2 * linux/arch/arm/boot/compressed/head.S
4 * Copyright (C) 1996-2002 Russell King
5 * Copyright (C) 2004 Hyok S. Choi (MPU support)
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.
11 #include <linux/linkage.h>
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.
22 #if defined(CONFIG_DEBUG_ICEDCC)
24 #if defined(CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K)
25 .macro loadsp, rb, tmp
28 mcr p14, 0, \ch, c0, c5, 0
30 #elif defined(CONFIG_CPU_V7)
31 .macro loadsp, rb, tmp
34 wait: mrc p14, 0, pc, c0, c1, 0
36 mcr p14, 0, \ch, c0, c5, 0
38 #elif defined(CONFIG_CPU_XSCALE)
39 .macro loadsp, rb, tmp
42 mcr p14, 0, \ch, c8, c0, 0
45 .macro loadsp, rb, tmp
48 mcr p14, 0, \ch, c1, c0, 0
54 #include <mach/debug-macro.S>
60 #if defined(CONFIG_ARCH_SA1100)
61 .macro loadsp, rb, tmp
62 mov \rb, #0x80000000 @ physical base address
63 #ifdef CONFIG_DEBUG_LL_SER3
64 add \rb, \rb, #0x00050000 @ Ser3
66 add \rb, \rb, #0x00010000 @ Ser1
69 #elif defined(CONFIG_ARCH_S3C2410)
70 .macro loadsp, rb, tmp
72 add \rb, \rb, #0x4000 * CONFIG_S3C_LOWLEVEL_UART_PORT
75 .macro loadsp, rb, tmp
93 .macro debug_reloc_start
96 kphex r6, 8 /* processor id */
98 kphex r7, 8 /* architecture id */
99 #ifdef CONFIG_CPU_CP15
101 mrc p15, 0, r0, c1, c0
102 kphex r0, 8 /* control reg */
105 kphex r5, 8 /* decompressed kernel start */
107 kphex r9, 8 /* decompressed kernel end */
109 kphex r4, 8 /* kernel execution address */
114 .macro debug_reloc_end
116 kphex r5, 8 /* end of kernel */
119 bl memdump /* dump 256 bytes at start of kernel */
123 .section ".start", #alloc, #execinstr
125 * sort out different calling conventions
128 .arm @ Always enter in ARM state
130 .type start,#function
136 THUMB( adr r12, BSYM(1f) )
139 .word 0x016f2818 @ Magic numbers to help the loader
140 .word start @ absolute load/run zImage address
141 .word _edata @ zImage end address
143 1: mov r7, r1 @ save architecture ID
144 mov r8, r2 @ save atags pointer
146 #ifndef __ARM_ARCH_2__
148 * Booting from Angel - need to enter SVC mode and disable
149 * FIQs/IRQs (numeric definitions from angel arm.h source).
150 * We only do this if we were in user mode on entry.
152 mrs r2, cpsr @ get current mode
153 tst r2, #3 @ not user?
155 mov r0, #0x17 @ angel_SWIreason_EnterSVC
156 ARM( swi 0x123456 ) @ angel_SWI_ARM
157 THUMB( svc 0xab ) @ angel_SWI_THUMB
159 mrs r2, cpsr @ turn off interrupts to
160 orr r2, r2, #0xc0 @ prevent angel from running
163 teqp pc, #0x0c000003 @ turn off interrupts
167 * Note that some cache flushing and other stuff may
168 * be needed here - is there an Angel SWI call for this?
172 * some architecture specific code can be inserted
173 * by the linker here, but it should preserve r7, r8, and r9.
178 #ifdef CONFIG_AUTO_ZRELADDR
179 @ determine final kernel image address
181 and r4, r4, #0xf8000000
182 add r4, r4, #TEXT_OFFSET
190 ldmia r0, {r1, r2, r3, r6, r10, r11, r12}
194 * We might be running at a different address. We need
195 * to fix up various pointers.
197 sub r0, r0, r1 @ calculate the delta offset
198 add r6, r6, r0 @ _edata
199 add r10, r10, r0 @ inflated kernel size location
202 * The kernel build system appends the size of the
203 * decompressed kernel at the end of the compressed data
204 * in little-endian form.
208 orr r9, r9, lr, lsl #8
211 orr r9, r9, lr, lsl #16
212 orr r9, r9, r10, lsl #24
214 #ifndef CONFIG_ZBOOT_ROM
215 /* malloc space is above the relocated stack (64k max) */
217 add r10, sp, #0x10000
220 * With ZBOOT_ROM the bss/stack is non relocatable,
221 * but someone could still run this code from RAM,
222 * in which case our reference is _edata.
228 * Check to see if we will overwrite ourselves.
229 * r4 = final kernel address
230 * r9 = size of decompressed image
231 * r10 = end of this image, including bss/stack/malloc space if non XIP
233 * r4 - 16k page directory >= r10 -> OK
234 * r4 + image length <= current position (pc) -> OK
246 * Relocate ourselves past the end of the decompressed kernel.
248 * r10 = end of the decompressed kernel
249 * Because we always copy ahead, we need to do it from the end and go
250 * backward in case the source and destination overlap.
253 * Bump to the next 256-byte boundary with the size of
254 * the relocation code added. This avoids overwriting
255 * ourself when the offset is small.
257 add r10, r10, #((reloc_code_end - restart + 256) & ~255)
260 /* Get start of code we want to copy and align it down. */
264 sub r9, r6, r5 @ size to copy
265 add r9, r9, #31 @ rounded up to a multiple
266 bic r9, r9, #31 @ ... of 32 bytes
270 1: ldmdb r6!, {r0 - r3, r10 - r12, lr}
272 stmdb r9!, {r0 - r3, r10 - r12, lr}
275 /* Preserve offset to relocated code. */
278 #ifndef CONFIG_ZBOOT_ROM
279 /* cache_clean_flush may use the stack, so relocate it */
285 adr r0, BSYM(restart)
291 * If delta is zero, we are running at the address we were linked at.
295 * r4 = kernel execution address
296 * r7 = architecture ID
307 #ifndef CONFIG_ZBOOT_ROM
309 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
310 * we need to fix up pointers into the BSS region.
311 * Note that the stack pointer has already been fixed up.
317 * Relocate all entries in the GOT table.
319 1: ldr r1, [r11, #0] @ relocate entries in the GOT
320 add r1, r1, r0 @ table. This fixes up the
321 str r1, [r11], #4 @ C references.
327 * Relocate entries in the GOT table. We only relocate
328 * the entries that are outside the (relocated) BSS region.
330 1: ldr r1, [r11, #0] @ relocate entries in the GOT
331 cmp r1, r2 @ entry < bss_start ||
332 cmphs r3, r1 @ _end < entry
333 addlo r1, r1, r0 @ table. This fixes up the
334 str r1, [r11], #4 @ C references.
339 not_relocated: mov r0, #0
340 1: str r0, [r2], #4 @ clear bss
348 * The C runtime environment should now be setup sufficiently.
349 * Set up some pointers, and start decompressing.
350 * r4 = kernel execution address
351 * r7 = architecture ID
355 mov r1, sp @ malloc space above stack
356 add r2, sp, #0x10000 @ 64k max
361 mov r0, #0 @ must be zero
362 mov r1, r7 @ restore architecture number
363 mov r2, r8 @ restore atags pointer
364 mov pc, r4 @ call kernel
369 .word __bss_start @ r2
372 .word input_data_end - 4 @ r10 (inflated size location)
373 .word _got_start @ r11
375 .word .L_user_stack_end @ sp
378 #ifdef CONFIG_ARCH_RPC
380 params: ldr r0, =0x10000100 @ params_phys for RPC
387 * Turn on the cache. We need to setup some page tables so that we
388 * can have both the I and D caches on.
390 * We place the page tables 16k down from the kernel execution address,
391 * and we hope that nothing else is using it. If we're using it, we
395 * r4 = kernel execution address
396 * r7 = architecture number
399 * r0, r1, r2, r3, r9, r10, r12 corrupted
400 * This routine must preserve:
404 cache_on: mov r3, #8 @ cache_on function
408 * Initialize the highest priority protection region, PR7
409 * to cover all 32bit address and cacheable and bufferable.
411 __armv4_mpu_cache_on:
412 mov r0, #0x3f @ 4G, the whole
413 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
414 mcr p15, 0, r0, c6, c7, 1
417 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
418 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
419 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
422 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
423 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
426 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
427 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
428 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
429 mrc p15, 0, r0, c1, c0, 0 @ read control reg
430 @ ...I .... ..D. WC.M
431 orr r0, r0, #0x002d @ .... .... ..1. 11.1
432 orr r0, r0, #0x1000 @ ...1 .... .... ....
434 mcr p15, 0, r0, c1, c0, 0 @ write control reg
437 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
438 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
441 __armv3_mpu_cache_on:
442 mov r0, #0x3f @ 4G, the whole
443 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
446 mcr p15, 0, r0, c2, c0, 0 @ cache on
447 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
450 mcr p15, 0, r0, c5, c0, 0 @ access permission
453 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
455 * ?? ARMv3 MMU does not allow reading the control register,
456 * does this really work on ARMv3 MPU?
458 mrc p15, 0, r0, c1, c0, 0 @ read control reg
459 @ .... .... .... WC.M
460 orr r0, r0, #0x000d @ .... .... .... 11.1
461 /* ?? this overwrites the value constructed above? */
463 mcr p15, 0, r0, c1, c0, 0 @ write control reg
465 /* ?? invalidate for the second time? */
466 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
469 __setup_mmu: sub r3, r4, #16384 @ Page directory size
470 bic r3, r3, #0xff @ Align the pointer
473 * Initialise the page tables, turning on the cacheable and bufferable
474 * bits for the RAM area only.
478 mov r9, r9, lsl #18 @ start of RAM
479 add r10, r9, #0x10000000 @ a reasonable RAM size
483 1: cmp r1, r9 @ if virt > start of RAM
484 orrhs r1, r1, #0x0c @ set cacheable, bufferable
485 cmp r1, r10 @ if virt > end of RAM
486 bichs r1, r1, #0x0c @ clear cacheable, bufferable
487 str r1, [r0], #4 @ 1:1 mapping
492 * If ever we are running from Flash, then we surely want the cache
493 * to be enabled also for our execution instance... We map 2MB of it
494 * so there is no map overlap problem for up to 1 MB compressed kernel.
495 * If the execution is in RAM then we would only be duplicating the above.
501 orr r1, r1, r2, lsl #20
502 add r0, r3, r2, lsl #2
509 __armv4_mmu_cache_on:
514 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
515 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
516 mrc p15, 0, r0, c1, c0, 0 @ read control reg
517 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
519 #ifdef CONFIG_CPU_ENDIAN_BE8
520 orr r0, r0, #1 << 25 @ big-endian page tables
522 bl __common_mmu_cache_on
524 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
528 __armv7_mmu_cache_on:
531 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
535 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
537 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
539 mrc p15, 0, r0, c1, c0, 0 @ read control reg
540 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
541 orr r0, r0, #0x003c @ write buffer
543 #ifdef CONFIG_CPU_ENDIAN_BE8
544 orr r0, r0, #1 << 25 @ big-endian page tables
546 orrne r0, r0, #1 @ MMU enabled
548 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
549 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
551 mcr p15, 0, r0, c1, c0, 0 @ load control register
552 mrc p15, 0, r0, c1, c0, 0 @ and read it back
554 mcr p15, 0, r0, c7, c5, 4 @ ISB
561 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
562 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
563 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
564 mrc p15, 0, r0, c1, c0, 0 @ read control reg
565 orr r0, r0, #0x1000 @ I-cache enable
566 bl __common_mmu_cache_on
568 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
575 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
576 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
578 bl __common_mmu_cache_on
580 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
583 __common_mmu_cache_on:
584 #ifndef CONFIG_THUMB2_KERNEL
586 orr r0, r0, #0x000d @ Write buffer, mmu
589 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
590 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
592 .align 5 @ cache line aligned
593 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
594 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
595 sub pc, lr, r0, lsr #32 @ properly flush pipeline
599 * Here follow the relocatable cache support functions for the
600 * various processors. This is a generic hook for locating an
601 * entry and jumping to an instruction at the specified offset
602 * from the start of the block. Please note this is all position
612 call_cache_fn: adr r12, proc_types
613 #ifdef CONFIG_CPU_CP15
614 mrc p15, 0, r9, c0, c0 @ get processor ID
616 ldr r9, =CONFIG_PROCESSOR_ID
618 1: ldr r1, [r12, #0] @ get value
619 ldr r2, [r12, #4] @ get mask
620 eor r1, r1, r9 @ (real ^ match)
622 ARM( addeq pc, r12, r3 ) @ call cache function
623 THUMB( addeq r12, r3 )
624 THUMB( moveq pc, r12 ) @ call cache function
629 * Table for cache operations. This is basically:
632 * - 'cache on' method instruction
633 * - 'cache off' method instruction
634 * - 'cache flush' method instruction
636 * We match an entry using: ((real_id ^ match) & mask) == 0
638 * Writethrough caches generally only need 'on' and 'off'
639 * methods. Writeback caches _must_ have the flush method
643 .type proc_types,#object
645 .word 0x41560600 @ ARM6/610
647 W(b) __arm6_mmu_cache_off @ works, but slow
648 W(b) __arm6_mmu_cache_off
651 @ b __arm6_mmu_cache_on @ untested
652 @ b __arm6_mmu_cache_off
653 @ b __armv3_mmu_cache_flush
655 .word 0x00000000 @ old ARM ID
664 .word 0x41007000 @ ARM7/710
666 W(b) __arm7_mmu_cache_off
667 W(b) __arm7_mmu_cache_off
671 .word 0x41807200 @ ARM720T (writethrough)
673 W(b) __armv4_mmu_cache_on
674 W(b) __armv4_mmu_cache_off
678 .word 0x41007400 @ ARM74x
680 W(b) __armv3_mpu_cache_on
681 W(b) __armv3_mpu_cache_off
682 W(b) __armv3_mpu_cache_flush
684 .word 0x41009400 @ ARM94x
686 W(b) __armv4_mpu_cache_on
687 W(b) __armv4_mpu_cache_off
688 W(b) __armv4_mpu_cache_flush
690 .word 0x00007000 @ ARM7 IDs
699 @ Everything from here on will be the new ID system.
701 .word 0x4401a100 @ sa110 / sa1100
703 W(b) __armv4_mmu_cache_on
704 W(b) __armv4_mmu_cache_off
705 W(b) __armv4_mmu_cache_flush
707 .word 0x6901b110 @ sa1110
709 W(b) __armv4_mmu_cache_on
710 W(b) __armv4_mmu_cache_off
711 W(b) __armv4_mmu_cache_flush
714 .word 0xffffff00 @ PXA9xx
715 W(b) __armv4_mmu_cache_on
716 W(b) __armv4_mmu_cache_off
717 W(b) __armv4_mmu_cache_flush
719 .word 0x56158000 @ PXA168
721 W(b) __armv4_mmu_cache_on
722 W(b) __armv4_mmu_cache_off
723 W(b) __armv5tej_mmu_cache_flush
725 .word 0x56050000 @ Feroceon
727 W(b) __armv4_mmu_cache_on
728 W(b) __armv4_mmu_cache_off
729 W(b) __armv5tej_mmu_cache_flush
731 #ifdef CONFIG_CPU_FEROCEON_OLD_ID
732 /* this conflicts with the standard ARMv5TE entry */
733 .long 0x41009260 @ Old Feroceon
735 b __armv4_mmu_cache_on
736 b __armv4_mmu_cache_off
737 b __armv5tej_mmu_cache_flush
740 .word 0x66015261 @ FA526
742 W(b) __fa526_cache_on
743 W(b) __armv4_mmu_cache_off
744 W(b) __fa526_cache_flush
746 @ These match on the architecture ID
748 .word 0x00020000 @ ARMv4T
750 W(b) __armv4_mmu_cache_on
751 W(b) __armv4_mmu_cache_off
752 W(b) __armv4_mmu_cache_flush
754 .word 0x00050000 @ ARMv5TE
756 W(b) __armv4_mmu_cache_on
757 W(b) __armv4_mmu_cache_off
758 W(b) __armv4_mmu_cache_flush
760 .word 0x00060000 @ ARMv5TEJ
762 W(b) __armv4_mmu_cache_on
763 W(b) __armv4_mmu_cache_off
764 W(b) __armv5tej_mmu_cache_flush
766 .word 0x0007b000 @ ARMv6
768 W(b) __armv4_mmu_cache_on
769 W(b) __armv4_mmu_cache_off
770 W(b) __armv6_mmu_cache_flush
772 .word 0x560f5810 @ Marvell PJ4 ARMv6
774 W(b) __armv4_mmu_cache_on
775 W(b) __armv4_mmu_cache_off
776 W(b) __armv6_mmu_cache_flush
778 .word 0x000f0000 @ new CPU Id
780 W(b) __armv7_mmu_cache_on
781 W(b) __armv7_mmu_cache_off
782 W(b) __armv7_mmu_cache_flush
784 .word 0 @ unrecognised type
793 .size proc_types, . - proc_types
796 * Turn off the Cache and MMU. ARMv3 does not support
797 * reading the control register, but ARMv4 does.
800 * r0, r1, r2, r3, r9, r12 corrupted
801 * This routine must preserve:
805 cache_off: mov r3, #12 @ cache_off function
808 __armv4_mpu_cache_off:
809 mrc p15, 0, r0, c1, c0
811 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
813 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
814 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
815 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
818 __armv3_mpu_cache_off:
819 mrc p15, 0, r0, c1, c0
821 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
823 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
826 __armv4_mmu_cache_off:
828 mrc p15, 0, r0, c1, c0
830 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
832 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
833 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
837 __armv7_mmu_cache_off:
838 mrc p15, 0, r0, c1, c0
844 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
846 bl __armv7_mmu_cache_flush
849 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
851 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
852 mcr p15, 0, r0, c7, c10, 4 @ DSB
853 mcr p15, 0, r0, c7, c5, 4 @ ISB
856 __arm6_mmu_cache_off:
857 mov r0, #0x00000030 @ ARM6 control reg.
858 b __armv3_mmu_cache_off
860 __arm7_mmu_cache_off:
861 mov r0, #0x00000070 @ ARM7 control reg.
862 b __armv3_mmu_cache_off
864 __armv3_mmu_cache_off:
865 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
867 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
868 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
872 * Clean and flush the cache to maintain consistency.
875 * r1, r2, r3, r9, r10, r11, r12 corrupted
876 * This routine must preserve:
884 __armv4_mpu_cache_flush:
887 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
888 mov r1, #7 << 5 @ 8 segments
889 1: orr r3, r1, #63 << 26 @ 64 entries
890 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
891 subs r3, r3, #1 << 26
892 bcs 2b @ entries 63 to 0
894 bcs 1b @ segments 7 to 0
897 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
898 mcr p15, 0, ip, c7, c10, 4 @ drain WB
903 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
904 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
905 mcr p15, 0, r1, c7, c10, 4 @ drain WB
908 __armv6_mmu_cache_flush:
910 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
911 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
912 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
913 mcr p15, 0, r1, c7, c10, 4 @ drain WB
916 __armv7_mmu_cache_flush:
917 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
918 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
921 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
924 mcr p15, 0, r10, c7, c10, 5 @ DMB
925 stmfd sp!, {r0-r7, r9-r11}
926 mrc p15, 1, r0, c0, c0, 1 @ read clidr
927 ands r3, r0, #0x7000000 @ extract loc from clidr
928 mov r3, r3, lsr #23 @ left align loc bit field
929 beq finished @ if loc is 0, then no need to clean
930 mov r10, #0 @ start clean at cache level 0
932 add r2, r10, r10, lsr #1 @ work out 3x current cache level
933 mov r1, r0, lsr r2 @ extract cache type bits from clidr
934 and r1, r1, #7 @ mask of the bits for current cache only
935 cmp r1, #2 @ see what cache we have at this level
936 blt skip @ skip if no cache, or just i-cache
937 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
938 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
939 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
940 and r2, r1, #7 @ extract the length of the cache lines
941 add r2, r2, #4 @ add 4 (line length offset)
943 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
944 clz r5, r4 @ find bit position of way size increment
946 ands r7, r7, r1, lsr #13 @ extract max number of the index size
948 mov r9, r4 @ create working copy of max way size
950 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
951 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
952 THUMB( lsl r6, r9, r5 )
953 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
954 THUMB( lsl r6, r7, r2 )
955 THUMB( orr r11, r11, r6 ) @ factor index number into r11
956 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
957 subs r9, r9, #1 @ decrement the way
959 subs r7, r7, #1 @ decrement the index
962 add r10, r10, #2 @ increment cache number
966 ldmfd sp!, {r0-r7, r9-r11}
967 mov r10, #0 @ swith back to cache level 0
968 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
970 mcr p15, 0, r10, c7, c10, 4 @ DSB
971 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
972 mcr p15, 0, r10, c7, c10, 4 @ DSB
973 mcr p15, 0, r10, c7, c5, 4 @ ISB
976 __armv5tej_mmu_cache_flush:
977 1: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
979 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
980 mcr p15, 0, r0, c7, c10, 4 @ drain WB
983 __armv4_mmu_cache_flush:
984 mov r2, #64*1024 @ default: 32K dcache size (*2)
985 mov r11, #32 @ default: 32 byte line size
986 mrc p15, 0, r3, c0, c0, 1 @ read cache type
987 teq r3, r9 @ cache ID register present?
992 mov r2, r2, lsl r1 @ base dcache size *2
993 tst r3, #1 << 14 @ test M bit
994 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
998 mov r11, r11, lsl r3 @ cache line size in bytes
1001 bic r1, r1, #63 @ align to longest cache line
1004 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
1005 THUMB( ldr r3, [r1] ) @ s/w flush D cache
1006 THUMB( add r1, r1, r11 )
1010 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1011 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
1012 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1015 __armv3_mmu_cache_flush:
1016 __armv3_mpu_cache_flush:
1018 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
1022 * Various debugging routines for printing hex characters and
1023 * memory, which again must be relocatable.
1027 .type phexbuf,#object
1029 .size phexbuf, . - phexbuf
1031 @ phex corrupts {r0, r1, r2, r3}
1032 phex: adr r3, phexbuf
1046 @ puts corrupts {r0, r1, r2, r3}
1048 1: ldrb r2, [r0], #1
1061 @ putc corrupts {r0, r1, r2, r3}
1068 @ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1069 memdump: mov r12, r0
1072 2: mov r0, r11, lsl #2
1080 ldr r0, [r12, r11, lsl #2]
1102 .section ".stack", "aw", %nobits
1103 .L_user_stack: .space 4096