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)
26 mcr p14, 0, \ch, c0, c1, 0
30 #include <asm/arch/debug-macro.S>
36 #if defined(CONFIG_ARCH_SA1100)
38 mov \rb, #0x80000000 @ physical base address
39 #ifdef CONFIG_DEBUG_LL_SER3
40 add \rb, \rb, #0x00050000 @ Ser3
42 add \rb, \rb, #0x00010000 @ Ser1
45 #elif defined(CONFIG_ARCH_IOP331)
48 orr \rb, \rb, #0x00ff0000
49 orr \rb, \rb, #0x0000f700 @ location of the UART
51 #elif defined(CONFIG_ARCH_S3C2410)
54 add \rb, \rb, #0x4000 * CONFIG_S3C2410_LOWLEVEL_UART_PORT
75 .macro debug_reloc_start
78 kphex r6, 8 /* processor id */
80 kphex r7, 8 /* architecture id */
82 mrc p15, 0, r0, c1, c0
83 kphex r0, 8 /* control reg */
85 kphex r5, 8 /* decompressed kernel start */
87 kphex r9, 8 /* decompressed kernel end */
89 kphex r4, 8 /* kernel execution address */
94 .macro debug_reloc_end
96 kphex r5, 8 /* end of kernel */
99 bl memdump /* dump 256 bytes at start of kernel */
103 .section ".start", #alloc, #execinstr
105 * sort out different calling conventions
109 .type start,#function
115 .word 0x016f2818 @ Magic numbers to help the loader
116 .word start @ absolute load/run zImage address
117 .word _edata @ zImage end address
118 1: mov r7, r1 @ save architecture ID
119 mov r8, r2 @ save atags pointer
121 #ifndef __ARM_ARCH_2__
123 * Booting from Angel - need to enter SVC mode and disable
124 * FIQs/IRQs (numeric definitions from angel arm.h source).
125 * We only do this if we were in user mode on entry.
127 mrs r2, cpsr @ get current mode
128 tst r2, #3 @ not user?
130 mov r0, #0x17 @ angel_SWIreason_EnterSVC
131 swi 0x123456 @ angel_SWI_ARM
133 mrs r2, cpsr @ turn off interrupts to
134 orr r2, r2, #0xc0 @ prevent angel from running
137 teqp pc, #0x0c000003 @ turn off interrupts
141 * Note that some cache flushing and other stuff may
142 * be needed here - is there an Angel SWI call for this?
146 * some architecture specific code can be inserted
147 * by the linker here, but it should preserve r7, r8, and r9.
152 ldmia r0, {r1, r2, r3, r4, r5, r6, ip, sp}
153 subs r0, r0, r1 @ calculate the delta offset
155 @ if delta is zero, we are
156 beq not_relocated @ running at the address we
160 * We're running at a different address. We need to fix
161 * up various pointers:
162 * r5 - zImage base address
170 #ifndef CONFIG_ZBOOT_ROM
172 * If we're running fully PIC === CONFIG_ZBOOT_ROM = n,
173 * we need to fix up pointers into the BSS region.
183 * Relocate all entries in the GOT table.
185 1: ldr r1, [r6, #0] @ relocate entries in the GOT
186 add r1, r1, r0 @ table. This fixes up the
187 str r1, [r6], #4 @ C references.
193 * Relocate entries in the GOT table. We only relocate
194 * the entries that are outside the (relocated) BSS region.
196 1: ldr r1, [r6, #0] @ relocate entries in the GOT
197 cmp r1, r2 @ entry < bss_start ||
198 cmphs r3, r1 @ _end < entry
199 addlo r1, r1, r0 @ table. This fixes up the
200 str r1, [r6], #4 @ C references.
205 not_relocated: mov r0, #0
206 1: str r0, [r2], #4 @ clear bss
214 * The C runtime environment should now be setup
215 * sufficiently. Turn the cache on, set up some
216 * pointers, and start decompressing.
220 mov r1, sp @ malloc space above stack
221 add r2, sp, #0x10000 @ 64k max
224 * Check to see if we will overwrite ourselves.
225 * r4 = final kernel address
226 * r5 = start of this image
227 * r2 = end of malloc space (and therefore this image)
230 * r4 + image length <= r5 -> OK
234 add r0, r4, #4096*1024 @ 4MB largest kernel size
238 mov r5, r2 @ decompress after malloc space
244 bic r0, r0, #127 @ align the kernel length
246 * r0 = decompressed kernel length
248 * r4 = kernel execution address
249 * r5 = decompressed kernel start
251 * r7 = architecture ID
255 add r1, r5, r0 @ end of decompressed kernel
259 1: ldmia r2!, {r9 - r14} @ copy relocation code
260 stmia r1!, {r9 - r14}
261 ldmia r2!, {r9 - r14}
262 stmia r1!, {r9 - r14}
267 add pc, r5, r0 @ call relocation code
270 * We're not in danger of overwriting ourselves. Do this the simple way.
272 * r4 = kernel execution address
273 * r7 = architecture ID
275 wont_overwrite: mov r0, r4
282 .word __bss_start @ r2
286 .word _got_start @ r6
288 .word user_stack+4096 @ sp
289 LC1: .word reloc_end - reloc_start
292 #ifdef CONFIG_ARCH_RPC
294 params: ldr r0, =params_phys
301 * Turn on the cache. We need to setup some page tables so that we
302 * can have both the I and D caches on.
304 * We place the page tables 16k down from the kernel execution address,
305 * and we hope that nothing else is using it. If we're using it, we
309 * r4 = kernel execution address
311 * r7 = architecture number
313 * r9 = run-time address of "start" (???)
315 * r1, r2, r3, r9, r10, r12 corrupted
316 * This routine must preserve:
320 cache_on: mov r3, #8 @ cache_on function
324 * Initialize the highest priority protection region, PR7
325 * to cover all 32bit address and cacheable and bufferable.
327 __armv4_mpu_cache_on:
328 mov r0, #0x3f @ 4G, the whole
329 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
330 mcr p15, 0, r0, c6, c7, 1
333 mcr p15, 0, r0, c2, c0, 0 @ D-cache on
334 mcr p15, 0, r0, c2, c0, 1 @ I-cache on
335 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
338 mcr p15, 0, r0, c5, c0, 1 @ I-access permission
339 mcr p15, 0, r0, c5, c0, 0 @ D-access permission
342 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
343 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
344 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
345 mrc p15, 0, r0, c1, c0, 0 @ read control reg
346 @ ...I .... ..D. WC.M
347 orr r0, r0, #0x002d @ .... .... ..1. 11.1
348 orr r0, r0, #0x1000 @ ...1 .... .... ....
350 mcr p15, 0, r0, c1, c0, 0 @ write control reg
353 mcr p15, 0, r0, c7, c5, 0 @ flush(inval) I-Cache
354 mcr p15, 0, r0, c7, c6, 0 @ flush(inval) D-Cache
357 __armv3_mpu_cache_on:
358 mov r0, #0x3f @ 4G, the whole
359 mcr p15, 0, r0, c6, c7, 0 @ PR7 Area Setting
362 mcr p15, 0, r0, c2, c0, 0 @ cache on
363 mcr p15, 0, r0, c3, c0, 0 @ write-buffer on
366 mcr p15, 0, r0, c5, c0, 0 @ access permission
369 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
370 mrc p15, 0, r0, c1, c0, 0 @ read control reg
371 @ .... .... .... WC.M
372 orr r0, r0, #0x000d @ .... .... .... 11.1
374 mcr p15, 0, r0, c1, c0, 0 @ write control reg
376 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
379 __setup_mmu: sub r3, r4, #16384 @ Page directory size
380 bic r3, r3, #0xff @ Align the pointer
383 * Initialise the page tables, turning on the cacheable and bufferable
384 * bits for the RAM area only.
388 mov r9, r9, lsl #18 @ start of RAM
389 add r10, r9, #0x10000000 @ a reasonable RAM size
393 1: cmp r1, r9 @ if virt > start of RAM
394 orrhs r1, r1, #0x0c @ set cacheable, bufferable
395 cmp r1, r10 @ if virt > end of RAM
396 bichs r1, r1, #0x0c @ clear cacheable, bufferable
397 str r1, [r0], #4 @ 1:1 mapping
402 * If ever we are running from Flash, then we surely want the cache
403 * to be enabled also for our execution instance... We map 2MB of it
404 * so there is no map overlap problem for up to 1 MB compressed kernel.
405 * If the execution is in RAM then we would only be duplicating the above.
410 orr r1, r1, r2, lsl #20
411 add r0, r3, r2, lsl #2
417 __armv4_mmu_cache_on:
421 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
422 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
423 mrc p15, 0, r0, c1, c0, 0 @ read control reg
424 orr r0, r0, #0x5000 @ I-cache enable, RR cache replacement
426 bl __common_mmu_cache_on
428 mcr p15, 0, r0, c8, c7, 0 @ flush I,D TLBs
435 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
436 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
438 bl __common_mmu_cache_on
440 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
443 __common_mmu_cache_on:
445 orr r0, r0, #0x000d @ Write buffer, mmu
448 mcr p15, 0, r3, c2, c0, 0 @ load page table pointer
449 mcr p15, 0, r1, c3, c0, 0 @ load domain access control
451 .align 5 @ cache line aligned
452 1: mcr p15, 0, r0, c1, c0, 0 @ load control register
453 mrc p15, 0, r0, c1, c0, 0 @ and read it back to
454 sub pc, lr, r0, lsr #32 @ properly flush pipeline
457 * All code following this line is relocatable. It is relocated by
458 * the above code to the end of the decompressed kernel image and
459 * executed there. During this time, we have no stacks.
461 * r0 = decompressed kernel length
463 * r4 = kernel execution address
464 * r5 = decompressed kernel start
466 * r7 = architecture ID
471 reloc_start: add r9, r5, r0
476 ldmia r5!, {r0, r2, r3, r10 - r14} @ relocate kernel
477 stmia r1!, {r0, r2, r3, r10 - r14}
484 call_kernel: bl cache_clean_flush
486 mov r0, #0 @ must be zero
487 mov r1, r7 @ restore architecture number
488 mov r2, r8 @ restore atags pointer
489 mov pc, r4 @ call kernel
492 * Here follow the relocatable cache support functions for the
493 * various processors. This is a generic hook for locating an
494 * entry and jumping to an instruction at the specified offset
495 * from the start of the block. Please note this is all position
505 call_cache_fn: adr r12, proc_types
506 mrc p15, 0, r6, c0, c0 @ get processor ID
507 1: ldr r1, [r12, #0] @ get value
508 ldr r2, [r12, #4] @ get mask
509 eor r1, r1, r6 @ (real ^ match)
511 addeq pc, r12, r3 @ call cache function
516 * Table for cache operations. This is basically:
519 * - 'cache on' method instruction
520 * - 'cache off' method instruction
521 * - 'cache flush' method instruction
523 * We match an entry using: ((real_id ^ match) & mask) == 0
525 * Writethrough caches generally only need 'on' and 'off'
526 * methods. Writeback caches _must_ have the flush method
529 .type proc_types,#object
531 .word 0x41560600 @ ARM6/610
533 b __arm6_mmu_cache_off @ works, but slow
534 b __arm6_mmu_cache_off
536 @ b __arm6_mmu_cache_on @ untested
537 @ b __arm6_mmu_cache_off
538 @ b __armv3_mmu_cache_flush
540 .word 0x00000000 @ old ARM ID
546 .word 0x41007000 @ ARM7/710
548 b __arm7_mmu_cache_off
549 b __arm7_mmu_cache_off
552 .word 0x41807200 @ ARM720T (writethrough)
554 b __armv4_mmu_cache_on
555 b __armv4_mmu_cache_off
558 .word 0x41007400 @ ARM74x
560 b __armv3_mpu_cache_on
561 b __armv3_mpu_cache_off
562 b __armv3_mpu_cache_flush
564 .word 0x41009400 @ ARM94x
566 b __armv4_mpu_cache_on
567 b __armv4_mpu_cache_off
568 b __armv4_mpu_cache_flush
570 .word 0x00007000 @ ARM7 IDs
576 @ Everything from here on will be the new ID system.
578 .word 0x4401a100 @ sa110 / sa1100
580 b __armv4_mmu_cache_on
581 b __armv4_mmu_cache_off
582 b __armv4_mmu_cache_flush
584 .word 0x6901b110 @ sa1110
586 b __armv4_mmu_cache_on
587 b __armv4_mmu_cache_off
588 b __armv4_mmu_cache_flush
590 @ These match on the architecture ID
592 .word 0x00020000 @ ARMv4T
594 b __armv4_mmu_cache_on
595 b __armv4_mmu_cache_off
596 b __armv4_mmu_cache_flush
598 .word 0x00050000 @ ARMv5TE
600 b __armv4_mmu_cache_on
601 b __armv4_mmu_cache_off
602 b __armv4_mmu_cache_flush
604 .word 0x00060000 @ ARMv5TEJ
606 b __armv4_mmu_cache_on
607 b __armv4_mmu_cache_off
608 b __armv4_mmu_cache_flush
610 .word 0x0007b000 @ ARMv6
612 b __armv4_mmu_cache_on
613 b __armv4_mmu_cache_off
614 b __armv6_mmu_cache_flush
616 .word 0 @ unrecognised type
622 .size proc_types, . - proc_types
625 * Turn off the Cache and MMU. ARMv3 does not support
626 * reading the control register, but ARMv4 does.
628 * On entry, r6 = processor ID
629 * On exit, r0, r1, r2, r3, r12 corrupted
630 * This routine must preserve: r4, r6, r7
633 cache_off: mov r3, #12 @ cache_off function
636 __armv4_mpu_cache_off:
637 mrc p15, 0, r0, c1, c0
639 mcr p15, 0, r0, c1, c0 @ turn MPU and cache off
641 mcr p15, 0, r0, c7, c10, 4 @ drain write buffer
642 mcr p15, 0, r0, c7, c6, 0 @ flush D-Cache
643 mcr p15, 0, r0, c7, c5, 0 @ flush I-Cache
646 __armv3_mpu_cache_off:
647 mrc p15, 0, r0, c1, c0
649 mcr p15, 0, r0, c1, c0, 0 @ turn MPU and cache off
651 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
654 __armv4_mmu_cache_off:
655 mrc p15, 0, r0, c1, c0
657 mcr p15, 0, r0, c1, c0 @ turn MMU and cache off
659 mcr p15, 0, r0, c7, c7 @ invalidate whole cache v4
660 mcr p15, 0, r0, c8, c7 @ invalidate whole TLB v4
663 __arm6_mmu_cache_off:
664 mov r0, #0x00000030 @ ARM6 control reg.
665 b __armv3_mmu_cache_off
667 __arm7_mmu_cache_off:
668 mov r0, #0x00000070 @ ARM7 control reg.
669 b __armv3_mmu_cache_off
671 __armv3_mmu_cache_off:
672 mcr p15, 0, r0, c1, c0, 0 @ turn MMU and cache off
674 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
675 mcr p15, 0, r0, c5, c0, 0 @ invalidate whole TLB v3
679 * Clean and flush the cache to maintain consistency.
684 * r1, r2, r3, r11, r12 corrupted
685 * This routine must preserve:
693 __armv4_mpu_cache_flush:
696 mcr p15, 0, ip, c7, c6, 0 @ invalidate D cache
697 mov r1, #7 << 5 @ 8 segments
698 1: orr r3, r1, #63 << 26 @ 64 entries
699 2: mcr p15, 0, r3, c7, c14, 2 @ clean & invalidate D index
700 subs r3, r3, #1 << 26
701 bcs 2b @ entries 63 to 0
703 bcs 1b @ segments 7 to 0
706 mcrne p15, 0, ip, c7, c5, 0 @ invalidate I cache
707 mcr p15, 0, ip, c7, c10, 4 @ drain WB
711 __armv6_mmu_cache_flush:
713 mcr p15, 0, r1, c7, c14, 0 @ clean+invalidate D
714 mcr p15, 0, r1, c7, c5, 0 @ invalidate I+BTB
715 mcr p15, 0, r1, c7, c15, 0 @ clean+invalidate unified
716 mcr p15, 0, r1, c7, c10, 4 @ drain WB
719 __armv4_mmu_cache_flush:
720 mov r2, #64*1024 @ default: 32K dcache size (*2)
721 mov r11, #32 @ default: 32 byte line size
722 mrc p15, 0, r3, c0, c0, 1 @ read cache type
723 teq r3, r6 @ cache ID register present?
728 mov r2, r2, lsl r1 @ base dcache size *2
729 tst r3, #1 << 14 @ test M bit
730 addne r2, r2, r2, lsr #1 @ +1/2 size if M == 1
734 mov r11, r11, lsl r3 @ cache line size in bytes
736 bic r1, pc, #63 @ align to longest cache line
738 1: ldr r3, [r1], r11 @ s/w flush D cache
742 mcr p15, 0, r1, c7, c5, 0 @ flush I cache
743 mcr p15, 0, r1, c7, c6, 0 @ flush D cache
744 mcr p15, 0, r1, c7, c10, 4 @ drain WB
747 __armv3_mmu_cache_flush:
748 __armv3_mpu_cache_flush:
750 mcr p15, 0, r0, c7, c0, 0 @ invalidate whole cache v3
754 * Various debugging routines for printing hex characters and
755 * memory, which again must be relocatable.
758 .type phexbuf,#object
760 .size phexbuf, . - phexbuf
762 phex: adr r3, phexbuf
799 2: mov r0, r11, lsl #2
807 ldr r0, [r12, r11, lsl #2]
828 .section ".stack", "w"
829 user_stack: .space 4096