]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/page-flags.h
f2fs: fix deadlock between quota writes and checkpoint
[mirror_ubuntu-jammy-kernel.git] / include / linux / page-flags.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4
LT
2/*
3 * Macros for manipulating and testing page->flags
4 */
5
6#ifndef PAGE_FLAGS_H
7#define PAGE_FLAGS_H
8
f886ed44 9#include <linux/types.h>
187f1882 10#include <linux/bug.h>
072bb0aa 11#include <linux/mmdebug.h>
9223b419 12#ifndef __GENERATING_BOUNDS_H
6d777953 13#include <linux/mm_types.h>
01fc0ac1 14#include <generated/bounds.h>
9223b419 15#endif /* !__GENERATING_BOUNDS_H */
f886ed44 16
1da177e4
LT
17/*
18 * Various page->flags bits:
19 *
6e2e07cd
DH
20 * PG_reserved is set for special pages. The "struct page" of such a page
21 * should in general not be touched (e.g. set dirty) except by its owner.
22 * Pages marked as PG_reserved include:
23 * - Pages part of the kernel image (including vDSO) and similar (e.g. BIOS,
24 * initrd, HW tables)
25 * - Pages reserved or allocated early during boot (before the page allocator
26 * was initialized). This includes (depending on the architecture) the
27 * initial vmemmap, initial page tables, crashkernel, elfcorehdr, and much
28 * much more. Once (if ever) freed, PG_reserved is cleared and they will
29 * be given to the page allocator.
30 * - Pages falling into physical memory gaps - not IORESOURCE_SYSRAM. Trying
31 * to read/write these pages might end badly. Don't touch!
32 * - The zero page(s)
33 * - Pages not added to the page allocator when onlining a section because
34 * they were excluded via the online_page_callback() or because they are
35 * PG_hwpoison.
36 * - Pages allocated in the context of kexec/kdump (loaded kernel image,
37 * control pages, vmcoreinfo)
38 * - MMIO/DMA pages. Some architectures don't allow to ioremap pages that are
39 * not marked PG_reserved (as they might be in use by somebody else who does
40 * not respect the caching strategy).
41 * - Pages part of an offline section (struct pages of offline sections should
42 * not be trusted as they will be initialized when first onlined).
43 * - MCA pages on ia64
44 * - Pages holding CPU notes for POWER Firmware Assisted Dump
45 * - Device memory (e.g. PMEM, DAX, HMM)
46 * Some PG_reserved pages will be excluded from the hibernation image.
47 * PG_reserved does in general not hinder anybody from dumping or swapping
48 * and is no longer required for remap_pfn_range(). ioremap might require it.
49 * Consequently, PG_reserved for a page mapped into user space can indicate
50 * the zero page, the vDSO, MMIO pages or device memory.
1da177e4 51 *
da6052f7
NP
52 * The PG_private bitflag is set on pagecache pages if they contain filesystem
53 * specific data (which is normally at page->private). It can be used by
54 * private allocations for its own usage.
1da177e4 55 *
da6052f7
NP
56 * During initiation of disk I/O, PG_locked is set. This bit is set before I/O
57 * and cleared when writeback _starts_ or when read _completes_. PG_writeback
58 * is set before writeback starts and cleared when it finishes.
59 *
60 * PG_locked also pins a page in pagecache, and blocks truncation of the file
61 * while it is held.
62 *
63 * page_waitqueue(page) is a wait queue of all tasks waiting for the page
64 * to become unlocked.
1da177e4 65 *
9de4f22a
HY
66 * PG_swapbacked is set when a page uses swap as a backing storage. This are
67 * usually PageAnon or shmem pages but please note that even anonymous pages
68 * might lose their PG_swapbacked flag when they simply can be dropped (e.g. as
69 * a result of MADV_FREE).
70 *
1da177e4
LT
71 * PG_uptodate tells whether the page's contents is valid. When a read
72 * completes, the page becomes uptodate, unless a disk I/O error happened.
73 *
da6052f7
NP
74 * PG_referenced, PG_reclaim are used for page reclaim for anonymous and
75 * file-backed pagecache (see mm/vmscan.c).
1da177e4
LT
76 *
77 * PG_error is set to indicate that an I/O error occurred on this page.
78 *
79 * PG_arch_1 is an architecture specific page state bit. The generic code
80 * guarantees that this bit is cleared for a page when it first is entered into
81 * the page cache.
82 *
d466f2fc
AK
83 * PG_hwpoison indicates that a page got corrupted in hardware and contains
84 * data with incorrect ECC bits that triggered a machine check. Accessing is
85 * not safe since it may cause another machine check. Don't touch!
1da177e4
LT
86 */
87
88/*
89 * Don't use the *_dontuse flags. Use the macros. Otherwise you'll break
91fc8ab3
AW
90 * locked- and dirty-page accounting.
91 *
92 * The page flags field is split into two parts, the main flags area
93 * which extends from the low bits upwards, and the fields area which
94 * extends from the high bits downwards.
95 *
96 * | FIELD | ... | FLAGS |
9223b419
CL
97 * N-1 ^ 0
98 * (NR_PAGEFLAGS)
91fc8ab3 99 *
9223b419
CL
100 * The fields area is reserved for fields mapping zone, node (for NUMA) and
101 * SPARSEMEM section (for variants of SPARSEMEM that require section ids like
102 * SPARSEMEM_EXTREME with !SPARSEMEM_VMEMMAP).
1da177e4 103 */
e2683181
CL
104enum pageflags {
105 PG_locked, /* Page is locked. Don't touch. */
e2683181
CL
106 PG_referenced,
107 PG_uptodate,
108 PG_dirty,
109 PG_lru,
110 PG_active,
1899ad18 111 PG_workingset,
b91e1302 112 PG_waiters, /* Page has waiters, check its waitqueue. Must be bit #7 and in the same byte as "PG_locked" */
1899ad18 113 PG_error,
e2683181
CL
114 PG_slab,
115 PG_owner_priv_1, /* Owner use. If pagecache, fs may use*/
e2683181
CL
116 PG_arch_1,
117 PG_reserved,
118 PG_private, /* If pagecache, has fs-private data */
266cf658 119 PG_private_2, /* If pagecache, has fs aux data */
e2683181 120 PG_writeback, /* Page is under writeback */
e20b8cca 121 PG_head, /* A head page */
e2683181
CL
122 PG_mappedtodisk, /* Has blocks allocated on-disk */
123 PG_reclaim, /* To be reclaimed asap */
b2e18538 124 PG_swapbacked, /* Page is backed by RAM/swap */
894bc310 125 PG_unevictable, /* Page is "unevictable" */
af8e3354 126#ifdef CONFIG_MMU
b291f000 127 PG_mlocked, /* Page is vma mlocked */
894bc310 128#endif
46cf98cd 129#ifdef CONFIG_ARCH_USES_PG_UNCACHED
602c4d11 130 PG_uncached, /* Page has been mapped as uncached */
d466f2fc
AK
131#endif
132#ifdef CONFIG_MEMORY_FAILURE
133 PG_hwpoison, /* hardware poisoned page. Don't touch */
e9da73d6 134#endif
33c3fc71
VD
135#if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT)
136 PG_young,
137 PG_idle,
f886ed44 138#endif
0cad47cf
AW
139 __NR_PAGEFLAGS,
140
141 /* Filesystems */
142 PG_checked = PG_owner_priv_1,
143
6326fec1
NP
144 /* SwapBacked */
145 PG_swapcache = PG_owner_priv_1, /* Swap page: swp_entry_t in private */
146
266cf658
DH
147 /* Two page bits are conscripted by FS-Cache to maintain local caching
148 * state. These bits are set on pages belonging to the netfs's inodes
149 * when those inodes are being locally cached.
150 */
151 PG_fscache = PG_private_2, /* page backed by cache */
152
0cad47cf 153 /* XEN */
d8ac3dd4 154 /* Pinned in Xen as a read-only pagetable page. */
0cad47cf 155 PG_pinned = PG_owner_priv_1,
d8ac3dd4 156 /* Pinned as part of domain save (see xen_mm_pin_all()). */
0cad47cf 157 PG_savepinned = PG_dirty,
d8ac3dd4
JH
158 /* Has a grant mapping of another (foreign) domain's page. */
159 PG_foreign = PG_owner_priv_1,
b877ac98
JG
160 /* Remapped by swiotlb-xen. */
161 PG_xen_remapped = PG_owner_priv_1,
8a38082d 162
9023cb7e 163 /* SLOB */
9023cb7e 164 PG_slob_free = PG_private,
53f9263b
KS
165
166 /* Compound pages. Stored in first tail page's flags */
167 PG_double_map = PG_private_2,
bda807d4
MK
168
169 /* non-lru isolated movable page */
170 PG_isolated = PG_reclaim,
36e66c55
AD
171
172 /* Only valid for buddy pages. Used to track pages that are reported */
173 PG_reported = PG_uptodate,
e2683181 174};
1da177e4 175
9223b419
CL
176#ifndef __GENERATING_BOUNDS_H
177
0e6d31a7
KS
178struct page; /* forward declaration */
179
180static inline struct page *compound_head(struct page *page)
181{
182 unsigned long head = READ_ONCE(page->compound_head);
183
184 if (unlikely(head & 1))
185 return (struct page *) (head - 1);
186 return page;
187}
188
4b0f3261 189static __always_inline int PageTail(struct page *page)
0e6d31a7
KS
190{
191 return READ_ONCE(page->compound_head) & 1;
192}
193
4b0f3261 194static __always_inline int PageCompound(struct page *page)
0e6d31a7
KS
195{
196 return test_bit(PG_head, &page->flags) || PageTail(page);
197}
198
f165b378
PT
199#define PAGE_POISON_PATTERN -1l
200static inline int PagePoisoned(const struct page *page)
201{
202 return page->flags == PAGE_POISON_PATTERN;
203}
204
f682a97a
AD
205#ifdef CONFIG_DEBUG_VM
206void page_init_poison(struct page *page, size_t size);
207#else
208static inline void page_init_poison(struct page *page, size_t size)
209{
210}
211#endif
212
95ad9755
KS
213/*
214 * Page flags policies wrt compound pages
215 *
f165b378
PT
216 * PF_POISONED_CHECK
217 * check if this struct page poisoned/uninitialized
218 *
95ad9755
KS
219 * PF_ANY:
220 * the page flag is relevant for small, head and tail pages.
221 *
222 * PF_HEAD:
223 * for compound page all operations related to the page flag applied to
224 * head page.
225 *
62906027
NP
226 * PF_ONLY_HEAD:
227 * for compound page, callers only ever operate on the head page.
228 *
95ad9755
KS
229 * PF_NO_TAIL:
230 * modifications of the page flag must be done on small or head pages,
231 * checks can be done on tail pages too.
232 *
233 * PF_NO_COMPOUND:
234 * the page flag is not relevant for compound pages.
235 */
f165b378
PT
236#define PF_POISONED_CHECK(page) ({ \
237 VM_BUG_ON_PGFLAGS(PagePoisoned(page), page); \
238 page; })
239#define PF_ANY(page, enforce) PF_POISONED_CHECK(page)
240#define PF_HEAD(page, enforce) PF_POISONED_CHECK(compound_head(page))
62906027
NP
241#define PF_ONLY_HEAD(page, enforce) ({ \
242 VM_BUG_ON_PGFLAGS(PageTail(page), page); \
f165b378 243 PF_POISONED_CHECK(page); })
95ad9755
KS
244#define PF_NO_TAIL(page, enforce) ({ \
245 VM_BUG_ON_PGFLAGS(enforce && PageTail(page), page); \
f165b378 246 PF_POISONED_CHECK(compound_head(page)); })
822cdd11 247#define PF_NO_COMPOUND(page, enforce) ({ \
95ad9755 248 VM_BUG_ON_PGFLAGS(enforce && PageCompound(page), page); \
f165b378 249 PF_POISONED_CHECK(page); })
95ad9755 250
f94a62e9
CL
251/*
252 * Macros to create function definitions for page flags
253 */
95ad9755 254#define TESTPAGEFLAG(uname, lname, policy) \
4b0f3261 255static __always_inline int Page##uname(struct page *page) \
95ad9755 256 { return test_bit(PG_##lname, &policy(page, 0)->flags); }
f94a62e9 257
95ad9755 258#define SETPAGEFLAG(uname, lname, policy) \
4b0f3261 259static __always_inline void SetPage##uname(struct page *page) \
95ad9755 260 { set_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 261
95ad9755 262#define CLEARPAGEFLAG(uname, lname, policy) \
4b0f3261 263static __always_inline void ClearPage##uname(struct page *page) \
95ad9755 264 { clear_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 265
95ad9755 266#define __SETPAGEFLAG(uname, lname, policy) \
4b0f3261 267static __always_inline void __SetPage##uname(struct page *page) \
95ad9755 268 { __set_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 269
95ad9755 270#define __CLEARPAGEFLAG(uname, lname, policy) \
4b0f3261 271static __always_inline void __ClearPage##uname(struct page *page) \
95ad9755 272 { __clear_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 273
95ad9755 274#define TESTSETFLAG(uname, lname, policy) \
4b0f3261 275static __always_inline int TestSetPage##uname(struct page *page) \
95ad9755 276 { return test_and_set_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 277
95ad9755 278#define TESTCLEARFLAG(uname, lname, policy) \
4b0f3261 279static __always_inline int TestClearPage##uname(struct page *page) \
95ad9755 280 { return test_and_clear_bit(PG_##lname, &policy(page, 1)->flags); }
f94a62e9 281
95ad9755
KS
282#define PAGEFLAG(uname, lname, policy) \
283 TESTPAGEFLAG(uname, lname, policy) \
284 SETPAGEFLAG(uname, lname, policy) \
285 CLEARPAGEFLAG(uname, lname, policy)
f94a62e9 286
95ad9755
KS
287#define __PAGEFLAG(uname, lname, policy) \
288 TESTPAGEFLAG(uname, lname, policy) \
289 __SETPAGEFLAG(uname, lname, policy) \
290 __CLEARPAGEFLAG(uname, lname, policy)
f94a62e9 291
95ad9755
KS
292#define TESTSCFLAG(uname, lname, policy) \
293 TESTSETFLAG(uname, lname, policy) \
294 TESTCLEARFLAG(uname, lname, policy)
f94a62e9 295
2f3e442c
JW
296#define TESTPAGEFLAG_FALSE(uname) \
297static inline int Page##uname(const struct page *page) { return 0; }
298
8a7a8544
LS
299#define SETPAGEFLAG_NOOP(uname) \
300static inline void SetPage##uname(struct page *page) { }
301
302#define CLEARPAGEFLAG_NOOP(uname) \
303static inline void ClearPage##uname(struct page *page) { }
304
305#define __CLEARPAGEFLAG_NOOP(uname) \
306static inline void __ClearPage##uname(struct page *page) { }
307
2f3e442c
JW
308#define TESTSETFLAG_FALSE(uname) \
309static inline int TestSetPage##uname(struct page *page) { return 0; }
310
8a7a8544
LS
311#define TESTCLEARFLAG_FALSE(uname) \
312static inline int TestClearPage##uname(struct page *page) { return 0; }
313
2f3e442c
JW
314#define PAGEFLAG_FALSE(uname) TESTPAGEFLAG_FALSE(uname) \
315 SETPAGEFLAG_NOOP(uname) CLEARPAGEFLAG_NOOP(uname)
316
317#define TESTSCFLAG_FALSE(uname) \
318 TESTSETFLAG_FALSE(uname) TESTCLEARFLAG_FALSE(uname)
319
48c935ad 320__PAGEFLAG(Locked, locked, PF_NO_TAIL)
62906027 321PAGEFLAG(Waiters, waiters, PF_ONLY_HEAD) __CLEARPAGEFLAG(Waiters, waiters, PF_ONLY_HEAD)
d72520ad 322PAGEFLAG(Error, error, PF_NO_TAIL) TESTCLEARFLAG(Error, error, PF_NO_TAIL)
8cb38fab
KS
323PAGEFLAG(Referenced, referenced, PF_HEAD)
324 TESTCLEARFLAG(Referenced, referenced, PF_HEAD)
325 __SETPAGEFLAG(Referenced, referenced, PF_HEAD)
df8c94d1
KS
326PAGEFLAG(Dirty, dirty, PF_HEAD) TESTSCFLAG(Dirty, dirty, PF_HEAD)
327 __CLEARPAGEFLAG(Dirty, dirty, PF_HEAD)
8cb38fab
KS
328PAGEFLAG(LRU, lru, PF_HEAD) __CLEARPAGEFLAG(LRU, lru, PF_HEAD)
329PAGEFLAG(Active, active, PF_HEAD) __CLEARPAGEFLAG(Active, active, PF_HEAD)
330 TESTCLEARFLAG(Active, active, PF_HEAD)
1899ad18
JW
331PAGEFLAG(Workingset, workingset, PF_HEAD)
332 TESTCLEARFLAG(Workingset, workingset, PF_HEAD)
dcb351cd
KS
333__PAGEFLAG(Slab, slab, PF_NO_TAIL)
334__PAGEFLAG(SlobFree, slob_free, PF_NO_TAIL)
df8c94d1 335PAGEFLAG(Checked, checked, PF_NO_COMPOUND) /* Used by some filesystems */
c13985fa
KS
336
337/* Xen */
338PAGEFLAG(Pinned, pinned, PF_NO_COMPOUND)
339 TESTSCFLAG(Pinned, pinned, PF_NO_COMPOUND)
340PAGEFLAG(SavePinned, savepinned, PF_NO_COMPOUND);
341PAGEFLAG(Foreign, foreign, PF_NO_COMPOUND);
b877ac98
JG
342PAGEFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
343 TESTCLEARFLAG(XenRemapped, xen_remapped, PF_NO_COMPOUND)
c13985fa 344
de09d31d
KS
345PAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
346 __CLEARPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
d483da5b 347 __SETPAGEFLAG(Reserved, reserved, PF_NO_COMPOUND)
da5efc40
KS
348PAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
349 __CLEARPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
350 __SETPAGEFLAG(SwapBacked, swapbacked, PF_NO_TAIL)
95ad9755 351
266cf658
DH
352/*
353 * Private page markings that may be used by the filesystem that owns the page
354 * for its own purposes.
355 * - PG_private and PG_private_2 cause releasepage() and co to be invoked
356 */
95ad9755
KS
357PAGEFLAG(Private, private, PF_ANY) __SETPAGEFLAG(Private, private, PF_ANY)
358 __CLEARPAGEFLAG(Private, private, PF_ANY)
359PAGEFLAG(Private2, private_2, PF_ANY) TESTSCFLAG(Private2, private_2, PF_ANY)
360PAGEFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
361 TESTCLEARFLAG(OwnerPriv1, owner_priv_1, PF_ANY)
266cf658 362
6a1e7f77
CL
363/*
364 * Only test-and-set exist for PG_writeback. The unconditional operators are
365 * risky: they bypass page accounting.
366 */
225311a4
HY
367TESTPAGEFLAG(Writeback, writeback, PF_NO_TAIL)
368 TESTSCFLAG(Writeback, writeback, PF_NO_TAIL)
e2f0a0db 369PAGEFLAG(MappedToDisk, mappedtodisk, PF_NO_TAIL)
6a1e7f77 370
579f8290 371/* PG_readahead is only used for reads; PG_reclaim is only for writes */
e2f0a0db
KS
372PAGEFLAG(Reclaim, reclaim, PF_NO_TAIL)
373 TESTCLEARFLAG(Reclaim, reclaim, PF_NO_TAIL)
df8c94d1
KS
374PAGEFLAG(Readahead, reclaim, PF_NO_COMPOUND)
375 TESTCLEARFLAG(Readahead, reclaim, PF_NO_COMPOUND)
6a1e7f77
CL
376
377#ifdef CONFIG_HIGHMEM
1da177e4 378/*
6a1e7f77
CL
379 * Must use a macro here due to header dependency issues. page_zone() is not
380 * available at this point.
1da177e4 381 */
3ca65c19 382#define PageHighMem(__p) is_highmem_idx(page_zonenum(__p))
6a1e7f77 383#else
ec7cade8 384PAGEFLAG_FALSE(HighMem)
6a1e7f77
CL
385#endif
386
387#ifdef CONFIG_SWAP
6326fec1
NP
388static __always_inline int PageSwapCache(struct page *page)
389{
38d8b4e6
HY
390#ifdef CONFIG_THP_SWAP
391 page = compound_head(page);
392#endif
6326fec1
NP
393 return PageSwapBacked(page) && test_bit(PG_swapcache, &page->flags);
394
395}
38d8b4e6
HY
396SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
397CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
6a1e7f77 398#else
ec7cade8 399PAGEFLAG_FALSE(SwapCache)
6a1e7f77
CL
400#endif
401
8cb38fab
KS
402PAGEFLAG(Unevictable, unevictable, PF_HEAD)
403 __CLEARPAGEFLAG(Unevictable, unevictable, PF_HEAD)
404 TESTCLEARFLAG(Unevictable, unevictable, PF_HEAD)
b291f000 405
af8e3354 406#ifdef CONFIG_MMU
e4f87d5d
KS
407PAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
408 __CLEARPAGEFLAG(Mlocked, mlocked, PF_NO_TAIL)
409 TESTSCFLAG(Mlocked, mlocked, PF_NO_TAIL)
894bc310 410#else
2f3e442c 411PAGEFLAG_FALSE(Mlocked) __CLEARPAGEFLAG_NOOP(Mlocked)
685eaade 412 TESTSCFLAG_FALSE(Mlocked)
894bc310
LS
413#endif
414
46cf98cd 415#ifdef CONFIG_ARCH_USES_PG_UNCACHED
b9d41817 416PAGEFLAG(Uncached, uncached, PF_NO_COMPOUND)
602c4d11 417#else
ec7cade8 418PAGEFLAG_FALSE(Uncached)
6a1e7f77 419#endif
1da177e4 420
d466f2fc 421#ifdef CONFIG_MEMORY_FAILURE
95ad9755
KS
422PAGEFLAG(HWPoison, hwpoison, PF_ANY)
423TESTSCFLAG(HWPoison, hwpoison, PF_ANY)
d466f2fc 424#define __PG_HWPOISON (1UL << PG_hwpoison)
d4ae9916 425extern bool set_hwpoison_free_buddy_page(struct page *page);
d466f2fc
AK
426#else
427PAGEFLAG_FALSE(HWPoison)
d4ae9916
NH
428static inline bool set_hwpoison_free_buddy_page(struct page *page)
429{
430 return 0;
431}
d466f2fc
AK
432#define __PG_HWPOISON 0
433#endif
434
33c3fc71 435#if defined(CONFIG_IDLE_PAGE_TRACKING) && defined(CONFIG_64BIT)
95ad9755
KS
436TESTPAGEFLAG(Young, young, PF_ANY)
437SETPAGEFLAG(Young, young, PF_ANY)
438TESTCLEARFLAG(Young, young, PF_ANY)
439PAGEFLAG(Idle, idle, PF_ANY)
33c3fc71
VD
440#endif
441
36e66c55
AD
442/*
443 * PageReported() is used to track reported free pages within the Buddy
444 * allocator. We can use the non-atomic version of the test and set
445 * operations as both should be shielded with the zone lock to prevent
446 * any possible races on the setting or clearing of the bit.
447 */
448__PAGEFLAG(Reported, reported, PF_NO_COMPOUND)
449
e8c6158f
KS
450/*
451 * On an anonymous page mapped into a user virtual memory area,
452 * page->mapping points to its anon_vma, not to a struct address_space;
453 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
454 *
455 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
bda807d4
MK
456 * the PAGE_MAPPING_MOVABLE bit may be set along with the PAGE_MAPPING_ANON
457 * bit; and then page->mapping points, not to an anon_vma, but to a private
e8c6158f
KS
458 * structure which KSM associates with that merged page. See ksm.h.
459 *
bda807d4
MK
460 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is used for non-lru movable
461 * page and then page->mapping points a struct address_space.
e8c6158f
KS
462 *
463 * Please note that, confusingly, "page_mapping" refers to the inode
464 * address_space which maps the page from disk; whereas "page_mapped"
465 * refers to user virtual address space into which the page is mapped.
466 */
bda807d4
MK
467#define PAGE_MAPPING_ANON 0x1
468#define PAGE_MAPPING_MOVABLE 0x2
469#define PAGE_MAPPING_KSM (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
470#define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_MOVABLE)
e8c6158f 471
bda807d4 472static __always_inline int PageMappingFlags(struct page *page)
17514574 473{
bda807d4 474 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) != 0;
17514574
MG
475}
476
4b0f3261 477static __always_inline int PageAnon(struct page *page)
e8c6158f 478{
822cdd11 479 page = compound_head(page);
bda807d4
MK
480 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
481}
482
483static __always_inline int __PageMovable(struct page *page)
484{
485 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
486 PAGE_MAPPING_MOVABLE;
e8c6158f
KS
487}
488
489#ifdef CONFIG_KSM
490/*
491 * A KSM page is one of those write-protected "shared pages" or "merged pages"
492 * which KSM maps into multiple mms, wherever identical anonymous page content
493 * is found in VM_MERGEABLE vmas. It's a PageAnon page, pointing not to any
494 * anon_vma, but to that page's node of the stable tree.
495 */
4b0f3261 496static __always_inline int PageKsm(struct page *page)
e8c6158f 497{
822cdd11 498 page = compound_head(page);
e8c6158f 499 return ((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) ==
bda807d4 500 PAGE_MAPPING_KSM;
e8c6158f
KS
501}
502#else
503TESTPAGEFLAG_FALSE(Ksm)
504#endif
505
1a9b5b7f
WF
506u64 stable_page_flags(struct page *page);
507
0ed361de
NP
508static inline int PageUptodate(struct page *page)
509{
d2998c4d
KS
510 int ret;
511 page = compound_head(page);
512 ret = test_bit(PG_uptodate, &(page)->flags);
0ed361de
NP
513 /*
514 * Must ensure that the data we read out of the page is loaded
515 * _after_ we've loaded page->flags to check for PageUptodate.
516 * We can skip the barrier if the page is not uptodate, because
517 * we wouldn't be reading anything from it.
518 *
519 * See SetPageUptodate() for the other side of the story.
520 */
521 if (ret)
522 smp_rmb();
523
524 return ret;
525}
526
4b0f3261 527static __always_inline void __SetPageUptodate(struct page *page)
0ed361de 528{
d2998c4d 529 VM_BUG_ON_PAGE(PageTail(page), page);
0ed361de 530 smp_wmb();
df8c94d1 531 __set_bit(PG_uptodate, &page->flags);
0ed361de
NP
532}
533
4b0f3261 534static __always_inline void SetPageUptodate(struct page *page)
2dcea57a 535{
d2998c4d 536 VM_BUG_ON_PAGE(PageTail(page), page);
0ed361de
NP
537 /*
538 * Memory barrier must be issued before setting the PG_uptodate bit,
539 * so that all previous stores issued in order to bring the page
540 * uptodate are actually visible before PageUptodate becomes true.
0ed361de
NP
541 */
542 smp_wmb();
df8c94d1 543 set_bit(PG_uptodate, &page->flags);
0ed361de
NP
544}
545
d2998c4d 546CLEARPAGEFLAG(Uptodate, uptodate, PF_NO_TAIL)
1da177e4 547
6a1e7f77 548int test_clear_page_writeback(struct page *page);
1c8349a1
NJ
549int __test_set_page_writeback(struct page *page, bool keep_write);
550
551#define test_set_page_writeback(page) \
552 __test_set_page_writeback(page, false)
553#define test_set_page_writeback_keepwrite(page) \
554 __test_set_page_writeback(page, true)
1da177e4 555
6a1e7f77
CL
556static inline void set_page_writeback(struct page *page)
557{
558 test_set_page_writeback(page);
559}
1da177e4 560
1c8349a1
NJ
561static inline void set_page_writeback_keepwrite(struct page *page)
562{
563 test_set_page_writeback_keepwrite(page);
564}
565
95ad9755 566__PAGEFLAG(Head, head, PF_ANY) CLEARPAGEFLAG(Head, head, PF_ANY)
e20b8cca 567
4b0f3261 568static __always_inline void set_compound_head(struct page *page, struct page *head)
ad4b3fb7 569{
1d798ca3 570 WRITE_ONCE(page->compound_head, (unsigned long)head + 1);
ad4b3fb7
CD
571}
572
4b0f3261 573static __always_inline void clear_compound_head(struct page *page)
6a1e7f77 574{
1d798ca3 575 WRITE_ONCE(page->compound_head, 0);
6a1e7f77 576}
6d777953 577
4e6af67e
AA
578#ifdef CONFIG_TRANSPARENT_HUGEPAGE
579static inline void ClearPageCompound(struct page *page)
580{
1d798ca3
KS
581 BUG_ON(!PageHead(page));
582 ClearPageHead(page);
4e6af67e
AA
583}
584#endif
585
d2a1a1f0 586#define PG_head_mask ((1UL << PG_head))
dfa7e20c 587
e8c6158f
KS
588#ifdef CONFIG_HUGETLB_PAGE
589int PageHuge(struct page *page);
590int PageHeadHuge(struct page *page);
7e1f049e 591bool page_huge_active(struct page *page);
e8c6158f
KS
592#else
593TESTPAGEFLAG_FALSE(Huge)
594TESTPAGEFLAG_FALSE(HeadHuge)
7e1f049e
NH
595
596static inline bool page_huge_active(struct page *page)
597{
598 return 0;
599}
e8c6158f
KS
600#endif
601
7e1f049e 602
936a5fe6 603#ifdef CONFIG_TRANSPARENT_HUGEPAGE
71e3aac0
AA
604/*
605 * PageHuge() only returns true for hugetlbfs pages, but not for
606 * normal or transparent huge pages.
607 *
608 * PageTransHuge() returns true for both transparent huge and
609 * hugetlbfs pages, but not normal pages. PageTransHuge() can only be
610 * called only in the core VM paths where hugetlbfs pages can't exist.
611 */
612static inline int PageTransHuge(struct page *page)
613{
309381fe 614 VM_BUG_ON_PAGE(PageTail(page), page);
71e3aac0
AA
615 return PageHead(page);
616}
617
385de357
DN
618/*
619 * PageTransCompound returns true for both transparent huge pages
620 * and hugetlbfs pages, so it should only be called when it's known
621 * that hugetlbfs pages aren't involved.
622 */
936a5fe6
AA
623static inline int PageTransCompound(struct page *page)
624{
625 return PageCompound(page);
626}
71e3aac0 627
127393fb
AA
628/*
629 * PageTransCompoundMap is the same as PageTransCompound, but it also
630 * guarantees the primary MMU has the entire compound page mapped
631 * through pmd_trans_huge, which in turn guarantees the secondary MMUs
632 * can also map the entire compound page. This allows the secondary
633 * MMUs to call get_user_pages() only once for each compound page and
634 * to immediately map the entire compound page with a single secondary
635 * MMU fault. If there will be a pmd split later, the secondary MMUs
636 * will get an update through the MMU notifier invalidation through
637 * split_huge_pmd().
638 *
639 * Unlike PageTransCompound, this is safe to be called only while
640 * split_huge_pmd() cannot run from under us, like if protected by the
169226f7 641 * MMU notifier, otherwise it may result in page->_mapcount check false
127393fb 642 * positives.
169226f7
YS
643 *
644 * We have to treat page cache THP differently since every subpage of it
645 * would get _mapcount inc'ed once it is PMD mapped. But, it may be PTE
646 * mapped in the current process so comparing subpage's _mapcount to
647 * compound_mapcount to filter out PTE mapped case.
127393fb
AA
648 */
649static inline int PageTransCompoundMap(struct page *page)
650{
169226f7
YS
651 struct page *head;
652
653 if (!PageTransCompound(page))
654 return 0;
655
656 if (PageAnon(page))
657 return atomic_read(&page->_mapcount) < 0;
658
659 head = compound_head(page);
660 /* File THP is PMD mapped and not PTE mapped */
661 return atomic_read(&page->_mapcount) ==
662 atomic_read(compound_mapcount_ptr(head));
127393fb
AA
663}
664
385de357
DN
665/*
666 * PageTransTail returns true for both transparent huge pages
667 * and hugetlbfs pages, so it should only be called when it's known
668 * that hugetlbfs pages aren't involved.
669 */
670static inline int PageTransTail(struct page *page)
671{
672 return PageTail(page);
673}
674
53f9263b
KS
675/*
676 * PageDoubleMap indicates that the compound page is mapped with PTEs as well
677 * as PMDs.
678 *
679 * This is required for optimization of rmap operations for THP: we can postpone
680 * per small page mapcount accounting (and its overhead from atomic operations)
681 * until the first PMD split.
682 *
683 * For the page PageDoubleMap means ->_mapcount in all sub-pages is offset up
684 * by one. This reference will go away with last compound_mapcount.
685 *
686 * See also __split_huge_pmd_locked() and page_remove_anon_compound_rmap().
687 */
688static inline int PageDoubleMap(struct page *page)
689{
690 return PageHead(page) && test_bit(PG_double_map, &page[1].flags);
691}
692
9a73f61b
KS
693static inline void SetPageDoubleMap(struct page *page)
694{
695 VM_BUG_ON_PAGE(!PageHead(page), page);
696 set_bit(PG_double_map, &page[1].flags);
697}
698
699static inline void ClearPageDoubleMap(struct page *page)
700{
701 VM_BUG_ON_PAGE(!PageHead(page), page);
702 clear_bit(PG_double_map, &page[1].flags);
703}
53f9263b
KS
704static inline int TestSetPageDoubleMap(struct page *page)
705{
706 VM_BUG_ON_PAGE(!PageHead(page), page);
707 return test_and_set_bit(PG_double_map, &page[1].flags);
708}
709
710static inline int TestClearPageDoubleMap(struct page *page)
711{
712 VM_BUG_ON_PAGE(!PageHead(page), page);
713 return test_and_clear_bit(PG_double_map, &page[1].flags);
714}
715
936a5fe6 716#else
d8c1bdeb
KS
717TESTPAGEFLAG_FALSE(TransHuge)
718TESTPAGEFLAG_FALSE(TransCompound)
127393fb 719TESTPAGEFLAG_FALSE(TransCompoundMap)
d8c1bdeb 720TESTPAGEFLAG_FALSE(TransTail)
9a73f61b 721PAGEFLAG_FALSE(DoubleMap)
53f9263b
KS
722 TESTSETFLAG_FALSE(DoubleMap)
723 TESTCLEARFLAG_FALSE(DoubleMap)
936a5fe6
AA
724#endif
725
e8c6158f 726/*
6e292b9b
MW
727 * For pages that are never mapped to userspace (and aren't PageSlab),
728 * page_type may be used. Because it is initialised to -1, we invert the
729 * sense of the bit, so __SetPageFoo *clears* the bit used for PageFoo, and
730 * __ClearPageFoo *sets* the bit used for PageFoo. We reserve a few high and
731 * low bits so that an underflow or overflow of page_mapcount() won't be
732 * mistaken for a page type value.
e8c6158f 733 */
6e292b9b
MW
734
735#define PAGE_TYPE_BASE 0xf0000000
736/* Reserve 0x0000007f to catch underflows of page_mapcount */
144552ff 737#define PAGE_MAPCOUNT_RESERVE -128
6e292b9b 738#define PG_buddy 0x00000080
ca215086 739#define PG_offline 0x00000100
6e292b9b 740#define PG_kmemcg 0x00000200
1d40a5ea 741#define PG_table 0x00000400
3972f6bb 742#define PG_guard 0x00000800
6e292b9b
MW
743
744#define PageType(page, flag) \
745 ((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
746
144552ff
AY
747static inline int page_has_type(struct page *page)
748{
749 return (int)page->page_type < PAGE_MAPCOUNT_RESERVE;
750}
751
6e292b9b 752#define PAGE_TYPE_OPS(uname, lname) \
632c0a1a
VD
753static __always_inline int Page##uname(struct page *page) \
754{ \
6e292b9b 755 return PageType(page, PG_##lname); \
632c0a1a
VD
756} \
757static __always_inline void __SetPage##uname(struct page *page) \
758{ \
6e292b9b
MW
759 VM_BUG_ON_PAGE(!PageType(page, 0), page); \
760 page->page_type &= ~PG_##lname; \
632c0a1a
VD
761} \
762static __always_inline void __ClearPage##uname(struct page *page) \
763{ \
764 VM_BUG_ON_PAGE(!Page##uname(page), page); \
6e292b9b 765 page->page_type |= PG_##lname; \
e8c6158f
KS
766}
767
632c0a1a 768/*
6e292b9b 769 * PageBuddy() indicates that the page is free and in the buddy system
632c0a1a
VD
770 * (see mm/page_alloc.c).
771 */
6e292b9b 772PAGE_TYPE_OPS(Buddy, buddy)
e8c6158f 773
632c0a1a 774/*
ca215086
DH
775 * PageOffline() indicates that the page is logically offline although the
776 * containing section is online. (e.g. inflated in a balloon driver or
777 * not onlined when onlining the section).
778 * The content of these pages is effectively stale. Such pages should not
779 * be touched (read/write/dump/save) except by their owner.
aa218795
DH
780 *
781 * If a driver wants to allow to offline unmovable PageOffline() pages without
782 * putting them back to the buddy, it can do so via the memory notifier by
783 * decrementing the reference count in MEM_GOING_OFFLINE and incrementing the
784 * reference count in MEM_CANCEL_OFFLINE. When offlining, the PageOffline()
785 * pages (now with a reference count of zero) are treated like free pages,
786 * allowing the containing memory block to get offlined. A driver that
787 * relies on this feature is aware that re-onlining the memory block will
788 * require to re-set the pages PageOffline() and not giving them to the
789 * buddy via online_page_callback_t.
632c0a1a 790 */
ca215086 791PAGE_TYPE_OPS(Offline, offline)
e8c6158f 792
4949148a
VD
793/*
794 * If kmemcg is enabled, the buddy allocator will set PageKmemcg() on
795 * pages allocated with __GFP_ACCOUNT. It gets cleared on page free.
796 */
6e292b9b 797PAGE_TYPE_OPS(Kmemcg, kmemcg)
4949148a 798
1d40a5ea
MW
799/*
800 * Marks pages in use as page tables.
801 */
802PAGE_TYPE_OPS(Table, table)
803
3972f6bb
VB
804/*
805 * Marks guardpages used with debug_pagealloc.
806 */
807PAGE_TYPE_OPS(Guard, guard)
808
832fc1de
NH
809extern bool is_free_buddy_page(struct page *page);
810
bda807d4
MK
811__PAGEFLAG(Isolated, isolated, PF_ANY);
812
072bb0aa
MG
813/*
814 * If network-based swap is enabled, sl*b must keep track of whether pages
815 * were allocated from pfmemalloc reserves.
816 */
817static inline int PageSlabPfmemalloc(struct page *page)
818{
309381fe 819 VM_BUG_ON_PAGE(!PageSlab(page), page);
072bb0aa
MG
820 return PageActive(page);
821}
822
823static inline void SetPageSlabPfmemalloc(struct page *page)
824{
309381fe 825 VM_BUG_ON_PAGE(!PageSlab(page), page);
072bb0aa
MG
826 SetPageActive(page);
827}
828
829static inline void __ClearPageSlabPfmemalloc(struct page *page)
830{
309381fe 831 VM_BUG_ON_PAGE(!PageSlab(page), page);
072bb0aa
MG
832 __ClearPageActive(page);
833}
834
835static inline void ClearPageSlabPfmemalloc(struct page *page)
836{
309381fe 837 VM_BUG_ON_PAGE(!PageSlab(page), page);
072bb0aa
MG
838 ClearPageActive(page);
839}
840
af8e3354 841#ifdef CONFIG_MMU
d2a1a1f0 842#define __PG_MLOCKED (1UL << PG_mlocked)
33925b25 843#else
b291f000 844#define __PG_MLOCKED 0
894bc310
LS
845#endif
846
dfa7e20c
RA
847/*
848 * Flags checked when a page is freed. Pages being freed should not have
849 * these flags set. It they are, there is a problem.
850 */
6326fec1
NP
851#define PAGE_FLAGS_CHECK_AT_FREE \
852 (1UL << PG_lru | 1UL << PG_locked | \
853 1UL << PG_private | 1UL << PG_private_2 | \
854 1UL << PG_writeback | 1UL << PG_reserved | \
855 1UL << PG_slab | 1UL << PG_active | \
856 1UL << PG_unevictable | __PG_MLOCKED)
dfa7e20c
RA
857
858/*
859 * Flags checked when a page is prepped for return by the page allocator.
f4c18e6f 860 * Pages being prepped should not have these flags set. It they are set,
79f4b7bf 861 * there has been a kernel bug or struct page corruption.
f4c18e6f
NH
862 *
863 * __PG_HWPOISON is exceptional because it needs to be kept beyond page's
864 * alloc-free cycle to prevent from reusing the page.
dfa7e20c 865 */
f4c18e6f 866#define PAGE_FLAGS_CHECK_AT_PREP \
d2a1a1f0 867 (((1UL << NR_PAGEFLAGS) - 1) & ~__PG_HWPOISON)
dfa7e20c 868
edcf4748 869#define PAGE_FLAGS_PRIVATE \
d2a1a1f0 870 (1UL << PG_private | 1UL << PG_private_2)
266cf658
DH
871/**
872 * page_has_private - Determine if page has private stuff
873 * @page: The page to be checked
874 *
875 * Determine if a page has private stuff, indicating that release routines
876 * should be invoked upon it.
877 */
edcf4748
JW
878static inline int page_has_private(struct page *page)
879{
880 return !!(page->flags & PAGE_FLAGS_PRIVATE);
881}
882
95ad9755
KS
883#undef PF_ANY
884#undef PF_HEAD
62906027 885#undef PF_ONLY_HEAD
95ad9755
KS
886#undef PF_NO_TAIL
887#undef PF_NO_COMPOUND
edcf4748 888#endif /* !__GENERATING_BOUNDS_H */
266cf658 889
1da177e4 890#endif /* PAGE_FLAGS_H */