000000000 /* TS continued */ |
Vivek Goyal | 1ab60e0 | 2007-05-02 19:27:07 +0200 | [diff] [blame] | 314 | gdt_end: |
Alexander van Heukelum | 7c53976 | 2008-04-08 12:54:30 +0200 | [diff] [blame] | 315 | |
| 316 | .bss |
| 317 | /* Stack and heap for uncompression */ |
| 318 | .balign 4 |
| 319 | boot_heap: |
| 320 | .fill BOOT_HEAP_SIZE, 1, 0 |
| 321 | boot_stack: |
| 322 | .fill BOOT_STACK_SIZE, 1, 0 |
| 323 | boot_stack_end: |