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 */ 18 typedef 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 26 enum { 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 #ifdef CONFIG_SLAB_OBJ_EXT 59 ___GFP_NO_OBJ_EXT_BIT, 60 #endif 61 ___GFP_LAST_BIT 62 }; 63 64 /* Plain integer GFP bitmasks. Do not use this directly. */ 65 #define ___GFP_DMA BIT(___GFP_DMA_BIT) 66 #define ___GFP_HIGHMEM BIT(___GFP_HIGHMEM_BIT) 67 #define ___GFP_DMA32 BIT(___GFP_DMA32_BIT) 68 #define ___GFP_MOVABLE BIT(___GFP_MOVABLE_BIT) 69 #define ___GFP_RECLAIMABLE BIT(___GFP_RECLAIMABLE_BIT) 70 #define ___GFP_HIGH BIT(___GFP_HIGH_BIT) 71 #define ___GFP_IO BIT(___GFP_IO_BIT) 72 #define ___GFP_FS BIT(___GFP_FS_BIT) 73 #define ___GFP_ZERO BIT(___GFP_ZERO_BIT) 74 /* 0x200u unused */ 75 #define ___GFP_DIRECT_RECLAIM BIT(___GFP_DIRECT_RECLAIM_BIT) 76 #define ___GFP_KSWAPD_RECLAIM BIT(___GFP_KSWAPD_RECLAIM_BIT) 77 #define ___GFP_WRITE BIT(___GFP_WRITE_BIT) 78 #define ___GFP_NOWARN BIT(___GFP_NOWARN_BIT) 79 #define ___GFP_RETRY_MAYFAIL BIT(___GFP_RETRY_MAYFAIL_BIT) 80 #define ___GFP_NOFAIL BIT(___GFP_NOFAIL_BIT) 81 #define ___GFP_NORETRY BIT(___GFP_NORETRY_BIT) 82 #define ___GFP_MEMALLOC BIT(___GFP_MEMALLOC_BIT) 83 #define ___GFP_COMP BIT(___GFP_COMP_BIT) 84 #define ___GFP_NOMEMALLOC BIT(___GFP_NOMEMALLOC_BIT) 85 #define ___GFP_HARDWALL BIT(___GFP_HARDWALL_BIT) 86 #define ___GFP_THISNODE BIT(___GFP_THISNODE_BIT) 87 #define ___GFP_ACCOUNT BIT(___GFP_ACCOUNT_BIT) 88 #define ___GFP_ZEROTAGS BIT(___GFP_ZEROTAGS_BIT) 89 #ifdef CONFIG_KASAN_HW_TAGS 90 #define ___GFP_SKIP_ZERO BIT(___GFP_SKIP_ZERO_BIT) 91 #define ___GFP_SKIP_KASAN BIT(___GFP_SKIP_KASAN_BIT) 92 #else 93 #define ___GFP_SKIP_ZERO 0 94 #define ___GFP_SKIP_KASAN 0 95 #endif 96 #ifdef CONFIG_LOCKDEP 97 #define ___GFP_NOLOCKDEP BIT(___GFP_NOLOCKDEP_BIT) 98 #else 99 #define ___GFP_NOLOCKDEP 0 100 #endif 101 #ifdef CONFIG_SLAB_OBJ_EXT 102 #define ___GFP_NO_OBJ_EXT BIT(___GFP_NO_OBJ_EXT_BIT) 103 #else 104 #define ___GFP_NO_OBJ_EXT 0 105 #endif 106 107 /* 108 * Physical address zone modifiers (see linux/mmzone.h - low four bits) 109 * 110 * Do not put any conditional on these. If necessary modify the definitions 111 * without the underscores and use them consistently. The definitions here may 112 * be used in bit comparisons. 113 */ 114 #define __GFP_DMA ((__force gfp_t)___GFP_DMA) 115 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM) 116 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32) 117 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */ 118 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE) 119 120 /** 121 * DOC: Page mobility and placement hints 122 * 123 * Page mobility and placement hints 124 * --------------------------------- 125 * 126 * These flags provide hints about how mobile the page is. Pages with similar 127 * mobility are placed within the same pageblocks to minimise problems due 128 * to external fragmentation. 129 * 130 * %__GFP_MOVABLE (also a zone modifier) indicates that the page can be 131 * moved by page migration during memory compaction or can be reclaimed. 132 * 133 * %__GFP_RECLAIMABLE is used for slab allocations that specify 134 * SLAB_RECLAIM_ACCOUNT and whose pages can be freed via shrinkers. 135 * 136 * %__GFP_WRITE indicates the caller intends to dirty the page. Where possible, 137 * these pages will be spread between local zones to avoid all the dirty 138 * pages being in one zone (fair zone allocation policy). 139 * 140 * %__GFP_HARDWALL enforces the cpuset memory allocation policy. 141 * 142 * %__GFP_THISNODE forces the allocation to be satisfied from the requested 143 * node with no fallbacks or placement policy enforcements. 144 * 145 * %__GFP_ACCOUNT causes the allocation to be accounted to kmemcg. 146 * 147 * %__GFP_NO_OBJ_EXT causes slab allocation to have no object extension. 148 */ 149 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) 150 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) 151 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) 152 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE) 153 #define __GFP_ACCOUNT ((__force gfp_t)___GFP_ACCOUNT) 154 #define __GFP_NO_OBJ_EXT ((__force gfp_t)___GFP_NO_OBJ_EXT) 155 156 /** 157 * DOC: Watermark modifiers 158 * 159 * Watermark modifiers -- controls access to emergency reserves 160 * ------------------------------------------------------------ 161 * 162 * %__GFP_HIGH indicates that the caller is high-priority and that granting 163 * the request is necessary before the system can make forward progress. 164 * For example creating an IO context to clean pages and requests 165 * from atomic context. 166 * 167 * %__GFP_MEMALLOC allows access to all memory. This should only be used when 168 * the caller guarantees the allocation will allow more memory to be freed 169 * very shortly e.g. process exiting or swapping. Users either should 170 * be the MM or co-ordinating closely with the VM (e.g. swap over NFS). 171 * Users of this flag have to be extremely careful to not deplete the reserve 172 * completely and implement a throttling mechanism which controls the 173 * consumption of the reserve based on the amount of freed memory. 174 * Usage of a pre-allocated pool (e.g. mempool) should be always considered 175 * before using this flag. 176 * 177 * %__GFP_NOMEMALLOC is used to explicitly forbid access to emergency reserves. 178 * This takes precedence over the %__GFP_MEMALLOC flag if both are set. 179 */ 180 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) 181 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC) 182 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) 183 184 /** 185 * DOC: Reclaim modifiers 186 * 187 * Reclaim modifiers 188 * ----------------- 189 * Please note that all the following flags are only applicable to sleepable 190 * allocations (e.g. %GFP_NOWAIT and %GFP_ATOMIC will ignore them). 191 * 192 * %__GFP_IO can start physical IO. 193 * 194 * %__GFP_FS can call down to the low-level FS. Clearing the flag avoids the 195 * allocator recursing into the filesystem which might already be holding 196 * locks. 197 * 198 * %__GFP_DIRECT_RECLAIM indicates that the caller may enter direct reclaim. 199 * This flag can be cleared to avoid unnecessary delays when a fallback 200 * option is available. 201 * 202 * %__GFP_KSWAPD_RECLAIM indicates that the caller wants to wake kswapd when 203 * the low watermark is reached and have it reclaim pages until the high 204 * watermark is reached. A caller may wish to clear this flag when fallback 205 * options are available and the reclaim is likely to disrupt the system. The 206 * canonical example is THP allocation where a fallback is cheap but 207 * reclaim/compaction may cause indirect stalls. 208 * 209 * %__GFP_RECLAIM is shorthand to allow/forbid both direct and kswapd reclaim. 210 * 211 * The default allocator behavior depends on the request size. We have a concept 212 * of so-called costly allocations (with order > %PAGE_ALLOC_COSTLY_ORDER). 213 * !costly allocations are too essential to fail so they are implicitly 214 * non-failing by default (with some exceptions like OOM victims might fail so 215 * the caller still has to check for failures) while costly requests try to be 216 * not disruptive and back off even without invoking the OOM killer. 217 * The following three modifiers might be used to override some of these 218 * implicit rules. 219 * 220 * %__GFP_NORETRY: The VM implementation will try only very lightweight 221 * memory direct reclaim to get some memory under memory pressure (thus 222 * it can sleep). It will avoid disruptive actions like OOM killer. The 223 * caller must handle the failure which is quite likely to happen under 224 * heavy memory pressure. The flag is suitable when failure can easily be 225 * handled at small cost, such as reduced throughput. 226 * 227 * %__GFP_RETRY_MAYFAIL: The VM implementation will retry memory reclaim 228 * procedures that have previously failed if there is some indication 229 * that progress has been made elsewhere. It can wait for other 230 * tasks to attempt high-level approaches to freeing memory such as 231 * compaction (which removes fragmentation) and page-out. 232 * There is still a definite limit to the number of retries, but it is 233 * a larger limit than with %__GFP_NORETRY. 234 * Allocations with this flag may fail, but only when there is 235 * genuinely little unused memory. While these allocations do not 236 * directly trigger the OOM killer, their failure indicates that 237 * the system is likely to need to use the OOM killer soon. The 238 * caller must handle failure, but can reasonably do so by failing 239 * a higher-level request, or completing it only in a much less 240 * efficient manner. 241 * If the allocation does fail, and the caller is in a position to 242 * free some non-essential memory, doing so could benefit the system 243 * as a whole. 244 * 245 * %__GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller 246 * cannot handle allocation failures. The allocation could block 247 * indefinitely but will never return with failure. Testing for 248 * failure is pointless. 249 * New users should be evaluated carefully (and the flag should be 250 * used only when there is no reasonable failure policy) but it is 251 * definitely preferable to use the flag rather than opencode endless 252 * loop around allocator. 253 * Using this flag for costly allocations is _highly_ discouraged. 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 and few other strict 315 * non-preemptive contexts (e.g. raw_spin_lock). The same applies to %GFP_NOWAIT. 316 * 317 * %GFP_KERNEL is typical for kernel-internal allocations. The caller requires 318 * %ZONE_NORMAL or a lower zone for direct access but can direct reclaim. 319 * 320 * %GFP_KERNEL_ACCOUNT is the same as GFP_KERNEL, except the allocation is 321 * accounted to kmemcg. 322 * 323 * %GFP_NOWAIT is for kernel allocations that should not stall for direct 324 * reclaim, start physical IO or use any filesystem callback. It is very 325 * likely to fail to allocate memory, even for very small allocations. 326 * 327 * %GFP_NOIO will use direct reclaim to discard clean pages or slab pages 328 * that do not require the starting of any physical IO. 329 * Please try to avoid using this flag directly and instead use 330 * memalloc_noio_{save,restore} to mark the whole scope which cannot 331 * perform any IO with a short explanation why. All allocation requests 332 * will inherit GFP_NOIO implicitly. 333 * 334 * %GFP_NOFS will use direct reclaim but will not use any filesystem interfaces. 335 * Please try to avoid using this flag directly and instead use 336 * memalloc_nofs_{save,restore} to mark the whole scope which cannot/shouldn't 337 * recurse into the FS layer with a short explanation why. All allocation 338 * requests will inherit GFP_NOFS implicitly. 339 * 340 * %GFP_USER is for userspace allocations that also need to be directly 341 * accessibly by the kernel or hardware. It is typically used by hardware 342 * for buffers that are mapped to userspace (e.g. graphics) that hardware 343 * still must DMA to. cpuset limits are enforced for these allocations. 344 * 345 * %GFP_DMA exists for historical reasons and should be avoided where possible. 346 * The flags indicates that the caller requires that the lowest zone be 347 * used (%ZONE_DMA or 16M on x86-64). Ideally, this would be removed but 348 * it would require careful auditing as some users really require it and 349 * others use the flag to avoid lowmem reserves in %ZONE_DMA and treat the 350 * lowest zone as a type of emergency reserve. 351 * 352 * %GFP_DMA32 is similar to %GFP_DMA except that the caller requires a 32-bit 353 * address. Note that kmalloc(..., GFP_DMA32) does not return DMA32 memory 354 * because the DMA32 kmalloc cache array is not implemented. 355 * (Reason: there is no such user in kernel). 356 * 357 * %GFP_HIGHUSER is for userspace allocations that may be mapped to userspace, 358 * do not need to be directly accessible by the kernel but that cannot 359 * move once in use. An example may be a hardware allocation that maps 360 * data directly into userspace but has no addressing limitations. 361 * 362 * %GFP_HIGHUSER_MOVABLE is for userspace allocations that the kernel does not 363 * need direct access to but can use kmap() when access is required. They 364 * are expected to be movable via page reclaim or page migration. Typically, 365 * pages on the LRU would also be allocated with %GFP_HIGHUSER_MOVABLE. 366 * 367 * %GFP_TRANSHUGE and %GFP_TRANSHUGE_LIGHT are used for THP allocations. They 368 * are compound allocations that will generally fail quickly if memory is not 369 * available and will not wake kswapd/kcompactd on failure. The _LIGHT 370 * version does not attempt reclaim/compaction at all and is by default used 371 * in page fault path, while the non-light is used by khugepaged. 372 */ 373 #define GFP_ATOMIC (__GFP_HIGH|__GFP_KSWAPD_RECLAIM) 374 #define GFP_KERNEL (__GFP_RECLAIM | __GFP_IO | __GFP_FS) 375 #define GFP_KERNEL_ACCOUNT (GFP_KERNEL | __GFP_ACCOUNT) 376 #define GFP_NOWAIT (__GFP_KSWAPD_RECLAIM | __GFP_NOWARN) 377 #define GFP_NOIO (__GFP_RECLAIM) 378 #define GFP_NOFS (__GFP_RECLAIM | __GFP_IO) 379 #define GFP_USER (__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL) 380 #define GFP_DMA __GFP_DMA 381 #define GFP_DMA32 __GFP_DMA32 382 #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM) 383 #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE | __GFP_SKIP_KASAN) 384 #define GFP_TRANSHUGE_LIGHT ((GFP_HIGHUSER_MOVABLE | __GFP_COMP | \ 385 __GFP_NOMEMALLOC | __GFP_NOWARN) & ~__GFP_RECLAIM) 386 #define GFP_TRANSHUGE (GFP_TRANSHUGE_LIGHT | __GFP_DIRECT_RECLAIM) 387 388 #endif /* __LINUX_GFP_TYPES_H */ 389