]>
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 | ||
5a9e3e15 | 20 | #include <linux/cache.h> |
c1cc1552 CM |
21 | #include <linux/export.h> |
22 | #include <linux/kernel.h> | |
23 | #include <linux/errno.h> | |
24 | #include <linux/init.h> | |
61bd93ce | 25 | #include <linux/libfdt.h> |
c1cc1552 CM |
26 | #include <linux/mman.h> |
27 | #include <linux/nodemask.h> | |
28 | #include <linux/memblock.h> | |
29 | #include <linux/fs.h> | |
2475ff9d | 30 | #include <linux/io.h> |
2077be67 | 31 | #include <linux/mm.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 | 43 | #include <asm/mmu_context.h> |
1404d6f1 | 44 | #include <asm/ptdump.h> |
c1cc1552 | 45 | |
dd006da2 AB |
46 | u64 idmap_t0sz = TCR_T0SZ(VA_BITS); |
47 | ||
5a9e3e15 | 48 | u64 kimage_voffset __ro_after_init; |
a7f8de16 AB |
49 | EXPORT_SYMBOL(kimage_voffset); |
50 | ||
c1cc1552 CM |
51 | /* |
52 | * Empty_zero_page is a special page that is used for zero-initialized data | |
53 | * and COW. | |
54 | */ | |
5227cfa7 | 55 | unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)] __page_aligned_bss; |
c1cc1552 CM |
56 | EXPORT_SYMBOL(empty_zero_page); |
57 | ||
f9040773 AB |
58 | static pte_t bm_pte[PTRS_PER_PTE] __page_aligned_bss; |
59 | static pmd_t bm_pmd[PTRS_PER_PMD] __page_aligned_bss __maybe_unused; | |
60 | static pud_t bm_pud[PTRS_PER_PUD] __page_aligned_bss __maybe_unused; | |
61 | ||
c1cc1552 CM |
62 | pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
63 | unsigned long size, pgprot_t vma_prot) | |
64 | { | |
65 | if (!pfn_valid(pfn)) | |
66 | return pgprot_noncached(vma_prot); | |
67 | else if (file->f_flags & O_SYNC) | |
68 | return pgprot_writecombine(vma_prot); | |
69 | return vma_prot; | |
70 | } | |
71 | EXPORT_SYMBOL(phys_mem_access_prot); | |
72 | ||
f4710445 | 73 | static phys_addr_t __init early_pgtable_alloc(void) |
c1cc1552 | 74 | { |
7142392d SP |
75 | phys_addr_t phys; |
76 | void *ptr; | |
77 | ||
21ab99c2 | 78 | phys = memblock_alloc(PAGE_SIZE, PAGE_SIZE); |
f4710445 MR |
79 | |
80 | /* | |
81 | * The FIX_{PGD,PUD,PMD} slots may be in active use, but the FIX_PTE | |
82 | * slot will be free, so we can (ab)use the FIX_PTE slot to initialise | |
83 | * any level of table. | |
84 | */ | |
85 | ptr = pte_set_fixmap(phys); | |
86 | ||
21ab99c2 MR |
87 | memset(ptr, 0, PAGE_SIZE); |
88 | ||
f4710445 MR |
89 | /* |
90 | * Implicit barriers also ensure the zeroed page is visible to the page | |
91 | * table walker | |
92 | */ | |
93 | pte_clear_fixmap(); | |
94 | ||
95 | return phys; | |
c1cc1552 CM |
96 | } |
97 | ||
e98216b5 AB |
98 | static bool pgattr_change_is_safe(u64 old, u64 new) |
99 | { | |
100 | /* | |
101 | * The following mapping attributes may be updated in live | |
102 | * kernel mappings without the need for break-before-make. | |
103 | */ | |
104 | static const pteval_t mask = PTE_PXN | PTE_RDONLY | PTE_WRITE; | |
105 | ||
106 | return old == 0 || new == 0 || ((old ^ new) & ~mask) == 0; | |
107 | } | |
108 | ||
da141706 | 109 | static void alloc_init_pte(pmd_t *pmd, unsigned long addr, |
667c2759 | 110 | unsigned long end, unsigned long pfn, |
da141706 | 111 | pgprot_t prot, |
0bfc445d AB |
112 | phys_addr_t (*pgtable_alloc)(void), |
113 | bool page_mappings_only) | |
c1cc1552 | 114 | { |
0bfc445d | 115 | pgprot_t __prot = prot; |
c1cc1552 CM |
116 | pte_t *pte; |
117 | ||
4133af6c CM |
118 | BUG_ON(pmd_sect(*pmd)); |
119 | if (pmd_none(*pmd)) { | |
132233a7 LA |
120 | phys_addr_t pte_phys; |
121 | BUG_ON(!pgtable_alloc); | |
122 | pte_phys = pgtable_alloc(); | |
f4710445 | 123 | pte = pte_set_fixmap(pte_phys); |
f4710445 | 124 | __pmd_populate(pmd, pte_phys, PMD_TYPE_TABLE); |
f4710445 | 125 | pte_clear_fixmap(); |
c1cc1552 | 126 | } |
a1c76574 | 127 | BUG_ON(pmd_bad(*pmd)); |
c1cc1552 | 128 | |
f4710445 | 129 | pte = pte_set_fixmap_offset(pmd, addr); |
c1cc1552 | 130 | do { |
e98216b5 AB |
131 | pte_t old_pte = *pte; |
132 | ||
0bfc445d AB |
133 | /* |
134 | * Set the contiguous bit for the subsequent group of PTEs if | |
135 | * its size and alignment are appropriate. | |
136 | */ | |
137 | if (((addr | PFN_PHYS(pfn)) & ~CONT_PTE_MASK) == 0) { | |
138 | if (end - addr >= CONT_PTE_SIZE && !page_mappings_only) | |
139 | __prot = __pgprot(pgprot_val(prot) | PTE_CONT); | |
140 | else | |
141 | __prot = prot; | |
142 | } | |
143 | ||
144 | set_pte(pte, pfn_pte(pfn, __prot)); | |
667c2759 | 145 | pfn++; |
e98216b5 AB |
146 | |
147 | /* | |
148 | * After the PTE entry has been populated once, we | |
149 | * only allow updates to the permission attributes. | |
150 | */ | |
151 | BUG_ON(!pgattr_change_is_safe(pte_val(old_pte), pte_val(*pte))); | |
152 | ||
667c2759 | 153 | } while (pte++, addr += PAGE_SIZE, addr != end); |
f4710445 MR |
154 | |
155 | pte_clear_fixmap(); | |
c1cc1552 CM |
156 | } |
157 | ||
11509a30 | 158 | static void alloc_init_pmd(pud_t *pud, unsigned long addr, unsigned long end, |
da141706 | 159 | phys_addr_t phys, pgprot_t prot, |
53e1b329 | 160 | phys_addr_t (*pgtable_alloc)(void), |
f14c66ce | 161 | bool page_mappings_only) |
c1cc1552 | 162 | { |
0bfc445d | 163 | pgprot_t __prot = prot; |
c1cc1552 CM |
164 | pmd_t *pmd; |
165 | unsigned long next; | |
166 | ||
167 | /* | |
168 | * Check for initial section mappings in the pgd/pud and remove them. | |
169 | */ | |
4133af6c CM |
170 | BUG_ON(pud_sect(*pud)); |
171 | if (pud_none(*pud)) { | |
132233a7 LA |
172 | phys_addr_t pmd_phys; |
173 | BUG_ON(!pgtable_alloc); | |
174 | pmd_phys = pgtable_alloc(); | |
f4710445 | 175 | pmd = pmd_set_fixmap(pmd_phys); |
f4710445 | 176 | __pud_populate(pud, pmd_phys, PUD_TYPE_TABLE); |
f4710445 | 177 | pmd_clear_fixmap(); |
c1cc1552 | 178 | } |
a1c76574 | 179 | BUG_ON(pud_bad(*pud)); |
c1cc1552 | 180 | |
f4710445 | 181 | pmd = pmd_set_fixmap_offset(pud, addr); |
c1cc1552 | 182 | do { |
e98216b5 AB |
183 | pmd_t old_pmd = *pmd; |
184 | ||
c1cc1552 | 185 | next = pmd_addr_end(addr, end); |
e98216b5 | 186 | |
c1cc1552 | 187 | /* try section mapping first */ |
83863f25 | 188 | if (((addr | next | phys) & ~SECTION_MASK) == 0 && |
f14c66ce | 189 | !page_mappings_only) { |
0bfc445d AB |
190 | /* |
191 | * Set the contiguous bit for the subsequent group of | |
192 | * PMDs if its size and alignment are appropriate. | |
193 | */ | |
194 | if (((addr | phys) & ~CONT_PMD_MASK) == 0) { | |
195 | if (end - addr >= CONT_PMD_SIZE) | |
196 | __prot = __pgprot(pgprot_val(prot) | | |
197 | PTE_CONT); | |
198 | else | |
199 | __prot = prot; | |
200 | } | |
201 | pmd_set_huge(pmd, phys, __prot); | |
e98216b5 | 202 | |
a55f9929 | 203 | /* |
e98216b5 AB |
204 | * After the PMD entry has been populated once, we |
205 | * only allow updates to the permission attributes. | |
a55f9929 | 206 | */ |
e98216b5 AB |
207 | BUG_ON(!pgattr_change_is_safe(pmd_val(old_pmd), |
208 | pmd_val(*pmd))); | |
a55f9929 | 209 | } else { |
667c2759 | 210 | alloc_init_pte(pmd, addr, next, __phys_to_pfn(phys), |
0bfc445d AB |
211 | prot, pgtable_alloc, |
212 | page_mappings_only); | |
e98216b5 AB |
213 | |
214 | BUG_ON(pmd_val(old_pmd) != 0 && | |
215 | pmd_val(old_pmd) != pmd_val(*pmd)); | |
a55f9929 | 216 | } |
c1cc1552 CM |
217 | phys += next - addr; |
218 | } while (pmd++, addr = next, addr != end); | |
f4710445 MR |
219 | |
220 | pmd_clear_fixmap(); | |
c1cc1552 CM |
221 | } |
222 | ||
da141706 LA |
223 | static inline bool use_1G_block(unsigned long addr, unsigned long next, |
224 | unsigned long phys) | |
225 | { | |
226 | if (PAGE_SHIFT != 12) | |
227 | return false; | |
228 | ||
229 | if (((addr | next | phys) & ~PUD_MASK) != 0) | |
230 | return false; | |
231 | ||
232 | return true; | |
233 | } | |
234 | ||
11509a30 | 235 | static void alloc_init_pud(pgd_t *pgd, unsigned long addr, unsigned long end, |
da141706 | 236 | phys_addr_t phys, pgprot_t prot, |
53e1b329 | 237 | phys_addr_t (*pgtable_alloc)(void), |
f14c66ce | 238 | bool page_mappings_only) |
c1cc1552 | 239 | { |
c79b954b | 240 | pud_t *pud; |
c1cc1552 CM |
241 | unsigned long next; |
242 | ||
c79b954b | 243 | if (pgd_none(*pgd)) { |
132233a7 LA |
244 | phys_addr_t pud_phys; |
245 | BUG_ON(!pgtable_alloc); | |
246 | pud_phys = pgtable_alloc(); | |
f4710445 | 247 | __pgd_populate(pgd, pud_phys, PUD_TYPE_TABLE); |
c79b954b JL |
248 | } |
249 | BUG_ON(pgd_bad(*pgd)); | |
250 | ||
f4710445 | 251 | pud = pud_set_fixmap_offset(pgd, addr); |
c1cc1552 | 252 | do { |
e98216b5 AB |
253 | pud_t old_pud = *pud; |
254 | ||
c1cc1552 | 255 | next = pud_addr_end(addr, end); |
206a2a73 SC |
256 | |
257 | /* | |
258 | * For 4K granule only, attempt to put down a 1GB block | |
259 | */ | |
f14c66ce | 260 | if (use_1G_block(addr, next, phys) && !page_mappings_only) { |
c661cb1c | 261 | pud_set_huge(pud, phys, prot); |
206a2a73 SC |
262 | |
263 | /* | |
e98216b5 AB |
264 | * After the PUD entry has been populated once, we |
265 | * only allow updates to the permission attributes. | |
206a2a73 | 266 | */ |
e98216b5 AB |
267 | BUG_ON(!pgattr_change_is_safe(pud_val(old_pud), |
268 | pud_val(*pud))); | |
206a2a73 | 269 | } else { |
11509a30 | 270 | alloc_init_pmd(pud, addr, next, phys, prot, |
f14c66ce | 271 | pgtable_alloc, page_mappings_only); |
e98216b5 AB |
272 | |
273 | BUG_ON(pud_val(old_pud) != 0 && | |
274 | pud_val(old_pud) != pud_val(*pud)); | |
206a2a73 | 275 | } |
c1cc1552 CM |
276 | phys += next - addr; |
277 | } while (pud++, addr = next, addr != end); | |
f4710445 MR |
278 | |
279 | pud_clear_fixmap(); | |
c1cc1552 CM |
280 | } |
281 | ||
40f87d31 AB |
282 | static void __create_pgd_mapping(pgd_t *pgdir, phys_addr_t phys, |
283 | unsigned long virt, phys_addr_t size, | |
284 | pgprot_t prot, | |
285 | phys_addr_t (*pgtable_alloc)(void), | |
f14c66ce | 286 | bool page_mappings_only) |
c1cc1552 CM |
287 | { |
288 | unsigned long addr, length, end, next; | |
40f87d31 | 289 | pgd_t *pgd = pgd_offset_raw(pgdir, virt); |
c1cc1552 | 290 | |
cc5d2b3b MR |
291 | /* |
292 | * If the virtual and physical address don't have the same offset | |
293 | * within a page, we cannot map the region as the caller expects. | |
294 | */ | |
295 | if (WARN_ON((phys ^ virt) & ~PAGE_MASK)) | |
296 | return; | |
297 | ||
9c4e08a3 | 298 | phys &= PAGE_MASK; |
c1cc1552 CM |
299 | addr = virt & PAGE_MASK; |
300 | length = PAGE_ALIGN(size + (virt & ~PAGE_MASK)); | |
301 | ||
c1cc1552 CM |
302 | end = addr + length; |
303 | do { | |
304 | next = pgd_addr_end(addr, end); | |
53e1b329 | 305 | alloc_init_pud(pgd, addr, next, phys, prot, pgtable_alloc, |
f14c66ce | 306 | page_mappings_only); |
c1cc1552 CM |
307 | phys += next - addr; |
308 | } while (pgd++, addr = next, addr != end); | |
309 | } | |
310 | ||
1378dc3d | 311 | static phys_addr_t pgd_pgtable_alloc(void) |
da141706 | 312 | { |
21ab99c2 | 313 | void *ptr = (void *)__get_free_page(PGALLOC_GFP); |
1378dc3d AB |
314 | if (!ptr || !pgtable_page_ctor(virt_to_page(ptr))) |
315 | BUG(); | |
21ab99c2 MR |
316 | |
317 | /* Ensure the zeroed page is visible to the page table walker */ | |
318 | dsb(ishst); | |
f4710445 | 319 | return __pa(ptr); |
da141706 LA |
320 | } |
321 | ||
132233a7 LA |
322 | /* |
323 | * This function can only be used to modify existing table entries, | |
324 | * without allocating new levels of table. Note that this permits the | |
325 | * creation of new section or page entries. | |
326 | */ | |
327 | static void __init create_mapping_noalloc(phys_addr_t phys, unsigned long virt, | |
da141706 | 328 | phys_addr_t size, pgprot_t prot) |
d7ecbddf MS |
329 | { |
330 | if (virt < VMALLOC_START) { | |
331 | pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n", | |
332 | &phys, virt); | |
333 | return; | |
334 | } | |
f14c66ce | 335 | __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL, false); |
d7ecbddf MS |
336 | } |
337 | ||
8ce837ce AB |
338 | void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys, |
339 | unsigned long virt, phys_addr_t size, | |
f14c66ce | 340 | pgprot_t prot, bool page_mappings_only) |
8ce837ce | 341 | { |
1378dc3d AB |
342 | BUG_ON(mm == &init_mm); |
343 | ||
11509a30 | 344 | __create_pgd_mapping(mm->pgd, phys, virt, size, prot, |
f14c66ce | 345 | pgd_pgtable_alloc, page_mappings_only); |
d7ecbddf MS |
346 | } |
347 | ||
da141706 LA |
348 | static void create_mapping_late(phys_addr_t phys, unsigned long virt, |
349 | phys_addr_t size, pgprot_t prot) | |
350 | { | |
351 | if (virt < VMALLOC_START) { | |
352 | pr_warn("BUG: not creating mapping for %pa at 0x%016lx - outside kernel range\n", | |
353 | &phys, virt); | |
354 | return; | |
355 | } | |
356 | ||
11509a30 | 357 | __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, |
f14c66ce | 358 | NULL, debug_pagealloc_enabled()); |
da141706 LA |
359 | } |
360 | ||
068a17a5 | 361 | static void __init __map_memblock(pgd_t *pgd, phys_addr_t start, phys_addr_t end) |
da141706 | 362 | { |
2077be67 LA |
363 | unsigned long kernel_start = __pa_symbol(_text); |
364 | unsigned long kernel_end = __pa_symbol(__init_begin); | |
068a17a5 | 365 | |
da141706 | 366 | /* |
f9040773 AB |
367 | * Take care not to create a writable alias for the |
368 | * read-only text and rodata sections of the kernel image. | |
da141706 | 369 | */ |
068a17a5 | 370 | |
9fdc14c5 | 371 | /* No overlap with the kernel text/rodata */ |
068a17a5 MR |
372 | if (end < kernel_start || start >= kernel_end) { |
373 | __create_pgd_mapping(pgd, start, __phys_to_virt(start), | |
374 | end - start, PAGE_KERNEL, | |
53e1b329 | 375 | early_pgtable_alloc, |
f14c66ce | 376 | debug_pagealloc_enabled()); |
068a17a5 | 377 | return; |
da141706 LA |
378 | } |
379 | ||
068a17a5 | 380 | /* |
9fdc14c5 | 381 | * This block overlaps the kernel text/rodata mappings. |
f9040773 | 382 | * Map the portion(s) which don't overlap. |
068a17a5 MR |
383 | */ |
384 | if (start < kernel_start) | |
385 | __create_pgd_mapping(pgd, start, | |
386 | __phys_to_virt(start), | |
387 | kernel_start - start, PAGE_KERNEL, | |
53e1b329 | 388 | early_pgtable_alloc, |
f14c66ce | 389 | debug_pagealloc_enabled()); |
068a17a5 MR |
390 | if (kernel_end < end) |
391 | __create_pgd_mapping(pgd, kernel_end, | |
392 | __phys_to_virt(kernel_end), | |
393 | end - kernel_end, PAGE_KERNEL, | |
53e1b329 | 394 | early_pgtable_alloc, |
f14c66ce | 395 | debug_pagealloc_enabled()); |
f9040773 AB |
396 | |
397 | /* | |
9fdc14c5 | 398 | * Map the linear alias of the [_text, __init_begin) interval as |
f9040773 AB |
399 | * read-only/non-executable. This makes the contents of the |
400 | * region accessible to subsystems such as hibernate, but | |
401 | * protects it from inadvertent modification or execution. | |
402 | */ | |
403 | __create_pgd_mapping(pgd, kernel_start, __phys_to_virt(kernel_start), | |
404 | kernel_end - kernel_start, PAGE_KERNEL_RO, | |
f14c66ce | 405 | early_pgtable_alloc, debug_pagealloc_enabled()); |
da141706 | 406 | } |
da141706 | 407 | |
068a17a5 | 408 | static void __init map_mem(pgd_t *pgd) |
c1cc1552 CM |
409 | { |
410 | struct memblock_region *reg; | |
f6bc87c3 | 411 | |
c1cc1552 CM |
412 | /* map all the memory banks */ |
413 | for_each_memblock(memory, reg) { | |
414 | phys_addr_t start = reg->base; | |
415 | phys_addr_t end = start + reg->size; | |
416 | ||
417 | if (start >= end) | |
418 | break; | |
68709f45 AB |
419 | if (memblock_is_nomap(reg)) |
420 | continue; | |
c1cc1552 | 421 | |
068a17a5 | 422 | __map_memblock(pgd, start, end); |
c1cc1552 CM |
423 | } |
424 | } | |
425 | ||
da141706 LA |
426 | void mark_rodata_ro(void) |
427 | { | |
2f39b5f9 | 428 | unsigned long section_size; |
f9040773 | 429 | |
9fdc14c5 | 430 | section_size = (unsigned long)_etext - (unsigned long)_text; |
2077be67 | 431 | create_mapping_late(__pa_symbol(_text), (unsigned long)_text, |
2f39b5f9 JL |
432 | section_size, PAGE_KERNEL_ROX); |
433 | /* | |
9fdc14c5 AB |
434 | * mark .rodata as read only. Use __init_begin rather than __end_rodata |
435 | * to cover NOTES and EXCEPTION_TABLE. | |
2f39b5f9 | 436 | */ |
9fdc14c5 | 437 | section_size = (unsigned long)__init_begin - (unsigned long)__start_rodata; |
2077be67 | 438 | create_mapping_late(__pa_symbol(__start_rodata), (unsigned long)__start_rodata, |
2f39b5f9 | 439 | section_size, PAGE_KERNEL_RO); |
e98216b5 AB |
440 | |
441 | /* flush the TLBs after updating live kernel mappings */ | |
442 | flush_tlb_all(); | |
1404d6f1 LA |
443 | |
444 | debug_checkwx(); | |
da141706 | 445 | } |
da141706 | 446 | |
2c09ec06 AB |
447 | static void __init map_kernel_segment(pgd_t *pgd, void *va_start, void *va_end, |
448 | pgprot_t prot, struct vm_struct *vma) | |
068a17a5 | 449 | { |
2077be67 | 450 | phys_addr_t pa_start = __pa_symbol(va_start); |
068a17a5 MR |
451 | unsigned long size = va_end - va_start; |
452 | ||
453 | BUG_ON(!PAGE_ALIGNED(pa_start)); | |
454 | BUG_ON(!PAGE_ALIGNED(size)); | |
455 | ||
456 | __create_pgd_mapping(pgd, pa_start, (unsigned long)va_start, size, prot, | |
f14c66ce | 457 | early_pgtable_alloc, debug_pagealloc_enabled()); |
f9040773 AB |
458 | |
459 | vma->addr = va_start; | |
460 | vma->phys_addr = pa_start; | |
461 | vma->size = size; | |
462 | vma->flags = VM_MAP; | |
463 | vma->caller = __builtin_return_address(0); | |
464 | ||
465 | vm_area_add_early(vma); | |
068a17a5 MR |
466 | } |
467 | ||
468 | /* | |
469 | * Create fine-grained mappings for the kernel. | |
470 | */ | |
471 | static void __init map_kernel(pgd_t *pgd) | |
472 | { | |
2f39b5f9 | 473 | static struct vm_struct vmlinux_text, vmlinux_rodata, vmlinux_init, vmlinux_data; |
068a17a5 | 474 | |
9fdc14c5 AB |
475 | map_kernel_segment(pgd, _text, _etext, PAGE_KERNEL_EXEC, &vmlinux_text); |
476 | map_kernel_segment(pgd, __start_rodata, __init_begin, PAGE_KERNEL, &vmlinux_rodata); | |
2c09ec06 AB |
477 | map_kernel_segment(pgd, __init_begin, __init_end, PAGE_KERNEL_EXEC, |
478 | &vmlinux_init); | |
479 | map_kernel_segment(pgd, _data, _end, PAGE_KERNEL, &vmlinux_data); | |
068a17a5 | 480 | |
f9040773 AB |
481 | if (!pgd_val(*pgd_offset_raw(pgd, FIXADDR_START))) { |
482 | /* | |
483 | * The fixmap falls in a separate pgd to the kernel, and doesn't | |
484 | * live in the carveout for the swapper_pg_dir. We can simply | |
485 | * re-use the existing dir for the fixmap. | |
486 | */ | |
487 | set_pgd(pgd_offset_raw(pgd, FIXADDR_START), | |
488 | *pgd_offset_k(FIXADDR_START)); | |
489 | } else if (CONFIG_PGTABLE_LEVELS > 3) { | |
490 | /* | |
491 | * The fixmap shares its top level pgd entry with the kernel | |
492 | * mapping. This can really only occur when we are running | |
493 | * with 16k/4 levels, so we can simply reuse the pud level | |
494 | * entry instead. | |
495 | */ | |
496 | BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES)); | |
497 | set_pud(pud_set_fixmap_offset(pgd, FIXADDR_START), | |
2077be67 | 498 | __pud(__pa_symbol(bm_pmd) | PUD_TYPE_TABLE)); |
f9040773 AB |
499 | pud_clear_fixmap(); |
500 | } else { | |
501 | BUG(); | |
502 | } | |
068a17a5 MR |
503 | |
504 | kasan_copy_shadow(pgd); | |
505 | } | |
506 | ||
c1cc1552 CM |
507 | /* |
508 | * paging_init() sets up the page tables, initialises the zone memory | |
509 | * maps and sets up the zero page. | |
510 | */ | |
511 | void __init paging_init(void) | |
512 | { | |
068a17a5 MR |
513 | phys_addr_t pgd_phys = early_pgtable_alloc(); |
514 | pgd_t *pgd = pgd_set_fixmap(pgd_phys); | |
515 | ||
516 | map_kernel(pgd); | |
517 | map_mem(pgd); | |
518 | ||
519 | /* | |
520 | * We want to reuse the original swapper_pg_dir so we don't have to | |
521 | * communicate the new address to non-coherent secondaries in | |
522 | * secondary_entry, and so cpu_switch_mm can generate the address with | |
523 | * adrp+add rather than a load from some global variable. | |
524 | * | |
525 | * To do this we need to go via a temporary pgd. | |
526 | */ | |
527 | cpu_replace_ttbr1(__va(pgd_phys)); | |
528 | memcpy(swapper_pg_dir, pgd, PAGE_SIZE); | |
2077be67 | 529 | cpu_replace_ttbr1(lm_alias(swapper_pg_dir)); |
068a17a5 MR |
530 | |
531 | pgd_clear_fixmap(); | |
532 | memblock_free(pgd_phys, PAGE_SIZE); | |
533 | ||
534 | /* | |
535 | * We only reuse the PGD from the swapper_pg_dir, not the pud + pmd | |
536 | * allocated with it. | |
537 | */ | |
2077be67 | 538 | memblock_free(__pa_symbol(swapper_pg_dir) + PAGE_SIZE, |
068a17a5 | 539 | SWAPPER_DIR_SIZE - PAGE_SIZE); |
c1cc1552 CM |
540 | } |
541 | ||
c1cc1552 CM |
542 | /* |
543 | * Check whether a kernel address is valid (derived from arch/x86/). | |
544 | */ | |
545 | int kern_addr_valid(unsigned long addr) | |
546 | { | |
547 | pgd_t *pgd; | |
548 | pud_t *pud; | |
549 | pmd_t *pmd; | |
550 | pte_t *pte; | |
551 | ||
552 | if ((((long)addr) >> VA_BITS) != -1UL) | |
553 | return 0; | |
554 | ||
555 | pgd = pgd_offset_k(addr); | |
556 | if (pgd_none(*pgd)) | |
557 | return 0; | |
558 | ||
559 | pud = pud_offset(pgd, addr); | |
560 | if (pud_none(*pud)) | |
561 | return 0; | |
562 | ||
206a2a73 SC |
563 | if (pud_sect(*pud)) |
564 | return pfn_valid(pud_pfn(*pud)); | |
565 | ||
c1cc1552 CM |
566 | pmd = pmd_offset(pud, addr); |
567 | if (pmd_none(*pmd)) | |
568 | return 0; | |
569 | ||
da6e4cb6 DA |
570 | if (pmd_sect(*pmd)) |
571 | return pfn_valid(pmd_pfn(*pmd)); | |
572 | ||
c1cc1552 CM |
573 | pte = pte_offset_kernel(pmd, addr); |
574 | if (pte_none(*pte)) | |
575 | return 0; | |
576 | ||
577 | return pfn_valid(pte_pfn(*pte)); | |
578 | } | |
579 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
b433dce0 | 580 | #if !ARM64_SWAPPER_USES_SECTION_MAPS |
0aad818b | 581 | int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node) |
c1cc1552 | 582 | { |
0aad818b | 583 | return vmemmap_populate_basepages(start, end, node); |
c1cc1552 | 584 | } |
b433dce0 | 585 | #else /* !ARM64_SWAPPER_USES_SECTION_MAPS */ |
0aad818b | 586 | int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node) |
c1cc1552 | 587 | { |
0aad818b | 588 | unsigned long addr = start; |
c1cc1552 CM |
589 | unsigned long next; |
590 | pgd_t *pgd; | |
591 | pud_t *pud; | |
592 | pmd_t *pmd; | |
593 | ||
594 | do { | |
595 | next = pmd_addr_end(addr, end); | |
596 | ||
597 | pgd = vmemmap_pgd_populate(addr, node); | |
598 | if (!pgd) | |
599 | return -ENOMEM; | |
600 | ||
601 | pud = vmemmap_pud_populate(pgd, addr, node); | |
602 | if (!pud) | |
603 | return -ENOMEM; | |
604 | ||
605 | pmd = pmd_offset(pud, addr); | |
606 | if (pmd_none(*pmd)) { | |
607 | void *p = NULL; | |
608 | ||
609 | p = vmemmap_alloc_block_buf(PMD_SIZE, node); | |
610 | if (!p) | |
611 | return -ENOMEM; | |
612 | ||
a501e324 | 613 | set_pmd(pmd, __pmd(__pa(p) | PROT_SECT_NORMAL)); |
c1cc1552 CM |
614 | } else |
615 | vmemmap_verify((pte_t *)pmd, node, addr, next); | |
616 | } while (addr = next, addr != end); | |
617 | ||
618 | return 0; | |
619 | } | |
620 | #endif /* CONFIG_ARM64_64K_PAGES */ | |
0aad818b | 621 | void vmemmap_free(unsigned long start, unsigned long end) |
0197518c TC |
622 | { |
623 | } | |
c1cc1552 | 624 | #endif /* CONFIG_SPARSEMEM_VMEMMAP */ |
af86e597 | 625 | |
af86e597 LA |
626 | static inline pud_t * fixmap_pud(unsigned long addr) |
627 | { | |
628 | pgd_t *pgd = pgd_offset_k(addr); | |
629 | ||
630 | BUG_ON(pgd_none(*pgd) || pgd_bad(*pgd)); | |
631 | ||
157962f5 | 632 | return pud_offset_kimg(pgd, addr); |
af86e597 LA |
633 | } |
634 | ||
635 | static inline pmd_t * fixmap_pmd(unsigned long addr) | |
636 | { | |
637 | pud_t *pud = fixmap_pud(addr); | |
638 | ||
639 | BUG_ON(pud_none(*pud) || pud_bad(*pud)); | |
640 | ||
157962f5 | 641 | return pmd_offset_kimg(pud, addr); |
af86e597 LA |
642 | } |
643 | ||
644 | static inline pte_t * fixmap_pte(unsigned long addr) | |
645 | { | |
157962f5 | 646 | return &bm_pte[pte_index(addr)]; |
af86e597 LA |
647 | } |
648 | ||
2077be67 LA |
649 | /* |
650 | * The p*d_populate functions call virt_to_phys implicitly so they can't be used | |
651 | * directly on kernel symbols (bm_p*d). This function is called too early to use | |
652 | * lm_alias so __p*d_populate functions must be used to populate with the | |
653 | * physical address from __pa_symbol. | |
654 | */ | |
af86e597 LA |
655 | void __init early_fixmap_init(void) |
656 | { | |
657 | pgd_t *pgd; | |
658 | pud_t *pud; | |
659 | pmd_t *pmd; | |
660 | unsigned long addr = FIXADDR_START; | |
661 | ||
662 | pgd = pgd_offset_k(addr); | |
f80fb3a3 | 663 | if (CONFIG_PGTABLE_LEVELS > 3 && |
2077be67 | 664 | !(pgd_none(*pgd) || pgd_page_paddr(*pgd) == __pa_symbol(bm_pud))) { |
f9040773 AB |
665 | /* |
666 | * We only end up here if the kernel mapping and the fixmap | |
667 | * share the top level pgd entry, which should only happen on | |
668 | * 16k/4 levels configurations. | |
669 | */ | |
670 | BUG_ON(!IS_ENABLED(CONFIG_ARM64_16K_PAGES)); | |
671 | pud = pud_offset_kimg(pgd, addr); | |
672 | } else { | |
2077be67 LA |
673 | if (pgd_none(*pgd)) |
674 | __pgd_populate(pgd, __pa_symbol(bm_pud), PUD_TYPE_TABLE); | |
f9040773 AB |
675 | pud = fixmap_pud(addr); |
676 | } | |
2077be67 LA |
677 | if (pud_none(*pud)) |
678 | __pud_populate(pud, __pa_symbol(bm_pmd), PMD_TYPE_TABLE); | |
157962f5 | 679 | pmd = fixmap_pmd(addr); |
2077be67 | 680 | __pmd_populate(pmd, __pa_symbol(bm_pte), PMD_TYPE_TABLE); |
af86e597 LA |
681 | |
682 | /* | |
683 | * The boot-ioremap range spans multiple pmds, for which | |
157962f5 | 684 | * we are not prepared: |
af86e597 LA |
685 | */ |
686 | BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT) | |
687 | != (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT)); | |
688 | ||
689 | if ((pmd != fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN))) | |
690 | || pmd != fixmap_pmd(fix_to_virt(FIX_BTMAP_END))) { | |
691 | WARN_ON(1); | |
692 | pr_warn("pmd %p != %p, %p\n", | |
693 | pmd, fixmap_pmd(fix_to_virt(FIX_BTMAP_BEGIN)), | |
694 | fixmap_pmd(fix_to_virt(FIX_BTMAP_END))); | |
695 | pr_warn("fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n", | |
696 | fix_to_virt(FIX_BTMAP_BEGIN)); | |
697 | pr_warn("fix_to_virt(FIX_BTMAP_END): %08lx\n", | |
698 | fix_to_virt(FIX_BTMAP_END)); | |
699 | ||
700 | pr_warn("FIX_BTMAP_END: %d\n", FIX_BTMAP_END); | |
701 | pr_warn("FIX_BTMAP_BEGIN: %d\n", FIX_BTMAP_BEGIN); | |
702 | } | |
703 | } | |
704 | ||
705 | void __set_fixmap(enum fixed_addresses idx, | |
706 | phys_addr_t phys, pgprot_t flags) | |
707 | { | |
708 | unsigned long addr = __fix_to_virt(idx); | |
709 | pte_t *pte; | |
710 | ||
b63dbef9 | 711 | BUG_ON(idx <= FIX_HOLE || idx >= __end_of_fixed_addresses); |
af86e597 LA |
712 | |
713 | pte = fixmap_pte(addr); | |
714 | ||
715 | if (pgprot_val(flags)) { | |
716 | set_pte(pte, pfn_pte(phys >> PAGE_SHIFT, flags)); | |
717 | } else { | |
718 | pte_clear(&init_mm, addr, pte); | |
719 | flush_tlb_kernel_range(addr, addr+PAGE_SIZE); | |
720 | } | |
721 | } | |
61bd93ce | 722 | |
f80fb3a3 | 723 | void *__init __fixmap_remap_fdt(phys_addr_t dt_phys, int *size, pgprot_t prot) |
61bd93ce AB |
724 | { |
725 | const u64 dt_virt_base = __fix_to_virt(FIX_FDT); | |
f80fb3a3 | 726 | int offset; |
61bd93ce AB |
727 | void *dt_virt; |
728 | ||
729 | /* | |
730 | * Check whether the physical FDT address is set and meets the minimum | |
731 | * alignment requirement. Since we are relying on MIN_FDT_ALIGN to be | |
04a84810 AB |
732 | * at least 8 bytes so that we can always access the magic and size |
733 | * fields of the FDT header after mapping the first chunk, double check | |
734 | * here if that is indeed the case. | |
61bd93ce AB |
735 | */ |
736 | BUILD_BUG_ON(MIN_FDT_ALIGN < 8); | |
737 | if (!dt_phys || dt_phys % MIN_FDT_ALIGN) | |
738 | return NULL; | |
739 | ||
740 | /* | |
741 | * Make sure that the FDT region can be mapped without the need to | |
742 | * allocate additional translation table pages, so that it is safe | |
132233a7 | 743 | * to call create_mapping_noalloc() this early. |
61bd93ce AB |
744 | * |
745 | * On 64k pages, the FDT will be mapped using PTEs, so we need to | |
746 | * be in the same PMD as the rest of the fixmap. | |
747 | * On 4k pages, we'll use section mappings for the FDT so we only | |
748 | * have to be in the same PUD. | |
749 | */ | |
750 | BUILD_BUG_ON(dt_virt_base % SZ_2M); | |
751 | ||
b433dce0 SP |
752 | BUILD_BUG_ON(__fix_to_virt(FIX_FDT_END) >> SWAPPER_TABLE_SHIFT != |
753 | __fix_to_virt(FIX_BTMAP_BEGIN) >> SWAPPER_TABLE_SHIFT); | |
61bd93ce | 754 | |
b433dce0 | 755 | offset = dt_phys % SWAPPER_BLOCK_SIZE; |
61bd93ce AB |
756 | dt_virt = (void *)dt_virt_base + offset; |
757 | ||
758 | /* map the first chunk so we can read the size from the header */ | |
132233a7 LA |
759 | create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), |
760 | dt_virt_base, SWAPPER_BLOCK_SIZE, prot); | |
61bd93ce | 761 | |
04a84810 | 762 | if (fdt_magic(dt_virt) != FDT_MAGIC) |
61bd93ce AB |
763 | return NULL; |
764 | ||
f80fb3a3 AB |
765 | *size = fdt_totalsize(dt_virt); |
766 | if (*size > MAX_FDT_SIZE) | |
61bd93ce AB |
767 | return NULL; |
768 | ||
f80fb3a3 | 769 | if (offset + *size > SWAPPER_BLOCK_SIZE) |
132233a7 | 770 | create_mapping_noalloc(round_down(dt_phys, SWAPPER_BLOCK_SIZE), dt_virt_base, |
f80fb3a3 | 771 | round_up(offset + *size, SWAPPER_BLOCK_SIZE), prot); |
61bd93ce | 772 | |
f80fb3a3 AB |
773 | return dt_virt; |
774 | } | |
61bd93ce | 775 | |
f80fb3a3 AB |
776 | void *__init fixmap_remap_fdt(phys_addr_t dt_phys) |
777 | { | |
778 | void *dt_virt; | |
779 | int size; | |
780 | ||
781 | dt_virt = __fixmap_remap_fdt(dt_phys, &size, PAGE_KERNEL_RO); | |
782 | if (!dt_virt) | |
783 | return NULL; | |
784 | ||
785 | memblock_reserve(dt_phys, size); | |
61bd93ce AB |
786 | return dt_virt; |
787 | } | |
324420bf AB |
788 | |
789 | int __init arch_ioremap_pud_supported(void) | |
790 | { | |
791 | /* only 4k granule supports level 1 block mappings */ | |
792 | return IS_ENABLED(CONFIG_ARM64_4K_PAGES); | |
793 | } | |
794 | ||
795 | int __init arch_ioremap_pmd_supported(void) | |
796 | { | |
797 | return 1; | |
798 | } | |
799 | ||
800 | int pud_set_huge(pud_t *pud, phys_addr_t phys, pgprot_t prot) | |
801 | { | |
802 | BUG_ON(phys & ~PUD_MASK); | |
803 | set_pud(pud, __pud(phys | PUD_TYPE_SECT | pgprot_val(mk_sect_prot(prot)))); | |
804 | return 1; | |
805 | } | |
806 | ||
807 | int pmd_set_huge(pmd_t *pmd, phys_addr_t phys, pgprot_t prot) | |
808 | { | |
809 | BUG_ON(phys & ~PMD_MASK); | |
810 | set_pmd(pmd, __pmd(phys | PMD_TYPE_SECT | pgprot_val(mk_sect_prot(prot)))); | |
811 | return 1; | |
812 | } | |
813 | ||
814 | int pud_clear_huge(pud_t *pud) | |
815 | { | |
816 | if (!pud_sect(*pud)) | |
817 | return 0; | |
818 | pud_clear(pud); | |
819 | return 1; | |
820 | } | |
821 | ||
822 | int pmd_clear_huge(pmd_t *pmd) | |
823 | { | |
824 | if (!pmd_sect(*pmd)) | |
825 | return 0; | |
826 | pmd_clear(pmd); | |
827 | return 1; | |
828 | } |