1 /*
2 * linux/arch/arm/kernel/head.S
3 *
4 * Copyright (C) 1994-2002 Russell King
5 * Copyright (c) 2003 ARM Limited
6 * All Rights Reserved
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
11 *
12 * Kernel startup code for all 32-bit CPUs
13 */
14 #include <linux/linkage.h>
15 #include <linux/init.h>
17 #include <asm/assembler.h>
18 #include <asm/cp15.h>
19 #include <asm/domain.h>
20 #include <asm/ptrace.h>
21 #include <asm/asm-offsets.h>
22 #include <asm/memory.h>
23 #include <asm/thread_info.h>
24 #include <asm/pgtable.h>
26 #if defined(CONFIG_DEBUG_LL) && !defined(CONFIG_DEBUG_SEMIHOSTING)
27 #include CONFIG_DEBUG_LL_INCLUDE
28 #endif
30 /*
31 * swapper_pg_dir is the virtual address of the initial page table.
32 * We place the page tables 16K below KERNEL_RAM_VADDR. Therefore, we must
33 * make sure that KERNEL_RAM_VADDR is correctly set. Currently, we expect
34 * the least significant 16 bits to be 0x8000, but we could probably
35 * relax this restriction to KERNEL_RAM_VADDR >= PAGE_OFFSET + 0x4000.
36 */
37 #define KERNEL_RAM_VADDR (PAGE_OFFSET + TEXT_OFFSET)
38 #if (KERNEL_RAM_VADDR & 0xffff) != 0x8000
39 #error KERNEL_RAM_VADDR must start at 0xXXXX8000
40 #endif
42 #ifdef CONFIG_ARM_LPAE
43 /* LPAE requires an additional page for the PGD */
44 #define PG_DIR_SIZE 0x5000
45 #define PMD_ORDER 3
46 #else
47 #define PG_DIR_SIZE 0x4000
48 #define PMD_ORDER 2
49 #endif
51 .globl swapper_pg_dir
52 .equ swapper_pg_dir, KERNEL_RAM_VADDR - PG_DIR_SIZE
54 .macro pgtbl, rd, phys
55 add \rd, \phys, #TEXT_OFFSET - PG_DIR_SIZE
56 .endm
58 /*
59 * Kernel startup entry point.
60 * ---------------------------
61 *
62 * This is normally called from the decompressor code. The requirements
63 * are: MMU = off, D-cache = off, I-cache = dont care, r0 = 0,
64 * r1 = machine nr, r2 = atags or dtb pointer.
65 *
66 * This code is mostly position independent, so if you link the kernel at
67 * 0xc0008000, you call this at __pa(0xc0008000).
68 *
69 * See linux/arch/arm/tools/mach-types for the complete list of machine
70 * numbers for r1.
71 *
72 * We're trying to keep crap to a minimum; DO NOT add any machine specific
73 * crap here - that's what the boot loader (or in extreme, well justified
74 * circumstances, zImage) is for.
75 */
76 .arm
78 __HEAD
79 ENTRY(stext)
81 THUMB( adr r9, BSYM(1f) ) @ Kernel is always entered in ARM.
82 THUMB( bx r9 ) @ If this is a Thumb-2 kernel,
83 THUMB( .thumb ) @ switch to Thumb now.
84 THUMB(1: )
86 #ifdef CONFIG_ARM_VIRT_EXT
87 bl __hyp_stub_install
88 #endif
89 @ ensure svc mode and all interrupts masked
90 safe_svcmode_maskall r9
92 mrc p15, 0, r9, c0, c0 @ get processor id
93 bl __lookup_processor_type @ r5=procinfo r9=cpuid
94 movs r10, r5 @ invalid processor (r5=0)?
95 THUMB( it eq ) @ force fixup-able long branch encoding
96 beq __error_p @ yes, error 'p'
98 #ifdef CONFIG_ARM_LPAE
99 mrc p15, 0, r3, c0, c1, 4 @ read ID_MMFR0
100 and r3, r3, #0xf @ extract VMSA support
101 cmp r3, #5 @ long-descriptor translation table format?
102 THUMB( it lo ) @ force fixup-able long branch encoding
103 blo __error_p @ only classic page table format
104 #endif
106 #ifndef CONFIG_XIP_KERNEL
107 adr r3, 2f
108 ldmia r3, {r4, r8}
109 sub r4, r3, r4 @ (PHYS_OFFSET - PAGE_OFFSET)
110 add r8, r8, r4 @ PHYS_OFFSET
111 #else
112 ldr r8, =PHYS_OFFSET @ always constant in this case
113 #endif
115 /*
116 * r1 = machine no, r2 = atags or dtb,
117 * r8 = phys_offset, r9 = cpuid, r10 = procinfo
118 */
119 bl __vet_atags
120 #ifdef CONFIG_SMP_ON_UP
121 bl __fixup_smp
122 #endif
123 #ifdef CONFIG_ARM_PATCH_PHYS_VIRT
124 bl __fixup_pv_table
125 #endif
126 bl __create_page_tables
128 /*
129 * The following calls CPU specific code in a position independent
130 * manner. See arch/arm/mm/proc-*.S for details. r10 = base of
131 * xxx_proc_info structure selected by __lookup_processor_type
132 * above. On return, the CPU will be ready for the MMU to be
133 * turned on, and r0 will hold the CPU control register value.
134 */
135 ldr r13, =__mmap_switched @ address to jump to after
136 @ mmu has been enabled
137 adr lr, BSYM(1f) @ return (PIC) address
138 mov r8, r4 @ set TTBR1 to swapper_pg_dir
139 ARM( add pc, r10, #PROCINFO_INITFUNC )
140 THUMB( add r12, r10, #PROCINFO_INITFUNC )
141 THUMB( mov pc, r12 )
142 1: b __enable_mmu
143 ENDPROC(stext)
144 .ltorg
145 #ifndef CONFIG_XIP_KERNEL
146 2: .long .
147 .long PAGE_OFFSET
148 #endif
150 /*
151 * Setup the initial page tables. We only setup the barest
152 * amount which are required to get the kernel running, which
153 * generally means mapping in the kernel code.
154 *
155 * r8 = phys_offset, r9 = cpuid, r10 = procinfo
156 *
157 * Returns:
158 * r0, r3, r5-r7 corrupted
159 * r4 = physical page table address
160 */
161 __create_page_tables:
162 pgtbl r4, r8 @ page table address
164 /*
165 * Clear the swapper page table
166 */
167 mov r0, r4
168 mov r3, #0
169 add r6, r0, #PG_DIR_SIZE
170 1: str r3, [r0], #4
171 str r3, [r0], #4
172 str r3, [r0], #4
173 str r3, [r0], #4
174 teq r0, r6
175 bne 1b
177 #ifdef CONFIG_ARM_LPAE
178 /*
179 * Build the PGD table (first level) to point to the PMD table. A PGD
180 * entry is 64-bit wide.
181 */
182 mov r0, r4
183 add r3, r4, #0x1000 @ first PMD table address
184 orr r3, r3, #3 @ PGD block type
185 mov r6, #4 @ PTRS_PER_PGD
186 mov r7, #1 << (55 - 32) @ L_PGD_SWAPPER
187 1: str r3, [r0], #4 @ set bottom PGD entry bits
188 str r7, [r0], #4 @ set top PGD entry bits
189 add r3, r3, #0x1000 @ next PMD table
190 subs r6, r6, #1
191 bne 1b
193 add r4, r4, #0x1000 @ point to the PMD tables
194 #endif
196 ldr r7, [r10, #PROCINFO_MM_MMUFLAGS] @ mm_mmuflags
198 /*
199 * Create identity mapping to cater for __enable_mmu.
200 * This identity mapping will be removed by paging_init().
201 */
202 adr r0, __turn_mmu_on_loc
203 ldmia r0, {r3, r5, r6}
204 sub r0, r0, r3 @ virt->phys offset
205 add r5, r5, r0 @ phys __turn_mmu_on
206 add r6, r6, r0 @ phys __turn_mmu_on_end
207 mov r5, r5, lsr #SECTION_SHIFT
208 mov r6, r6, lsr #SECTION_SHIFT
210 1: orr r3, r7, r5, lsl #SECTION_SHIFT @ flags + kernel base
211 str r3, [r4, r5, lsl #PMD_ORDER] @ identity mapping
212 cmp r5, r6
213 addlo r5, r5, #1 @ next section
214 blo 1b
216 /*
217 * Map our RAM from the start to the end of the kernel .bss section.
218 */
219 add r0, r4, #PAGE_OFFSET >> (SECTION_SHIFT - PMD_ORDER)
220 ldr r6, =(_end - 1)
221 orr r3, r8, r7
222 add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER)
223 1: str r3, [r0], #1 << PMD_ORDER
224 add r3, r3, #1 << SECTION_SHIFT
225 cmp r0, r6
226 bls 1b
228 #ifdef CONFIG_XIP_KERNEL
229 /*
230 * Map the kernel image separately as it is not located in RAM.
231 */
232 #define XIP_START XIP_VIRT_ADDR(CONFIG_XIP_PHYS_ADDR)
233 mov r3, pc
234 mov r3, r3, lsr #SECTION_SHIFT
235 orr r3, r7, r3, lsl #SECTION_SHIFT
236 add r0, r4, #(XIP_START & 0xff000000) >> (SECTION_SHIFT - PMD_ORDER)
237 str r3, [r0, #((XIP_START & 0x00f00000) >> SECTION_SHIFT) << PMD_ORDER]!
238 ldr r6, =(_edata_loc - 1)
239 add r0, r0, #1 << PMD_ORDER
240 add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER)
241 1: cmp r0, r6
242 add r3, r3, #1 << SECTION_SHIFT
243 strls r3, [r0], #1 << PMD_ORDER
244 bls 1b
245 #endif
247 /*
248 * Then map boot params address in r2 if specified.
249 * We map 2 sections in case the ATAGs/DTB crosses a section boundary.
250 */
251 mov r0, r2, lsr #SECTION_SHIFT
252 movs r0, r0, lsl #SECTION_SHIFT
253 subne r3, r0, r8
254 addne r3, r3, #PAGE_OFFSET
255 addne r3, r4, r3, lsr #(SECTION_SHIFT - PMD_ORDER)
256 orrne r6, r7, r0
257 strne r6, [r3], #1 << PMD_ORDER
258 addne r6, r6, #1 << SECTION_SHIFT
259 strne r6, [r3]
261 #ifdef CONFIG_DEBUG_LL
262 #if !defined(CONFIG_DEBUG_ICEDCC) && !defined(CONFIG_DEBUG_SEMIHOSTING)
263 /*
264 * Map in IO space for serial debugging.
265 * This allows debug messages to be output
266 * via a serial console before paging_init.
267 */
268 addruart r7, r3, r0
270 mov r3, r3, lsr #SECTION_SHIFT
271 mov r3, r3, lsl #PMD_ORDER
273 add r0, r4, r3
274 mov r3, r7, lsr #SECTION_SHIFT
275 ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags
276 orr r3, r7, r3, lsl #SECTION_SHIFT
277 #ifdef CONFIG_ARM_LPAE
278 mov r7, #1 << (54 - 32) @ XN
279 #else
280 orr r3, r3, #PMD_SECT_XN
281 #endif
282 str r3, [r0], #4
283 #ifdef CONFIG_ARM_LPAE
284 str r7, [r0], #4
285 #endif
287 #else /* CONFIG_DEBUG_ICEDCC || CONFIG_DEBUG_SEMIHOSTING */
288 /* we don't need any serial debugging mappings */
289 ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags
290 #endif
292 #if defined(CONFIG_ARCH_NETWINDER) || defined(CONFIG_ARCH_CATS)
293 /*
294 * If we're using the NetWinder or CATS, we also need to map
295 * in the 16550-type serial port for the debug messages
296 */
297 add r0, r4, #0xff000000 >> (SECTION_SHIFT - PMD_ORDER)
298 orr r3, r7, #0x7c000000
299 str r3, [r0]
300 #endif
301 #ifdef CONFIG_ARCH_RPC
302 /*
303 * Map in screen at 0x02000000 & SCREEN2_BASE
304 * Similar reasons here - for debug. This is
305 * only for Acorn RiscPC architectures.
306 */
307 add r0, r4, #0x02000000 >> (SECTION_SHIFT - PMD_ORDER)
308 orr r3, r7, #0x02000000
309 str r3, [r0]
310 add r0, r4, #0xd8000000 >> (SECTION_SHIFT - PMD_ORDER)
311 str r3, [r0]
312 #endif
313 #endif
314 #ifdef CONFIG_ARM_LPAE
315 sub r4, r4, #0x1000 @ point to the PGD table
316 #endif
317 mov pc, lr
318 ENDPROC(__create_page_tables)
319 .ltorg
320 .align
321 __turn_mmu_on_loc:
322 .long .
323 .long __turn_mmu_on
324 .long __turn_mmu_on_end
326 #if defined(CONFIG_SMP)
327 __CPUINIT
328 ENTRY(secondary_startup)
329 /*
330 * Common entry point for secondary CPUs.
331 *
332 * Ensure that we're in SVC mode, and IRQs are disabled. Lookup
333 * the processor type - there is no need to check the machine type
334 * as it has already been validated by the primary processor.
335 */
336 #ifdef CONFIG_ARM_VIRT_EXT
337 bl __hyp_stub_install_secondary
338 #endif
339 safe_svcmode_maskall r9
341 mrc p15, 0, r9, c0, c0 @ get processor id
342 bl __lookup_processor_type
343 movs r10, r5 @ invalid processor?
344 moveq r0, #'p' @ yes, error 'p'
345 THUMB( it eq ) @ force fixup-able long branch encoding
346 beq __error_p
348 /*
349 * Use the page tables supplied from __cpu_up.
350 */
351 adr r4, __secondary_data
352 ldmia r4, {r5, r7, r12} @ address to jump to after
353 sub lr, r4, r5 @ mmu has been enabled
354 ldr r4, [r7, lr] @ get secondary_data.pgdir
355 add r7, r7, #4
356 ldr r8, [r7, lr] @ get secondary_data.swapper_pg_dir
357 adr lr, BSYM(__enable_mmu) @ return address
358 mov r13, r12 @ __secondary_switched address
359 ARM( add pc, r10, #PROCINFO_INITFUNC ) @ initialise processor
360 @ (return control reg)
361 THUMB( add r12, r10, #PROCINFO_INITFUNC )
362 THUMB( mov pc, r12 )
363 ENDPROC(secondary_startup)
365 /*
366 * r6 = &secondary_data
367 */
368 ENTRY(__secondary_switched)
369 ldr sp, [r7, #4] @ get secondary_data.stack
370 mov fp, #0
371 b secondary_start_kernel
372 ENDPROC(__secondary_switched)
374 .align
376 .type __secondary_data, %object
377 __secondary_data:
378 .long .
379 .long secondary_data
380 .long __secondary_switched
381 #endif /* defined(CONFIG_SMP) */
385 /*
386 * Setup common bits before finally enabling the MMU. Essentially
387 * this is just loading the page table pointer and domain access
388 * registers.
389 *
390 * r0 = cp#15 control register
391 * r1 = machine ID
392 * r2 = atags or dtb pointer
393 * r4 = page table pointer
394 * r9 = processor ID
395 * r13 = *virtual* address to jump to upon completion
396 */
397 __enable_mmu:
398 #if defined(CONFIG_ALIGNMENT_TRAP) && __LINUX_ARM_ARCH__ < 6
399 orr r0, r0, #CR_A
400 #else
401 bic r0, r0, #CR_A
402 #endif
403 #ifdef CONFIG_CPU_DCACHE_DISABLE
404 bic r0, r0, #CR_C
405 #endif
406 #ifdef CONFIG_CPU_BPREDICT_DISABLE
407 bic r0, r0, #CR_Z
408 #endif
409 #ifdef CONFIG_CPU_ICACHE_DISABLE
410 bic r0, r0, #CR_I
411 #endif
412 #ifdef CONFIG_ARM_LPAE
413 mov r5, #0
414 mcrr p15, 0, r4, r5, c2 @ load TTBR0
415 #else
416 mov r5, #(domain_val(DOMAIN_USER, DOMAIN_MANAGER) | \
417 domain_val(DOMAIN_KERNEL, DOMAIN_MANAGER) | \
418 domain_val(DOMAIN_TABLE, DOMAIN_MANAGER) | \
419 domain_val(DOMAIN_IO, DOMAIN_CLIENT))
420 mcr p15, 0, r5, c3, c0, 0 @ load domain access register
421 mcr p15, 0, r4, c2, c0, 0 @ load page table pointer
422 #endif
423 b __turn_mmu_on
424 ENDPROC(__enable_mmu)
426 /*
427 * Enable the MMU. This completely changes the structure of the visible
428 * memory space. You will not be able to trace execution through this.
429 * If you have an enquiry about this, *please* check the linux-arm-kernel
430 * mailing list archives BEFORE sending another post to the list.
431 *
432 * r0 = cp#15 control register
433 * r1 = machine ID
434 * r2 = atags or dtb pointer
435 * r9 = processor ID
436 * r13 = *virtual* address to jump to upon completion
437 *
438 * other registers depend on the function called upon completion
439 */
440 .align 5
441 .pushsection .idmap.text, "ax"
442 ENTRY(__turn_mmu_on)
443 mov r0, r0
444 instr_sync
445 mcr p15, 0, r0, c1, c0, 0 @ write control reg
446 mrc p15, 0, r3, c0, c0, 0 @ read id reg
447 instr_sync
448 mov r3, r3
449 mov r3, r13
450 mov pc, r3
451 __turn_mmu_on_end:
452 ENDPROC(__turn_mmu_on)
453 .popsection
456 #ifdef CONFIG_SMP_ON_UP
457 __INIT
458 __fixup_smp:
459 and r3, r9, #0x000f0000 @ architecture version
460 teq r3, #0x000f0000 @ CPU ID supported?
461 bne __fixup_smp_on_up @ no, assume UP
463 bic r3, r9, #0x00ff0000
464 bic r3, r3, #0x0000000f @ mask 0xff00fff0
465 mov r4, #0x41000000
466 orr r4, r4, #0x0000b000
467 orr r4, r4, #0x00000020 @ val 0x4100b020
468 teq r3, r4 @ ARM 11MPCore?
469 moveq pc, lr @ yes, assume SMP
471 mrc p15, 0, r0, c0, c0, 5 @ read MPIDR
472 and r0, r0, #0xc0000000 @ multiprocessing extensions and
473 teq r0, #0x80000000 @ not part of a uniprocessor system?
474 moveq pc, lr @ yes, assume SMP
476 __fixup_smp_on_up:
477 adr r0, 1f
478 ldmia r0, {r3 - r5}
479 sub r3, r0, r3
480 add r4, r4, r3
481 add r5, r5, r3
482 b __do_fixup_smp_on_up
483 ENDPROC(__fixup_smp)
485 .align
486 1: .word .
487 .word __smpalt_begin
488 .word __smpalt_end
490 .pushsection .data
491 .globl smp_on_up
492 smp_on_up:
493 ALT_SMP(.long 1)
494 ALT_UP(.long 0)
495 .popsection
496 #endif
498 .text
499 __do_fixup_smp_on_up:
500 cmp r4, r5
501 movhs pc, lr
502 ldmia r4!, {r0, r6}
503 ARM( str r6, [r0, r3] )
504 THUMB( add r0, r0, r3 )
505 #ifdef __ARMEB__
506 THUMB( mov r6, r6, ror #16 ) @ Convert word order for big-endian.
507 #endif
508 THUMB( strh r6, [r0], #2 ) @ For Thumb-2, store as two halfwords
509 THUMB( mov r6, r6, lsr #16 ) @ to be robust against misaligned r3.
510 THUMB( strh r6, [r0] )
511 b __do_fixup_smp_on_up
512 ENDPROC(__do_fixup_smp_on_up)
514 ENTRY(fixup_smp)
515 stmfd sp!, {r4 - r6, lr}
516 mov r4, r0
517 add r5, r0, r1
518 mov r3, #0
519 bl __do_fixup_smp_on_up
520 ldmfd sp!, {r4 - r6, pc}
521 ENDPROC(fixup_smp)
523 #ifdef CONFIG_ARM_PATCH_PHYS_VIRT
525 /* __fixup_pv_table - patch the stub instructions with the delta between
526 * PHYS_OFFSET and PAGE_OFFSET, which is assumed to be 16MiB aligned and
527 * can be expressed by an immediate shifter operand. The stub instruction
528 * has a form of '(add|sub) rd, rn, #imm'.
529 */
530 __HEAD
531 __fixup_pv_table:
532 adr r0, 1f
533 ldmia r0, {r3-r5, r7}
534 sub r3, r0, r3 @ PHYS_OFFSET - PAGE_OFFSET
535 add r4, r4, r3 @ adjust table start address
536 add r5, r5, r3 @ adjust table end address
537 add r7, r7, r3 @ adjust __pv_phys_offset address
538 str r8, [r7] @ save computed PHYS_OFFSET to __pv_phys_offset
539 mov r6, r3, lsr #24 @ constant for add/sub instructions
540 teq r3, r6, lsl #24 @ must be 16MiB aligned
541 THUMB( it ne @ cross section branch )
542 bne __error
543 str r6, [r7, #4] @ save to __pv_offset
544 b __fixup_a_pv_table
545 ENDPROC(__fixup_pv_table)
547 .align
548 1: .long .
549 .long __pv_table_begin
550 .long __pv_table_end
551 2: .long __pv_phys_offset
553 .text
554 __fixup_a_pv_table:
555 #ifdef CONFIG_THUMB2_KERNEL
556 lsls r6, #24
557 beq 2f
558 clz r7, r6
559 lsr r6, #24
560 lsl r6, r7
561 bic r6, #0x0080
562 lsrs r7, #1
563 orrcs r6, #0x0080
564 orr r6, r6, r7, lsl #12
565 orr r6, #0x4000
566 b 2f
567 1: add r7, r3
568 ldrh ip, [r7, #2]
569 and ip, 0x8f00
570 orr ip, r6 @ mask in offset bits 31-24
571 strh ip, [r7, #2]
572 2: cmp r4, r5
573 ldrcc r7, [r4], #4 @ use branch for delay slot
574 bcc 1b
575 bx lr
576 #else
577 b 2f
578 1: ldr ip, [r7, r3]
579 bic ip, ip, #0x000000ff
580 orr ip, ip, r6 @ mask in offset bits 31-24
581 str ip, [r7, r3]
582 2: cmp r4, r5
583 ldrcc r7, [r4], #4 @ use branch for delay slot
584 bcc 1b
585 mov pc, lr
586 #endif
587 ENDPROC(__fixup_a_pv_table)
589 ENTRY(fixup_pv_table)
590 stmfd sp!, {r4 - r7, lr}
591 ldr r2, 2f @ get address of __pv_phys_offset
592 mov r3, #0 @ no offset
593 mov r4, r0 @ r0 = table start
594 add r5, r0, r1 @ r1 = table size
595 ldr r6, [r2, #4] @ get __pv_offset
596 bl __fixup_a_pv_table
597 ldmfd sp!, {r4 - r7, pc}
598 ENDPROC(fixup_pv_table)
600 .align
601 2: .long __pv_phys_offset
603 .data
604 .globl __pv_phys_offset
605 .type __pv_phys_offset, %object
606 __pv_phys_offset:
607 .long 0
608 .size __pv_phys_offset, . - __pv_phys_offset
609 __pv_offset:
610 .long 0
611 #endif
613 #include "head-common.S"