1 #ifndef _ASM_GENERIC_PGTABLE_H
2 #define _ASM_GENERIC_PGTABLE_H
7 #include <linux/mm_types.h>
11 * On almost all architectures and configurations, 0 can be used as the
12 * upper ceiling to free_pgtables(): on many architectures it has the same
13 * effect as using TASK_SIZE. However, there is one configuration which
14 * must impose a more careful limit, to avoid freeing kernel pgtables.
16 #ifndef USER_PGTABLES_CEILING
17 #define USER_PGTABLES_CEILING 0UL
20 #ifndef __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
21 extern int ptep_set_access_flags(struct vm_area_struct
*vma
,
22 unsigned long address
, pte_t
*ptep
,
23 pte_t entry
, int dirty
);
26 #ifndef __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
27 extern int pmdp_set_access_flags(struct vm_area_struct
*vma
,
28 unsigned long address
, pmd_t
*pmdp
,
29 pmd_t entry
, int dirty
);
32 #ifndef __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
33 static inline int ptep_test_and_clear_young(struct vm_area_struct
*vma
,
34 unsigned long address
,
42 set_pte_at(vma
->vm_mm
, address
, ptep
, pte_mkold(pte
));
47 #ifndef __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
48 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
49 static inline int pmdp_test_and_clear_young(struct vm_area_struct
*vma
,
50 unsigned long address
,
58 set_pmd_at(vma
->vm_mm
, address
, pmdp
, pmd_mkold(pmd
));
61 #else /* CONFIG_TRANSPARENT_HUGEPAGE */
62 static inline int pmdp_test_and_clear_young(struct vm_area_struct
*vma
,
63 unsigned long address
,
69 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
72 #ifndef __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
73 int ptep_clear_flush_young(struct vm_area_struct
*vma
,
74 unsigned long address
, pte_t
*ptep
);
77 #ifndef __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
78 int pmdp_clear_flush_young(struct vm_area_struct
*vma
,
79 unsigned long address
, pmd_t
*pmdp
);
82 #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR
83 static inline pte_t
ptep_get_and_clear(struct mm_struct
*mm
,
84 unsigned long address
,
88 pte_clear(mm
, address
, ptep
);
93 #ifndef __HAVE_ARCH_PMDP_GET_AND_CLEAR
94 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
95 static inline pmd_t
pmdp_get_and_clear(struct mm_struct
*mm
,
96 unsigned long address
,
103 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
106 #ifndef __HAVE_ARCH_PMDP_GET_AND_CLEAR_FULL
107 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
108 static inline pmd_t
pmdp_get_and_clear_full(struct mm_struct
*mm
,
109 unsigned long address
, pmd_t
*pmdp
,
112 return pmdp_get_and_clear(mm
, address
, pmdp
);
114 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
117 #ifndef __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
118 static inline pte_t
ptep_get_and_clear_full(struct mm_struct
*mm
,
119 unsigned long address
, pte_t
*ptep
,
123 pte
= ptep_get_and_clear(mm
, address
, ptep
);
129 * Some architectures may be able to avoid expensive synchronization
130 * primitives when modifications are made to PTE's which are already
131 * not present, or in the process of an address space destruction.
133 #ifndef __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
134 static inline void pte_clear_not_present_full(struct mm_struct
*mm
,
135 unsigned long address
,
139 pte_clear(mm
, address
, ptep
);
143 #ifndef __HAVE_ARCH_PTEP_CLEAR_FLUSH
144 extern pte_t
ptep_clear_flush(struct vm_area_struct
*vma
,
145 unsigned long address
,
149 #ifndef __HAVE_ARCH_PMDP_CLEAR_FLUSH
150 extern pmd_t
pmdp_clear_flush(struct vm_area_struct
*vma
,
151 unsigned long address
,
155 #ifndef __HAVE_ARCH_PTEP_SET_WRPROTECT
157 static inline void ptep_set_wrprotect(struct mm_struct
*mm
, unsigned long address
, pte_t
*ptep
)
159 pte_t old_pte
= *ptep
;
160 set_pte_at(mm
, address
, ptep
, pte_wrprotect(old_pte
));
164 #ifndef __HAVE_ARCH_PMDP_SET_WRPROTECT
165 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
166 static inline void pmdp_set_wrprotect(struct mm_struct
*mm
,
167 unsigned long address
, pmd_t
*pmdp
)
169 pmd_t old_pmd
= *pmdp
;
170 set_pmd_at(mm
, address
, pmdp
, pmd_wrprotect(old_pmd
));
172 #else /* CONFIG_TRANSPARENT_HUGEPAGE */
173 static inline void pmdp_set_wrprotect(struct mm_struct
*mm
,
174 unsigned long address
, pmd_t
*pmdp
)
178 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
181 #ifndef __HAVE_ARCH_PMDP_SPLITTING_FLUSH
182 extern void pmdp_splitting_flush(struct vm_area_struct
*vma
,
183 unsigned long address
, pmd_t
*pmdp
);
186 #ifndef __HAVE_ARCH_PGTABLE_DEPOSIT
187 extern void pgtable_trans_huge_deposit(struct mm_struct
*mm
, pmd_t
*pmdp
,
191 #ifndef __HAVE_ARCH_PGTABLE_WITHDRAW
192 extern pgtable_t
pgtable_trans_huge_withdraw(struct mm_struct
*mm
, pmd_t
*pmdp
);
195 #ifndef __HAVE_ARCH_PMDP_INVALIDATE
196 extern void pmdp_invalidate(struct vm_area_struct
*vma
, unsigned long address
,
200 #ifndef __HAVE_ARCH_PTE_SAME
201 static inline int pte_same(pte_t pte_a
, pte_t pte_b
)
203 return pte_val(pte_a
) == pte_val(pte_b
);
207 #ifndef __HAVE_ARCH_PTE_UNUSED
209 * Some architectures provide facilities to virtualization guests
210 * so that they can flag allocated pages as unused. This allows the
211 * host to transparently reclaim unused pages. This function returns
212 * whether the pte's page is unused.
214 static inline int pte_unused(pte_t pte
)
220 #ifndef __HAVE_ARCH_PMD_SAME
221 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
222 static inline int pmd_same(pmd_t pmd_a
, pmd_t pmd_b
)
224 return pmd_val(pmd_a
) == pmd_val(pmd_b
);
226 #else /* CONFIG_TRANSPARENT_HUGEPAGE */
227 static inline int pmd_same(pmd_t pmd_a
, pmd_t pmd_b
)
232 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
235 #ifndef __HAVE_ARCH_PGD_OFFSET_GATE
236 #define pgd_offset_gate(mm, addr) pgd_offset(mm, addr)
239 #ifndef __HAVE_ARCH_MOVE_PTE
240 #define move_pte(pte, prot, old_addr, new_addr) (pte)
243 #ifndef pte_accessible
244 # define pte_accessible(mm, pte) ((void)(pte), 1)
247 #ifndef flush_tlb_fix_spurious_fault
248 #define flush_tlb_fix_spurious_fault(vma, address) flush_tlb_page(vma, address)
251 #ifndef pgprot_noncached
252 #define pgprot_noncached(prot) (prot)
255 #ifndef pgprot_writecombine
256 #define pgprot_writecombine pgprot_noncached
259 #ifndef pgprot_device
260 #define pgprot_device pgprot_noncached
263 #ifndef pgprot_modify
264 #define pgprot_modify pgprot_modify
265 static inline pgprot_t
pgprot_modify(pgprot_t oldprot
, pgprot_t newprot
)
267 if (pgprot_val(oldprot
) == pgprot_val(pgprot_noncached(oldprot
)))
268 newprot
= pgprot_noncached(newprot
);
269 if (pgprot_val(oldprot
) == pgprot_val(pgprot_writecombine(oldprot
)))
270 newprot
= pgprot_writecombine(newprot
);
271 if (pgprot_val(oldprot
) == pgprot_val(pgprot_device(oldprot
)))
272 newprot
= pgprot_device(newprot
);
278 * When walking page tables, get the address of the next boundary,
279 * or the end address of the range if that comes earlier. Although no
280 * vma end wraps to 0, rounded up __boundary may wrap to 0 throughout.
283 #define pgd_addr_end(addr, end) \
284 ({ unsigned long __boundary = ((addr) + PGDIR_SIZE) & PGDIR_MASK; \
285 (__boundary - 1 < (end) - 1)? __boundary: (end); \
289 #define pud_addr_end(addr, end) \
290 ({ unsigned long __boundary = ((addr) + PUD_SIZE) & PUD_MASK; \
291 (__boundary - 1 < (end) - 1)? __boundary: (end); \
296 #define pmd_addr_end(addr, end) \
297 ({ unsigned long __boundary = ((addr) + PMD_SIZE) & PMD_MASK; \
298 (__boundary - 1 < (end) - 1)? __boundary: (end); \
303 * When walking page tables, we usually want to skip any p?d_none entries;
304 * and any p?d_bad entries - reporting the error before resetting to none.
305 * Do the tests inline, but report and clear the bad entry in mm/memory.c.
307 void pgd_clear_bad(pgd_t
*);
308 void pud_clear_bad(pud_t
*);
309 void pmd_clear_bad(pmd_t
*);
311 static inline int pgd_none_or_clear_bad(pgd_t
*pgd
)
315 if (unlikely(pgd_bad(*pgd
))) {
322 static inline int pud_none_or_clear_bad(pud_t
*pud
)
326 if (unlikely(pud_bad(*pud
))) {
333 static inline int pmd_none_or_clear_bad(pmd_t
*pmd
)
337 if (unlikely(pmd_bad(*pmd
))) {
344 static inline pte_t
__ptep_modify_prot_start(struct mm_struct
*mm
,
349 * Get the current pte state, but zero it out to make it
350 * non-present, preventing the hardware from asynchronously
353 return ptep_get_and_clear(mm
, addr
, ptep
);
356 static inline void __ptep_modify_prot_commit(struct mm_struct
*mm
,
358 pte_t
*ptep
, pte_t pte
)
361 * The pte is non-present, so there's no hardware state to
364 set_pte_at(mm
, addr
, ptep
, pte
);
367 #ifndef __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
369 * Start a pte protection read-modify-write transaction, which
370 * protects against asynchronous hardware modifications to the pte.
371 * The intention is not to prevent the hardware from making pte
372 * updates, but to prevent any updates it may make from being lost.
374 * This does not protect against other software modifications of the
375 * pte; the appropriate pte lock must be held over the transation.
377 * Note that this interface is intended to be batchable, meaning that
378 * ptep_modify_prot_commit may not actually update the pte, but merely
379 * queue the update to be done at some later time. The update must be
380 * actually committed before the pte lock is released, however.
382 static inline pte_t
ptep_modify_prot_start(struct mm_struct
*mm
,
386 return __ptep_modify_prot_start(mm
, addr
, ptep
);
390 * Commit an update to a pte, leaving any hardware-controlled bits in
391 * the PTE unmodified.
393 static inline void ptep_modify_prot_commit(struct mm_struct
*mm
,
395 pte_t
*ptep
, pte_t pte
)
397 __ptep_modify_prot_commit(mm
, addr
, ptep
, pte
);
399 #endif /* __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION */
400 #endif /* CONFIG_MMU */
403 * A facility to provide lazy MMU batching. This allows PTE updates and
404 * page invalidations to be delayed until a call to leave lazy MMU mode
405 * is issued. Some architectures may benefit from doing this, and it is
406 * beneficial for both shadow and direct mode hypervisors, which may batch
407 * the PTE updates which happen during this window. Note that using this
408 * interface requires that read hazards be removed from the code. A read
409 * hazard could result in the direct mode hypervisor case, since the actual
410 * write to the page tables may not yet have taken place, so reads though
411 * a raw PTE pointer after it has been modified are not guaranteed to be
412 * up to date. This mode can only be entered and left under the protection of
413 * the page table locks for all page tables which may be modified. In the UP
414 * case, this is required so that preemption is disabled, and in the SMP case,
415 * it must synchronize the delayed page table writes properly on other CPUs.
417 #ifndef __HAVE_ARCH_ENTER_LAZY_MMU_MODE
418 #define arch_enter_lazy_mmu_mode() do {} while (0)
419 #define arch_leave_lazy_mmu_mode() do {} while (0)
420 #define arch_flush_lazy_mmu_mode() do {} while (0)
424 * A facility to provide batching of the reload of page tables and
425 * other process state with the actual context switch code for
426 * paravirtualized guests. By convention, only one of the batched
427 * update (lazy) modes (CPU, MMU) should be active at any given time,
428 * entry should never be nested, and entry and exits should always be
429 * paired. This is for sanity of maintaining and reasoning about the
430 * kernel code. In this case, the exit (end of the context switch) is
431 * in architecture-specific code, and so doesn't need a generic
434 #ifndef __HAVE_ARCH_START_CONTEXT_SWITCH
435 #define arch_start_context_switch(prev) do {} while (0)
438 #ifndef CONFIG_HAVE_ARCH_SOFT_DIRTY
439 static inline int pte_soft_dirty(pte_t pte
)
444 static inline int pmd_soft_dirty(pmd_t pmd
)
449 static inline pte_t
pte_mksoft_dirty(pte_t pte
)
454 static inline pmd_t
pmd_mksoft_dirty(pmd_t pmd
)
459 static inline pte_t
pte_swp_mksoft_dirty(pte_t pte
)
464 static inline int pte_swp_soft_dirty(pte_t pte
)
469 static inline pte_t
pte_swp_clear_soft_dirty(pte_t pte
)
475 #ifndef __HAVE_PFNMAP_TRACKING
477 * Interfaces that can be used by architecture code to keep track of
478 * memory type of pfn mappings specified by the remap_pfn_range,
483 * track_pfn_remap is called when a _new_ pfn mapping is being established
484 * by remap_pfn_range() for physical range indicated by pfn and size.
486 static inline int track_pfn_remap(struct vm_area_struct
*vma
, pgprot_t
*prot
,
487 unsigned long pfn
, unsigned long addr
,
494 * track_pfn_insert is called when a _new_ single pfn is established
495 * by vm_insert_pfn().
497 static inline int track_pfn_insert(struct vm_area_struct
*vma
, pgprot_t
*prot
,
504 * track_pfn_copy is called when vma that is covering the pfnmap gets
505 * copied through copy_page_range().
507 static inline int track_pfn_copy(struct vm_area_struct
*vma
)
513 * untrack_pfn_vma is called while unmapping a pfnmap for a region.
514 * untrack can be called for a specific region indicated by pfn and size or
515 * can be for the entire vma (in which case pfn, size are zero).
517 static inline void untrack_pfn(struct vm_area_struct
*vma
,
518 unsigned long pfn
, unsigned long size
)
522 extern int track_pfn_remap(struct vm_area_struct
*vma
, pgprot_t
*prot
,
523 unsigned long pfn
, unsigned long addr
,
525 extern int track_pfn_insert(struct vm_area_struct
*vma
, pgprot_t
*prot
,
527 extern int track_pfn_copy(struct vm_area_struct
*vma
);
528 extern void untrack_pfn(struct vm_area_struct
*vma
, unsigned long pfn
,
532 #ifdef __HAVE_COLOR_ZERO_PAGE
533 static inline int is_zero_pfn(unsigned long pfn
)
535 extern unsigned long zero_pfn
;
536 unsigned long offset_from_zero_pfn
= pfn
- zero_pfn
;
537 return offset_from_zero_pfn
<= (zero_page_mask
>> PAGE_SHIFT
);
540 #define my_zero_pfn(addr) page_to_pfn(ZERO_PAGE(addr))
543 static inline int is_zero_pfn(unsigned long pfn
)
545 extern unsigned long zero_pfn
;
546 return pfn
== zero_pfn
;
549 static inline unsigned long my_zero_pfn(unsigned long addr
)
551 extern unsigned long zero_pfn
;
558 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
559 static inline int pmd_trans_huge(pmd_t pmd
)
563 static inline int pmd_trans_splitting(pmd_t pmd
)
567 #ifndef __HAVE_ARCH_PMD_WRITE
568 static inline int pmd_write(pmd_t pmd
)
573 #endif /* __HAVE_ARCH_PMD_WRITE */
574 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
576 #ifndef pmd_read_atomic
577 static inline pmd_t
pmd_read_atomic(pmd_t
*pmdp
)
580 * Depend on compiler for an atomic pmd read. NOTE: this is
581 * only going to work, if the pmdval_t isn't larger than
588 #ifndef pmd_move_must_withdraw
589 static inline int pmd_move_must_withdraw(spinlock_t
*new_pmd_ptl
,
590 spinlock_t
*old_pmd_ptl
)
593 * With split pmd lock we also need to move preallocated
594 * PTE page table if new_pmd is on different PMD page table.
596 return new_pmd_ptl
!= old_pmd_ptl
;
601 * This function is meant to be used by sites walking pagetables with
602 * the mmap_sem hold in read mode to protect against MADV_DONTNEED and
603 * transhuge page faults. MADV_DONTNEED can convert a transhuge pmd
604 * into a null pmd and the transhuge page fault can convert a null pmd
605 * into an hugepmd or into a regular pmd (if the hugepage allocation
606 * fails). While holding the mmap_sem in read mode the pmd becomes
607 * stable and stops changing under us only if it's not null and not a
608 * transhuge pmd. When those races occurs and this function makes a
609 * difference vs the standard pmd_none_or_clear_bad, the result is
610 * undefined so behaving like if the pmd was none is safe (because it
611 * can return none anyway). The compiler level barrier() is critically
612 * important to compute the two checks atomically on the same pmdval.
614 * For 32bit kernels with a 64bit large pmd_t this automatically takes
615 * care of reading the pmd atomically to avoid SMP race conditions
616 * against pmd_populate() when the mmap_sem is hold for reading by the
617 * caller (a special atomic read not done by "gcc" as in the generic
618 * version above, is also needed when THP is disabled because the page
619 * fault can populate the pmd from under us).
621 static inline int pmd_none_or_trans_huge_or_clear_bad(pmd_t
*pmd
)
623 pmd_t pmdval
= pmd_read_atomic(pmd
);
625 * The barrier will stabilize the pmdval in a register or on
626 * the stack so that it will stop changing under the code.
628 * When CONFIG_TRANSPARENT_HUGEPAGE=y on x86 32bit PAE,
629 * pmd_read_atomic is allowed to return a not atomic pmdval
630 * (for example pointing to an hugepage that has never been
631 * mapped in the pmd). The below checks will only care about
632 * the low part of the pmd with 32bit PAE x86 anyway, with the
633 * exception of pmd_none(). So the important thing is that if
634 * the low part of the pmd is found null, the high part will
635 * be also null or the pmd_none() check below would be
638 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
641 if (pmd_none(pmdval
) || pmd_trans_huge(pmdval
))
643 if (unlikely(pmd_bad(pmdval
))) {
651 * This is a noop if Transparent Hugepage Support is not built into
652 * the kernel. Otherwise it is equivalent to
653 * pmd_none_or_trans_huge_or_clear_bad(), and shall only be called in
654 * places that already verified the pmd is not none and they want to
655 * walk ptes while holding the mmap sem in read mode (write mode don't
656 * need this). If THP is not enabled, the pmd can't go away under the
657 * code even if MADV_DONTNEED runs, but if THP is enabled we need to
658 * run a pmd_trans_unstable before walking the ptes after
659 * split_huge_page_pmd returns (because it may have run when the pmd
660 * become null, but then a page fault can map in a THP and not a
663 static inline int pmd_trans_unstable(pmd_t
*pmd
)
665 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
666 return pmd_none_or_trans_huge_or_clear_bad(pmd
);
672 #ifndef CONFIG_NUMA_BALANCING
674 * Technically a PTE can be PROTNONE even when not doing NUMA balancing but
675 * the only case the kernel cares is for NUMA balancing and is only ever set
676 * when the VMA is accessible. For PROT_NONE VMAs, the PTEs are not marked
677 * _PAGE_PROTNONE so by by default, implement the helper as "always no". It
678 * is the responsibility of the caller to distinguish between PROT_NONE
679 * protections and NUMA hinting fault protections.
681 static inline int pte_protnone(pte_t pte
)
686 static inline int pmd_protnone(pmd_t pmd
)
690 #endif /* CONFIG_NUMA_BALANCING */
692 #endif /* CONFIG_MMU */
694 #endif /* !__ASSEMBLY__ */
696 #ifndef io_remap_pfn_range
697 #define io_remap_pfn_range remap_pfn_range
700 #endif /* _ASM_GENERIC_PGTABLE_H */
This page took 0.058067 seconds and 5 git commands to generate.