Merge branch 'master' into next
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / arch / arm / boot / compressed / head.S
1 /*
2 * linux/arch/arm/boot/compressed/head.S
3 *
4 * Copyright (C) 1996-2002 Russell King
5 * Copyright (C) 2004 Hyok S. Choi (MPU support)
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11 #include <linux/linkage.h>
12
13 /*
14 * Debugging stuff
15 *
16 * Note that these macros must not contain any code which is not
17 * 100% relocatable. Any attempt to do so will result in a crash.
18 * Please select one of the following when turning on debugging.
19 */
20 #ifdef DEBUG
21
22 #if defined(CONFIG_DEBUG_ICEDCC)
23
24 #if defined(CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K) || defined(CONFIG_CPU_V7)
25 .macro loadsp, rb, tmp
26 .endm
27 .macro writeb, ch, rb
28 mcr p14, 0, \ch, c0, c5, 0
29 .endm
30 #elif defined(CONFIG_CPU_XSCALE)
31 .macro loadsp, rb, tmp
32 .endm
33 .macro writeb, ch, rb
34 mcr p14, 0, \ch, c8, c0, 0
35 .endm
36 #else
37 .macro loadsp, rb, tmp
38 .endm
39 .macro writeb, ch, rb
40 mcr p14, 0, \ch, c1, c0, 0
41 .endm
42 #endif
43
44 #else
45
46 #include <mach/debug-macro.S>
47
48 .macro writeb, ch, rb
49 senduart \ch, \rb
50 .endm
51
52 #if defined(CONFIG_ARCH_SA1100)
53 .macro loadsp, rb, tmp
54 mov \rb, #0x80000000 @ physical base address
55 #ifdef CONFIG_DEBUG_LL_SER3
56 add \rb, \rb, #0x00050000 @ Ser3
57 #else
58 add \rb, \rb, #0x00010000 @ Ser1
59 #endif
60 .endm
61 #elif defined(CONFIG_ARCH_S3C2410)
62 .macro loadsp, rb, tmp
63 mov \rb, #0x50000000
64 add \rb, \rb, #0x4000 * CONFIG_S3C_LOWLEVEL_UART_PORT
65 .endm
66 #else
67 .macro loadsp, rb, tmp
68 addruart \rb, \tmp
69 .endm
70 #endif
71 #endif
72 #endif
73
74 .macro kputc,val
75 mov r0, \val
76 bl putc
77 .endm
78
79 .macro kphex,val,len
80 mov r0, \val
81 mov r1, #\len
82 bl phex
83 .endm
84
85 .macro debug_reloc_start
86 #ifdef DEBUG
87 kputc #'\n'
88 kphex r6, 8 /* processor id */
89 kputc #':'
90 kphex r7, 8 /* architecture id */
91 #ifdef CONFIG_CPU_CP15
92 kputc #':'
93 mrc p15, 0, r0, c1, c0
94 kphex r0, 8 /* control reg */
95 #endif
96 kputc #'\n'
97 kphex r5, 8 /* decompressed kernel start */
98 kputc #'-'
99 kphex r9, 8 /* decompressed kernel end */
100 kputc #'>'
101 kphex r4, 8 /* kernel execution address */
102 kputc #'\n'
103 #endif
104 .endm
105
106 .macro debug_reloc_end
107 #ifdef DEBUG
108 kphex r5, 8 /* end of kernel */
109 kputc #'\n'
110 mov r0, r4
111 bl memdump /* dump 256 bytes at start of kernel */
112 #endif
113 .endm
114
115 .section ".start", #alloc, #execinstr
116 /*
117 * sort out different calling conventions
118 */
119 .align
120 .arm @ Always enter in ARM state
121 start:
122 .type start,#function
123 .rept 7
124 mov r0, r0
125 .endr
126 ARM( mov r0, r0 )
127 ARM( b 1f )
128 THUMB( adr r12, BSYM(1f) )
129 THUMB( bx r12 )
130
131 .word 0x016f2818 @ Magic numbers to help the loader
132 .word start @ absolute load/run zImage address
133 .word _edata @ zImage end address
134 THUMB( .thumb )
135 1: mov r7, r1 @ save architecture ID
136 mov r8, r2 @ save atags pointer
137
138 #ifndef __ARM_ARCH_2__
139 /*
140 * Booting from Angel - need to enter SVC mode and disable
141 * FIQs/IRQs (numeric definitions from angel arm.h source).
142 * We only do this if we were in user mode on entry.
143 */
144 mrs r2, cpsr @ get current mode
145 tst r2, #3 @ not user?
146 bne not_angel
147 mov r0, #0x17 @ angel_SWIreason_EnterSVC
148 ARM( swi 0x123456 ) @ angel_SWI_ARM
149 THUMB( svc 0xab ) @ angel_SWI_THUMB
150 not_angel:
151 mrs r2, cpsr @ turn off interrupts to
152 orr r2, r2, #0xc0 @ prevent angel from running
153 msr cpsr_c, r2
154 #else
155 teqp pc, #0x0c000003 @ turn off interrupts
156 #endif
157
158 /*
159 * Note that some cache flushing and other stuff may
160 * be needed here - is there an Angel SWI call for this?
161 */
162
163 /*
164 * some architecture specific code can be inserted
165 * by the linker here, but it should preserve r7, r8, and r9.
166 */
167
168 .text
169
170 #ifdef CONFIG_AUTO_ZRELADDR
171 @ determine final kernel image address
172 mov r4, pc
173 and r4, r4, #0xf8000000
174 add r4, r4, #TEXT_OFFSET
175 #else
176 ldr r4, =zreladdr
177 #endif
178
179 bl cache_on
180
181 restart: adr r0, LC0
182 ldmia r0, {r1, r2, r3, r6, r9, r11, r12}
183 ldr sp, [r0, #28]
184
185 /*
186 * We might be running at a different address. We need
187 * to fix up various pointers.
188 */
189 sub r0, r0, r1 @ calculate the delta offset
190 add r6, r6, r0 @ _edata
191
192 #ifndef CONFIG_ZBOOT_ROM
193 /* malloc space is above the relocated stack (64k max) */
194 add sp, sp, r0
195 add r10, sp, #0x10000
196 #else
197 /*
198 * With ZBOOT_ROM the bss/stack is non relocatable,
199 * but someone could still run this code from RAM,
200 * in which case our reference is _edata.
201 */
202 mov r10, r6
203 #endif
204
205 /*
206 * Check to see if we will overwrite ourselves.
207 * r4 = final kernel address
208 * r9 = size of decompressed image
209 * r10 = end of this image, including bss/stack/malloc space if non XIP
210 * We basically want:
211 * r4 - 16k page directory >= r10 -> OK
212 * r4 + image length <= current position (pc) -> OK
213 */
214 add r10, r10, #16384
215 cmp r4, r10
216 bhs wont_overwrite
217 add r10, r4, r9
218 ARM( cmp r10, pc )
219 THUMB( mov lr, pc )
220 THUMB( cmp r10, lr )
221 bls wont_overwrite
222
223 /*
224 * Relocate ourselves past the end of the decompressed kernel.
225 * r6 = _edata
226 * r10 = end of the decompressed kernel
227 * Because we always copy ahead, we need to do it from the end and go
228 * backward in case the source and destination overlap.
229 */
230 /*
231 * Bump to the next 256-byte boundary with the size of
232 * the relocation code added. This avoids overwriting
233 * ourself when the offset is small.
234 */
235 add r10, r10, #((reloc_code_end - restart + 256) & ~255)
236 bic r10, r10, #255
237
238 /* Get start of code we want to copy and align it down. */
239 adr r5, restart
240 bic r5, r5, #31
241
242 sub r9, r6, r5 @ size to copy
243 add r9, r9, #31 @ rounded up to a multiple
244 bic r9, r9, #31 @ ... of 32 bytes
245 add r6, r9, r5
246 add r9, r9, r10
247
248 1: ldmdb r6!, {r0 - r3, r10 - r12, lr}
249 cmp r6, r5
250 stmdb r9!, {r0 - r3, r10 - r12, lr}
251 bhi 1b
252
253 /* Preserve offset to relocated code. */
254 sub r6, r9, r6
255
256 #ifndef CONFIG_ZBOOT_ROM
257 /* cache_clean_flush may use the stack, so relocate it */
258 add sp, sp, r6
259 #endif
260
261 bl cache_clean_flush
262
263 adr r0, BSYM(restart)
264 add r0, r0, r6
265 mov pc, r0
266
267 wont_overwrite:
268 /*
269 * If delta is zero, we are running at the address we were linked at.
270 * r0 = delta
271 * r2 = BSS start
272 * r3 = BSS end
273 * r4 = kernel execution address
274 * r7 = architecture ID
275 * r8 = atags pointer
276 * r11 = GOT start
277 * r12 = GOT end
278 * sp = stack pointer
279 */
280 teq r0, #0
281 beq not_relocated
282 add r11, r11, r0
283 add r12, r12, r0
284
285 #ifndef CONFIG_ZBOOT_ROM
286 /*
287 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
288 * we need to fix up pointers into the BSS region.
289 * Note that the stack pointer has already been fixed up.
290 */
291 add r2, r2, r0
292 add r3, r3, r0
293
294 /*
295 * Relocate all entries in the GOT table.
296 */
297 1: ldr r1, [r11, #0] @ relocate entries in the GOT
298 add r1, r1, r0 @ table. This fixes up the
299 str r1, [r11], #4 @ C references.
300 cmp r11, r12
301 blo 1b
302 #else
303
304 /*
305 * Relocate entries in the GOT table. We only relocate
306 * the entries that are outside the (relocated) BSS region.
307 */
308 1: ldr r1, [r11, #0] @ relocate entries in the GOT
309 cmp r1, r2 @ entry < bss_start ||
310 cmphs r3, r1 @ _end < entry
311 addlo r1, r1, r0 @ table. This fixes up the
312 str r1, [r11], #4 @ C references.
313 cmp r11, r12
314 blo 1b
315 #endif
316
317 not_relocated: mov r0, #0
318 1: str r0, [r2], #4 @ clear bss
319 str r0, [r2], #4
320 str r0, [r2], #4
321 str r0, [r2], #4
322 cmp r2, r3
323 blo 1b
324
325 /*
326 * The C runtime environment should now be setup sufficiently.
327 * Set up some pointers, and start decompressing.
328 * r4 = kernel execution address
329 * r7 = architecture ID
330 * r8 = atags pointer
331 */
332 mov r0, r4
333 mov r1, sp @ malloc space above stack
334 add r2, sp, #0x10000 @ 64k max
335 mov r3, r7
336 bl decompress_kernel
337 bl cache_clean_flush
338 bl cache_off
339 mov r0, #0 @ must be zero
340 mov r1, r7 @ restore architecture number
341 mov r2, r8 @ restore atags pointer
342 mov pc, r4 @ call kernel
343
344 .align 2
345 .type LC0, #object
346 LC0: .word LC0 @ r1
347 .word __bss_start @ r2
348 .word _end @ r3
349 .word _edata @ r6
350 .word _image_size @ r9
351 .word _got_start @ r11
352 .word _got_end @ ip
353 .word user_stack_end @ sp
354 .size LC0, . - LC0
355
356 #ifdef CONFIG_ARCH_RPC
357 .globl params
358 params: ldr r0, =0x10000100 @ params_phys for RPC
359 mov pc, lr
360 .ltorg
361 .align
362 #endif
363
364 /*
365 * Turn on the cache. We need to setup some page tables so that we
366 * can have both the I and D caches on.
367 *
368 * We place the page tables 16k down from the kernel execution address,
369 * and we hope that nothing else is using it. If we're using it, we
370 * will go pop!
371 *
372 * On entry,
373 * r4 = kernel execution address
374 * r7 = architecture number
375 * r8 = atags pointer
376 * On exit,
377 * r0, r1, r2, r3, r9, r10, r12 corrupted
378 * This routine must preserve:
379 * r4, r7, r8
380 */
381 .align 5
382 cache_on: mov r3, #8 @ cache_on function
383 b call_cache_fn
384
385 /*
386 * Initialize the highest priority protection region, PR7
387 * to cover all 32bit address and cacheable and bufferable.
388 */
389 __armv4_mpu_cache_on:
390 mov r0, #0x3f @ 4G, the whole
391 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
392 mcr p15, 0, r0, c6, c7, 1
393
394 mov r0, #0x80 @ PR7
395 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
396 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
397 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
398
399 mov r0, #0xc000
400 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
401 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
402
403 mov r0, #0
404 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
405 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
406 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
407 mrc p15, 0, r0, c1, c0, 0 @ read control reg
408 @ ...I .... ..D. WC.M
409 orr r0, r0, #0x002d @ .... .... ..1. 11.1
410 orr r0, r0, #0x1000 @ ...1 .... .... ....
411
412 mcr p15, 0, r0, c1, c0, 0 @ write control reg
413
414 mov r0, #0
415 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
416 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
417 mov pc, lr
418
419 __armv3_mpu_cache_on:
420 mov r0, #0x3f @ 4G, the whole
421 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
422
423 mov r0, #0x80 @ PR7
424 mcr p15, 0, r0, c2, c0, 0 @ cache on
425 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
426
427 mov r0, #0xc000
428 mcr p15, 0, r0, c5, c0, 0 @ access permission
429
430 mov r0, #0
431 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
432 /*
433 * ?? ARMv3 MMU does not allow reading the control register,
434 * does this really work on ARMv3 MPU?
435 */
436 mrc p15, 0, r0, c1, c0, 0 @ read control reg
437 @ .... .... .... WC.M
438 orr r0, r0, #0x000d @ .... .... .... 11.1
439 /* ?? this overwrites the value constructed above? */
440 mov r0, #0
441 mcr p15, 0, r0, c1, c0, 0 @ write control reg
442
443 /* ?? invalidate for the second time? */
444 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
445 mov pc, lr
446
447 __setup_mmu: sub r3, r4, #16384 @ Page directory size
448 bic r3, r3, #0xff @ Align the pointer
449 bic r3, r3, #0x3f00
450 /*
451 * Initialise the page tables, turning on the cacheable and bufferable
452 * bits for the RAM area only.
453 */
454 mov r0, r3
455 mov r9, r0, lsr #18
456 mov r9, r9, lsl #18 @ start of RAM
457 add r10, r9, #0x10000000 @ a reasonable RAM size
458 mov r1, #0x12
459 orr r1, r1, #3 << 10
460 add r2, r3, #16384
461 1: cmp r1, r9 @ if virt > start of RAM
462 orrhs r1, r1, #0x0c @ set cacheable, bufferable
463 cmp r1, r10 @ if virt > end of RAM
464 bichs r1, r1, #0x0c @ clear cacheable, bufferable
465 str r1, [r0], #4 @ 1:1 mapping
466 add r1, r1, #1048576
467 teq r0, r2
468 bne 1b
469 /*
470 * If ever we are running from Flash, then we surely want the cache
471 * to be enabled also for our execution instance... We map 2MB of it
472 * so there is no map overlap problem for up to 1 MB compressed kernel.
473 * If the execution is in RAM then we would only be duplicating the above.
474 */
475 mov r1, #0x1e
476 orr r1, r1, #3 << 10
477 mov r2, pc
478 mov r2, r2, lsr #20
479 orr r1, r1, r2, lsl #20
480 add r0, r3, r2, lsl #2
481 str r1, [r0], #4
482 add r1, r1, #1048576
483 str r1, [r0]
484 mov pc, lr
485 ENDPROC(__setup_mmu)
486
487 __armv4_mmu_cache_on:
488 mov r12, lr
489 #ifdef CONFIG_MMU
490 bl __setup_mmu
491 mov r0, #0
492 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
493 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
494 mrc p15, 0, r0, c1, c0, 0 @ read control reg
495 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
496 orr r0, r0, #0x0030
497 #ifdef CONFIG_CPU_ENDIAN_BE8
498 orr r0, r0, #1 << 25 @ big-endian page tables
499 #endif
500 bl __common_mmu_cache_on
501 mov r0, #0
502 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
503 #endif
504 mov pc, r12
505
506 __armv7_mmu_cache_on:
507 mov r12, lr
508 #ifdef CONFIG_MMU
509 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
510 tst r11, #0xf @ VMSA
511 blne __setup_mmu
512 mov r0, #0
513 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
514 tst r11, #0xf @ VMSA
515 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
516 #endif
517 mrc p15, 0, r0, c1, c0, 0 @ read control reg
518 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
519 orr r0, r0, #0x003c @ write buffer
520 #ifdef CONFIG_MMU
521 #ifdef CONFIG_CPU_ENDIAN_BE8
522 orr r0, r0, #1 << 25 @ big-endian page tables
523 #endif
524 orrne r0, r0, #1 @ MMU enabled
525 movne r1, #-1
526 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
527 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
528 #endif
529 mcr p15, 0, r0, c1, c0, 0 @ load control register
530 mrc p15, 0, r0, c1, c0, 0 @ and read it back
531 mov r0, #0
532 mcr p15, 0, r0, c7, c5, 4 @ ISB
533 mov pc, r12
534
535 __fa526_cache_on:
536 mov r12, lr
537 bl __setup_mmu
538 mov r0, #0
539 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
540 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
541 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
542 mrc p15, 0, r0, c1, c0, 0 @ read control reg
543 orr r0, r0, #0x1000 @ I-cache enable
544 bl __common_mmu_cache_on
545 mov r0, #0
546 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
547 mov pc, r12
548
549 __arm6_mmu_cache_on:
550 mov r12, lr
551 bl __setup_mmu
552 mov r0, #0
553 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
554 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
555 mov r0, #0x30
556 bl __common_mmu_cache_on
557 mov r0, #0
558 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
559 mov pc, r12
560
561 __common_mmu_cache_on:
562 #ifndef CONFIG_THUMB2_KERNEL
563 #ifndef DEBUG
564 orr r0, r0, #0x000d @ Write buffer, mmu
565 #endif
566 mov r1, #-1
567 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
568 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
569 b 1f
570 .align 5 @ cache line aligned
571 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
572 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
573 sub pc, lr, r0, lsr #32 @ properly flush pipeline
574 #endif
575
576 /*
577 * Here follow the relocatable cache support functions for the
578 * various processors. This is a generic hook for locating an
579 * entry and jumping to an instruction at the specified offset
580 * from the start of the block. Please note this is all position
581 * independent code.
582 *
583 * r1 = corrupted
584 * r2 = corrupted
585 * r3 = block offset
586 * r9 = corrupted
587 * r12 = corrupted
588 */
589
590 call_cache_fn: adr r12, proc_types
591 #ifdef CONFIG_CPU_CP15
592 mrc p15, 0, r9, c0, c0 @ get processor ID
593 #else
594 ldr r9, =CONFIG_PROCESSOR_ID
595 #endif
596 1: ldr r1, [r12, #0] @ get value
597 ldr r2, [r12, #4] @ get mask
598 eor r1, r1, r9 @ (real ^ match)
599 tst r1, r2 @ & mask
600 ARM( addeq pc, r12, r3 ) @ call cache function
601 THUMB( addeq r12, r3 )
602 THUMB( moveq pc, r12 ) @ call cache function
603 add r12, r12, #4*5
604 b 1b
605
606 /*
607 * Table for cache operations. This is basically:
608 * - CPU ID match
609 * - CPU ID mask
610 * - 'cache on' method instruction
611 * - 'cache off' method instruction
612 * - 'cache flush' method instruction
613 *
614 * We match an entry using: ((real_id ^ match) & mask) == 0
615 *
616 * Writethrough caches generally only need 'on' and 'off'
617 * methods. Writeback caches _must_ have the flush method
618 * defined.
619 */
620 .align 2
621 .type proc_types,#object
622 proc_types:
623 .word 0x41560600 @ ARM6/610
624 .word 0xffffffe0
625 W(b) __arm6_mmu_cache_off @ works, but slow
626 W(b) __arm6_mmu_cache_off
627 mov pc, lr
628 THUMB( nop )
629 @ b __arm6_mmu_cache_on @ untested
630 @ b __arm6_mmu_cache_off
631 @ b __armv3_mmu_cache_flush
632
633 .word 0x00000000 @ old ARM ID
634 .word 0x0000f000
635 mov pc, lr
636 THUMB( nop )
637 mov pc, lr
638 THUMB( nop )
639 mov pc, lr
640 THUMB( nop )
641
642 .word 0x41007000 @ ARM7/710
643 .word 0xfff8fe00
644 W(b) __arm7_mmu_cache_off
645 W(b) __arm7_mmu_cache_off
646 mov pc, lr
647 THUMB( nop )
648
649 .word 0x41807200 @ ARM720T (writethrough)
650 .word 0xffffff00
651 W(b) __armv4_mmu_cache_on
652 W(b) __armv4_mmu_cache_off
653 mov pc, lr
654 THUMB( nop )
655
656 .word 0x41007400 @ ARM74x
657 .word 0xff00ff00
658 W(b) __armv3_mpu_cache_on
659 W(b) __armv3_mpu_cache_off
660 W(b) __armv3_mpu_cache_flush
661
662 .word 0x41009400 @ ARM94x
663 .word 0xff00ff00
664 W(b) __armv4_mpu_cache_on
665 W(b) __armv4_mpu_cache_off
666 W(b) __armv4_mpu_cache_flush
667
668 .word 0x00007000 @ ARM7 IDs
669 .word 0x0000f000
670 mov pc, lr
671 THUMB( nop )
672 mov pc, lr
673 THUMB( nop )
674 mov pc, lr
675 THUMB( nop )
676
677 @ Everything from here on will be the new ID system.
678
679 .word 0x4401a100 @ sa110 / sa1100
680 .word 0xffffffe0
681 W(b) __armv4_mmu_cache_on
682 W(b) __armv4_mmu_cache_off
683 W(b) __armv4_mmu_cache_flush
684
685 .word 0x6901b110 @ sa1110
686 .word 0xfffffff0
687 W(b) __armv4_mmu_cache_on
688 W(b) __armv4_mmu_cache_off
689 W(b) __armv4_mmu_cache_flush
690
691 .word 0x56056900
692 .word 0xffffff00 @ PXA9xx
693 W(b) __armv4_mmu_cache_on
694 W(b) __armv4_mmu_cache_off
695 W(b) __armv4_mmu_cache_flush
696
697 .word 0x56158000 @ PXA168
698 .word 0xfffff000
699 W(b) __armv4_mmu_cache_on
700 W(b) __armv4_mmu_cache_off
701 W(b) __armv5tej_mmu_cache_flush
702
703 .word 0x56050000 @ Feroceon
704 .word 0xff0f0000
705 W(b) __armv4_mmu_cache_on
706 W(b) __armv4_mmu_cache_off
707 W(b) __armv5tej_mmu_cache_flush
708
709 #ifdef CONFIG_CPU_FEROCEON_OLD_ID
710 /* this conflicts with the standard ARMv5TE entry */
711 .long 0x41009260 @ Old Feroceon
712 .long 0xff00fff0
713 b __armv4_mmu_cache_on
714 b __armv4_mmu_cache_off
715 b __armv5tej_mmu_cache_flush
716 #endif
717
718 .word 0x66015261 @ FA526
719 .word 0xff01fff1
720 W(b) __fa526_cache_on
721 W(b) __armv4_mmu_cache_off
722 W(b) __fa526_cache_flush
723
724 @ These match on the architecture ID
725
726 .word 0x00020000 @ ARMv4T
727 .word 0x000f0000
728 W(b) __armv4_mmu_cache_on
729 W(b) __armv4_mmu_cache_off
730 W(b) __armv4_mmu_cache_flush
731
732 .word 0x00050000 @ ARMv5TE
733 .word 0x000f0000
734 W(b) __armv4_mmu_cache_on
735 W(b) __armv4_mmu_cache_off
736 W(b) __armv4_mmu_cache_flush
737
738 .word 0x00060000 @ ARMv5TEJ
739 .word 0x000f0000
740 W(b) __armv4_mmu_cache_on
741 W(b) __armv4_mmu_cache_off
742 W(b) __armv5tej_mmu_cache_flush
743
744 .word 0x0007b000 @ ARMv6
745 .word 0x000ff000
746 W(b) __armv4_mmu_cache_on
747 W(b) __armv4_mmu_cache_off
748 W(b) __armv6_mmu_cache_flush
749
750 .word 0x560f5810 @ Marvell PJ4 ARMv6
751 .word 0xff0ffff0
752 W(b) __armv4_mmu_cache_on
753 W(b) __armv4_mmu_cache_off
754 W(b) __armv6_mmu_cache_flush
755
756 .word 0x000f0000 @ new CPU Id
757 .word 0x000f0000
758 W(b) __armv7_mmu_cache_on
759 W(b) __armv7_mmu_cache_off
760 W(b) __armv7_mmu_cache_flush
761
762 .word 0 @ unrecognised type
763 .word 0
764 mov pc, lr
765 THUMB( nop )
766 mov pc, lr
767 THUMB( nop )
768 mov pc, lr
769 THUMB( nop )
770
771 .size proc_types, . - proc_types
772
773 /*
774 * Turn off the Cache and MMU. ARMv3 does not support
775 * reading the control register, but ARMv4 does.
776 *
777 * On exit,
778 * r0, r1, r2, r3, r9, r12 corrupted
779 * This routine must preserve:
780 * r4, r7, r8
781 */
782 .align 5
783 cache_off: mov r3, #12 @ cache_off function
784 b call_cache_fn
785
786 __armv4_mpu_cache_off:
787 mrc p15, 0, r0, c1, c0
788 bic r0, r0, #0x000d
789 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
790 mov r0, #0
791 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
792 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
793 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
794 mov pc, lr
795
796 __armv3_mpu_cache_off:
797 mrc p15, 0, r0, c1, c0
798 bic r0, r0, #0x000d
799 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
800 mov r0, #0
801 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
802 mov pc, lr
803
804 __armv4_mmu_cache_off:
805 #ifdef CONFIG_MMU
806 mrc p15, 0, r0, c1, c0
807 bic r0, r0, #0x000d
808 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
809 mov r0, #0
810 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
811 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
812 #endif
813 mov pc, lr
814
815 __armv7_mmu_cache_off:
816 mrc p15, 0, r0, c1, c0
817 #ifdef CONFIG_MMU
818 bic r0, r0, #0x000d
819 #else
820 bic r0, r0, #0x000c
821 #endif
822 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
823 mov r12, lr
824 bl __armv7_mmu_cache_flush
825 mov r0, #0
826 #ifdef CONFIG_MMU
827 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
828 #endif
829 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
830 mcr p15, 0, r0, c7, c10, 4 @ DSB
831 mcr p15, 0, r0, c7, c5, 4 @ ISB
832 mov pc, r12
833
834 __arm6_mmu_cache_off:
835 mov r0, #0x00000030 @ ARM6 control reg.
836 b __armv3_mmu_cache_off
837
838 __arm7_mmu_cache_off:
839 mov r0, #0x00000070 @ ARM7 control reg.
840 b __armv3_mmu_cache_off
841
842 __armv3_mmu_cache_off:
843 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
844 mov r0, #0
845 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
846 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
847 mov pc, lr
848
849 /*
850 * Clean and flush the cache to maintain consistency.
851 *
852 * On exit,
853 * r1, r2, r3, r9, r10, r11, r12 corrupted
854 * This routine must preserve:
855 * r4, r6, r7, r8
856 */
857 .align 5
858 cache_clean_flush:
859 mov r3, #16
860 b call_cache_fn
861
862 __armv4_mpu_cache_flush:
863 mov r2, #1
864 mov r3, #0
865 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
866 mov r1, #7 << 5 @ 8 segments
867 1: orr r3, r1, #63 << 26 @ 64 entries
868 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
869 subs r3, r3, #1 << 26
870 bcs 2b @ entries 63 to 0
871 subs r1, r1, #1 << 5
872 bcs 1b @ segments 7 to 0
873
874 teq r2, #0
875 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
876 mcr p15, 0, ip, c7, c10, 4 @ drain WB
877 mov pc, lr
878
879 __fa526_cache_flush:
880 mov r1, #0
881 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
882 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
883 mcr p15, 0, r1, c7, c10, 4 @ drain WB
884 mov pc, lr
885
886 __armv6_mmu_cache_flush:
887 mov r1, #0
888 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
889 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
890 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
891 mcr p15, 0, r1, c7, c10, 4 @ drain WB
892 mov pc, lr
893
894 __armv7_mmu_cache_flush:
895 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
896 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
897 mov r10, #0
898 beq hierarchical
899 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
900 b iflush
901 hierarchical:
902 mcr p15, 0, r10, c7, c10, 5 @ DMB
903 stmfd sp!, {r0-r7, r9-r11}
904 mrc p15, 1, r0, c0, c0, 1 @ read clidr
905 ands r3, r0, #0x7000000 @ extract loc from clidr
906 mov r3, r3, lsr #23 @ left align loc bit field
907 beq finished @ if loc is 0, then no need to clean
908 mov r10, #0 @ start clean at cache level 0
909 loop1:
910 add r2, r10, r10, lsr #1 @ work out 3x current cache level
911 mov r1, r0, lsr r2 @ extract cache type bits from clidr
912 and r1, r1, #7 @ mask of the bits for current cache only
913 cmp r1, #2 @ see what cache we have at this level
914 blt skip @ skip if no cache, or just i-cache
915 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
916 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
917 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
918 and r2, r1, #7 @ extract the length of the cache lines
919 add r2, r2, #4 @ add 4 (line length offset)
920 ldr r4, =0x3ff
921 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
922 clz r5, r4 @ find bit position of way size increment
923 ldr r7, =0x7fff
924 ands r7, r7, r1, lsr #13 @ extract max number of the index size
925 loop2:
926 mov r9, r4 @ create working copy of max way size
927 loop3:
928 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
929 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
930 THUMB( lsl r6, r9, r5 )
931 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
932 THUMB( lsl r6, r7, r2 )
933 THUMB( orr r11, r11, r6 ) @ factor index number into r11
934 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
935 subs r9, r9, #1 @ decrement the way
936 bge loop3
937 subs r7, r7, #1 @ decrement the index
938 bge loop2
939 skip:
940 add r10, r10, #2 @ increment cache number
941 cmp r3, r10
942 bgt loop1
943 finished:
944 ldmfd sp!, {r0-r7, r9-r11}
945 mov r10, #0 @ swith back to cache level 0
946 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
947 iflush:
948 mcr p15, 0, r10, c7, c10, 4 @ DSB
949 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
950 mcr p15, 0, r10, c7, c10, 4 @ DSB
951 mcr p15, 0, r10, c7, c5, 4 @ ISB
952 mov pc, lr
953
954 __armv5tej_mmu_cache_flush:
955 1: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
956 bne 1b
957 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
958 mcr p15, 0, r0, c7, c10, 4 @ drain WB
959 mov pc, lr
960
961 __armv4_mmu_cache_flush:
962 mov r2, #64*1024 @ default: 32K dcache size (*2)
963 mov r11, #32 @ default: 32 byte line size
964 mrc p15, 0, r3, c0, c0, 1 @ read cache type
965 teq r3, r9 @ cache ID register present?
966 beq no_cache_id
967 mov r1, r3, lsr #18
968 and r1, r1, #7
969 mov r2, #1024
970 mov r2, r2, lsl r1 @ base dcache size *2
971 tst r3, #1 << 14 @ test M bit
972 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
973 mov r3, r3, lsr #12
974 and r3, r3, #3
975 mov r11, #8
976 mov r11, r11, lsl r3 @ cache line size in bytes
977 no_cache_id:
978 mov r1, pc
979 bic r1, r1, #63 @ align to longest cache line
980 add r2, r1, r2
981 1:
982 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
983 THUMB( ldr r3, [r1] ) @ s/w flush D cache
984 THUMB( add r1, r1, r11 )
985 teq r1, r2
986 bne 1b
987
988 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
989 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
990 mcr p15, 0, r1, c7, c10, 4 @ drain WB
991 mov pc, lr
992
993 __armv3_mmu_cache_flush:
994 __armv3_mpu_cache_flush:
995 mov r1, #0
996 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
997 mov pc, lr
998
999 /*
1000 * Various debugging routines for printing hex characters and
1001 * memory, which again must be relocatable.
1002 */
1003 #ifdef DEBUG
1004 .align 2
1005 .type phexbuf,#object
1006 phexbuf: .space 12
1007 .size phexbuf, . - phexbuf
1008
1009 @ phex corrupts {r0, r1, r2, r3}
1010 phex: adr r3, phexbuf
1011 mov r2, #0
1012 strb r2, [r3, r1]
1013 1: subs r1, r1, #1
1014 movmi r0, r3
1015 bmi puts
1016 and r2, r0, #15
1017 mov r0, r0, lsr #4
1018 cmp r2, #10
1019 addge r2, r2, #7
1020 add r2, r2, #'0'
1021 strb r2, [r3, r1]
1022 b 1b
1023
1024 @ puts corrupts {r0, r1, r2, r3}
1025 puts: loadsp r3, r1
1026 1: ldrb r2, [r0], #1
1027 teq r2, #0
1028 moveq pc, lr
1029 2: writeb r2, r3
1030 mov r1, #0x00020000
1031 3: subs r1, r1, #1
1032 bne 3b
1033 teq r2, #'\n'
1034 moveq r2, #'\r'
1035 beq 2b
1036 teq r0, #0
1037 bne 1b
1038 mov pc, lr
1039 @ putc corrupts {r0, r1, r2, r3}
1040 putc:
1041 mov r2, r0
1042 mov r0, #0
1043 loadsp r3, r1
1044 b 2b
1045
1046 @ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1047 memdump: mov r12, r0
1048 mov r10, lr
1049 mov r11, #0
1050 2: mov r0, r11, lsl #2
1051 add r0, r0, r12
1052 mov r1, #8
1053 bl phex
1054 mov r0, #':'
1055 bl putc
1056 1: mov r0, #' '
1057 bl putc
1058 ldr r0, [r12, r11, lsl #2]
1059 mov r1, #8
1060 bl phex
1061 and r0, r11, #7
1062 teq r0, #3
1063 moveq r0, #' '
1064 bleq putc
1065 and r0, r11, #7
1066 add r11, r11, #1
1067 teq r0, #7
1068 bne 1b
1069 mov r0, #'\n'
1070 bl putc
1071 cmp r11, #64
1072 blt 2b
1073 mov pc, r10
1074 #endif
1075
1076 .ltorg
1077 reloc_code_end:
1078
1079 .align
1080 .section ".stack", "aw", %nobits
1081 user_stack: .space 4096
1082 user_stack_end: