pgtable.h 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587
  1. #ifndef _ASM_X86_PGTABLE_H
  2. #define _ASM_X86_PGTABLE_H
  3. #include <asm/page.h>
  4. #include <asm/pgtable_types.h>
  5. /*
  6. * Macro to mark a page protection value as UC-
  7. */
  8. #define pgprot_noncached(prot) \
  9. ((boot_cpu_data.x86 > 3) \
  10. ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS)) \
  11. : (prot))
  12. #ifndef __ASSEMBLY__
  13. /*
  14. * ZERO_PAGE is a global shared page that is always zero: used
  15. * for zero-mapped memory areas etc..
  16. */
  17. extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  18. #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  19. extern spinlock_t pgd_lock;
  20. extern struct list_head pgd_list;
  21. /*
  22. * The following only work if pte_present() is true.
  23. * Undefined behaviour if not..
  24. */
  25. static inline int pte_dirty(pte_t pte)
  26. {
  27. return pte_flags(pte) & _PAGE_DIRTY;
  28. }
  29. static inline int pte_young(pte_t pte)
  30. {
  31. return pte_flags(pte) & _PAGE_ACCESSED;
  32. }
  33. static inline int pte_write(pte_t pte)
  34. {
  35. return pte_flags(pte) & _PAGE_RW;
  36. }
  37. static inline int pte_file(pte_t pte)
  38. {
  39. return pte_flags(pte) & _PAGE_FILE;
  40. }
  41. static inline int pte_huge(pte_t pte)
  42. {
  43. return pte_flags(pte) & _PAGE_PSE;
  44. }
  45. static inline int pte_global(pte_t pte)
  46. {
  47. return pte_flags(pte) & _PAGE_GLOBAL;
  48. }
  49. static inline int pte_exec(pte_t pte)
  50. {
  51. return !(pte_flags(pte) & _PAGE_NX);
  52. }
  53. static inline int pte_special(pte_t pte)
  54. {
  55. return pte_flags(pte) & _PAGE_SPECIAL;
  56. }
  57. static inline unsigned long pte_pfn(pte_t pte)
  58. {
  59. return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
  60. }
  61. #define pte_page(pte) pfn_to_page(pte_pfn(pte))
  62. static inline int pmd_large(pmd_t pte)
  63. {
  64. return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
  65. (_PAGE_PSE | _PAGE_PRESENT);
  66. }
  67. static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
  68. {
  69. pteval_t v = native_pte_val(pte);
  70. return native_make_pte(v | set);
  71. }
  72. static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
  73. {
  74. pteval_t v = native_pte_val(pte);
  75. return native_make_pte(v & ~clear);
  76. }
  77. static inline pte_t pte_mkclean(pte_t pte)
  78. {
  79. return pte_clear_flags(pte, _PAGE_DIRTY);
  80. }
  81. static inline pte_t pte_mkold(pte_t pte)
  82. {
  83. return pte_clear_flags(pte, _PAGE_ACCESSED);
  84. }
  85. static inline pte_t pte_wrprotect(pte_t pte)
  86. {
  87. return pte_clear_flags(pte, _PAGE_RW);
  88. }
  89. static inline pte_t pte_mkexec(pte_t pte)
  90. {
  91. return pte_clear_flags(pte, _PAGE_NX);
  92. }
  93. static inline pte_t pte_mkdirty(pte_t pte)
  94. {
  95. return pte_set_flags(pte, _PAGE_DIRTY);
  96. }
  97. static inline pte_t pte_mkyoung(pte_t pte)
  98. {
  99. return pte_set_flags(pte, _PAGE_ACCESSED);
  100. }
  101. static inline pte_t pte_mkwrite(pte_t pte)
  102. {
  103. return pte_set_flags(pte, _PAGE_RW);
  104. }
  105. static inline pte_t pte_mkhuge(pte_t pte)
  106. {
  107. return pte_set_flags(pte, _PAGE_PSE);
  108. }
  109. static inline pte_t pte_clrhuge(pte_t pte)
  110. {
  111. return pte_clear_flags(pte, _PAGE_PSE);
  112. }
  113. static inline pte_t pte_mkglobal(pte_t pte)
  114. {
  115. return pte_set_flags(pte, _PAGE_GLOBAL);
  116. }
  117. static inline pte_t pte_clrglobal(pte_t pte)
  118. {
  119. return pte_clear_flags(pte, _PAGE_GLOBAL);
  120. }
  121. static inline pte_t pte_mkspecial(pte_t pte)
  122. {
  123. return pte_set_flags(pte, _PAGE_SPECIAL);
  124. }
  125. static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
  126. {
  127. return __pte((((phys_addr_t)page_nr << PAGE_SHIFT) |
  128. pgprot_val(pgprot)) & __supported_pte_mask);
  129. }
  130. static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
  131. {
  132. return __pmd((((phys_addr_t)page_nr << PAGE_SHIFT) |
  133. pgprot_val(pgprot)) & __supported_pte_mask);
  134. }
  135. static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
  136. {
  137. pteval_t val = pte_val(pte);
  138. /*
  139. * Chop off the NX bit (if present), and add the NX portion of
  140. * the newprot (if present):
  141. */
  142. val &= _PAGE_CHG_MASK;
  143. val |= pgprot_val(newprot) & (~_PAGE_CHG_MASK) & __supported_pte_mask;
  144. return __pte(val);
  145. }
  146. /* mprotect needs to preserve PAT bits when updating vm_page_prot */
  147. #define pgprot_modify pgprot_modify
  148. static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
  149. {
  150. pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
  151. pgprotval_t addbits = pgprot_val(newprot);
  152. return __pgprot(preservebits | addbits);
  153. }
  154. #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
  155. #define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask)
  156. static inline int is_new_memtype_allowed(unsigned long flags,
  157. unsigned long new_flags)
  158. {
  159. /*
  160. * Certain new memtypes are not allowed with certain
  161. * requested memtype:
  162. * - request is uncached, return cannot be write-back
  163. * - request is write-combine, return cannot be write-back
  164. */
  165. if ((flags == _PAGE_CACHE_UC_MINUS &&
  166. new_flags == _PAGE_CACHE_WB) ||
  167. (flags == _PAGE_CACHE_WC &&
  168. new_flags == _PAGE_CACHE_WB)) {
  169. return 0;
  170. }
  171. return 1;
  172. }
  173. #ifdef CONFIG_PARAVIRT
  174. #include <asm/paravirt.h>
  175. #else /* !CONFIG_PARAVIRT */
  176. #define set_pte(ptep, pte) native_set_pte(ptep, pte)
  177. #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  178. #define set_pte_present(mm, addr, ptep, pte) \
  179. native_set_pte_present(mm, addr, ptep, pte)
  180. #define set_pte_atomic(ptep, pte) \
  181. native_set_pte_atomic(ptep, pte)
  182. #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
  183. #ifndef __PAGETABLE_PUD_FOLDED
  184. #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
  185. #define pgd_clear(pgd) native_pgd_clear(pgd)
  186. #endif
  187. #ifndef set_pud
  188. # define set_pud(pudp, pud) native_set_pud(pudp, pud)
  189. #endif
  190. #ifndef __PAGETABLE_PMD_FOLDED
  191. #define pud_clear(pud) native_pud_clear(pud)
  192. #endif
  193. #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
  194. #define pmd_clear(pmd) native_pmd_clear(pmd)
  195. #define pte_update(mm, addr, ptep) do { } while (0)
  196. #define pte_update_defer(mm, addr, ptep) do { } while (0)
  197. static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
  198. {
  199. native_pagetable_setup_start(base);
  200. }
  201. static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
  202. {
  203. native_pagetable_setup_done(base);
  204. }
  205. #endif /* CONFIG_PARAVIRT */
  206. #endif /* __ASSEMBLY__ */
  207. #ifdef CONFIG_X86_32
  208. # include "pgtable_32.h"
  209. #else
  210. # include "pgtable_64.h"
  211. #endif
  212. #ifndef __ASSEMBLY__
  213. #include <linux/mm_types.h>
  214. static inline int pte_none(pte_t pte)
  215. {
  216. return !pte.pte;
  217. }
  218. #define __HAVE_ARCH_PTE_SAME
  219. static inline int pte_same(pte_t a, pte_t b)
  220. {
  221. return a.pte == b.pte;
  222. }
  223. static inline int pte_present(pte_t a)
  224. {
  225. return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
  226. }
  227. static inline int pmd_present(pmd_t pmd)
  228. {
  229. return pmd_flags(pmd) & _PAGE_PRESENT;
  230. }
  231. static inline int pmd_none(pmd_t pmd)
  232. {
  233. /* Only check low word on 32-bit platforms, since it might be
  234. out of sync with upper half. */
  235. return (unsigned long)native_pmd_val(pmd) == 0;
  236. }
  237. static inline unsigned long pmd_page_vaddr(pmd_t pmd)
  238. {
  239. return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
  240. }
  241. /*
  242. * Currently stuck as a macro due to indirect forward reference to
  243. * linux/mmzone.h's __section_mem_map_addr() definition:
  244. */
  245. #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
  246. /*
  247. * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
  248. *
  249. * this macro returns the index of the entry in the pmd page which would
  250. * control the given virtual address
  251. */
  252. static inline unsigned pmd_index(unsigned long address)
  253. {
  254. return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
  255. }
  256. /*
  257. * Conversion functions: convert a page and protection to a page entry,
  258. * and a page entry and page directory to the page they refer to.
  259. *
  260. * (Currently stuck as a macro because of indirect forward reference
  261. * to linux/mm.h:page_to_nid())
  262. */
  263. #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
  264. /*
  265. * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
  266. *
  267. * this function returns the index of the entry in the pte page which would
  268. * control the given virtual address
  269. */
  270. static inline unsigned pte_index(unsigned long address)
  271. {
  272. return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
  273. }
  274. static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
  275. {
  276. return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
  277. }
  278. static inline int pmd_bad(pmd_t pmd)
  279. {
  280. return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
  281. }
  282. static inline unsigned long pages_to_mb(unsigned long npg)
  283. {
  284. return npg >> (20 - PAGE_SHIFT);
  285. }
  286. #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
  287. remap_pfn_range(vma, vaddr, pfn, size, prot)
  288. #if PAGETABLE_LEVELS == 2
  289. static inline int pud_large(pud_t pud)
  290. {
  291. return 0;
  292. }
  293. #endif
  294. #if PAGETABLE_LEVELS > 2
  295. static inline int pud_none(pud_t pud)
  296. {
  297. return native_pud_val(pud) == 0;
  298. }
  299. static inline int pud_present(pud_t pud)
  300. {
  301. return pud_flags(pud) & _PAGE_PRESENT;
  302. }
  303. static inline unsigned long pud_page_vaddr(pud_t pud)
  304. {
  305. return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
  306. }
  307. /*
  308. * Currently stuck as a macro due to indirect forward reference to
  309. * linux/mmzone.h's __section_mem_map_addr() definition:
  310. */
  311. #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
  312. /* Find an entry in the second-level page table.. */
  313. static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
  314. {
  315. return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
  316. }
  317. static inline unsigned long pmd_pfn(pmd_t pmd)
  318. {
  319. return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
  320. }
  321. static inline int pud_large(pud_t pud)
  322. {
  323. return (pud_flags(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
  324. (_PAGE_PSE | _PAGE_PRESENT);
  325. }
  326. static inline int pud_bad(pud_t pud)
  327. {
  328. return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
  329. }
  330. #endif /* PAGETABLE_LEVELS > 2 */
  331. #if PAGETABLE_LEVELS > 3
  332. static inline int pgd_present(pgd_t pgd)
  333. {
  334. return pgd_flags(pgd) & _PAGE_PRESENT;
  335. }
  336. static inline unsigned long pgd_page_vaddr(pgd_t pgd)
  337. {
  338. return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
  339. }
  340. /*
  341. * Currently stuck as a macro due to indirect forward reference to
  342. * linux/mmzone.h's __section_mem_map_addr() definition:
  343. */
  344. #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
  345. /* to find an entry in a page-table-directory. */
  346. static inline unsigned pud_index(unsigned long address)
  347. {
  348. return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
  349. }
  350. static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
  351. {
  352. return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
  353. }
  354. static inline int pgd_bad(pgd_t pgd)
  355. {
  356. return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
  357. }
  358. static inline int pgd_none(pgd_t pgd)
  359. {
  360. return !native_pgd_val(pgd);
  361. }
  362. #endif /* PAGETABLE_LEVELS > 3 */
  363. #endif /* __ASSEMBLY__ */
  364. /*
  365. * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
  366. *
  367. * this macro returns the index of the entry in the pgd page which would
  368. * control the given virtual address
  369. */
  370. #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
  371. /*
  372. * pgd_offset() returns a (pgd_t *)
  373. * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
  374. */
  375. #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
  376. /*
  377. * a shortcut which implies the use of the kernel's pgd, instead
  378. * of a process's
  379. */
  380. #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
  381. #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
  382. #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
  383. #ifndef __ASSEMBLY__
  384. /* local pte updates need not use xchg for locking */
  385. static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
  386. {
  387. pte_t res = *ptep;
  388. /* Pure native function needs no input for mm, addr */
  389. native_pte_clear(NULL, 0, ptep);
  390. return res;
  391. }
  392. static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
  393. pte_t *ptep , pte_t pte)
  394. {
  395. native_set_pte(ptep, pte);
  396. }
  397. #ifndef CONFIG_PARAVIRT
  398. /*
  399. * Rules for using pte_update - it must be called after any PTE update which
  400. * has not been done using the set_pte / clear_pte interfaces. It is used by
  401. * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
  402. * updates should either be sets, clears, or set_pte_atomic for P->P
  403. * transitions, which means this hook should only be called for user PTEs.
  404. * This hook implies a P->P protection or access change has taken place, which
  405. * requires a subsequent TLB flush. The notification can optionally be delayed
  406. * until the TLB flush event by using the pte_update_defer form of the
  407. * interface, but care must be taken to assure that the flush happens while
  408. * still holding the same page table lock so that the shadow and primary pages
  409. * do not become out of sync on SMP.
  410. */
  411. #define pte_update(mm, addr, ptep) do { } while (0)
  412. #define pte_update_defer(mm, addr, ptep) do { } while (0)
  413. #endif
  414. /*
  415. * We only update the dirty/accessed state if we set
  416. * the dirty bit by hand in the kernel, since the hardware
  417. * will do the accessed bit for us, and we don't want to
  418. * race with other CPU's that might be updating the dirty
  419. * bit at the same time.
  420. */
  421. struct vm_area_struct;
  422. #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
  423. extern int ptep_set_access_flags(struct vm_area_struct *vma,
  424. unsigned long address, pte_t *ptep,
  425. pte_t entry, int dirty);
  426. #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
  427. extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
  428. unsigned long addr, pte_t *ptep);
  429. #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
  430. extern int ptep_clear_flush_young(struct vm_area_struct *vma,
  431. unsigned long address, pte_t *ptep);
  432. #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
  433. static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
  434. pte_t *ptep)
  435. {
  436. pte_t pte = native_ptep_get_and_clear(ptep);
  437. pte_update(mm, addr, ptep);
  438. return pte;
  439. }
  440. #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
  441. static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
  442. unsigned long addr, pte_t *ptep,
  443. int full)
  444. {
  445. pte_t pte;
  446. if (full) {
  447. /*
  448. * Full address destruction in progress; paravirt does not
  449. * care about updates and native needs no locking
  450. */
  451. pte = native_local_ptep_get_and_clear(ptep);
  452. } else {
  453. pte = ptep_get_and_clear(mm, addr, ptep);
  454. }
  455. return pte;
  456. }
  457. #define __HAVE_ARCH_PTEP_SET_WRPROTECT
  458. static inline void ptep_set_wrprotect(struct mm_struct *mm,
  459. unsigned long addr, pte_t *ptep)
  460. {
  461. clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
  462. pte_update(mm, addr, ptep);
  463. }
  464. /*
  465. * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
  466. *
  467. * dst - pointer to pgd range anwhere on a pgd page
  468. * src - ""
  469. * count - the number of pgds to copy.
  470. *
  471. * dst and src can be on the same page, but the range must not overlap,
  472. * and must not cross a page boundary.
  473. */
  474. static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
  475. {
  476. memcpy(dst, src, count * sizeof(pgd_t));
  477. }
  478. #include <asm-generic/pgtable.h>
  479. #endif /* __ASSEMBLY__ */
  480. #endif /* _ASM_X86_PGTABLE_H */