1/* 2 * linux/arch/arm/mm/proc-xsc3.S 3 * 4 * Original Author: Matthew Gilbert 5 * Current Maintainer: Deepak Saxena <dsaxena@plexity.net> 6 * 7 * Copyright 2004 (C) Intel Corp. 8 * Copyright 2005 (c) MontaVista Software, Inc. 9 * 10 * This program is free software; you can redistribute it and/or modify 11 * it under the terms of the GNU General Public License version 2 as 12 * published by the Free Software Foundation. 13 * 14 * MMU functions for the Intel XScale3 Core (XSC3). The XSC3 core is an 15 * extension to Intel's original XScale core that adds the following 16 * features: 17 * 18 * - ARMv6 Supersections 19 * - Low Locality Reference pages (replaces mini-cache) 20 * - 36-bit addressing 21 * - L2 cache 22 * - Cache-coherency if chipset supports it 23 * 24 * Based on orignal XScale code by Nicolas Pitre 25 */ 26 27#include <linux/linkage.h> 28#include <linux/init.h> 29#include <asm/assembler.h> 30#include <asm/elf.h> 31#include <asm/hardware.h> 32#include <asm/pgtable.h> 33#include <asm/pgtable-hwdef.h> 34#include <asm/page.h> 35#include <asm/ptrace.h> 36#include "proc-macros.S" 37 38/* 39 * This is the maximum size of an area which will be flushed. If the 40 * area is larger than this, then we flush the whole cache. 41 */ 42#define MAX_AREA_SIZE 32768 43 44/* 45 * The cache line size of the I and D cache. 46 */ 47#define CACHELINESIZE 32 48 49/* 50 * The size of the data cache. 51 */ 52#define CACHESIZE 32768 53 54/* 55 * Run with L2 enabled. 56 */ 57#define L2_CACHE_ENABLE 1 58 59/* 60 * This macro is used to wait for a CP15 write and is needed 61 * when we have to ensure that the last operation to the co-pro 62 * was completed before continuing with operation. 63 */ 64 .macro cpwait_ret, lr, rd 65 mrc p15, 0, \rd, c2, c0, 0 @ arbitrary read of cp15 66 sub pc, \lr, \rd, LSR #32 @ wait for completion and 67 @ flush instruction pipeline 68 .endm 69 70/* 71 * This macro cleans & invalidates the entire xsc3 dcache by set & way. 72 */ 73 74 .macro clean_d_cache rd, rs 75 mov \rd, #0x1f00 76 orr \rd, \rd, #0x00e0 771: mcr p15, 0, \rd, c7, c14, 2 @ clean/inv set/way 78 adds \rd, \rd, #0x40000000 79 bcc 1b 80 subs \rd, \rd, #0x20 81 bpl 1b 82 .endm 83 84 .text 85 86/* 87 * cpu_xsc3_proc_init() 88 * 89 * Nothing too exciting at the moment 90 */ 91ENTRY(cpu_xsc3_proc_init) 92 mov pc, lr 93 94/* 95 * cpu_xsc3_proc_fin() 96 */ 97ENTRY(cpu_xsc3_proc_fin) 98 str lr, [sp, #-4]! 99 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE 100 msr cpsr_c, r0 101 bl xsc3_flush_kern_cache_all @ clean caches 102 mrc p15, 0, r0, c1, c0, 0 @ ctrl register 103 bic r0, r0, #0x1800 @ ...IZ........... 104 bic r0, r0, #0x0006 @ .............CA. 105 mcr p15, 0, r0, c1, c0, 0 @ disable caches 106 ldr pc, [sp], #4 107 108/* 109 * cpu_xsc3_reset(loc) 110 * 111 * Perform a soft reset of the system. Put the CPU into the 112 * same state as it would be if it had been reset, and branch 113 * to what would be the reset vector. 114 * 115 * loc: location to jump to for soft reset 116 */ 117 .align 5 118ENTRY(cpu_xsc3_reset) 119 mov r1, #PSR_F_BIT|PSR_I_BIT|SVC_MODE 120 msr cpsr_c, r1 @ reset CPSR 121 mrc p15, 0, r1, c1, c0, 0 @ ctrl register 122 bic r1, r1, #0x0086 @ ........B....CA. 123 bic r1, r1, #0x3900 @ ..VIZ..S........ 124 mcr p15, 0, r1, c1, c0, 0 @ ctrl register 125 mcr p15, 0, ip, c7, c7, 0 @ invalidate I,D caches & BTB 126 bic r1, r1, #0x0001 @ ...............M 127 mcr p15, 0, r1, c1, c0, 0 @ ctrl register 128 @ CAUTION: MMU turned off from this point. We count on the pipeline 129 @ already containing those two last instructions to survive. 130 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs 131 mov pc, r0 132 133/* 134 * cpu_xsc3_do_idle() 135 * 136 * Cause the processor to idle 137 * 138 * For now we do nothing but go to idle mode for every case 139 * 140 * XScale supports clock switching, but using idle mode support 141 * allows external hardware to react to system state changes. 142 143 MMG: Come back to this one. 144 */ 145 .align 5 146 147ENTRY(cpu_xsc3_do_idle) 148 mov r0, #1 149 mcr p14, 0, r0, c7, c0, 0 @ Go to IDLE 150 mov pc, lr 151 152/* ================================= CACHE ================================ */ 153 154/* 155 * flush_user_cache_all() 156 * 157 * Invalidate all cache entries in a particular address 158 * space. 159 */ 160ENTRY(xsc3_flush_user_cache_all) 161 /* FALLTHROUGH */ 162 163/* 164 * flush_kern_cache_all() 165 * 166 * Clean and invalidate the entire cache. 167 */ 168ENTRY(xsc3_flush_kern_cache_all) 169 mov r2, #VM_EXEC 170 mov ip, #0 171__flush_whole_cache: 172 clean_d_cache r0, r1 173 tst r2, #VM_EXEC 174 mcrne p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB 175 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write Buffer 176 mcrne p15, 0, ip, c7, c5, 4 @ Prefetch Flush 177 mov pc, lr 178 179/* 180 * flush_user_cache_range(start, end, vm_flags) 181 * 182 * Invalidate a range of cache entries in the specified 183 * address space. 184 * 185 * - start - start address (may not be aligned) 186 * - end - end address (exclusive, may not be aligned) 187 * - vma - vma_area_struct describing address space 188 */ 189 .align 5 190ENTRY(xsc3_flush_user_cache_range) 191 mov ip, #0 192 sub r3, r1, r0 @ calculate total size 193 cmp r3, #MAX_AREA_SIZE 194 bhs __flush_whole_cache 195 1961: tst r2, #VM_EXEC 197 mcrne p15, 0, r0, c7, c5, 1 @ Invalidate I cache line 198 mcr p15, 0, r0, c7, c14, 1 @ Clean/invalidate D cache line 199 add r0, r0, #CACHELINESIZE 200 cmp r0, r1 201 blo 1b 202 tst r2, #VM_EXEC 203 mcrne p15, 0, ip, c7, c5, 6 @ Invalidate BTB 204 mcrne p15, 0, ip, c7, c10, 4 @ Drain Write Buffer 205 mcrne p15, 0, ip, c7, c5, 4 @ Prefetch Flush 206 mov pc, lr 207 208/* 209 * coherent_kern_range(start, end) 210 * 211 * Ensure coherency between the Icache and the Dcache in the 212 * region described by start. If you have non-snooping 213 * Harvard caches, you need to implement this function. 214 * 215 * - start - virtual start address 216 * - end - virtual end address 217 * 218 * Note: single I-cache line invalidation isn't used here since 219 * it also trashes the mini I-cache used by JTAG debuggers. 220 */ 221ENTRY(xsc3_coherent_kern_range) 222/* FALLTHROUGH */ 223ENTRY(xsc3_coherent_user_range) 224 bic r0, r0, #CACHELINESIZE - 1 2251: mcr p15, 0, r0, c7, c10, 1 @ clean D entry 226 add r0, r0, #CACHELINESIZE 227 cmp r0, r1 228 blo 1b 229 mov r0, #0 230 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB 231 mcr p15, 0, r0, c7, c10, 4 @ Drain Write Buffer 232 mcr p15, 0, r0, c7, c5, 4 @ Prefetch Flush 233 mov pc, lr 234 235/* 236 * flush_kern_dcache_page(void *page) 237 * 238 * Ensure no D cache aliasing occurs, either with itself or 239 * the I cache 240 * 241 * - addr - page aligned address 242 */ 243ENTRY(xsc3_flush_kern_dcache_page) 244 add r1, r0, #PAGE_SZ 2451: mcr p15, 0, r0, c7, c14, 1 @ Clean/Invalidate D Cache line 246 add r0, r0, #CACHELINESIZE 247 cmp r0, r1 248 blo 1b 249 mov r0, #0 250 mcr p15, 0, r0, c7, c5, 0 @ Invalidate I cache & BTB 251 mcr p15, 0, r0, c7, c10, 4 @ Drain Write Buffer 252 mcr p15, 0, r0, c7, c5, 4 @ Prefetch Flush 253 mov pc, lr 254 255/* 256 * dma_inv_range(start, end) 257 * 258 * Invalidate (discard) the specified virtual address range. 259 * May not write back any entries. If 'start' or 'end' 260 * are not cache line aligned, those lines must be written 261 * back. 262 * 263 * - start - virtual start address 264 * - end - virtual end address 265 */ 266ENTRY(xsc3_dma_inv_range) 267 tst r0, #CACHELINESIZE - 1 268 bic r0, r0, #CACHELINESIZE - 1 269 mcrne p15, 0, r0, c7, c10, 1 @ clean L1 D entry 270 mcrne p15, 1, r0, c7, c11, 1 @ clean L2 D entry 271 tst r1, #CACHELINESIZE - 1 272 mcrne p15, 0, r1, c7, c10, 1 @ clean L1 D entry 273 mcrne p15, 1, r1, c7, c11, 1 @ clean L2 D entry 2741: mcr p15, 0, r0, c7, c6, 1 @ invalidate L1 D entry 275 mcr p15, 1, r0, c7, c7, 1 @ Invalidate L2 D cache line 276 add r0, r0, #CACHELINESIZE 277 cmp r0, r1 278 blo 1b 279 mcr p15, 0, r0, c7, c10, 4 @ Drain Write Buffer 280 mov pc, lr 281 282/* 283 * dma_clean_range(start, end) 284 * 285 * Clean the specified virtual address range. 286 * 287 * - start - virtual start address 288 * - end - virtual end address 289 */ 290ENTRY(xsc3_dma_clean_range) 291 bic r0, r0, #CACHELINESIZE - 1 2921: mcr p15, 0, r0, c7, c10, 1 @ clean L1 D entry 293 mcr p15, 1, r0, c7, c11, 1 @ clean L2 D entry 294 add r0, r0, #CACHELINESIZE 295 cmp r0, r1 296 blo 1b 297 mcr p15, 0, r0, c7, c10, 4 @ Drain Write Buffer 298 mov pc, lr 299 300/* 301 * dma_flush_range(start, end) 302 * 303 * Clean and invalidate the specified virtual address range. 304 * 305 * - start - virtual start address 306 * - end - virtual end address 307 */ 308ENTRY(xsc3_dma_flush_range) 309 bic r0, r0, #CACHELINESIZE - 1 3101: mcr p15, 0, r0, c7, c14, 1 @ Clean/invalidate L1 D cache line 311 mcr p15, 1, r0, c7, c11, 1 @ Clean L2 D cache line 312 mcr p15, 1, r0, c7, c7, 1 @ Invalidate L2 D cache line 313 add r0, r0, #CACHELINESIZE 314 cmp r0, r1 315 blo 1b 316 mcr p15, 0, r0, c7, c10, 4 @ Drain Write Buffer 317 mov pc, lr 318 319ENTRY(xsc3_cache_fns) 320 .long xsc3_flush_kern_cache_all 321 .long xsc3_flush_user_cache_all 322 .long xsc3_flush_user_cache_range 323 .long xsc3_coherent_kern_range 324 .long xsc3_coherent_user_range 325 .long xsc3_flush_kern_dcache_page 326 .long xsc3_dma_inv_range 327 .long xsc3_dma_clean_range 328 .long xsc3_dma_flush_range 329 330ENTRY(cpu_xsc3_dcache_clean_area) 3311: mcr p15, 0, r0, c7, c10, 1 @ clean D entry 332 add r0, r0, #CACHELINESIZE 333 subs r1, r1, #CACHELINESIZE 334 bhi 1b 335 mov pc, lr 336 337/* =============================== PageTable ============================== */ 338 339/* 340 * cpu_xsc3_switch_mm(pgd) 341 * 342 * Set the translation base pointer to be as described by pgd. 343 * 344 * pgd: new page tables 345 */ 346 .align 5 347ENTRY(cpu_xsc3_switch_mm) 348 clean_d_cache r1, r2 349 mcr p15, 0, ip, c7, c5, 0 @ Invalidate I cache & BTB 350 mcr p15, 0, ip, c7, c10, 4 @ Drain Write Buffer 351 mcr p15, 0, ip, c7, c5, 4 @ Prefetch Flush 352#ifdef L2_CACHE_ENABLE 353 orr r0, r0, #0x18 @ cache the page table in L2 354#endif 355 mcr p15, 0, r0, c2, c0, 0 @ load page table pointer 356 mcr p15, 0, ip, c8, c7, 0 @ invalidate I & D TLBs 357 cpwait_ret lr, ip 358 359/* 360 * cpu_xsc3_set_pte_ext(ptep, pte, ext) 361 * 362 * Set a PTE and flush it out 363 * 364 */ 365 .align 5 366ENTRY(cpu_xsc3_set_pte_ext) 367 str r1, [r0], #-2048 @ linux version 368 369 bic r2, r1, #0xff0 @ Keep C, B bits 370 orr r2, r2, #PTE_TYPE_EXT @ extended page 371 tst r1, #L_PTE_SHARED @ Shared? 372 orrne r2, r2, #0x200 373 374 eor r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY 375 376 tst r3, #L_PTE_USER @ User? 377 orrne r2, r2, #PTE_EXT_AP_URO_SRW @ yes -> user r/o, system r/w 378 379 tst r3, #L_PTE_WRITE | L_PTE_DIRTY @ Write and Dirty? 380 orreq r2, r2, #PTE_EXT_AP_UNO_SRW @ yes -> user n/a, system r/w 381 @ combined with user -> user r/w 382 383#if L2_CACHE_ENABLE 384 @ If its cacheable it needs to be in L2 also. 385 eor ip, r1, #L_PTE_CACHEABLE 386 tst ip, #L_PTE_CACHEABLE 387 orreq r2, r2, #PTE_EXT_TEX(0x5) 388#endif 389 390 tst r3, #L_PTE_PRESENT | L_PTE_YOUNG @ Present and Young? 391 movne r2, #0 @ no -> fault 392 393 str r2, [r0] @ hardware version 394 mov ip, #0 395 mcr p15, 0, r0, c7, c10, 1 @ Clean D cache line mcr 396 mcr p15, 0, ip, c7, c10, 4 @ Drain Write Buffer 397 mov pc, lr 398 399 .ltorg 400 401 .align 402 403 __INIT 404 405 .type __xsc3_setup, #function 406__xsc3_setup: 407 mov r0, #PSR_F_BIT|PSR_I_BIT|SVC_MODE 408 msr cpsr_c, r0 409 mcr p15, 0, ip, c7, c7, 0 @ invalidate I, D caches & BTB 410 mcr p15, 0, ip, c7, c10, 4 @ Drain Write Buffer 411 mcr p15, 0, ip, c7, c5, 4 @ Prefetch Flush 412 mcr p15, 0, ip, c8, c7, 0 @ invalidate I, D TLBs 413#if L2_CACHE_ENABLE 414 orr r4, r4, #0x18 @ cache the page table in L2 415#endif 416 mcr p15, 0, r4, c2, c0, 0 @ load page table pointer 417 mov r0, #1 @ Allow access to CP0 and CP13 418 orr r0, r0, #1 << 13 @ Its undefined whether this 419 mcr p15, 0, r0, c15, c1, 0 @ affects USR or SVC modes 420 mrc p15, 0, r0, c1, c0, 1 @ get auxiliary control reg 421 and r0, r0, #2 @ preserve bit P bit setting 422#if L2_CACHE_ENABLE 423 orr r0, r0, #(1 << 10) @ enable L2 for LLR cache 424#endif 425 mcr p15, 0, r0, c1, c0, 1 @ set auxiliary control reg 426 427 adr r5, xsc3_crval 428 ldmia r5, {r5, r6} 429 mrc p15, 0, r0, c1, c0, 0 @ get control register 430 bic r0, r0, r5 @ .... .... .... ..A. 431 orr r0, r0, r6 @ .... .... .... .C.M 432 orr r0, r0, #0x00000800 @ ..VI Z..S .... .... 433#if L2_CACHE_ENABLE 434 orr r0, r0, #0x04000000 @ L2 enable 435#endif 436 mov pc, lr 437 438 .size __xsc3_setup, . - __xsc3_setup 439 440 .type xsc3_crval, #object 441xsc3_crval: 442 crval clear=0x04003b02, mmuset=0x00003105, ucset=0x00001100 443 444 __INITDATA 445 446/* 447 * Purpose : Function pointers used to access above functions - all calls 448 * come through these 449 */ 450 451 .type xsc3_processor_functions, #object 452ENTRY(xsc3_processor_functions) 453 .word v5t_early_abort 454 .word cpu_xsc3_proc_init 455 .word cpu_xsc3_proc_fin 456 .word cpu_xsc3_reset 457 .word cpu_xsc3_do_idle 458 .word cpu_xsc3_dcache_clean_area 459 .word cpu_xsc3_switch_mm 460 .word cpu_xsc3_set_pte_ext 461 .size xsc3_processor_functions, . - xsc3_processor_functions 462 463 .section ".rodata" 464 465 .type cpu_arch_name, #object 466cpu_arch_name: 467 .asciz "armv5te" 468 .size cpu_arch_name, . - cpu_arch_name 469 470 .type cpu_elf_name, #object 471cpu_elf_name: 472 .asciz "v5" 473 .size cpu_elf_name, . - cpu_elf_name 474 475 .type cpu_xsc3_name, #object 476cpu_xsc3_name: 477 .asciz "XScale-Core3" 478 .size cpu_xsc3_name, . - cpu_xsc3_name 479 480 .align 481 482 .section ".proc.info.init", #alloc, #execinstr 483 484 .type __xsc3_proc_info,#object 485__xsc3_proc_info: 486 .long 0x69056000 487 .long 0xffffe000 488 .long PMD_TYPE_SECT | \ 489 PMD_SECT_BUFFERABLE | \ 490 PMD_SECT_CACHEABLE | \ 491 PMD_SECT_AP_WRITE | \ 492 PMD_SECT_AP_READ 493 .long PMD_TYPE_SECT | \ 494 PMD_SECT_AP_WRITE | \ 495 PMD_SECT_AP_READ 496 b __xsc3_setup 497 .long cpu_arch_name 498 .long cpu_elf_name 499 .long HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP 500 .long cpu_xsc3_name 501 .long xsc3_processor_functions 502 .long v4wbi_tlb_fns 503 .long xsc3_mc_user_fns 504 .long xsc3_cache_fns 505 .size __xsc3_proc_info, . - __xsc3_proc_info 506