blob: e9b8e5926befb7b38b03218ffd0f9e6d2d75d454 [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
Ben Hutchingsfbd71842011-02-27 05:41:35 +00007#include <linux/mm_types.h>
8
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#ifndef __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
Andrea Arcangelie2cda322011-01-13 15:46:40 -080010extern int ptep_set_access_flags(struct vm_area_struct *vma,
11 unsigned long address, pte_t *ptep,
12 pte_t entry, int dirty);
13#endif
14
15#ifndef __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
16extern int pmdp_set_access_flags(struct vm_area_struct *vma,
17 unsigned long address, pmd_t *pmdp,
18 pmd_t entry, int dirty);
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#endif
20
21#ifndef __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
Andrea Arcangelie2cda322011-01-13 15:46:40 -080022static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
23 unsigned long address,
24 pte_t *ptep)
25{
26 pte_t pte = *ptep;
27 int r = 1;
28 if (!pte_young(pte))
29 r = 0;
30 else
31 set_pte_at(vma->vm_mm, address, ptep, pte_mkold(pte));
32 return r;
33}
34#endif
35
36#ifndef __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
37#ifdef CONFIG_TRANSPARENT_HUGEPAGE
38static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
39 unsigned long address,
40 pmd_t *pmdp)
41{
42 pmd_t pmd = *pmdp;
43 int r = 1;
44 if (!pmd_young(pmd))
45 r = 0;
46 else
47 set_pmd_at(vma->vm_mm, address, pmdp, pmd_mkold(pmd));
48 return r;
49}
50#else /* CONFIG_TRANSPARENT_HUGEPAGE */
51static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
52 unsigned long address,
53 pmd_t *pmdp)
54{
55 BUG();
56 return 0;
57}
58#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#endif
60
61#ifndef __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
Andrea Arcangelie2cda322011-01-13 15:46:40 -080062int ptep_clear_flush_young(struct vm_area_struct *vma,
63 unsigned long address, pte_t *ptep);
64#endif
65
66#ifndef __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
67int pmdp_clear_flush_young(struct vm_area_struct *vma,
68 unsigned long address, pmd_t *pmdp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#endif
70
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR
Andrea Arcangelie2cda322011-01-13 15:46:40 -080072static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
73 unsigned long address,
74 pte_t *ptep)
75{
76 pte_t pte = *ptep;
77 pte_clear(mm, address, ptep);
78 return pte;
79}
80#endif
81
82#ifndef __HAVE_ARCH_PMDP_GET_AND_CLEAR
83#ifdef CONFIG_TRANSPARENT_HUGEPAGE
84static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
85 unsigned long address,
86 pmd_t *pmdp)
87{
88 pmd_t pmd = *pmdp;
89 pmd_clear(mm, address, pmdp);
90 return pmd;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091})
Andrea Arcangelie2cda322011-01-13 15:46:40 -080092#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#endif
94
Zachary Amsdena6003882005-09-03 15:55:04 -070095#ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
Andrea Arcangelie2cda322011-01-13 15:46:40 -080096static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
97 unsigned long address, pte_t *ptep,
98 int full)
99{
100 pte_t pte;
101 pte = ptep_get_and_clear(mm, address, ptep);
102 return pte;
103}
Zachary Amsdena6003882005-09-03 15:55:04 -0700104#endif
105
Zachary Amsden9888a1c2006-09-30 23:29:31 -0700106/*
107 * Some architectures may be able to avoid expensive synchronization
108 * primitives when modifications are made to PTE's which are already
109 * not present, or in the process of an address space destruction.
110 */
111#ifndef __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800112static inline void pte_clear_not_present_full(struct mm_struct *mm,
113 unsigned long address,
114 pte_t *ptep,
115 int full)
116{
117 pte_clear(mm, address, ptep);
118}
Zachary Amsdena6003882005-09-03 15:55:04 -0700119#endif
120
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121#ifndef __HAVE_ARCH_PTEP_CLEAR_FLUSH
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800122extern pte_t ptep_clear_flush(struct vm_area_struct *vma,
123 unsigned long address,
124 pte_t *ptep);
125#endif
126
127#ifndef __HAVE_ARCH_PMDP_CLEAR_FLUSH
128extern pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
129 unsigned long address,
130 pmd_t *pmdp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131#endif
132
133#ifndef __HAVE_ARCH_PTEP_SET_WRPROTECT
Tim Schmielau8c65b4a2005-11-07 00:59:43 -0800134struct mm_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep)
136{
137 pte_t old_pte = *ptep;
138 set_pte_at(mm, address, ptep, pte_wrprotect(old_pte));
139}
140#endif
141
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800142#ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
143#ifdef CONFIG_TRANSPARENT_HUGEPAGE
144static inline void pmdp_set_wrprotect(struct mm_struct *mm,
145 unsigned long address, pmd_t *pmdp)
146{
147 pmd_t old_pmd = *pmdp;
148 set_pmd_at(mm, address, pmdp, pmd_wrprotect(old_pmd));
149}
150#else /* CONFIG_TRANSPARENT_HUGEPAGE */
151static inline void pmdp_set_wrprotect(struct mm_struct *mm,
152 unsigned long address, pmd_t *pmdp)
153{
154 BUG();
155}
156#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
157#endif
158
159#ifndef __HAVE_ARCH_PMDP_SPLITTING_FLUSH
Andrea Arcangelib3697c02011-01-16 13:10:39 -0800160extern pmd_t pmdp_splitting_flush(struct vm_area_struct *vma,
161 unsigned long address,
162 pmd_t *pmdp);
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800163#endif
164
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165#ifndef __HAVE_ARCH_PTE_SAME
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800166static inline int pte_same(pte_t pte_a, pte_t pte_b)
167{
168 return pte_val(pte_a) == pte_val(pte_b);
169}
170#endif
171
172#ifndef __HAVE_ARCH_PMD_SAME
173#ifdef CONFIG_TRANSPARENT_HUGEPAGE
174static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b)
175{
176 return pmd_val(pmd_a) == pmd_val(pmd_b);
177}
178#else /* CONFIG_TRANSPARENT_HUGEPAGE */
179static inline int pmd_same(pmd_t pmd_a, pmd_t pmd_b)
180{
181 BUG();
182 return 0;
183}
184#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185#endif
186
Martin Schwidefsky2d425522011-05-23 10:24:39 +0200187#ifndef __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
188#define page_test_and_clear_dirty(pfn, mapped) (0)
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200189#endif
190
Martin Schwidefsky2d425522011-05-23 10:24:39 +0200191#ifndef __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
Abhijit Karmarkarb4955ce2005-06-21 17:15:13 -0700192#define pte_maybe_dirty(pte) pte_dirty(pte)
193#else
194#define pte_maybe_dirty(pte) (1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195#endif
196
197#ifndef __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
Martin Schwidefsky2d425522011-05-23 10:24:39 +0200198#define page_test_and_clear_young(pfn) (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199#endif
200
201#ifndef __HAVE_ARCH_PGD_OFFSET_GATE
202#define pgd_offset_gate(mm, addr) pgd_offset(mm, addr)
203#endif
204
David S. Miller0b0968a2006-06-01 17:47:25 -0700205#ifndef __HAVE_ARCH_MOVE_PTE
Nick Piggin8b1f3122005-09-27 21:45:18 -0700206#define move_pte(pte, prot, old_addr, new_addr) (pte)
Nick Piggin8b1f3122005-09-27 21:45:18 -0700207#endif
208
Shaohua Li61c77322010-08-16 09:16:55 +0800209#ifndef flush_tlb_fix_spurious_fault
210#define flush_tlb_fix_spurious_fault(vma, address) flush_tlb_page(vma, address)
211#endif
212
Paul Mundt0634a632009-06-23 13:51:19 +0200213#ifndef pgprot_noncached
214#define pgprot_noncached(prot) (prot)
215#endif
216
venkatesh.pallipadi@intel.com2520bd32008-12-18 11:41:32 -0800217#ifndef pgprot_writecombine
218#define pgprot_writecombine pgprot_noncached
219#endif
220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221/*
Hugh Dickins8f6c99c2005-04-19 13:29:17 -0700222 * When walking page tables, get the address of the next boundary,
223 * or the end address of the range if that comes earlier. Although no
224 * vma end wraps to 0, rounded up __boundary may wrap to 0 throughout.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 */
226
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227#define pgd_addr_end(addr, end) \
228({ unsigned long __boundary = ((addr) + PGDIR_SIZE) & PGDIR_MASK; \
229 (__boundary - 1 < (end) - 1)? __boundary: (end); \
230})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
232#ifndef pud_addr_end
233#define pud_addr_end(addr, end) \
234({ unsigned long __boundary = ((addr) + PUD_SIZE) & PUD_MASK; \
235 (__boundary - 1 < (end) - 1)? __boundary: (end); \
236})
237#endif
238
239#ifndef pmd_addr_end
240#define pmd_addr_end(addr, end) \
241({ unsigned long __boundary = ((addr) + PMD_SIZE) & PMD_MASK; \
242 (__boundary - 1 < (end) - 1)? __boundary: (end); \
243})
244#endif
245
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246/*
247 * When walking page tables, we usually want to skip any p?d_none entries;
248 * and any p?d_bad entries - reporting the error before resetting to none.
249 * Do the tests inline, but report and clear the bad entry in mm/memory.c.
250 */
251void pgd_clear_bad(pgd_t *);
252void pud_clear_bad(pud_t *);
253void pmd_clear_bad(pmd_t *);
254
255static inline int pgd_none_or_clear_bad(pgd_t *pgd)
256{
257 if (pgd_none(*pgd))
258 return 1;
259 if (unlikely(pgd_bad(*pgd))) {
260 pgd_clear_bad(pgd);
261 return 1;
262 }
263 return 0;
264}
265
266static inline int pud_none_or_clear_bad(pud_t *pud)
267{
268 if (pud_none(*pud))
269 return 1;
270 if (unlikely(pud_bad(*pud))) {
271 pud_clear_bad(pud);
272 return 1;
273 }
274 return 0;
275}
276
277static inline int pmd_none_or_clear_bad(pmd_t *pmd)
278{
279 if (pmd_none(*pmd))
280 return 1;
281 if (unlikely(pmd_bad(*pmd))) {
282 pmd_clear_bad(pmd);
283 return 1;
284 }
285 return 0;
286}
Greg Ungerer95352392007-08-10 13:01:20 -0700287
Jeremy Fitzhardinge1ea07042008-06-16 04:30:00 -0700288static inline pte_t __ptep_modify_prot_start(struct mm_struct *mm,
289 unsigned long addr,
290 pte_t *ptep)
291{
292 /*
293 * Get the current pte state, but zero it out to make it
294 * non-present, preventing the hardware from asynchronously
295 * updating it.
296 */
297 return ptep_get_and_clear(mm, addr, ptep);
298}
299
300static inline void __ptep_modify_prot_commit(struct mm_struct *mm,
301 unsigned long addr,
302 pte_t *ptep, pte_t pte)
303{
304 /*
305 * The pte is non-present, so there's no hardware state to
306 * preserve.
307 */
308 set_pte_at(mm, addr, ptep, pte);
309}
310
311#ifndef __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
312/*
313 * Start a pte protection read-modify-write transaction, which
314 * protects against asynchronous hardware modifications to the pte.
315 * The intention is not to prevent the hardware from making pte
316 * updates, but to prevent any updates it may make from being lost.
317 *
318 * This does not protect against other software modifications of the
319 * pte; the appropriate pte lock must be held over the transation.
320 *
321 * Note that this interface is intended to be batchable, meaning that
322 * ptep_modify_prot_commit may not actually update the pte, but merely
323 * queue the update to be done at some later time. The update must be
324 * actually committed before the pte lock is released, however.
325 */
326static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
327 unsigned long addr,
328 pte_t *ptep)
329{
330 return __ptep_modify_prot_start(mm, addr, ptep);
331}
332
333/*
334 * Commit an update to a pte, leaving any hardware-controlled bits in
335 * the PTE unmodified.
336 */
337static inline void ptep_modify_prot_commit(struct mm_struct *mm,
338 unsigned long addr,
339 pte_t *ptep, pte_t pte)
340{
341 __ptep_modify_prot_commit(mm, addr, ptep, pte);
342}
343#endif /* __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION */
Sebastian Siewiorfe1a6872008-07-15 22:28:46 +0200344#endif /* CONFIG_MMU */
Jeremy Fitzhardinge1ea07042008-06-16 04:30:00 -0700345
Greg Ungerer95352392007-08-10 13:01:20 -0700346/*
347 * A facility to provide lazy MMU batching. This allows PTE updates and
348 * page invalidations to be delayed until a call to leave lazy MMU mode
349 * is issued. Some architectures may benefit from doing this, and it is
350 * beneficial for both shadow and direct mode hypervisors, which may batch
351 * the PTE updates which happen during this window. Note that using this
352 * interface requires that read hazards be removed from the code. A read
353 * hazard could result in the direct mode hypervisor case, since the actual
354 * write to the page tables may not yet have taken place, so reads though
355 * a raw PTE pointer after it has been modified are not guaranteed to be
356 * up to date. This mode can only be entered and left under the protection of
357 * the page table locks for all page tables which may be modified. In the UP
358 * case, this is required so that preemption is disabled, and in the SMP case,
359 * it must synchronize the delayed page table writes properly on other CPUs.
360 */
361#ifndef __HAVE_ARCH_ENTER_LAZY_MMU_MODE
362#define arch_enter_lazy_mmu_mode() do {} while (0)
363#define arch_leave_lazy_mmu_mode() do {} while (0)
364#define arch_flush_lazy_mmu_mode() do {} while (0)
365#endif
366
367/*
Jeremy Fitzhardinge7fd7d832009-02-17 23:24:03 -0800368 * A facility to provide batching of the reload of page tables and
369 * other process state with the actual context switch code for
370 * paravirtualized guests. By convention, only one of the batched
371 * update (lazy) modes (CPU, MMU) should be active at any given time,
372 * entry should never be nested, and entry and exits should always be
373 * paired. This is for sanity of maintaining and reasoning about the
374 * kernel code. In this case, the exit (end of the context switch) is
375 * in architecture-specific code, and so doesn't need a generic
376 * definition.
Greg Ungerer95352392007-08-10 13:01:20 -0700377 */
Jeremy Fitzhardinge7fd7d832009-02-17 23:24:03 -0800378#ifndef __HAVE_ARCH_START_CONTEXT_SWITCH
Jeremy Fitzhardinge224101e2009-02-18 11:18:57 -0800379#define arch_start_context_switch(prev) do {} while (0)
Greg Ungerer95352392007-08-10 13:01:20 -0700380#endif
381
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800382#ifndef __HAVE_PFNMAP_TRACKING
383/*
384 * Interface that can be used by architecture code to keep track of
385 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
386 *
387 * track_pfn_vma_new is called when a _new_ pfn mapping is being established
388 * for physical range indicated by pfn and size.
389 */
venkatesh.pallipadi@intel.come4b866e2009-01-09 16:13:11 -0800390static inline int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800391 unsigned long pfn, unsigned long size)
392{
393 return 0;
394}
395
396/*
397 * Interface that can be used by architecture code to keep track of
398 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
399 *
400 * track_pfn_vma_copy is called when vma that is covering the pfnmap gets
401 * copied through copy_page_range().
402 */
403static inline int track_pfn_vma_copy(struct vm_area_struct *vma)
404{
405 return 0;
406}
407
408/*
409 * Interface that can be used by architecture code to keep track of
410 * memory type of pfn mappings (remap_pfn_range, vm_insert_pfn)
411 *
412 * untrack_pfn_vma is called while unmapping a pfnmap for a region.
413 * untrack can be called for a specific region indicated by pfn and size or
414 * can be for the entire vma (in which case size can be zero).
415 */
416static inline void untrack_pfn_vma(struct vm_area_struct *vma,
417 unsigned long pfn, unsigned long size)
418{
419}
420#else
venkatesh.pallipadi@intel.come4b866e2009-01-09 16:13:11 -0800421extern int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
venkatesh.pallipadi@intel.com34801ba2008-12-19 13:47:29 -0800422 unsigned long pfn, unsigned long size);
423extern int track_pfn_vma_copy(struct vm_area_struct *vma);
424extern void untrack_pfn_vma(struct vm_area_struct *vma, unsigned long pfn,
425 unsigned long size);
426#endif
427
Andrea Arcangeli5f6e8da2011-01-13 15:46:40 -0800428#ifndef CONFIG_TRANSPARENT_HUGEPAGE
429static inline int pmd_trans_huge(pmd_t pmd)
430{
431 return 0;
432}
433static inline int pmd_trans_splitting(pmd_t pmd)
434{
435 return 0;
436}
Andrea Arcangelie2cda322011-01-13 15:46:40 -0800437#ifndef __HAVE_ARCH_PMD_WRITE
438static inline int pmd_write(pmd_t pmd)
439{
440 BUG();
441 return 0;
442}
443#endif /* __HAVE_ARCH_PMD_WRITE */
Andrea Arcangeli5f6e8da2011-01-13 15:46:40 -0800444#endif
445
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446#endif /* !__ASSEMBLY__ */
447
448#endif /* _ASM_GENERIC_PGTABLE_H */