blob: 31b6188df221f6114eb5c54bdc83673e8130e525 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _ASM_GENERIC_PGTABLE_H
2#define _ASM_GENERIC_PGTABLE_H
3
Rusty Russell673eae82006-09-25 23:32:29 -07004#ifndef __ASSEMBLY__
Greg Ungerer95352392007-08-10 13:01:20 -07005#ifdef CONFIG_MMU
Rusty Russell673eae82006-09-25 23:32:29 -07006
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#ifndef __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
Andrea Arcangelie2cda322011-01-13 15:46:40 -08008extern int ptep_set_access_flags(struct vm_area_struct *vma,
9 unsigned long address, pte_t *ptep,
10 pte_t entry, int dirty);
11#endif
12
13#ifndef __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
14extern int pmdp_set_access_flags(struct vm_area_struct *vma,
15 unsigned long address, pmd_t *pmdp,
16 pmd_t entry, int dirty);
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#endif
18
19#ifndef __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
Andrea Arcangelie2cda322011-01-13 15:46:40 -080020static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
21 unsigned long address,
22 pte_t *ptep)
23{
24 pte_t pte = *ptep;
25 int r = 1;
26 if (!pte_young(pte))
27 r = 0;
28 else
29 set_pte_at(vma->vm_mm, address, ptep, pte_mkold(pte));
30 return r;
31}
32#endif
33
34#ifndef __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
35#ifdef CONFIG_TRANSPARENT_HUGEPAGE
36static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
37 unsigned long address,
38 pmd_t *pmdp)
39{
40 pmd_t pmd = *pmdp;
41 int r = 1;
42 if (!pmd_young(pmd))
43 r = 0;
44 else
45 set_pmd_at(vma->vm_mm, address, pmdp, pmd_mkold(pmd));
46 return r;
47}
48#else /* CONFIG_TRANSPARENT_HUGEPAGE */
49static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
50 unsigned long address,
51 pmd_t *pmdp)
52{
53 BUG();
54 return 0;
55}
56#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#endif
58
59#ifndef __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
Andrea Arcangelie2cda322011-01-13 15:46:40 -080060int ptep_clear_flush_young(struct vm_area_struct *vma,
61 unsigned long address, pte_t *ptep);
62#endif
63
64#ifndef __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
65int pmdp_clear_flush_young(struct vm_area_struct *vma,
66 unsigned long address, pmd_t *pmdp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#endif
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR
Andrea Arcangelie2cda322011-01-13 15:46:40 -080070static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
71 unsigned long address,
72 pte_t *ptep)
73{
74 pte_t pte = *ptep;
75 pte_clear(mm, address, ptep);
76 return pte;
77}
78#endif
79
80#ifndef __HAVE_ARCH_PMDP_GET_AND_CLEAR
81#ifdef CONFIG_TRANSPARENT_HUGEPAGE
82static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
83 unsigned long address,
84 pmd_t *pmdp)
85{
86 pmd_t pmd = *pmdp;
87 pmd_clear(mm, address, pmdp);
88 return pmd;
Linus Torvalds1da177e2005-04-16 15:20:36 -070089})
Andrea Arcangelie2cda322011-01-13 15:46:40 -080090#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -070091#endif
92
Zachary Amsdena6003882005-09-03 15:55:04 -070093#ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
Andrea Arcangelie2cda322011-01-13 15:46:40 -080094static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
95 unsigned long address, pte_t *ptep,
96 int full)
97{
98 pte_t pte;
99 pte = ptep_get_and_clear(mm, address, ptep);
100 return pte;
101}
Zachary Amsdena6003882005-09-03 15:55:04 -0700102#endif
103
Zachary Amsden9888a1c2006-09-30 23:29:31 -0700104/*
105 * Some architectures may be able to avoid expensive synchronization
106 * primitives when modifications are made to PTE's which are already
107 * not present, or in the process of an address space destruction.
108 */
109#ifndef __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800110static inline void pte_clear_not_present_full(struct mm_struct *mm,
111 unsigned long address,
112 pte_t *ptep,
113 int full)
114{
115 pte_clear(mm, address, ptep);
116}
Zachary Amsdena6003882005-09-03 15:55:04 -0700117#endif
118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119#ifndef __HAVE_ARCH_PTEP_CLEAR_FLUSH
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800120extern pte_t ptep_clear_flush(struct vm_area_struct *vma,
121 unsigned long address,
122 pte_t *ptep);
123#endif
124
125#ifndef __HAVE_ARCH_PMDP_CLEAR_FLUSH
126extern pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
127 unsigned long address,
128 pmd_t *pmdp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129#endif
130
131#ifndef __HAVE_ARCH_PTEP_SET_WRPROTECT
Tim Schmielau8c65b4a2005-11-07 00:59:43 -0800132struct mm_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep)
134{
135 pte_t old_pte = *ptep;
136 set_pte_at(mm, address, ptep, pte_wrprotect(old_pte));
137}
138#endif
139
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800140#ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
141#ifdef CONFIG_TRANSPARENT_HUGEPAGE
142static inline void pmdp_set_wrprotect(struct mm_struct *mm,
143 unsigned long address, pmd_t *pmdp)
144{
145 pmd_t old_pmd = *pmdp;
146 set_pmd_at(mm, address, pmdp, pmd_wrprotect(old_pmd));
147}
148#else /* CONFIG_TRANSPARENT_HUGEPAGE */
149static inline void pmdp_set_wrprotect(struct mm_struct *mm,
150 unsigned long address, pmd_t *pmdp)
151{
152 BUG();
153}
154#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
155#endif
156
157#ifndef __HAVE_ARCH_PMDP_SPLITTING_FLUSH
Andrea Arcangelib3697c02011-01-16 13:10:39 -0800158extern pmd_t pmdp_splitting_flush(struct vm_area_struct *vma,
159 unsigned long address,
160 pmd_t *pmdp);
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800161#endif
162
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163#ifndef __HAVE_ARCH_PTE_SAME
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800164static inline int pte_same(pte_t pte_a, pte_t pte_b)
165{
166 return pte_val(pte_a) == pte_val(pte_b);
167}
168#endif
169
170#ifndef __HAVE_ARCH_PMD_SAME
171#ifdef CONFIG_TRANSPARENT_HUGEPAGE
172static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b)
173{
174 return pmd_val(pmd_a) == pmd_val(pmd_b);
175}
176#else /* CONFIG_TRANSPARENT_HUGEPAGE */
177static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b)
178{
179 BUG();
180 return 0;
181}
182#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183#endif
184
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200185#ifndef __HAVE_ARCH_PAGE_TEST_DIRTY
186#define page_test_dirty(page) (0)
187#endif
188
189#ifndef __HAVE_ARCH_PAGE_CLEAR_DIRTY
Martin Schwidefskye2b8d7a2010-10-25 16:10:14 +0200190#define page_clear_dirty(page, mapped) do { } while (0)
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200191#endif
192
193#ifndef __HAVE_ARCH_PAGE_TEST_DIRTY
Abhijit Karmarkarb4955ce2005-06-21 17:15:13 -0700194#define pte_maybe_dirty(pte) pte_dirty(pte)
195#else
196#define pte_maybe_dirty(pte) (1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197#endif
198
199#ifndef __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
200#define page_test_and_clear_young(page) (0)
201#endif
202
203#ifndef __HAVE_ARCH_PGD_OFFSET_GATE
204#define pgd_offset_gate(mm, addr) pgd_offset(mm, addr)
205#endif
206
David S. Miller0b0968a2006-06-01 17:47:25 -0700207#ifndef __HAVE_ARCH_MOVE_PTE
Nick Piggin8b1f3122005-09-27 21:45:18 -0700208#define move_pte(pte, prot, old_addr, new_addr) (pte)
Nick Piggin8b1f3122005-09-27 21:45:18 -0700209#endif
210
Shaohua Li61c77322010-08-16 09:16:55 +0800211#ifndef flush_tlb_fix_spurious_fault
212#define flush_tlb_fix_spurious_fault(vma, address) flush_tlb_page(vma, address)
213#endif
214
Paul Mundt0634a632009-06-23 13:51:19 +0200215#ifndef pgprot_noncached
216#define pgprot_noncached(prot) (prot)
217#endif
218
venkatesh.pallipadi@intel.com2520bd32008-12-18 11:41:32 -0800219#ifndef pgprot_writecombine
220#define pgprot_writecombine pgprot_noncached
221#endif
222
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223/*
Hugh Dickins8f6c99c2005-04-19 13:29:17 -0700224 * When walking page tables, get the address of the next boundary,
225 * or the end address of the range if that comes earlier. Although no
226 * vma end wraps to 0, rounded up __boundary may wrap to 0 throughout.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 */
228
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229#define pgd_addr_end(addr, end) \
230({ unsigned long __boundary = ((addr) + PGDIR_SIZE) & PGDIR_MASK; \
231 (__boundary - 1 < (end) - 1)? __boundary: (end); \
232})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
234#ifndef pud_addr_end
235#define pud_addr_end(addr, end) \
236({ unsigned long __boundary = ((addr) + PUD_SIZE) & PUD_MASK; \
237 (__boundary - 1 < (end) - 1)? __boundary: (end); \
238})
239#endif
240
241#ifndef pmd_addr_end
242#define pmd_addr_end(addr, end) \
243({ unsigned long __boundary = ((addr) + PMD_SIZE) & PMD_MASK; \
244 (__boundary - 1 < (end) - 1)? __boundary: (end); \
245})
246#endif
247
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248/*
249 * When walking page tables, we usually want to skip any p?d_none entries;
250 * and any p?d_bad entries - reporting the error before resetting to none.
251 * Do the tests inline, but report and clear the bad entry in mm/memory.c.
252 */
253void pgd_clear_bad(pgd_t *);
254void pud_clear_bad(pud_t *);
255void pmd_clear_bad(pmd_t *);
256
257static inline int pgd_none_or_clear_bad(pgd_t *pgd)
258{
259 if (pgd_none(*pgd))
260 return 1;
261 if (unlikely(pgd_bad(*pgd))) {
262 pgd_clear_bad(pgd);
263 return 1;
264 }
265 return 0;
266}
267
268static inline int pud_none_or_clear_bad(pud_t *pud)
269{
270 if (pud_none(*pud))
271 return 1;
272 if (unlikely(pud_bad(*pud))) {
273 pud_clear_bad(pud);
274 return 1;
275 }
276 return 0;
277}
278
279static inline int pmd_none_or_clear_bad(pmd_t *pmd)
280{
281 if (pmd_none(*pmd))
282 return 1;
283 if (unlikely(pmd_bad(*pmd))) {
284 pmd_clear_bad(pmd);
285 return 1;
286 }
287 return 0;
288}
Greg Ungerer95352392007-08-10 13:01:20 -0700289
Jeremy Fitzhardinge1ea07042008-06-16 04:30:00 -0700290static inline pte_t __ptep_modify_prot_start(struct mm_struct *mm,
291 unsigned long addr,
292 pte_t *ptep)
293{
294 /*
295 * Get the current pte state, but zero it out to make it
296 * non-present, preventing the hardware from asynchronously
297 * updating it.
298 */
299 return ptep_get_and_clear(mm, addr, ptep);
300}
301
302static inline void __ptep_modify_prot_commit(struct mm_struct *mm,
303 unsigned long addr,
304 pte_t *ptep, pte_t pte)
305{
306 /*
307 * The pte is non-present, so there's no hardware state to
308 * preserve.
309 */
310 set_pte_at(mm, addr, ptep, pte);
311}
312
313#ifndef __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
314/*
315 * Start a pte protection read-modify-write transaction, which
316 * protects against asynchronous hardware modifications to the pte.
317 * The intention is not to prevent the hardware from making pte
318 * updates, but to prevent any updates it may make from being lost.
319 *
320 * This does not protect against other software modifications of the
321 * pte; the appropriate pte lock must be held over the transation.
322 *
323 * Note that this interface is intended to be batchable, meaning that
324 * ptep_modify_prot_commit may not actually update the pte, but merely
325 * queue the update to be done at some later time. The update must be
326 * actually committed before the pte lock is released, however.
327 */
328static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
329 unsigned long addr,
330 pte_t *ptep)
331{
332 return __ptep_modify_prot_start(mm, addr, ptep);
333}
334
335/*
336 * Commit an update to a pte, leaving any hardware-controlled bits in
337 * the PTE unmodified.
338 */
339static inline void ptep_modify_prot_commit(struct mm_struct *mm,
340 unsigned long addr,
341 pte_t *ptep, pte_t pte)
342{
343 __ptep_modify_prot_commit(mm, addr, ptep, pte);
344}
345#endif /* __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION */
Sebastian Siewiorfe1a6872008-07-15 22:28:46 +0200346#endif /* CONFIG_MMU */
Jeremy Fitzhardinge1ea07042008-06-16 04:30:00 -0700347
Greg Ungerer95352392007-08-10 13:01:20 -0700348/*
349 * A facility to provide lazy MMU batching. This allows PTE updates and
350 * page invalidations to be delayed until a call to leave lazy MMU mode
351 * is issued. Some architectures may benefit from doing this, and it is
352 * beneficial for both shadow and direct mode hypervisors, which may batch
353 * the PTE updates which happen during this window. Note that using this
354 * interface requires that read hazards be removed from the code. A read
355 * hazard could result in the direct mode hypervisor case, since the actual
356 * write to the page tables may not yet have taken place, so reads though
357 * a raw PTE pointer after it has been modified are not guaranteed to be
358 * up to date. This mode can only be entered and left under the protection of
359 * the page table locks for all page tables which may be modified. In the UP
360 * case, this is required so that preemption is disabled, and in the SMP case,
361 * it must synchronize the delayed page table writes properly on other CPUs.
362 */
363#ifndef __HAVE_ARCH_ENTER_LAZY_MMU_MODE
364#define arch_enter_lazy_mmu_mode() do {} while (0)
365#define arch_leave_lazy_mmu_mode() do {} while (0)
366#define arch_flush_lazy_mmu_mode() do {} while (0)
367#endif
368
369/*
Jeremy Fitzhardinge7fd7d832009-02-17 23:24:03 -0800370 * A facility to provide batching of the reload of page tables and
371 * other process state with the actual context switch code for
372 * paravirtualized guests. By convention, only one of the batched
373 * update (lazy) modes (CPU, MMU) should be active at any given time,
374 * entry should never be nested, and entry and exits should always be
375 * paired. This is for sanity of maintaining and reasoning about the
376 * kernel code. In this case, the exit (end of the context switch) is
377 * in architecture-specific code, and so doesn't need a generic
378 * definition.
Greg Ungerer95352392007-08-10 13:01:20 -0700379 */
Jeremy Fitzhardinge7fd7d832009-02-17 23:24:03 -0800380#ifndef __HAVE_ARCH_START_CONTEXT_SWITCH
Jeremy Fitzhardinge224101e2009-02-18 11:18:57 -0800381#define arch_start_context_switch(prev) do {} while (0)
Greg Ungerer95352392007-08-10 13:01:20 -0700382#endif
383
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800384#ifndef __HAVE_PFNMAP_TRACKING
385/*
386 * Interface that can be used by architecture code to keep track of
387 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
388 *
389 * track_pfn_vma_new is called when a _new_ pfn mapping is being established
390 * for physical range indicated by pfn and size.
391 */
venkatesh.pallipadi@intel.come4b866e2009-01-09 16:13:11 -0800392static inline int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800393 unsigned long pfn, unsigned long size)
394{
395 return 0;
396}
397
398/*
399 * Interface that can be used by architecture code to keep track of
400 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
401 *
402 * track_pfn_vma_copy is called when vma that is covering the pfnmap gets
403 * copied through copy_page_range().
404 */
405static inline int track_pfn_vma_copy(struct vm_area_struct *vma)
406{
407 return 0;
408}
409
410/*
411 * Interface that can be used by architecture code to keep track of
412 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
413 *
414 * untrack_pfn_vma is called while unmapping a pfnmap for a region.
415 * untrack can be called for a specific region indicated by pfn and size or
416 * can be for the entire vma (in which case size can be zero).
417 */
418static inline void untrack_pfn_vma(struct vm_area_struct *vma,
419 unsigned long pfn, unsigned long size)
420{
421}
422#else
venkatesh.pallipadi@intel.come4b866e2009-01-09 16:13:11 -0800423extern int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800424 unsigned long pfn, unsigned long size);
425extern int track_pfn_vma_copy(struct vm_area_struct *vma);
426extern void untrack_pfn_vma(struct vm_area_struct *vma, unsigned long pfn,
427 unsigned long size);
428#endif
429
Andrea Arcangeli5f6e8da2011-01-13 15:46:40 -0800430#ifndef CONFIG_TRANSPARENT_HUGEPAGE
431static inline int pmd_trans_huge(pmd_t pmd)
432{
433 return 0;
434}
435static inline int pmd_trans_splitting(pmd_t pmd)
436{
437 return 0;
438}
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800439#ifndef __HAVE_ARCH_PMD_WRITE
440static inline int pmd_write(pmd_t pmd)
441{
442 BUG();
443 return 0;
444}
445#endif /* __HAVE_ARCH_PMD_WRITE */
Andrea Arcangeli5f6e8da2011-01-13 15:46:40 -0800446#endif
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448#endif /* !__ASSEMBLY__ */
449
450#endif /* _ASM_GENERIC_PGTABLE_H */