Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef __LINUX_GFP_TYPES_H
3#define __LINUX_GFP_TYPES_H
4
5#include <linux/bits.h>
6
7/* The typedef is in types.h but we want the documentation here */
8#if 0
9/**
10 * typedef gfp_t - Memory allocation flags.
11 *
12 * GFP flags are commonly used throughout Linux to indicate how memory
13 * should be allocated. The GFP acronym stands for get_free_pages(),
14 * the underlying memory allocation function. Not every GFP flag is
15 * supported by every function which may allocate memory. Most users
16 * will want to use a plain ``GFP_KERNEL``.
17 */
18typedef unsigned int __bitwise gfp_t;
19#endif
20
21/*
22 * In case of changes, please don't forget to update
23 * include/trace/events/mmflags.h and tools/perf/builtin-kmem.c
24 */
25
26enum {
27 ___GFP_DMA_BIT,
28 ___GFP_HIGHMEM_BIT,
29 ___GFP_DMA32_BIT,
30 ___GFP_MOVABLE_BIT,
31 ___GFP_RECLAIMABLE_BIT,
32 ___GFP_HIGH_BIT,
33 ___GFP_IO_BIT,
34 ___GFP_FS_BIT,
35 ___GFP_ZERO_BIT,
36 ___GFP_UNUSED_BIT, /* 0x200u unused */
37 ___GFP_DIRECT_RECLAIM_BIT,
38 ___GFP_KSWAPD_RECLAIM_BIT,
39 ___GFP_WRITE_BIT,
40 ___GFP_NOWARN_BIT,
41 ___GFP_RETRY_MAYFAIL_BIT,
42 ___GFP_NOFAIL_BIT,
43 ___GFP_NORETRY_BIT,
44 ___GFP_MEMALLOC_BIT,
45 ___GFP_COMP_BIT,
46 ___GFP_NOMEMALLOC_BIT,
47 ___GFP_HARDWALL_BIT,
48 ___GFP_THISNODE_BIT,
49 ___GFP_ACCOUNT_BIT,
50 ___GFP_ZEROTAGS_BIT,
51#ifdef CONFIG_KASAN_HW_TAGS
52 ___GFP_SKIP_ZERO_BIT,
53 ___GFP_SKIP_KASAN_BIT,
54#endif
55#ifdef CONFIG_LOCKDEP
56 ___GFP_NOLOCKDEP_BIT,
57#endif
58 ___GFP_NO_OBJ_EXT_BIT,
59 ___GFP_LAST_BIT
60};
61
62/* Plain integer GFP bitmasks. Do not use this directly. */
63#define ___GFP_DMA BIT(___GFP_DMA_BIT)
64#define ___GFP_HIGHMEM BIT(___GFP_HIGHMEM_BIT)
65#define ___GFP_DMA32 BIT(___GFP_DMA32_BIT)
66#define ___GFP_MOVABLE BIT(___GFP_MOVABLE_BIT)
67#define ___GFP_RECLAIMABLE BIT(___GFP_RECLAIMABLE_BIT)
68#define ___GFP_HIGH BIT(___GFP_HIGH_BIT)
69#define ___GFP_IO BIT(___GFP_IO_BIT)
70#define ___GFP_FS BIT(___GFP_FS_BIT)
71#define ___GFP_ZERO BIT(___GFP_ZERO_BIT)
72/* 0x200u unused */
73#define ___GFP_DIRECT_RECLAIM BIT(___GFP_DIRECT_RECLAIM_BIT)
74#define ___GFP_KSWAPD_RECLAIM BIT(___GFP_KSWAPD_RECLAIM_BIT)
75#define ___GFP_WRITE BIT(___GFP_WRITE_BIT)
76#define ___GFP_NOWARN BIT(___GFP_NOWARN_BIT)
77#define ___GFP_RETRY_MAYFAIL BIT(___GFP_RETRY_MAYFAIL_BIT)
78#define ___GFP_NOFAIL BIT(___GFP_NOFAIL_BIT)
79#define ___GFP_NORETRY BIT(___GFP_NORETRY_BIT)
80#define ___GFP_MEMALLOC BIT(___GFP_MEMALLOC_BIT)
81#define ___GFP_COMP BIT(___GFP_COMP_BIT)
82#define ___GFP_NOMEMALLOC BIT(___GFP_NOMEMALLOC_BIT)
83#define ___GFP_HARDWALL BIT(___GFP_HARDWALL_BIT)
84#define ___GFP_THISNODE BIT(___GFP_THISNODE_BIT)
85#define ___GFP_ACCOUNT BIT(___GFP_ACCOUNT_BIT)
86#define ___GFP_ZEROTAGS BIT(___GFP_ZEROTAGS_BIT)
87#ifdef CONFIG_KASAN_HW_TAGS
88#define ___GFP_SKIP_ZERO BIT(___GFP_SKIP_ZERO_BIT)
89#define ___GFP_SKIP_KASAN BIT(___GFP_SKIP_KASAN_BIT)
90#else
91#define ___GFP_SKIP_ZERO 0
92#define ___GFP_SKIP_KASAN 0
93#endif
94#ifdef CONFIG_LOCKDEP
95#define ___GFP_NOLOCKDEP BIT(___GFP_NOLOCKDEP_BIT)
96#else
97#define ___GFP_NOLOCKDEP 0
98#endif
99#define ___GFP_NO_OBJ_EXT BIT(___GFP_NO_OBJ_EXT_BIT)
100
101/*
102 * Physical address zone modifiers (see linux/mmzone.h - low four bits)
103 *
104 * Do not put any conditional on these. If necessary modify the definitions
105 * without the underscores and use them consistently. The definitions here may
106 * be used in bit comparisons.
107 */
108#define __GFP_DMA ((__force gfp_t)___GFP_DMA)
109#define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
110#define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
111#define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */
112#define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
113
114/**
115 * DOC: Page mobility and placement hints
116 *
117 * Page mobility and placement hints
118 * ---------------------------------
119 *
120 * These flags provide hints about how mobile the page is. Pages with similar
121 * mobility are placed within the same pageblocks to minimise problems due
122 * to external fragmentation.
123 *
124 * %__GFP_MOVABLE (also a zone modifier) indicates that the page can be
125 * moved by page migration during memory compaction or can be reclaimed.
126 *
127 * %__GFP_RECLAIMABLE is used for slab allocations that specify
128 * SLAB_RECLAIM_ACCOUNT and whose pages can be freed via shrinkers.
129 *
130 * %__GFP_WRITE indicates the caller intends to dirty the page. Where possible,
131 * these pages will be spread between local zones to avoid all the dirty
132 * pages being in one zone (fair zone allocation policy).
133 *
134 * %__GFP_HARDWALL enforces the cpuset memory allocation policy.
135 *
136 * %__GFP_THISNODE forces the allocation to be satisfied from the requested
137 * node with no fallbacks or placement policy enforcements.
138 *
139 * %__GFP_ACCOUNT causes the allocation to be accounted to kmemcg.
140 *
141 * %__GFP_NO_OBJ_EXT causes slab allocation to have no object extension.
142 * mark_obj_codetag_empty() should be called upon freeing for objects allocated
143 * with this flag to indicate that their NULL tags are expected and normal.
144 */
145#define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE)
146#define __GFP_WRITE ((__force gfp_t)___GFP_WRITE)
147#define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL)
148#define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)
149#define __GFP_ACCOUNT ((__force gfp_t)___GFP_ACCOUNT)
150#define __GFP_NO_OBJ_EXT ((__force gfp_t)___GFP_NO_OBJ_EXT)
151
152/**
153 * DOC: Watermark modifiers
154 *
155 * Watermark modifiers -- controls access to emergency reserves
156 * ------------------------------------------------------------
157 *
158 * %__GFP_HIGH indicates that the caller is high-priority and that granting
159 * the request is necessary before the system can make forward progress.
160 * For example creating an IO context to clean pages and requests
161 * from atomic context.
162 *
163 * %__GFP_MEMALLOC allows access to all memory. This should only be used when
164 * the caller guarantees the allocation will allow more memory to be freed
165 * very shortly e.g. process exiting or swapping. Users either should
166 * be the MM or co-ordinating closely with the VM (e.g. swap over NFS).
167 * Users of this flag have to be extremely careful to not deplete the reserve
168 * completely and implement a throttling mechanism which controls the
169 * consumption of the reserve based on the amount of freed memory.
170 * Usage of a pre-allocated pool (e.g. mempool) should be always considered
171 * before using this flag.
172 *
173 * %__GFP_NOMEMALLOC is used to explicitly forbid access to emergency reserves.
174 * This takes precedence over the %__GFP_MEMALLOC flag if both are set.
175 */
176#define __GFP_HIGH ((__force gfp_t)___GFP_HIGH)
177#define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)
178#define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC)
179
180/**
181 * DOC: Reclaim modifiers
182 *
183 * Reclaim modifiers
184 * -----------------
185 * Please note that all the following flags are only applicable to sleepable
186 * allocations (e.g. %GFP_NOWAIT and %GFP_ATOMIC will ignore them).
187 *
188 * %__GFP_IO can start physical IO.
189 *
190 * %__GFP_FS can call down to the low-level FS. Clearing the flag avoids the
191 * allocator recursing into the filesystem which might already be holding
192 * locks.
193 *
194 * %__GFP_DIRECT_RECLAIM indicates that the caller may enter direct reclaim.
195 * This flag can be cleared to avoid unnecessary delays when a fallback
196 * option is available.
197 *
198 * %__GFP_KSWAPD_RECLAIM indicates that the caller wants to wake kswapd when
199 * the low watermark is reached and have it reclaim pages until the high
200 * watermark is reached. A caller may wish to clear this flag when fallback
201 * options are available and the reclaim is likely to disrupt the system. The
202 * canonical example is THP allocation where a fallback is cheap but
203 * reclaim/compaction may cause indirect stalls.
204 *
205 * %__GFP_RECLAIM is shorthand to allow/forbid both direct and kswapd reclaim.
206 *
207 * The default allocator behavior depends on the request size. We have a concept
208 * of so-called costly allocations (with order > %PAGE_ALLOC_COSTLY_ORDER).
209 * !costly allocations are too essential to fail so they are implicitly
210 * non-failing by default (with some exceptions like OOM victims might fail so
211 * the caller still has to check for failures) while costly requests try to be
212 * not disruptive and back off even without invoking the OOM killer.
213 * The following three modifiers might be used to override some of these
214 * implicit rules. Please note that all of them must be used along with
215 * %__GFP_DIRECT_RECLAIM flag.
216 *
217 * %__GFP_NORETRY: The VM implementation will try only very lightweight
218 * memory direct reclaim to get some memory under memory pressure (thus
219 * it can sleep). It will avoid disruptive actions like OOM killer. The
220 * caller must handle the failure which is quite likely to happen under
221 * heavy memory pressure. The flag is suitable when failure can easily be
222 * handled at small cost, such as reduced throughput.
223 *
224 * %__GFP_RETRY_MAYFAIL: The VM implementation will retry memory reclaim
225 * procedures that have previously failed if there is some indication
226 * that progress has been made elsewhere. It can wait for other
227 * tasks to attempt high-level approaches to freeing memory such as
228 * compaction (which removes fragmentation) and page-out.
229 * There is still a definite limit to the number of retries, but it is
230 * a larger limit than with %__GFP_NORETRY.
231 * Allocations with this flag may fail, but only when there is
232 * genuinely little unused memory. While these allocations do not
233 * directly trigger the OOM killer, their failure indicates that
234 * the system is likely to need to use the OOM killer soon. The
235 * caller must handle failure, but can reasonably do so by failing
236 * a higher-level request, or completing it only in a much less
237 * efficient manner.
238 * If the allocation does fail, and the caller is in a position to
239 * free some non-essential memory, doing so could benefit the system
240 * as a whole.
241 *
242 * %__GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
243 * cannot handle allocation failures. The allocation could block
244 * indefinitely but will never return with failure. Testing for
245 * failure is pointless.
246 * It _must_ be blockable and used together with __GFP_DIRECT_RECLAIM.
247 * It should _never_ be used in non-sleepable contexts.
248 * New users should be evaluated carefully (and the flag should be
249 * used only when there is no reasonable failure policy) but it is
250 * definitely preferable to use the flag rather than opencode endless
251 * loop around allocator.
252 * Allocating pages from the buddy with __GFP_NOFAIL and order > 1 is
253 * not supported. Please consider using kvmalloc() instead.
254 */
255#define __GFP_IO ((__force gfp_t)___GFP_IO)
256#define __GFP_FS ((__force gfp_t)___GFP_FS)
257#define __GFP_DIRECT_RECLAIM ((__force gfp_t)___GFP_DIRECT_RECLAIM) /* Caller can reclaim */
258#define __GFP_KSWAPD_RECLAIM ((__force gfp_t)___GFP_KSWAPD_RECLAIM) /* kswapd can wake */
259#define __GFP_RECLAIM ((__force gfp_t)(___GFP_DIRECT_RECLAIM|___GFP_KSWAPD_RECLAIM))
260#define __GFP_RETRY_MAYFAIL ((__force gfp_t)___GFP_RETRY_MAYFAIL)
261#define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL)
262#define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY)
263
264/**
265 * DOC: Action modifiers
266 *
267 * Action modifiers
268 * ----------------
269 *
270 * %__GFP_NOWARN suppresses allocation failure reports.
271 *
272 * %__GFP_COMP address compound page metadata.
273 *
274 * %__GFP_ZERO returns a zeroed page on success.
275 *
276 * %__GFP_ZEROTAGS zeroes memory tags at allocation time if the memory itself
277 * is being zeroed (either via __GFP_ZERO or via init_on_alloc, provided that
278 * __GFP_SKIP_ZERO is not set). This flag is intended for optimization: setting
279 * memory tags at the same time as zeroing memory has minimal additional
280 * performance impact.
281 *
282 * %__GFP_SKIP_KASAN makes KASAN skip unpoisoning on page allocation.
283 * Used for userspace and vmalloc pages; the latter are unpoisoned by
284 * kasan_unpoison_vmalloc instead. For userspace pages, results in
285 * poisoning being skipped as well, see should_skip_kasan_poison for
286 * details. Only effective in HW_TAGS mode.
287 */
288#define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN)
289#define __GFP_COMP ((__force gfp_t)___GFP_COMP)
290#define __GFP_ZERO ((__force gfp_t)___GFP_ZERO)
291#define __GFP_ZEROTAGS ((__force gfp_t)___GFP_ZEROTAGS)
292#define __GFP_SKIP_ZERO ((__force gfp_t)___GFP_SKIP_ZERO)
293#define __GFP_SKIP_KASAN ((__force gfp_t)___GFP_SKIP_KASAN)
294
295/* Disable lockdep for GFP context tracking */
296#define __GFP_NOLOCKDEP ((__force gfp_t)___GFP_NOLOCKDEP)
297
298/* Room for N __GFP_FOO bits */
299#define __GFP_BITS_SHIFT ___GFP_LAST_BIT
300#define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
301
302/**
303 * DOC: Useful GFP flag combinations
304 *
305 * Useful GFP flag combinations
306 * ----------------------------
307 *
308 * Useful GFP flag combinations that are commonly used. It is recommended
309 * that subsystems start with one of these combinations and then set/clear
310 * %__GFP_FOO flags as necessary.
311 *
312 * %GFP_ATOMIC users can not sleep and need the allocation to succeed. A lower
313 * watermark is applied to allow access to "atomic reserves".
314 * The current implementation doesn't support NMI, nor contexts that disable
315 * preemption under PREEMPT_RT. This includes raw_spin_lock() and plain
316 * preempt_disable() - see "Memory allocation" in
317 * Documentation/core-api/real-time/differences.rst for more info.
318 *
319 * %GFP_KERNEL is typical for kernel-internal allocations. The caller requires
320 * %ZONE_NORMAL or a lower zone for direct access but can direct reclaim.
321 *
322 * %GFP_KERNEL_ACCOUNT is the same as GFP_KERNEL, except the allocation is
323 * accounted to kmemcg.
324 *
325 * %GFP_NOWAIT is for kernel allocations that should not stall for direct
326 * reclaim, start physical IO or use any filesystem callback. It is very
327 * likely to fail to allocate memory, even for very small allocations.
328 * The same restrictions on calling contexts apply as for %GFP_ATOMIC.
329 *
330 * %GFP_NOIO will use direct reclaim to discard clean pages or slab pages
331 * that do not require the starting of any physical IO.
332 * Please try to avoid using this flag directly and instead use
333 * memalloc_noio_{save,restore} to mark the whole scope which cannot
334 * perform any IO with a short explanation why. All allocation requests
335 * will inherit GFP_NOIO implicitly.
336 *
337 * %GFP_NOFS will use direct reclaim but will not use any filesystem interfaces.
338 * Please try to avoid using this flag directly and instead use
339 * memalloc_nofs_{save,restore} to mark the whole scope which cannot/shouldn't
340 * recurse into the FS layer with a short explanation why. All allocation
341 * requests will inherit GFP_NOFS implicitly.
342 *
343 * %GFP_USER is for userspace allocations that also need to be directly
344 * accessibly by the kernel or hardware. It is typically used by hardware
345 * for buffers that are mapped to userspace (e.g. graphics) that hardware
346 * still must DMA to. cpuset limits are enforced for these allocations.
347 *
348 * %GFP_DMA exists for historical reasons and should be avoided where possible.
349 * The flags indicates that the caller requires that the lowest zone be
350 * used (%ZONE_DMA or 16M on x86-64). Ideally, this would be removed but
351 * it would require careful auditing as some users really require it and
352 * others use the flag to avoid lowmem reserves in %ZONE_DMA and treat the
353 * lowest zone as a type of emergency reserve.
354 *
355 * %GFP_DMA32 is similar to %GFP_DMA except that the caller requires a 32-bit
356 * address. Note that kmalloc(..., GFP_DMA32) does not return DMA32 memory
357 * because the DMA32 kmalloc cache array is not implemented.
358 * (Reason: there is no such user in kernel).
359 *
360 * %GFP_HIGHUSER is for userspace allocations that may be mapped to userspace,
361 * do not need to be directly accessible by the kernel but that cannot
362 * move once in use. An example may be a hardware allocation that maps
363 * data directly into userspace but has no addressing limitations.
364 *
365 * %GFP_HIGHUSER_MOVABLE is for userspace allocations that the kernel does not
366 * need direct access to but can use kmap() when access is required. They
367 * are expected to be movable via page reclaim or page migration. Typically,
368 * pages on the LRU would also be allocated with %GFP_HIGHUSER_MOVABLE.
369 *
370 * %GFP_TRANSHUGE and %GFP_TRANSHUGE_LIGHT are used for THP allocations. They
371 * are compound allocations that will generally fail quickly if memory is not
372 * available and will not wake kswapd/kcompactd on failure. The _LIGHT
373 * version does not attempt reclaim/compaction at all and is by default used
374 * in page fault path, while the non-light is used by khugepaged.
375 */
376#define GFP_ATOMIC (__GFP_HIGH|__GFP_KSWAPD_RECLAIM)
377#define GFP_KERNEL (__GFP_RECLAIM | __GFP_IO | __GFP_FS)
378#define GFP_KERNEL_ACCOUNT (GFP_KERNEL | __GFP_ACCOUNT)
379#define GFP_NOWAIT (__GFP_KSWAPD_RECLAIM | __GFP_NOWARN)
380#define GFP_NOIO (__GFP_RECLAIM)
381#define GFP_NOFS (__GFP_RECLAIM | __GFP_IO)
382#define GFP_USER (__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
383#define GFP_DMA __GFP_DMA
384#define GFP_DMA32 __GFP_DMA32
385#define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
386#define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE | __GFP_SKIP_KASAN)
387#define GFP_TRANSHUGE_LIGHT ((GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
388 __GFP_NOMEMALLOC | __GFP_NOWARN) & ~__GFP_RECLAIM)
389#define GFP_TRANSHUGE (GFP_TRANSHUGE_LIGHT | __GFP_DIRECT_RECLAIM)
390
391#endif /* __LINUX_GFP_TYPES_H */