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)
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, r5, r6, r9, 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 r5, r5, r0 @ _start
199 add r6, r6, r0 @ _edata
201 #ifndef CONFIG_ZBOOT_ROM
202 /* malloc space is above the relocated stack (64k max) */
204 add r10, sp, #0x10000
207 * With ZBOOT_ROM the bss/stack is non relocatable,
208 * but someone could still run this code from RAM,
209 * in which case our reference is _edata.
215 * Check to see if we will overwrite ourselves.
216 * r4 = final kernel address
217 * r5 = start of this image
218 * r9 = size of decompressed image
219 * r10 = end of this image, including bss/stack/malloc space if non XIP
222 * r4 + image length <= r5 -> OK
231 * Relocate ourselves past the end of the decompressed kernel.
232 * r5 = start of this image
234 * r10 = end of the decompressed kernel
235 * Because we always copy ahead, we need to do it from the end and go
236 * backward in case the source and destination overlap.
238 /* Round up to next 256-byte boundary. */
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
248 1: ldmdb r6!, {r0 - r3, r10 - r12, lr}
250 stmdb r9!, {r0 - r3, r10 - r12, lr}
253 /* Preserve offset to relocated code. */
258 adr r0, BSYM(restart)
264 * If delta is zero, we are running at the address we were linked at.
268 * r4 = kernel execution address
269 * r7 = architecture ID
280 #ifndef CONFIG_ZBOOT_ROM
282 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
283 * we need to fix up pointers into the BSS region.
284 * Note that the stack pointer has already been fixed up.
290 * Relocate all entries in the GOT table.
292 1: ldr r1, [r11, #0] @ relocate entries in the GOT
293 add r1, r1, r0 @ table. This fixes up the
294 str r1, [r11], #4 @ C references.
300 * Relocate entries in the GOT table. We only relocate
301 * the entries that are outside the (relocated) BSS region.
303 1: ldr r1, [r11, #0] @ relocate entries in the GOT
304 cmp r1, r2 @ entry < bss_start ||
305 cmphs r3, r1 @ _end < entry
306 addlo r1, r1, r0 @ table. This fixes up the
307 str r1, [r11], #4 @ C references.
312 not_relocated: mov r0, #0
313 1: str r0, [r2], #4 @ clear bss
321 * The C runtime environment should now be setup sufficiently.
322 * Set up some pointers, and start decompressing.
323 * r4 = kernel execution address
324 * r7 = architecture ID
328 mov r1, sp @ malloc space above stack
329 add r2, sp, #0x10000 @ 64k max
334 mov r0, #0 @ must be zero
335 mov r1, r7 @ restore architecture number
336 mov r2, r8 @ restore atags pointer
337 mov pc, r4 @ call kernel
342 .word __bss_start @ r2
346 .word _image_size @ r9
347 .word _got_start @ r11
349 .word user_stack_end @ sp
352 #ifdef CONFIG_ARCH_RPC
354 params: ldr r0, =0x10000100 @ params_phys for RPC
361 * Turn on the cache. We need to setup some page tables so that we
362 * can have both the I and D caches on.
364 * We place the page tables 16k down from the kernel execution address,
365 * and we hope that nothing else is using it. If we're using it, we
369 * r4 = kernel execution address
370 * r7 = architecture number
373 * r0, r1, r2, r3, r9, r10, r12 corrupted
374 * This routine must preserve:
378 cache_on: mov r3, #8 @ cache_on function
382 * Initialize the highest priority protection region, PR7
383 * to cover all 32bit address and cacheable and bufferable.
385 __armv4_mpu_cache_on:
386 mov r0, #0x3f @ 4G, the whole
387 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
388 mcr p15, 0, r0, c6, c7, 1
391 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
392 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
393 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
396 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
397 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
400 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
401 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
402 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
403 mrc p15, 0, r0, c1, c0, 0 @ read control reg
404 @ ...I .... ..D. WC.M
405 orr r0, r0, #0x002d @ .... .... ..1. 11.1
406 orr r0, r0, #0x1000 @ ...1 .... .... ....
408 mcr p15, 0, r0, c1, c0, 0 @ write control reg
411 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
412 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
415 __armv3_mpu_cache_on:
416 mov r0, #0x3f @ 4G, the whole
417 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
420 mcr p15, 0, r0, c2, c0, 0 @ cache on
421 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
424 mcr p15, 0, r0, c5, c0, 0 @ access permission
427 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
429 * ?? ARMv3 MMU does not allow reading the control register,
430 * does this really work on ARMv3 MPU?
432 mrc p15, 0, r0, c1, c0, 0 @ read control reg
433 @ .... .... .... WC.M
434 orr r0, r0, #0x000d @ .... .... .... 11.1
435 /* ?? this overwrites the value constructed above? */
437 mcr p15, 0, r0, c1, c0, 0 @ write control reg
439 /* ?? invalidate for the second time? */
440 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
443 __setup_mmu: sub r3, r4, #16384 @ Page directory size
444 bic r3, r3, #0xff @ Align the pointer
447 * Initialise the page tables, turning on the cacheable and bufferable
448 * bits for the RAM area only.
452 mov r9, r9, lsl #18 @ start of RAM
453 add r10, r9, #0x10000000 @ a reasonable RAM size
457 1: cmp r1, r9 @ if virt > start of RAM
458 orrhs r1, r1, #0x0c @ set cacheable, bufferable
459 cmp r1, r10 @ if virt > end of RAM
460 bichs r1, r1, #0x0c @ clear cacheable, bufferable
461 str r1, [r0], #4 @ 1:1 mapping
466 * If ever we are running from Flash, then we surely want the cache
467 * to be enabled also for our execution instance... We map 2MB of it
468 * so there is no map overlap problem for up to 1 MB compressed kernel.
469 * If the execution is in RAM then we would only be duplicating the above.
475 orr r1, r1, r2, lsl #20
476 add r0, r3, r2, lsl #2
483 __armv4_mmu_cache_on:
488 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
489 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
490 mrc p15, 0, r0, c1, c0, 0 @ read control reg
491 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
493 #ifdef CONFIG_CPU_ENDIAN_BE8
494 orr r0, r0, #1 << 25 @ big-endian page tables
496 bl __common_mmu_cache_on
498 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
502 __armv7_mmu_cache_on:
505 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
509 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
511 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
513 mrc p15, 0, r0, c1, c0, 0 @ read control reg
514 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
515 orr r0, r0, #0x003c @ write buffer
517 #ifdef CONFIG_CPU_ENDIAN_BE8
518 orr r0, r0, #1 << 25 @ big-endian page tables
520 orrne r0, r0, #1 @ MMU enabled
522 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
523 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
525 mcr p15, 0, r0, c1, c0, 0 @ load control register
526 mrc p15, 0, r0, c1, c0, 0 @ and read it back
528 mcr p15, 0, r0, c7, c5, 4 @ ISB
535 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
536 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
537 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
538 mrc p15, 0, r0, c1, c0, 0 @ read control reg
539 orr r0, r0, #0x1000 @ I-cache enable
540 bl __common_mmu_cache_on
542 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
549 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
550 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
552 bl __common_mmu_cache_on
554 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
557 __common_mmu_cache_on:
558 #ifndef CONFIG_THUMB2_KERNEL
560 orr r0, r0, #0x000d @ Write buffer, mmu
563 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
564 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
566 .align 5 @ cache line aligned
567 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
568 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
569 sub pc, lr, r0, lsr #32 @ properly flush pipeline
573 * Here follow the relocatable cache support functions for the
574 * various processors. This is a generic hook for locating an
575 * entry and jumping to an instruction at the specified offset
576 * from the start of the block. Please note this is all position
586 call_cache_fn: adr r12, proc_types
587 #ifdef CONFIG_CPU_CP15
588 mrc p15, 0, r9, c0, c0 @ get processor ID
590 ldr r9, =CONFIG_PROCESSOR_ID
592 1: ldr r1, [r12, #0] @ get value
593 ldr r2, [r12, #4] @ get mask
594 eor r1, r1, r9 @ (real ^ match)
596 ARM( addeq pc, r12, r3 ) @ call cache function
597 THUMB( addeq r12, r3 )
598 THUMB( moveq pc, r12 ) @ call cache function
603 * Table for cache operations. This is basically:
606 * - 'cache on' method instruction
607 * - 'cache off' method instruction
608 * - 'cache flush' method instruction
610 * We match an entry using: ((real_id ^ match) & mask) == 0
612 * Writethrough caches generally only need 'on' and 'off'
613 * methods. Writeback caches _must_ have the flush method
617 .type proc_types,#object
619 .word 0x41560600 @ ARM6/610
621 W(b) __arm6_mmu_cache_off @ works, but slow
622 W(b) __arm6_mmu_cache_off
625 @ b __arm6_mmu_cache_on @ untested
626 @ b __arm6_mmu_cache_off
627 @ b __armv3_mmu_cache_flush
629 .word 0x00000000 @ old ARM ID
638 .word 0x41007000 @ ARM7/710
640 W(b) __arm7_mmu_cache_off
641 W(b) __arm7_mmu_cache_off
645 .word 0x41807200 @ ARM720T (writethrough)
647 W(b) __armv4_mmu_cache_on
648 W(b) __armv4_mmu_cache_off
652 .word 0x41007400 @ ARM74x
654 W(b) __armv3_mpu_cache_on
655 W(b) __armv3_mpu_cache_off
656 W(b) __armv3_mpu_cache_flush
658 .word 0x41009400 @ ARM94x
660 W(b) __armv4_mpu_cache_on
661 W(b) __armv4_mpu_cache_off
662 W(b) __armv4_mpu_cache_flush
664 .word 0x00007000 @ ARM7 IDs
673 @ Everything from here on will be the new ID system.
675 .word 0x4401a100 @ sa110 / sa1100
677 W(b) __armv4_mmu_cache_on
678 W(b) __armv4_mmu_cache_off
679 W(b) __armv4_mmu_cache_flush
681 .word 0x6901b110 @ sa1110
683 W(b) __armv4_mmu_cache_on
684 W(b) __armv4_mmu_cache_off
685 W(b) __armv4_mmu_cache_flush
688 .word 0xffffff00 @ PXA9xx
689 W(b) __armv4_mmu_cache_on
690 W(b) __armv4_mmu_cache_off
691 W(b) __armv4_mmu_cache_flush
693 .word 0x56158000 @ PXA168
695 W(b) __armv4_mmu_cache_on
696 W(b) __armv4_mmu_cache_off
697 W(b) __armv5tej_mmu_cache_flush
699 .word 0x56050000 @ Feroceon
701 W(b) __armv4_mmu_cache_on
702 W(b) __armv4_mmu_cache_off
703 W(b) __armv5tej_mmu_cache_flush
705 #ifdef CONFIG_CPU_FEROCEON_OLD_ID
706 /* this conflicts with the standard ARMv5TE entry */
707 .long 0x41009260 @ Old Feroceon
709 b __armv4_mmu_cache_on
710 b __armv4_mmu_cache_off
711 b __armv5tej_mmu_cache_flush
714 .word 0x66015261 @ FA526
716 W(b) __fa526_cache_on
717 W(b) __armv4_mmu_cache_off
718 W(b) __fa526_cache_flush
720 @ These match on the architecture ID
722 .word 0x00020000 @ ARMv4T
724 W(b) __armv4_mmu_cache_on
725 W(b) __armv4_mmu_cache_off
726 W(b) __armv4_mmu_cache_flush
728 .word 0x00050000 @ ARMv5TE
730 W(b) __armv4_mmu_cache_on
731 W(b) __armv4_mmu_cache_off
732 W(b) __armv4_mmu_cache_flush
734 .word 0x00060000 @ ARMv5TEJ
736 W(b) __armv4_mmu_cache_on
737 W(b) __armv4_mmu_cache_off
738 W(b) __armv5tej_mmu_cache_flush
740 .word 0x0007b000 @ ARMv6
742 W(b) __armv4_mmu_cache_on
743 W(b) __armv4_mmu_cache_off
744 W(b) __armv6_mmu_cache_flush
746 .word 0x560f5810 @ Marvell PJ4 ARMv6
748 W(b) __armv4_mmu_cache_on
749 W(b) __armv4_mmu_cache_off
750 W(b) __armv6_mmu_cache_flush
752 .word 0x000f0000 @ new CPU Id
754 W(b) __armv7_mmu_cache_on
755 W(b) __armv7_mmu_cache_off
756 W(b) __armv7_mmu_cache_flush
758 .word 0 @ unrecognised type
767 .size proc_types, . - proc_types
770 * Turn off the Cache and MMU. ARMv3 does not support
771 * reading the control register, but ARMv4 does.
774 * r0, r1, r2, r3, r9, r12 corrupted
775 * This routine must preserve:
779 cache_off: mov r3, #12 @ cache_off function
782 __armv4_mpu_cache_off:
783 mrc p15, 0, r0, c1, c0
785 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
787 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
788 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
789 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
792 __armv3_mpu_cache_off:
793 mrc p15, 0, r0, c1, c0
795 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
797 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
800 __armv4_mmu_cache_off:
802 mrc p15, 0, r0, c1, c0
804 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
806 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
807 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
811 __armv7_mmu_cache_off:
812 mrc p15, 0, r0, c1, c0
818 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
820 bl __armv7_mmu_cache_flush
823 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
825 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
826 mcr p15, 0, r0, c7, c10, 4 @ DSB
827 mcr p15, 0, r0, c7, c5, 4 @ ISB
830 __arm6_mmu_cache_off:
831 mov r0, #0x00000030 @ ARM6 control reg.
832 b __armv3_mmu_cache_off
834 __arm7_mmu_cache_off:
835 mov r0, #0x00000070 @ ARM7 control reg.
836 b __armv3_mmu_cache_off
838 __armv3_mmu_cache_off:
839 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
841 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
842 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
846 * Clean and flush the cache to maintain consistency.
849 * r1, r2, r3, r9, r10, r11, r12 corrupted
850 * This routine must preserve:
858 __armv4_mpu_cache_flush:
861 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
862 mov r1, #7 << 5 @ 8 segments
863 1: orr r3, r1, #63 << 26 @ 64 entries
864 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
865 subs r3, r3, #1 << 26
866 bcs 2b @ entries 63 to 0
868 bcs 1b @ segments 7 to 0
871 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
872 mcr p15, 0, ip, c7, c10, 4 @ drain WB
877 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
878 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
879 mcr p15, 0, r1, c7, c10, 4 @ drain WB
882 __armv6_mmu_cache_flush:
884 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
885 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
886 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
887 mcr p15, 0, r1, c7, c10, 4 @ drain WB
890 __armv7_mmu_cache_flush:
891 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
892 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
895 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
898 mcr p15, 0, r10, c7, c10, 5 @ DMB
899 stmfd sp!, {r0-r7, r9-r11}
900 mrc p15, 1, r0, c0, c0, 1 @ read clidr
901 ands r3, r0, #0x7000000 @ extract loc from clidr
902 mov r3, r3, lsr #23 @ left align loc bit field
903 beq finished @ if loc is 0, then no need to clean
904 mov r10, #0 @ start clean at cache level 0
906 add r2, r10, r10, lsr #1 @ work out 3x current cache level
907 mov r1, r0, lsr r2 @ extract cache type bits from clidr
908 and r1, r1, #7 @ mask of the bits for current cache only
909 cmp r1, #2 @ see what cache we have at this level
910 blt skip @ skip if no cache, or just i-cache
911 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
912 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
913 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
914 and r2, r1, #7 @ extract the length of the cache lines
915 add r2, r2, #4 @ add 4 (line length offset)
917 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
918 clz r5, r4 @ find bit position of way size increment
920 ands r7, r7, r1, lsr #13 @ extract max number of the index size
922 mov r9, r4 @ create working copy of max way size
924 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
925 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
926 THUMB( lsl r6, r9, r5 )
927 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
928 THUMB( lsl r6, r7, r2 )
929 THUMB( orr r11, r11, r6 ) @ factor index number into r11
930 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
931 subs r9, r9, #1 @ decrement the way
933 subs r7, r7, #1 @ decrement the index
936 add r10, r10, #2 @ increment cache number
940 ldmfd sp!, {r0-r7, r9-r11}
941 mov r10, #0 @ swith back to cache level 0
942 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
944 mcr p15, 0, r10, c7, c10, 4 @ DSB
945 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
946 mcr p15, 0, r10, c7, c10, 4 @ DSB
947 mcr p15, 0, r10, c7, c5, 4 @ ISB
950 __armv5tej_mmu_cache_flush:
951 1: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
953 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
954 mcr p15, 0, r0, c7, c10, 4 @ drain WB
957 __armv4_mmu_cache_flush:
958 mov r2, #64*1024 @ default: 32K dcache size (*2)
959 mov r11, #32 @ default: 32 byte line size
960 mrc p15, 0, r3, c0, c0, 1 @ read cache type
961 teq r3, r9 @ cache ID register present?
966 mov r2, r2, lsl r1 @ base dcache size *2
967 tst r3, #1 << 14 @ test M bit
968 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
972 mov r11, r11, lsl r3 @ cache line size in bytes
975 bic r1, r1, #63 @ align to longest cache line
978 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
979 THUMB( ldr r3, [r1] ) @ s/w flush D cache
980 THUMB( add r1, r1, r11 )
984 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
985 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
986 mcr p15, 0, r1, c7, c10, 4 @ drain WB
989 __armv3_mmu_cache_flush:
990 __armv3_mpu_cache_flush:
992 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
996 * Various debugging routines for printing hex characters and
997 * memory, which again must be relocatable.
1001 .type phexbuf,#object
1003 .size phexbuf, . - phexbuf
1005 @ phex corrupts {r0, r1, r2, r3}
1006 phex: adr r3, phexbuf
1020 @ puts corrupts {r0, r1, r2, r3}
1022 1: ldrb r2, [r0], #1
1035 @ putc corrupts {r0, r1, r2, r3}
1042 @ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1043 memdump: mov r12, r0
1046 2: mov r0, r11, lsl #2
1054 ldr r0, [r12, r11, lsl #2]
1075 .section ".stack", "aw", %nobits
1076 user_stack: .space 4096