Loading...
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_S3C24XX)
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
121start:
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 )
1351: 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
150not_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
181restart: adr r0, LC0
182 ldmia r0, {r1, r2, r3, r6, r10, 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 add r10, r10, r0 @ inflated kernel size location
192
193 /*
194 * The kernel build system appends the size of the
195 * decompressed kernel at the end of the compressed data
196 * in little-endian form.
197 */
198 ldrb r9, [r10, #0]
199 ldrb lr, [r10, #1]
200 orr r9, r9, lr, lsl #8
201 ldrb lr, [r10, #2]
202 ldrb r10, [r10, #3]
203 orr r9, r9, lr, lsl #16
204 orr r9, r9, r10, lsl #24
205
206#ifndef CONFIG_ZBOOT_ROM
207 /* malloc space is above the relocated stack (64k max) */
208 add sp, sp, r0
209 add r10, sp, #0x10000
210#else
211 /*
212 * With ZBOOT_ROM the bss/stack is non relocatable,
213 * but someone could still run this code from RAM,
214 * in which case our reference is _edata.
215 */
216 mov r10, r6
217#endif
218
219 mov r5, #0 @ init dtb size to 0
220#ifdef CONFIG_ARM_APPENDED_DTB
221/*
222 * r0 = delta
223 * r2 = BSS start
224 * r3 = BSS end
225 * r4 = final kernel address
226 * r5 = appended dtb size (still unknown)
227 * r6 = _edata
228 * r7 = architecture ID
229 * r8 = atags/device tree pointer
230 * r9 = size of decompressed image
231 * r10 = end of this image, including bss/stack/malloc space if non XIP
232 * r11 = GOT start
233 * r12 = GOT end
234 * sp = stack pointer
235 *
236 * if there are device trees (dtb) appended to zImage, advance r10 so that the
237 * dtb data will get relocated along with the kernel if necessary.
238 */
239
240 ldr lr, [r6, #0]
241#ifndef __ARMEB__
242 ldr r1, =0xedfe0dd0 @ sig is 0xd00dfeed big endian
243#else
244 ldr r1, =0xd00dfeed
245#endif
246 cmp lr, r1
247 bne dtb_check_done @ not found
248
249#ifdef CONFIG_ARM_ATAG_DTB_COMPAT
250 /*
251 * OK... Let's do some funky business here.
252 * If we do have a DTB appended to zImage, and we do have
253 * an ATAG list around, we want the later to be translated
254 * and folded into the former here. To be on the safe side,
255 * let's temporarily move the stack away into the malloc
256 * area. No GOT fixup has occurred yet, but none of the
257 * code we're about to call uses any global variable.
258 */
259 add sp, sp, #0x10000
260 stmfd sp!, {r0-r3, ip, lr}
261 mov r0, r8
262 mov r1, r6
263 sub r2, sp, r6
264 bl atags_to_fdt
265
266 /*
267 * If returned value is 1, there is no ATAG at the location
268 * pointed by r8. Try the typical 0x100 offset from start
269 * of RAM and hope for the best.
270 */
271 cmp r0, #1
272 sub r0, r4, #TEXT_OFFSET
273 add r0, r0, #0x100
274 mov r1, r6
275 sub r2, sp, r6
276 bleq atags_to_fdt
277
278 ldmfd sp!, {r0-r3, ip, lr}
279 sub sp, sp, #0x10000
280#endif
281
282 mov r8, r6 @ use the appended device tree
283
284 /*
285 * Make sure that the DTB doesn't end up in the final
286 * kernel's .bss area. To do so, we adjust the decompressed
287 * kernel size to compensate if that .bss size is larger
288 * than the relocated code.
289 */
290 ldr r5, =_kernel_bss_size
291 adr r1, wont_overwrite
292 sub r1, r6, r1
293 subs r1, r5, r1
294 addhi r9, r9, r1
295
296 /* Get the dtb's size */
297 ldr r5, [r6, #4]
298#ifndef __ARMEB__
299 /* convert r5 (dtb size) to little endian */
300 eor r1, r5, r5, ror #16
301 bic r1, r1, #0x00ff0000
302 mov r5, r5, ror #8
303 eor r5, r5, r1, lsr #8
304#endif
305
306 /* preserve 64-bit alignment */
307 add r5, r5, #7
308 bic r5, r5, #7
309
310 /* relocate some pointers past the appended dtb */
311 add r6, r6, r5
312 add r10, r10, r5
313 add sp, sp, r5
314dtb_check_done:
315#endif
316
317/*
318 * Check to see if we will overwrite ourselves.
319 * r4 = final kernel address
320 * r9 = size of decompressed image
321 * r10 = end of this image, including bss/stack/malloc space if non XIP
322 * We basically want:
323 * r4 - 16k page directory >= r10 -> OK
324 * r4 + image length <= address of wont_overwrite -> OK
325 */
326 add r10, r10, #16384
327 cmp r4, r10
328 bhs wont_overwrite
329 add r10, r4, r9
330 adr r9, wont_overwrite
331 cmp r10, r9
332 bls wont_overwrite
333
334/*
335 * Relocate ourselves past the end of the decompressed kernel.
336 * r6 = _edata
337 * r10 = end of the decompressed kernel
338 * Because we always copy ahead, we need to do it from the end and go
339 * backward in case the source and destination overlap.
340 */
341 /*
342 * Bump to the next 256-byte boundary with the size of
343 * the relocation code added. This avoids overwriting
344 * ourself when the offset is small.
345 */
346 add r10, r10, #((reloc_code_end - restart + 256) & ~255)
347 bic r10, r10, #255
348
349 /* Get start of code we want to copy and align it down. */
350 adr r5, restart
351 bic r5, r5, #31
352
353 sub r9, r6, r5 @ size to copy
354 add r9, r9, #31 @ rounded up to a multiple
355 bic r9, r9, #31 @ ... of 32 bytes
356 add r6, r9, r5
357 add r9, r9, r10
358
3591: ldmdb r6!, {r0 - r3, r10 - r12, lr}
360 cmp r6, r5
361 stmdb r9!, {r0 - r3, r10 - r12, lr}
362 bhi 1b
363
364 /* Preserve offset to relocated code. */
365 sub r6, r9, r6
366
367#ifndef CONFIG_ZBOOT_ROM
368 /* cache_clean_flush may use the stack, so relocate it */
369 add sp, sp, r6
370#endif
371
372 bl cache_clean_flush
373
374 adr r0, BSYM(restart)
375 add r0, r0, r6
376 mov pc, r0
377
378wont_overwrite:
379/*
380 * If delta is zero, we are running at the address we were linked at.
381 * r0 = delta
382 * r2 = BSS start
383 * r3 = BSS end
384 * r4 = kernel execution address
385 * r5 = appended dtb size (0 if not present)
386 * r7 = architecture ID
387 * r8 = atags pointer
388 * r11 = GOT start
389 * r12 = GOT end
390 * sp = stack pointer
391 */
392 orrs r1, r0, r5
393 beq not_relocated
394
395 add r11, r11, r0
396 add r12, r12, r0
397
398#ifndef CONFIG_ZBOOT_ROM
399 /*
400 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
401 * we need to fix up pointers into the BSS region.
402 * Note that the stack pointer has already been fixed up.
403 */
404 add r2, r2, r0
405 add r3, r3, r0
406
407 /*
408 * Relocate all entries in the GOT table.
409 * Bump bss entries to _edata + dtb size
410 */
4111: ldr r1, [r11, #0] @ relocate entries in the GOT
412 add r1, r1, r0 @ This fixes up C references
413 cmp r1, r2 @ if entry >= bss_start &&
414 cmphs r3, r1 @ bss_end > entry
415 addhi r1, r1, r5 @ entry += dtb size
416 str r1, [r11], #4 @ next entry
417 cmp r11, r12
418 blo 1b
419
420 /* bump our bss pointers too */
421 add r2, r2, r5
422 add r3, r3, r5
423
424#else
425
426 /*
427 * Relocate entries in the GOT table. We only relocate
428 * the entries that are outside the (relocated) BSS region.
429 */
4301: ldr r1, [r11, #0] @ relocate entries in the GOT
431 cmp r1, r2 @ entry < bss_start ||
432 cmphs r3, r1 @ _end < entry
433 addlo r1, r1, r0 @ table. This fixes up the
434 str r1, [r11], #4 @ C references.
435 cmp r11, r12
436 blo 1b
437#endif
438
439not_relocated: mov r0, #0
4401: str r0, [r2], #4 @ clear bss
441 str r0, [r2], #4
442 str r0, [r2], #4
443 str r0, [r2], #4
444 cmp r2, r3
445 blo 1b
446
447/*
448 * The C runtime environment should now be setup sufficiently.
449 * Set up some pointers, and start decompressing.
450 * r4 = kernel execution address
451 * r7 = architecture ID
452 * r8 = atags pointer
453 */
454 mov r0, r4
455 mov r1, sp @ malloc space above stack
456 add r2, sp, #0x10000 @ 64k max
457 mov r3, r7
458 bl decompress_kernel
459 bl cache_clean_flush
460 bl cache_off
461 mov r0, #0 @ must be zero
462 mov r1, r7 @ restore architecture number
463 mov r2, r8 @ restore atags pointer
464 ARM( mov pc, r4 ) @ call kernel
465 THUMB( bx r4 ) @ entry point is always ARM
466
467 .align 2
468 .type LC0, #object
469LC0: .word LC0 @ r1
470 .word __bss_start @ r2
471 .word _end @ r3
472 .word _edata @ r6
473 .word input_data_end - 4 @ r10 (inflated size location)
474 .word _got_start @ r11
475 .word _got_end @ ip
476 .word .L_user_stack_end @ sp
477 .size LC0, . - LC0
478
479#ifdef CONFIG_ARCH_RPC
480 .globl params
481params: ldr r0, =0x10000100 @ params_phys for RPC
482 mov pc, lr
483 .ltorg
484 .align
485#endif
486
487/*
488 * Turn on the cache. We need to setup some page tables so that we
489 * can have both the I and D caches on.
490 *
491 * We place the page tables 16k down from the kernel execution address,
492 * and we hope that nothing else is using it. If we're using it, we
493 * will go pop!
494 *
495 * On entry,
496 * r4 = kernel execution address
497 * r7 = architecture number
498 * r8 = atags pointer
499 * On exit,
500 * r0, r1, r2, r3, r9, r10, r12 corrupted
501 * This routine must preserve:
502 * r4, r7, r8
503 */
504 .align 5
505cache_on: mov r3, #8 @ cache_on function
506 b call_cache_fn
507
508/*
509 * Initialize the highest priority protection region, PR7
510 * to cover all 32bit address and cacheable and bufferable.
511 */
512__armv4_mpu_cache_on:
513 mov r0, #0x3f @ 4G, the whole
514 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
515 mcr p15, 0, r0, c6, c7, 1
516
517 mov r0, #0x80 @ PR7
518 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
519 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
520 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
521
522 mov r0, #0xc000
523 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
524 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
525
526 mov r0, #0
527 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
528 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
529 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
530 mrc p15, 0, r0, c1, c0, 0 @ read control reg
531 @ ...I .... ..D. WC.M
532 orr r0, r0, #0x002d @ .... .... ..1. 11.1
533 orr r0, r0, #0x1000 @ ...1 .... .... ....
534
535 mcr p15, 0, r0, c1, c0, 0 @ write control reg
536
537 mov r0, #0
538 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
539 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
540 mov pc, lr
541
542__armv3_mpu_cache_on:
543 mov r0, #0x3f @ 4G, the whole
544 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
545
546 mov r0, #0x80 @ PR7
547 mcr p15, 0, r0, c2, c0, 0 @ cache on
548 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
549
550 mov r0, #0xc000
551 mcr p15, 0, r0, c5, c0, 0 @ access permission
552
553 mov r0, #0
554 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
555 /*
556 * ?? ARMv3 MMU does not allow reading the control register,
557 * does this really work on ARMv3 MPU?
558 */
559 mrc p15, 0, r0, c1, c0, 0 @ read control reg
560 @ .... .... .... WC.M
561 orr r0, r0, #0x000d @ .... .... .... 11.1
562 /* ?? this overwrites the value constructed above? */
563 mov r0, #0
564 mcr p15, 0, r0, c1, c0, 0 @ write control reg
565
566 /* ?? invalidate for the second time? */
567 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
568 mov pc, lr
569
570#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
571#define CB_BITS 0x08
572#else
573#define CB_BITS 0x0c
574#endif
575
576__setup_mmu: sub r3, r4, #16384 @ Page directory size
577 bic r3, r3, #0xff @ Align the pointer
578 bic r3, r3, #0x3f00
579/*
580 * Initialise the page tables, turning on the cacheable and bufferable
581 * bits for the RAM area only.
582 */
583 mov r0, r3
584 mov r9, r0, lsr #18
585 mov r9, r9, lsl #18 @ start of RAM
586 add r10, r9, #0x10000000 @ a reasonable RAM size
587 mov r1, #0x12 @ XN|U + section mapping
588 orr r1, r1, #3 << 10 @ AP=11
589 add r2, r3, #16384
5901: cmp r1, r9 @ if virt > start of RAM
591 cmphs r10, r1 @ && end of RAM > virt
592 bic r1, r1, #0x1c @ clear XN|U + C + B
593 orrlo r1, r1, #0x10 @ Set XN|U for non-RAM
594 orrhs r1, r1, r6 @ set RAM section settings
595 str r1, [r0], #4 @ 1:1 mapping
596 add r1, r1, #1048576
597 teq r0, r2
598 bne 1b
599/*
600 * If ever we are running from Flash, then we surely want the cache
601 * to be enabled also for our execution instance... We map 2MB of it
602 * so there is no map overlap problem for up to 1 MB compressed kernel.
603 * If the execution is in RAM then we would only be duplicating the above.
604 */
605 orr r1, r6, #0x04 @ ensure B is set for this
606 orr r1, r1, #3 << 10
607 mov r2, pc
608 mov r2, r2, lsr #20
609 orr r1, r1, r2, lsl #20
610 add r0, r3, r2, lsl #2
611 str r1, [r0], #4
612 add r1, r1, #1048576
613 str r1, [r0]
614 mov pc, lr
615ENDPROC(__setup_mmu)
616
617__arm926ejs_mmu_cache_on:
618#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
619 mov r0, #4 @ put dcache in WT mode
620 mcr p15, 7, r0, c15, c0, 0
621#endif
622
623__armv4_mmu_cache_on:
624 mov r12, lr
625#ifdef CONFIG_MMU
626 mov r6, #CB_BITS | 0x12 @ U
627 bl __setup_mmu
628 mov r0, #0
629 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
630 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
631 mrc p15, 0, r0, c1, c0, 0 @ read control reg
632 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
633 orr r0, r0, #0x0030
634#ifdef CONFIG_CPU_ENDIAN_BE8
635 orr r0, r0, #1 << 25 @ big-endian page tables
636#endif
637 bl __common_mmu_cache_on
638 mov r0, #0
639 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
640#endif
641 mov pc, r12
642
643__armv7_mmu_cache_on:
644 mov r12, lr
645#ifdef CONFIG_MMU
646 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
647 tst r11, #0xf @ VMSA
648 movne r6, #CB_BITS | 0x02 @ !XN
649 blne __setup_mmu
650 mov r0, #0
651 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
652 tst r11, #0xf @ VMSA
653 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
654#endif
655 mrc p15, 0, r0, c1, c0, 0 @ read control reg
656 bic r0, r0, #1 << 28 @ clear SCTLR.TRE
657 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
658 orr r0, r0, #0x003c @ write buffer
659#ifdef CONFIG_MMU
660#ifdef CONFIG_CPU_ENDIAN_BE8
661 orr r0, r0, #1 << 25 @ big-endian page tables
662#endif
663 mrcne p15, 0, r6, c2, c0, 2 @ read ttb control reg
664 orrne r0, r0, #1 @ MMU enabled
665 movne r1, #0xfffffffd @ domain 0 = client
666 bic r6, r6, #1 << 31 @ 32-bit translation system
667 bic r6, r6, #3 << 0 @ use only ttbr0
668 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
669 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
670 mcrne p15, 0, r6, c2, c0, 2 @ load ttb control
671#endif
672 mcr p15, 0, r0, c7, c5, 4 @ ISB
673 mcr p15, 0, r0, c1, c0, 0 @ load control register
674 mrc p15, 0, r0, c1, c0, 0 @ and read it back
675 mov r0, #0
676 mcr p15, 0, r0, c7, c5, 4 @ ISB
677 mov pc, r12
678
679__fa526_cache_on:
680 mov r12, lr
681 mov r6, #CB_BITS | 0x12 @ U
682 bl __setup_mmu
683 mov r0, #0
684 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
685 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
686 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
687 mrc p15, 0, r0, c1, c0, 0 @ read control reg
688 orr r0, r0, #0x1000 @ I-cache enable
689 bl __common_mmu_cache_on
690 mov r0, #0
691 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
692 mov pc, r12
693
694__common_mmu_cache_on:
695#ifndef CONFIG_THUMB2_KERNEL
696#ifndef DEBUG
697 orr r0, r0, #0x000d @ Write buffer, mmu
698#endif
699 mov r1, #-1
700 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
701 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
702 b 1f
703 .align 5 @ cache line aligned
7041: mcr p15, 0, r0, c1, c0, 0 @ load control register
705 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
706 sub pc, lr, r0, lsr #32 @ properly flush pipeline
707#endif
708
709#define PROC_ENTRY_SIZE (4*5)
710
711/*
712 * Here follow the relocatable cache support functions for the
713 * various processors. This is a generic hook for locating an
714 * entry and jumping to an instruction at the specified offset
715 * from the start of the block. Please note this is all position
716 * independent code.
717 *
718 * r1 = corrupted
719 * r2 = corrupted
720 * r3 = block offset
721 * r9 = corrupted
722 * r12 = corrupted
723 */
724
725call_cache_fn: adr r12, proc_types
726#ifdef CONFIG_CPU_CP15
727 mrc p15, 0, r9, c0, c0 @ get processor ID
728#else
729 ldr r9, =CONFIG_PROCESSOR_ID
730#endif
7311: ldr r1, [r12, #0] @ get value
732 ldr r2, [r12, #4] @ get mask
733 eor r1, r1, r9 @ (real ^ match)
734 tst r1, r2 @ & mask
735 ARM( addeq pc, r12, r3 ) @ call cache function
736 THUMB( addeq r12, r3 )
737 THUMB( moveq pc, r12 ) @ call cache function
738 add r12, r12, #PROC_ENTRY_SIZE
739 b 1b
740
741/*
742 * Table for cache operations. This is basically:
743 * - CPU ID match
744 * - CPU ID mask
745 * - 'cache on' method instruction
746 * - 'cache off' method instruction
747 * - 'cache flush' method instruction
748 *
749 * We match an entry using: ((real_id ^ match) & mask) == 0
750 *
751 * Writethrough caches generally only need 'on' and 'off'
752 * methods. Writeback caches _must_ have the flush method
753 * defined.
754 */
755 .align 2
756 .type proc_types,#object
757proc_types:
758 .word 0x00000000 @ old ARM ID
759 .word 0x0000f000
760 mov pc, lr
761 THUMB( nop )
762 mov pc, lr
763 THUMB( nop )
764 mov pc, lr
765 THUMB( nop )
766
767 .word 0x41007000 @ ARM7/710
768 .word 0xfff8fe00
769 mov pc, lr
770 THUMB( nop )
771 mov pc, lr
772 THUMB( nop )
773 mov pc, lr
774 THUMB( nop )
775
776 .word 0x41807200 @ ARM720T (writethrough)
777 .word 0xffffff00
778 W(b) __armv4_mmu_cache_on
779 W(b) __armv4_mmu_cache_off
780 mov pc, lr
781 THUMB( nop )
782
783 .word 0x41007400 @ ARM74x
784 .word 0xff00ff00
785 W(b) __armv3_mpu_cache_on
786 W(b) __armv3_mpu_cache_off
787 W(b) __armv3_mpu_cache_flush
788
789 .word 0x41009400 @ ARM94x
790 .word 0xff00ff00
791 W(b) __armv4_mpu_cache_on
792 W(b) __armv4_mpu_cache_off
793 W(b) __armv4_mpu_cache_flush
794
795 .word 0x41069260 @ ARM926EJ-S (v5TEJ)
796 .word 0xff0ffff0
797 W(b) __arm926ejs_mmu_cache_on
798 W(b) __armv4_mmu_cache_off
799 W(b) __armv5tej_mmu_cache_flush
800
801 .word 0x00007000 @ ARM7 IDs
802 .word 0x0000f000
803 mov pc, lr
804 THUMB( nop )
805 mov pc, lr
806 THUMB( nop )
807 mov pc, lr
808 THUMB( nop )
809
810 @ Everything from here on will be the new ID system.
811
812 .word 0x4401a100 @ sa110 / sa1100
813 .word 0xffffffe0
814 W(b) __armv4_mmu_cache_on
815 W(b) __armv4_mmu_cache_off
816 W(b) __armv4_mmu_cache_flush
817
818 .word 0x6901b110 @ sa1110
819 .word 0xfffffff0
820 W(b) __armv4_mmu_cache_on
821 W(b) __armv4_mmu_cache_off
822 W(b) __armv4_mmu_cache_flush
823
824 .word 0x56056900
825 .word 0xffffff00 @ PXA9xx
826 W(b) __armv4_mmu_cache_on
827 W(b) __armv4_mmu_cache_off
828 W(b) __armv4_mmu_cache_flush
829
830 .word 0x56158000 @ PXA168
831 .word 0xfffff000
832 W(b) __armv4_mmu_cache_on
833 W(b) __armv4_mmu_cache_off
834 W(b) __armv5tej_mmu_cache_flush
835
836 .word 0x56050000 @ Feroceon
837 .word 0xff0f0000
838 W(b) __armv4_mmu_cache_on
839 W(b) __armv4_mmu_cache_off
840 W(b) __armv5tej_mmu_cache_flush
841
842#ifdef CONFIG_CPU_FEROCEON_OLD_ID
843 /* this conflicts with the standard ARMv5TE entry */
844 .long 0x41009260 @ Old Feroceon
845 .long 0xff00fff0
846 b __armv4_mmu_cache_on
847 b __armv4_mmu_cache_off
848 b __armv5tej_mmu_cache_flush
849#endif
850
851 .word 0x66015261 @ FA526
852 .word 0xff01fff1
853 W(b) __fa526_cache_on
854 W(b) __armv4_mmu_cache_off
855 W(b) __fa526_cache_flush
856
857 @ These match on the architecture ID
858
859 .word 0x00020000 @ ARMv4T
860 .word 0x000f0000
861 W(b) __armv4_mmu_cache_on
862 W(b) __armv4_mmu_cache_off
863 W(b) __armv4_mmu_cache_flush
864
865 .word 0x00050000 @ ARMv5TE
866 .word 0x000f0000
867 W(b) __armv4_mmu_cache_on
868 W(b) __armv4_mmu_cache_off
869 W(b) __armv4_mmu_cache_flush
870
871 .word 0x00060000 @ ARMv5TEJ
872 .word 0x000f0000
873 W(b) __armv4_mmu_cache_on
874 W(b) __armv4_mmu_cache_off
875 W(b) __armv5tej_mmu_cache_flush
876
877 .word 0x0007b000 @ ARMv6
878 .word 0x000ff000
879 W(b) __armv4_mmu_cache_on
880 W(b) __armv4_mmu_cache_off
881 W(b) __armv6_mmu_cache_flush
882
883 .word 0x000f0000 @ new CPU Id
884 .word 0x000f0000
885 W(b) __armv7_mmu_cache_on
886 W(b) __armv7_mmu_cache_off
887 W(b) __armv7_mmu_cache_flush
888
889 .word 0 @ unrecognised type
890 .word 0
891 mov pc, lr
892 THUMB( nop )
893 mov pc, lr
894 THUMB( nop )
895 mov pc, lr
896 THUMB( nop )
897
898 .size proc_types, . - proc_types
899
900 /*
901 * If you get a "non-constant expression in ".if" statement"
902 * error from the assembler on this line, check that you have
903 * not accidentally written a "b" instruction where you should
904 * have written W(b).
905 */
906 .if (. - proc_types) % PROC_ENTRY_SIZE != 0
907 .error "The size of one or more proc_types entries is wrong."
908 .endif
909
910/*
911 * Turn off the Cache and MMU. ARMv3 does not support
912 * reading the control register, but ARMv4 does.
913 *
914 * On exit,
915 * r0, r1, r2, r3, r9, r12 corrupted
916 * This routine must preserve:
917 * r4, r7, r8
918 */
919 .align 5
920cache_off: mov r3, #12 @ cache_off function
921 b call_cache_fn
922
923__armv4_mpu_cache_off:
924 mrc p15, 0, r0, c1, c0
925 bic r0, r0, #0x000d
926 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
927 mov r0, #0
928 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
929 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
930 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
931 mov pc, lr
932
933__armv3_mpu_cache_off:
934 mrc p15, 0, r0, c1, c0
935 bic r0, r0, #0x000d
936 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
937 mov r0, #0
938 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
939 mov pc, lr
940
941__armv4_mmu_cache_off:
942#ifdef CONFIG_MMU
943 mrc p15, 0, r0, c1, c0
944 bic r0, r0, #0x000d
945 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
946 mov r0, #0
947 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
948 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
949#endif
950 mov pc, lr
951
952__armv7_mmu_cache_off:
953 mrc p15, 0, r0, c1, c0
954#ifdef CONFIG_MMU
955 bic r0, r0, #0x000d
956#else
957 bic r0, r0, #0x000c
958#endif
959 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
960 mov r12, lr
961 bl __armv7_mmu_cache_flush
962 mov r0, #0
963#ifdef CONFIG_MMU
964 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
965#endif
966 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
967 mcr p15, 0, r0, c7, c10, 4 @ DSB
968 mcr p15, 0, r0, c7, c5, 4 @ ISB
969 mov pc, r12
970
971/*
972 * Clean and flush the cache to maintain consistency.
973 *
974 * On exit,
975 * r1, r2, r3, r9, r10, r11, r12 corrupted
976 * This routine must preserve:
977 * r4, r6, r7, r8
978 */
979 .align 5
980cache_clean_flush:
981 mov r3, #16
982 b call_cache_fn
983
984__armv4_mpu_cache_flush:
985 mov r2, #1
986 mov r3, #0
987 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
988 mov r1, #7 << 5 @ 8 segments
9891: orr r3, r1, #63 << 26 @ 64 entries
9902: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
991 subs r3, r3, #1 << 26
992 bcs 2b @ entries 63 to 0
993 subs r1, r1, #1 << 5
994 bcs 1b @ segments 7 to 0
995
996 teq r2, #0
997 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
998 mcr p15, 0, ip, c7, c10, 4 @ drain WB
999 mov pc, lr
1000
1001__fa526_cache_flush:
1002 mov r1, #0
1003 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
1004 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1005 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1006 mov pc, lr
1007
1008__armv6_mmu_cache_flush:
1009 mov r1, #0
1010 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
1011 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
1012 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
1013 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1014 mov pc, lr
1015
1016__armv7_mmu_cache_flush:
1017 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
1018 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
1019 mov r10, #0
1020 beq hierarchical
1021 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
1022 b iflush
1023hierarchical:
1024 mcr p15, 0, r10, c7, c10, 5 @ DMB
1025 stmfd sp!, {r0-r7, r9-r11}
1026 mrc p15, 1, r0, c0, c0, 1 @ read clidr
1027 ands r3, r0, #0x7000000 @ extract loc from clidr
1028 mov r3, r3, lsr #23 @ left align loc bit field
1029 beq finished @ if loc is 0, then no need to clean
1030 mov r10, #0 @ start clean at cache level 0
1031loop1:
1032 add r2, r10, r10, lsr #1 @ work out 3x current cache level
1033 mov r1, r0, lsr r2 @ extract cache type bits from clidr
1034 and r1, r1, #7 @ mask of the bits for current cache only
1035 cmp r1, #2 @ see what cache we have at this level
1036 blt skip @ skip if no cache, or just i-cache
1037 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1038 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
1039 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
1040 and r2, r1, #7 @ extract the length of the cache lines
1041 add r2, r2, #4 @ add 4 (line length offset)
1042 ldr r4, =0x3ff
1043 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
1044 clz r5, r4 @ find bit position of way size increment
1045 ldr r7, =0x7fff
1046 ands r7, r7, r1, lsr #13 @ extract max number of the index size
1047loop2:
1048 mov r9, r4 @ create working copy of max way size
1049loop3:
1050 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
1051 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
1052 THUMB( lsl r6, r9, r5 )
1053 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
1054 THUMB( lsl r6, r7, r2 )
1055 THUMB( orr r11, r11, r6 ) @ factor index number into r11
1056 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
1057 subs r9, r9, #1 @ decrement the way
1058 bge loop3
1059 subs r7, r7, #1 @ decrement the index
1060 bge loop2
1061skip:
1062 add r10, r10, #2 @ increment cache number
1063 cmp r3, r10
1064 bgt loop1
1065finished:
1066 ldmfd sp!, {r0-r7, r9-r11}
1067 mov r10, #0 @ swith back to cache level 0
1068 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1069iflush:
1070 mcr p15, 0, r10, c7, c10, 4 @ DSB
1071 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
1072 mcr p15, 0, r10, c7, c10, 4 @ DSB
1073 mcr p15, 0, r10, c7, c5, 4 @ ISB
1074 mov pc, lr
1075
1076__armv5tej_mmu_cache_flush:
10771: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
1078 bne 1b
1079 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
1080 mcr p15, 0, r0, c7, c10, 4 @ drain WB
1081 mov pc, lr
1082
1083__armv4_mmu_cache_flush:
1084 mov r2, #64*1024 @ default: 32K dcache size (*2)
1085 mov r11, #32 @ default: 32 byte line size
1086 mrc p15, 0, r3, c0, c0, 1 @ read cache type
1087 teq r3, r9 @ cache ID register present?
1088 beq no_cache_id
1089 mov r1, r3, lsr #18
1090 and r1, r1, #7
1091 mov r2, #1024
1092 mov r2, r2, lsl r1 @ base dcache size *2
1093 tst r3, #1 << 14 @ test M bit
1094 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
1095 mov r3, r3, lsr #12
1096 and r3, r3, #3
1097 mov r11, #8
1098 mov r11, r11, lsl r3 @ cache line size in bytes
1099no_cache_id:
1100 mov r1, pc
1101 bic r1, r1, #63 @ align to longest cache line
1102 add r2, r1, r2
11031:
1104 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
1105 THUMB( ldr r3, [r1] ) @ s/w flush D cache
1106 THUMB( add r1, r1, r11 )
1107 teq r1, r2
1108 bne 1b
1109
1110 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1111 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
1112 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1113 mov pc, lr
1114
1115__armv3_mmu_cache_flush:
1116__armv3_mpu_cache_flush:
1117 mov r1, #0
1118 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
1119 mov pc, lr
1120
1121/*
1122 * Various debugging routines for printing hex characters and
1123 * memory, which again must be relocatable.
1124 */
1125#ifdef DEBUG
1126 .align 2
1127 .type phexbuf,#object
1128phexbuf: .space 12
1129 .size phexbuf, . - phexbuf
1130
1131@ phex corrupts {r0, r1, r2, r3}
1132phex: adr r3, phexbuf
1133 mov r2, #0
1134 strb r2, [r3, r1]
11351: subs r1, r1, #1
1136 movmi r0, r3
1137 bmi puts
1138 and r2, r0, #15
1139 mov r0, r0, lsr #4
1140 cmp r2, #10
1141 addge r2, r2, #7
1142 add r2, r2, #'0'
1143 strb r2, [r3, r1]
1144 b 1b
1145
1146@ puts corrupts {r0, r1, r2, r3}
1147puts: loadsp r3, r1
11481: ldrb r2, [r0], #1
1149 teq r2, #0
1150 moveq pc, lr
11512: writeb r2, r3
1152 mov r1, #0x00020000
11533: subs r1, r1, #1
1154 bne 3b
1155 teq r2, #'\n'
1156 moveq r2, #'\r'
1157 beq 2b
1158 teq r0, #0
1159 bne 1b
1160 mov pc, lr
1161@ putc corrupts {r0, r1, r2, r3}
1162putc:
1163 mov r2, r0
1164 mov r0, #0
1165 loadsp r3, r1
1166 b 2b
1167
1168@ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1169memdump: mov r12, r0
1170 mov r10, lr
1171 mov r11, #0
11722: mov r0, r11, lsl #2
1173 add r0, r0, r12
1174 mov r1, #8
1175 bl phex
1176 mov r0, #':'
1177 bl putc
11781: mov r0, #' '
1179 bl putc
1180 ldr r0, [r12, r11, lsl #2]
1181 mov r1, #8
1182 bl phex
1183 and r0, r11, #7
1184 teq r0, #3
1185 moveq r0, #' '
1186 bleq putc
1187 and r0, r11, #7
1188 add r11, r11, #1
1189 teq r0, #7
1190 bne 1b
1191 mov r0, #'\n'
1192 bl putc
1193 cmp r11, #64
1194 blt 2b
1195 mov pc, r10
1196#endif
1197
1198 .ltorg
1199reloc_code_end:
1200
1201 .align
1202 .section ".stack", "aw", %nobits
1203.L_user_stack: .space 4096
1204.L_user_stack_end: