| blob | 15bc187de14ac8ce738128f2219f044713c269e3 |
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>
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
22 #if defined(CONFIG_DEBUG_ICEDCC)
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
44 #else
46 #include <mach/debug-macro.S>
48 .macro writeb, ch, rb
49 senduart \ch, \rb
50 .endm
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
74 .macro kputc,val
75 mov r0, \val
76 bl putc
77 .endm
79 .macro kphex,val,len
80 mov r0, \val
81 mov r1, #\len
82 bl phex
83 .endm
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
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
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 )
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
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
158 /*
159 * Note that some cache flushing and other stuff may
160 * be needed here - is there an Angel SWI call for this?
161 */
163 /*
164 * some architecture specific code can be inserted
165 * by the linker here, but it should preserve r7, r8, and r9.
166 */
168 .text
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
179 bl cache_on
181 restart: adr r0, LC0
182 ldmia r0, {r1, r2, r3, r6, r10, r11, r12}
183 ldr sp, [r0, #28]
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
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
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
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 */
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
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
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
278 ldmfd sp!, {r0-r3, ip, lr}
279 sub sp, sp, #0x10000
280 #endif
282 mov r8, r6 @ use the appended device tree
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
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
306 /* preserve 64-bit alignment */
307 add r5, r5, #7
308 bic r5, r5, #7
310 /* relocate some pointers past the appended dtb */
311 add r6, r6, r5
312 add r10, r10, r5
313 add sp, sp, r5
314 dtb_check_done:
315 #endif
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
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
349 /* Get start of code we want to copy and align it down. */
350 adr r5, restart
351 bic r5, r5, #31
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
359 1: ldmdb r6!, {r0 - r3, r10 - r12, lr}
360 cmp r6, r5
361 stmdb r9!, {r0 - r3, r10 - r12, lr}
362 bhi 1b
364 /* Preserve offset to relocated code. */
365 sub r6, r9, r6
367 #ifndef CONFIG_ZBOOT_ROM
368 /* cache_clean_flush may use the stack, so relocate it */
369 add sp, sp, r6
370 #endif
372 bl cache_clean_flush
374 adr r0, BSYM(restart)
375 add r0, r0, r6
376 mov pc, r0
378 wont_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
395 add r11, r11, r0
396 add r12, r12, r0
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
407 /*
408 * Relocate all entries in the GOT table.
409 * Bump bss entries to _edata + dtb size
410 */
411 1: 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
420 /* bump our bss pointers too */
421 add r2, r2, r5
422 add r3, r3, r5
424 #else
426 /*
427 * Relocate entries in the GOT table. We only relocate
428 * the entries that are outside the (relocated) BSS region.
429 */
430 1: 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
439 not_relocated: mov r0, #0
440 1: 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
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
467 .align 2
468 .type LC0, #object
469 LC0: .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
479 #ifdef CONFIG_ARCH_RPC
480 .globl params
481 params: ldr r0, =0x10000100 @ params_phys for RPC
482 mov pc, lr
483 .ltorg
484 .align
485 #endif
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
505 cache_on: mov r3, #8 @ cache_on function
506 b call_cache_fn
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
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
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
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 .... .... ....
535 mcr p15, 0, r0, c1, c0, 0 @ write control reg
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
542 __armv3_mpu_cache_on:
543 mov r0, #0x3f @ 4G, the whole
544 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
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
550 mov r0, #0xc000
551 mcr p15, 0, r0, c5, c0, 0 @ access permission
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
566 /* ?? invalidate for the second time? */
567 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
568 mov pc, lr
570 __setup_mmu: sub r3, r4, #16384 @ Page directory size
571 bic r3, r3, #0xff @ Align the pointer
572 bic r3, r3, #0x3f00
573 /*
574 * Initialise the page tables, turning on the cacheable and bufferable
575 * bits for the RAM area only.
576 */
577 mov r0, r3
578 mov r9, r0, lsr #18
579 mov r9, r9, lsl #18 @ start of RAM
580 add r10, r9, #0x10000000 @ a reasonable RAM size
581 mov r1, #0x12
582 orr r1, r1, #3 << 10
583 add r2, r3, #16384
584 1: cmp r1, r9 @ if virt > start of RAM
585 #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
586 orrhs r1, r1, #0x08 @ set cacheable
587 #else
588 orrhs r1, r1, #0x0c @ set cacheable, bufferable
589 #endif
590 cmp r1, r10 @ if virt > end of RAM
591 bichs r1, r1, #0x0c @ clear cacheable, bufferable
592 str r1, [r0], #4 @ 1:1 mapping
593 add r1, r1, #1048576
594 teq r0, r2
595 bne 1b
596 /*
597 * If ever we are running from Flash, then we surely want the cache
598 * to be enabled also for our execution instance... We map 2MB of it
599 * so there is no map overlap problem for up to 1 MB compressed kernel.
600 * If the execution is in RAM then we would only be duplicating the above.
601 */
602 mov r1, #0x1e
603 orr r1, r1, #3 << 10
604 mov r2, pc
605 mov r2, r2, lsr #20
606 orr r1, r1, r2, lsl #20
607 add r0, r3, r2, lsl #2
608 str r1, [r0], #4
609 add r1, r1, #1048576
610 str r1, [r0]
611 mov pc, lr
612 ENDPROC(__setup_mmu)
614 __arm926ejs_mmu_cache_on:
615 #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
616 mov r0, #4 @ put dcache in WT mode
617 mcr p15, 7, r0, c15, c0, 0
618 #endif
620 __armv4_mmu_cache_on:
621 mov r12, lr
622 #ifdef CONFIG_MMU
623 bl __setup_mmu
624 mov r0, #0
625 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
626 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
627 mrc p15, 0, r0, c1, c0, 0 @ read control reg
628 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
629 orr r0, r0, #0x0030
630 #ifdef CONFIG_CPU_ENDIAN_BE8
631 orr r0, r0, #1 << 25 @ big-endian page tables
632 #endif
633 bl __common_mmu_cache_on
634 mov r0, #0
635 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
636 #endif
637 mov pc, r12
639 __armv7_mmu_cache_on:
640 mov r12, lr
641 #ifdef CONFIG_MMU
642 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
643 tst r11, #0xf @ VMSA
644 blne __setup_mmu
645 mov r0, #0
646 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
647 tst r11, #0xf @ VMSA
648 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
649 #endif
650 mrc p15, 0, r0, c1, c0, 0 @ read control reg
651 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
652 orr r0, r0, #0x003c @ write buffer
653 #ifdef CONFIG_MMU
654 #ifdef CONFIG_CPU_ENDIAN_BE8
655 orr r0, r0, #1 << 25 @ big-endian page tables
656 #endif
657 orrne r0, r0, #1 @ MMU enabled
658 movne r1, #-1
659 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
660 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
661 #endif
662 mcr p15, 0, r0, c7, c5, 4 @ ISB
663 mcr p15, 0, r0, c1, c0, 0 @ load control register
664 mrc p15, 0, r0, c1, c0, 0 @ and read it back
665 mov r0, #0
666 mcr p15, 0, r0, c7, c5, 4 @ ISB
667 mov pc, r12
669 __fa526_cache_on:
670 mov r12, lr
671 bl __setup_mmu
672 mov r0, #0
673 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
674 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
675 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
676 mrc p15, 0, r0, c1, c0, 0 @ read control reg
677 orr r0, r0, #0x1000 @ I-cache enable
678 bl __common_mmu_cache_on
679 mov r0, #0
680 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
681 mov pc, r12
683 __arm6_mmu_cache_on:
684 mov r12, lr
685 bl __setup_mmu
686 mov r0, #0
687 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
688 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
689 mov r0, #0x30
690 bl __common_mmu_cache_on
691 mov r0, #0
692 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
693 mov pc, r12
695 __common_mmu_cache_on:
696 #ifndef CONFIG_THUMB2_KERNEL
697 #ifndef DEBUG
698 orr r0, r0, #0x000d @ Write buffer, mmu
699 #endif
700 mov r1, #-1
701 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
702 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
703 b 1f
704 .align 5 @ cache line aligned
705 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
706 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
707 sub pc, lr, r0, lsr #32 @ properly flush pipeline
708 #endif
710 #define PROC_ENTRY_SIZE (4*5)
712 /*
713 * Here follow the relocatable cache support functions for the
714 * various processors. This is a generic hook for locating an
715 * entry and jumping to an instruction at the specified offset
716 * from the start of the block. Please note this is all position
717 * independent code.
718 *
719 * r1 = corrupted
720 * r2 = corrupted
721 * r3 = block offset
722 * r9 = corrupted
723 * r12 = corrupted
724 */
726 call_cache_fn: adr r12, proc_types
727 #ifdef CONFIG_CPU_CP15
728 mrc p15, 0, r9, c0, c0 @ get processor ID
729 #else
730 ldr r9, =CONFIG_PROCESSOR_ID
731 #endif
732 1: ldr r1, [r12, #0] @ get value
733 ldr r2, [r12, #4] @ get mask
734 eor r1, r1, r9 @ (real ^ match)
735 tst r1, r2 @ & mask
736 ARM( addeq pc, r12, r3 ) @ call cache function
737 THUMB( addeq r12, r3 )
738 THUMB( moveq pc, r12 ) @ call cache function
739 add r12, r12, #PROC_ENTRY_SIZE
740 b 1b
742 /*
743 * Table for cache operations. This is basically:
744 * - CPU ID match
745 * - CPU ID mask
746 * - 'cache on' method instruction
747 * - 'cache off' method instruction
748 * - 'cache flush' method instruction
749 *
750 * We match an entry using: ((real_id ^ match) & mask) == 0
751 *
752 * Writethrough caches generally only need 'on' and 'off'
753 * methods. Writeback caches _must_ have the flush method
754 * defined.
755 */
756 .align 2
757 .type proc_types,#object
758 proc_types:
759 .word 0x41560600 @ ARM6/610
760 .word 0xffffffe0
761 W(b) __arm6_mmu_cache_off @ works, but slow
762 W(b) __arm6_mmu_cache_off
763 mov pc, lr
764 THUMB( nop )
765 @ b __arm6_mmu_cache_on @ untested
766 @ b __arm6_mmu_cache_off
767 @ b __armv3_mmu_cache_flush
769 .word 0x00000000 @ old ARM ID
770 .word 0x0000f000
771 mov pc, lr
772 THUMB( nop )
773 mov pc, lr
774 THUMB( nop )
775 mov pc, lr
776 THUMB( nop )
778 .word 0x41007000 @ ARM7/710
779 .word 0xfff8fe00
780 W(b) __arm7_mmu_cache_off
781 W(b) __arm7_mmu_cache_off
782 mov pc, lr
783 THUMB( nop )
785 .word 0x41807200 @ ARM720T (writethrough)
786 .word 0xffffff00
787 W(b) __armv4_mmu_cache_on
788 W(b) __armv4_mmu_cache_off
789 mov pc, lr
790 THUMB( nop )
792 .word 0x41007400 @ ARM74x
793 .word 0xff00ff00
794 W(b) __armv3_mpu_cache_on
795 W(b) __armv3_mpu_cache_off
796 W(b) __armv3_mpu_cache_flush
798 .word 0x41009400 @ ARM94x
799 .word 0xff00ff00
800 W(b) __armv4_mpu_cache_on
801 W(b) __armv4_mpu_cache_off
802 W(b) __armv4_mpu_cache_flush
804 .word 0x41069260 @ ARM926EJ-S (v5TEJ)
805 .word 0xff0ffff0
806 W(b) __arm926ejs_mmu_cache_on
807 W(b) __armv4_mmu_cache_off
808 W(b) __armv5tej_mmu_cache_flush
810 .word 0x00007000 @ ARM7 IDs
811 .word 0x0000f000
812 mov pc, lr
813 THUMB( nop )
814 mov pc, lr
815 THUMB( nop )
816 mov pc, lr
817 THUMB( nop )
819 @ Everything from here on will be the new ID system.
821 .word 0x4401a100 @ sa110 / sa1100
822 .word 0xffffffe0
823 W(b) __armv4_mmu_cache_on
824 W(b) __armv4_mmu_cache_off
825 W(b) __armv4_mmu_cache_flush
827 .word 0x6901b110 @ sa1110
828 .word 0xfffffff0
829 W(b) __armv4_mmu_cache_on
830 W(b) __armv4_mmu_cache_off
831 W(b) __armv4_mmu_cache_flush
833 .word 0x56056900
834 .word 0xffffff00 @ PXA9xx
835 W(b) __armv4_mmu_cache_on
836 W(b) __armv4_mmu_cache_off
837 W(b) __armv4_mmu_cache_flush
839 .word 0x56158000 @ PXA168
840 .word 0xfffff000
841 W(b) __armv4_mmu_cache_on
842 W(b) __armv4_mmu_cache_off
843 W(b) __armv5tej_mmu_cache_flush
845 .word 0x56050000 @ Feroceon
846 .word 0xff0f0000
847 W(b) __armv4_mmu_cache_on
848 W(b) __armv4_mmu_cache_off
849 W(b) __armv5tej_mmu_cache_flush
851 #ifdef CONFIG_CPU_FEROCEON_OLD_ID
852 /* this conflicts with the standard ARMv5TE entry */
853 .long 0x41009260 @ Old Feroceon
854 .long 0xff00fff0
855 b __armv4_mmu_cache_on
856 b __armv4_mmu_cache_off
857 b __armv5tej_mmu_cache_flush
858 #endif
860 .word 0x66015261 @ FA526
861 .word 0xff01fff1
862 W(b) __fa526_cache_on
863 W(b) __armv4_mmu_cache_off
864 W(b) __fa526_cache_flush
866 @ These match on the architecture ID
868 .word 0x00020000 @ ARMv4T
869 .word 0x000f0000
870 W(b) __armv4_mmu_cache_on
871 W(b) __armv4_mmu_cache_off
872 W(b) __armv4_mmu_cache_flush
874 .word 0x00050000 @ ARMv5TE
875 .word 0x000f0000
876 W(b) __armv4_mmu_cache_on
877 W(b) __armv4_mmu_cache_off
878 W(b) __armv4_mmu_cache_flush
880 .word 0x00060000 @ ARMv5TEJ
881 .word 0x000f0000
882 W(b) __armv4_mmu_cache_on
883 W(b) __armv4_mmu_cache_off
884 W(b) __armv5tej_mmu_cache_flush
886 .word 0x0007b000 @ ARMv6
887 .word 0x000ff000
888 W(b) __armv4_mmu_cache_on
889 W(b) __armv4_mmu_cache_off
890 W(b) __armv6_mmu_cache_flush
892 .word 0x000f0000 @ new CPU Id
893 .word 0x000f0000
894 W(b) __armv7_mmu_cache_on
895 W(b) __armv7_mmu_cache_off
896 W(b) __armv7_mmu_cache_flush
898 .word 0 @ unrecognised type
899 .word 0
900 mov pc, lr
901 THUMB( nop )
902 mov pc, lr
903 THUMB( nop )
904 mov pc, lr
905 THUMB( nop )
907 .size proc_types, . - proc_types
909 /*
910 * If you get a "non-constant expression in ".if" statement"
911 * error from the assembler on this line, check that you have
912 * not accidentally written a "b" instruction where you should
913 * have written W(b).
914 */
915 .if (. - proc_types) % PROC_ENTRY_SIZE != 0
916 .error "The size of one or more proc_types entries is wrong."
917 .endif
919 /*
920 * Turn off the Cache and MMU. ARMv3 does not support
921 * reading the control register, but ARMv4 does.
922 *
923 * On exit,
924 * r0, r1, r2, r3, r9, r12 corrupted
925 * This routine must preserve:
926 * r4, r7, r8
927 */
928 .align 5
929 cache_off: mov r3, #12 @ cache_off function
930 b call_cache_fn
932 __armv4_mpu_cache_off:
933 mrc p15, 0, r0, c1, c0
934 bic r0, r0, #0x000d
935 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
936 mov r0, #0
937 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
938 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
939 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
940 mov pc, lr
942 __armv3_mpu_cache_off:
943 mrc p15, 0, r0, c1, c0
944 bic r0, r0, #0x000d
945 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
946 mov r0, #0
947 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
948 mov pc, lr
950 __armv4_mmu_cache_off:
951 #ifdef CONFIG_MMU
952 mrc p15, 0, r0, c1, c0
953 bic r0, r0, #0x000d
954 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
955 mov r0, #0
956 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
957 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
958 #endif
959 mov pc, lr
961 __armv7_mmu_cache_off:
962 mrc p15, 0, r0, c1, c0
963 #ifdef CONFIG_MMU
964 bic r0, r0, #0x000d
965 #else
966 bic r0, r0, #0x000c
967 #endif
968 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
969 mov r12, lr
970 bl __armv7_mmu_cache_flush
971 mov r0, #0
972 #ifdef CONFIG_MMU
973 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
974 #endif
975 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
976 mcr p15, 0, r0, c7, c10, 4 @ DSB
977 mcr p15, 0, r0, c7, c5, 4 @ ISB
978 mov pc, r12
980 __arm6_mmu_cache_off:
981 mov r0, #0x00000030 @ ARM6 control reg.
982 b __armv3_mmu_cache_off
984 __arm7_mmu_cache_off:
985 mov r0, #0x00000070 @ ARM7 control reg.
986 b __armv3_mmu_cache_off
988 __armv3_mmu_cache_off:
989 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
990 mov r0, #0
991 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
992 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
993 mov pc, lr
995 /*
996 * Clean and flush the cache to maintain consistency.
997 *
998 * On exit,
999 * r1, r2, r3, r9, r10, r11, r12 corrupted
1000 * This routine must preserve:
1001 * r4, r6, r7, r8
1002 */
1003 .align 5
1004 cache_clean_flush:
1005 mov r3, #16
1006 b call_cache_fn
1008 __armv4_mpu_cache_flush:
1009 mov r2, #1
1010 mov r3, #0
1011 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
1012 mov r1, #7 << 5 @ 8 segments
1013 1: orr r3, r1, #63 << 26 @ 64 entries
1014 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
1015 subs r3, r3, #1 << 26
1016 bcs 2b @ entries 63 to 0
1017 subs r1, r1, #1 << 5
1018 bcs 1b @ segments 7 to 0
1020 teq r2, #0
1021 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
1022 mcr p15, 0, ip, c7, c10, 4 @ drain WB
1023 mov pc, lr
1025 __fa526_cache_flush:
1026 mov r1, #0
1027 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
1028 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1029 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1030 mov pc, lr
1032 __armv6_mmu_cache_flush:
1033 mov r1, #0
1034 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
1035 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
1036 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
1037 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1038 mov pc, lr
1040 __armv7_mmu_cache_flush:
1041 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
1042 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
1043 mov r10, #0
1044 beq hierarchical
1045 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
1046 b iflush
1047 hierarchical:
1048 mcr p15, 0, r10, c7, c10, 5 @ DMB
1049 stmfd sp!, {r0-r7, r9-r11}
1050 mrc p15, 1, r0, c0, c0, 1 @ read clidr
1051 ands r3, r0, #0x7000000 @ extract loc from clidr
1052 mov r3, r3, lsr #23 @ left align loc bit field
1053 beq finished @ if loc is 0, then no need to clean
1054 mov r10, #0 @ start clean at cache level 0
1055 loop1:
1056 add r2, r10, r10, lsr #1 @ work out 3x current cache level
1057 mov r1, r0, lsr r2 @ extract cache type bits from clidr
1058 and r1, r1, #7 @ mask of the bits for current cache only
1059 cmp r1, #2 @ see what cache we have at this level
1060 blt skip @ skip if no cache, or just i-cache
1061 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1062 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
1063 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
1064 and r2, r1, #7 @ extract the length of the cache lines
1065 add r2, r2, #4 @ add 4 (line length offset)
1066 ldr r4, =0x3ff
1067 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
1068 clz r5, r4 @ find bit position of way size increment
1069 ldr r7, =0x7fff
1070 ands r7, r7, r1, lsr #13 @ extract max number of the index size
1071 loop2:
1072 mov r9, r4 @ create working copy of max way size
1073 loop3:
1074 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
1075 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
1076 THUMB( lsl r6, r9, r5 )
1077 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
1078 THUMB( lsl r6, r7, r2 )
1079 THUMB( orr r11, r11, r6 ) @ factor index number into r11
1080 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
1081 subs r9, r9, #1 @ decrement the way
1082 bge loop3
1083 subs r7, r7, #1 @ decrement the index
1084 bge loop2
1085 skip:
1086 add r10, r10, #2 @ increment cache number
1087 cmp r3, r10
1088 bgt loop1
1089 finished:
1090 ldmfd sp!, {r0-r7, r9-r11}
1091 mov r10, #0 @ swith back to cache level 0
1092 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1093 iflush:
1094 mcr p15, 0, r10, c7, c10, 4 @ DSB
1095 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
1096 mcr p15, 0, r10, c7, c10, 4 @ DSB
1097 mcr p15, 0, r10, c7, c5, 4 @ ISB
1098 mov pc, lr
1100 __armv5tej_mmu_cache_flush:
1101 1: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
1102 bne 1b
1103 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
1104 mcr p15, 0, r0, c7, c10, 4 @ drain WB
1105 mov pc, lr
1107 __armv4_mmu_cache_flush:
1108 mov r2, #64*1024 @ default: 32K dcache size (*2)
1109 mov r11, #32 @ default: 32 byte line size
1110 mrc p15, 0, r3, c0, c0, 1 @ read cache type
1111 teq r3, r9 @ cache ID register present?
1112 beq no_cache_id
1113 mov r1, r3, lsr #18
1114 and r1, r1, #7
1115 mov r2, #1024
1116 mov r2, r2, lsl r1 @ base dcache size *2
1117 tst r3, #1 << 14 @ test M bit
1118 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
1119 mov r3, r3, lsr #12
1120 and r3, r3, #3
1121 mov r11, #8
1122 mov r11, r11, lsl r3 @ cache line size in bytes
1123 no_cache_id:
1124 mov r1, pc
1125 bic r1, r1, #63 @ align to longest cache line
1126 add r2, r1, r2
1127 1:
1128 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
1129 THUMB( ldr r3, [r1] ) @ s/w flush D cache
1130 THUMB( add r1, r1, r11 )
1131 teq r1, r2
1132 bne 1b
1134 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1135 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
1136 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1137 mov pc, lr
1139 __armv3_mmu_cache_flush:
1140 __armv3_mpu_cache_flush:
1141 mov r1, #0
1142 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
1143 mov pc, lr
1145 /*
1146 * Various debugging routines for printing hex characters and
1147 * memory, which again must be relocatable.
1148 */
1149 #ifdef DEBUG
1150 .align 2
1151 .type phexbuf,#object
1152 phexbuf: .space 12
1153 .size phexbuf, . - phexbuf
1155 @ phex corrupts {r0, r1, r2, r3}
1156 phex: adr r3, phexbuf
1157 mov r2, #0
1158 strb r2, [r3, r1]
1159 1: subs r1, r1, #1
1160 movmi r0, r3
1161 bmi puts
1162 and r2, r0, #15
1163 mov r0, r0, lsr #4
1164 cmp r2, #10
1165 addge r2, r2, #7
1166 add r2, r2, #'0'
1167 strb r2, [r3, r1]
1168 b 1b
1170 @ puts corrupts {r0, r1, r2, r3}
1171 puts: loadsp r3, r1
1172 1: ldrb r2, [r0], #1
1173 teq r2, #0
1174 moveq pc, lr
1175 2: writeb r2, r3
1176 mov r1, #0x00020000
1177 3: subs r1, r1, #1
1178 bne 3b
1179 teq r2, #'\n'
1180 moveq r2, #'\r'
1181 beq 2b
1182 teq r0, #0
1183 bne 1b
1184 mov pc, lr
1185 @ putc corrupts {r0, r1, r2, r3}
1186 putc:
1187 mov r2, r0
1188 mov r0, #0
1189 loadsp r3, r1
1190 b 2b
1192 @ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1193 memdump: mov r12, r0
1194 mov r10, lr
1195 mov r11, #0
1196 2: mov r0, r11, lsl #2
1197 add r0, r0, r12
1198 mov r1, #8
1199 bl phex
1200 mov r0, #':'
1201 bl putc
1202 1: mov r0, #' '
1203 bl putc
1204 ldr r0, [r12, r11, lsl #2]
1205 mov r1, #8
1206 bl phex
1207 and r0, r11, #7
1208 teq r0, #3
1209 moveq r0, #' '
1210 bleq putc
1211 and r0, r11, #7
1212 add r11, r11, #1
1213 teq r0, #7
1214 bne 1b
1215 mov r0, #'\n'
1216 bl putc
1217 cmp r11, #64
1218 blt 2b
1219 mov pc, r10
1220 #endif
1222 .ltorg
1223 reloc_code_end:
1225 .align
1226 .section ".stack", "aw", %nobits
1227 .L_user_stack: .space 4096
1228 .L_user_stack_end: