1 /*- 2 * Copyright (c) 2002-2005, 2009, 2013 Jeffrey Roberson <jeff@FreeBSD.org> 3 * Copyright (c) 2004, 2005 Bosko Milekic <bmilekic@FreeBSD.org> 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice unmodified, this list of conditions, and the following 11 * disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 * 27 * $FreeBSD$ 28 * 29 */ 30 31 /* 32 * This file includes definitions, structures, prototypes, and inlines that 33 * should not be used outside of the actual implementation of UMA. 34 */ 35 36 /* 37 * Here's a quick description of the relationship between the objects: 38 * 39 * Kegs contain lists of slabs which are stored in either the full bin, empty 40 * bin, or partially allocated bin, to reduce fragmentation. They also contain 41 * the user supplied value for size, which is adjusted for alignment purposes 42 * and rsize is the result of that. The Keg also stores information for 43 * managing a hash of page addresses that maps pages to uma_slab_t structures 44 * for pages that don't have embedded uma_slab_t's. 45 * 46 * The uma_slab_t may be embedded in a UMA_SLAB_SIZE chunk of memory or it may 47 * be allocated off the page from a special slab zone. The free list within a 48 * slab is managed with a bitmask. For item sizes that would yield more than 49 * 10% memory waste we potentially allocate a separate uma_slab_t if this will 50 * improve the number of items per slab that will fit. 51 * 52 * Other potential space optimizations are storing the 8bit of linkage in space 53 * wasted between items due to alignment problems. This may yield a much better 54 * memory footprint for certain sizes of objects. Another alternative is to 55 * increase the UMA_SLAB_SIZE, or allow for dynamic slab sizes. I prefer 56 * dynamic slab sizes because we could stick with 8 bit indices and only use 57 * large slab sizes for zones with a lot of waste per slab. This may create 58 * inefficiencies in the vm subsystem due to fragmentation in the address space. 59 * 60 * The only really gross cases, with regards to memory waste, are for those 61 * items that are just over half the page size. You can get nearly 50% waste, 62 * so you fall back to the memory footprint of the power of two allocator. I 63 * have looked at memory allocation sizes on many of the machines available to 64 * me, and there does not seem to be an abundance of allocations at this range 65 * so at this time it may not make sense to optimize for it. This can, of 66 * course, be solved with dynamic slab sizes. 67 * 68 * Kegs may serve multiple Zones but by far most of the time they only serve 69 * one. When a Zone is created, a Keg is allocated and setup for it. While 70 * the backing Keg stores slabs, the Zone caches Buckets of items allocated 71 * from the slabs. Each Zone is equipped with an init/fini and ctor/dtor 72 * pair, as well as with its own set of small per-CPU caches, layered above 73 * the Zone's general Bucket cache. 74 * 75 * The PCPU caches are protected by critical sections, and may be accessed 76 * safely only from their associated CPU, while the Zones backed by the same 77 * Keg all share a common Keg lock (to coalesce contention on the backing 78 * slabs). The backing Keg typically only serves one Zone but in the case of 79 * multiple Zones, one of the Zones is considered the Master Zone and all 80 * Zone-related stats from the Keg are done in the Master Zone. For an 81 * example of a Multi-Zone setup, refer to the Mbuf allocation code. 82 */ 83 84 /* 85 * This is the representation for normal (Non OFFPAGE slab) 86 * 87 * i == item 88 * s == slab pointer 89 * 90 * <---------------- Page (UMA_SLAB_SIZE) ------------------> 91 * ___________________________________________________________ 92 * | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ___________ | 93 * ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i| |slab header|| 94 * ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_| |___________|| 95 * |___________________________________________________________| 96 * 97 * 98 * This is an OFFPAGE slab. These can be larger than UMA_SLAB_SIZE. 99 * 100 * ___________________________________________________________ 101 * | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ | 102 * ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i| | 103 * ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_| | 104 * |___________________________________________________________| 105 * ___________ ^ 106 * |slab header| | 107 * |___________|---* 108 * 109 */ 110 111 #ifndef VM_UMA_INT_H 112 #define VM_UMA_INT_H 113 114 #define UMA_SLAB_SIZE PAGE_SIZE /* How big are our slabs? */ 115 #define UMA_SLAB_MASK (PAGE_SIZE - 1) /* Mask to get back to the page */ 116 #define UMA_SLAB_SHIFT PAGE_SHIFT /* Number of bits PAGE_MASK */ 117 118 #define UMA_BOOT_PAGES 64 /* Pages allocated for startup */ 119 120 /* Max waste percentage before going to off page slab management */ 121 #define UMA_MAX_WASTE 10 122 123 /* 124 * I doubt there will be many cases where this is exceeded. This is the initial 125 * size of the hash table for uma_slabs that are managed off page. This hash 126 * does expand by powers of two. Currently it doesn't get smaller. 127 */ 128 #define UMA_HASH_SIZE_INIT 32 129 130 /* 131 * I should investigate other hashing algorithms. This should yield a low 132 * number of collisions if the pages are relatively contiguous. 133 */ 134 135 #define UMA_HASH(h, s) ((((uintptr_t)s) >> UMA_SLAB_SHIFT) & (h)->uh_hashmask) 136 137 #define UMA_HASH_INSERT(h, s, mem) \ 138 SLIST_INSERT_HEAD(&(h)->uh_slab_hash[UMA_HASH((h), \ 139 (mem))], (s), us_hlink) 140 #define UMA_HASH_REMOVE(h, s, mem) \ 141 SLIST_REMOVE(&(h)->uh_slab_hash[UMA_HASH((h), \ 142 (mem))], (s), uma_slab, us_hlink) 143 144 /* Hash table for freed address -> slab translation */ 145 146 SLIST_HEAD(slabhead, uma_slab); 147 148 struct uma_hash { 149 struct slabhead *uh_slab_hash; /* Hash table for slabs */ 150 int uh_hashsize; /* Current size of the hash table */ 151 int uh_hashmask; /* Mask used during hashing */ 152 }; 153 154 /* 155 * align field or structure to cache line 156 */ 157 #if defined(__amd64__) 158 #define UMA_ALIGN __aligned(CACHE_LINE_SIZE) 159 #else 160 #define UMA_ALIGN 161 #endif 162 163 /* 164 * Structures for per cpu queues. 165 */ 166 167 struct uma_bucket { 168 LIST_ENTRY(uma_bucket) ub_link; /* Link into the zone */ 169 int16_t ub_cnt; /* Count of free items. */ 170 int16_t ub_entries; /* Max items. */ 171 void *ub_bucket[]; /* actual allocation storage */ 172 }; 173 174 typedef struct uma_bucket * uma_bucket_t; 175 176 struct uma_cache { 177 uma_bucket_t uc_freebucket; /* Bucket we're freeing to */ 178 uma_bucket_t uc_allocbucket; /* Bucket to allocate from */ 179 uint64_t uc_allocs; /* Count of allocations */ 180 uint64_t uc_frees; /* Count of frees */ 181 } UMA_ALIGN; 182 183 typedef struct uma_cache * uma_cache_t; 184 185 /* 186 * Keg management structure 187 * 188 * TODO: Optimize for cache line size 189 * 190 */ 191 struct uma_keg { 192 struct mtx_padalign uk_lock; /* Lock for the keg */ 193 struct uma_hash uk_hash; 194 195 LIST_HEAD(,uma_zone) uk_zones; /* Keg's zones */ 196 LIST_HEAD(,uma_slab) uk_part_slab; /* partially allocated slabs */ 197 LIST_HEAD(,uma_slab) uk_free_slab; /* empty slab list */ 198 LIST_HEAD(,uma_slab) uk_full_slab; /* full slabs */ 199 200 uint32_t uk_align; /* Alignment mask */ 201 uint32_t uk_pages; /* Total page count */ 202 uint32_t uk_free; /* Count of items free in slabs */ 203 uint32_t uk_reserve; /* Number of reserved items. */ 204 uint32_t uk_size; /* Requested size of each item */ 205 uint32_t uk_rsize; /* Real size of each item */ 206 uint32_t uk_maxpages; /* Maximum number of pages to alloc */ 207 208 uma_init uk_init; /* Keg's init routine */ 209 uma_fini uk_fini; /* Keg's fini routine */ 210 uma_alloc uk_allocf; /* Allocation function */ 211 uma_free uk_freef; /* Free routine */ 212 213 u_long uk_offset; /* Next free offset from base KVA */ 214 vm_offset_t uk_kva; /* Zone base KVA */ 215 uma_zone_t uk_slabzone; /* Slab zone backing us, if OFFPAGE */ 216 217 uint16_t uk_slabsize; /* Slab size for this keg */ 218 uint16_t uk_pgoff; /* Offset to uma_slab struct */ 219 uint16_t uk_ppera; /* pages per allocation from backend */ 220 uint16_t uk_ipers; /* Items per slab */ 221 uint32_t uk_flags; /* Internal flags */ 222 223 /* Least used fields go to the last cache line. */ 224 const char *uk_name; /* Name of creating zone. */ 225 LIST_ENTRY(uma_keg) uk_link; /* List of all kegs */ 226 }; 227 typedef struct uma_keg * uma_keg_t; 228 229 /* 230 * Free bits per-slab. 231 */ 232 #define SLAB_SETSIZE (PAGE_SIZE / UMA_SMALLEST_UNIT) 233 BITSET_DEFINE(slabbits, SLAB_SETSIZE); 234 235 /* 236 * The slab structure manages a single contiguous allocation from backing 237 * store and subdivides it into individually allocatable items. 238 */ 239 struct uma_slab { 240 uma_keg_t us_keg; /* Keg we live in */ 241 union { 242 LIST_ENTRY(uma_slab) _us_link; /* slabs in zone */ 243 unsigned long _us_size; /* Size of allocation */ 244 } us_type; 245 SLIST_ENTRY(uma_slab) us_hlink; /* Link for hash table */ 246 uint8_t *us_data; /* First item */ 247 struct slabbits us_free; /* Free bitmask. */ 248 #ifdef INVARIANTS 249 struct slabbits us_debugfree; /* Debug bitmask. */ 250 #endif 251 uint16_t us_freecount; /* How many are free? */ 252 uint8_t us_flags; /* Page flags see uma.h */ 253 uint8_t us_pad; /* Pad to 32bits, unused. */ 254 }; 255 256 #define us_link us_type._us_link 257 #define us_size us_type._us_size 258 259 /* 260 * The slab structure for UMA_ZONE_REFCNT zones for whose items we 261 * maintain reference counters in the slab for. 262 */ 263 struct uma_slab_refcnt { 264 struct uma_slab us_head; /* slab header data */ 265 uint32_t us_refcnt[0]; /* Actually larger. */ 266 }; 267 268 typedef struct uma_slab * uma_slab_t; 269 typedef struct uma_slab_refcnt * uma_slabrefcnt_t; 270 typedef uma_slab_t (*uma_slaballoc)(uma_zone_t, uma_keg_t, int); 271 272 struct uma_klink { 273 LIST_ENTRY(uma_klink) kl_link; 274 uma_keg_t kl_keg; 275 }; 276 typedef struct uma_klink *uma_klink_t; 277 278 /* 279 * Zone management structure 280 * 281 * TODO: Optimize for cache line size 282 * 283 */ 284 struct uma_zone { 285 struct mtx_padalign uz_lock; /* Lock for the zone */ 286 struct mtx_padalign *uz_lockptr; 287 const char *uz_name; /* Text name of the zone */ 288 289 LIST_ENTRY(uma_zone) uz_link; /* List of all zones in keg */ 290 LIST_HEAD(,uma_bucket) uz_buckets; /* full buckets */ 291 292 LIST_HEAD(,uma_klink) uz_kegs; /* List of kegs. */ 293 struct uma_klink uz_klink; /* klink for first keg. */ 294 295 uma_slaballoc uz_slab; /* Allocate a slab from the backend. */ 296 uma_ctor uz_ctor; /* Constructor for each allocation */ 297 uma_dtor uz_dtor; /* Destructor */ 298 uma_init uz_init; /* Initializer for each item */ 299 uma_fini uz_fini; /* Finalizer for each item. */ 300 uma_import uz_import; /* Import new memory to cache. */ 301 uma_release uz_release; /* Release memory from cache. */ 302 void *uz_arg; /* Import/release argument. */ 303 304 uint32_t uz_flags; /* Flags inherited from kegs */ 305 uint32_t uz_size; /* Size inherited from kegs */ 306 307 volatile u_long uz_allocs UMA_ALIGN; /* Total number of allocations */ 308 volatile u_long uz_fails; /* Total number of alloc failures */ 309 volatile u_long uz_frees; /* Total number of frees */ 310 uint64_t uz_sleeps; /* Total number of alloc sleeps */ 311 uint16_t uz_count; /* Highest amount of items in bucket */ 312 313 /* The next three fields are used to print a rate-limited warnings. */ 314 const char *uz_warning; /* Warning to print on failure */ 315 struct timeval uz_ratecheck; /* Warnings rate-limiting */ 316 317 /* 318 * This HAS to be the last item because we adjust the zone size 319 * based on NCPU and then allocate the space for the zones. 320 */ 321 struct uma_cache uz_cpu[1]; /* Per cpu caches */ 322 }; 323 324 /* 325 * These flags must not overlap with the UMA_ZONE flags specified in uma.h. 326 */ 327 #define UMA_ZFLAG_MULTI 0x04000000 /* Multiple kegs in the zone. */ 328 #define UMA_ZFLAG_DRAINING 0x08000000 /* Running zone_drain. */ 329 #define UMA_ZFLAG_BUCKET 0x10000000 /* Bucket zone. */ 330 #define UMA_ZFLAG_INTERNAL 0x20000000 /* No offpage no PCPU. */ 331 #define UMA_ZFLAG_FULL 0x40000000 /* Reached uz_maxpages */ 332 #define UMA_ZFLAG_CACHEONLY 0x80000000 /* Don't ask VM for buckets. */ 333 334 #define UMA_ZFLAG_INHERIT \ 335 (UMA_ZFLAG_INTERNAL | UMA_ZFLAG_CACHEONLY | UMA_ZFLAG_BUCKET) 336 337 static inline uma_keg_t 338 zone_first_keg(uma_zone_t zone) 339 { 340 uma_klink_t klink; 341 342 klink = LIST_FIRST(&zone->uz_kegs); 343 return (klink != NULL) ? klink->kl_keg : NULL; 344 } 345 346 #undef UMA_ALIGN 347 348 #ifdef _KERNEL 349 /* Internal prototypes */ 350 static __inline uma_slab_t hash_sfind(struct uma_hash *hash, uint8_t *data); 351 void *uma_large_malloc(int size, int wait); 352 void uma_large_free(uma_slab_t slab); 353 354 /* Lock Macros */ 355 356 #define KEG_LOCK_INIT(k, lc) \ 357 do { \ 358 if ((lc)) \ 359 mtx_init(&(k)->uk_lock, (k)->uk_name, \ 360 (k)->uk_name, MTX_DEF | MTX_DUPOK); \ 361 else \ 362 mtx_init(&(k)->uk_lock, (k)->uk_name, \ 363 "UMA zone", MTX_DEF | MTX_DUPOK); \ 364 } while (0) 365 366 #define KEG_LOCK_FINI(k) mtx_destroy(&(k)->uk_lock) 367 #define KEG_LOCK(k) mtx_lock(&(k)->uk_lock) 368 #define KEG_UNLOCK(k) mtx_unlock(&(k)->uk_lock) 369 370 #define ZONE_LOCK_INIT(z, lc) \ 371 do { \ 372 if ((lc)) \ 373 mtx_init(&(z)->uz_lock, (z)->uz_name, \ 374 (z)->uz_name, MTX_DEF | MTX_DUPOK); \ 375 else \ 376 mtx_init(&(z)->uz_lock, (z)->uz_name, \ 377 "UMA zone", MTX_DEF | MTX_DUPOK); \ 378 } while (0) 379 380 #define ZONE_LOCK(z) mtx_lock((z)->uz_lockptr) 381 #define ZONE_TRYLOCK(z) mtx_trylock((z)->uz_lockptr) 382 #define ZONE_UNLOCK(z) mtx_unlock((z)->uz_lockptr) 383 #define ZONE_LOCK_FINI(z) mtx_destroy(&(z)->uz_lock) 384 385 /* 386 * Find a slab within a hash table. This is used for OFFPAGE zones to lookup 387 * the slab structure. 388 * 389 * Arguments: 390 * hash The hash table to search. 391 * data The base page of the item. 392 * 393 * Returns: 394 * A pointer to a slab if successful, else NULL. 395 */ 396 static __inline uma_slab_t 397 hash_sfind(struct uma_hash *hash, uint8_t *data) 398 { 399 uma_slab_t slab; 400 int hval; 401 402 hval = UMA_HASH(hash, data); 403 404 SLIST_FOREACH(slab, &hash->uh_slab_hash[hval], us_hlink) { 405 if ((uint8_t *)slab->us_data == data) 406 return (slab); 407 } 408 return (NULL); 409 } 410 411 static __inline uma_slab_t 412 vtoslab(vm_offset_t va) 413 { 414 vm_page_t p; 415 uma_slab_t slab; 416 417 p = PHYS_TO_VM_PAGE(pmap_kextract(va)); 418 slab = (uma_slab_t )p->object; 419 420 if (p->flags & PG_SLAB) 421 return (slab); 422 else 423 return (NULL); 424 } 425 426 static __inline void 427 vsetslab(vm_offset_t va, uma_slab_t slab) 428 { 429 vm_page_t p; 430 431 p = PHYS_TO_VM_PAGE(pmap_kextract(va)); 432 p->object = (vm_object_t)slab; 433 p->flags |= PG_SLAB; 434 } 435 436 static __inline void 437 vsetobj(vm_offset_t va, vm_object_t obj) 438 { 439 vm_page_t p; 440 441 p = PHYS_TO_VM_PAGE(pmap_kextract(va)); 442 p->object = obj; 443 p->flags &= ~PG_SLAB; 444 } 445 446 /* 447 * The following two functions may be defined by architecture specific code 448 * if they can provide more effecient allocation functions. This is useful 449 * for using direct mapped addresses. 450 */ 451 void *uma_small_alloc(uma_zone_t zone, int bytes, uint8_t *pflag, int wait); 452 void uma_small_free(void *mem, int size, uint8_t flags); 453 #endif /* _KERNEL */ 454 455 #endif /* VM_UMA_INT_H */ 456