blob: daeaa8fe1bbd6995f39f9940d7dc412585997c25 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#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>
Rusty Russell082edb72009-03-13 23:43:37 +10307#include <linux/topology.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07008
9struct vm_area_struct;
10
11/*
12 * GFP bitmasks..
Christoph Lametere53ef382006-09-25 23:31:14 -070013 *
14 * Zone modifiers (see linux/mmzone.h - low three bits)
15 *
Christoph Lametere53ef382006-09-25 23:31:14 -070016 * Do not put any conditional on these. If necessary modify the definitions
17 * without the underscores and use the consistently. The definitions here may
18 * be used in bit comparisons.
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 */
Al Viroaf4ca452005-10-21 02:55:38 -040020#define __GFP_DMA ((__force gfp_t)0x01u)
21#define __GFP_HIGHMEM ((__force gfp_t)0x02u)
Christoph Lametere53ef382006-09-25 23:31:14 -070022#define __GFP_DMA32 ((__force gfp_t)0x04u)
Linus Torvalds1da177e2005-04-16 15:20:36 -070023
24/*
25 * Action modifiers - doesn't change the zoning
26 *
27 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
28 * _might_ fail. This depends upon the particular VM implementation.
29 *
30 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
31 * cannot handle allocation failures.
32 *
33 * __GFP_NORETRY: The VM implementation must not retry indefinitely.
Mel Gorman769848c2007-07-17 04:03:05 -070034 *
35 * __GFP_MOVABLE: Flag that this page will be movable by the page migration
36 * mechanism or reclaimed
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 */
Al Viroaf4ca452005-10-21 02:55:38 -040038#define __GFP_WAIT ((__force gfp_t)0x10u) /* Can wait and reschedule? */
39#define __GFP_HIGH ((__force gfp_t)0x20u) /* Should access emergency pools? */
40#define __GFP_IO ((__force gfp_t)0x40u) /* Can start physical IO? */
41#define __GFP_FS ((__force gfp_t)0x80u) /* Can call down to low-level FS? */
42#define __GFP_COLD ((__force gfp_t)0x100u) /* Cache-cold page required */
43#define __GFP_NOWARN ((__force gfp_t)0x200u) /* Suppress page allocation failure warning */
Nishanth Aravamudanab857d0932008-04-29 00:58:23 -070044#define __GFP_REPEAT ((__force gfp_t)0x400u) /* See above */
45#define __GFP_NOFAIL ((__force gfp_t)0x800u) /* See above */
46#define __GFP_NORETRY ((__force gfp_t)0x1000u)/* See above */
Al Viroaf4ca452005-10-21 02:55:38 -040047#define __GFP_COMP ((__force gfp_t)0x4000u)/* Add compound page metadata */
48#define __GFP_ZERO ((__force gfp_t)0x8000u)/* Return zeroed page on success */
49#define __GFP_NOMEMALLOC ((__force gfp_t)0x10000u) /* Don't use emergency reserves */
Paul Jackson2d6c6662005-11-13 16:06:44 -080050#define __GFP_HARDWALL ((__force gfp_t)0x20000u) /* Enforce hardwall cpuset memory allocs */
Christoph Lameter9b819d22006-09-25 23:31:40 -070051#define __GFP_THISNODE ((__force gfp_t)0x40000u)/* No fallback, no policies */
Mel Gormane12ba742007-10-16 01:25:52 -070052#define __GFP_RECLAIMABLE ((__force gfp_t)0x80000u) /* Page is reclaimable */
53#define __GFP_MOVABLE ((__force gfp_t)0x100000u) /* Page is movable */
Vegard Nossum2dff4402008-05-31 15:56:17 +020054#define __GFP_NOTRACK ((__force gfp_t)0x200000u) /* Don't track with kmemcheck */
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
Vegard Nossum2dff4402008-05-31 15:56:17 +020056/*
57 * This may seem redundant, but it's a way of annotating false positives vs.
58 * allocations that simply cannot be supported (e.g. page tables).
59 */
60#define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
61
62#define __GFP_BITS_SHIFT 22 /* Room for 22 __GFP_FOO bits */
Al Viroaf4ca452005-10-21 02:55:38 -040063#define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Jeff Dike7b04d712006-04-10 22:53:27 -070065/* This equals 0, but use constants in case they ever change */
66#define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH)
Paul Jackson4eac9152006-01-11 12:17:19 -080067/* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#define GFP_ATOMIC (__GFP_HIGH)
69#define GFP_NOIO (__GFP_WAIT)
70#define GFP_NOFS (__GFP_WAIT | __GFP_IO)
71#define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
Mel Gormane12ba742007-10-16 01:25:52 -070072#define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
73 __GFP_RECLAIMABLE)
Paul Jacksonf90b1d22005-09-06 15:18:10 -070074#define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
75#define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \
76 __GFP_HIGHMEM)
Mel Gorman769848c2007-07-17 04:03:05 -070077#define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
78 __GFP_HARDWALL | __GFP_HIGHMEM | \
79 __GFP_MOVABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -070080
Christoph Lameter77f700d2006-09-27 01:50:07 -070081#ifdef CONFIG_NUMA
Christoph Lameter980128f2006-09-25 23:31:46 -070082#define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)
Christoph Lameter77f700d2006-09-27 01:50:07 -070083#else
Al Virof2e97df2007-02-09 16:38:55 +000084#define GFP_THISNODE ((__force gfp_t)0)
Christoph Lameter77f700d2006-09-27 01:50:07 -070085#endif
86
Christoph Lameter6cb06222007-10-16 01:25:41 -070087/* This mask makes up all the page movable related flags */
Mel Gormane12ba742007-10-16 01:25:52 -070088#define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
Christoph Lameter6cb06222007-10-16 01:25:41 -070089
90/* Control page allocator reclaim behavior */
91#define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
92 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
93 __GFP_NORETRY|__GFP_NOMEMALLOC)
94
95/* Control allocation constraints */
96#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
97
98/* Do not use these with a slab allocator */
99#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
Christoph Lameter980128f2006-09-25 23:31:46 -0700100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101/* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
102 platforms, used as appropriate on others */
103
104#define GFP_DMA __GFP_DMA
105
Andi Kleena2f1b422005-11-05 17:25:53 +0100106/* 4GB DMA on some platforms */
107#define GFP_DMA32 __GFP_DMA32
108
Mel Gorman467c9962007-10-16 01:26:02 -0700109/* Convert GFP flags to their corresponding migrate type */
110static inline int allocflags_to_migratetype(gfp_t gfp_flags)
111{
112 WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
113
114 if (unlikely(page_group_by_mobility_disabled))
115 return MIGRATE_UNMOVABLE;
116
117 /* Group based on mobility */
118 return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
119 ((gfp_flags & __GFP_RECLAIMABLE) != 0);
120}
Andi Kleena2f1b422005-11-05 17:25:53 +0100121
Christoph Lameter19655d32006-09-25 23:31:19 -0700122static inline enum zone_type gfp_zone(gfp_t flags)
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700123{
Christoph Lameter4b51d662007-02-10 01:43:10 -0800124#ifdef CONFIG_ZONE_DMA
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700125 if (flags & __GFP_DMA)
KAMEZAWA Hiroyuki8cece85ec2008-04-28 02:13:36 -0700126 return ZONE_DMA;
Christoph Lameter4b51d662007-02-10 01:43:10 -0800127#endif
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700128#ifdef CONFIG_ZONE_DMA32
129 if (flags & __GFP_DMA32)
KAMEZAWA Hiroyuki8cece85ec2008-04-28 02:13:36 -0700130 return ZONE_DMA32;
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700131#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700132 if ((flags & (__GFP_HIGHMEM | __GFP_MOVABLE)) ==
133 (__GFP_HIGHMEM | __GFP_MOVABLE))
KAMEZAWA Hiroyuki8cece85ec2008-04-28 02:13:36 -0700134 return ZONE_MOVABLE;
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700135#ifdef CONFIG_HIGHMEM
136 if (flags & __GFP_HIGHMEM)
KAMEZAWA Hiroyuki8cece85ec2008-04-28 02:13:36 -0700137 return ZONE_HIGHMEM;
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700138#endif
KAMEZAWA Hiroyuki8cece85ec2008-04-28 02:13:36 -0700139 return ZONE_NORMAL;
Christoph Lameter4e4785b2006-09-25 23:31:17 -0700140}
141
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142/*
143 * There is only one page-allocator function, and two main namespaces to
144 * it. The alloc_page*() variants return 'struct page *' and as such
145 * can allocate highmem pages, the *get*page*() variants return
146 * virtual kernel addresses to the allocated page(s).
147 */
148
Mel Gorman54a6eb52008-04-28 02:12:16 -0700149static inline int gfp_zonelist(gfp_t flags)
150{
151 if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE))
152 return 1;
153
154 return 0;
155}
156
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157/*
158 * We get the zone list from the current node and the gfp_mask.
159 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
Mel Gorman54a6eb52008-04-28 02:12:16 -0700160 * There are two zonelists per node, one for all zones with memory and
161 * one containing just zones from the node the zonelist belongs to.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 *
163 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
164 * optimized to &contig_page_data at compile-time.
165 */
Mel Gorman0e884602008-04-28 02:12:14 -0700166static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
167{
Mel Gorman54a6eb52008-04-28 02:12:16 -0700168 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
Mel Gorman0e884602008-04-28 02:12:14 -0700169}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
171#ifndef HAVE_ARCH_FREE_PAGE
172static inline void arch_free_page(struct page *page, int order) { }
173#endif
Nick Piggincc1025092006-12-06 20:32:00 -0800174#ifndef HAVE_ARCH_ALLOC_PAGE
175static inline void arch_alloc_page(struct page *page, int order) { }
176#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
KOSAKI Motohiroe4048e52008-07-23 21:27:01 -0700178struct page *
179__alloc_pages_internal(gfp_t gfp_mask, unsigned int order,
180 struct zonelist *zonelist, nodemask_t *nodemask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
KOSAKI Motohiroe4048e52008-07-23 21:27:01 -0700182static inline struct page *
183__alloc_pages(gfp_t gfp_mask, unsigned int order,
184 struct zonelist *zonelist)
185{
186 return __alloc_pages_internal(gfp_mask, order, zonelist, NULL);
187}
188
189static inline struct page *
190__alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
191 struct zonelist *zonelist, nodemask_t *nodemask)
192{
193 return __alloc_pages_internal(gfp_mask, order, zonelist, nodemask);
194}
195
Mel Gorman19770b32008-04-28 02:12:18 -0700196
Al Virodd0fc662005-10-07 07:46:04 +0100197static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 unsigned int order)
199{
200 if (unlikely(order >= MAX_ORDER))
201 return NULL;
202
Andi Kleen819a6922006-01-11 22:43:45 +0100203 /* Unknown node is current node */
204 if (nid < 0)
205 nid = numa_node_id();
206
Mel Gorman0e884602008-04-28 02:12:14 -0700207 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208}
209
210#ifdef CONFIG_NUMA
Al Virodd0fc662005-10-07 07:46:04 +0100211extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
213static inline struct page *
Al Virodd0fc662005-10-07 07:46:04 +0100214alloc_pages(gfp_t gfp_mask, unsigned int order)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 if (unlikely(order >= MAX_ORDER))
217 return NULL;
218
219 return alloc_pages_current(gfp_mask, order);
220}
Al Virodd0fc662005-10-07 07:46:04 +0100221extern struct page *alloc_page_vma(gfp_t gfp_mask,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 struct vm_area_struct *vma, unsigned long addr);
223#else
224#define alloc_pages(gfp_mask, order) \
225 alloc_pages_node(numa_node_id(), gfp_mask, order)
226#define alloc_page_vma(gfp_mask, vma, addr) alloc_pages(gfp_mask, 0)
227#endif
228#define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
229
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800230extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
231extern unsigned long get_zeroed_page(gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232
Timur Tabi2be0ffe2008-07-23 21:28:11 -0700233void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
234void free_pages_exact(void *virt, size_t size);
235
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236#define __get_free_page(gfp_mask) \
237 __get_free_pages((gfp_mask),0)
238
239#define __get_dma_pages(gfp_mask, order) \
240 __get_free_pages((gfp_mask) | GFP_DMA,(order))
241
Harvey Harrisonb3c97522008-02-13 15:03:15 -0800242extern void __free_pages(struct page *page, unsigned int order);
243extern void free_pages(unsigned long addr, unsigned int order);
244extern void free_hot_page(struct page *page);
245extern void free_cold_page(struct page *page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247#define __free_page(page) __free_pages((page), 0)
248#define free_page(addr) free_pages((addr),0)
249
250void page_alloc_init(void);
Christoph Lameter4037d452007-05-09 02:35:14 -0700251void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
Christoph Lameter9f8f2172008-02-04 22:29:11 -0800252void drain_all_pages(void);
253void drain_local_pages(void *dummy);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
255#endif /* __LINUX_GFP_H */