lh | 9ed821d | 2023-04-07 01:36:19 -0700 | [diff] [blame] | 1 | /******************************************************************************* |
| 2 | * °æÈ¨ËùÓÐ (C)2016, ÖÐÐËͨѶ¹É·ÝÓÐÏÞ¹«Ë¾¡£ |
| 3 | * |
| 4 | * ÎļþÃû³Æ: head.S |
| 5 | * Îļþ±êʶ: head.S |
| 6 | * ÄÚÈÝÕªÒª: °æ±¾½âѹÆô¶¯´úÂë |
| 7 | * |
| 8 | * ÐÞ¸ÄÈÕÆÚ °æ±¾ºÅ Ð޸ıê¼Ç ÐÞ¸ÄÈË ÐÞ¸ÄÄÚÈÝ |
| 9 | * ------------------------------------------------------------------------------ |
| 10 | * 2016/09/12 V1.0 Create µËÄþÒ ´´½¨ |
| 11 | * |
| 12 | *******************************************************************************/ |
| 13 | |
| 14 | /******************************************************************************* |
| 15 | * Í·Îļþ * |
| 16 | *******************************************************************************/ |
| 17 | |
| 18 | /******************************************************************************* |
| 19 | * ºê¶¨Òå * |
| 20 | *******************************************************************************/ |
| 21 | #ifdef __thumb2__ |
| 22 | #define ARM(x...) |
| 23 | #define THUMB(x...) x |
| 24 | #define W(instr) instr.w |
| 25 | #else |
| 26 | #define ARM(x...) x |
| 27 | #define THUMB(x...) |
| 28 | #define W(instr) instr |
| 29 | #endif |
| 30 | #define ARM_BE8(x...) |
| 31 | |
| 32 | #define END(name) \ |
| 33 | .size name, .-name |
| 34 | |
| 35 | #define ENDPROC(name) \ |
| 36 | .type name, %function; \ |
| 37 | END(name) |
| 38 | |
| 39 | #define CYGOPT_HAL_ARM_MMU |
| 40 | #define CONFIG_CPU_CP15 |
| 41 | |
| 42 | #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH |
| 43 | #define CB_BITS 0x08 |
| 44 | #else |
| 45 | #define CB_BITS 0x0c |
| 46 | #endif |
| 47 | |
| 48 | /******************************************************************************* |
| 49 | * Íⲿº¯ÊýÉùÃ÷ * |
| 50 | *******************************************************************************/ |
| 51 | .extern decompress_kernel |
| 52 | |
| 53 | /******************************************************************************* |
| 54 | * Íⲿ±äÁ¿ÉùÃ÷ * |
| 55 | *******************************************************************************/ |
| 56 | .extern image_start |
| 57 | |
| 58 | /******************************************************************************* |
| 59 | * È«¾Öº¯ÊýʵÏÖ * |
| 60 | *******************************************************************************/ |
| 61 | |
| 62 | .section ".start", #alloc, #execinstr |
| 63 | .align |
| 64 | .arm @ Always enter in ARM state |
| 65 | .global _start |
| 66 | .type _start, function |
| 67 | |
| 68 | _start: |
| 69 | .rept 8 |
| 70 | mov r0, r0 |
| 71 | .endr |
| 72 | |
| 73 | .text |
| 74 | /* move to SVC MODE */ |
| 75 | mrs r0, cpsr |
| 76 | bic r0, #0x1f |
| 77 | orr r0, r0, #0xd3 |
| 78 | bic r0, #(1<<8) /* unmask Asynchronous abort */ |
| 79 | msr cpsr_cxsf, r0 |
| 80 | |
| 81 | /* Control Register Setup */ |
| 82 | mrc p15, 0, r0, c1, c0, 0 |
| 83 | bic r0, r0, #(1<<0) /* MMU disabled */ |
| 84 | orr r0, r0, #(1<<1) /* Alignment fault checking enabled */ |
| 85 | bic r0, r0, #(1<<2) /* Data Cache disabled */ |
| 86 | orr r0, r0, #(1<<11) /* Branch prediction enabled */ |
| 87 | bic r0, r0, #(1<<12) /* Instruction Cache disabled */ |
| 88 | bic r0, r0, #(1<<13) /* USE VBAR to set the vector base address */ |
| 89 | DSB /* Ensure all previous loads/stores have completed */ |
| 90 | mcr p15, 0, r0, c1, c0, 0 |
| 91 | ISB |
| 92 | |
| 93 | adr r0, LC0 |
| 94 | ldmia r0, {r1, r2, r3, r6, r10, r11, r12} |
| 95 | ldr sp, [r0, #28] |
| 96 | |
| 97 | /* |
| 98 | * We might be running at a different address. We need |
| 99 | * to fix up various pointers. |
| 100 | */ |
| 101 | sub r0, r0, r1 @ calculate the delta offset |
| 102 | add r2, r2, r0 @ __bss_start |
| 103 | add r3, r3, r0 @ __bss_end |
| 104 | add r6, r6, r0 @ _edata |
| 105 | add r10, r10, r0 @ inflated kernel size location |
| 106 | add r11, r11, r0 @ got_start |
| 107 | add r12, r12, r0 @ got_end |
| 108 | add sp, sp, r0 @ sp |
| 109 | |
| 110 | /*¡¡½âѹºóÄÚºËµÄÆô¶¯µØÖ· */ |
| 111 | ldr r4, =image_start |
| 112 | add r4, r4, r0 |
| 113 | ldr r4, [r4] |
| 114 | |
| 115 | /* |
| 116 | * The kernel build system appends the size of the |
| 117 | * decompressed kernel at the end of the compressed data |
| 118 | * in little-endian form. |
| 119 | */ |
| 120 | ldrb r9, [r10, #0] |
| 121 | ldrb lr, [r10, #1] |
| 122 | orr r9, r9, lr, lsl #8 |
| 123 | ldrb lr, [r10, #2] |
| 124 | ldrb r10, [r10, #3] |
| 125 | orr r9, r9, lr, lsl #16 |
| 126 | orr r9, r9, r10, lsl #24 |
| 127 | |
| 128 | add r10, r4, r9 |
| 129 | adr r9, _clear_bss |
| 130 | cmp r10, r9 |
| 131 | dead_loop: |
| 132 | bgt dead_loop |
| 133 | |
| 134 | /* |
| 135 | * Relocate all entries in the GOT table. |
| 136 | * Bump bss entries to _edata + dtb size |
| 137 | */ |
| 138 | mov r5, #0 |
| 139 | 1: |
| 140 | ldr r1, [r11, #0] @ relocate entries in the GOT |
| 141 | add r1, r1, r0 @ This fixes up C references |
| 142 | cmp r1, r2 @ if entry >= bss_start && |
| 143 | cmphs r3, r1 @ bss_end > entry |
| 144 | addhi r1, r1, r5 @ entry += dtb size |
| 145 | str r1, [r11], #4 @ next entry |
| 146 | cmp r11, r12 |
| 147 | blo 1b |
| 148 | |
| 149 | /* bump our bss pointers too */ |
| 150 | add r2, r2, r5 |
| 151 | add r3, r3, r5 |
| 152 | |
| 153 | /* |
| 154 | * BSS¶ÎÇåÁã |
| 155 | */ |
| 156 | mov r0, #0 |
| 157 | _clear_bss: |
| 158 | str r0, [r2], #4 |
| 159 | cmp r3, r2 |
| 160 | bhi _clear_bss |
| 161 | |
| 162 | bl cache_on |
| 163 | /* |
| 164 | * decompress kernel |
| 165 | */ |
| 166 | mov r0, r4 |
| 167 | mov r1, sp @ malloc space above stack |
| 168 | add r2, sp, #0x10000 @ 64k max |
| 169 | bl decompress_kernel |
| 170 | bl cache_clean_flush |
| 171 | bl cache_off |
| 172 | |
| 173 | /* |
| 174 | * enter kernel |
| 175 | */ |
| 176 | mov r0, #0 |
| 177 | bx r4 |
| 178 | |
| 179 | .align 2 |
| 180 | .type LC0, #object |
| 181 | |
| 182 | LC0: |
| 183 | .word LC0 @ r1 |
| 184 | .word __bss_start @ r2 |
| 185 | .word __bss_end @ r3 |
| 186 | .word _edata @ r6 |
| 187 | .word input_data_end - 4 @ r10 (inflated size location) |
| 188 | .word _got_start @ r11 |
| 189 | .word _got_end @ ip |
| 190 | .word .L_user_stack_end @ sp |
| 191 | .size LC0, . - LC0 |
| 192 | |
| 193 | __setup_mmu: |
| 194 | lsr r3, r4, #0x14 |
| 195 | lsl r3, r3, #0x14 |
| 196 | sub r3, r3, #16384 @ Page directory size |
| 197 | bic r3, r3, #0xff @ Align the pointer |
| 198 | bic r3, r3, #0x3f00 |
| 199 | /* |
| 200 | * Initialise the page tables, turning on the cacheable and bufferable |
| 201 | * bits for the RAM area only. |
| 202 | */ |
| 203 | mov r0, r3 |
| 204 | mov r9, r0, lsr #18 |
| 205 | mov r9, r9, lsl #18 @ start of RAM |
| 206 | add r10, r9, #0x10000000 @ a reasonable RAM size |
| 207 | mov r1, #0x12 @ XN|U + section mapping |
| 208 | orr r1, r1, #3 << 10 @ AP=11 |
| 209 | add r2, r3, #16384 |
| 210 | 1: cmp r1, r9 @ if virt > start of RAM |
| 211 | cmphs r10, r1 @ && end of RAM > virt |
| 212 | bic r1, r1, #0x1c @ clear XN|U + C + B |
| 213 | orrlo r1, r1, #0x10 @ Set XN|U for non-RAM |
| 214 | orrhs r1, r1, r6 @ set RAM section settings |
| 215 | str r1, [r0], #4 @ 1:1 mapping |
| 216 | add r1, r1, #1048576 |
| 217 | teq r0, r2 |
| 218 | bne 1b |
| 219 | /* |
| 220 | * If ever we are running from Flash, then we surely want the cache |
| 221 | * to be enabled also for our execution instance... We map 2MB of it |
| 222 | * so there is no map overlap problem for up to 1 MB compressed kernel. |
| 223 | * If the execution is in RAM then we would only be duplicating the above. |
| 224 | */ |
| 225 | orr r1, r6, #0x04 @ ensure B is set for this |
| 226 | orr r1, r1, #3 << 10 |
| 227 | mov r2, pc |
| 228 | mov r2, r2, lsr #20 |
| 229 | orr r1, r1, r2, lsl #20 |
| 230 | add r0, r3, r2, lsl #2 |
| 231 | str r1, [r0], #4 |
| 232 | add r1, r1, #1048576 |
| 233 | str r1, [r0] |
| 234 | mov pc, lr |
| 235 | ENDPROC(__setup_mmu) |
| 236 | |
| 237 | __armv7_mmu_cache_on: |
| 238 | mov r12, lr |
| 239 | #ifdef CYGOPT_HAL_ARM_MMU |
| 240 | mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0 |
| 241 | tst r11, #0xf @ VMSA |
| 242 | movne r6, #CB_BITS | 0x02 @ !XN |
| 243 | blne __setup_mmu |
| 244 | mov r0, #0 |
| 245 | mcr p15, 0, r0, c7, c10, 4 @ drain write buffer |
| 246 | tst r11, #0xf @ VMSA |
| 247 | mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs |
| 248 | #endif |
| 249 | mrc p15, 0, r0, c1, c0, 0 @ read control reg |
| 250 | bic r0, r0, #1 << 28 @ clear SCTLR.TRE |
| 251 | orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement |
| 252 | orr r0, r0, #0x003c @ write buffer |
| 253 | bic r0, r0, #2 @ A (no unaligned access fault) |
| 254 | orr r0, r0, #1 << 22 @ U (v6 unaligned access model) |
| 255 | @ (needed for ARM1176) |
| 256 | #ifdef CYGOPT_HAL_ARM_MMU |
| 257 | ARM_BE8( orr r0, r0, #1 << 25 ) @ big-endian page tables |
| 258 | mrcne p15, 0, r6, c2, c0, 2 @ read ttb control reg |
| 259 | orrne r0, r0, #1 @ MMU enabled |
| 260 | movne r1, #0xfffffffd @ domain 0 = client |
| 261 | bic r6, r6, #1 << 31 @ 32-bit translation system |
| 262 | bic r6, r6, #3 << 0 @ use only ttbr0 |
| 263 | mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer |
| 264 | mcrne p15, 0, r1, c3, c0, 0 @ load domain access control |
| 265 | mcrne p15, 0, r6, c2, c0, 2 @ load ttb control |
| 266 | #endif |
| 267 | mcr p15, 0, r0, c7, c5, 4 @ ISB |
| 268 | mcr p15, 0, r0, c1, c0, 0 @ load control register |
| 269 | mrc p15, 0, r0, c1, c0, 0 @ and read it back |
| 270 | mov r0, #0 |
| 271 | mcr p15, 0, r0, c7, c5, 4 @ ISB |
| 272 | mov pc, r12 |
| 273 | |
| 274 | #define PROC_ENTRY_SIZE (4*5) |
| 275 | |
| 276 | /* cache on */ |
| 277 | .align 5 |
| 278 | cache_on: mov r3, #8 @ cache_on function |
| 279 | b call_cache_fn |
| 280 | |
| 281 | call_cache_fn: adr r12, proc_types |
| 282 | #ifdef CONFIG_CPU_CP15 |
| 283 | mrc p15, 0, r9, c0, c0 @ get processor ID |
| 284 | #elif defined(CONFIG_CPU_V7M) |
| 285 | /* |
| 286 | * On v7-M the processor id is located in the V7M_SCB_CPUID |
| 287 | * register, but as cache handling is IMPLEMENTATION DEFINED on |
| 288 | * v7-M (if existant at all) we just return early here. |
| 289 | * If V7M_SCB_CPUID were used the cpu ID functions (i.e. |
| 290 | * __armv7_mmu_cache_{on,off,flush}) would be selected which |
| 291 | * use cp15 registers that are not implemented on v7-M. |
| 292 | */ |
| 293 | bx lr |
| 294 | #else |
| 295 | ldr r9, =CONFIG_PROCESSOR_ID |
| 296 | #endif |
| 297 | 1: ldr r1, [r12, #0] @ get value |
| 298 | ldr r2, [r12, #4] @ get mask |
| 299 | eor r1, r1, r9 @ (real ^ match) |
| 300 | tst r1, r2 @ & mask |
| 301 | ARM( addeq pc, r12, r3 ) @ call cache function |
| 302 | THUMB( addeq r12, r3 ) |
| 303 | THUMB( moveq pc, r12 ) @ call cache function |
| 304 | add r12, r12, #PROC_ENTRY_SIZE |
| 305 | b 1b |
| 306 | |
| 307 | /* |
| 308 | * Table for cache operations. This is basically: |
| 309 | * - CPU ID match |
| 310 | * - CPU ID mask |
| 311 | * - 'cache on' method instruction |
| 312 | * - 'cache off' method instruction |
| 313 | * - 'cache flush' method instruction |
| 314 | * |
| 315 | * We match an entry using: ((real_id ^ match) & mask) == 0 |
| 316 | * |
| 317 | * Writethrough caches generally only need 'on' and 'off' |
| 318 | * methods. Writeback caches _must_ have the flush method |
| 319 | * defined. |
| 320 | */ |
| 321 | .align 2 |
| 322 | .type proc_types,#object |
| 323 | proc_types: |
| 324 | .word 0x000f0000 @ new CPU Id |
| 325 | .word 0x000f0000 |
| 326 | W(b) __armv7_mmu_cache_on |
| 327 | W(b) __armv7_mmu_cache_off |
| 328 | W(b) __armv7_mmu_cache_flush |
| 329 | |
| 330 | .word 0 @ unrecognised type |
| 331 | .word 0 |
| 332 | mov pc, lr |
| 333 | THUMB( nop ) |
| 334 | mov pc, lr |
| 335 | THUMB( nop ) |
| 336 | mov pc, lr |
| 337 | THUMB( nop ) |
| 338 | |
| 339 | .size proc_types, . - proc_types |
| 340 | |
| 341 | /* |
| 342 | * If you get a "non-constant expression in ".if" statement" |
| 343 | * error from the assembler on this line, check that you have |
| 344 | * not accidentally written a "b" instruction where you should |
| 345 | * have written W(b). |
| 346 | */ |
| 347 | .if (. - proc_types) % PROC_ENTRY_SIZE != 0 |
| 348 | .error "The size of one or more proc_types entries is wrong." |
| 349 | .endif |
| 350 | |
| 351 | .align 5 |
| 352 | cache_off: mov r3, #12 @ cache_off function |
| 353 | b call_cache_fn |
| 354 | |
| 355 | __armv7_mmu_cache_off: |
| 356 | mrc p15, 0, r0, c1, c0 |
| 357 | #ifdef CYGOPT_HAL_ARM_MMU |
| 358 | bic r0, r0, #0x000d |
| 359 | #else |
| 360 | bic r0, r0, #0x000c |
| 361 | #endif |
| 362 | mcr p15, 0, r0, c1, c0 @ turn MMU and cache off |
| 363 | mov r12, lr |
| 364 | bl __armv7_mmu_cache_flush |
| 365 | mov r0, #0 |
| 366 | #ifdef CYGOPT_HAL_ARM_MMU |
| 367 | mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB |
| 368 | #endif |
| 369 | mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC |
| 370 | mcr p15, 0, r0, c7, c10, 4 @ DSB |
| 371 | mcr p15, 0, r0, c7, c5, 4 @ ISB |
| 372 | mov pc, r12 |
| 373 | |
| 374 | .align 5 |
| 375 | cache_clean_flush: |
| 376 | mov r3, #16 |
| 377 | b call_cache_fn |
| 378 | |
| 379 | __armv7_mmu_cache_flush: |
| 380 | tst r4, #1 |
| 381 | bne _iflush |
| 382 | mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1 |
| 383 | tst r10, #0xf << 16 @ hierarchical cache (ARMv7) |
| 384 | mov r10, #0 |
| 385 | beq hierarchical |
| 386 | mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D |
| 387 | b _iflush |
| 388 | hierarchical: |
| 389 | mcr p15, 0, r10, c7, c10, 5 @ DMB |
| 390 | stmfd sp!, {r0-r7, r9-r11} |
| 391 | mrc p15, 1, r0, c0, c0, 1 |
| 392 | ands r3, r0, #0x7000000 |
| 393 | mov r3, r3, lsr #23 |
| 394 | beq _finished |
| 395 | mov r10, #0 |
| 396 | _loop1: |
| 397 | add r2, r10, r10, lsr #1 |
| 398 | mov r1, r0, lsr r2 |
| 399 | and r1, r1, #7 |
| 400 | cmp r1, #2 |
| 401 | blt _skip |
| 402 | mcr p15, 2, r10, c0, c0, 0 |
| 403 | mcr p15, 0, r10, c7, c5, 4 |
| 404 | mrc p15, 1, r1, c0, c0, 0 |
| 405 | and r2, r1, #7 |
| 406 | add r2, r2, #4 |
| 407 | ldr r4, =0x3ff |
| 408 | ands r4, r4, r1, lsr #3 |
| 409 | clz r5, r4 |
| 410 | ldr r7, =0x7fff |
| 411 | ands r7, r7, r1, lsr #13 |
| 412 | _loop2: |
| 413 | mov r9, r4 |
| 414 | _loop3: |
| 415 | ARM( orr r11, r10, r9, lsl r5 ) |
| 416 | ARM( orr r11, r11, r7, lsl r2 ) |
| 417 | THUMB( lsl r6, r9, r5 ) |
| 418 | THUMB( orr r11, r10, r6 ) |
| 419 | THUMB( lsl r6, r7, r2 ) |
| 420 | THUMB( orr r11, r11, r6 ) |
| 421 | mcr p15, 0, r11, c7, c14, 2 |
| 422 | subs r9, r9, #1 |
| 423 | bge _loop3 |
| 424 | subs r7, r7, #1 |
| 425 | bge _loop2 |
| 426 | _skip: |
| 427 | add r10, r10, #2 |
| 428 | cmp r3, r10 |
| 429 | bgt _loop1 |
| 430 | _finished: |
| 431 | ldmfd sp!, {r0-r7, r9-r11} |
| 432 | mov r10, #0 @ swith back to cache level 0 |
| 433 | mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr |
| 434 | _iflush: |
| 435 | mcr p15, 0, r10, c7, c10, 4 @ DSB |
| 436 | mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB |
| 437 | mcr p15, 0, r10, c7, c10, 4 @ DSB |
| 438 | mcr p15, 0, r10, c7, c5, 4 @ ISB |
| 439 | mov pc, lr |
| 440 | |
| 441 | .align |
| 442 | .section ".stack", "aw", %nobits |
| 443 | |
| 444 | .L_user_stack: .space 4096 |
| 445 | .L_user_stack_end: |
| 446 | |