1/* 2 * linux/arch/arm/kernel/head.S 3 * 4 * Copyright (C) 1994-2002 Russell King 5 * Copyright (c) 2003 ARM Limited 6 * All Rights Reserved 7 * 8 * This program is free software; you can redistribute it and/or modify 9 * it under the terms of the GNU General Public License version 2 as 10 * published by the Free Software Foundation. 11 * 12 * Kernel startup code for all 32-bit CPUs 13 */ 14#include <linux/linkage.h> 15#include <linux/init.h> 16 17#include <asm/assembler.h> 18#include <asm/domain.h> 19#include <asm/ptrace.h> 20#include <asm/asm-offsets.h> 21#include <asm/memory.h> 22#include <asm/thread_info.h> 23#include <asm/system.h> 24#include <asm/pgtable.h> 25 26#ifdef CONFIG_DEBUG_LL 27#include <mach/debug-macro.S> 28#endif 29 30/* 31 * swapper_pg_dir is the virtual address of the initial page table. 32 * We place the page tables 16K below KERNEL_RAM_VADDR. Therefore, we must 33 * make sure that KERNEL_RAM_VADDR is correctly set. Currently, we expect 34 * the least significant 16 bits to be 0x8000, but we could probably 35 * relax this restriction to KERNEL_RAM_VADDR >= PAGE_OFFSET + 0x4000. 36 */ 37#define KERNEL_RAM_VADDR (PAGE_OFFSET + TEXT_OFFSET) 38#if (KERNEL_RAM_VADDR & 0xffff) != 0x8000 39#error KERNEL_RAM_VADDR must start at 0xXXXX8000 40#endif 41 42#define PG_DIR_SIZE 0x4000 43#define PMD_ORDER 2 44 45 .globl swapper_pg_dir 46 .equ swapper_pg_dir, KERNEL_RAM_VADDR - PG_DIR_SIZE 47 48 .macro pgtbl, rd, phys 49 add \rd, \phys, #TEXT_OFFSET - PG_DIR_SIZE 50 .endm 51 52#ifdef CONFIG_XIP_KERNEL 53#define KERNEL_START XIP_VIRT_ADDR(CONFIG_XIP_PHYS_ADDR) 54#define KERNEL_END _edata_loc 55#else 56#define KERNEL_START KERNEL_RAM_VADDR 57#define KERNEL_END _end 58#endif 59 60/* 61 * Kernel startup entry point. 62 * --------------------------- 63 * 64 * This is normally called from the decompressor code. The requirements 65 * are: MMU = off, D-cache = off, I-cache = dont care, r0 = 0, 66 * r1 = machine nr, r2 = atags or dtb pointer. 67 * 68 * This code is mostly position independent, so if you link the kernel at 69 * 0xc0008000, you call this at __pa(0xc0008000). 70 * 71 * See linux/arch/arm/tools/mach-types for the complete list of machine 72 * numbers for r1. 73 * 74 * We're trying to keep crap to a minimum; DO NOT add any machine specific 75 * crap here - that's what the boot loader (or in extreme, well justified 76 * circumstances, zImage) is for. 77 */ 78 .arm 79 80 __HEAD 81ENTRY(stext) 82 83 THUMB( adr r9, BSYM(1f) ) @ Kernel is always entered in ARM. 84 THUMB( bx r9 ) @ If this is a Thumb-2 kernel, 85 THUMB( .thumb ) @ switch to Thumb now. 86 THUMB(1: ) 87 88 setmode PSR_F_BIT | PSR_I_BIT | SVC_MODE, r9 @ ensure svc mode 89 @ and irqs disabled 90 mrc p15, 0, r9, c0, c0 @ get processor id 91 bl __lookup_processor_type @ r5=procinfo r9=cpuid 92 movs r10, r5 @ invalid processor (r5=0)? 93 THUMB( it eq ) @ force fixup-able long branch encoding 94 beq __error_p @ yes, error 'p' 95 96#ifndef CONFIG_XIP_KERNEL 97 adr r3, 2f 98 ldmia r3, {r4, r8} 99 sub r4, r3, r4 @ (PHYS_OFFSET - PAGE_OFFSET) 100 add r8, r8, r4 @ PHYS_OFFSET 101#else 102 ldr r8, =PHYS_OFFSET @ always constant in this case 103#endif 104 105 /* 106 * r1 = machine no, r2 = atags or dtb, 107 * r8 = phys_offset, r9 = cpuid, r10 = procinfo 108 */ 109 bl __vet_atags 110#ifdef CONFIG_SMP_ON_UP 111 bl __fixup_smp 112#endif 113#ifdef CONFIG_ARM_PATCH_PHYS_VIRT 114 bl __fixup_pv_table 115#endif 116 bl __create_page_tables 117 118 /* 119 * The following calls CPU specific code in a position independent 120 * manner. See arch/arm/mm/proc-*.S for details. r10 = base of 121 * xxx_proc_info structure selected by __lookup_processor_type 122 * above. On return, the CPU will be ready for the MMU to be 123 * turned on, and r0 will hold the CPU control register value. 124 */ 125 ldr r13, =__mmap_switched @ address to jump to after 126 @ mmu has been enabled 127 adr lr, BSYM(1f) @ return (PIC) address 128 mov r8, r4 @ set TTBR1 to swapper_pg_dir 129 ARM( add pc, r10, #PROCINFO_INITFUNC ) 130 THUMB( add r12, r10, #PROCINFO_INITFUNC ) 131 THUMB( mov pc, r12 ) 1321: b __enable_mmu 133ENDPROC(stext) 134 .ltorg 135#ifndef CONFIG_XIP_KERNEL 1362: .long . 137 .long PAGE_OFFSET 138#endif 139 140/* 141 * Setup the initial page tables. We only setup the barest 142 * amount which are required to get the kernel running, which 143 * generally means mapping in the kernel code. 144 * 145 * r8 = phys_offset, r9 = cpuid, r10 = procinfo 146 * 147 * Returns: 148 * r0, r3, r5-r7 corrupted 149 * r4 = physical page table address 150 */ 151__create_page_tables: 152 pgtbl r4, r8 @ page table address 153 154 /* 155 * Clear the swapper page table 156 */ 157 mov r0, r4 158 mov r3, #0 159 add r6, r0, #PG_DIR_SIZE 1601: str r3, [r0], #4 161 str r3, [r0], #4 162 str r3, [r0], #4 163 str r3, [r0], #4 164 teq r0, r6 165 bne 1b 166 167 ldr r7, [r10, #PROCINFO_MM_MMUFLAGS] @ mm_mmuflags 168 169 /* 170 * Create identity mapping to cater for __enable_mmu. 171 * This identity mapping will be removed by paging_init(). 172 */ 173 adr r0, __enable_mmu_loc 174 ldmia r0, {r3, r5, r6} 175 sub r0, r0, r3 @ virt->phys offset 176 add r5, r5, r0 @ phys __enable_mmu 177 add r6, r6, r0 @ phys __enable_mmu_end 178 mov r5, r5, lsr #SECTION_SHIFT 179 mov r6, r6, lsr #SECTION_SHIFT 180 1811: orr r3, r7, r5, lsl #SECTION_SHIFT @ flags + kernel base 182 str r3, [r4, r5, lsl #PMD_ORDER] @ identity mapping 183 cmp r5, r6 184 addlo r5, r5, #1 @ next section 185 blo 1b 186 187 /* 188 * Now setup the pagetables for our kernel direct 189 * mapped region. 190 */ 191 mov r3, pc 192 mov r3, r3, lsr #SECTION_SHIFT 193 orr r3, r7, r3, lsl #SECTION_SHIFT 194 add r0, r4, #(KERNEL_START & 0xff000000) >> (SECTION_SHIFT - PMD_ORDER) 195 str r3, [r0, #((KERNEL_START & 0x00f00000) >> SECTION_SHIFT) << PMD_ORDER]! 196 ldr r6, =(KERNEL_END - 1) 197 add r0, r0, #1 << PMD_ORDER 198 add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER) 1991: cmp r0, r6 200 add r3, r3, #1 << SECTION_SHIFT 201 strls r3, [r0], #1 << PMD_ORDER 202 bls 1b 203 204#ifdef CONFIG_XIP_KERNEL 205 /* 206 * Map some ram to cover our .data and .bss areas. 207 */ 208 add r3, r8, #TEXT_OFFSET 209 orr r3, r3, r7 210 add r0, r4, #(KERNEL_RAM_VADDR & 0xff000000) >> (SECTION_SHIFT - PMD_ORDER) 211 str r3, [r0, #(KERNEL_RAM_VADDR & 0x00f00000) >> (SECTION_SHIFT - PMD_ORDER)]! 212 ldr r6, =(_end - 1) 213 add r0, r0, #4 214 add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER) 2151: cmp r0, r6 216 add r3, r3, #1 << 20 217 strls r3, [r0], #4 218 bls 1b 219#endif 220 221 /* 222 * Then map boot params address in r2 or 223 * the first 1MB of ram if boot params address is not specified. 224 */ 225 mov r0, r2, lsr #SECTION_SHIFT 226 movs r0, r0, lsl #SECTION_SHIFT 227 moveq r0, r8 228 sub r3, r0, r8 229 add r3, r3, #PAGE_OFFSET 230 add r3, r4, r3, lsr #(SECTION_SHIFT - PMD_ORDER) 231 orr r6, r7, r0 232 str r6, [r3] 233 234#ifdef CONFIG_DEBUG_LL 235#ifndef CONFIG_DEBUG_ICEDCC 236 /* 237 * Map in IO space for serial debugging. 238 * This allows debug messages to be output 239 * via a serial console before paging_init. 240 */ 241 addruart r7, r3, r0 242 243 mov r3, r3, lsr #SECTION_SHIFT 244 mov r3, r3, lsl #PMD_ORDER 245 246 add r0, r4, r3 247 rsb r3, r3, #0x4000 @ PTRS_PER_PGD*sizeof(long) 248 cmp r3, #0x0800 @ limit to 512MB 249 movhi r3, #0x0800 250 add r6, r0, r3 251 mov r3, r7, lsr #SECTION_SHIFT 252 ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags 253 orr r3, r7, r3, lsl #SECTION_SHIFT 2541: str r3, [r0], #4 255 add r3, r3, #1 << SECTION_SHIFT 256 cmp r0, r6 257 blo 1b 258 259#else /* CONFIG_DEBUG_ICEDCC */ 260 /* we don't need any serial debugging mappings for ICEDCC */ 261 ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags 262#endif /* !CONFIG_DEBUG_ICEDCC */ 263 264#if defined(CONFIG_ARCH_NETWINDER) || defined(CONFIG_ARCH_CATS) 265 /* 266 * If we're using the NetWinder or CATS, we also need to map 267 * in the 16550-type serial port for the debug messages 268 */ 269 add r0, r4, #0xff000000 >> (SECTION_SHIFT - PMD_ORDER) 270 orr r3, r7, #0x7c000000 271 str r3, [r0] 272#endif 273#ifdef CONFIG_ARCH_RPC 274 /* 275 * Map in screen at 0x02000000 & SCREEN2_BASE 276 * Similar reasons here - for debug. This is 277 * only for Acorn RiscPC architectures. 278 */ 279 add r0, r4, #0x02000000 >> (SECTION_SHIFT - PMD_ORDER) 280 orr r3, r7, #0x02000000 281 str r3, [r0] 282 add r0, r4, #0xd8000000 >> (SECTION_SHIFT - PMD_ORDER) 283 str r3, [r0] 284#endif 285#endif 286 mov pc, lr 287ENDPROC(__create_page_tables) 288 .ltorg 289 .align 290__enable_mmu_loc: 291 .long . 292 .long __enable_mmu 293 .long __enable_mmu_end 294 295#if defined(CONFIG_SMP) 296 __CPUINIT 297ENTRY(secondary_startup) 298 /* 299 * Common entry point for secondary CPUs. 300 * 301 * Ensure that we're in SVC mode, and IRQs are disabled. Lookup 302 * the processor type - there is no need to check the machine type 303 * as it has already been validated by the primary processor. 304 */ 305 setmode PSR_F_BIT | PSR_I_BIT | SVC_MODE, r9 306 mrc p15, 0, r9, c0, c0 @ get processor id 307 bl __lookup_processor_type 308 movs r10, r5 @ invalid processor? 309 moveq r0, #'p' @ yes, error 'p' 310 THUMB( it eq ) @ force fixup-able long branch encoding 311 beq __error_p 312 313 /* 314 * Use the page tables supplied from __cpu_up. 315 */ 316 adr r4, __secondary_data 317 ldmia r4, {r5, r7, r12} @ address to jump to after 318 sub lr, r4, r5 @ mmu has been enabled 319 ldr r4, [r7, lr] @ get secondary_data.pgdir 320 add r7, r7, #4 321 ldr r8, [r7, lr] @ get secondary_data.swapper_pg_dir 322 adr lr, BSYM(__enable_mmu) @ return address 323 mov r13, r12 @ __secondary_switched address 324 ARM( add pc, r10, #PROCINFO_INITFUNC ) @ initialise processor 325 @ (return control reg) 326 THUMB( add r12, r10, #PROCINFO_INITFUNC ) 327 THUMB( mov pc, r12 ) 328ENDPROC(secondary_startup) 329 330 /* 331 * r6 = &secondary_data 332 */ 333ENTRY(__secondary_switched) 334 ldr sp, [r7, #4] @ get secondary_data.stack 335 mov fp, #0 336 b secondary_start_kernel 337ENDPROC(__secondary_switched) 338 339 .align 340 341 .type __secondary_data, %object 342__secondary_data: 343 .long . 344 .long secondary_data 345 .long __secondary_switched 346#endif /* defined(CONFIG_SMP) */ 347 348 349 350/* 351 * Setup common bits before finally enabling the MMU. Essentially 352 * this is just loading the page table pointer and domain access 353 * registers. 354 * 355 * r0 = cp#15 control register 356 * r1 = machine ID 357 * r2 = atags or dtb pointer 358 * r4 = page table pointer 359 * r9 = processor ID 360 * r13 = *virtual* address to jump to upon completion 361 */ 362__enable_mmu: 363#ifdef CONFIG_ALIGNMENT_TRAP 364 orr r0, r0, #CR_A 365#else 366 bic r0, r0, #CR_A 367#endif 368#ifdef CONFIG_CPU_DCACHE_DISABLE 369 bic r0, r0, #CR_C 370#endif 371#ifdef CONFIG_CPU_BPREDICT_DISABLE 372 bic r0, r0, #CR_Z 373#endif 374#ifdef CONFIG_CPU_ICACHE_DISABLE 375 bic r0, r0, #CR_I 376#endif 377 mov r5, #(domain_val(DOMAIN_USER, DOMAIN_MANAGER) | \ 378 domain_val(DOMAIN_KERNEL, DOMAIN_MANAGER) | \ 379 domain_val(DOMAIN_TABLE, DOMAIN_MANAGER) | \ 380 domain_val(DOMAIN_IO, DOMAIN_CLIENT)) 381 mcr p15, 0, r5, c3, c0, 0 @ load domain access register 382 mcr p15, 0, r4, c2, c0, 0 @ load page table pointer 383 b __turn_mmu_on 384ENDPROC(__enable_mmu) 385 386/* 387 * Enable the MMU. This completely changes the structure of the visible 388 * memory space. You will not be able to trace execution through this. 389 * If you have an enquiry about this, *please* check the linux-arm-kernel 390 * mailing list archives BEFORE sending another post to the list. 391 * 392 * r0 = cp#15 control register 393 * r1 = machine ID 394 * r2 = atags or dtb pointer 395 * r9 = processor ID 396 * r13 = *virtual* address to jump to upon completion 397 * 398 * other registers depend on the function called upon completion 399 */ 400 .align 5 401__turn_mmu_on: 402 mov r0, r0 403 mcr p15, 0, r0, c1, c0, 0 @ write control reg 404 mrc p15, 0, r3, c0, c0, 0 @ read id reg 405 mov r3, r3 406 mov r3, r13 407 mov pc, r3 408__enable_mmu_end: 409ENDPROC(__turn_mmu_on) 410 411 412#ifdef CONFIG_SMP_ON_UP 413 __INIT 414__fixup_smp: 415 and r3, r9, #0x000f0000 @ architecture version 416 teq r3, #0x000f0000 @ CPU ID supported? 417 bne __fixup_smp_on_up @ no, assume UP 418 419 bic r3, r9, #0x00ff0000 420 bic r3, r3, #0x0000000f @ mask 0xff00fff0 421 mov r4, #0x41000000 422 orr r4, r4, #0x0000b000 423 orr r4, r4, #0x00000020 @ val 0x4100b020 424 teq r3, r4 @ ARM 11MPCore? 425 moveq pc, lr @ yes, assume SMP 426 427 mrc p15, 0, r0, c0, c0, 5 @ read MPIDR 428 and r0, r0, #0xc0000000 @ multiprocessing extensions and 429 teq r0, #0x80000000 @ not part of a uniprocessor system? 430 moveq pc, lr @ yes, assume SMP 431 432__fixup_smp_on_up: 433 adr r0, 1f 434 ldmia r0, {r3 - r5} 435 sub r3, r0, r3 436 add r4, r4, r3 437 add r5, r5, r3 438 b __do_fixup_smp_on_up 439ENDPROC(__fixup_smp) 440 441 .align 4421: .word . 443 .word __smpalt_begin 444 .word __smpalt_end 445 446 .pushsection .data 447 .globl smp_on_up 448smp_on_up: 449 ALT_SMP(.long 1) 450 ALT_UP(.long 0) 451 .popsection 452#endif 453 454 .text 455__do_fixup_smp_on_up: 456 cmp r4, r5 457 movhs pc, lr 458 ldmia r4!, {r0, r6} 459 ARM( str r6, [r0, r3] ) 460 THUMB( add r0, r0, r3 ) 461#ifdef __ARMEB__ 462 THUMB( mov r6, r6, ror #16 ) @ Convert word order for big-endian. 463#endif 464 THUMB( strh r6, [r0], #2 ) @ For Thumb-2, store as two halfwords 465 THUMB( mov r6, r6, lsr #16 ) @ to be robust against misaligned r3. 466 THUMB( strh r6, [r0] ) 467 b __do_fixup_smp_on_up 468ENDPROC(__do_fixup_smp_on_up) 469 470ENTRY(fixup_smp) 471 stmfd sp!, {r4 - r6, lr} 472 mov r4, r0 473 add r5, r0, r1 474 mov r3, #0 475 bl __do_fixup_smp_on_up 476 ldmfd sp!, {r4 - r6, pc} 477ENDPROC(fixup_smp) 478 479#ifdef CONFIG_ARM_PATCH_PHYS_VIRT 480 481/* __fixup_pv_table - patch the stub instructions with the delta between 482 * PHYS_OFFSET and PAGE_OFFSET, which is assumed to be 16MiB aligned and 483 * can be expressed by an immediate shifter operand. The stub instruction 484 * has a form of '(add|sub) rd, rn, #imm'. 485 */ 486 __HEAD 487__fixup_pv_table: 488 adr r0, 1f 489 ldmia r0, {r3-r5, r7} 490 sub r3, r0, r3 @ PHYS_OFFSET - PAGE_OFFSET 491 add r4, r4, r3 @ adjust table start address 492 add r5, r5, r3 @ adjust table end address 493 add r7, r7, r3 @ adjust __pv_phys_offset address 494 str r8, [r7] @ save computed PHYS_OFFSET to __pv_phys_offset 495 mov r6, r3, lsr #24 @ constant for add/sub instructions 496 teq r3, r6, lsl #24 @ must be 16MiB aligned 497THUMB( it ne @ cross section branch ) 498 bne __error 499 str r6, [r7, #4] @ save to __pv_offset 500 b __fixup_a_pv_table 501ENDPROC(__fixup_pv_table) 502 503 .align 5041: .long . 505 .long __pv_table_begin 506 .long __pv_table_end 5072: .long __pv_phys_offset 508 509 .text 510__fixup_a_pv_table: 511#ifdef CONFIG_THUMB2_KERNEL 512 lsls r6, #24 513 beq 2f 514 clz r7, r6 515 lsr r6, #24 516 lsl r6, r7 517 bic r6, #0x0080 518 lsrs r7, #1 519 orrcs r6, #0x0080 520 orr r6, r6, r7, lsl #12 521 orr r6, #0x4000 522 b 2f 5231: add r7, r3 524 ldrh ip, [r7, #2] 525 and ip, 0x8f00 526 orr ip, r6 @ mask in offset bits 31-24 527 strh ip, [r7, #2] 5282: cmp r4, r5 529 ldrcc r7, [r4], #4 @ use branch for delay slot 530 bcc 1b 531 bx lr 532#else 533 b 2f 5341: ldr ip, [r7, r3] 535 bic ip, ip, #0x000000ff 536 orr ip, ip, r6 @ mask in offset bits 31-24 537 str ip, [r7, r3] 5382: cmp r4, r5 539 ldrcc r7, [r4], #4 @ use branch for delay slot 540 bcc 1b 541 mov pc, lr 542#endif 543ENDPROC(__fixup_a_pv_table) 544 545ENTRY(fixup_pv_table) 546 stmfd sp!, {r4 - r7, lr} 547 ldr r2, 2f @ get address of __pv_phys_offset 548 mov r3, #0 @ no offset 549 mov r4, r0 @ r0 = table start 550 add r5, r0, r1 @ r1 = table size 551 ldr r6, [r2, #4] @ get __pv_offset 552 bl __fixup_a_pv_table 553 ldmfd sp!, {r4 - r7, pc} 554ENDPROC(fixup_pv_table) 555 556 .align 5572: .long __pv_phys_offset 558 559 .data 560 .globl __pv_phys_offset 561 .type __pv_phys_offset, %object 562__pv_phys_offset: 563 .long 0 564 .size __pv_phys_offset, . - __pv_phys_offset 565__pv_offset: 566 .long 0 567#endif 568 569#include "head-common.S" 570