pgtable.h 25 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846
  1. /*
  2. * include/asm-s390/pgtable.h
  3. *
  4. * S390 version
  5. * Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
  6. * Author(s): Hartmut Penner (hp@de.ibm.com)
  7. * Ulrich Weigand (weigand@de.ibm.com)
  8. * Martin Schwidefsky (schwidefsky@de.ibm.com)
  9. *
  10. * Derived from "include/asm-i386/pgtable.h"
  11. */
  12. #ifndef _ASM_S390_PGTABLE_H
  13. #define _ASM_S390_PGTABLE_H
  14. #include <asm-generic/4level-fixup.h>
  15. /*
  16. * The Linux memory management assumes a three-level page table setup. For
  17. * s390 31 bit we "fold" the mid level into the top-level page table, so
  18. * that we physically have the same two-level page table as the s390 mmu
  19. * expects in 31 bit mode. For s390 64 bit we use three of the five levels
  20. * the hardware provides (region first and region second tables are not
  21. * used).
  22. *
  23. * The "pgd_xxx()" functions are trivial for a folded two-level
  24. * setup: the pgd is never bad, and a pmd always exists (as it's folded
  25. * into the pgd entry)
  26. *
  27. * This file contains the functions and defines necessary to modify and use
  28. * the S390 page table tree.
  29. */
  30. #ifndef __ASSEMBLY__
  31. #include <linux/mm_types.h>
  32. #include <asm/bug.h>
  33. #include <asm/processor.h>
  34. struct vm_area_struct; /* forward declaration (include/linux/mm.h) */
  35. struct mm_struct;
  36. extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
  37. extern void paging_init(void);
  38. /*
  39. * The S390 doesn't have any external MMU info: the kernel page
  40. * tables contain all the necessary information.
  41. */
  42. #define update_mmu_cache(vma, address, pte) do { } while (0)
  43. /*
  44. * ZERO_PAGE is a global shared page that is always zero: used
  45. * for zero-mapped memory areas etc..
  46. */
  47. extern char empty_zero_page[PAGE_SIZE];
  48. #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  49. #endif /* !__ASSEMBLY__ */
  50. /*
  51. * PMD_SHIFT determines the size of the area a second-level page
  52. * table can map
  53. * PGDIR_SHIFT determines what a third-level page table entry can map
  54. */
  55. #ifndef __s390x__
  56. # define PMD_SHIFT 22
  57. # define PGDIR_SHIFT 22
  58. #else /* __s390x__ */
  59. # define PMD_SHIFT 21
  60. # define PGDIR_SHIFT 31
  61. #endif /* __s390x__ */
  62. #define PMD_SIZE (1UL << PMD_SHIFT)
  63. #define PMD_MASK (~(PMD_SIZE-1))
  64. #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
  65. #define PGDIR_MASK (~(PGDIR_SIZE-1))
  66. /*
  67. * entries per page directory level: the S390 is two-level, so
  68. * we don't really have any PMD directory physically.
  69. * for S390 segment-table entries are combined to one PGD
  70. * that leads to 1024 pte per pgd
  71. */
  72. #ifndef __s390x__
  73. # define PTRS_PER_PTE 1024
  74. # define PTRS_PER_PMD 1
  75. # define PTRS_PER_PGD 512
  76. #else /* __s390x__ */
  77. # define PTRS_PER_PTE 512
  78. # define PTRS_PER_PMD 1024
  79. # define PTRS_PER_PGD 2048
  80. #endif /* __s390x__ */
  81. #define FIRST_USER_ADDRESS 0
  82. #define pte_ERROR(e) \
  83. printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
  84. #define pmd_ERROR(e) \
  85. printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
  86. #define pgd_ERROR(e) \
  87. printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
  88. #ifndef __ASSEMBLY__
  89. /*
  90. * Just any arbitrary offset to the start of the vmalloc VM area: the
  91. * current 8MB value just means that there will be a 8MB "hole" after the
  92. * physical memory until the kernel virtual memory starts. That means that
  93. * any out-of-bounds memory accesses will hopefully be caught.
  94. * The vmalloc() routines leaves a hole of 4kB between each vmalloced
  95. * area for the same reason. ;)
  96. */
  97. extern unsigned long vmalloc_end;
  98. #define VMALLOC_OFFSET (8*1024*1024)
  99. #define VMALLOC_START (((unsigned long) high_memory + VMALLOC_OFFSET) \
  100. & ~(VMALLOC_OFFSET-1))
  101. #define VMALLOC_END vmalloc_end
  102. /*
  103. * We need some free virtual space to be able to do vmalloc.
  104. * VMALLOC_MIN_SIZE defines the minimum size of the vmalloc
  105. * area. On a machine with 2GB memory we make sure that we
  106. * have at least 128MB free space for vmalloc. On a machine
  107. * with 4TB we make sure we have at least 128GB.
  108. */
  109. #ifndef __s390x__
  110. #define VMALLOC_MIN_SIZE 0x8000000UL
  111. #define VMALLOC_END_INIT 0x80000000UL
  112. #else /* __s390x__ */
  113. #define VMALLOC_MIN_SIZE 0x2000000000UL
  114. #define VMALLOC_END_INIT 0x40000000000UL
  115. #endif /* __s390x__ */
  116. /*
  117. * A 31 bit pagetable entry of S390 has following format:
  118. * | PFRA | | OS |
  119. * 0 0IP0
  120. * 00000000001111111111222222222233
  121. * 01234567890123456789012345678901
  122. *
  123. * I Page-Invalid Bit: Page is not available for address-translation
  124. * P Page-Protection Bit: Store access not possible for page
  125. *
  126. * A 31 bit segmenttable entry of S390 has following format:
  127. * | P-table origin | |PTL
  128. * 0 IC
  129. * 00000000001111111111222222222233
  130. * 01234567890123456789012345678901
  131. *
  132. * I Segment-Invalid Bit: Segment is not available for address-translation
  133. * C Common-Segment Bit: Segment is not private (PoP 3-30)
  134. * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
  135. *
  136. * The 31 bit segmenttable origin of S390 has following format:
  137. *
  138. * |S-table origin | | STL |
  139. * X **GPS
  140. * 00000000001111111111222222222233
  141. * 01234567890123456789012345678901
  142. *
  143. * X Space-Switch event:
  144. * G Segment-Invalid Bit: *
  145. * P Private-Space Bit: Segment is not private (PoP 3-30)
  146. * S Storage-Alteration:
  147. * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
  148. *
  149. * A 64 bit pagetable entry of S390 has following format:
  150. * | PFRA |0IP0| OS |
  151. * 0000000000111111111122222222223333333333444444444455555555556666
  152. * 0123456789012345678901234567890123456789012345678901234567890123
  153. *
  154. * I Page-Invalid Bit: Page is not available for address-translation
  155. * P Page-Protection Bit: Store access not possible for page
  156. *
  157. * A 64 bit segmenttable entry of S390 has following format:
  158. * | P-table origin | TT
  159. * 0000000000111111111122222222223333333333444444444455555555556666
  160. * 0123456789012345678901234567890123456789012345678901234567890123
  161. *
  162. * I Segment-Invalid Bit: Segment is not available for address-translation
  163. * C Common-Segment Bit: Segment is not private (PoP 3-30)
  164. * P Page-Protection Bit: Store access not possible for page
  165. * TT Type 00
  166. *
  167. * A 64 bit region table entry of S390 has following format:
  168. * | S-table origin | TF TTTL
  169. * 0000000000111111111122222222223333333333444444444455555555556666
  170. * 0123456789012345678901234567890123456789012345678901234567890123
  171. *
  172. * I Segment-Invalid Bit: Segment is not available for address-translation
  173. * TT Type 01
  174. * TF
  175. * TL Table lenght
  176. *
  177. * The 64 bit regiontable origin of S390 has following format:
  178. * | region table origon | DTTL
  179. * 0000000000111111111122222222223333333333444444444455555555556666
  180. * 0123456789012345678901234567890123456789012345678901234567890123
  181. *
  182. * X Space-Switch event:
  183. * G Segment-Invalid Bit:
  184. * P Private-Space Bit:
  185. * S Storage-Alteration:
  186. * R Real space
  187. * TL Table-Length:
  188. *
  189. * A storage key has the following format:
  190. * | ACC |F|R|C|0|
  191. * 0 3 4 5 6 7
  192. * ACC: access key
  193. * F : fetch protection bit
  194. * R : referenced bit
  195. * C : changed bit
  196. */
  197. /* Hardware bits in the page table entry */
  198. #define _PAGE_RO 0x200 /* HW read-only bit */
  199. #define _PAGE_INVALID 0x400 /* HW invalid bit */
  200. #define _PAGE_SWT 0x001 /* SW pte type bit t */
  201. #define _PAGE_SWX 0x002 /* SW pte type bit x */
  202. /* Six different types of pages. */
  203. #define _PAGE_TYPE_EMPTY 0x400
  204. #define _PAGE_TYPE_NONE 0x401
  205. #define _PAGE_TYPE_SWAP 0x403
  206. #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */
  207. #define _PAGE_TYPE_RO 0x200
  208. #define _PAGE_TYPE_RW 0x000
  209. /*
  210. * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
  211. * pte_none and pte_file to find out the pte type WITHOUT holding the page
  212. * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
  213. * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
  214. * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
  215. * This change is done while holding the lock, but the intermediate step
  216. * of a previously valid pte with the hw invalid bit set can be observed by
  217. * handle_pte_fault. That makes it necessary that all valid pte types with
  218. * the hw invalid bit set must be distinguishable from the four pte types
  219. * empty, none, swap and file.
  220. *
  221. * irxt ipte irxt
  222. * _PAGE_TYPE_EMPTY 1000 -> 1000
  223. * _PAGE_TYPE_NONE 1001 -> 1001
  224. * _PAGE_TYPE_SWAP 1011 -> 1011
  225. * _PAGE_TYPE_FILE 11?1 -> 11?1
  226. * _PAGE_TYPE_RO 0100 -> 1100
  227. * _PAGE_TYPE_RW 0000 -> 1000
  228. *
  229. * pte_none is true for bits combinations 1000, 1100
  230. * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
  231. * pte_file is true for bits combinations 1101, 1111
  232. * swap pte is 1011 and 0001, 0011, 0101, 0111, 1010 and 1110 are invalid.
  233. */
  234. #ifndef __s390x__
  235. /* Bits in the segment table entry */
  236. #define _PAGE_TABLE_LEN 0xf /* only full page-tables */
  237. #define _PAGE_TABLE_COM 0x10 /* common page-table */
  238. #define _PAGE_TABLE_INV 0x20 /* invalid page-table */
  239. #define _SEG_PRESENT 0x001 /* Software (overlap with PTL) */
  240. /* Bits int the storage key */
  241. #define _PAGE_CHANGED 0x02 /* HW changed bit */
  242. #define _PAGE_REFERENCED 0x04 /* HW referenced bit */
  243. #define _USER_SEG_TABLE_LEN 0x7f /* user-segment-table up to 2 GB */
  244. #define _KERNEL_SEG_TABLE_LEN 0x7f /* kernel-segment-table up to 2 GB */
  245. /*
  246. * User and Kernel pagetables are identical
  247. */
  248. #define _PAGE_TABLE _PAGE_TABLE_LEN
  249. #define _KERNPG_TABLE _PAGE_TABLE_LEN
  250. /*
  251. * The Kernel segment-tables includes the User segment-table
  252. */
  253. #define _SEGMENT_TABLE (_USER_SEG_TABLE_LEN|0x80000000|0x100)
  254. #define _KERNSEG_TABLE _KERNEL_SEG_TABLE_LEN
  255. #define USER_STD_MASK 0x00000080UL
  256. #else /* __s390x__ */
  257. /* Bits in the segment table entry */
  258. #define _PMD_ENTRY_INV 0x20 /* invalid segment table entry */
  259. #define _PMD_ENTRY 0x00
  260. /* Bits in the region third table entry */
  261. #define _PGD_ENTRY_INV 0x20 /* invalid region table entry */
  262. #define _PGD_ENTRY 0x07
  263. /*
  264. * User and kernel page directory
  265. */
  266. #define _REGION_THIRD 0x4
  267. #define _REGION_THIRD_LEN 0x3
  268. #define _REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100)
  269. #define _KERN_REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN)
  270. #define USER_STD_MASK 0x0000000000000080UL
  271. /* Bits in the storage key */
  272. #define _PAGE_CHANGED 0x02 /* HW changed bit */
  273. #define _PAGE_REFERENCED 0x04 /* HW referenced bit */
  274. #endif /* __s390x__ */
  275. /*
  276. * Page protection definitions.
  277. */
  278. #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE)
  279. #define PAGE_RO __pgprot(_PAGE_TYPE_RO)
  280. #define PAGE_RW __pgprot(_PAGE_TYPE_RW)
  281. #define PAGE_KERNEL PAGE_RW
  282. #define PAGE_COPY PAGE_RO
  283. /*
  284. * The S390 can't do page protection for execute, and considers that the
  285. * same are read. Also, write permissions imply read permissions. This is
  286. * the closest we can get..
  287. */
  288. /*xwr*/
  289. #define __P000 PAGE_NONE
  290. #define __P001 PAGE_RO
  291. #define __P010 PAGE_RO
  292. #define __P011 PAGE_RO
  293. #define __P100 PAGE_RO
  294. #define __P101 PAGE_RO
  295. #define __P110 PAGE_RO
  296. #define __P111 PAGE_RO
  297. #define __S000 PAGE_NONE
  298. #define __S001 PAGE_RO
  299. #define __S010 PAGE_RW
  300. #define __S011 PAGE_RW
  301. #define __S100 PAGE_RO
  302. #define __S101 PAGE_RO
  303. #define __S110 PAGE_RW
  304. #define __S111 PAGE_RW
  305. /*
  306. * Certain architectures need to do special things when PTEs
  307. * within a page table are directly modified. Thus, the following
  308. * hook is made available.
  309. */
  310. static inline void set_pte(pte_t *pteptr, pte_t pteval)
  311. {
  312. *pteptr = pteval;
  313. }
  314. #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
  315. /*
  316. * pgd/pmd/pte query functions
  317. */
  318. #ifndef __s390x__
  319. static inline int pgd_present(pgd_t pgd) { return 1; }
  320. static inline int pgd_none(pgd_t pgd) { return 0; }
  321. static inline int pgd_bad(pgd_t pgd) { return 0; }
  322. static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; }
  323. static inline int pmd_none(pmd_t pmd) { return pmd_val(pmd) & _PAGE_TABLE_INV; }
  324. static inline int pmd_bad(pmd_t pmd)
  325. {
  326. return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE;
  327. }
  328. #else /* __s390x__ */
  329. static inline int pgd_present(pgd_t pgd)
  330. {
  331. return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY;
  332. }
  333. static inline int pgd_none(pgd_t pgd)
  334. {
  335. return pgd_val(pgd) & _PGD_ENTRY_INV;
  336. }
  337. static inline int pgd_bad(pgd_t pgd)
  338. {
  339. return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY;
  340. }
  341. static inline int pmd_present(pmd_t pmd)
  342. {
  343. return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY;
  344. }
  345. static inline int pmd_none(pmd_t pmd)
  346. {
  347. return pmd_val(pmd) & _PMD_ENTRY_INV;
  348. }
  349. static inline int pmd_bad(pmd_t pmd)
  350. {
  351. return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY;
  352. }
  353. #endif /* __s390x__ */
  354. static inline int pte_none(pte_t pte)
  355. {
  356. return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
  357. }
  358. static inline int pte_present(pte_t pte)
  359. {
  360. unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
  361. return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
  362. (!(pte_val(pte) & _PAGE_INVALID) &&
  363. !(pte_val(pte) & _PAGE_SWT));
  364. }
  365. static inline int pte_file(pte_t pte)
  366. {
  367. unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
  368. return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
  369. }
  370. #define pte_same(a,b) (pte_val(a) == pte_val(b))
  371. /*
  372. * query functions pte_write/pte_dirty/pte_young only work if
  373. * pte_present() is true. Undefined behaviour if not..
  374. */
  375. static inline int pte_write(pte_t pte)
  376. {
  377. return (pte_val(pte) & _PAGE_RO) == 0;
  378. }
  379. static inline int pte_dirty(pte_t pte)
  380. {
  381. /* A pte is neither clean nor dirty on s/390. The dirty bit
  382. * is in the storage key. See page_test_and_clear_dirty for
  383. * details.
  384. */
  385. return 0;
  386. }
  387. static inline int pte_young(pte_t pte)
  388. {
  389. /* A pte is neither young nor old on s/390. The young bit
  390. * is in the storage key. See page_test_and_clear_young for
  391. * details.
  392. */
  393. return 0;
  394. }
  395. static inline int pte_read(pte_t pte)
  396. {
  397. /* All pages are readable since we don't use the fetch
  398. * protection bit in the storage key.
  399. */
  400. return 1;
  401. }
  402. /*
  403. * pgd/pmd/pte modification functions
  404. */
  405. #ifndef __s390x__
  406. static inline void pgd_clear(pgd_t * pgdp) { }
  407. static inline void pmd_clear(pmd_t * pmdp)
  408. {
  409. pmd_val(pmdp[0]) = _PAGE_TABLE_INV;
  410. pmd_val(pmdp[1]) = _PAGE_TABLE_INV;
  411. pmd_val(pmdp[2]) = _PAGE_TABLE_INV;
  412. pmd_val(pmdp[3]) = _PAGE_TABLE_INV;
  413. }
  414. #else /* __s390x__ */
  415. static inline void pgd_clear(pgd_t * pgdp)
  416. {
  417. pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY;
  418. }
  419. static inline void pmd_clear(pmd_t * pmdp)
  420. {
  421. pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
  422. pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
  423. }
  424. #endif /* __s390x__ */
  425. static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
  426. {
  427. pte_val(*ptep) = _PAGE_TYPE_EMPTY;
  428. }
  429. /*
  430. * The following pte modification functions only work if
  431. * pte_present() is true. Undefined behaviour if not..
  432. */
  433. static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
  434. {
  435. pte_val(pte) &= PAGE_MASK;
  436. pte_val(pte) |= pgprot_val(newprot);
  437. return pte;
  438. }
  439. static inline pte_t pte_wrprotect(pte_t pte)
  440. {
  441. /* Do not clobber _PAGE_TYPE_NONE pages! */
  442. if (!(pte_val(pte) & _PAGE_INVALID))
  443. pte_val(pte) |= _PAGE_RO;
  444. return pte;
  445. }
  446. static inline pte_t pte_mkwrite(pte_t pte)
  447. {
  448. pte_val(pte) &= ~_PAGE_RO;
  449. return pte;
  450. }
  451. static inline pte_t pte_mkclean(pte_t pte)
  452. {
  453. /* The only user of pte_mkclean is the fork() code.
  454. We must *not* clear the *physical* page dirty bit
  455. just because fork() wants to clear the dirty bit in
  456. *one* of the page's mappings. So we just do nothing. */
  457. return pte;
  458. }
  459. static inline pte_t pte_mkdirty(pte_t pte)
  460. {
  461. /* We do not explicitly set the dirty bit because the
  462. * sske instruction is slow. It is faster to let the
  463. * next instruction set the dirty bit.
  464. */
  465. return pte;
  466. }
  467. static inline pte_t pte_mkold(pte_t pte)
  468. {
  469. /* S/390 doesn't keep its dirty/referenced bit in the pte.
  470. * There is no point in clearing the real referenced bit.
  471. */
  472. return pte;
  473. }
  474. static inline pte_t pte_mkyoung(pte_t pte)
  475. {
  476. /* S/390 doesn't keep its dirty/referenced bit in the pte.
  477. * There is no point in setting the real referenced bit.
  478. */
  479. return pte;
  480. }
  481. static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
  482. {
  483. return 0;
  484. }
  485. static inline int
  486. ptep_clear_flush_young(struct vm_area_struct *vma,
  487. unsigned long address, pte_t *ptep)
  488. {
  489. /* No need to flush TLB; bits are in storage key */
  490. return ptep_test_and_clear_young(vma, address, ptep);
  491. }
  492. static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
  493. {
  494. return 0;
  495. }
  496. static inline int
  497. ptep_clear_flush_dirty(struct vm_area_struct *vma,
  498. unsigned long address, pte_t *ptep)
  499. {
  500. /* No need to flush TLB; bits are in storage key */
  501. return ptep_test_and_clear_dirty(vma, address, ptep);
  502. }
  503. static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
  504. {
  505. pte_t pte = *ptep;
  506. pte_clear(mm, addr, ptep);
  507. return pte;
  508. }
  509. static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
  510. {
  511. if (!(pte_val(*ptep) & _PAGE_INVALID)) {
  512. #ifndef __s390x__
  513. /* S390 has 1mb segments, we are emulating 4MB segments */
  514. pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
  515. #else
  516. /* ipte in zarch mode can do the math */
  517. pte_t *pto = ptep;
  518. #endif
  519. asm volatile(
  520. " ipte %2,%3"
  521. : "=m" (*ptep) : "m" (*ptep),
  522. "a" (pto), "a" (address));
  523. }
  524. pte_val(*ptep) = _PAGE_TYPE_EMPTY;
  525. }
  526. static inline pte_t
  527. ptep_clear_flush(struct vm_area_struct *vma,
  528. unsigned long address, pte_t *ptep)
  529. {
  530. pte_t pte = *ptep;
  531. __ptep_ipte(address, ptep);
  532. return pte;
  533. }
  534. static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
  535. {
  536. pte_t old_pte = *ptep;
  537. set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
  538. }
  539. static inline void
  540. ptep_establish(struct vm_area_struct *vma,
  541. unsigned long address, pte_t *ptep,
  542. pte_t entry)
  543. {
  544. ptep_clear_flush(vma, address, ptep);
  545. set_pte(ptep, entry);
  546. }
  547. #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
  548. ptep_establish(__vma, __address, __ptep, __entry)
  549. /*
  550. * Test and clear dirty bit in storage key.
  551. * We can't clear the changed bit atomically. This is a potential
  552. * race against modification of the referenced bit. This function
  553. * should therefore only be called if it is not mapped in any
  554. * address space.
  555. */
  556. static inline int page_test_and_clear_dirty(struct page *page)
  557. {
  558. unsigned long physpage = page_to_phys(page);
  559. int skey = page_get_storage_key(physpage);
  560. if (skey & _PAGE_CHANGED)
  561. page_set_storage_key(physpage, skey & ~_PAGE_CHANGED);
  562. return skey & _PAGE_CHANGED;
  563. }
  564. /*
  565. * Test and clear referenced bit in storage key.
  566. */
  567. static inline int page_test_and_clear_young(struct page *page)
  568. {
  569. unsigned long physpage = page_to_phys(page);
  570. int ccode;
  571. asm volatile(
  572. " rrbe 0,%1\n"
  573. " ipm %0\n"
  574. " srl %0,28\n"
  575. : "=d" (ccode) : "a" (physpage) : "cc" );
  576. return ccode & 2;
  577. }
  578. /*
  579. * Conversion functions: convert a page and protection to a page entry,
  580. * and a page entry and page directory to the page they refer to.
  581. */
  582. static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
  583. {
  584. pte_t __pte;
  585. pte_val(__pte) = physpage + pgprot_val(pgprot);
  586. return __pte;
  587. }
  588. static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
  589. {
  590. unsigned long physpage = page_to_phys(page);
  591. return mk_pte_phys(physpage, pgprot);
  592. }
  593. static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
  594. {
  595. unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
  596. return mk_pte_phys(physpage, pgprot);
  597. }
  598. #ifdef __s390x__
  599. static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot)
  600. {
  601. unsigned long physpage = __pa((pfn) << PAGE_SHIFT);
  602. return __pmd(physpage + pgprot_val(pgprot));
  603. }
  604. #endif /* __s390x__ */
  605. #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
  606. #define pte_page(x) pfn_to_page(pte_pfn(x))
  607. #define pmd_page_vaddr(pmd) (pmd_val(pmd) & PAGE_MASK)
  608. #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
  609. #define pgd_page_vaddr(pgd) (pgd_val(pgd) & PAGE_MASK)
  610. #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
  611. /* to find an entry in a page-table-directory */
  612. #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
  613. #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
  614. /* to find an entry in a kernel page-table-directory */
  615. #define pgd_offset_k(address) pgd_offset(&init_mm, address)
  616. #ifndef __s390x__
  617. /* Find an entry in the second-level page table.. */
  618. static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
  619. {
  620. return (pmd_t *) dir;
  621. }
  622. #else /* __s390x__ */
  623. /* Find an entry in the second-level page table.. */
  624. #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
  625. #define pmd_offset(dir,addr) \
  626. ((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(addr))
  627. #endif /* __s390x__ */
  628. /* Find an entry in the third-level page table.. */
  629. #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
  630. #define pte_offset_kernel(pmd, address) \
  631. ((pte_t *) pmd_page_vaddr(*(pmd)) + pte_index(address))
  632. #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
  633. #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
  634. #define pte_unmap(pte) do { } while (0)
  635. #define pte_unmap_nested(pte) do { } while (0)
  636. /*
  637. * 31 bit swap entry format:
  638. * A page-table entry has some bits we have to treat in a special way.
  639. * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
  640. * exception will occur instead of a page translation exception. The
  641. * specifiation exception has the bad habit not to store necessary
  642. * information in the lowcore.
  643. * Bit 21 and bit 22 are the page invalid bit and the page protection
  644. * bit. We set both to indicate a swapped page.
  645. * Bit 30 and 31 are used to distinguish the different page types. For
  646. * a swapped page these bits need to be zero.
  647. * This leaves the bits 1-19 and bits 24-29 to store type and offset.
  648. * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
  649. * plus 24 for the offset.
  650. * 0| offset |0110|o|type |00|
  651. * 0 0000000001111111111 2222 2 22222 33
  652. * 0 1234567890123456789 0123 4 56789 01
  653. *
  654. * 64 bit swap entry format:
  655. * A page-table entry has some bits we have to treat in a special way.
  656. * Bits 52 and bit 55 have to be zero, otherwise an specification
  657. * exception will occur instead of a page translation exception. The
  658. * specifiation exception has the bad habit not to store necessary
  659. * information in the lowcore.
  660. * Bit 53 and bit 54 are the page invalid bit and the page protection
  661. * bit. We set both to indicate a swapped page.
  662. * Bit 62 and 63 are used to distinguish the different page types. For
  663. * a swapped page these bits need to be zero.
  664. * This leaves the bits 0-51 and bits 56-61 to store type and offset.
  665. * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
  666. * plus 56 for the offset.
  667. * | offset |0110|o|type |00|
  668. * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
  669. * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
  670. */
  671. #ifndef __s390x__
  672. #define __SWP_OFFSET_MASK (~0UL >> 12)
  673. #else
  674. #define __SWP_OFFSET_MASK (~0UL >> 11)
  675. #endif
  676. static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
  677. {
  678. pte_t pte;
  679. offset &= __SWP_OFFSET_MASK;
  680. pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
  681. ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
  682. return pte;
  683. }
  684. #define __swp_type(entry) (((entry).val >> 2) & 0x1f)
  685. #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
  686. #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
  687. #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
  688. #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
  689. #ifndef __s390x__
  690. # define PTE_FILE_MAX_BITS 26
  691. #else /* __s390x__ */
  692. # define PTE_FILE_MAX_BITS 59
  693. #endif /* __s390x__ */
  694. #define pte_to_pgoff(__pte) \
  695. ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
  696. #define pgoff_to_pte(__off) \
  697. ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
  698. | _PAGE_TYPE_FILE })
  699. #endif /* !__ASSEMBLY__ */
  700. #define kern_addr_valid(addr) (1)
  701. extern int add_shared_memory(unsigned long start, unsigned long size);
  702. extern int remove_shared_memory(unsigned long start, unsigned long size);
  703. /*
  704. * No page table caches to initialise
  705. */
  706. #define pgtable_cache_init() do { } while (0)
  707. #define __HAVE_ARCH_MEMMAP_INIT
  708. extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
  709. #define __HAVE_ARCH_PTEP_ESTABLISH
  710. #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
  711. #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
  712. #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
  713. #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
  714. #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
  715. #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
  716. #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
  717. #define __HAVE_ARCH_PTEP_SET_WRPROTECT
  718. #define __HAVE_ARCH_PTE_SAME
  719. #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
  720. #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
  721. #include <asm-generic/pgtable.h>
  722. #endif /* _S390_PAGE_H */