mmu.c 27.4 KB
Newer Older
C
Catalin Marinas 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 * Based on arch/arm/mm/mmu.c
 *
 * Copyright (C) 1995-2005 Russell King
 * Copyright (C) 2012 ARM Ltd.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

20
#include <linux/cache.h>
C
Catalin Marinas 已提交
21 22 23 24
#include <linux/export.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/init.h>
25 26
#include <linux/ioport.h>
#include <linux/kexec.h>
27
#include <linux/libfdt.h>
C
Catalin Marinas 已提交
28 29 30 31
#include <linux/mman.h>
#include <linux/nodemask.h>
#include <linux/memblock.h>
#include <linux/fs.h>
32
#include <linux/io.h>
33
#include <linux/mm.h>
34
#include <linux/vmalloc.h>
C
Catalin Marinas 已提交
35

36
#include <asm/barrier.h>
C
Catalin Marinas 已提交
37
#include <asm/cputype.h>
38
#include <asm/fixmap.h>
39
#include <asm/kasan.h>
40
#include <asm/kernel-pgtable.h>
C
Catalin Marinas 已提交
41 42 43 44
#include <asm/sections.h>
#include <asm/setup.h>
#include <asm/sizes.h>
#include <asm/tlb.h>
45
#include <asm/memblock.h>
C
Catalin Marinas 已提交
46
#include <asm/mmu_context.h>
47
#include <asm/ptdump.h>
48
#include <asm/tlbflush.h>
C
Catalin Marinas 已提交
49

50
#define NO_BLOCK_MAPPINGS	BIT(0)
51
#define NO_CONT_MAPPINGS	BIT(1)
52

53
u64 idmap_t0sz = TCR_T0SZ(VA_BITS);
54
u64 idmap_ptrs_per_pgd = PTRS_PER_PGD;
55
u64 vabits_user __ro_after_init;
56
EXPORT_SYMBOL(vabits_user);
57

58
u64 kimage_voffset __ro_after_init;
59 60
EXPORT_SYMBOL(kimage_voffset);

C
Catalin Marinas 已提交
61 62 63 64
/*
 * Empty_zero_page is a special page that is used for zero-initialized data
 * and COW.
 */
65
unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)] __page_aligned_bss;
C
Catalin Marinas 已提交
66 67
EXPORT_SYMBOL(empty_zero_page);

68 69 70 71
static pte_t bm_pte[PTRS_PER_PTE] __page_aligned_bss;
static pmd_t bm_pmd[PTRS_PER_PMD] __page_aligned_bss __maybe_unused;
static pud_t bm_pud[PTRS_PER_PUD] __page_aligned_bss __maybe_unused;

72 73 74 75 76 77 78
static DEFINE_SPINLOCK(swapper_pgdir_lock);

void set_swapper_pgd(pgd_t *pgdp, pgd_t pgd)
{
	pgd_t *fixmap_pgdp;

	spin_lock(&swapper_pgdir_lock);
79
	fixmap_pgdp = pgd_set_fixmap(__pa_symbol(pgdp));
80 81 82 83 84 85 86 87 88 89
	WRITE_ONCE(*fixmap_pgdp, pgd);
	/*
	 * We need dsb(ishst) here to ensure the page-table-walker sees
	 * our new entry before set_p?d() returns. The fixmap's
	 * flush_tlb_kernel_range() via clear_fixmap() does this for us.
	 */
	pgd_clear_fixmap();
	spin_unlock(&swapper_pgdir_lock);
}

C
Catalin Marinas 已提交
90 91 92 93 94 95 96 97 98 99 100
pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
			      unsigned long size, pgprot_t vma_prot)
{
	if (!pfn_valid(pfn))
		return pgprot_noncached(vma_prot);
	else if (file->f_flags & O_SYNC)
		return pgprot_writecombine(vma_prot);
	return vma_prot;
}
EXPORT_SYMBOL(phys_mem_access_prot);

101
static phys_addr_t __init early_pgtable_alloc(void)
C
Catalin Marinas 已提交
102
{
103 104 105
	phys_addr_t phys;
	void *ptr;

106
	phys = memblock_phys_alloc(PAGE_SIZE, PAGE_SIZE);
107 108 109 110 111 112 113 114

	/*
	 * The FIX_{PGD,PUD,PMD} slots may be in active use, but the FIX_PTE
	 * slot will be free, so we can (ab)use the FIX_PTE slot to initialise
	 * any level of table.
	 */
	ptr = pte_set_fixmap(phys);

115 116
	memset(ptr, 0, PAGE_SIZE);

117 118 119 120 121 122 123
	/*
	 * Implicit barriers also ensure the zeroed page is visible to the page
	 * table walker
	 */
	pte_clear_fixmap();

	return phys;
C
Catalin Marinas 已提交
124 125
}

126 127 128 129 130 131
static bool pgattr_change_is_safe(u64 old, u64 new)
{
	/*
	 * The following mapping attributes may be updated in live
	 * kernel mappings without the need for break-before-make.
	 */
132
	static const pteval_t mask = PTE_PXN | PTE_RDONLY | PTE_WRITE | PTE_NG;
133

134 135 136 137 138 139 140 141
	/* creating or taking down mappings is always safe */
	if (old == 0 || new == 0)
		return true;

	/* live contiguous mappings may not be manipulated at all */
	if ((old | new) & PTE_CONT)
		return false;

142 143 144
	/* Transitioning from Non-Global to Global is unsafe */
	if (old & ~new & PTE_NG)
		return false;
145

146
	return ((old ^ new) & ~mask) == 0;
147 148
}

149
static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end,
150
		     phys_addr_t phys, pgprot_t prot)
C
Catalin Marinas 已提交
151
{
152
	pte_t *ptep;
C
Catalin Marinas 已提交
153

154
	ptep = pte_set_fixmap_offset(pmdp, addr);
C
Catalin Marinas 已提交
155
	do {
156
		pte_t old_pte = READ_ONCE(*ptep);
157

158
		set_pte(ptep, pfn_pte(__phys_to_pfn(phys), prot));
159 160 161 162 163

		/*
		 * After the PTE entry has been populated once, we
		 * only allow updates to the permission attributes.
		 */
164 165
		BUG_ON(!pgattr_change_is_safe(pte_val(old_pte),
					      READ_ONCE(pte_val(*ptep))));
166

167
		phys += PAGE_SIZE;
168
	} while (ptep++, addr += PAGE_SIZE, addr != end);
169 170

	pte_clear_fixmap();
C
Catalin Marinas 已提交
171 172
}

173
static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr,
174 175 176 177
				unsigned long end, phys_addr_t phys,
				pgprot_t prot,
				phys_addr_t (*pgtable_alloc)(void),
				int flags)
C
Catalin Marinas 已提交
178 179
{
	unsigned long next;
180
	pmd_t pmd = READ_ONCE(*pmdp);
C
Catalin Marinas 已提交
181

182 183
	BUG_ON(pmd_sect(pmd));
	if (pmd_none(pmd)) {
184
		phys_addr_t pte_phys;
185
		BUG_ON(!pgtable_alloc);
186
		pte_phys = pgtable_alloc();
187 188
		__pmd_populate(pmdp, pte_phys, PMD_TYPE_TABLE);
		pmd = READ_ONCE(*pmdp);
C
Catalin Marinas 已提交
189
	}
190
	BUG_ON(pmd_bad(pmd));
191 192 193 194 195 196 197 198 199 200 201

	do {
		pgprot_t __prot = prot;

		next = pte_cont_addr_end(addr, end);

		/* use a contiguous mapping if the range is suitably aligned */
		if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) &&
		    (flags & NO_CONT_MAPPINGS) == 0)
			__prot = __pgprot(pgprot_val(prot) | PTE_CONT);

202
		init_pte(pmdp, addr, next, phys, __prot);
203 204 205 206 207

		phys += next - addr;
	} while (addr = next, addr != end);
}

208
static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end,
209 210 211 212
		     phys_addr_t phys, pgprot_t prot,
		     phys_addr_t (*pgtable_alloc)(void), int flags)
{
	unsigned long next;
213
	pmd_t *pmdp;
C
Catalin Marinas 已提交
214

215
	pmdp = pmd_set_fixmap_offset(pudp, addr);
C
Catalin Marinas 已提交
216
	do {
217
		pmd_t old_pmd = READ_ONCE(*pmdp);
218

C
Catalin Marinas 已提交
219
		next = pmd_addr_end(addr, end);
220

C
Catalin Marinas 已提交
221
		/* try section mapping first */
222
		if (((addr | next | phys) & ~SECTION_MASK) == 0 &&
223
		    (flags & NO_BLOCK_MAPPINGS) == 0) {
224
			pmd_set_huge(pmdp, phys, prot);
225

226
			/*
227 228
			 * After the PMD entry has been populated once, we
			 * only allow updates to the permission attributes.
229
			 */
230
			BUG_ON(!pgattr_change_is_safe(pmd_val(old_pmd),
231
						      READ_ONCE(pmd_val(*pmdp))));
232
		} else {
233
			alloc_init_cont_pte(pmdp, addr, next, phys, prot,
234
					    pgtable_alloc, flags);
235 236

			BUG_ON(pmd_val(old_pmd) != 0 &&
237
			       pmd_val(old_pmd) != READ_ONCE(pmd_val(*pmdp)));
238
		}
C
Catalin Marinas 已提交
239
		phys += next - addr;
240
	} while (pmdp++, addr = next, addr != end);
241 242

	pmd_clear_fixmap();
C
Catalin Marinas 已提交
243 244
}

245
static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr,
246 247 248 249 250
				unsigned long end, phys_addr_t phys,
				pgprot_t prot,
				phys_addr_t (*pgtable_alloc)(void), int flags)
{
	unsigned long next;
251
	pud_t pud = READ_ONCE(*pudp);
252 253 254 255

	/*
	 * Check for initial section mappings in the pgd/pud.
	 */
256 257
	BUG_ON(pud_sect(pud));
	if (pud_none(pud)) {
258 259 260
		phys_addr_t pmd_phys;
		BUG_ON(!pgtable_alloc);
		pmd_phys = pgtable_alloc();
261 262
		__pud_populate(pudp, pmd_phys, PUD_TYPE_TABLE);
		pud = READ_ONCE(*pudp);
263
	}
264
	BUG_ON(pud_bad(pud));
265 266 267 268 269 270 271 272 273 274 275

	do {
		pgprot_t __prot = prot;

		next = pmd_cont_addr_end(addr, end);

		/* use a contiguous mapping if the range is suitably aligned */
		if ((((addr | next | phys) & ~CONT_PMD_MASK) == 0) &&
		    (flags & NO_CONT_MAPPINGS) == 0)
			__prot = __pgprot(pgprot_val(prot) | PTE_CONT);

276
		init_pmd(pudp, addr, next, phys, __prot, pgtable_alloc, flags);
277 278 279 280 281

		phys += next - addr;
	} while (addr = next, addr != end);
}

282 283 284 285 286 287 288 289 290 291 292 293
static inline bool use_1G_block(unsigned long addr, unsigned long next,
			unsigned long phys)
{
	if (PAGE_SHIFT != 12)
		return false;

	if (((addr | next | phys) & ~PUD_MASK) != 0)
		return false;

	return true;
}

294 295 296 297
static void alloc_init_pud(pgd_t *pgdp, unsigned long addr, unsigned long end,
			   phys_addr_t phys, pgprot_t prot,
			   phys_addr_t (*pgtable_alloc)(void),
			   int flags)
C
Catalin Marinas 已提交
298 299
{
	unsigned long next;
300 301
	pud_t *pudp;
	pgd_t pgd = READ_ONCE(*pgdp);
C
Catalin Marinas 已提交
302

303
	if (pgd_none(pgd)) {
304 305 306
		phys_addr_t pud_phys;
		BUG_ON(!pgtable_alloc);
		pud_phys = pgtable_alloc();
307 308
		__pgd_populate(pgdp, pud_phys, PUD_TYPE_TABLE);
		pgd = READ_ONCE(*pgdp);
309
	}
310
	BUG_ON(pgd_bad(pgd));
311

312
	pudp = pud_set_fixmap_offset(pgdp, addr);
C
Catalin Marinas 已提交
313
	do {
314
		pud_t old_pud = READ_ONCE(*pudp);
315

C
Catalin Marinas 已提交
316
		next = pud_addr_end(addr, end);
317 318 319 320

		/*
		 * For 4K granule only, attempt to put down a 1GB block
		 */
321 322
		if (use_1G_block(addr, next, phys) &&
		    (flags & NO_BLOCK_MAPPINGS) == 0) {
323
			pud_set_huge(pudp, phys, prot);
324 325

			/*
326 327
			 * After the PUD entry has been populated once, we
			 * only allow updates to the permission attributes.
328
			 */
329
			BUG_ON(!pgattr_change_is_safe(pud_val(old_pud),
330
						      READ_ONCE(pud_val(*pudp))));
331
		} else {
332
			alloc_init_cont_pmd(pudp, addr, next, phys, prot,
333
					    pgtable_alloc, flags);
334 335

			BUG_ON(pud_val(old_pud) != 0 &&
336
			       pud_val(old_pud) != READ_ONCE(pud_val(*pudp)));
337
		}
C
Catalin Marinas 已提交
338
		phys += next - addr;
339
	} while (pudp++, addr = next, addr != end);
340 341

	pud_clear_fixmap();
C
Catalin Marinas 已提交
342 343
}

344 345 346 347
static void __create_pgd_mapping(pgd_t *pgdir, phys_addr_t phys,
				 unsigned long virt, phys_addr_t size,
				 pgprot_t prot,
				 phys_addr_t (*pgtable_alloc)(void),
348
				 int flags)
C
Catalin Marinas 已提交
349 350
{
	unsigned long addr, length, end, next;
351
	pgd_t *pgdp = pgd_offset_raw(pgdir, virt);
C
Catalin Marinas 已提交
352

353 354 355 356 357 358 359
	/*
	 * If the virtual and physical address don't have the same offset
	 * within a page, we cannot map the region as the caller expects.
	 */
	if (WARN_ON((phys ^ virt) & ~PAGE_MASK))
		return;

360
	phys &= PAGE_MASK;
C
Catalin Marinas 已提交
361 362 363 364 365 366
	addr = virt & PAGE_MASK;
	length = PAGE_ALIGN(size + (virt & ~PAGE_MASK));

	end = addr + length;
	do {
		next = pgd_addr_end(addr, end);
367
		alloc_init_pud(pgdp, addr, next, phys, prot, pgtable_alloc,
368
			       flags);
C
Catalin Marinas 已提交
369
		phys += next - addr;
370
	} while (pgdp++, addr = next, addr != end);
C
Catalin Marinas 已提交
371 372
}

373
static phys_addr_t pgd_pgtable_alloc(void)
374
{
375
	void *ptr = (void *)__get_free_page(PGALLOC_GFP);
376 377
	if (!ptr || !pgtable_page_ctor(virt_to_page(ptr)))
		BUG();
378 379 380

	/* Ensure the zeroed page is visible to the page table walker */
	dsb(ishst);
381
	return __pa(ptr);
382 383
}

384 385 386 387 388 389
/*
 * This function can only be used to modify existing table entries,
 * without allocating new levels of table. Note that this permits the
 * creation of new section or page entries.
 */
static void __init create_mapping_noalloc(phys_addr_t phys, unsigned long virt,
390
				  phys_addr_t size, pgprot_t prot)
391 392 393 394 395 396
{
	if (virt < VMALLOC_START) {
		pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n",
			&phys, virt);
		return;
	}
397 398
	__create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL,
			     NO_CONT_MAPPINGS);
399 400
}

401 402
void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
			       unsigned long virt, phys_addr_t size,
403
			       pgprot_t prot, bool page_mappings_only)
404
{
405 406
	int flags = 0;

407 408
	BUG_ON(mm == &init_mm);

409
	if (page_mappings_only)
410
		flags = NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS;
411

412
	__create_pgd_mapping(mm->pgd, phys, virt, size, prot,
413
			     pgd_pgtable_alloc, flags);
414 415
}

416 417
static void update_mapping_prot(phys_addr_t phys, unsigned long virt,
				phys_addr_t size, pgprot_t prot)
418 419
{
	if (virt < VMALLOC_START) {
420
		pr_warn("BUG: not updating mapping for %pa at 0x%016lx - outside kernel range\n",
421 422 423 424
			&phys, virt);
		return;
	}

425 426
	__create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL,
			     NO_CONT_MAPPINGS);
427 428 429

	/* flush the TLBs after updating live kernel mappings */
	flush_tlb_kernel_range(virt, virt + size);
430 431
}

432
static void __init __map_memblock(pgd_t *pgdp, phys_addr_t start,
433 434
				  phys_addr_t end, pgprot_t prot, int flags)
{
435
	__create_pgd_mapping(pgdp, start, __phys_to_virt(start), end - start,
436 437 438 439 440 441 442 443 444 445 446 447 448
			     prot, early_pgtable_alloc, flags);
}

void __init mark_linear_text_alias_ro(void)
{
	/*
	 * Remove the write permissions from the linear alias of .text/.rodata
	 */
	update_mapping_prot(__pa_symbol(_text), (unsigned long)lm_alias(_text),
			    (unsigned long)__init_begin - (unsigned long)_text,
			    PAGE_KERNEL_RO);
}

449
static void __init map_mem(pgd_t *pgdp)
450
{
451 452
	phys_addr_t kernel_start = __pa_symbol(_text);
	phys_addr_t kernel_end = __pa_symbol(__init_begin);
453
	struct memblock_region *reg;
454 455
	int flags = 0;

456
	if (rodata_full || debug_pagealloc_enabled())
457
		flags = NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS;
458

459
	/*
460 461
	 * Take care not to create a writable alias for the
	 * read-only text and rodata sections of the kernel image.
462 463
	 * So temporarily mark them as NOMAP to skip mappings in
	 * the following for-loop
464
	 */
465 466 467 468 469 470
	memblock_mark_nomap(kernel_start, kernel_end - kernel_start);
#ifdef CONFIG_KEXEC_CORE
	if (crashk_res.end)
		memblock_mark_nomap(crashk_res.start,
				    resource_size(&crashk_res));
#endif
471

472 473 474 475
	/* map all the memory banks */
	for_each_memblock(memory, reg) {
		phys_addr_t start = reg->base;
		phys_addr_t end = start + reg->size;
476

477 478 479 480 481
		if (start >= end)
			break;
		if (memblock_is_nomap(reg))
			continue;

482
		__map_memblock(pgdp, start, end, PAGE_KERNEL, flags);
483
	}
484 485

	/*
486 487 488 489 490 491
	 * Map the linear alias of the [_text, __init_begin) interval
	 * as non-executable now, and remove the write permission in
	 * mark_linear_text_alias_ro() below (which will be called after
	 * alternative patching has completed). This makes the contents
	 * of the region accessible to subsystems such as hibernate,
	 * but protects it from inadvertent modification or execution.
492 493
	 * Note that contiguous mappings cannot be remapped in this way,
	 * so we should avoid them here.
494
	 */
495
	__map_memblock(pgdp, kernel_start, kernel_end,
496 497
		       PAGE_KERNEL, NO_CONT_MAPPINGS);
	memblock_clear_nomap(kernel_start, kernel_end - kernel_start);
498

499
#ifdef CONFIG_KEXEC_CORE
500
	/*
501 502 503
	 * Use page-level mappings here so that we can shrink the region
	 * in page granularity and put back unused memory to buddy system
	 * through /sys/kernel/kexec_crash_size interface.
504
	 */
505
	if (crashk_res.end) {
506
		__map_memblock(pgdp, crashk_res.start, crashk_res.end + 1,
507 508 509 510
			       PAGE_KERNEL,
			       NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS);
		memblock_clear_nomap(crashk_res.start,
				     resource_size(&crashk_res));
C
Catalin Marinas 已提交
511
	}
512
#endif
C
Catalin Marinas 已提交
513 514
}

515 516
void mark_rodata_ro(void)
{
J
Jeremy Linton 已提交
517
	unsigned long section_size;
518

J
Jeremy Linton 已提交
519
	/*
520 521
	 * mark .rodata as read only. Use __init_begin rather than __end_rodata
	 * to cover NOTES and EXCEPTION_TABLE.
J
Jeremy Linton 已提交
522
	 */
523
	section_size = (unsigned long)__init_begin - (unsigned long)__start_rodata;
524
	update_mapping_prot(__pa_symbol(__start_rodata), (unsigned long)__start_rodata,
J
Jeremy Linton 已提交
525
			    section_size, PAGE_KERNEL_RO);
526

527
	debug_checkwx();
528 529
}

530
static void __init map_kernel_segment(pgd_t *pgdp, void *va_start, void *va_end,
531
				      pgprot_t prot, struct vm_struct *vma,
532
				      int flags, unsigned long vm_flags)
533
{
534
	phys_addr_t pa_start = __pa_symbol(va_start);
535 536 537 538 539
	unsigned long size = va_end - va_start;

	BUG_ON(!PAGE_ALIGNED(pa_start));
	BUG_ON(!PAGE_ALIGNED(size));

540
	__create_pgd_mapping(pgdp, pa_start, (unsigned long)va_start, size, prot,
541
			     early_pgtable_alloc, flags);
542

543 544 545
	if (!(vm_flags & VM_NO_GUARD))
		size += PAGE_SIZE;

546 547 548
	vma->addr	= va_start;
	vma->phys_addr	= pa_start;
	vma->size	= size;
549
	vma->flags	= VM_MAP | vm_flags;
550 551 552
	vma->caller	= __builtin_return_address(0);

	vm_area_add_early(vma);
553 554
}

555 556
static int __init parse_rodata(char *arg)
{
557 558 559 560 561 562 563 564 565 566 567 568 569
	int ret = strtobool(arg, &rodata_enabled);
	if (!ret) {
		rodata_full = false;
		return 0;
	}

	/* permit 'full' in addition to boolean options */
	if (strcmp(arg, "full"))
		return -EINVAL;

	rodata_enabled = true;
	rodata_full = true;
	return 0;
570 571 572
}
early_param("rodata", parse_rodata);

573 574 575 576 577 578 579 580 581 582 583 584 585 586
#ifdef CONFIG_UNMAP_KERNEL_AT_EL0
static int __init map_entry_trampoline(void)
{
	pgprot_t prot = rodata_enabled ? PAGE_KERNEL_ROX : PAGE_KERNEL_EXEC;
	phys_addr_t pa_start = __pa_symbol(__entry_tramp_text_start);

	/* The trampoline is always mapped and can therefore be global */
	pgprot_val(prot) &= ~PTE_NG;

	/* Map only the text into the trampoline page table */
	memset(tramp_pg_dir, 0, PGD_SIZE);
	__create_pgd_mapping(tramp_pg_dir, pa_start, TRAMP_VALIAS, PAGE_SIZE,
			     prot, pgd_pgtable_alloc, 0);

587
	/* Map both the text and data into the kernel page table */
588
	__set_fixmap(FIX_ENTRY_TRAMP_TEXT, pa_start, prot);
589 590 591 592 593 594 595 596
	if (IS_ENABLED(CONFIG_RANDOMIZE_BASE)) {
		extern char __entry_tramp_data_start[];

		__set_fixmap(FIX_ENTRY_TRAMP_DATA,
			     __pa_symbol(__entry_tramp_data_start),
			     PAGE_KERNEL_RO);
	}

597 598 599 600 601
	return 0;
}
core_initcall(map_entry_trampoline);
#endif

602 603 604
/*
 * Create fine-grained mappings for the kernel.
 */
605
static void __init map_kernel(pgd_t *pgdp)
606
{
607 608
	static struct vm_struct vmlinux_text, vmlinux_rodata, vmlinux_inittext,
				vmlinux_initdata, vmlinux_data;
609

610 611 612 613 614 615 616
	/*
	 * External debuggers may need to write directly to the text
	 * mapping to install SW breakpoints. Allow this (only) when
	 * explicitly requested with rodata=off.
	 */
	pgprot_t text_prot = rodata_enabled ? PAGE_KERNEL_ROX : PAGE_KERNEL_EXEC;

617 618 619 620
	/*
	 * Only rodata will be remapped with different permissions later on,
	 * all other segments are allowed to use contiguous mappings.
	 */
621
	map_kernel_segment(pgdp, _text, _etext, text_prot, &vmlinux_text, 0,
622
			   VM_NO_GUARD);
623
	map_kernel_segment(pgdp, __start_rodata, __inittext_begin, PAGE_KERNEL,
624
			   &vmlinux_rodata, NO_CONT_MAPPINGS, VM_NO_GUARD);
625
	map_kernel_segment(pgdp, __inittext_begin, __inittext_end, text_prot,
626
			   &vmlinux_inittext, 0, VM_NO_GUARD);
627
	map_kernel_segment(pgdp, __initdata_begin, __initdata_end, PAGE_KERNEL,
628
			   &vmlinux_initdata, 0, VM_NO_GUARD);
629
	map_kernel_segment(pgdp, _data, _end, PAGE_KERNEL, &vmlinux_data, 0, 0);
630

631
	if (!READ_ONCE(pgd_val(*pgd_offset_raw(pgdp, FIXADDR_START)))) {
632 633 634 635 636
		/*
		 * The fixmap falls in a separate pgd to the kernel, and doesn't
		 * live in the carveout for the swapper_pg_dir. We can simply
		 * re-use the existing dir for the fixmap.
		 */
637 638
		set_pgd(pgd_offset_raw(pgdp, FIXADDR_START),
			READ_ONCE(*pgd_offset_k(FIXADDR_START)));
639 640 641 642 643 644 645 646
	} else if (CONFIG_PGTABLE_LEVELS > 3) {
		/*
		 * The fixmap shares its top level pgd entry with the kernel
		 * mapping. This can really only occur when we are running
		 * with 16k/4 levels, so we can simply reuse the pud level
		 * entry instead.
		 */
		BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES));
647 648
		pud_populate(&init_mm,
			     pud_set_fixmap_offset(pgdp, FIXADDR_START),
649
			     lm_alias(bm_pmd));
650 651 652 653
		pud_clear_fixmap();
	} else {
		BUG();
	}
654

655
	kasan_copy_shadow(pgdp);
656 657
}

C
Catalin Marinas 已提交
658 659 660 661 662 663
/*
 * paging_init() sets up the page tables, initialises the zone memory
 * maps and sets up the zero page.
 */
void __init paging_init(void)
{
664
	pgd_t *pgdp = pgd_set_fixmap(__pa_symbol(swapper_pg_dir));
665

666 667
	map_kernel(pgdp);
	map_mem(pgdp);
668 669 670

	pgd_clear_fixmap();

671
	cpu_replace_ttbr1(lm_alias(swapper_pg_dir));
672
	init_mm.pgd = swapper_pg_dir;
673

674 675
	memblock_free(__pa_symbol(init_pg_dir),
		      __pa_symbol(init_pg_end) - __pa_symbol(init_pg_dir));
676 677

	memblock_allow_resize();
C
Catalin Marinas 已提交
678 679 680 681 682 683 684
}

/*
 * Check whether a kernel address is valid (derived from arch/x86/).
 */
int kern_addr_valid(unsigned long addr)
{
685 686 687 688
	pgd_t *pgdp;
	pud_t *pudp, pud;
	pmd_t *pmdp, pmd;
	pte_t *ptep, pte;
C
Catalin Marinas 已提交
689 690 691 692

	if ((((long)addr) >> VA_BITS) != -1UL)
		return 0;

693 694
	pgdp = pgd_offset_k(addr);
	if (pgd_none(READ_ONCE(*pgdp)))
C
Catalin Marinas 已提交
695 696
		return 0;

697 698 699
	pudp = pud_offset(pgdp, addr);
	pud = READ_ONCE(*pudp);
	if (pud_none(pud))
C
Catalin Marinas 已提交
700 701
		return 0;

702 703
	if (pud_sect(pud))
		return pfn_valid(pud_pfn(pud));
704

705 706 707
	pmdp = pmd_offset(pudp, addr);
	pmd = READ_ONCE(*pmdp);
	if (pmd_none(pmd))
C
Catalin Marinas 已提交
708 709
		return 0;

710 711
	if (pmd_sect(pmd))
		return pfn_valid(pmd_pfn(pmd));
712

713 714 715
	ptep = pte_offset_kernel(pmdp, addr);
	pte = READ_ONCE(*ptep);
	if (pte_none(pte))
C
Catalin Marinas 已提交
716 717
		return 0;

718
	return pfn_valid(pte_pfn(pte));
C
Catalin Marinas 已提交
719 720
}
#ifdef CONFIG_SPARSEMEM_VMEMMAP
721
#if !ARM64_SWAPPER_USES_SECTION_MAPS
722 723
int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,
		struct vmem_altmap *altmap)
C
Catalin Marinas 已提交
724
{
725
	return vmemmap_populate_basepages(start, end, node);
C
Catalin Marinas 已提交
726
}
727
#else	/* !ARM64_SWAPPER_USES_SECTION_MAPS */
728 729
int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,
		struct vmem_altmap *altmap)
C
Catalin Marinas 已提交
730
{
731
	unsigned long addr = start;
C
Catalin Marinas 已提交
732
	unsigned long next;
733 734 735
	pgd_t *pgdp;
	pud_t *pudp;
	pmd_t *pmdp;
C
Catalin Marinas 已提交
736 737 738 739

	do {
		next = pmd_addr_end(addr, end);

740 741
		pgdp = vmemmap_pgd_populate(addr, node);
		if (!pgdp)
C
Catalin Marinas 已提交
742 743
			return -ENOMEM;

744 745
		pudp = vmemmap_pud_populate(pgdp, addr, node);
		if (!pudp)
C
Catalin Marinas 已提交
746 747
			return -ENOMEM;

748 749
		pmdp = pmd_offset(pudp, addr);
		if (pmd_none(READ_ONCE(*pmdp))) {
C
Catalin Marinas 已提交
750 751 752 753 754 755
			void *p = NULL;

			p = vmemmap_alloc_block_buf(PMD_SIZE, node);
			if (!p)
				return -ENOMEM;

756
			pmd_set_huge(pmdp, __pa(p), __pgprot(PROT_SECT_NORMAL));
C
Catalin Marinas 已提交
757
		} else
758
			vmemmap_verify((pte_t *)pmdp, node, addr, next);
C
Catalin Marinas 已提交
759 760 761 762 763
	} while (addr = next, addr != end);

	return 0;
}
#endif	/* CONFIG_ARM64_64K_PAGES */
764 765
void vmemmap_free(unsigned long start, unsigned long end,
		struct vmem_altmap *altmap)
766 767
{
}
C
Catalin Marinas 已提交
768
#endif	/* CONFIG_SPARSEMEM_VMEMMAP */
769 770 771

static inline pud_t * fixmap_pud(unsigned long addr)
{
772 773
	pgd_t *pgdp = pgd_offset_k(addr);
	pgd_t pgd = READ_ONCE(*pgdp);
774

775
	BUG_ON(pgd_none(pgd) || pgd_bad(pgd));
776

777
	return pud_offset_kimg(pgdp, addr);
778 779 780 781
}

static inline pmd_t * fixmap_pmd(unsigned long addr)
{
782 783
	pud_t *pudp = fixmap_pud(addr);
	pud_t pud = READ_ONCE(*pudp);
784

785
	BUG_ON(pud_none(pud) || pud_bad(pud));
786

787
	return pmd_offset_kimg(pudp, addr);
788 789 790 791
}

static inline pte_t * fixmap_pte(unsigned long addr)
{
792
	return &bm_pte[pte_index(addr)];
793 794
}

795 796 797 798 799 800
/*
 * The p*d_populate functions call virt_to_phys implicitly so they can't be used
 * directly on kernel symbols (bm_p*d). This function is called too early to use
 * lm_alias so __p*d_populate functions must be used to populate with the
 * physical address from __pa_symbol.
 */
801 802
void __init early_fixmap_init(void)
{
803 804 805
	pgd_t *pgdp, pgd;
	pud_t *pudp;
	pmd_t *pmdp;
806 807
	unsigned long addr = FIXADDR_START;

808 809
	pgdp = pgd_offset_k(addr);
	pgd = READ_ONCE(*pgdp);
810
	if (CONFIG_PGTABLE_LEVELS > 3 &&
811
	    !(pgd_none(pgd) || pgd_page_paddr(pgd) == __pa_symbol(bm_pud))) {
812 813 814 815 816 817
		/*
		 * We only end up here if the kernel mapping and the fixmap
		 * share the top level pgd entry, which should only happen on
		 * 16k/4 levels configurations.
		 */
		BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES));
818
		pudp = pud_offset_kimg(pgdp, addr);
819
	} else {
820 821 822
		if (pgd_none(pgd))
			__pgd_populate(pgdp, __pa_symbol(bm_pud), PUD_TYPE_TABLE);
		pudp = fixmap_pud(addr);
823
	}
824 825 826 827
	if (pud_none(READ_ONCE(*pudp)))
		__pud_populate(pudp, __pa_symbol(bm_pmd), PMD_TYPE_TABLE);
	pmdp = fixmap_pmd(addr);
	__pmd_populate(pmdp, __pa_symbol(bm_pte), PMD_TYPE_TABLE);
828 829 830

	/*
	 * The boot-ioremap range spans multiple pmds, for which
831
	 * we are not prepared:
832 833 834 835
	 */
	BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT)
		     != (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT));

836 837
	if ((pmdp != fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)))
	     || pmdp != fixmap_pmd(fix_to_virt(FIX_BTMAP_END))) {
838
		WARN_ON(1);
839 840
		pr_warn("pmdp %p != %p, %p\n",
			pmdp, fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)),
841 842 843 844 845 846 847 848 849 850 851
			fixmap_pmd(fix_to_virt(FIX_BTMAP_END)));
		pr_warn("fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n",
			fix_to_virt(FIX_BTMAP_BEGIN));
		pr_warn("fix_to_virt(FIX_BTMAP_END):   %08lx\n",
			fix_to_virt(FIX_BTMAP_END));

		pr_warn("FIX_BTMAP_END:       %d\n", FIX_BTMAP_END);
		pr_warn("FIX_BTMAP_BEGIN:     %d\n", FIX_BTMAP_BEGIN);
	}
}

852 853 854 855
/*
 * Unusually, this is also called in IRQ context (ghes_iounmap_irq) so if we
 * ever need to use IPIs for TLB broadcasting, then we're in trouble here.
 */
856 857 858 859
void __set_fixmap(enum fixed_addresses idx,
			       phys_addr_t phys, pgprot_t flags)
{
	unsigned long addr = __fix_to_virt(idx);
860
	pte_t *ptep;
861

862
	BUG_ON(idx <= FIX_HOLE || idx >= __end_of_fixed_addresses);
863

864
	ptep = fixmap_pte(addr);
865 866

	if (pgprot_val(flags)) {
867
		set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, flags));
868
	} else {
869
		pte_clear(&init_mm, addr, ptep);
870 871 872
		flush_tlb_kernel_range(addr, addr+PAGE_SIZE);
	}
}
873

874
void *__init __fixmap_remap_fdt(phys_addr_t dt_phys, int *size, pgprot_t prot)
875 876
{
	const u64 dt_virt_base = __fix_to_virt(FIX_FDT);
877
	int offset;
878 879 880 881 882
	void *dt_virt;

	/*
	 * Check whether the physical FDT address is set and meets the minimum
	 * alignment requirement. Since we are relying on MIN_FDT_ALIGN to be
883 884 885
	 * at least 8 bytes so that we can always access the magic and size
	 * fields of the FDT header after mapping the first chunk, double check
	 * here if that is indeed the case.
886 887 888 889 890 891 892 893
	 */
	BUILD_BUG_ON(MIN_FDT_ALIGN < 8);
	if (!dt_phys || dt_phys % MIN_FDT_ALIGN)
		return NULL;

	/*
	 * Make sure that the FDT region can be mapped without the need to
	 * allocate additional translation table pages, so that it is safe
894
	 * to call create_mapping_noalloc() this early.
895 896 897 898 899 900 901 902
	 *
	 * On 64k pages, the FDT will be mapped using PTEs, so we need to
	 * be in the same PMD as the rest of the fixmap.
	 * On 4k pages, we'll use section mappings for the FDT so we only
	 * have to be in the same PUD.
	 */
	BUILD_BUG_ON(dt_virt_base % SZ_2M);

903 904
	BUILD_BUG_ON(__fix_to_virt(FIX_FDT_END) >> SWAPPER_TABLE_SHIFT !=
		     __fix_to_virt(FIX_BTMAP_BEGIN) >> SWAPPER_TABLE_SHIFT);
905

906
	offset = dt_phys % SWAPPER_BLOCK_SIZE;
907 908 909
	dt_virt = (void *)dt_virt_base + offset;

	/* map the first chunk so we can read the size from the header */
910 911
	create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE),
			dt_virt_base, SWAPPER_BLOCK_SIZE, prot);
912

913
	if (fdt_magic(dt_virt) != FDT_MAGIC)
914 915
		return NULL;

916 917
	*size = fdt_totalsize(dt_virt);
	if (*size > MAX_FDT_SIZE)
918 919
		return NULL;

920
	if (offset + *size > SWAPPER_BLOCK_SIZE)
921
		create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), dt_virt_base,
922
			       round_up(offset + *size, SWAPPER_BLOCK_SIZE), prot);
923

924 925
	return dt_virt;
}
926

927 928 929 930 931 932 933 934 935 936
void *__init fixmap_remap_fdt(phys_addr_t dt_phys)
{
	void *dt_virt;
	int size;

	dt_virt = __fixmap_remap_fdt(dt_phys, &size, PAGE_KERNEL_RO);
	if (!dt_virt)
		return NULL;

	memblock_reserve(dt_phys, size);
937 938
	return dt_virt;
}
939 940 941 942 943 944 945 946 947 948 949 950

int __init arch_ioremap_pud_supported(void)
{
	/* only 4k granule supports level 1 block mappings */
	return IS_ENABLED(CONFIG_ARM64_4K_PAGES);
}

int __init arch_ioremap_pmd_supported(void)
{
	return 1;
}

951
int pud_set_huge(pud_t *pudp, phys_addr_t phys, pgprot_t prot)
952
{
953 954
	pgprot_t sect_prot = __pgprot(PUD_TYPE_SECT |
					pgprot_val(mk_sect_prot(prot)));
955
	pud_t new_pud = pfn_pud(__phys_to_pfn(phys), sect_prot);
956

957 958 959
	/* Only allow permission changes for now */
	if (!pgattr_change_is_safe(READ_ONCE(pud_val(*pudp)),
				   pud_val(new_pud)))
960 961
		return 0;

962
	BUG_ON(phys & ~PUD_MASK);
963
	set_pud(pudp, new_pud);
964 965 966
	return 1;
}

967
int pmd_set_huge(pmd_t *pmdp, phys_addr_t phys, pgprot_t prot)
968
{
969 970
	pgprot_t sect_prot = __pgprot(PMD_TYPE_SECT |
					pgprot_val(mk_sect_prot(prot)));
971
	pmd_t new_pmd = pfn_pmd(__phys_to_pfn(phys), sect_prot);
972

973 974 975
	/* Only allow permission changes for now */
	if (!pgattr_change_is_safe(READ_ONCE(pmd_val(*pmdp)),
				   pmd_val(new_pmd)))
976 977
		return 0;

978
	BUG_ON(phys & ~PMD_MASK);
979
	set_pmd(pmdp, new_pmd);
980 981 982
	return 1;
}

983
int pud_clear_huge(pud_t *pudp)
984
{
985
	if (!pud_sect(READ_ONCE(*pudp)))
986
		return 0;
987
	pud_clear(pudp);
988 989 990
	return 1;
}

991
int pmd_clear_huge(pmd_t *pmdp)
992
{
993
	if (!pmd_sect(READ_ONCE(*pmdp)))
994
		return 0;
995
	pmd_clear(pmdp);
996 997
	return 1;
}
998

999
int pmd_free_pte_page(pmd_t *pmdp, unsigned long addr)
1000
{
1001 1002 1003 1004 1005
	pte_t *table;
	pmd_t pmd;

	pmd = READ_ONCE(*pmdp);

1006 1007 1008
	if (!pmd_present(pmd))
		return 1;
	if (!pmd_table(pmd)) {
1009 1010 1011 1012 1013 1014 1015 1016 1017
		VM_WARN_ON(!pmd_table(pmd));
		return 1;
	}

	table = pte_offset_kernel(pmdp, addr);
	pmd_clear(pmdp);
	__flush_tlb_kernel_pgtable(addr);
	pte_free_kernel(NULL, table);
	return 1;
1018 1019
}

1020
int pud_free_pmd_page(pud_t *pudp, unsigned long addr)
1021
{
1022 1023 1024 1025 1026 1027 1028
	pmd_t *table;
	pmd_t *pmdp;
	pud_t pud;
	unsigned long next, end;

	pud = READ_ONCE(*pudp);

1029 1030 1031
	if (!pud_present(pud))
		return 1;
	if (!pud_table(pud)) {
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
		VM_WARN_ON(!pud_table(pud));
		return 1;
	}

	table = pmd_offset(pudp, addr);
	pmdp = table;
	next = addr;
	end = addr + PUD_SIZE;
	do {
		pmd_free_pte_page(pmdp, next);
	} while (pmdp++, next += PMD_SIZE, next != end);

	pud_clear(pudp);
	__flush_tlb_kernel_pgtable(addr);
	pmd_free(NULL, table);
	return 1;
1048
}
R
Robin Murphy 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065

#ifdef CONFIG_MEMORY_HOTPLUG
int arch_add_memory(int nid, u64 start, u64 size, struct vmem_altmap *altmap,
		    bool want_memblock)
{
	int flags = 0;

	if (rodata_full || debug_pagealloc_enabled())
		flags = NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS;

	__create_pgd_mapping(swapper_pg_dir, start, __phys_to_virt(start),
			     size, PAGE_KERNEL, pgd_pgtable_alloc, flags);

	return __add_pages(nid, start >> PAGE_SHIFT, size >> PAGE_SHIFT,
			   altmap, want_memblock);
}
#endif