Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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> |
Rusty Russell | 082edb7 | 2009-03-13 23:43:37 +1030 | [diff] [blame] | 7 | #include <linux/topology.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 8 | |
| 9 | struct vm_area_struct; |
| 10 | |
| 11 | /* |
| 12 | * GFP bitmasks.. |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 13 | * |
| 14 | * Zone modifiers (see linux/mmzone.h - low three bits) |
| 15 | * |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 16 | * 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 19 | */ |
Al Viro | af4ca45 | 2005-10-21 02:55:38 -0400 | [diff] [blame] | 20 | #define __GFP_DMA ((__force gfp_t)0x01u) |
| 21 | #define __GFP_HIGHMEM ((__force gfp_t)0x02u) |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 22 | #define __GFP_DMA32 ((__force gfp_t)0x04u) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 23 | |
| 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 Gorman | 769848c | 2007-07-17 04:03:05 -0700 | [diff] [blame] | 34 | * |
| 35 | * __GFP_MOVABLE: Flag that this page will be movable by the page migration |
| 36 | * mechanism or reclaimed |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | */ |
Al Viro | af4ca45 | 2005-10-21 02:55:38 -0400 | [diff] [blame] | 38 | #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 Aravamudan | ab857d0 | 2008-04-29 00:58:23 -0700 | [diff] [blame] | 44 | #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 Viro | af4ca45 | 2005-10-21 02:55:38 -0400 | [diff] [blame] | 47 | #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 Jackson | 2d6c666 | 2005-11-13 16:06:44 -0800 | [diff] [blame] | 50 | #define __GFP_HARDWALL ((__force gfp_t)0x20000u) /* Enforce hardwall cpuset memory allocs */ |
Christoph Lameter | 9b819d2 | 2006-09-25 23:31:40 -0700 | [diff] [blame] | 51 | #define __GFP_THISNODE ((__force gfp_t)0x40000u)/* No fallback, no policies */ |
Mel Gorman | e12ba74 | 2007-10-16 01:25:52 -0700 | [diff] [blame] | 52 | #define __GFP_RECLAIMABLE ((__force gfp_t)0x80000u) /* Page is reclaimable */ |
| 53 | #define __GFP_MOVABLE ((__force gfp_t)0x100000u) /* Page is movable */ |
Vegard Nossum | b1eeab6 | 2008-11-25 16:55:53 +0100 | [diff] [blame^] | 54 | |
| 55 | #ifdef CONFIG_KMEMCHECK |
Vegard Nossum | 2dff440 | 2008-05-31 15:56:17 +0200 | [diff] [blame] | 56 | #define __GFP_NOTRACK ((__force gfp_t)0x200000u) /* Don't track with kmemcheck */ |
Vegard Nossum | b1eeab6 | 2008-11-25 16:55:53 +0100 | [diff] [blame^] | 57 | #else |
| 58 | #define __GFP_NOTRACK ((__force gfp_t)0) |
| 59 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | |
Vegard Nossum | 2dff440 | 2008-05-31 15:56:17 +0200 | [diff] [blame] | 61 | /* |
| 62 | * This may seem redundant, but it's a way of annotating false positives vs. |
| 63 | * allocations that simply cannot be supported (e.g. page tables). |
| 64 | */ |
| 65 | #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK) |
| 66 | |
| 67 | #define __GFP_BITS_SHIFT 22 /* Room for 22 __GFP_FOO bits */ |
Al Viro | af4ca45 | 2005-10-21 02:55:38 -0400 | [diff] [blame] | 68 | #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | |
Jeff Dike | 7b04d71 | 2006-04-10 22:53:27 -0700 | [diff] [blame] | 70 | /* This equals 0, but use constants in case they ever change */ |
| 71 | #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH) |
Paul Jackson | 4eac915 | 2006-01-11 12:17:19 -0800 | [diff] [blame] | 72 | /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 73 | #define GFP_ATOMIC (__GFP_HIGH) |
| 74 | #define GFP_NOIO (__GFP_WAIT) |
| 75 | #define GFP_NOFS (__GFP_WAIT | __GFP_IO) |
| 76 | #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS) |
Mel Gorman | e12ba74 | 2007-10-16 01:25:52 -0700 | [diff] [blame] | 77 | #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \ |
| 78 | __GFP_RECLAIMABLE) |
Paul Jackson | f90b1d2 | 2005-09-06 15:18:10 -0700 | [diff] [blame] | 79 | #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL) |
| 80 | #define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \ |
| 81 | __GFP_HIGHMEM) |
Mel Gorman | 769848c | 2007-07-17 04:03:05 -0700 | [diff] [blame] | 82 | #define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \ |
| 83 | __GFP_HARDWALL | __GFP_HIGHMEM | \ |
| 84 | __GFP_MOVABLE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 85 | |
Christoph Lameter | 77f700d | 2006-09-27 01:50:07 -0700 | [diff] [blame] | 86 | #ifdef CONFIG_NUMA |
Christoph Lameter | 980128f | 2006-09-25 23:31:46 -0700 | [diff] [blame] | 87 | #define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY) |
Christoph Lameter | 77f700d | 2006-09-27 01:50:07 -0700 | [diff] [blame] | 88 | #else |
Al Viro | f2e97df | 2007-02-09 16:38:55 +0000 | [diff] [blame] | 89 | #define GFP_THISNODE ((__force gfp_t)0) |
Christoph Lameter | 77f700d | 2006-09-27 01:50:07 -0700 | [diff] [blame] | 90 | #endif |
| 91 | |
Christoph Lameter | 6cb0622 | 2007-10-16 01:25:41 -0700 | [diff] [blame] | 92 | /* This mask makes up all the page movable related flags */ |
Mel Gorman | e12ba74 | 2007-10-16 01:25:52 -0700 | [diff] [blame] | 93 | #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE) |
Christoph Lameter | 6cb0622 | 2007-10-16 01:25:41 -0700 | [diff] [blame] | 94 | |
| 95 | /* Control page allocator reclaim behavior */ |
| 96 | #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\ |
| 97 | __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\ |
| 98 | __GFP_NORETRY|__GFP_NOMEMALLOC) |
| 99 | |
| 100 | /* Control allocation constraints */ |
| 101 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) |
| 102 | |
| 103 | /* Do not use these with a slab allocator */ |
| 104 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) |
Christoph Lameter | 980128f | 2006-09-25 23:31:46 -0700 | [diff] [blame] | 105 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 106 | /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some |
| 107 | platforms, used as appropriate on others */ |
| 108 | |
| 109 | #define GFP_DMA __GFP_DMA |
| 110 | |
Andi Kleen | a2f1b42 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 111 | /* 4GB DMA on some platforms */ |
| 112 | #define GFP_DMA32 __GFP_DMA32 |
| 113 | |
Mel Gorman | 467c996 | 2007-10-16 01:26:02 -0700 | [diff] [blame] | 114 | /* Convert GFP flags to their corresponding migrate type */ |
| 115 | static inline int allocflags_to_migratetype(gfp_t gfp_flags) |
| 116 | { |
| 117 | WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK); |
| 118 | |
| 119 | if (unlikely(page_group_by_mobility_disabled)) |
| 120 | return MIGRATE_UNMOVABLE; |
| 121 | |
| 122 | /* Group based on mobility */ |
| 123 | return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) | |
| 124 | ((gfp_flags & __GFP_RECLAIMABLE) != 0); |
| 125 | } |
Andi Kleen | a2f1b42 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 126 | |
Christoph Lameter | 19655d3 | 2006-09-25 23:31:19 -0700 | [diff] [blame] | 127 | static inline enum zone_type gfp_zone(gfp_t flags) |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 128 | { |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 129 | #ifdef CONFIG_ZONE_DMA |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 130 | if (flags & __GFP_DMA) |
KAMEZAWA Hiroyuki | 8cece85 | 2008-04-28 02:13:36 -0700 | [diff] [blame] | 131 | return ZONE_DMA; |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 132 | #endif |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 133 | #ifdef CONFIG_ZONE_DMA32 |
| 134 | if (flags & __GFP_DMA32) |
KAMEZAWA Hiroyuki | 8cece85 | 2008-04-28 02:13:36 -0700 | [diff] [blame] | 135 | return ZONE_DMA32; |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 136 | #endif |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 137 | if ((flags & (__GFP_HIGHMEM | __GFP_MOVABLE)) == |
| 138 | (__GFP_HIGHMEM | __GFP_MOVABLE)) |
KAMEZAWA Hiroyuki | 8cece85 | 2008-04-28 02:13:36 -0700 | [diff] [blame] | 139 | return ZONE_MOVABLE; |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 140 | #ifdef CONFIG_HIGHMEM |
| 141 | if (flags & __GFP_HIGHMEM) |
KAMEZAWA Hiroyuki | 8cece85 | 2008-04-28 02:13:36 -0700 | [diff] [blame] | 142 | return ZONE_HIGHMEM; |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 143 | #endif |
KAMEZAWA Hiroyuki | 8cece85 | 2008-04-28 02:13:36 -0700 | [diff] [blame] | 144 | return ZONE_NORMAL; |
Christoph Lameter | 4e4785b | 2006-09-25 23:31:17 -0700 | [diff] [blame] | 145 | } |
| 146 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 147 | /* |
| 148 | * There is only one page-allocator function, and two main namespaces to |
| 149 | * it. The alloc_page*() variants return 'struct page *' and as such |
| 150 | * can allocate highmem pages, the *get*page*() variants return |
| 151 | * virtual kernel addresses to the allocated page(s). |
| 152 | */ |
| 153 | |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 154 | static inline int gfp_zonelist(gfp_t flags) |
| 155 | { |
| 156 | if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE)) |
| 157 | return 1; |
| 158 | |
| 159 | return 0; |
| 160 | } |
| 161 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 162 | /* |
| 163 | * We get the zone list from the current node and the gfp_mask. |
| 164 | * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones. |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 165 | * There are two zonelists per node, one for all zones with memory and |
| 166 | * one containing just zones from the node the zonelist belongs to. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 167 | * |
| 168 | * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets |
| 169 | * optimized to &contig_page_data at compile-time. |
| 170 | */ |
Mel Gorman | 0e88460 | 2008-04-28 02:12:14 -0700 | [diff] [blame] | 171 | static inline struct zonelist *node_zonelist(int nid, gfp_t flags) |
| 172 | { |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 173 | return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags); |
Mel Gorman | 0e88460 | 2008-04-28 02:12:14 -0700 | [diff] [blame] | 174 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | |
| 176 | #ifndef HAVE_ARCH_FREE_PAGE |
| 177 | static inline void arch_free_page(struct page *page, int order) { } |
| 178 | #endif |
Nick Piggin | cc102509 | 2006-12-06 20:32:00 -0800 | [diff] [blame] | 179 | #ifndef HAVE_ARCH_ALLOC_PAGE |
| 180 | static inline void arch_alloc_page(struct page *page, int order) { } |
| 181 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | |
KOSAKI Motohiro | e4048e5 | 2008-07-23 21:27:01 -0700 | [diff] [blame] | 183 | struct page * |
| 184 | __alloc_pages_internal(gfp_t gfp_mask, unsigned int order, |
| 185 | struct zonelist *zonelist, nodemask_t *nodemask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 186 | |
KOSAKI Motohiro | e4048e5 | 2008-07-23 21:27:01 -0700 | [diff] [blame] | 187 | static inline struct page * |
| 188 | __alloc_pages(gfp_t gfp_mask, unsigned int order, |
| 189 | struct zonelist *zonelist) |
| 190 | { |
| 191 | return __alloc_pages_internal(gfp_mask, order, zonelist, NULL); |
| 192 | } |
| 193 | |
| 194 | static inline struct page * |
| 195 | __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, |
| 196 | struct zonelist *zonelist, nodemask_t *nodemask) |
| 197 | { |
| 198 | return __alloc_pages_internal(gfp_mask, order, zonelist, nodemask); |
| 199 | } |
| 200 | |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 201 | |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 202 | static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 203 | unsigned int order) |
| 204 | { |
| 205 | if (unlikely(order >= MAX_ORDER)) |
| 206 | return NULL; |
| 207 | |
Andi Kleen | 819a692 | 2006-01-11 22:43:45 +0100 | [diff] [blame] | 208 | /* Unknown node is current node */ |
| 209 | if (nid < 0) |
| 210 | nid = numa_node_id(); |
| 211 | |
Mel Gorman | 0e88460 | 2008-04-28 02:12:14 -0700 | [diff] [blame] | 212 | return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | } |
| 214 | |
| 215 | #ifdef CONFIG_NUMA |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 216 | extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | |
| 218 | static inline struct page * |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 219 | alloc_pages(gfp_t gfp_mask, unsigned int order) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 220 | { |
| 221 | if (unlikely(order >= MAX_ORDER)) |
| 222 | return NULL; |
| 223 | |
| 224 | return alloc_pages_current(gfp_mask, order); |
| 225 | } |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 226 | extern struct page *alloc_page_vma(gfp_t gfp_mask, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 227 | struct vm_area_struct *vma, unsigned long addr); |
| 228 | #else |
| 229 | #define alloc_pages(gfp_mask, order) \ |
| 230 | alloc_pages_node(numa_node_id(), gfp_mask, order) |
| 231 | #define alloc_page_vma(gfp_mask, vma, addr) alloc_pages(gfp_mask, 0) |
| 232 | #endif |
| 233 | #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0) |
| 234 | |
Harvey Harrison | b3c9752 | 2008-02-13 15:03:15 -0800 | [diff] [blame] | 235 | extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order); |
| 236 | extern unsigned long get_zeroed_page(gfp_t gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 237 | |
Timur Tabi | 2be0ffe | 2008-07-23 21:28:11 -0700 | [diff] [blame] | 238 | void *alloc_pages_exact(size_t size, gfp_t gfp_mask); |
| 239 | void free_pages_exact(void *virt, size_t size); |
| 240 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 241 | #define __get_free_page(gfp_mask) \ |
| 242 | __get_free_pages((gfp_mask),0) |
| 243 | |
| 244 | #define __get_dma_pages(gfp_mask, order) \ |
| 245 | __get_free_pages((gfp_mask) | GFP_DMA,(order)) |
| 246 | |
Harvey Harrison | b3c9752 | 2008-02-13 15:03:15 -0800 | [diff] [blame] | 247 | extern void __free_pages(struct page *page, unsigned int order); |
| 248 | extern void free_pages(unsigned long addr, unsigned int order); |
| 249 | extern void free_hot_page(struct page *page); |
| 250 | extern void free_cold_page(struct page *page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 251 | |
| 252 | #define __free_page(page) __free_pages((page), 0) |
| 253 | #define free_page(addr) free_pages((addr),0) |
| 254 | |
| 255 | void page_alloc_init(void); |
Christoph Lameter | 4037d45 | 2007-05-09 02:35:14 -0700 | [diff] [blame] | 256 | void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp); |
Christoph Lameter | 9f8f217 | 2008-02-04 22:29:11 -0800 | [diff] [blame] | 257 | void drain_all_pages(void); |
| 258 | void drain_local_pages(void *dummy); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | |
| 260 | #endif /* __LINUX_GFP_H */ |