mmu.c 25.7 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>
C
Catalin Marinas 已提交
48

49
#define NO_BLOCK_MAPPINGS	BIT(0)
50
#define NO_CONT_MAPPINGS	BIT(1)
51

52
u64 idmap_t0sz = TCR_T0SZ(VA_BITS);
53
u64 idmap_ptrs_per_pgd = PTRS_PER_PGD;
54

55
u64 kimage_voffset __ro_after_init;
56 57
EXPORT_SYMBOL(kimage_voffset);

C
Catalin Marinas 已提交
58 59 60 61
/*
 * Empty_zero_page is a special page that is used for zero-initialized data
 * and COW.
 */
62
unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)] __page_aligned_bss;
C
Catalin Marinas 已提交
63 64
EXPORT_SYMBOL(empty_zero_page);

65 66 67 68
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;

C
Catalin Marinas 已提交
69 70 71 72 73 74 75 76 77 78 79
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);

80
static phys_addr_t __init early_pgtable_alloc(void)
C
Catalin Marinas 已提交
81
{
82 83 84
	phys_addr_t phys;
	void *ptr;

85
	phys = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
86 87 88 89 90 91 92 93

	/*
	 * 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);

94 95
	memset(ptr, 0, PAGE_SIZE);

96 97 98 99 100 101 102
	/*
	 * 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
103 104
}

105 106 107 108 109 110 111 112
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.
	 */
	static const pteval_t mask = PTE_PXN | PTE_RDONLY | PTE_WRITE;

113 114 115 116 117 118 119 120
	/* 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;

121 122 123 124
	/* Transitioning from Global to Non-Global is safe */
	if (((old ^ new) == PTE_NG) && (new & PTE_NG))
		return true;

125
	return ((old ^ new) & ~mask) == 0;
126 127
}

128
static void init_pte(pmd_t *pmdp, unsigned long addr, unsigned long end,
129
		     phys_addr_t phys, pgprot_t prot)
C
Catalin Marinas 已提交
130
{
131
	pte_t *ptep;
C
Catalin Marinas 已提交
132

133
	ptep = pte_set_fixmap_offset(pmdp, addr);
C
Catalin Marinas 已提交
134
	do {
135
		pte_t old_pte = READ_ONCE(*ptep);
136

137
		set_pte(ptep, pfn_pte(__phys_to_pfn(phys), prot));
138 139 140 141 142

		/*
		 * After the PTE entry has been populated once, we
		 * only allow updates to the permission attributes.
		 */
143 144
		BUG_ON(!pgattr_change_is_safe(pte_val(old_pte),
					      READ_ONCE(pte_val(*ptep))));
145

146
		phys += PAGE_SIZE;
147
	} while (ptep++, addr += PAGE_SIZE, addr != end);
148 149

	pte_clear_fixmap();
C
Catalin Marinas 已提交
150 151
}

152
static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr,
153 154 155 156
				unsigned long end, phys_addr_t phys,
				pgprot_t prot,
				phys_addr_t (*pgtable_alloc)(void),
				int flags)
C
Catalin Marinas 已提交
157 158
{
	unsigned long next;
159
	pmd_t pmd = READ_ONCE(*pmdp);
C
Catalin Marinas 已提交
160

161 162
	BUG_ON(pmd_sect(pmd));
	if (pmd_none(pmd)) {
163
		phys_addr_t pte_phys;
164
		BUG_ON(!pgtable_alloc);
165
		pte_phys = pgtable_alloc();
166 167
		__pmd_populate(pmdp, pte_phys, PMD_TYPE_TABLE);
		pmd = READ_ONCE(*pmdp);
C
Catalin Marinas 已提交
168
	}
169
	BUG_ON(pmd_bad(pmd));
170 171 172 173 174 175 176 177 178 179 180

	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);

181
		init_pte(pmdp, addr, next, phys, __prot);
182 183 184 185 186

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

187
static void init_pmd(pud_t *pudp, unsigned long addr, unsigned long end,
188 189 190 191
		     phys_addr_t phys, pgprot_t prot,
		     phys_addr_t (*pgtable_alloc)(void), int flags)
{
	unsigned long next;
192
	pmd_t *pmdp;
C
Catalin Marinas 已提交
193

194
	pmdp = pmd_set_fixmap_offset(pudp, addr);
C
Catalin Marinas 已提交
195
	do {
196
		pmd_t old_pmd = READ_ONCE(*pmdp);
197

C
Catalin Marinas 已提交
198
		next = pmd_addr_end(addr, end);
199

C
Catalin Marinas 已提交
200
		/* try section mapping first */
201
		if (((addr | next | phys) & ~SECTION_MASK) == 0 &&
202
		    (flags & NO_BLOCK_MAPPINGS) == 0) {
203
			pmd_set_huge(pmdp, phys, prot);
204

205
			/*
206 207
			 * After the PMD entry has been populated once, we
			 * only allow updates to the permission attributes.
208
			 */
209
			BUG_ON(!pgattr_change_is_safe(pmd_val(old_pmd),
210
						      READ_ONCE(pmd_val(*pmdp))));
211
		} else {
212
			alloc_init_cont_pte(pmdp, addr, next, phys, prot,
213
					    pgtable_alloc, flags);
214 215

			BUG_ON(pmd_val(old_pmd) != 0 &&
216
			       pmd_val(old_pmd) != READ_ONCE(pmd_val(*pmdp)));
217
		}
C
Catalin Marinas 已提交
218
		phys += next - addr;
219
	} while (pmdp++, addr = next, addr != end);
220 221

	pmd_clear_fixmap();
C
Catalin Marinas 已提交
222 223
}

224
static void alloc_init_cont_pmd(pud_t *pudp, unsigned long addr,
225 226 227 228 229
				unsigned long end, phys_addr_t phys,
				pgprot_t prot,
				phys_addr_t (*pgtable_alloc)(void), int flags)
{
	unsigned long next;
230
	pud_t pud = READ_ONCE(*pudp);
231 232 233 234

	/*
	 * Check for initial section mappings in the pgd/pud.
	 */
235 236
	BUG_ON(pud_sect(pud));
	if (pud_none(pud)) {
237 238 239
		phys_addr_t pmd_phys;
		BUG_ON(!pgtable_alloc);
		pmd_phys = pgtable_alloc();
240 241
		__pud_populate(pudp, pmd_phys, PUD_TYPE_TABLE);
		pud = READ_ONCE(*pudp);
242
	}
243
	BUG_ON(pud_bad(pud));
244 245 246 247 248 249 250 251 252 253 254

	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);

255
		init_pmd(pudp, addr, next, phys, __prot, pgtable_alloc, flags);
256 257 258 259 260

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

261 262 263 264 265 266 267 268 269 270 271 272
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;
}

273 274 275 276
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
277 278
{
	unsigned long next;
279 280
	pud_t *pudp;
	pgd_t pgd = READ_ONCE(*pgdp);
C
Catalin Marinas 已提交
281

282
	if (pgd_none(pgd)) {
283 284 285
		phys_addr_t pud_phys;
		BUG_ON(!pgtable_alloc);
		pud_phys = pgtable_alloc();
286 287
		__pgd_populate(pgdp, pud_phys, PUD_TYPE_TABLE);
		pgd = READ_ONCE(*pgdp);
288
	}
289
	BUG_ON(pgd_bad(pgd));
290

291
	pudp = pud_set_fixmap_offset(pgdp, addr);
C
Catalin Marinas 已提交
292
	do {
293
		pud_t old_pud = READ_ONCE(*pudp);
294

C
Catalin Marinas 已提交
295
		next = pud_addr_end(addr, end);
296 297 298 299

		/*
		 * For 4K granule only, attempt to put down a 1GB block
		 */
300 301
		if (use_1G_block(addr, next, phys) &&
		    (flags & NO_BLOCK_MAPPINGS) == 0) {
302
			pud_set_huge(pudp, phys, prot);
303 304

			/*
305 306
			 * After the PUD entry has been populated once, we
			 * only allow updates to the permission attributes.
307
			 */
308
			BUG_ON(!pgattr_change_is_safe(pud_val(old_pud),
309
						      READ_ONCE(pud_val(*pudp))));
310
		} else {
311
			alloc_init_cont_pmd(pudp, addr, next, phys, prot,
312
					    pgtable_alloc, flags);
313 314

			BUG_ON(pud_val(old_pud) != 0 &&
315
			       pud_val(old_pud) != READ_ONCE(pud_val(*pudp)));
316
		}
C
Catalin Marinas 已提交
317
		phys += next - addr;
318
	} while (pudp++, addr = next, addr != end);
319 320

	pud_clear_fixmap();
C
Catalin Marinas 已提交
321 322
}

323 324 325 326
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),
327
				 int flags)
C
Catalin Marinas 已提交
328 329
{
	unsigned long addr, length, end, next;
330
	pgd_t *pgdp = pgd_offset_raw(pgdir, virt);
C
Catalin Marinas 已提交
331

332 333 334 335 336 337 338
	/*
	 * 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;

339
	phys &= PAGE_MASK;
C
Catalin Marinas 已提交
340 341 342 343 344 345
	addr = virt & PAGE_MASK;
	length = PAGE_ALIGN(size + (virt & ~PAGE_MASK));

	end = addr + length;
	do {
		next = pgd_addr_end(addr, end);
346
		alloc_init_pud(pgdp, addr, next, phys, prot, pgtable_alloc,
347
			       flags);
C
Catalin Marinas 已提交
348
		phys += next - addr;
349
	} while (pgdp++, addr = next, addr != end);
C
Catalin Marinas 已提交
350 351
}

352
static phys_addr_t pgd_pgtable_alloc(void)
353
{
354
	void *ptr = (void *)__get_free_page(PGALLOC_GFP);
355 356
	if (!ptr || !pgtable_page_ctor(virt_to_page(ptr)))
		BUG();
357 358 359

	/* Ensure the zeroed page is visible to the page table walker */
	dsb(ishst);
360
	return __pa(ptr);
361 362
}

363 364 365 366 367 368
/*
 * 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,
369
				  phys_addr_t size, pgprot_t prot)
370 371 372 373 374 375
{
	if (virt < VMALLOC_START) {
		pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n",
			&phys, virt);
		return;
	}
376 377
	__create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL,
			     NO_CONT_MAPPINGS);
378 379
}

380 381
void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
			       unsigned long virt, phys_addr_t size,
382
			       pgprot_t prot, bool page_mappings_only)
383
{
384 385
	int flags = 0;

386 387
	BUG_ON(mm == &init_mm);

388
	if (page_mappings_only)
389
		flags = NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS;
390

391
	__create_pgd_mapping(mm->pgd, phys, virt, size, prot,
392
			     pgd_pgtable_alloc, flags);
393 394
}

395 396
static void update_mapping_prot(phys_addr_t phys, unsigned long virt,
				phys_addr_t size, pgprot_t prot)
397 398
{
	if (virt < VMALLOC_START) {
399
		pr_warn("BUG: not updating mapping for %pa at 0x%016lx - outside kernel range\n",
400 401 402 403
			&phys, virt);
		return;
	}

404 405
	__create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL,
			     NO_CONT_MAPPINGS);
406 407 408

	/* flush the TLBs after updating live kernel mappings */
	flush_tlb_kernel_range(virt, virt + size);
409 410
}

411
static void __init __map_memblock(pgd_t *pgdp, phys_addr_t start,
412 413
				  phys_addr_t end, pgprot_t prot, int flags)
{
414
	__create_pgd_mapping(pgdp, start, __phys_to_virt(start), end - start,
415 416 417 418 419 420 421 422 423 424 425 426 427
			     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);
}

428
static void __init map_mem(pgd_t *pgdp)
429
{
430 431
	phys_addr_t kernel_start = __pa_symbol(_text);
	phys_addr_t kernel_end = __pa_symbol(__init_begin);
432
	struct memblock_region *reg;
433 434 435
	int flags = 0;

	if (debug_pagealloc_enabled())
436
		flags = NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS;
437

438
	/*
439 440
	 * Take care not to create a writable alias for the
	 * read-only text and rodata sections of the kernel image.
441 442
	 * So temporarily mark them as NOMAP to skip mappings in
	 * the following for-loop
443
	 */
444 445 446 447 448 449
	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
450

451 452 453 454
	/* map all the memory banks */
	for_each_memblock(memory, reg) {
		phys_addr_t start = reg->base;
		phys_addr_t end = start + reg->size;
455

456 457 458 459 460
		if (start >= end)
			break;
		if (memblock_is_nomap(reg))
			continue;

461
		__map_memblock(pgdp, start, end, PAGE_KERNEL, flags);
462
	}
463 464

	/*
465 466 467 468 469 470
	 * 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.
471 472
	 * Note that contiguous mappings cannot be remapped in this way,
	 * so we should avoid them here.
473
	 */
474
	__map_memblock(pgdp, kernel_start, kernel_end,
475 476
		       PAGE_KERNEL, NO_CONT_MAPPINGS);
	memblock_clear_nomap(kernel_start, kernel_end - kernel_start);
477

478
#ifdef CONFIG_KEXEC_CORE
479
	/*
480 481 482
	 * 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.
483
	 */
484
	if (crashk_res.end) {
485
		__map_memblock(pgdp, crashk_res.start, crashk_res.end + 1,
486 487 488 489
			       PAGE_KERNEL,
			       NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS);
		memblock_clear_nomap(crashk_res.start,
				     resource_size(&crashk_res));
C
Catalin Marinas 已提交
490
	}
491
#endif
C
Catalin Marinas 已提交
492 493
}

494 495
void mark_rodata_ro(void)
{
J
Jeremy Linton 已提交
496
	unsigned long section_size;
497

J
Jeremy Linton 已提交
498
	/*
499 500
	 * mark .rodata as read only. Use __init_begin rather than __end_rodata
	 * to cover NOTES and EXCEPTION_TABLE.
J
Jeremy Linton 已提交
501
	 */
502
	section_size = (unsigned long)__init_begin - (unsigned long)__start_rodata;
503
	update_mapping_prot(__pa_symbol(__start_rodata), (unsigned long)__start_rodata,
J
Jeremy Linton 已提交
504
			    section_size, PAGE_KERNEL_RO);
505

506
	debug_checkwx();
507 508
}

509
static void __init map_kernel_segment(pgd_t *pgdp, void *va_start, void *va_end,
510
				      pgprot_t prot, struct vm_struct *vma,
511
				      int flags, unsigned long vm_flags)
512
{
513
	phys_addr_t pa_start = __pa_symbol(va_start);
514 515 516 517 518
	unsigned long size = va_end - va_start;

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

519
	__create_pgd_mapping(pgdp, pa_start, (unsigned long)va_start, size, prot,
520
			     early_pgtable_alloc, flags);
521

522 523 524
	if (!(vm_flags & VM_NO_GUARD))
		size += PAGE_SIZE;

525 526 527
	vma->addr	= va_start;
	vma->phys_addr	= pa_start;
	vma->size	= size;
528
	vma->flags	= VM_MAP | vm_flags;
529 530 531
	vma->caller	= __builtin_return_address(0);

	vm_area_add_early(vma);
532 533
}

534 535 536 537 538 539
static int __init parse_rodata(char *arg)
{
	return strtobool(arg, &rodata_enabled);
}
early_param("rodata", parse_rodata);

540 541 542 543 544 545 546 547 548 549 550 551 552 553
#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);

554
	/* Map both the text and data into the kernel page table */
555
	__set_fixmap(FIX_ENTRY_TRAMP_TEXT, pa_start, prot);
556 557 558 559 560 561 562 563
	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);
	}

564 565 566 567 568
	return 0;
}
core_initcall(map_entry_trampoline);
#endif

569 570 571
/*
 * Create fine-grained mappings for the kernel.
 */
572
static void __init map_kernel(pgd_t *pgdp)
573
{
574 575
	static struct vm_struct vmlinux_text, vmlinux_rodata, vmlinux_inittext,
				vmlinux_initdata, vmlinux_data;
576

577 578 579 580 581 582 583
	/*
	 * 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;

584 585 586 587
	/*
	 * Only rodata will be remapped with different permissions later on,
	 * all other segments are allowed to use contiguous mappings.
	 */
588
	map_kernel_segment(pgdp, _text, _etext, text_prot, &vmlinux_text, 0,
589
			   VM_NO_GUARD);
590
	map_kernel_segment(pgdp, __start_rodata, __inittext_begin, PAGE_KERNEL,
591
			   &vmlinux_rodata, NO_CONT_MAPPINGS, VM_NO_GUARD);
592
	map_kernel_segment(pgdp, __inittext_begin, __inittext_end, text_prot,
593
			   &vmlinux_inittext, 0, VM_NO_GUARD);
594
	map_kernel_segment(pgdp, __initdata_begin, __initdata_end, PAGE_KERNEL,
595
			   &vmlinux_initdata, 0, VM_NO_GUARD);
596
	map_kernel_segment(pgdp, _data, _end, PAGE_KERNEL, &vmlinux_data, 0, 0);
597

598
	if (!READ_ONCE(pgd_val(*pgd_offset_raw(pgdp, FIXADDR_START)))) {
599 600 601 602 603
		/*
		 * 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.
		 */
604 605
		set_pgd(pgd_offset_raw(pgdp, FIXADDR_START),
			READ_ONCE(*pgd_offset_k(FIXADDR_START)));
606 607 608 609 610 611 612 613
	} 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));
614 615
		pud_populate(&init_mm,
			     pud_set_fixmap_offset(pgdp, FIXADDR_START),
616
			     lm_alias(bm_pmd));
617 618 619 620
		pud_clear_fixmap();
	} else {
		BUG();
	}
621

622
	kasan_copy_shadow(pgdp);
623 624
}

C
Catalin Marinas 已提交
625 626 627 628 629 630
/*
 * paging_init() sets up the page tables, initialises the zone memory
 * maps and sets up the zero page.
 */
void __init paging_init(void)
{
631
	phys_addr_t pgd_phys = early_pgtable_alloc();
632
	pgd_t *pgdp = pgd_set_fixmap(pgd_phys);
633

634 635
	map_kernel(pgdp);
	map_mem(pgdp);
636 637 638 639 640 641 642 643 644 645

	/*
	 * We want to reuse the original swapper_pg_dir so we don't have to
	 * communicate the new address to non-coherent secondaries in
	 * secondary_entry, and so cpu_switch_mm can generate the address with
	 * adrp+add rather than a load from some global variable.
	 *
	 * To do this we need to go via a temporary pgd.
	 */
	cpu_replace_ttbr1(__va(pgd_phys));
646
	memcpy(swapper_pg_dir, pgdp, PGD_SIZE);
647
	cpu_replace_ttbr1(lm_alias(swapper_pg_dir));
648 649 650 651 652 653 654 655

	pgd_clear_fixmap();
	memblock_free(pgd_phys, PAGE_SIZE);

	/*
	 * We only reuse the PGD from the swapper_pg_dir, not the pud + pmd
	 * allocated with it.
	 */
656
	memblock_free(__pa_symbol(swapper_pg_dir) + PAGE_SIZE,
657 658
		      __pa_symbol(swapper_pg_end) - __pa_symbol(swapper_pg_dir)
		      - PAGE_SIZE);
C
Catalin Marinas 已提交
659 660 661 662 663 664 665
}

/*
 * Check whether a kernel address is valid (derived from arch/x86/).
 */
int kern_addr_valid(unsigned long addr)
{
666 667 668 669
	pgd_t *pgdp;
	pud_t *pudp, pud;
	pmd_t *pmdp, pmd;
	pte_t *ptep, pte;
C
Catalin Marinas 已提交
670 671 672 673

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

674 675
	pgdp = pgd_offset_k(addr);
	if (pgd_none(READ_ONCE(*pgdp)))
C
Catalin Marinas 已提交
676 677
		return 0;

678 679 680
	pudp = pud_offset(pgdp, addr);
	pud = READ_ONCE(*pudp);
	if (pud_none(pud))
C
Catalin Marinas 已提交
681 682
		return 0;

683 684
	if (pud_sect(pud))
		return pfn_valid(pud_pfn(pud));
685

686 687 688
	pmdp = pmd_offset(pudp, addr);
	pmd = READ_ONCE(*pmdp);
	if (pmd_none(pmd))
C
Catalin Marinas 已提交
689 690
		return 0;

691 692
	if (pmd_sect(pmd))
		return pfn_valid(pmd_pfn(pmd));
693

694 695 696
	ptep = pte_offset_kernel(pmdp, addr);
	pte = READ_ONCE(*ptep);
	if (pte_none(pte))
C
Catalin Marinas 已提交
697 698
		return 0;

699
	return pfn_valid(pte_pfn(pte));
C
Catalin Marinas 已提交
700 701
}
#ifdef CONFIG_SPARSEMEM_VMEMMAP
702
#if !ARM64_SWAPPER_USES_SECTION_MAPS
703 704
int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,
		struct vmem_altmap *altmap)
C
Catalin Marinas 已提交
705
{
706
	return vmemmap_populate_basepages(start, end, node);
C
Catalin Marinas 已提交
707
}
708
#else	/* !ARM64_SWAPPER_USES_SECTION_MAPS */
709 710
int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,
		struct vmem_altmap *altmap)
C
Catalin Marinas 已提交
711
{
712
	unsigned long addr = start;
C
Catalin Marinas 已提交
713
	unsigned long next;
714 715 716
	pgd_t *pgdp;
	pud_t *pudp;
	pmd_t *pmdp;
C
Catalin Marinas 已提交
717 718 719 720

	do {
		next = pmd_addr_end(addr, end);

721 722
		pgdp = vmemmap_pgd_populate(addr, node);
		if (!pgdp)
C
Catalin Marinas 已提交
723 724
			return -ENOMEM;

725 726
		pudp = vmemmap_pud_populate(pgdp, addr, node);
		if (!pudp)
C
Catalin Marinas 已提交
727 728
			return -ENOMEM;

729 730
		pmdp = pmd_offset(pudp, addr);
		if (pmd_none(READ_ONCE(*pmdp))) {
C
Catalin Marinas 已提交
731 732 733 734 735 736
			void *p = NULL;

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

737
			pmd_set_huge(pmdp, __pa(p), __pgprot(PROT_SECT_NORMAL));
C
Catalin Marinas 已提交
738
		} else
739
			vmemmap_verify((pte_t *)pmdp, node, addr, next);
C
Catalin Marinas 已提交
740 741 742 743 744
	} while (addr = next, addr != end);

	return 0;
}
#endif	/* CONFIG_ARM64_64K_PAGES */
745 746
void vmemmap_free(unsigned long start, unsigned long end,
		struct vmem_altmap *altmap)
747 748
{
}
C
Catalin Marinas 已提交
749
#endif	/* CONFIG_SPARSEMEM_VMEMMAP */
750 751 752

static inline pud_t * fixmap_pud(unsigned long addr)
{
753 754
	pgd_t *pgdp = pgd_offset_k(addr);
	pgd_t pgd = READ_ONCE(*pgdp);
755

756
	BUG_ON(pgd_none(pgd) || pgd_bad(pgd));
757

758
	return pud_offset_kimg(pgdp, addr);
759 760 761 762
}

static inline pmd_t * fixmap_pmd(unsigned long addr)
{
763 764
	pud_t *pudp = fixmap_pud(addr);
	pud_t pud = READ_ONCE(*pudp);
765

766
	BUG_ON(pud_none(pud) || pud_bad(pud));
767

768
	return pmd_offset_kimg(pudp, addr);
769 770 771 772
}

static inline pte_t * fixmap_pte(unsigned long addr)
{
773
	return &bm_pte[pte_index(addr)];
774 775
}

776 777 778 779 780 781
/*
 * 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.
 */
782 783
void __init early_fixmap_init(void)
{
784 785 786
	pgd_t *pgdp, pgd;
	pud_t *pudp;
	pmd_t *pmdp;
787 788
	unsigned long addr = FIXADDR_START;

789 790
	pgdp = pgd_offset_k(addr);
	pgd = READ_ONCE(*pgdp);
791
	if (CONFIG_PGTABLE_LEVELS > 3 &&
792
	    !(pgd_none(pgd) || pgd_page_paddr(pgd) == __pa_symbol(bm_pud))) {
793 794 795 796 797 798
		/*
		 * 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));
799
		pudp = pud_offset_kimg(pgdp, addr);
800
	} else {
801 802 803
		if (pgd_none(pgd))
			__pgd_populate(pgdp, __pa_symbol(bm_pud), PUD_TYPE_TABLE);
		pudp = fixmap_pud(addr);
804
	}
805 806 807 808
	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);
809 810 811

	/*
	 * The boot-ioremap range spans multiple pmds, for which
812
	 * we are not prepared:
813 814 815 816
	 */
	BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT)
		     != (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT));

817 818
	if ((pmdp != fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)))
	     || pmdp != fixmap_pmd(fix_to_virt(FIX_BTMAP_END))) {
819
		WARN_ON(1);
820 821
		pr_warn("pmdp %p != %p, %p\n",
			pmdp, fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)),
822 823 824 825 826 827 828 829 830 831 832
			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);
	}
}

833 834 835 836
/*
 * 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.
 */
837 838 839 840
void __set_fixmap(enum fixed_addresses idx,
			       phys_addr_t phys, pgprot_t flags)
{
	unsigned long addr = __fix_to_virt(idx);
841
	pte_t *ptep;
842

843
	BUG_ON(idx <= FIX_HOLE || idx >= __end_of_fixed_addresses);
844

845
	ptep = fixmap_pte(addr);
846 847

	if (pgprot_val(flags)) {
848
		set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, flags));
849
	} else {
850
		pte_clear(&init_mm, addr, ptep);
851 852 853
		flush_tlb_kernel_range(addr, addr+PAGE_SIZE);
	}
}
854

855
void *__init __fixmap_remap_fdt(phys_addr_t dt_phys, int *size, pgprot_t prot)
856 857
{
	const u64 dt_virt_base = __fix_to_virt(FIX_FDT);
858
	int offset;
859 860 861 862 863
	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
864 865 866
	 * 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.
867 868 869 870 871 872 873 874
	 */
	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
875
	 * to call create_mapping_noalloc() this early.
876 877 878 879 880 881 882 883
	 *
	 * 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);

884 885
	BUILD_BUG_ON(__fix_to_virt(FIX_FDT_END) >> SWAPPER_TABLE_SHIFT !=
		     __fix_to_virt(FIX_BTMAP_BEGIN) >> SWAPPER_TABLE_SHIFT);
886

887
	offset = dt_phys % SWAPPER_BLOCK_SIZE;
888 889 890
	dt_virt = (void *)dt_virt_base + offset;

	/* map the first chunk so we can read the size from the header */
891 892
	create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE),
			dt_virt_base, SWAPPER_BLOCK_SIZE, prot);
893

894
	if (fdt_magic(dt_virt) != FDT_MAGIC)
895 896
		return NULL;

897 898
	*size = fdt_totalsize(dt_virt);
	if (*size > MAX_FDT_SIZE)
899 900
		return NULL;

901
	if (offset + *size > SWAPPER_BLOCK_SIZE)
902
		create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), dt_virt_base,
903
			       round_up(offset + *size, SWAPPER_BLOCK_SIZE), prot);
904

905 906
	return dt_virt;
}
907

908 909 910 911 912 913 914 915 916 917
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);
918 919
	return dt_virt;
}
920 921 922 923 924 925 926 927 928 929 930 931

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;
}

932
int pud_set_huge(pud_t *pudp, phys_addr_t phys, pgprot_t prot)
933
{
934 935
	pgprot_t sect_prot = __pgprot(PUD_TYPE_SECT |
					pgprot_val(mk_sect_prot(prot)));
936 937 938 939 940

	/* ioremap_page_range doesn't honour BBM */
	if (pud_present(READ_ONCE(*pudp)))
		return 0;

941
	BUG_ON(phys & ~PUD_MASK);
942
	set_pud(pudp, pfn_pud(__phys_to_pfn(phys), sect_prot));
943 944 945
	return 1;
}

946
int pmd_set_huge(pmd_t *pmdp, phys_addr_t phys, pgprot_t prot)
947
{
948 949
	pgprot_t sect_prot = __pgprot(PMD_TYPE_SECT |
					pgprot_val(mk_sect_prot(prot)));
950 951 952 953 954

	/* ioremap_page_range doesn't honour BBM */
	if (pmd_present(READ_ONCE(*pmdp)))
		return 0;

955
	BUG_ON(phys & ~PMD_MASK);
956
	set_pmd(pmdp, pfn_pmd(__phys_to_pfn(phys), sect_prot));
957 958 959
	return 1;
}

960
int pud_clear_huge(pud_t *pudp)
961
{
962
	if (!pud_sect(READ_ONCE(*pudp)))
963
		return 0;
964
	pud_clear(pudp);
965 966 967
	return 1;
}

968
int pmd_clear_huge(pmd_t *pmdp)
969
{
970
	if (!pmd_sect(READ_ONCE(*pmdp)))
971
		return 0;
972
	pmd_clear(pmdp);
973 974
	return 1;
}