Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1#ifndef __LINUX_GFP_H
2#define __LINUX_GFP_H
3
4#include <linux/mmzone.h>
5#include <linux/stddef.h>
6#include <linux/linkage.h>
7#include <linux/topology.h>
8#include <linux/mmdebug.h>
9
10struct vm_area_struct;
11
12/* Plain integer GFP bitmasks. Do not use this directly. */
13#define ___GFP_DMA 0x01u
14#define ___GFP_HIGHMEM 0x02u
15#define ___GFP_DMA32 0x04u
16#define ___GFP_MOVABLE 0x08u
17#define ___GFP_WAIT 0x10u
18#define ___GFP_HIGH 0x20u
19#define ___GFP_IO 0x40u
20#define ___GFP_FS 0x80u
21#define ___GFP_COLD 0x100u
22#define ___GFP_NOWARN 0x200u
23#define ___GFP_REPEAT 0x400u
24#define ___GFP_NOFAIL 0x800u
25#define ___GFP_NORETRY 0x1000u
26#define ___GFP_COMP 0x4000u
27#define ___GFP_ZERO 0x8000u
28#define ___GFP_NOMEMALLOC 0x10000u
29#define ___GFP_HARDWALL 0x20000u
30#define ___GFP_THISNODE 0x40000u
31#define ___GFP_RECLAIMABLE 0x80000u
32#ifdef CONFIG_KMEMCHECK
33#define ___GFP_NOTRACK 0x200000u
34#else
35#define ___GFP_NOTRACK 0
36#endif
37
38/*
39 * GFP bitmasks..
40 *
41 * Zone modifiers (see linux/mmzone.h - low three bits)
42 *
43 * Do not put any conditional on these. If necessary modify the definitions
44 * without the underscores and use them consistently. The definitions here may
45 * be used in bit comparisons.
46 */
47#define __GFP_DMA ((__force gfp_t)___GFP_DMA)
48#define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
49#define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
50#define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */
51#define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
52/*
53 * Action modifiers - doesn't change the zoning
54 *
55 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
56 * _might_ fail. This depends upon the particular VM implementation.
57 *
58 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
59 * cannot handle allocation failures. This modifier is deprecated and no new
60 * users should be added.
61 *
62 * __GFP_NORETRY: The VM implementation must not retry indefinitely.
63 *
64 * __GFP_MOVABLE: Flag that this page will be movable by the page migration
65 * mechanism or reclaimed
66 */
67#define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */
68#define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */
69#define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */
70#define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */
71#define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */
72#define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */
73#define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */
74#define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */
75#define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */
76#define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */
77#define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */
78#define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves */
79#define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */
80#define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */
81#define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */
82#define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */
83
84/*
85 * This may seem redundant, but it's a way of annotating false positives vs.
86 * allocations that simply cannot be supported (e.g. page tables).
87 */
88#define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
89
90#define __GFP_BITS_SHIFT 22 /* Room for 22 __GFP_FOO bits */
91#define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
92
93/* This equals 0, but use constants in case they ever change */
94#define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH)
95/* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
96#define GFP_ATOMIC (__GFP_HIGH)
97#define GFP_NOIO (__GFP_WAIT)
98#define GFP_NOFS (__GFP_WAIT | __GFP_IO)
99#define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
100#define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
101 __GFP_RECLAIMABLE)
102#define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
103#define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \
104 __GFP_HIGHMEM)
105#define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
106 __GFP_HARDWALL | __GFP_HIGHMEM | \
107 __GFP_MOVABLE)
108#define GFP_IOFS (__GFP_IO | __GFP_FS)
109
110#ifdef CONFIG_NUMA
111#define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)
112#else
113#define GFP_THISNODE ((__force gfp_t)0)
114#endif
115
116/* This mask makes up all the page movable related flags */
117#define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
118
119/* Control page allocator reclaim behavior */
120#define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
121 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
122 __GFP_NORETRY|__GFP_NOMEMALLOC)
123
124/* Control slab gfp mask during early boot */
125#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
126
127/* Control allocation constraints */
128#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
129
130/* Do not use these with a slab allocator */
131#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
132
133/* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
134 platforms, used as appropriate on others */
135
136#define GFP_DMA __GFP_DMA
137
138/* 4GB DMA on some platforms */
139#define GFP_DMA32 __GFP_DMA32
140
141/* Convert GFP flags to their corresponding migrate type */
142static inline int allocflags_to_migratetype(gfp_t gfp_flags)
143{
144 WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
145
146 if (unlikely(page_group_by_mobility_disabled))
147 return MIGRATE_UNMOVABLE;
148
149 /* Group based on mobility */
150 return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
151 ((gfp_flags & __GFP_RECLAIMABLE) != 0);
152}
153
154#ifdef CONFIG_HIGHMEM
155#define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
156#else
157#define OPT_ZONE_HIGHMEM ZONE_NORMAL
158#endif
159
160#ifdef CONFIG_ZONE_DMA
161#define OPT_ZONE_DMA ZONE_DMA
162#else
163#define OPT_ZONE_DMA ZONE_NORMAL
164#endif
165
166#ifdef CONFIG_ZONE_DMA32
167#define OPT_ZONE_DMA32 ZONE_DMA32
168#else
169#define OPT_ZONE_DMA32 ZONE_NORMAL
170#endif
171
172/*
173 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
174 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
175 * and there are 16 of them to cover all possible combinations of
176 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
177 *
178 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
179 * But GFP_MOVABLE is not only a zone specifier but also an allocation
180 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
181 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
182 *
183 * bit result
184 * =================
185 * 0x0 => NORMAL
186 * 0x1 => DMA or NORMAL
187 * 0x2 => HIGHMEM or NORMAL
188 * 0x3 => BAD (DMA+HIGHMEM)
189 * 0x4 => DMA32 or DMA or NORMAL
190 * 0x5 => BAD (DMA+DMA32)
191 * 0x6 => BAD (HIGHMEM+DMA32)
192 * 0x7 => BAD (HIGHMEM+DMA32+DMA)
193 * 0x8 => NORMAL (MOVABLE+0)
194 * 0x9 => DMA or NORMAL (MOVABLE+DMA)
195 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
196 * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
197 * 0xc => DMA32 (MOVABLE+HIGHMEM+DMA32)
198 * 0xd => BAD (MOVABLE+DMA32+DMA)
199 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
200 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
201 *
202 * ZONES_SHIFT must be <= 2 on 32 bit platforms.
203 */
204
205#if 16 * ZONES_SHIFT > BITS_PER_LONG
206#error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
207#endif
208
209#define GFP_ZONE_TABLE ( \
210 (ZONE_NORMAL << 0 * ZONES_SHIFT) \
211 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
212 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
213 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
214 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
215 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
216 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
217 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
218)
219
220/*
221 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
222 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
223 * entry starting with bit 0. Bit is set if the combination is not
224 * allowed.
225 */
226#define GFP_ZONE_BAD ( \
227 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
228 | 1 << (___GFP_DMA | ___GFP_DMA32) \
229 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
230 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
231 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
232 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
233 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
234 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
235)
236
237static inline enum zone_type gfp_zone(gfp_t flags)
238{
239 enum zone_type z;
240 int bit = (__force int) (flags & GFP_ZONEMASK);
241
242 z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
243 ((1 << ZONES_SHIFT) - 1);
244
245 if (__builtin_constant_p(bit))
246 MAYBE_BUILD_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
247 else {
248#ifdef CONFIG_DEBUG_VM
249 BUG_ON((GFP_ZONE_BAD >> bit) & 1);
250#endif
251 }
252 return z;
253}
254
255/*
256 * There is only one page-allocator function, and two main namespaces to
257 * it. The alloc_page*() variants return 'struct page *' and as such
258 * can allocate highmem pages, the *get*page*() variants return
259 * virtual kernel addresses to the allocated page(s).
260 */
261
262static inline int gfp_zonelist(gfp_t flags)
263{
264 if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE))
265 return 1;
266
267 return 0;
268}
269
270/*
271 * We get the zone list from the current node and the gfp_mask.
272 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
273 * There are two zonelists per node, one for all zones with memory and
274 * one containing just zones from the node the zonelist belongs to.
275 *
276 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
277 * optimized to &contig_page_data at compile-time.
278 */
279static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
280{
281 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
282}
283
284#ifndef HAVE_ARCH_FREE_PAGE
285static inline void arch_free_page(struct page *page, int order) { }
286#endif
287#ifndef HAVE_ARCH_ALLOC_PAGE
288static inline void arch_alloc_page(struct page *page, int order) { }
289#endif
290
291struct page *
292__alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
293 struct zonelist *zonelist, nodemask_t *nodemask);
294
295static inline struct page *
296__alloc_pages(gfp_t gfp_mask, unsigned int order,
297 struct zonelist *zonelist)
298{
299 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
300}
301
302static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
303 unsigned int order)
304{
305 /* Unknown node is current node */
306 if (nid < 0)
307 nid = numa_node_id();
308
309 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
310}
311
312static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask,
313 unsigned int order)
314{
315 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
316
317 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
318}
319
320#ifdef CONFIG_NUMA
321extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
322
323static inline struct page *
324alloc_pages(gfp_t gfp_mask, unsigned int order)
325{
326 return alloc_pages_current(gfp_mask, order);
327}
328extern struct page *alloc_page_vma(gfp_t gfp_mask,
329 struct vm_area_struct *vma, unsigned long addr);
330#else
331#define alloc_pages(gfp_mask, order) \
332 alloc_pages_node(numa_node_id(), gfp_mask, order)
333#define alloc_page_vma(gfp_mask, vma, addr) alloc_pages(gfp_mask, 0)
334#endif
335#define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
336
337extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
338extern unsigned long get_zeroed_page(gfp_t gfp_mask);
339
340void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
341void free_pages_exact(void *virt, size_t size);
342
343#define __get_free_page(gfp_mask) \
344 __get_free_pages((gfp_mask), 0)
345
346#define __get_dma_pages(gfp_mask, order) \
347 __get_free_pages((gfp_mask) | GFP_DMA, (order))
348
349extern void __free_pages(struct page *page, unsigned int order);
350extern void free_pages(unsigned long addr, unsigned int order);
351extern void free_hot_cold_page(struct page *page, int cold);
352
353#define __free_page(page) __free_pages((page), 0)
354#define free_page(addr) free_pages((addr), 0)
355
356void page_alloc_init(void);
357void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
358void drain_all_pages(void);
359void drain_local_pages(void *dummy);
360
361extern gfp_t gfp_allowed_mask;
362
363extern void pm_restrict_gfp_mask(void);
364extern void pm_restore_gfp_mask(void);
365
366#endif /* __LINUX_GFP_H */