blob: 6178be0242f2584e9f732ef53f3cf71062ebe1cb [file] [log] [blame]
Russell Kingd111e8f2006-09-27 15:27:33 +01001/*
2 * linux/arch/arm/mm/mmu.c
3 *
4 * Copyright (C) 1995-2005 Russell King
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
Russell Kingae8f1542006-09-27 15:38:34 +010010#include <linux/module.h>
Russell Kingd111e8f2006-09-27 15:27:33 +010011#include <linux/kernel.h>
12#include <linux/errno.h>
13#include <linux/init.h>
14#include <linux/bootmem.h>
15#include <linux/mman.h>
16#include <linux/nodemask.h>
17
18#include <asm/mach-types.h>
19#include <asm/setup.h>
20#include <asm/sizes.h>
21#include <asm/tlb.h>
22
23#include <asm/mach/arch.h>
24#include <asm/mach/map.h>
25
26#include "mm.h"
27
28DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
29
Russell King6ae5a6e2006-09-30 10:50:05 +010030extern void _stext, _etext, __data_start, _end;
Russell Kingd111e8f2006-09-27 15:27:33 +010031extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
32
33/*
34 * empty_zero_page is a special page that is used for
35 * zero-initialized data and COW.
36 */
37struct page *empty_zero_page;
38
39/*
40 * The pmd table for the upper-most set of pages.
41 */
42pmd_t *top_pmd;
43
Russell Kingae8f1542006-09-27 15:38:34 +010044#define CPOLICY_UNCACHED 0
45#define CPOLICY_BUFFERED 1
46#define CPOLICY_WRITETHROUGH 2
47#define CPOLICY_WRITEBACK 3
48#define CPOLICY_WRITEALLOC 4
49
50static unsigned int cachepolicy __initdata = CPOLICY_WRITEBACK;
51static unsigned int ecc_mask __initdata = 0;
Imre_Deak44b18692007-02-11 13:45:13 +010052pgprot_t pgprot_user;
Russell Kingae8f1542006-09-27 15:38:34 +010053pgprot_t pgprot_kernel;
54
Imre_Deak44b18692007-02-11 13:45:13 +010055EXPORT_SYMBOL(pgprot_user);
Russell Kingae8f1542006-09-27 15:38:34 +010056EXPORT_SYMBOL(pgprot_kernel);
57
58struct cachepolicy {
59 const char policy[16];
60 unsigned int cr_mask;
61 unsigned int pmd;
62 unsigned int pte;
63};
64
65static struct cachepolicy cache_policies[] __initdata = {
66 {
67 .policy = "uncached",
68 .cr_mask = CR_W|CR_C,
69 .pmd = PMD_SECT_UNCACHED,
70 .pte = 0,
71 }, {
72 .policy = "buffered",
73 .cr_mask = CR_C,
74 .pmd = PMD_SECT_BUFFERED,
75 .pte = PTE_BUFFERABLE,
76 }, {
77 .policy = "writethrough",
78 .cr_mask = 0,
79 .pmd = PMD_SECT_WT,
80 .pte = PTE_CACHEABLE,
81 }, {
82 .policy = "writeback",
83 .cr_mask = 0,
84 .pmd = PMD_SECT_WB,
85 .pte = PTE_BUFFERABLE|PTE_CACHEABLE,
86 }, {
87 .policy = "writealloc",
88 .cr_mask = 0,
89 .pmd = PMD_SECT_WBWA,
90 .pte = PTE_BUFFERABLE|PTE_CACHEABLE,
91 }
92};
93
94/*
95 * These are useful for identifing cache coherency
96 * problems by allowing the cache or the cache and
97 * writebuffer to be turned off. (Note: the write
98 * buffer should not be on and the cache off).
99 */
100static void __init early_cachepolicy(char **p)
101{
102 int i;
103
104 for (i = 0; i < ARRAY_SIZE(cache_policies); i++) {
105 int len = strlen(cache_policies[i].policy);
106
107 if (memcmp(*p, cache_policies[i].policy, len) == 0) {
108 cachepolicy = i;
109 cr_alignment &= ~cache_policies[i].cr_mask;
110 cr_no_alignment &= ~cache_policies[i].cr_mask;
111 *p += len;
112 break;
113 }
114 }
115 if (i == ARRAY_SIZE(cache_policies))
116 printk(KERN_ERR "ERROR: unknown or unsupported cache policy\n");
117 flush_cache_all();
118 set_cr(cr_alignment);
119}
120__early_param("cachepolicy=", early_cachepolicy);
121
122static void __init early_nocache(char **__unused)
123{
124 char *p = "buffered";
125 printk(KERN_WARNING "nocache is deprecated; use cachepolicy=%s\n", p);
126 early_cachepolicy(&p);
127}
128__early_param("nocache", early_nocache);
129
130static void __init early_nowrite(char **__unused)
131{
132 char *p = "uncached";
133 printk(KERN_WARNING "nowb is deprecated; use cachepolicy=%s\n", p);
134 early_cachepolicy(&p);
135}
136__early_param("nowb", early_nowrite);
137
138static void __init early_ecc(char **p)
139{
140 if (memcmp(*p, "on", 2) == 0) {
141 ecc_mask = PMD_PROTECTION;
142 *p += 2;
143 } else if (memcmp(*p, "off", 3) == 0) {
144 ecc_mask = 0;
145 *p += 3;
146 }
147}
148__early_param("ecc=", early_ecc);
149
150static int __init noalign_setup(char *__unused)
151{
152 cr_alignment &= ~CR_A;
153 cr_no_alignment &= ~CR_A;
154 set_cr(cr_alignment);
155 return 1;
156}
157__setup("noalign", noalign_setup);
158
Russell King255d1f82006-12-18 00:12:47 +0000159#ifndef CONFIG_SMP
160void adjust_cr(unsigned long mask, unsigned long set)
161{
162 unsigned long flags;
163
164 mask &= ~CR_A;
165
166 set &= mask;
167
168 local_irq_save(flags);
169
170 cr_no_alignment = (cr_no_alignment & ~mask) | set;
171 cr_alignment = (cr_alignment & ~mask) | set;
172
173 set_cr((get_cr() & ~mask) | set);
174
175 local_irq_restore(flags);
176}
177#endif
178
Russell King2497f0a2007-04-21 09:59:44 +0100179struct mem_type {
Russell Kingae8f1542006-09-27 15:38:34 +0100180 unsigned int prot_pte;
181 unsigned int prot_l1;
182 unsigned int prot_sect;
183 unsigned int domain;
184};
185
Russell King2497f0a2007-04-21 09:59:44 +0100186static struct mem_type mem_types[] __initdata = {
Russell Kingae8f1542006-09-27 15:38:34 +0100187 [MT_DEVICE] = {
188 .prot_pte = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
189 L_PTE_WRITE,
190 .prot_l1 = PMD_TYPE_TABLE,
191 .prot_sect = PMD_TYPE_SECT | PMD_BIT4 | PMD_SECT_UNCACHED |
192 PMD_SECT_AP_WRITE,
193 .domain = DOMAIN_IO,
194 },
195 [MT_CACHECLEAN] = {
196 .prot_sect = PMD_TYPE_SECT | PMD_BIT4,
197 .domain = DOMAIN_KERNEL,
198 },
199 [MT_MINICLEAN] = {
200 .prot_sect = PMD_TYPE_SECT | PMD_BIT4 | PMD_SECT_MINICACHE,
201 .domain = DOMAIN_KERNEL,
202 },
203 [MT_LOW_VECTORS] = {
204 .prot_pte = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
205 L_PTE_EXEC,
206 .prot_l1 = PMD_TYPE_TABLE,
207 .domain = DOMAIN_USER,
208 },
209 [MT_HIGH_VECTORS] = {
210 .prot_pte = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
211 L_PTE_USER | L_PTE_EXEC,
212 .prot_l1 = PMD_TYPE_TABLE,
213 .domain = DOMAIN_USER,
214 },
215 [MT_MEMORY] = {
216 .prot_sect = PMD_TYPE_SECT | PMD_BIT4 | PMD_SECT_AP_WRITE,
217 .domain = DOMAIN_KERNEL,
218 },
219 [MT_ROM] = {
220 .prot_sect = PMD_TYPE_SECT | PMD_BIT4,
221 .domain = DOMAIN_KERNEL,
222 },
223 [MT_IXP2000_DEVICE] = { /* IXP2400 requires XCB=101 for on-chip I/O */
224 .prot_pte = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
225 L_PTE_WRITE,
226 .prot_l1 = PMD_TYPE_TABLE,
227 .prot_sect = PMD_TYPE_SECT | PMD_BIT4 | PMD_SECT_UNCACHED |
228 PMD_SECT_AP_WRITE | PMD_SECT_BUFFERABLE |
229 PMD_SECT_TEX(1),
230 .domain = DOMAIN_IO,
231 },
232 [MT_NONSHARED_DEVICE] = {
233 .prot_l1 = PMD_TYPE_TABLE,
234 .prot_sect = PMD_TYPE_SECT | PMD_BIT4 | PMD_SECT_NONSHARED_DEV |
235 PMD_SECT_AP_WRITE,
236 .domain = DOMAIN_IO,
237 }
238};
239
240/*
241 * Adjust the PMD section entries according to the CPU in use.
242 */
243static void __init build_mem_type_table(void)
244{
245 struct cachepolicy *cp;
246 unsigned int cr = get_cr();
247 unsigned int user_pgprot, kern_pgprot;
248 int cpu_arch = cpu_architecture();
249 int i;
250
251#if defined(CONFIG_CPU_DCACHE_DISABLE)
252 if (cachepolicy > CPOLICY_BUFFERED)
253 cachepolicy = CPOLICY_BUFFERED;
254#elif defined(CONFIG_CPU_DCACHE_WRITETHROUGH)
255 if (cachepolicy > CPOLICY_WRITETHROUGH)
256 cachepolicy = CPOLICY_WRITETHROUGH;
257#endif
258 if (cpu_arch < CPU_ARCH_ARMv5) {
259 if (cachepolicy >= CPOLICY_WRITEALLOC)
260 cachepolicy = CPOLICY_WRITEBACK;
261 ecc_mask = 0;
262 }
263
264 /*
265 * Xscale must not have PMD bit 4 set for section mappings.
266 */
267 if (cpu_is_xscale())
268 for (i = 0; i < ARRAY_SIZE(mem_types); i++)
269 mem_types[i].prot_sect &= ~PMD_BIT4;
270
271 /*
272 * ARMv5 and lower, excluding Xscale, bit 4 must be set for
273 * page tables.
274 */
275 if (cpu_arch < CPU_ARCH_ARMv6 && !cpu_is_xscale())
276 for (i = 0; i < ARRAY_SIZE(mem_types); i++)
277 if (mem_types[i].prot_l1)
278 mem_types[i].prot_l1 |= PMD_BIT4;
279
280 cp = &cache_policies[cachepolicy];
281 kern_pgprot = user_pgprot = cp->pte;
282
283 /*
284 * Enable CPU-specific coherency if supported.
285 * (Only available on XSC3 at the moment.)
286 */
287 if (arch_is_coherent()) {
288 if (cpu_is_xsc3()) {
289 mem_types[MT_MEMORY].prot_sect |= PMD_SECT_S;
Lennert Buytenhek0e5fdca72006-12-02 00:03:47 +0100290 mem_types[MT_MEMORY].prot_pte |= L_PTE_SHARED;
Russell Kingae8f1542006-09-27 15:38:34 +0100291 }
292 }
293
294 /*
295 * ARMv6 and above have extended page tables.
296 */
297 if (cpu_arch >= CPU_ARCH_ARMv6 && (cr & CR_XP)) {
298 /*
299 * bit 4 becomes XN which we must clear for the
300 * kernel memory mapping.
301 */
302 mem_types[MT_MEMORY].prot_sect &= ~PMD_SECT_XN;
303 mem_types[MT_ROM].prot_sect &= ~PMD_SECT_XN;
304
305 /*
306 * Mark cache clean areas and XIP ROM read only
307 * from SVC mode and no access from userspace.
308 */
309 mem_types[MT_ROM].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
310 mem_types[MT_MINICLEAN].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
311 mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_APX|PMD_SECT_AP_WRITE;
312
313 /*
314 * Mark the device area as "shared device"
315 */
316 mem_types[MT_DEVICE].prot_pte |= L_PTE_BUFFERABLE;
317 mem_types[MT_DEVICE].prot_sect |= PMD_SECT_BUFFERED;
318
Russell Kingae8f1542006-09-27 15:38:34 +0100319#ifdef CONFIG_SMP
320 /*
321 * Mark memory with the "shared" attribute for SMP systems
322 */
323 user_pgprot |= L_PTE_SHARED;
324 kern_pgprot |= L_PTE_SHARED;
325 mem_types[MT_MEMORY].prot_sect |= PMD_SECT_S;
326#endif
327 }
328
329 for (i = 0; i < 16; i++) {
330 unsigned long v = pgprot_val(protection_map[i]);
331 v = (v & ~(L_PTE_BUFFERABLE|L_PTE_CACHEABLE)) | user_pgprot;
332 protection_map[i] = __pgprot(v);
333 }
334
335 mem_types[MT_LOW_VECTORS].prot_pte |= kern_pgprot;
336 mem_types[MT_HIGH_VECTORS].prot_pte |= kern_pgprot;
337
338 if (cpu_arch >= CPU_ARCH_ARMv5) {
339#ifndef CONFIG_SMP
340 /*
341 * Only use write-through for non-SMP systems
342 */
343 mem_types[MT_LOW_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
344 mem_types[MT_HIGH_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
345#endif
346 } else {
347 mem_types[MT_MINICLEAN].prot_sect &= ~PMD_SECT_TEX(1);
348 }
349
Imre_Deak44b18692007-02-11 13:45:13 +0100350 pgprot_user = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | user_pgprot);
Russell Kingae8f1542006-09-27 15:38:34 +0100351 pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
352 L_PTE_DIRTY | L_PTE_WRITE |
353 L_PTE_EXEC | kern_pgprot);
354
355 mem_types[MT_LOW_VECTORS].prot_l1 |= ecc_mask;
356 mem_types[MT_HIGH_VECTORS].prot_l1 |= ecc_mask;
357 mem_types[MT_MEMORY].prot_sect |= ecc_mask | cp->pmd;
358 mem_types[MT_ROM].prot_sect |= cp->pmd;
359
360 switch (cp->pmd) {
361 case PMD_SECT_WT:
362 mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_WT;
363 break;
364 case PMD_SECT_WB:
365 case PMD_SECT_WBWA:
366 mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_WB;
367 break;
368 }
369 printk("Memory policy: ECC %sabled, Data cache %s\n",
370 ecc_mask ? "en" : "dis", cp->policy);
Russell King2497f0a2007-04-21 09:59:44 +0100371
372 for (i = 0; i < ARRAY_SIZE(mem_types); i++) {
373 struct mem_type *t = &mem_types[i];
374 if (t->prot_l1)
375 t->prot_l1 |= PMD_DOMAIN(t->domain);
376 if (t->prot_sect)
377 t->prot_sect |= PMD_DOMAIN(t->domain);
378 }
Russell Kingae8f1542006-09-27 15:38:34 +0100379}
380
381#define vectors_base() (vectors_high() ? 0xffff0000 : 0)
382
383/*
384 * Create a SECTION PGD between VIRT and PHYS in domain
385 * DOMAIN with protection PROT. This operates on half-
386 * pgdir entry increments.
387 */
388static inline void
389alloc_init_section(unsigned long virt, unsigned long phys, int prot)
390{
391 pmd_t *pmdp = pmd_off_k(virt);
392
393 if (virt & (1 << 20))
394 pmdp++;
395
396 *pmdp = __pmd(phys | prot);
397 flush_pmd_entry(pmdp);
398}
399
400/*
401 * Create a SUPER SECTION PGD between VIRT and PHYS with protection PROT
402 */
403static inline void
404alloc_init_supersection(unsigned long virt, unsigned long phys, int prot)
405{
406 int i;
407
408 for (i = 0; i < 16; i += 1) {
409 alloc_init_section(virt, phys, prot | PMD_SECT_SUPER);
410
411 virt += (PGDIR_SIZE / 2);
412 }
413}
414
415/*
416 * Add a PAGE mapping between VIRT and PHYS in domain
417 * DOMAIN with protection PROT. Note that due to the
418 * way we map the PTEs, we must allocate two PTE_SIZE'd
419 * blocks - one for the Linux pte table, and one for
420 * the hardware pte table.
421 */
422static inline void
423alloc_init_page(unsigned long virt, unsigned long phys, unsigned int prot_l1, pgprot_t prot)
424{
425 pmd_t *pmdp = pmd_off_k(virt);
426 pte_t *ptep;
427
428 if (pmd_none(*pmdp)) {
429 ptep = alloc_bootmem_low_pages(2 * PTRS_PER_PTE *
430 sizeof(pte_t));
431
432 __pmd_populate(pmdp, __pa(ptep) | prot_l1);
433 }
434 ptep = pte_offset_kernel(pmdp, virt);
435
Russell Kingad1ae2f2006-12-13 14:34:43 +0000436 set_pte_ext(ptep, pfn_pte(phys >> PAGE_SHIFT, prot), 0);
Russell Kingae8f1542006-09-27 15:38:34 +0100437}
438
439/*
440 * Create the page directory entries and any necessary
441 * page tables for the mapping specified by `md'. We
442 * are able to cope here with varying sizes and address
443 * offsets, and we take full advantage of sections and
444 * supersections.
445 */
446void __init create_mapping(struct map_desc *md)
447{
448 unsigned long virt, length;
449 int prot_sect, prot_l1, domain;
450 pgprot_t prot_pte;
451 unsigned long off = (u32)__pfn_to_phys(md->pfn);
452
453 if (md->virtual != vectors_base() && md->virtual < TASK_SIZE) {
454 printk(KERN_WARNING "BUG: not creating mapping for "
455 "0x%08llx at 0x%08lx in user region\n",
456 __pfn_to_phys((u64)md->pfn), md->virtual);
457 return;
458 }
459
460 if ((md->type == MT_DEVICE || md->type == MT_ROM) &&
461 md->virtual >= PAGE_OFFSET && md->virtual < VMALLOC_END) {
462 printk(KERN_WARNING "BUG: mapping for 0x%08llx at 0x%08lx "
463 "overlaps vmalloc space\n",
464 __pfn_to_phys((u64)md->pfn), md->virtual);
465 }
466
467 domain = mem_types[md->type].domain;
468 prot_pte = __pgprot(mem_types[md->type].prot_pte);
Russell King2497f0a2007-04-21 09:59:44 +0100469 prot_l1 = mem_types[md->type].prot_l1;
470 prot_sect = mem_types[md->type].prot_sect;
Russell Kingae8f1542006-09-27 15:38:34 +0100471
472 /*
473 * Catch 36-bit addresses
474 */
475 if(md->pfn >= 0x100000) {
476 if(domain) {
477 printk(KERN_ERR "MM: invalid domain in supersection "
478 "mapping for 0x%08llx at 0x%08lx\n",
479 __pfn_to_phys((u64)md->pfn), md->virtual);
480 return;
481 }
482 if((md->virtual | md->length | __pfn_to_phys(md->pfn))
483 & ~SUPERSECTION_MASK) {
484 printk(KERN_ERR "MM: cannot create mapping for "
485 "0x%08llx at 0x%08lx invalid alignment\n",
486 __pfn_to_phys((u64)md->pfn), md->virtual);
487 return;
488 }
489
490 /*
491 * Shift bits [35:32] of address into bits [23:20] of PMD
492 * (See ARMv6 spec).
493 */
494 off |= (((md->pfn >> (32 - PAGE_SHIFT)) & 0xF) << 20);
495 }
496
497 virt = md->virtual;
498 off -= virt;
499 length = md->length;
500
501 if (mem_types[md->type].prot_l1 == 0 &&
502 (virt & 0xfffff || (virt + off) & 0xfffff || (virt + length) & 0xfffff)) {
503 printk(KERN_WARNING "BUG: map for 0x%08lx at 0x%08lx can not "
504 "be mapped using pages, ignoring.\n",
505 __pfn_to_phys(md->pfn), md->virtual);
506 return;
507 }
508
509 while ((virt & 0xfffff || (virt + off) & 0xfffff) && length >= PAGE_SIZE) {
510 alloc_init_page(virt, virt + off, prot_l1, prot_pte);
511
512 virt += PAGE_SIZE;
513 length -= PAGE_SIZE;
514 }
515
516 /* N.B. ARMv6 supersections are only defined to work with domain 0.
517 * Since domain assignments can in fact be arbitrary, the
518 * 'domain == 0' check below is required to insure that ARMv6
519 * supersections are only allocated for domain 0 regardless
520 * of the actual domain assignments in use.
521 */
522 if ((cpu_architecture() >= CPU_ARCH_ARMv6 || cpu_is_xsc3())
523 && domain == 0) {
524 /*
525 * Align to supersection boundary if !high pages.
526 * High pages have already been checked for proper
527 * alignment above and they will fail the SUPSERSECTION_MASK
528 * check because of the way the address is encoded into
529 * offset.
530 */
531 if (md->pfn <= 0x100000) {
532 while ((virt & ~SUPERSECTION_MASK ||
533 (virt + off) & ~SUPERSECTION_MASK) &&
534 length >= (PGDIR_SIZE / 2)) {
535 alloc_init_section(virt, virt + off, prot_sect);
536
537 virt += (PGDIR_SIZE / 2);
538 length -= (PGDIR_SIZE / 2);
539 }
540 }
541
542 while (length >= SUPERSECTION_SIZE) {
543 alloc_init_supersection(virt, virt + off, prot_sect);
544
545 virt += SUPERSECTION_SIZE;
546 length -= SUPERSECTION_SIZE;
547 }
548 }
549
550 /*
551 * A section mapping covers half a "pgdir" entry.
552 */
553 while (length >= (PGDIR_SIZE / 2)) {
554 alloc_init_section(virt, virt + off, prot_sect);
555
556 virt += (PGDIR_SIZE / 2);
557 length -= (PGDIR_SIZE / 2);
558 }
559
560 while (length >= PAGE_SIZE) {
561 alloc_init_page(virt, virt + off, prot_l1, prot_pte);
562
563 virt += PAGE_SIZE;
564 length -= PAGE_SIZE;
565 }
566}
567
568/*
569 * Create the architecture specific mappings
570 */
571void __init iotable_init(struct map_desc *io_desc, int nr)
572{
573 int i;
574
575 for (i = 0; i < nr; i++)
576 create_mapping(io_desc + i);
577}
578
Russell Kingd111e8f2006-09-27 15:27:33 +0100579static inline void prepare_page_table(struct meminfo *mi)
580{
581 unsigned long addr;
582
583 /*
584 * Clear out all the mappings below the kernel image.
585 */
586 for (addr = 0; addr < MODULE_START; addr += PGDIR_SIZE)
587 pmd_clear(pmd_off_k(addr));
588
589#ifdef CONFIG_XIP_KERNEL
590 /* The XIP kernel is mapped in the module area -- skip over it */
591 addr = ((unsigned long)&_etext + PGDIR_SIZE - 1) & PGDIR_MASK;
592#endif
593 for ( ; addr < PAGE_OFFSET; addr += PGDIR_SIZE)
594 pmd_clear(pmd_off_k(addr));
595
596 /*
597 * Clear out all the kernel space mappings, except for the first
598 * memory bank, up to the end of the vmalloc region.
599 */
600 for (addr = __phys_to_virt(mi->bank[0].start + mi->bank[0].size);
601 addr < VMALLOC_END; addr += PGDIR_SIZE)
602 pmd_clear(pmd_off_k(addr));
603}
604
605/*
606 * Reserve the various regions of node 0
607 */
608void __init reserve_node_zero(pg_data_t *pgdat)
609{
610 unsigned long res_size = 0;
611
612 /*
613 * Register the kernel text and data with bootmem.
614 * Note that this can only be in node 0.
615 */
616#ifdef CONFIG_XIP_KERNEL
617 reserve_bootmem_node(pgdat, __pa(&__data_start), &_end - &__data_start);
618#else
619 reserve_bootmem_node(pgdat, __pa(&_stext), &_end - &_stext);
620#endif
621
622 /*
623 * Reserve the page tables. These are already in use,
624 * and can only be in node 0.
625 */
626 reserve_bootmem_node(pgdat, __pa(swapper_pg_dir),
627 PTRS_PER_PGD * sizeof(pgd_t));
628
629 /*
630 * Hmm... This should go elsewhere, but we really really need to
631 * stop things allocating the low memory; ideally we need a better
632 * implementation of GFP_DMA which does not assume that DMA-able
633 * memory starts at zero.
634 */
635 if (machine_is_integrator() || machine_is_cintegrator())
636 res_size = __pa(swapper_pg_dir) - PHYS_OFFSET;
637
638 /*
639 * These should likewise go elsewhere. They pre-reserve the
640 * screen memory region at the start of main system memory.
641 */
642 if (machine_is_edb7211())
643 res_size = 0x00020000;
644 if (machine_is_p720t())
645 res_size = 0x00014000;
646
Ben Dooksbbf6f282006-12-07 20:47:58 +0100647 /* H1940 and RX3715 need to reserve this for suspend */
648
649 if (machine_is_h1940() || machine_is_rx3715()) {
Ben Dooks90733412006-12-06 01:50:24 +0100650 reserve_bootmem_node(pgdat, 0x30003000, 0x1000);
651 reserve_bootmem_node(pgdat, 0x30081000, 0x1000);
652 }
653
Russell Kingd111e8f2006-09-27 15:27:33 +0100654#ifdef CONFIG_SA1111
655 /*
656 * Because of the SA1111 DMA bug, we want to preserve our
657 * precious DMA-able memory...
658 */
659 res_size = __pa(swapper_pg_dir) - PHYS_OFFSET;
660#endif
661 if (res_size)
662 reserve_bootmem_node(pgdat, PHYS_OFFSET, res_size);
663}
664
665/*
666 * Set up device the mappings. Since we clear out the page tables for all
667 * mappings above VMALLOC_END, we will remove any debug device mappings.
668 * This means you have to be careful how you debug this function, or any
669 * called function. This means you can't use any function or debugging
670 * method which may touch any device, otherwise the kernel _will_ crash.
671 */
672static void __init devicemaps_init(struct machine_desc *mdesc)
673{
674 struct map_desc map;
675 unsigned long addr;
676 void *vectors;
677
678 /*
679 * Allocate the vector page early.
680 */
681 vectors = alloc_bootmem_low_pages(PAGE_SIZE);
682 BUG_ON(!vectors);
683
684 for (addr = VMALLOC_END; addr; addr += PGDIR_SIZE)
685 pmd_clear(pmd_off_k(addr));
686
687 /*
688 * Map the kernel if it is XIP.
689 * It is always first in the modulearea.
690 */
691#ifdef CONFIG_XIP_KERNEL
692 map.pfn = __phys_to_pfn(CONFIG_XIP_PHYS_ADDR & SECTION_MASK);
693 map.virtual = MODULE_START;
694 map.length = ((unsigned long)&_etext - map.virtual + ~SECTION_MASK) & SECTION_MASK;
695 map.type = MT_ROM;
696 create_mapping(&map);
697#endif
698
699 /*
700 * Map the cache flushing regions.
701 */
702#ifdef FLUSH_BASE
703 map.pfn = __phys_to_pfn(FLUSH_BASE_PHYS);
704 map.virtual = FLUSH_BASE;
705 map.length = SZ_1M;
706 map.type = MT_CACHECLEAN;
707 create_mapping(&map);
708#endif
709#ifdef FLUSH_BASE_MINICACHE
710 map.pfn = __phys_to_pfn(FLUSH_BASE_PHYS + SZ_1M);
711 map.virtual = FLUSH_BASE_MINICACHE;
712 map.length = SZ_1M;
713 map.type = MT_MINICLEAN;
714 create_mapping(&map);
715#endif
716
717 /*
718 * Create a mapping for the machine vectors at the high-vectors
719 * location (0xffff0000). If we aren't using high-vectors, also
720 * create a mapping at the low-vectors virtual address.
721 */
722 map.pfn = __phys_to_pfn(virt_to_phys(vectors));
723 map.virtual = 0xffff0000;
724 map.length = PAGE_SIZE;
725 map.type = MT_HIGH_VECTORS;
726 create_mapping(&map);
727
728 if (!vectors_high()) {
729 map.virtual = 0;
730 map.type = MT_LOW_VECTORS;
731 create_mapping(&map);
732 }
733
734 /*
735 * Ask the machine support to map in the statically mapped devices.
736 */
737 if (mdesc->map_io)
738 mdesc->map_io();
739
740 /*
741 * Finally flush the caches and tlb to ensure that we're in a
742 * consistent state wrt the writebuffer. This also ensures that
743 * any write-allocated cache lines in the vector page are written
744 * back. After this point, we can start to touch devices again.
745 */
746 local_flush_tlb_all();
747 flush_cache_all();
748}
749
750/*
751 * paging_init() sets up the page tables, initialises the zone memory
752 * maps, and sets up the zero page, bad page and bad page tables.
753 */
754void __init paging_init(struct meminfo *mi, struct machine_desc *mdesc)
755{
756 void *zero_page;
757
758 build_mem_type_table();
759 prepare_page_table(mi);
760 bootmem_init(mi);
761 devicemaps_init(mdesc);
762
763 top_pmd = pmd_off_k(0xffff0000);
764
765 /*
766 * allocate the zero page. Note that we count on this going ok.
767 */
768 zero_page = alloc_bootmem_low_pages(PAGE_SIZE);
769 memzero(zero_page, PAGE_SIZE);
770 empty_zero_page = virt_to_page(zero_page);
771 flush_dcache_page(empty_zero_page);
772}
Russell Kingae8f1542006-09-27 15:38:34 +0100773
774/*
775 * In order to soft-boot, we need to insert a 1:1 mapping in place of
776 * the user-mode pages. This will then ensure that we have predictable
777 * results when turning the mmu off
778 */
779void setup_mm_for_reboot(char mode)
780{
781 unsigned long base_pmdval;
782 pgd_t *pgd;
783 int i;
784
785 if (current->mm && current->mm->pgd)
786 pgd = current->mm->pgd;
787 else
788 pgd = init_mm.pgd;
789
790 base_pmdval = PMD_SECT_AP_WRITE | PMD_SECT_AP_READ | PMD_TYPE_SECT;
791 if (cpu_architecture() <= CPU_ARCH_ARMv5TEJ && !cpu_is_xscale())
792 base_pmdval |= PMD_BIT4;
793
794 for (i = 0; i < FIRST_USER_PGD_NR + USER_PTRS_PER_PGD; i++, pgd++) {
795 unsigned long pmdval = (i << PGDIR_SHIFT) | base_pmdval;
796 pmd_t *pmd;
797
798 pmd = pmd_off(pgd, i << PGDIR_SHIFT);
799 pmd[0] = __pmd(pmdval);
800 pmd[1] = __pmd(pmdval + (1 << (PGDIR_SHIFT - 1)));
801 flush_pmd_entry(pmd);
802 }
803}