Commit | Line | Data |
---|---|---|
c1cc1552 CM |
1 | /* |
2 | * Based on arch/arm/mm/mmu.c | |
3 | * | |
4 | * Copyright (C) 1995-2005 Russell King | |
5 | * Copyright (C) 2012 ARM Ltd. | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program. If not, see <http://www.gnu.org/licenses/>. | |
18 | */ | |
19 | ||
20 | #include <linux/export.h> | |
21 | #include <linux/kernel.h> | |
22 | #include <linux/errno.h> | |
23 | #include <linux/init.h> | |
61bd93ce | 24 | #include <linux/libfdt.h> |
c1cc1552 CM |
25 | #include <linux/mman.h> |
26 | #include <linux/nodemask.h> | |
27 | #include <linux/memblock.h> | |
28 | #include <linux/fs.h> | |
2475ff9d | 29 | #include <linux/io.h> |
41089357 | 30 | #include <linux/slab.h> |
da141706 | 31 | #include <linux/stop_machine.h> |
c1cc1552 | 32 | |
21ab99c2 | 33 | #include <asm/barrier.h> |
c1cc1552 | 34 | #include <asm/cputype.h> |
af86e597 | 35 | #include <asm/fixmap.h> |
068a17a5 | 36 | #include <asm/kasan.h> |
b433dce0 | 37 | #include <asm/kernel-pgtable.h> |
c1cc1552 CM |
38 | #include <asm/sections.h> |
39 | #include <asm/setup.h> | |
40 | #include <asm/sizes.h> | |
41 | #include <asm/tlb.h> | |
c79b954b | 42 | #include <asm/memblock.h> |
c1cc1552 CM |
43 | #include <asm/mmu_context.h> |
44 | ||
45 | #include "mm.h" | |
46 | ||
dd006da2 AB |
47 | u64 idmap_t0sz = TCR_T0SZ(VA_BITS); |
48 | ||
a7f8de16 AB |
49 | u64 kimage_voffset __read_mostly; |
50 | EXPORT_SYMBOL(kimage_voffset); | |
51 | ||
c1cc1552 CM |
52 | /* |
53 | * Empty_zero_page is a special page that is used for zero-initialized data | |
54 | * and COW. | |
55 | */ | |
5227cfa7 | 56 | unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)] __page_aligned_bss; |
c1cc1552 CM |
57 | EXPORT_SYMBOL(empty_zero_page); |
58 | ||
f9040773 AB |
59 | static pte_t bm_pte[PTRS_PER_PTE] __page_aligned_bss; |
60 | static pmd_t bm_pmd[PTRS_PER_PMD] __page_aligned_bss __maybe_unused; | |
61 | static pud_t bm_pud[PTRS_PER_PUD] __page_aligned_bss __maybe_unused; | |
62 | ||
c1cc1552 CM |
63 | pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
64 | unsigned long size, pgprot_t vma_prot) | |
65 | { | |
66 | if (!pfn_valid(pfn)) | |
67 | return pgprot_noncached(vma_prot); | |
68 | else if (file->f_flags & O_SYNC) | |
69 | return pgprot_writecombine(vma_prot); | |
70 | return vma_prot; | |
71 | } | |
72 | EXPORT_SYMBOL(phys_mem_access_prot); | |
73 | ||
f4710445 | 74 | static phys_addr_t __init early_pgtable_alloc(void) |
c1cc1552 | 75 | { |
7142392d SP |
76 | phys_addr_t phys; |
77 | void *ptr; | |
78 | ||
21ab99c2 | 79 | phys = memblock_alloc(PAGE_SIZE, PAGE_SIZE); |
7142392d | 80 | BUG_ON(!phys); |
f4710445 MR |
81 | |
82 | /* | |
83 | * The FIX_{PGD,PUD,PMD} slots may be in active use, but the FIX_PTE | |
84 | * slot will be free, so we can (ab)use the FIX_PTE slot to initialise | |
85 | * any level of table. | |
86 | */ | |
87 | ptr = pte_set_fixmap(phys); | |
88 | ||
21ab99c2 MR |
89 | memset(ptr, 0, PAGE_SIZE); |
90 | ||
f4710445 MR |
91 | /* |
92 | * Implicit barriers also ensure the zeroed page is visible to the page | |
93 | * table walker | |
94 | */ | |
95 | pte_clear_fixmap(); | |
96 | ||
97 | return phys; | |
c1cc1552 CM |
98 | } |
99 | ||
da141706 LA |
100 | /* |
101 | * remap a PMD into pages | |
102 | */ | |
103 | static void split_pmd(pmd_t *pmd, pte_t *pte) | |
104 | { | |
105 | unsigned long pfn = pmd_pfn(*pmd); | |
106 | int i = 0; | |
107 | ||
108 | do { | |
109 | /* | |
110 | * Need to have the least restrictive permissions available | |
667c2759 | 111 | * permissions will be fixed up later |
da141706 | 112 | */ |
667c2759 | 113 | set_pte(pte, pfn_pte(pfn, PAGE_KERNEL_EXEC)); |
da141706 LA |
114 | pfn++; |
115 | } while (pte++, i++, i < PTRS_PER_PTE); | |
116 | } | |
117 | ||
118 | static void alloc_init_pte(pmd_t *pmd, unsigned long addr, | |
667c2759 | 119 | unsigned long end, unsigned long pfn, |
da141706 | 120 | pgprot_t prot, |
f4710445 | 121 | phys_addr_t (*pgtable_alloc)(void)) |
c1cc1552 CM |
122 | { |
123 | pte_t *pte; | |
124 | ||
a1c76574 | 125 | if (pmd_none(*pmd) || pmd_sect(*pmd)) { |
132233a7 LA |
126 | phys_addr_t pte_phys; |
127 | BUG_ON(!pgtable_alloc); | |
128 | pte_phys = pgtable_alloc(); | |
f4710445 | 129 | pte = pte_set_fixmap(pte_phys); |
da141706 LA |
130 | if (pmd_sect(*pmd)) |
131 | split_pmd(pmd, pte); | |
f4710445 | 132 | __pmd_populate(pmd, pte_phys, PMD_TYPE_TABLE); |
da141706 | 133 | flush_tlb_all(); |
f4710445 | 134 | pte_clear_fixmap(); |
c1cc1552 | 135 | } |
a1c76574 | 136 | BUG_ON(pmd_bad(*pmd)); |
c1cc1552 | 137 | |
f4710445 | 138 | pte = pte_set_fixmap_offset(pmd, addr); |
c1cc1552 | 139 | do { |
667c2759 CM |
140 | set_pte(pte, pfn_pte(pfn, prot)); |
141 | pfn++; | |
142 | } while (pte++, addr += PAGE_SIZE, addr != end); | |
f4710445 MR |
143 | |
144 | pte_clear_fixmap(); | |
c1cc1552 CM |
145 | } |
146 | ||
9a17a213 | 147 | static void split_pud(pud_t *old_pud, pmd_t *pmd) |
da141706 LA |
148 | { |
149 | unsigned long addr = pud_pfn(*old_pud) << PAGE_SHIFT; | |
150 | pgprot_t prot = __pgprot(pud_val(*old_pud) ^ addr); | |
151 | int i = 0; | |
152 | ||
153 | do { | |
1e43ba9c | 154 | set_pmd(pmd, __pmd(addr | pgprot_val(prot))); |
da141706 LA |
155 | addr += PMD_SIZE; |
156 | } while (pmd++, i++, i < PTRS_PER_PMD); | |
157 | } | |
158 | ||
83863f25 LA |
159 | #ifdef CONFIG_DEBUG_PAGEALLOC |
160 | static bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void)) | |
161 | { | |
162 | ||
163 | /* | |
164 | * If debug_page_alloc is enabled we must map the linear map | |
165 | * using pages. However, other mappings created by | |
166 | * create_mapping_noalloc must use sections in some cases. Allow | |
167 | * sections to be used in those cases, where no pgtable_alloc | |
168 | * function is provided. | |
169 | */ | |
170 | return !pgtable_alloc || !debug_pagealloc_enabled(); | |
171 | } | |
172 | #else | |
173 | static bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void)) | |
174 | { | |
175 | return true; | |
176 | } | |
177 | #endif | |
178 | ||
11509a30 | 179 | static void alloc_init_pmd(pud_t *pud, unsigned long addr, unsigned long end, |
da141706 | 180 | phys_addr_t phys, pgprot_t prot, |
f4710445 | 181 | phys_addr_t (*pgtable_alloc)(void)) |
c1cc1552 CM |
182 | { |
183 | pmd_t *pmd; | |
184 | unsigned long next; | |
185 | ||
186 | /* | |
187 | * Check for initial section mappings in the pgd/pud and remove them. | |
188 | */ | |
a1c76574 | 189 | if (pud_none(*pud) || pud_sect(*pud)) { |
132233a7 LA |
190 | phys_addr_t pmd_phys; |
191 | BUG_ON(!pgtable_alloc); | |
192 | pmd_phys = pgtable_alloc(); | |
f4710445 | 193 | pmd = pmd_set_fixmap(pmd_phys); |
da141706 LA |
194 | if (pud_sect(*pud)) { |
195 | /* | |
196 | * need to have the 1G of mappings continue to be | |
197 | * present | |
198 | */ | |
199 | split_pud(pud, pmd); | |
200 | } | |
f4710445 | 201 | __pud_populate(pud, pmd_phys, PUD_TYPE_TABLE); |
da141706 | 202 | flush_tlb_all(); |
f4710445 | 203 | pmd_clear_fixmap(); |
c1cc1552 | 204 | } |
a1c76574 | 205 | BUG_ON(pud_bad(*pud)); |
c1cc1552 | 206 | |
f4710445 | 207 | pmd = pmd_set_fixmap_offset(pud, addr); |
c1cc1552 CM |
208 | do { |
209 | next = pmd_addr_end(addr, end); | |
210 | /* try section mapping first */ | |
83863f25 LA |
211 | if (((addr | next | phys) & ~SECTION_MASK) == 0 && |
212 | block_mappings_allowed(pgtable_alloc)) { | |
a55f9929 | 213 | pmd_t old_pmd =*pmd; |
c661cb1c | 214 | pmd_set_huge(pmd, phys, prot); |
a55f9929 CM |
215 | /* |
216 | * Check for previous table entries created during | |
217 | * boot (__create_page_tables) and flush them. | |
218 | */ | |
523d6e9f | 219 | if (!pmd_none(old_pmd)) { |
a55f9929 | 220 | flush_tlb_all(); |
523d6e9f | 221 | if (pmd_table(old_pmd)) { |
316b39db | 222 | phys_addr_t table = pmd_page_paddr(old_pmd); |
41089357 CM |
223 | if (!WARN_ON_ONCE(slab_is_available())) |
224 | memblock_free(table, PAGE_SIZE); | |
523d6e9f | 225 | } |
226 | } | |
a55f9929 | 227 | } else { |
667c2759 | 228 | alloc_init_pte(pmd, addr, next, __phys_to_pfn(phys), |
21ab99c2 | 229 | prot, pgtable_alloc); |
a55f9929 | 230 | } |
c1cc1552 CM |
231 | phys += next - addr; |
232 | } while (pmd++, addr = next, addr != end); | |
f4710445 MR |
233 | |
234 | pmd_clear_fixmap(); | |
c1cc1552 CM |
235 | } |
236 | ||
da141706 LA |
237 | static inline bool use_1G_block(unsigned long addr, unsigned long next, |
238 | unsigned long phys) | |
239 | { | |
240 | if (PAGE_SHIFT != 12) | |
241 | return false; | |
242 | ||
243 | if (((addr | next | phys) & ~PUD_MASK) != 0) | |
244 | return false; | |
245 | ||
246 | return true; | |
247 | } | |
248 | ||
11509a30 | 249 | static void alloc_init_pud(pgd_t *pgd, unsigned long addr, unsigned long end, |
da141706 | 250 | phys_addr_t phys, pgprot_t prot, |
f4710445 | 251 | phys_addr_t (*pgtable_alloc)(void)) |
c1cc1552 | 252 | { |
c79b954b | 253 | pud_t *pud; |
c1cc1552 CM |
254 | unsigned long next; |
255 | ||
c79b954b | 256 | if (pgd_none(*pgd)) { |
132233a7 LA |
257 | phys_addr_t pud_phys; |
258 | BUG_ON(!pgtable_alloc); | |
259 | pud_phys = pgtable_alloc(); | |
f4710445 | 260 | __pgd_populate(pgd, pud_phys, PUD_TYPE_TABLE); |
c79b954b JL |
261 | } |
262 | BUG_ON(pgd_bad(*pgd)); | |
263 | ||
f4710445 | 264 | pud = pud_set_fixmap_offset(pgd, addr); |
c1cc1552 CM |
265 | do { |
266 | next = pud_addr_end(addr, end); | |
206a2a73 SC |
267 | |
268 | /* | |
269 | * For 4K granule only, attempt to put down a 1GB block | |
270 | */ | |
83863f25 LA |
271 | if (use_1G_block(addr, next, phys) && |
272 | block_mappings_allowed(pgtable_alloc)) { | |
206a2a73 | 273 | pud_t old_pud = *pud; |
c661cb1c | 274 | pud_set_huge(pud, phys, prot); |
206a2a73 SC |
275 | |
276 | /* | |
277 | * If we have an old value for a pud, it will | |
278 | * be pointing to a pmd table that we no longer | |
279 | * need (from swapper_pg_dir). | |
280 | * | |
281 | * Look up the old pmd table and free it. | |
282 | */ | |
283 | if (!pud_none(old_pud)) { | |
206a2a73 | 284 | flush_tlb_all(); |
523d6e9f | 285 | if (pud_table(old_pud)) { |
316b39db | 286 | phys_addr_t table = pud_page_paddr(old_pud); |
41089357 CM |
287 | if (!WARN_ON_ONCE(slab_is_available())) |
288 | memblock_free(table, PAGE_SIZE); | |
523d6e9f | 289 | } |
206a2a73 SC |
290 | } |
291 | } else { | |
11509a30 | 292 | alloc_init_pmd(pud, addr, next, phys, prot, |
21ab99c2 | 293 | pgtable_alloc); |
206a2a73 | 294 | } |
c1cc1552 CM |
295 | phys += next - addr; |
296 | } while (pud++, addr = next, addr != end); | |
f4710445 MR |
297 | |
298 | pud_clear_fixmap(); | |
c1cc1552 CM |
299 | } |
300 | ||
301 | /* | |
302 | * Create the page directory entries and any necessary page tables for the | |
303 | * mapping specified by 'md'. | |
304 | */ | |
11509a30 | 305 | static void init_pgd(pgd_t *pgd, phys_addr_t phys, unsigned long virt, |
da141706 | 306 | phys_addr_t size, pgprot_t prot, |
f4710445 | 307 | phys_addr_t (*pgtable_alloc)(void)) |
c1cc1552 CM |
308 | { |
309 | unsigned long addr, length, end, next; | |
c1cc1552 | 310 | |
cc5d2b3b MR |
311 | /* |
312 | * If the virtual and physical address don't have the same offset | |
313 | * within a page, we cannot map the region as the caller expects. | |
314 | */ | |
315 | if (WARN_ON((phys ^ virt) & ~PAGE_MASK)) | |
316 | return; | |
317 | ||
9c4e08a3 | 318 | phys &= PAGE_MASK; |
c1cc1552 CM |
319 | addr = virt & PAGE_MASK; |
320 | length = PAGE_ALIGN(size + (virt & ~PAGE_MASK)); | |
321 | ||
c1cc1552 CM |
322 | end = addr + length; |
323 | do { | |
324 | next = pgd_addr_end(addr, end); | |
11509a30 | 325 | alloc_init_pud(pgd, addr, next, phys, prot, pgtable_alloc); |
c1cc1552 CM |
326 | phys += next - addr; |
327 | } while (pgd++, addr = next, addr != end); | |
328 | } | |
329 | ||
f4710445 | 330 | static phys_addr_t late_pgtable_alloc(void) |
da141706 | 331 | { |
21ab99c2 | 332 | void *ptr = (void *)__get_free_page(PGALLOC_GFP); |
da141706 | 333 | BUG_ON(!ptr); |
21ab99c2 MR |
334 | |
335 | /* Ensure the zeroed page is visible to the page table walker */ | |
336 | dsb(ishst); | |
f4710445 | 337 | return __pa(ptr); |
da141706 LA |
338 | } |
339 | ||
11509a30 MR |
340 | static void __create_pgd_mapping(pgd_t *pgdir, phys_addr_t phys, |
341 | unsigned long virt, phys_addr_t size, | |
342 | pgprot_t prot, | |
343 | phys_addr_t (*alloc)(void)) | |
344 | { | |
345 | init_pgd(pgd_offset_raw(pgdir, virt), phys, virt, size, prot, alloc); | |
346 | } | |
347 | ||
132233a7 LA |
348 | /* |
349 | * This function can only be used to modify existing table entries, | |
350 | * without allocating new levels of table. Note that this permits the | |
351 | * creation of new section or page entries. | |
352 | */ | |
353 | static void __init create_mapping_noalloc(phys_addr_t phys, unsigned long virt, | |
da141706 | 354 | phys_addr_t size, pgprot_t prot) |
d7ecbddf MS |
355 | { |
356 | if (virt < VMALLOC_START) { | |
357 | pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n", | |
358 | &phys, virt); | |
359 | return; | |
360 | } | |
11509a30 | 361 | __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, |
132233a7 | 362 | NULL); |
d7ecbddf MS |
363 | } |
364 | ||
8ce837ce AB |
365 | void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys, |
366 | unsigned long virt, phys_addr_t size, | |
367 | pgprot_t prot) | |
368 | { | |
11509a30 MR |
369 | __create_pgd_mapping(mm->pgd, phys, virt, size, prot, |
370 | late_pgtable_alloc); | |
d7ecbddf MS |
371 | } |
372 | ||
da141706 LA |
373 | static void create_mapping_late(phys_addr_t phys, unsigned long virt, |
374 | phys_addr_t size, pgprot_t prot) | |
375 | { | |
376 | if (virt < VMALLOC_START) { | |
377 | pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n", | |
378 | &phys, virt); | |
379 | return; | |
380 | } | |
381 | ||
11509a30 MR |
382 | __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, |
383 | late_pgtable_alloc); | |
da141706 LA |
384 | } |
385 | ||
068a17a5 | 386 | static void __init __map_memblock(pgd_t *pgd, phys_addr_t start, phys_addr_t end) |
da141706 | 387 | { |
7eb90f2f | 388 | unsigned long kernel_start = __pa(_text); |
f9040773 | 389 | unsigned long kernel_end = __pa(_etext); |
068a17a5 | 390 | |
da141706 | 391 | /* |
f9040773 AB |
392 | * Take care not to create a writable alias for the |
393 | * read-only text and rodata sections of the kernel image. | |
da141706 | 394 | */ |
068a17a5 | 395 | |
f9040773 | 396 | /* No overlap with the kernel text */ |
068a17a5 MR |
397 | if (end < kernel_start || start >= kernel_end) { |
398 | __create_pgd_mapping(pgd, start, __phys_to_virt(start), | |
399 | end - start, PAGE_KERNEL, | |
400 | early_pgtable_alloc); | |
401 | return; | |
da141706 LA |
402 | } |
403 | ||
068a17a5 | 404 | /* |
f9040773 AB |
405 | * This block overlaps the kernel text mapping. |
406 | * Map the portion(s) which don't overlap. | |
068a17a5 MR |
407 | */ |
408 | if (start < kernel_start) | |
409 | __create_pgd_mapping(pgd, start, | |
410 | __phys_to_virt(start), | |
411 | kernel_start - start, PAGE_KERNEL, | |
412 | early_pgtable_alloc); | |
413 | if (kernel_end < end) | |
414 | __create_pgd_mapping(pgd, kernel_end, | |
415 | __phys_to_virt(kernel_end), | |
416 | end - kernel_end, PAGE_KERNEL, | |
417 | early_pgtable_alloc); | |
f9040773 AB |
418 | |
419 | /* | |
7eb90f2f | 420 | * Map the linear alias of the [_text, _etext) interval as |
f9040773 AB |
421 | * read-only/non-executable. This makes the contents of the |
422 | * region accessible to subsystems such as hibernate, but | |
423 | * protects it from inadvertent modification or execution. | |
424 | */ | |
425 | __create_pgd_mapping(pgd, kernel_start, __phys_to_virt(kernel_start), | |
426 | kernel_end - kernel_start, PAGE_KERNEL_RO, | |
427 | early_pgtable_alloc); | |
da141706 | 428 | } |
da141706 | 429 | |
068a17a5 | 430 | static void __init map_mem(pgd_t *pgd) |
c1cc1552 CM |
431 | { |
432 | struct memblock_region *reg; | |
f6bc87c3 | 433 | |
c1cc1552 CM |
434 | /* map all the memory banks */ |
435 | for_each_memblock(memory, reg) { | |
436 | phys_addr_t start = reg->base; | |
437 | phys_addr_t end = start + reg->size; | |
438 | ||
439 | if (start >= end) | |
440 | break; | |
68709f45 AB |
441 | if (memblock_is_nomap(reg)) |
442 | continue; | |
c1cc1552 | 443 | |
068a17a5 | 444 | __map_memblock(pgd, start, end); |
c1cc1552 CM |
445 | } |
446 | } | |
447 | ||
da141706 LA |
448 | void mark_rodata_ro(void) |
449 | { | |
2f39b5f9 | 450 | unsigned long section_size; |
f9040773 | 451 | |
7eb90f2f AB |
452 | section_size = (unsigned long)__start_rodata - (unsigned long)_text; |
453 | create_mapping_late(__pa(_text), (unsigned long)_text, | |
2f39b5f9 JL |
454 | section_size, PAGE_KERNEL_ROX); |
455 | /* | |
456 | * mark .rodata as read only. Use _etext rather than __end_rodata to | |
457 | * cover NOTES and EXCEPTION_TABLE. | |
458 | */ | |
459 | section_size = (unsigned long)_etext - (unsigned long)__start_rodata; | |
460 | create_mapping_late(__pa(__start_rodata), (unsigned long)__start_rodata, | |
461 | section_size, PAGE_KERNEL_RO); | |
da141706 | 462 | } |
da141706 LA |
463 | |
464 | void fixup_init(void) | |
465 | { | |
f9040773 AB |
466 | /* |
467 | * Unmap the __init region but leave the VM area in place. This | |
468 | * prevents the region from being reused for kernel modules, which | |
469 | * is not supported by kallsyms. | |
470 | */ | |
471 | unmap_kernel_range((u64)__init_begin, (u64)(__init_end - __init_begin)); | |
da141706 LA |
472 | } |
473 | ||
2c09ec06 AB |
474 | static void __init map_kernel_segment(pgd_t *pgd, void *va_start, void *va_end, |
475 | pgprot_t prot, struct vm_struct *vma) | |
068a17a5 MR |
476 | { |
477 | phys_addr_t pa_start = __pa(va_start); | |
478 | unsigned long size = va_end - va_start; | |
479 | ||
480 | BUG_ON(!PAGE_ALIGNED(pa_start)); | |
481 | BUG_ON(!PAGE_ALIGNED(size)); | |
482 | ||
483 | __create_pgd_mapping(pgd, pa_start, (unsigned long)va_start, size, prot, | |
484 | early_pgtable_alloc); | |
f9040773 AB |
485 | |
486 | vma->addr = va_start; | |
487 | vma->phys_addr = pa_start; | |
488 | vma->size = size; | |
489 | vma->flags = VM_MAP; | |
490 | vma->caller = __builtin_return_address(0); | |
491 | ||
492 | vm_area_add_early(vma); | |
068a17a5 MR |
493 | } |
494 | ||
495 | /* | |
496 | * Create fine-grained mappings for the kernel. | |
497 | */ | |
498 | static void __init map_kernel(pgd_t *pgd) | |
499 | { | |
2f39b5f9 | 500 | static struct vm_struct vmlinux_text, vmlinux_rodata, vmlinux_init, vmlinux_data; |
068a17a5 | 501 | |
7eb90f2f | 502 | map_kernel_segment(pgd, _text, __start_rodata, PAGE_KERNEL_EXEC, &vmlinux_text); |
2c09ec06 AB |
503 | map_kernel_segment(pgd, __start_rodata, _etext, PAGE_KERNEL, &vmlinux_rodata); |
504 | map_kernel_segment(pgd, __init_begin, __init_end, PAGE_KERNEL_EXEC, | |
505 | &vmlinux_init); | |
506 | map_kernel_segment(pgd, _data, _end, PAGE_KERNEL, &vmlinux_data); | |
068a17a5 | 507 | |
f9040773 AB |
508 | if (!pgd_val(*pgd_offset_raw(pgd, FIXADDR_START))) { |
509 | /* | |
510 | * The fixmap falls in a separate pgd to the kernel, and doesn't | |
511 | * live in the carveout for the swapper_pg_dir. We can simply | |
512 | * re-use the existing dir for the fixmap. | |
513 | */ | |
514 | set_pgd(pgd_offset_raw(pgd, FIXADDR_START), | |
515 | *pgd_offset_k(FIXADDR_START)); | |
516 | } else if (CONFIG_PGTABLE_LEVELS > 3) { | |
517 | /* | |
518 | * The fixmap shares its top level pgd entry with the kernel | |
519 | * mapping. This can really only occur when we are running | |
520 | * with 16k/4 levels, so we can simply reuse the pud level | |
521 | * entry instead. | |
522 | */ | |
523 | BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES)); | |
524 | set_pud(pud_set_fixmap_offset(pgd, FIXADDR_START), | |
525 | __pud(__pa(bm_pmd) | PUD_TYPE_TABLE)); | |
526 | pud_clear_fixmap(); | |
527 | } else { | |
528 | BUG(); | |
529 | } | |
068a17a5 MR |
530 | |
531 | kasan_copy_shadow(pgd); | |
532 | } | |
533 | ||
c1cc1552 CM |
534 | /* |
535 | * paging_init() sets up the page tables, initialises the zone memory | |
536 | * maps and sets up the zero page. | |
537 | */ | |
538 | void __init paging_init(void) | |
539 | { | |
068a17a5 MR |
540 | phys_addr_t pgd_phys = early_pgtable_alloc(); |
541 | pgd_t *pgd = pgd_set_fixmap(pgd_phys); | |
542 | ||
543 | map_kernel(pgd); | |
544 | map_mem(pgd); | |
545 | ||
546 | /* | |
547 | * We want to reuse the original swapper_pg_dir so we don't have to | |
548 | * communicate the new address to non-coherent secondaries in | |
549 | * secondary_entry, and so cpu_switch_mm can generate the address with | |
550 | * adrp+add rather than a load from some global variable. | |
551 | * | |
552 | * To do this we need to go via a temporary pgd. | |
553 | */ | |
554 | cpu_replace_ttbr1(__va(pgd_phys)); | |
555 | memcpy(swapper_pg_dir, pgd, PAGE_SIZE); | |
556 | cpu_replace_ttbr1(swapper_pg_dir); | |
557 | ||
558 | pgd_clear_fixmap(); | |
559 | memblock_free(pgd_phys, PAGE_SIZE); | |
560 | ||
561 | /* | |
562 | * We only reuse the PGD from the swapper_pg_dir, not the pud + pmd | |
563 | * allocated with it. | |
564 | */ | |
565 | memblock_free(__pa(swapper_pg_dir) + PAGE_SIZE, | |
566 | SWAPPER_DIR_SIZE - PAGE_SIZE); | |
c1cc1552 CM |
567 | } |
568 | ||
c1cc1552 CM |
569 | /* |
570 | * Check whether a kernel address is valid (derived from arch/x86/). | |
571 | */ | |
572 | int kern_addr_valid(unsigned long addr) | |
573 | { | |
574 | pgd_t *pgd; | |
575 | pud_t *pud; | |
576 | pmd_t *pmd; | |
577 | pte_t *pte; | |
578 | ||
579 | if ((((long)addr) >> VA_BITS) != -1UL) | |
580 | return 0; | |
581 | ||
582 | pgd = pgd_offset_k(addr); | |
583 | if (pgd_none(*pgd)) | |
584 | return 0; | |
585 | ||
586 | pud = pud_offset(pgd, addr); | |
587 | if (pud_none(*pud)) | |
588 | return 0; | |
589 | ||
206a2a73 SC |
590 | if (pud_sect(*pud)) |
591 | return pfn_valid(pud_pfn(*pud)); | |
592 | ||
c1cc1552 CM |
593 | pmd = pmd_offset(pud, addr); |
594 | if (pmd_none(*pmd)) | |
595 | return 0; | |
596 | ||
da6e4cb6 DA |
597 | if (pmd_sect(*pmd)) |
598 | return pfn_valid(pmd_pfn(*pmd)); | |
599 | ||
c1cc1552 CM |
600 | pte = pte_offset_kernel(pmd, addr); |
601 | if (pte_none(*pte)) | |
602 | return 0; | |
603 | ||
604 | return pfn_valid(pte_pfn(*pte)); | |
605 | } | |
606 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
b433dce0 | 607 | #if !ARM64_SWAPPER_USES_SECTION_MAPS |
0aad818b | 608 | int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node) |
c1cc1552 | 609 | { |
0aad818b | 610 | return vmemmap_populate_basepages(start, end, node); |
c1cc1552 | 611 | } |
b433dce0 | 612 | #else /* !ARM64_SWAPPER_USES_SECTION_MAPS */ |
0aad818b | 613 | int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node) |
c1cc1552 | 614 | { |
0aad818b | 615 | unsigned long addr = start; |
c1cc1552 CM |
616 | unsigned long next; |
617 | pgd_t *pgd; | |
618 | pud_t *pud; | |
619 | pmd_t *pmd; | |
620 | ||
621 | do { | |
622 | next = pmd_addr_end(addr, end); | |
623 | ||
624 | pgd = vmemmap_pgd_populate(addr, node); | |
625 | if (!pgd) | |
626 | return -ENOMEM; | |
627 | ||
628 | pud = vmemmap_pud_populate(pgd, addr, node); | |
629 | if (!pud) | |
630 | return -ENOMEM; | |
631 | ||
632 | pmd = pmd_offset(pud, addr); | |
633 | if (pmd_none(*pmd)) { | |
634 | void *p = NULL; | |
635 | ||
636 | p = vmemmap_alloc_block_buf(PMD_SIZE, node); | |
637 | if (!p) | |
638 | return -ENOMEM; | |
639 | ||
a501e324 | 640 | set_pmd(pmd, __pmd(__pa(p) | PROT_SECT_NORMAL)); |
c1cc1552 CM |
641 | } else |
642 | vmemmap_verify((pte_t *)pmd, node, addr, next); | |
643 | } while (addr = next, addr != end); | |
644 | ||
645 | return 0; | |
646 | } | |
647 | #endif /* CONFIG_ARM64_64K_PAGES */ | |
0aad818b | 648 | void vmemmap_free(unsigned long start, unsigned long end) |
0197518c TC |
649 | { |
650 | } | |
c1cc1552 | 651 | #endif /* CONFIG_SPARSEMEM_VMEMMAP */ |
af86e597 | 652 | |
af86e597 LA |
653 | static inline pud_t * fixmap_pud(unsigned long addr) |
654 | { | |
655 | pgd_t *pgd = pgd_offset_k(addr); | |
656 | ||
657 | BUG_ON(pgd_none(*pgd) || pgd_bad(*pgd)); | |
658 | ||
157962f5 | 659 | return pud_offset_kimg(pgd, addr); |
af86e597 LA |
660 | } |
661 | ||
662 | static inline pmd_t * fixmap_pmd(unsigned long addr) | |
663 | { | |
664 | pud_t *pud = fixmap_pud(addr); | |
665 | ||
666 | BUG_ON(pud_none(*pud) || pud_bad(*pud)); | |
667 | ||
157962f5 | 668 | return pmd_offset_kimg(pud, addr); |
af86e597 LA |
669 | } |
670 | ||
671 | static inline pte_t * fixmap_pte(unsigned long addr) | |
672 | { | |
157962f5 | 673 | return &bm_pte[pte_index(addr)]; |
af86e597 LA |
674 | } |
675 | ||
676 | void __init early_fixmap_init(void) | |
677 | { | |
678 | pgd_t *pgd; | |
679 | pud_t *pud; | |
680 | pmd_t *pmd; | |
681 | unsigned long addr = FIXADDR_START; | |
682 | ||
683 | pgd = pgd_offset_k(addr); | |
f80fb3a3 AB |
684 | if (CONFIG_PGTABLE_LEVELS > 3 && |
685 | !(pgd_none(*pgd) || pgd_page_paddr(*pgd) == __pa(bm_pud))) { | |
f9040773 AB |
686 | /* |
687 | * We only end up here if the kernel mapping and the fixmap | |
688 | * share the top level pgd entry, which should only happen on | |
689 | * 16k/4 levels configurations. | |
690 | */ | |
691 | BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES)); | |
692 | pud = pud_offset_kimg(pgd, addr); | |
693 | } else { | |
694 | pgd_populate(&init_mm, pgd, bm_pud); | |
695 | pud = fixmap_pud(addr); | |
696 | } | |
af86e597 | 697 | pud_populate(&init_mm, pud, bm_pmd); |
157962f5 | 698 | pmd = fixmap_pmd(addr); |
af86e597 LA |
699 | pmd_populate_kernel(&init_mm, pmd, bm_pte); |
700 | ||
701 | /* | |
702 | * The boot-ioremap range spans multiple pmds, for which | |
157962f5 | 703 | * we are not prepared: |
af86e597 LA |
704 | */ |
705 | BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT) | |
706 | != (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT)); | |
707 | ||
708 | if ((pmd != fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN))) | |
709 | || pmd != fixmap_pmd(fix_to_virt(FIX_BTMAP_END))) { | |
710 | WARN_ON(1); | |
711 | pr_warn("pmd %p != %p, %p\n", | |
712 | pmd, fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)), | |
713 | fixmap_pmd(fix_to_virt(FIX_BTMAP_END))); | |
714 | pr_warn("fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n", | |
715 | fix_to_virt(FIX_BTMAP_BEGIN)); | |
716 | pr_warn("fix_to_virt(FIX_BTMAP_END): %08lx\n", | |
717 | fix_to_virt(FIX_BTMAP_END)); | |
718 | ||
719 | pr_warn("FIX_BTMAP_END: %d\n", FIX_BTMAP_END); | |
720 | pr_warn("FIX_BTMAP_BEGIN: %d\n", FIX_BTMAP_BEGIN); | |
721 | } | |
722 | } | |
723 | ||
724 | void __set_fixmap(enum fixed_addresses idx, | |
725 | phys_addr_t phys, pgprot_t flags) | |
726 | { | |
727 | unsigned long addr = __fix_to_virt(idx); | |
728 | pte_t *pte; | |
729 | ||
b63dbef9 | 730 | BUG_ON(idx <= FIX_HOLE || idx >= __end_of_fixed_addresses); |
af86e597 LA |
731 | |
732 | pte = fixmap_pte(addr); | |
733 | ||
734 | if (pgprot_val(flags)) { | |
735 | set_pte(pte, pfn_pte(phys >> PAGE_SHIFT, flags)); | |
736 | } else { | |
737 | pte_clear(&init_mm, addr, pte); | |
738 | flush_tlb_kernel_range(addr, addr+PAGE_SIZE); | |
739 | } | |
740 | } | |
61bd93ce | 741 | |
f80fb3a3 | 742 | void *__init __fixmap_remap_fdt(phys_addr_t dt_phys, int *size, pgprot_t prot) |
61bd93ce AB |
743 | { |
744 | const u64 dt_virt_base = __fix_to_virt(FIX_FDT); | |
f80fb3a3 | 745 | int offset; |
61bd93ce AB |
746 | void *dt_virt; |
747 | ||
748 | /* | |
749 | * Check whether the physical FDT address is set and meets the minimum | |
750 | * alignment requirement. Since we are relying on MIN_FDT_ALIGN to be | |
751 | * at least 8 bytes so that we can always access the size field of the | |
752 | * FDT header after mapping the first chunk, double check here if that | |
753 | * is indeed the case. | |
754 | */ | |
755 | BUILD_BUG_ON(MIN_FDT_ALIGN < 8); | |
756 | if (!dt_phys || dt_phys % MIN_FDT_ALIGN) | |
757 | return NULL; | |
758 | ||
759 | /* | |
760 | * Make sure that the FDT region can be mapped without the need to | |
761 | * allocate additional translation table pages, so that it is safe | |
132233a7 | 762 | * to call create_mapping_noalloc() this early. |
61bd93ce AB |
763 | * |
764 | * On 64k pages, the FDT will be mapped using PTEs, so we need to | |
765 | * be in the same PMD as the rest of the fixmap. | |
766 | * On 4k pages, we'll use section mappings for the FDT so we only | |
767 | * have to be in the same PUD. | |
768 | */ | |
769 | BUILD_BUG_ON(dt_virt_base % SZ_2M); | |
770 | ||
b433dce0 SP |
771 | BUILD_BUG_ON(__fix_to_virt(FIX_FDT_END) >> SWAPPER_TABLE_SHIFT != |
772 | __fix_to_virt(FIX_BTMAP_BEGIN) >> SWAPPER_TABLE_SHIFT); | |
61bd93ce | 773 | |
b433dce0 | 774 | offset = dt_phys % SWAPPER_BLOCK_SIZE; |
61bd93ce AB |
775 | dt_virt = (void *)dt_virt_base + offset; |
776 | ||
777 | /* map the first chunk so we can read the size from the header */ | |
132233a7 LA |
778 | create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), |
779 | dt_virt_base, SWAPPER_BLOCK_SIZE, prot); | |
61bd93ce AB |
780 | |
781 | if (fdt_check_header(dt_virt) != 0) | |
782 | return NULL; | |
783 | ||
f80fb3a3 AB |
784 | *size = fdt_totalsize(dt_virt); |
785 | if (*size > MAX_FDT_SIZE) | |
61bd93ce AB |
786 | return NULL; |
787 | ||
f80fb3a3 | 788 | if (offset + *size > SWAPPER_BLOCK_SIZE) |
132233a7 | 789 | create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), dt_virt_base, |
f80fb3a3 | 790 | round_up(offset + *size, SWAPPER_BLOCK_SIZE), prot); |
61bd93ce | 791 | |
f80fb3a3 AB |
792 | return dt_virt; |
793 | } | |
61bd93ce | 794 | |
f80fb3a3 AB |
795 | void *__init fixmap_remap_fdt(phys_addr_t dt_phys) |
796 | { | |
797 | void *dt_virt; | |
798 | int size; | |
799 | ||
800 | dt_virt = __fixmap_remap_fdt(dt_phys, &size, PAGE_KERNEL_RO); | |
801 | if (!dt_virt) | |
802 | return NULL; | |
803 | ||
804 | memblock_reserve(dt_phys, size); | |
61bd93ce AB |
805 | return dt_virt; |
806 | } | |
324420bf AB |
807 | |
808 | int __init arch_ioremap_pud_supported(void) | |
809 | { | |
810 | /* only 4k granule supports level 1 block mappings */ | |
811 | return IS_ENABLED(CONFIG_ARM64_4K_PAGES); | |
812 | } | |
813 | ||
814 | int __init arch_ioremap_pmd_supported(void) | |
815 | { | |
816 | return 1; | |
817 | } | |
818 | ||
819 | int pud_set_huge(pud_t *pud, phys_addr_t phys, pgprot_t prot) | |
820 | { | |
821 | BUG_ON(phys & ~PUD_MASK); | |
822 | set_pud(pud, __pud(phys | PUD_TYPE_SECT | pgprot_val(mk_sect_prot(prot)))); | |
823 | return 1; | |
824 | } | |
825 | ||
826 | int pmd_set_huge(pmd_t *pmd, phys_addr_t phys, pgprot_t prot) | |
827 | { | |
828 | BUG_ON(phys & ~PMD_MASK); | |
829 | set_pmd(pmd, __pmd(phys | PMD_TYPE_SECT | pgprot_val(mk_sect_prot(prot)))); | |
830 | return 1; | |
831 | } | |
832 | ||
833 | int pud_clear_huge(pud_t *pud) | |
834 | { | |
835 | if (!pud_sect(*pud)) | |
836 | return 0; | |
837 | pud_clear(pud); | |
838 | return 1; | |
839 | } | |
840 | ||
841 | int pmd_clear_huge(pmd_t *pmd) | |
842 | { | |
843 | if (!pmd_sect(*pmd)) | |
844 | return 0; | |
845 | pmd_clear(pmd); | |
846 | return 1; | |
847 | } |