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) || defined(CONFIG_CPU_V7)
25 .macro loadsp, rb, tmp
28 mcr p14, 0, \ch, c0, c5, 0
30 #elif defined(CONFIG_CPU_XSCALE)
31 .macro loadsp, rb, tmp
34 mcr p14, 0, \ch, c8, c0, 0
37 .macro loadsp, rb, tmp
40 mcr p14, 0, \ch, c1, c0, 0
46 #include <mach/debug-macro.S>
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
58 add \rb, \rb, #0x00010000 @ Ser1
61 #elif defined(CONFIG_ARCH_S3C2410)
62 .macro loadsp, rb, tmp
64 add \rb, \rb, #0x4000 * CONFIG_S3C_LOWLEVEL_UART_PORT
67 .macro loadsp, rb, tmp
85 .macro debug_reloc_start
88 kphex r6, 8 /* processor id */
90 kphex r7, 8 /* architecture id */
91 #ifdef CONFIG_CPU_CP15
93 mrc p15, 0, r0, c1, c0
94 kphex r0, 8 /* control reg */
97 kphex r5, 8 /* decompressed kernel start */
99 kphex r9, 8 /* decompressed kernel end */
101 kphex r4, 8 /* kernel execution address */
106 .macro debug_reloc_end
108 kphex r5, 8 /* end of kernel */
111 bl memdump /* dump 256 bytes at start of kernel */
115 .section ".start", #alloc, #execinstr
117 * sort out different calling conventions
120 .arm @ Always enter in ARM state
122 .type start,#function
128 THUMB( adr r12, BSYM(1f) )
131 .word 0x016f2818 @ Magic numbers to help the loader
132 .word start @ absolute load/run zImage address
133 .word _edata @ zImage end address
135 1: mov r7, r1 @ save architecture ID
136 mov r8, r2 @ save atags pointer
138 #ifndef __ARM_ARCH_2__
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.
144 mrs r2, cpsr @ get current mode
145 tst r2, #3 @ not user?
147 mov r0, #0x17 @ angel_SWIreason_EnterSVC
148 ARM( swi 0x123456 ) @ angel_SWI_ARM
149 THUMB( svc 0xab ) @ angel_SWI_THUMB
151 mrs r2, cpsr @ turn off interrupts to
152 orr r2, r2, #0xc0 @ prevent angel from running
155 teqp pc, #0x0c000003 @ turn off interrupts
159 * Note that some cache flushing and other stuff may
160 * be needed here - is there an Angel SWI call for this?
164 * some architecture specific code can be inserted
165 * by the linker here, but it should preserve r7, r8, and r9.
170 #ifdef CONFIG_AUTO_ZRELADDR
171 @ determine final kernel image address
173 and r4, r4, #0xf8000000
174 add r4, r4, #TEXT_OFFSET
182 ldmia r0, {r1, r2, r3, r6, r10, r11, r12}
186 * We might be running at a different address. We need
187 * to fix up various pointers.
189 sub r0, r0, r1 @ calculate the delta offset
190 add r6, r6, r0 @ _edata
191 add r10, r10, r0 @ inflated kernel size location
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.
200 orr r9, r9, lr, lsl #8
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) */
209 add r10, sp, #0x10000
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.
219 mov r5, #0 @ init dtb size to 0
220 #ifdef CONFIG_ARM_APPENDED_DTB
225 * r4 = final kernel address
226 * r5 = appended dtb size (still unknown)
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
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.
242 ldr r1, =0xedfe0dd0 @ sig is 0xd00dfeed big endian
247 bne dtb_check_done @ not found
249 #ifdef CONFIG_ARM_ATAG_DTB_COMPAT
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.
260 stmfd sp!, {r0-r3, ip, lr}
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.
272 sub r0, r4, #(TEXT_OFFSET - 0x100)
277 ldmfd sp!, {r0-r3, ip, lr}
281 mov r8, r6 @ use the appended device tree
284 * Make sure that the DTB doesn't end up in the final
285 * kernel's .bss area. To do so, we adjust the decompressed
286 * kernel size to compensate if that .bss size is larger
287 * than the relocated code.
289 ldr r5, =_kernel_bss_size
290 adr r1, wont_overwrite
295 /* Get the dtb's size */
298 /* convert r5 (dtb size) to little endian */
299 eor r1, r5, r5, ror #16
300 bic r1, r1, #0x00ff0000
302 eor r5, r5, r1, lsr #8
305 /* preserve 64-bit alignment */
309 /* relocate some pointers past the appended dtb */
317 * Check to see if we will overwrite ourselves.
318 * r4 = final kernel address
319 * r9 = size of decompressed image
320 * r10 = end of this image, including bss/stack/malloc space if non XIP
322 * r4 - 16k page directory >= r10 -> OK
323 * r4 + image length <= address of wont_overwrite -> OK
329 adr r9, wont_overwrite
334 * Relocate ourselves past the end of the decompressed kernel.
336 * r10 = end of the decompressed kernel
337 * Because we always copy ahead, we need to do it from the end and go
338 * backward in case the source and destination overlap.
341 * Bump to the next 256-byte boundary with the size of
342 * the relocation code added. This avoids overwriting
343 * ourself when the offset is small.
345 add r10, r10, #((reloc_code_end - restart + 256) & ~255)
348 /* Get start of code we want to copy and align it down. */
352 sub r9, r6, r5 @ size to copy
353 add r9, r9, #31 @ rounded up to a multiple
354 bic r9, r9, #31 @ ... of 32 bytes
358 1: ldmdb r6!, {r0 - r3, r10 - r12, lr}
360 stmdb r9!, {r0 - r3, r10 - r12, lr}
363 /* Preserve offset to relocated code. */
366 #ifndef CONFIG_ZBOOT_ROM
367 /* cache_clean_flush may use the stack, so relocate it */
373 adr r0, BSYM(restart)
379 * If delta is zero, we are running at the address we were linked at.
383 * r4 = kernel execution address
384 * r5 = appended dtb size (0 if not present)
385 * r7 = architecture ID
397 #ifndef CONFIG_ZBOOT_ROM
399 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
400 * we need to fix up pointers into the BSS region.
401 * Note that the stack pointer has already been fixed up.
407 * Relocate all entries in the GOT table.
408 * Bump bss entries to _edata + dtb size
410 1: ldr r1, [r11, #0] @ relocate entries in the GOT
411 add r1, r1, r0 @ This fixes up C references
412 cmp r1, r2 @ if entry >= bss_start &&
413 cmphs r3, r1 @ bss_end > entry
414 addhi r1, r1, r5 @ entry += dtb size
415 str r1, [r11], #4 @ next entry
419 /* bump our bss pointers too */
426 * Relocate entries in the GOT table. We only relocate
427 * the entries that are outside the (relocated) BSS region.
429 1: ldr r1, [r11, #0] @ relocate entries in the GOT
430 cmp r1, r2 @ entry < bss_start ||
431 cmphs r3, r1 @ _end < entry
432 addlo r1, r1, r0 @ table. This fixes up the
433 str r1, [r11], #4 @ C references.
438 not_relocated: mov r0, #0
439 1: str r0, [r2], #4 @ clear bss
447 * The C runtime environment should now be setup sufficiently.
448 * Set up some pointers, and start decompressing.
449 * r4 = kernel execution address
450 * r7 = architecture ID
454 mov r1, sp @ malloc space above stack
455 add r2, sp, #0x10000 @ 64k max
460 mov r0, #0 @ must be zero
461 mov r1, r7 @ restore architecture number
462 mov r2, r8 @ restore atags pointer
463 ARM( mov pc, r4 ) @ call kernel
464 THUMB( bx r4 ) @ entry point is always ARM
469 .word __bss_start @ r2
472 .word input_data_end - 4 @ r10 (inflated size location)
473 .word _got_start @ r11
475 .word .L_user_stack_end @ sp
478 #ifdef CONFIG_ARCH_RPC
480 params: ldr r0, =0x10000100 @ params_phys for RPC
487 * Turn on the cache. We need to setup some page tables so that we
488 * can have both the I and D caches on.
490 * We place the page tables 16k down from the kernel execution address,
491 * and we hope that nothing else is using it. If we're using it, we
495 * r4 = kernel execution address
496 * r7 = architecture number
499 * r0, r1, r2, r3, r9, r10, r12 corrupted
500 * This routine must preserve:
504 cache_on: mov r3, #8 @ cache_on function
508 * Initialize the highest priority protection region, PR7
509 * to cover all 32bit address and cacheable and bufferable.
511 __armv4_mpu_cache_on:
512 mov r0, #0x3f @ 4G, the whole
513 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
514 mcr p15, 0, r0, c6, c7, 1
517 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
518 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
519 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
522 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
523 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
526 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
527 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
528 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
529 mrc p15, 0, r0, c1, c0, 0 @ read control reg
530 @ ...I .... ..D. WC.M
531 orr r0, r0, #0x002d @ .... .... ..1. 11.1
532 orr r0, r0, #0x1000 @ ...1 .... .... ....
534 mcr p15, 0, r0, c1, c0, 0 @ write control reg
537 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
538 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
541 __armv3_mpu_cache_on:
542 mov r0, #0x3f @ 4G, the whole
543 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
546 mcr p15, 0, r0, c2, c0, 0 @ cache on
547 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
550 mcr p15, 0, r0, c5, c0, 0 @ access permission
553 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
555 * ?? ARMv3 MMU does not allow reading the control register,
556 * does this really work on ARMv3 MPU?
558 mrc p15, 0, r0, c1, c0, 0 @ read control reg
559 @ .... .... .... WC.M
560 orr r0, r0, #0x000d @ .... .... .... 11.1
561 /* ?? this overwrites the value constructed above? */
563 mcr p15, 0, r0, c1, c0, 0 @ write control reg
565 /* ?? invalidate for the second time? */
566 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
569 __setup_mmu: sub r3, r4, #16384 @ Page directory size
570 bic r3, r3, #0xff @ Align the pointer
573 * Initialise the page tables, turning on the cacheable and bufferable
574 * bits for the RAM area only.
578 mov r9, r9, lsl #18 @ start of RAM
579 add r10, r9, #0x10000000 @ a reasonable RAM size
583 1: cmp r1, r9 @ if virt > start of RAM
584 #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
585 orrhs r1, r1, #0x08 @ set cacheable
587 orrhs r1, r1, #0x0c @ set cacheable, bufferable
589 cmp r1, r10 @ if virt > end of RAM
590 bichs r1, r1, #0x0c @ clear cacheable, bufferable
591 str r1, [r0], #4 @ 1:1 mapping
596 * If ever we are running from Flash, then we surely want the cache
597 * to be enabled also for our execution instance... We map 2MB of it
598 * so there is no map overlap problem for up to 1 MB compressed kernel.
599 * If the execution is in RAM then we would only be duplicating the above.
605 orr r1, r1, r2, lsl #20
606 add r0, r3, r2, lsl #2
613 __arm926ejs_mmu_cache_on:
614 #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
615 mov r0, #4 @ put dcache in WT mode
616 mcr p15, 7, r0, c15, c0, 0
619 __armv4_mmu_cache_on:
624 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
625 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
626 mrc p15, 0, r0, c1, c0, 0 @ read control reg
627 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
629 #ifdef CONFIG_CPU_ENDIAN_BE8
630 orr r0, r0, #1 << 25 @ big-endian page tables
632 bl __common_mmu_cache_on
634 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
638 __armv7_mmu_cache_on:
641 mrc p15, 0, r11, c0, c1, 4 @ read ID_MMFR0
645 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
647 mcrne p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
649 mrc p15, 0, r0, c1, c0, 0 @ read control reg
650 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
651 orr r0, r0, #0x003c @ write buffer
653 #ifdef CONFIG_CPU_ENDIAN_BE8
654 orr r0, r0, #1 << 25 @ big-endian page tables
656 orrne r0, r0, #1 @ MMU enabled
658 mcrne p15, 0, r3, c2, c0, 0 @ load page table pointer
659 mcrne p15, 0, r1, c3, c0, 0 @ load domain access control
661 mcr p15, 0, r0, c1, c0, 0 @ load control register
662 mrc p15, 0, r0, c1, c0, 0 @ and read it back
664 mcr p15, 0, r0, c7, c5, 4 @ ISB
671 mcr p15, 0, r0, c7, c7, 0 @ Invalidate whole cache
672 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
673 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
674 mrc p15, 0, r0, c1, c0, 0 @ read control reg
675 orr r0, r0, #0x1000 @ I-cache enable
676 bl __common_mmu_cache_on
678 mcr p15, 0, r0, c8, c7, 0 @ flush UTLB
685 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
686 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
688 bl __common_mmu_cache_on
690 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
693 __common_mmu_cache_on:
694 #ifndef CONFIG_THUMB2_KERNEL
696 orr r0, r0, #0x000d @ Write buffer, mmu
699 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
700 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
702 .align 5 @ cache line aligned
703 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
704 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
705 sub pc, lr, r0, lsr #32 @ properly flush pipeline
708 #define PROC_ENTRY_SIZE (4*5)
711 * Here follow the relocatable cache support functions for the
712 * various processors. This is a generic hook for locating an
713 * entry and jumping to an instruction at the specified offset
714 * from the start of the block. Please note this is all position
724 call_cache_fn: adr r12, proc_types
725 #ifdef CONFIG_CPU_CP15
726 mrc p15, 0, r9, c0, c0 @ get processor ID
728 ldr r9, =CONFIG_PROCESSOR_ID
730 1: ldr r1, [r12, #0] @ get value
731 ldr r2, [r12, #4] @ get mask
732 eor r1, r1, r9 @ (real ^ match)
734 ARM( addeq pc, r12, r3 ) @ call cache function
735 THUMB( addeq r12, r3 )
736 THUMB( moveq pc, r12 ) @ call cache function
737 add r12, r12, #PROC_ENTRY_SIZE
741 * Table for cache operations. This is basically:
744 * - 'cache on' method instruction
745 * - 'cache off' method instruction
746 * - 'cache flush' method instruction
748 * We match an entry using: ((real_id ^ match) & mask) == 0
750 * Writethrough caches generally only need 'on' and 'off'
751 * methods. Writeback caches _must_ have the flush method
755 .type proc_types,#object
757 .word 0x41560600 @ ARM6/610
759 W(b) __arm6_mmu_cache_off @ works, but slow
760 W(b) __arm6_mmu_cache_off
763 @ b __arm6_mmu_cache_on @ untested
764 @ b __arm6_mmu_cache_off
765 @ b __armv3_mmu_cache_flush
767 .word 0x00000000 @ old ARM ID
776 .word 0x41007000 @ ARM7/710
778 W(b) __arm7_mmu_cache_off
779 W(b) __arm7_mmu_cache_off
783 .word 0x41807200 @ ARM720T (writethrough)
785 W(b) __armv4_mmu_cache_on
786 W(b) __armv4_mmu_cache_off
790 .word 0x41007400 @ ARM74x
792 W(b) __armv3_mpu_cache_on
793 W(b) __armv3_mpu_cache_off
794 W(b) __armv3_mpu_cache_flush
796 .word 0x41009400 @ ARM94x
798 W(b) __armv4_mpu_cache_on
799 W(b) __armv4_mpu_cache_off
800 W(b) __armv4_mpu_cache_flush
802 .word 0x41069260 @ ARM926EJ-S (v5TEJ)
804 W(b) __arm926ejs_mmu_cache_on
805 W(b) __armv4_mmu_cache_off
806 W(b) __armv5tej_mmu_cache_flush
808 .word 0x00007000 @ ARM7 IDs
817 @ Everything from here on will be the new ID system.
819 .word 0x4401a100 @ sa110 / sa1100
821 W(b) __armv4_mmu_cache_on
822 W(b) __armv4_mmu_cache_off
823 W(b) __armv4_mmu_cache_flush
825 .word 0x6901b110 @ sa1110
827 W(b) __armv4_mmu_cache_on
828 W(b) __armv4_mmu_cache_off
829 W(b) __armv4_mmu_cache_flush
832 .word 0xffffff00 @ PXA9xx
833 W(b) __armv4_mmu_cache_on
834 W(b) __armv4_mmu_cache_off
835 W(b) __armv4_mmu_cache_flush
837 .word 0x56158000 @ PXA168
839 W(b) __armv4_mmu_cache_on
840 W(b) __armv4_mmu_cache_off
841 W(b) __armv5tej_mmu_cache_flush
843 .word 0x56050000 @ Feroceon
845 W(b) __armv4_mmu_cache_on
846 W(b) __armv4_mmu_cache_off
847 W(b) __armv5tej_mmu_cache_flush
849 #ifdef CONFIG_CPU_FEROCEON_OLD_ID
850 /* this conflicts with the standard ARMv5TE entry */
851 .long 0x41009260 @ Old Feroceon
853 b __armv4_mmu_cache_on
854 b __armv4_mmu_cache_off
855 b __armv5tej_mmu_cache_flush
858 .word 0x66015261 @ FA526
860 W(b) __fa526_cache_on
861 W(b) __armv4_mmu_cache_off
862 W(b) __fa526_cache_flush
864 @ These match on the architecture ID
866 .word 0x00020000 @ ARMv4T
868 W(b) __armv4_mmu_cache_on
869 W(b) __armv4_mmu_cache_off
870 W(b) __armv4_mmu_cache_flush
872 .word 0x00050000 @ ARMv5TE
874 W(b) __armv4_mmu_cache_on
875 W(b) __armv4_mmu_cache_off
876 W(b) __armv4_mmu_cache_flush
878 .word 0x00060000 @ ARMv5TEJ
880 W(b) __armv4_mmu_cache_on
881 W(b) __armv4_mmu_cache_off
882 W(b) __armv5tej_mmu_cache_flush
884 .word 0x0007b000 @ ARMv6
886 W(b) __armv4_mmu_cache_on
887 W(b) __armv4_mmu_cache_off
888 W(b) __armv6_mmu_cache_flush
890 .word 0x000f0000 @ new CPU Id
892 W(b) __armv7_mmu_cache_on
893 W(b) __armv7_mmu_cache_off
894 W(b) __armv7_mmu_cache_flush
896 .word 0 @ unrecognised type
905 .size proc_types, . - proc_types
908 * If you get a "non-constant expression in ".if" statement"
909 * error from the assembler on this line, check that you have
910 * not accidentally written a "b" instruction where you should
913 .if (. - proc_types) % PROC_ENTRY_SIZE != 0
914 .error "The size of one or more proc_types entries is wrong."
918 * Turn off the Cache and MMU. ARMv3 does not support
919 * reading the control register, but ARMv4 does.
922 * r0, r1, r2, r3, r9, r12 corrupted
923 * This routine must preserve:
927 cache_off: mov r3, #12 @ cache_off function
930 __armv4_mpu_cache_off:
931 mrc p15, 0, r0, c1, c0
933 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
935 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
936 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
937 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
940 __armv3_mpu_cache_off:
941 mrc p15, 0, r0, c1, c0
943 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
945 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
948 __armv4_mmu_cache_off:
950 mrc p15, 0, r0, c1, c0
952 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
954 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
955 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
959 __armv7_mmu_cache_off:
960 mrc p15, 0, r0, c1, c0
966 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
968 bl __armv7_mmu_cache_flush
971 mcr p15, 0, r0, c8, c7, 0 @ invalidate whole TLB
973 mcr p15, 0, r0, c7, c5, 6 @ invalidate BTC
974 mcr p15, 0, r0, c7, c10, 4 @ DSB
975 mcr p15, 0, r0, c7, c5, 4 @ ISB
978 __arm6_mmu_cache_off:
979 mov r0, #0x00000030 @ ARM6 control reg.
980 b __armv3_mmu_cache_off
982 __arm7_mmu_cache_off:
983 mov r0, #0x00000070 @ ARM7 control reg.
984 b __armv3_mmu_cache_off
986 __armv3_mmu_cache_off:
987 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
989 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
990 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
994 * Clean and flush the cache to maintain consistency.
997 * r1, r2, r3, r9, r10, r11, r12 corrupted
998 * This routine must preserve:
1006 __armv4_mpu_cache_flush:
1009 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
1010 mov r1, #7 << 5 @ 8 segments
1011 1: orr r3, r1, #63 << 26 @ 64 entries
1012 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
1013 subs r3, r3, #1 << 26
1014 bcs 2b @ entries 63 to 0
1015 subs r1, r1, #1 << 5
1016 bcs 1b @ segments 7 to 0
1019 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
1020 mcr p15, 0, ip, c7, c10, 4 @ drain WB
1023 __fa526_cache_flush:
1025 mcr p15, 0, r1, c7, c14, 0 @ clean and invalidate D cache
1026 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1027 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1030 __armv6_mmu_cache_flush:
1032 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
1033 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
1034 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
1035 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1038 __armv7_mmu_cache_flush:
1039 mrc p15, 0, r10, c0, c1, 5 @ read ID_MMFR1
1040 tst r10, #0xf << 16 @ hierarchical cache (ARMv7)
1043 mcr p15, 0, r10, c7, c14, 0 @ clean+invalidate D
1046 mcr p15, 0, r10, c7, c10, 5 @ DMB
1047 stmfd sp!, {r0-r7, r9-r11}
1048 mrc p15, 1, r0, c0, c0, 1 @ read clidr
1049 ands r3, r0, #0x7000000 @ extract loc from clidr
1050 mov r3, r3, lsr #23 @ left align loc bit field
1051 beq finished @ if loc is 0, then no need to clean
1052 mov r10, #0 @ start clean at cache level 0
1054 add r2, r10, r10, lsr #1 @ work out 3x current cache level
1055 mov r1, r0, lsr r2 @ extract cache type bits from clidr
1056 and r1, r1, #7 @ mask of the bits for current cache only
1057 cmp r1, #2 @ see what cache we have at this level
1058 blt skip @ skip if no cache, or just i-cache
1059 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1060 mcr p15, 0, r10, c7, c5, 4 @ isb to sych the new cssr&csidr
1061 mrc p15, 1, r1, c0, c0, 0 @ read the new csidr
1062 and r2, r1, #7 @ extract the length of the cache lines
1063 add r2, r2, #4 @ add 4 (line length offset)
1065 ands r4, r4, r1, lsr #3 @ find maximum number on the way size
1066 clz r5, r4 @ find bit position of way size increment
1068 ands r7, r7, r1, lsr #13 @ extract max number of the index size
1070 mov r9, r4 @ create working copy of max way size
1072 ARM( orr r11, r10, r9, lsl r5 ) @ factor way and cache number into r11
1073 ARM( orr r11, r11, r7, lsl r2 ) @ factor index number into r11
1074 THUMB( lsl r6, r9, r5 )
1075 THUMB( orr r11, r10, r6 ) @ factor way and cache number into r11
1076 THUMB( lsl r6, r7, r2 )
1077 THUMB( orr r11, r11, r6 ) @ factor index number into r11
1078 mcr p15, 0, r11, c7, c14, 2 @ clean & invalidate by set/way
1079 subs r9, r9, #1 @ decrement the way
1081 subs r7, r7, #1 @ decrement the index
1084 add r10, r10, #2 @ increment cache number
1088 ldmfd sp!, {r0-r7, r9-r11}
1089 mov r10, #0 @ swith back to cache level 0
1090 mcr p15, 2, r10, c0, c0, 0 @ select current cache level in cssr
1092 mcr p15, 0, r10, c7, c10, 4 @ DSB
1093 mcr p15, 0, r10, c7, c5, 0 @ invalidate I+BTB
1094 mcr p15, 0, r10, c7, c10, 4 @ DSB
1095 mcr p15, 0, r10, c7, c5, 4 @ ISB
1098 __armv5tej_mmu_cache_flush:
1099 1: mrc p15, 0, r15, c7, c14, 3 @ test,clean,invalidate D cache
1101 mcr p15, 0, r0, c7, c5, 0 @ flush I cache
1102 mcr p15, 0, r0, c7, c10, 4 @ drain WB
1105 __armv4_mmu_cache_flush:
1106 mov r2, #64*1024 @ default: 32K dcache size (*2)
1107 mov r11, #32 @ default: 32 byte line size
1108 mrc p15, 0, r3, c0, c0, 1 @ read cache type
1109 teq r3, r9 @ cache ID register present?
1114 mov r2, r2, lsl r1 @ base dcache size *2
1115 tst r3, #1 << 14 @ test M bit
1116 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
1120 mov r11, r11, lsl r3 @ cache line size in bytes
1123 bic r1, r1, #63 @ align to longest cache line
1126 ARM( ldr r3, [r1], r11 ) @ s/w flush D cache
1127 THUMB( ldr r3, [r1] ) @ s/w flush D cache
1128 THUMB( add r1, r1, r11 )
1132 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
1133 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
1134 mcr p15, 0, r1, c7, c10, 4 @ drain WB
1137 __armv3_mmu_cache_flush:
1138 __armv3_mpu_cache_flush:
1140 mcr p15, 0, r1, c7, c0, 0 @ invalidate whole cache v3
1144 * Various debugging routines for printing hex characters and
1145 * memory, which again must be relocatable.
1149 .type phexbuf,#object
1151 .size phexbuf, . - phexbuf
1153 @ phex corrupts {r0, r1, r2, r3}
1154 phex: adr r3, phexbuf
1168 @ puts corrupts {r0, r1, r2, r3}
1170 1: ldrb r2, [r0], #1
1183 @ putc corrupts {r0, r1, r2, r3}
1190 @ memdump corrupts {r0, r1, r2, r3, r10, r11, r12, lr}
1191 memdump: mov r12, r0
1194 2: mov r0, r11, lsl #2
1202 ldr r0, [r12, r11, lsl #2]
1224 .section ".stack", "aw", %nobits
1225 .L_user_stack: .space 4096