1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_SWAPOPS_H
3 #define _LINUX_SWAPOPS_H
5 #include <linux/radix-tree.h>
7 #include <linux/mm_types.h>
12 #include <linux/swapfile.h>
13 #endif /* CONFIG_SWAP */
16 * swapcache pages are stored in the swapper_space radix tree. We want to
17 * get good packing density in that tree, so the index should be dense in
20 * We arrange the `type' and `offset' fields so that `type' is at the six
21 * high-order bits of the swp_entry_t and `offset' is right-aligned in the
22 * remaining bits. Although `type' itself needs only five bits, we allow for
23 * shmem/tmpfs to shift it all up a further one bit: see swp_to_radix_entry().
25 * swp_entry_t's are *never* stored anywhere in their arch-dependent format.
27 #define SWP_TYPE_SHIFT (BITS_PER_XA_VALUE - MAX_SWAPFILES_SHIFT)
28 #define SWP_OFFSET_MASK ((1UL << SWP_TYPE_SHIFT) - 1)
31 * Definitions only for PFN swap entries (see is_pfn_swap_entry()). To
32 * store PFN, we only need SWP_PFN_BITS bits. Each of the pfn swap entries
33 * can use the extra bits to store other information besides PFN.
35 #ifdef MAX_PHYSMEM_BITS
36 #define SWP_PFN_BITS (MAX_PHYSMEM_BITS - PAGE_SHIFT)
37 #else /* MAX_PHYSMEM_BITS */
38 #define SWP_PFN_BITS min_t(int, \
39 sizeof(phys_addr_t) * 8 - PAGE_SHIFT, \
41 #endif /* MAX_PHYSMEM_BITS */
42 #define SWP_PFN_MASK (BIT(SWP_PFN_BITS) - 1)
45 * Migration swap entry specific bitfield definitions. Layout:
47 * |----------+--------------------|
48 * | swp_type | swp_offset |
49 * |----------+--------+-+-+-------|
50 * | | resv |D|A| PFN |
51 * |----------+--------+-+-+-------|
53 * @SWP_MIG_YOUNG_BIT: Whether the page used to have young bit set (bit A)
54 * @SWP_MIG_DIRTY_BIT: Whether the page used to have dirty bit set (bit D)
56 * Note: A/D bits will be stored in migration entries iff there're enough
57 * free bits in arch specific swp offset. By default we'll ignore A/D bits
58 * when migrating a page. Please refer to migration_entry_supports_ad()
59 * for more information. If there're more bits besides PFN and A/D bits,
60 * they should be reserved and always be zeros.
62 #define SWP_MIG_YOUNG_BIT (SWP_PFN_BITS)
63 #define SWP_MIG_DIRTY_BIT (SWP_PFN_BITS + 1)
64 #define SWP_MIG_TOTAL_BITS (SWP_PFN_BITS + 2)
66 #define SWP_MIG_YOUNG BIT(SWP_MIG_YOUNG_BIT)
67 #define SWP_MIG_DIRTY BIT(SWP_MIG_DIRTY_BIT)
69 static inline bool is_pfn_swap_entry(swp_entry_t entry);
71 /* Clear all flags but only keep swp_entry_t related information */
72 static inline pte_t pte_swp_clear_flags(pte_t pte)
74 if (pte_swp_exclusive(pte))
75 pte = pte_swp_clear_exclusive(pte);
76 if (pte_swp_soft_dirty(pte))
77 pte = pte_swp_clear_soft_dirty(pte);
78 if (pte_swp_uffd_wp(pte))
79 pte = pte_swp_clear_uffd_wp(pte);
84 * Store a type+offset into a swp_entry_t in an arch-independent format
86 static inline swp_entry_t swp_entry(unsigned long type, pgoff_t offset)
90 ret.val = (type << SWP_TYPE_SHIFT) | (offset & SWP_OFFSET_MASK);
95 * Extract the `type' field from a swp_entry_t. The swp_entry_t is in
96 * arch-independent format
98 static inline unsigned swp_type(swp_entry_t entry)
100 return (entry.val >> SWP_TYPE_SHIFT);
104 * Extract the `offset' field from a swp_entry_t. The swp_entry_t is in
105 * arch-independent format
107 static inline pgoff_t swp_offset(swp_entry_t entry)
109 return entry.val & SWP_OFFSET_MASK;
113 * This should only be called upon a pfn swap entry to get the PFN stored
114 * in the swap entry. Please refers to is_pfn_swap_entry() for definition
117 static inline unsigned long swp_offset_pfn(swp_entry_t entry)
119 VM_BUG_ON(!is_pfn_swap_entry(entry));
120 return swp_offset(entry) & SWP_PFN_MASK;
123 /* check whether a pte points to a swap entry */
124 static inline int is_swap_pte(pte_t pte)
126 return !pte_none(pte) && !pte_present(pte);
130 * Convert the arch-dependent pte representation of a swp_entry_t into an
131 * arch-independent swp_entry_t.
133 static inline swp_entry_t pte_to_swp_entry(pte_t pte)
135 swp_entry_t arch_entry;
137 pte = pte_swp_clear_flags(pte);
138 arch_entry = __pte_to_swp_entry(pte);
139 return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry));
143 * Convert the arch-independent representation of a swp_entry_t into the
144 * arch-dependent pte representation.
146 static inline pte_t swp_entry_to_pte(swp_entry_t entry)
148 swp_entry_t arch_entry;
150 arch_entry = __swp_entry(swp_type(entry), swp_offset(entry));
151 return __swp_entry_to_pte(arch_entry);
154 static inline swp_entry_t radix_to_swp_entry(void *arg)
158 entry.val = xa_to_value(arg);
162 static inline void *swp_to_radix_entry(swp_entry_t entry)
164 return xa_mk_value(entry.val);
167 #if IS_ENABLED(CONFIG_DEVICE_PRIVATE)
168 static inline swp_entry_t make_readable_device_private_entry(pgoff_t offset)
170 return swp_entry(SWP_DEVICE_READ, offset);
173 static inline swp_entry_t make_writable_device_private_entry(pgoff_t offset)
175 return swp_entry(SWP_DEVICE_WRITE, offset);
178 static inline bool is_device_private_entry(swp_entry_t entry)
180 int type = swp_type(entry);
181 return type == SWP_DEVICE_READ || type == SWP_DEVICE_WRITE;
184 static inline bool is_writable_device_private_entry(swp_entry_t entry)
186 return unlikely(swp_type(entry) == SWP_DEVICE_WRITE);
189 static inline swp_entry_t make_readable_device_exclusive_entry(pgoff_t offset)
191 return swp_entry(SWP_DEVICE_EXCLUSIVE_READ, offset);
194 static inline swp_entry_t make_writable_device_exclusive_entry(pgoff_t offset)
196 return swp_entry(SWP_DEVICE_EXCLUSIVE_WRITE, offset);
199 static inline bool is_device_exclusive_entry(swp_entry_t entry)
201 return swp_type(entry) == SWP_DEVICE_EXCLUSIVE_READ ||
202 swp_type(entry) == SWP_DEVICE_EXCLUSIVE_WRITE;
205 static inline bool is_writable_device_exclusive_entry(swp_entry_t entry)
207 return unlikely(swp_type(entry) == SWP_DEVICE_EXCLUSIVE_WRITE);
209 #else /* CONFIG_DEVICE_PRIVATE */
210 static inline swp_entry_t make_readable_device_private_entry(pgoff_t offset)
212 return swp_entry(0, 0);
215 static inline swp_entry_t make_writable_device_private_entry(pgoff_t offset)
217 return swp_entry(0, 0);
220 static inline bool is_device_private_entry(swp_entry_t entry)
225 static inline bool is_writable_device_private_entry(swp_entry_t entry)
230 static inline swp_entry_t make_readable_device_exclusive_entry(pgoff_t offset)
232 return swp_entry(0, 0);
235 static inline swp_entry_t make_writable_device_exclusive_entry(pgoff_t offset)
237 return swp_entry(0, 0);
240 static inline bool is_device_exclusive_entry(swp_entry_t entry)
245 static inline bool is_writable_device_exclusive_entry(swp_entry_t entry)
249 #endif /* CONFIG_DEVICE_PRIVATE */
251 #ifdef CONFIG_MIGRATION
252 static inline int is_migration_entry(swp_entry_t entry)
254 return unlikely(swp_type(entry) == SWP_MIGRATION_READ ||
255 swp_type(entry) == SWP_MIGRATION_READ_EXCLUSIVE ||
256 swp_type(entry) == SWP_MIGRATION_WRITE);
259 static inline int is_writable_migration_entry(swp_entry_t entry)
261 return unlikely(swp_type(entry) == SWP_MIGRATION_WRITE);
264 static inline int is_readable_migration_entry(swp_entry_t entry)
266 return unlikely(swp_type(entry) == SWP_MIGRATION_READ);
269 static inline int is_readable_exclusive_migration_entry(swp_entry_t entry)
271 return unlikely(swp_type(entry) == SWP_MIGRATION_READ_EXCLUSIVE);
274 static inline swp_entry_t make_readable_migration_entry(pgoff_t offset)
276 return swp_entry(SWP_MIGRATION_READ, offset);
279 static inline swp_entry_t make_readable_exclusive_migration_entry(pgoff_t offset)
281 return swp_entry(SWP_MIGRATION_READ_EXCLUSIVE, offset);
284 static inline swp_entry_t make_writable_migration_entry(pgoff_t offset)
286 return swp_entry(SWP_MIGRATION_WRITE, offset);
290 * Returns whether the host has large enough swap offset field to support
291 * carrying over pgtable A/D bits for page migrations. The result is
292 * pretty much arch specific.
294 static inline bool migration_entry_supports_ad(void)
297 return swap_migration_ad_supported;
298 #else /* CONFIG_SWAP */
300 #endif /* CONFIG_SWAP */
303 static inline swp_entry_t make_migration_entry_young(swp_entry_t entry)
305 if (migration_entry_supports_ad())
306 return swp_entry(swp_type(entry),
307 swp_offset(entry) | SWP_MIG_YOUNG);
311 static inline bool is_migration_entry_young(swp_entry_t entry)
313 if (migration_entry_supports_ad())
314 return swp_offset(entry) & SWP_MIG_YOUNG;
315 /* Keep the old behavior of aging page after migration */
319 static inline swp_entry_t make_migration_entry_dirty(swp_entry_t entry)
321 if (migration_entry_supports_ad())
322 return swp_entry(swp_type(entry),
323 swp_offset(entry) | SWP_MIG_DIRTY);
327 static inline bool is_migration_entry_dirty(swp_entry_t entry)
329 if (migration_entry_supports_ad())
330 return swp_offset(entry) & SWP_MIG_DIRTY;
331 /* Keep the old behavior of clean page after migration */
335 extern void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
336 unsigned long address);
337 extern void migration_entry_wait_huge(struct vm_area_struct *vma, pte_t *pte);
338 #else /* CONFIG_MIGRATION */
339 static inline swp_entry_t make_readable_migration_entry(pgoff_t offset)
341 return swp_entry(0, 0);
344 static inline swp_entry_t make_readable_exclusive_migration_entry(pgoff_t offset)
346 return swp_entry(0, 0);
349 static inline swp_entry_t make_writable_migration_entry(pgoff_t offset)
351 return swp_entry(0, 0);
354 static inline int is_migration_entry(swp_entry_t swp)
359 static inline void migration_entry_wait(struct mm_struct *mm, pmd_t *pmd,
360 unsigned long address) { }
361 static inline void migration_entry_wait_huge(struct vm_area_struct *vma,
363 static inline int is_writable_migration_entry(swp_entry_t entry)
367 static inline int is_readable_migration_entry(swp_entry_t entry)
372 static inline swp_entry_t make_migration_entry_young(swp_entry_t entry)
377 static inline bool is_migration_entry_young(swp_entry_t entry)
382 static inline swp_entry_t make_migration_entry_dirty(swp_entry_t entry)
387 static inline bool is_migration_entry_dirty(swp_entry_t entry)
391 #endif /* CONFIG_MIGRATION */
393 typedef unsigned long pte_marker;
395 #define PTE_MARKER_UFFD_WP BIT(0)
396 #define PTE_MARKER_SWAPIN_ERROR BIT(1)
397 #define PTE_MARKER_MASK (BIT(2) - 1)
399 static inline swp_entry_t make_pte_marker_entry(pte_marker marker)
401 return swp_entry(SWP_PTE_MARKER, marker);
404 static inline bool is_pte_marker_entry(swp_entry_t entry)
406 return swp_type(entry) == SWP_PTE_MARKER;
409 static inline pte_marker pte_marker_get(swp_entry_t entry)
411 return swp_offset(entry) & PTE_MARKER_MASK;
414 static inline bool is_pte_marker(pte_t pte)
416 return is_swap_pte(pte) && is_pte_marker_entry(pte_to_swp_entry(pte));
419 static inline pte_t make_pte_marker(pte_marker marker)
421 return swp_entry_to_pte(make_pte_marker_entry(marker));
424 static inline swp_entry_t make_swapin_error_entry(void)
426 return make_pte_marker_entry(PTE_MARKER_SWAPIN_ERROR);
429 static inline int is_swapin_error_entry(swp_entry_t entry)
431 return is_pte_marker_entry(entry) &&
432 (pte_marker_get(entry) & PTE_MARKER_SWAPIN_ERROR);
436 * This is a special version to check pte_none() just to cover the case when
437 * the pte is a pte marker. It existed because in many cases the pte marker
438 * should be seen as a none pte; it's just that we have stored some information
439 * onto the none pte so it becomes not-none any more.
441 * It should be used when the pte is file-backed, ram-based and backing
442 * userspace pages, like shmem. It is not needed upon pgtables that do not
443 * support pte markers at all. For example, it's not needed on anonymous
444 * memory, kernel-only memory (including when the system is during-boot),
445 * non-ram based generic file-system. It's fine to be used even there, but the
446 * extra pte marker check will be pure overhead.
448 static inline int pte_none_mostly(pte_t pte)
450 return pte_none(pte) || is_pte_marker(pte);
453 static inline struct page *pfn_swap_entry_to_page(swp_entry_t entry)
455 struct page *p = pfn_to_page(swp_offset_pfn(entry));
458 * Any use of migration entries may only occur while the
459 * corresponding page is locked
461 BUG_ON(is_migration_entry(entry) && !PageLocked(p));
467 * A pfn swap entry is a special type of swap entry that always has a pfn stored
468 * in the swap offset. They are used to represent unaddressable device memory
469 * and to restrict access to a page undergoing migration.
471 static inline bool is_pfn_swap_entry(swp_entry_t entry)
473 /* Make sure the swp offset can always store the needed fields */
474 BUILD_BUG_ON(SWP_TYPE_SHIFT < SWP_PFN_BITS);
476 return is_migration_entry(entry) || is_device_private_entry(entry) ||
477 is_device_exclusive_entry(entry);
480 struct page_vma_mapped_walk;
482 #ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
483 extern int set_pmd_migration_entry(struct page_vma_mapped_walk *pvmw,
486 extern void remove_migration_pmd(struct page_vma_mapped_walk *pvmw,
489 extern void pmd_migration_entry_wait(struct mm_struct *mm, pmd_t *pmd);
491 static inline swp_entry_t pmd_to_swp_entry(pmd_t pmd)
493 swp_entry_t arch_entry;
495 if (pmd_swp_soft_dirty(pmd))
496 pmd = pmd_swp_clear_soft_dirty(pmd);
497 if (pmd_swp_uffd_wp(pmd))
498 pmd = pmd_swp_clear_uffd_wp(pmd);
499 arch_entry = __pmd_to_swp_entry(pmd);
500 return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry));
503 static inline pmd_t swp_entry_to_pmd(swp_entry_t entry)
505 swp_entry_t arch_entry;
507 arch_entry = __swp_entry(swp_type(entry), swp_offset(entry));
508 return __swp_entry_to_pmd(arch_entry);
511 static inline int is_pmd_migration_entry(pmd_t pmd)
513 return is_swap_pmd(pmd) && is_migration_entry(pmd_to_swp_entry(pmd));
515 #else /* CONFIG_ARCH_ENABLE_THP_MIGRATION */
516 static inline int set_pmd_migration_entry(struct page_vma_mapped_walk *pvmw,
522 static inline void remove_migration_pmd(struct page_vma_mapped_walk *pvmw,
528 static inline void pmd_migration_entry_wait(struct mm_struct *m, pmd_t *p) { }
530 static inline swp_entry_t pmd_to_swp_entry(pmd_t pmd)
532 return swp_entry(0, 0);
535 static inline pmd_t swp_entry_to_pmd(swp_entry_t entry)
540 static inline int is_pmd_migration_entry(pmd_t pmd)
544 #endif /* CONFIG_ARCH_ENABLE_THP_MIGRATION */
546 #ifdef CONFIG_MEMORY_FAILURE
549 * Support for hardware poisoned pages
551 static inline swp_entry_t make_hwpoison_entry(struct page *page)
553 BUG_ON(!PageLocked(page));
554 return swp_entry(SWP_HWPOISON, page_to_pfn(page));
557 static inline int is_hwpoison_entry(swp_entry_t entry)
559 return swp_type(entry) == SWP_HWPOISON;
564 static inline swp_entry_t make_hwpoison_entry(struct page *page)
566 return swp_entry(0, 0);
569 static inline int is_hwpoison_entry(swp_entry_t swp)
575 static inline int non_swap_entry(swp_entry_t entry)
577 return swp_type(entry) >= MAX_SWAPFILES;
580 #endif /* CONFIG_MMU */
581 #endif /* _LINUX_SWAPOPS_H */