xref: /freebsd/sys/vm/uma_int.h (revision 484482b8ad59a8c68aaa1f0551a983c10b983a22)
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 #include <sys/_task.h>
32 
33 /*
34  * This file includes definitions, structures, prototypes, and inlines that
35  * should not be used outside of the actual implementation of UMA.
36  */
37 
38 /*
39  * Here's a quick description of the relationship between the objects:
40  *
41  * Kegs contain lists of slabs which are stored in either the full bin, empty
42  * bin, or partially allocated bin, to reduce fragmentation.  They also contain
43  * the user supplied value for size, which is adjusted for alignment purposes
44  * and rsize is the result of that.  The Keg also stores information for
45  * managing a hash of page addresses that maps pages to uma_slab_t structures
46  * for pages that don't have embedded uma_slab_t's.
47  *
48  * The uma_slab_t may be embedded in a UMA_SLAB_SIZE chunk of memory or it may
49  * be allocated off the page from a special slab zone.  The free list within a
50  * slab is managed with a bitmask.  For item sizes that would yield more than
51  * 10% memory waste we potentially allocate a separate uma_slab_t if this will
52  * improve the number of items per slab that will fit.
53  *
54  * The only really gross cases, with regards to memory waste, are for those
55  * items that are just over half the page size.   You can get nearly 50% waste,
56  * so you fall back to the memory footprint of the power of two allocator. I
57  * have looked at memory allocation sizes on many of the machines available to
58  * me, and there does not seem to be an abundance of allocations at this range
59  * so at this time it may not make sense to optimize for it.  This can, of
60  * course, be solved with dynamic slab sizes.
61  *
62  * Kegs may serve multiple Zones but by far most of the time they only serve
63  * one.  When a Zone is created, a Keg is allocated and setup for it.  While
64  * the backing Keg stores slabs, the Zone caches Buckets of items allocated
65  * from the slabs.  Each Zone is equipped with an init/fini and ctor/dtor
66  * pair, as well as with its own set of small per-CPU caches, layered above
67  * the Zone's general Bucket cache.
68  *
69  * The PCPU caches are protected by critical sections, and may be accessed
70  * safely only from their associated CPU, while the Zones backed by the same
71  * Keg all share a common Keg lock (to coalesce contention on the backing
72  * slabs).  The backing Keg typically only serves one Zone but in the case of
73  * multiple Zones, one of the Zones is considered the Master Zone and all
74  * Zone-related stats from the Keg are done in the Master Zone.  For an
75  * example of a Multi-Zone setup, refer to the Mbuf allocation code.
76  */
77 
78 /*
79  *	This is the representation for normal (Non OFFPAGE slab)
80  *
81  *	i == item
82  *	s == slab pointer
83  *
84  *	<----------------  Page (UMA_SLAB_SIZE) ------------------>
85  *	___________________________________________________________
86  *     | _  _  _  _  _  _  _  _  _  _  _  _  _  _  _   ___________ |
87  *     ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i| |slab header||
88  *     ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_| |___________||
89  *     |___________________________________________________________|
90  *
91  *
92  *	This is an OFFPAGE slab. These can be larger than UMA_SLAB_SIZE.
93  *
94  *	___________________________________________________________
95  *     | _  _  _  _  _  _  _  _  _  _  _  _  _  _  _  _  _  _  _   |
96  *     ||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i||i|  |
97  *     ||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_||_|  |
98  *     |___________________________________________________________|
99  *       ___________    ^
100  *	|slab header|   |
101  *	|___________|---*
102  *
103  */
104 
105 #ifndef VM_UMA_INT_H
106 #define VM_UMA_INT_H
107 
108 #define UMA_SLAB_SIZE	PAGE_SIZE	/* How big are our slabs? */
109 #define UMA_SLAB_MASK	(PAGE_SIZE - 1)	/* Mask to get back to the page */
110 #define UMA_SLAB_SHIFT	PAGE_SHIFT	/* Number of bits PAGE_MASK */
111 
112 #define UMA_BOOT_PAGES		64	/* Pages allocated for startup */
113 #define UMA_BOOT_PAGES_ZONES	32	/* Multiplier for pages to reserve */
114 					/* if uma_zone > PAGE_SIZE */
115 
116 /* Max waste percentage before going to off page slab management */
117 #define UMA_MAX_WASTE	10
118 
119 /*
120  * I doubt there will be many cases where this is exceeded. This is the initial
121  * size of the hash table for uma_slabs that are managed off page. This hash
122  * does expand by powers of two.  Currently it doesn't get smaller.
123  */
124 #define UMA_HASH_SIZE_INIT	32
125 
126 /*
127  * I should investigate other hashing algorithms.  This should yield a low
128  * number of collisions if the pages are relatively contiguous.
129  */
130 
131 #define UMA_HASH(h, s) ((((uintptr_t)s) >> UMA_SLAB_SHIFT) & (h)->uh_hashmask)
132 
133 #define UMA_HASH_INSERT(h, s, mem)					\
134 		SLIST_INSERT_HEAD(&(h)->uh_slab_hash[UMA_HASH((h),	\
135 		    (mem))], (s), us_hlink)
136 #define UMA_HASH_REMOVE(h, s, mem)					\
137 		SLIST_REMOVE(&(h)->uh_slab_hash[UMA_HASH((h),		\
138 		    (mem))], (s), uma_slab, us_hlink)
139 
140 /* Hash table for freed address -> slab translation */
141 
142 SLIST_HEAD(slabhead, uma_slab);
143 
144 struct uma_hash {
145 	struct slabhead	*uh_slab_hash;	/* Hash table for slabs */
146 	int		uh_hashsize;	/* Current size of the hash table */
147 	int		uh_hashmask;	/* Mask used during hashing */
148 };
149 
150 /*
151  * align field or structure to cache line
152  */
153 #if defined(__amd64__)
154 #define UMA_ALIGN	__aligned(CACHE_LINE_SIZE)
155 #else
156 #define UMA_ALIGN
157 #endif
158 
159 /*
160  * Structures for per cpu queues.
161  */
162 
163 struct uma_bucket {
164 	LIST_ENTRY(uma_bucket)	ub_link;	/* Link into the zone */
165 	int16_t	ub_cnt;				/* Count of free items. */
166 	int16_t	ub_entries;			/* Max items. */
167 	void	*ub_bucket[];			/* actual allocation storage */
168 };
169 
170 typedef struct uma_bucket * uma_bucket_t;
171 
172 struct uma_cache {
173 	uma_bucket_t	uc_freebucket;	/* Bucket we're freeing to */
174 	uma_bucket_t	uc_allocbucket;	/* Bucket to allocate from */
175 	uint64_t	uc_allocs;	/* Count of allocations */
176 	uint64_t	uc_frees;	/* Count of frees */
177 } UMA_ALIGN;
178 
179 typedef struct uma_cache * uma_cache_t;
180 
181 /*
182  * Keg management structure
183  *
184  * TODO: Optimize for cache line size
185  *
186  */
187 struct uma_keg {
188 	struct mtx_padalign	uk_lock;	/* Lock for the keg */
189 	struct uma_hash	uk_hash;
190 
191 	LIST_HEAD(,uma_zone)	uk_zones;	/* Keg's zones */
192 	LIST_HEAD(,uma_slab)	uk_part_slab;	/* partially allocated slabs */
193 	LIST_HEAD(,uma_slab)	uk_free_slab;	/* empty slab list */
194 	LIST_HEAD(,uma_slab)	uk_full_slab;	/* full slabs */
195 
196 	uint32_t	uk_align;	/* Alignment mask */
197 	uint32_t	uk_pages;	/* Total page count */
198 	uint32_t	uk_free;	/* Count of items free in slabs */
199 	uint32_t	uk_reserve;	/* Number of reserved items. */
200 	uint32_t	uk_size;	/* Requested size of each item */
201 	uint32_t	uk_rsize;	/* Real size of each item */
202 	uint32_t	uk_maxpages;	/* Maximum number of pages to alloc */
203 
204 	uma_init	uk_init;	/* Keg's init routine */
205 	uma_fini	uk_fini;	/* Keg's fini routine */
206 	uma_alloc	uk_allocf;	/* Allocation function */
207 	uma_free	uk_freef;	/* Free routine */
208 
209 	u_long		uk_offset;	/* Next free offset from base KVA */
210 	vm_offset_t	uk_kva;		/* Zone base KVA */
211 	uma_zone_t	uk_slabzone;	/* Slab zone backing us, if OFFPAGE */
212 
213 	uint16_t	uk_pgoff;	/* Offset to uma_slab struct */
214 	uint16_t	uk_ppera;	/* pages per allocation from backend */
215 	uint16_t	uk_ipers;	/* Items per slab */
216 	uint32_t	uk_flags;	/* Internal flags */
217 
218 	/* Least used fields go to the last cache line. */
219 	const char	*uk_name;		/* Name of creating zone. */
220 	LIST_ENTRY(uma_keg)	uk_link;	/* List of all kegs */
221 };
222 typedef struct uma_keg	* uma_keg_t;
223 
224 /*
225  * Free bits per-slab.
226  */
227 #define	SLAB_SETSIZE	(PAGE_SIZE / UMA_SMALLEST_UNIT)
228 BITSET_DEFINE(slabbits, SLAB_SETSIZE);
229 
230 /*
231  * The slab structure manages a single contiguous allocation from backing
232  * store and subdivides it into individually allocatable items.
233  */
234 struct uma_slab {
235 	uma_keg_t	us_keg;			/* Keg we live in */
236 	union {
237 		LIST_ENTRY(uma_slab)	_us_link;	/* slabs in zone */
238 		unsigned long	_us_size;	/* Size of allocation */
239 	} us_type;
240 	SLIST_ENTRY(uma_slab)	us_hlink;	/* Link for hash table */
241 	uint8_t		*us_data;		/* First item */
242 	struct slabbits	us_free;		/* Free bitmask. */
243 #ifdef INVARIANTS
244 	struct slabbits	us_debugfree;		/* Debug bitmask. */
245 #endif
246 	uint16_t	us_freecount;		/* How many are free? */
247 	uint8_t		us_flags;		/* Page flags see uma.h */
248 	uint8_t		us_pad;			/* Pad to 32bits, unused. */
249 };
250 
251 #define	us_link	us_type._us_link
252 #define	us_size	us_type._us_size
253 
254 typedef struct uma_slab * uma_slab_t;
255 typedef uma_slab_t (*uma_slaballoc)(uma_zone_t, uma_keg_t, int);
256 
257 struct uma_klink {
258 	LIST_ENTRY(uma_klink)	kl_link;
259 	uma_keg_t		kl_keg;
260 };
261 typedef struct uma_klink *uma_klink_t;
262 
263 /*
264  * Zone management structure
265  *
266  * TODO: Optimize for cache line size
267  *
268  */
269 struct uma_zone {
270 	struct mtx_padalign	uz_lock;	/* Lock for the zone */
271 	struct mtx_padalign	*uz_lockptr;
272 	const char		*uz_name;	/* Text name of the zone */
273 
274 	LIST_ENTRY(uma_zone)	uz_link;	/* List of all zones in keg */
275 	LIST_HEAD(,uma_bucket)	uz_buckets;	/* full buckets */
276 
277 	LIST_HEAD(,uma_klink)	uz_kegs;	/* List of kegs. */
278 	struct uma_klink	uz_klink;	/* klink for first keg. */
279 
280 	uma_slaballoc	uz_slab;	/* Allocate a slab from the backend. */
281 	uma_ctor	uz_ctor;	/* Constructor for each allocation */
282 	uma_dtor	uz_dtor;	/* Destructor */
283 	uma_init	uz_init;	/* Initializer for each item */
284 	uma_fini	uz_fini;	/* Finalizer for each item. */
285 	uma_import	uz_import;	/* Import new memory to cache. */
286 	uma_release	uz_release;	/* Release memory from cache. */
287 	void		*uz_arg;	/* Import/release argument. */
288 
289 	uint32_t	uz_flags;	/* Flags inherited from kegs */
290 	uint32_t	uz_size;	/* Size inherited from kegs */
291 
292 	volatile u_long	uz_allocs UMA_ALIGN; /* Total number of allocations */
293 	volatile u_long	uz_fails;	/* Total number of alloc failures */
294 	volatile u_long	uz_frees;	/* Total number of frees */
295 	uint64_t	uz_sleeps;	/* Total number of alloc sleeps */
296 	uint16_t	uz_count;	/* Amount of items in full bucket */
297 	uint16_t	uz_count_min;	/* Minimal amount of items there */
298 
299 	/* The next two fields are used to print a rate-limited warnings. */
300 	const char	*uz_warning;	/* Warning to print on failure */
301 	struct timeval	uz_ratecheck;	/* Warnings rate-limiting */
302 
303 	struct task	uz_maxaction;	/* Task to run when at limit */
304 
305 	/*
306 	 * This HAS to be the last item because we adjust the zone size
307 	 * based on NCPU and then allocate the space for the zones.
308 	 */
309 	struct uma_cache	uz_cpu[1]; /* Per cpu caches */
310 };
311 
312 /*
313  * These flags must not overlap with the UMA_ZONE flags specified in uma.h.
314  */
315 #define	UMA_ZFLAG_MULTI		0x04000000	/* Multiple kegs in the zone. */
316 #define	UMA_ZFLAG_DRAINING	0x08000000	/* Running zone_drain. */
317 #define	UMA_ZFLAG_BUCKET	0x10000000	/* Bucket zone. */
318 #define UMA_ZFLAG_INTERNAL	0x20000000	/* No offpage no PCPU. */
319 #define UMA_ZFLAG_FULL		0x40000000	/* Reached uz_maxpages */
320 #define UMA_ZFLAG_CACHEONLY	0x80000000	/* Don't ask VM for buckets. */
321 
322 #define	UMA_ZFLAG_INHERIT						\
323     (UMA_ZFLAG_INTERNAL | UMA_ZFLAG_CACHEONLY | UMA_ZFLAG_BUCKET)
324 
325 static inline uma_keg_t
326 zone_first_keg(uma_zone_t zone)
327 {
328 	uma_klink_t klink;
329 
330 	klink = LIST_FIRST(&zone->uz_kegs);
331 	return (klink != NULL) ? klink->kl_keg : NULL;
332 }
333 
334 #undef UMA_ALIGN
335 
336 #ifdef _KERNEL
337 /* Internal prototypes */
338 static __inline uma_slab_t hash_sfind(struct uma_hash *hash, uint8_t *data);
339 void *uma_large_malloc(vm_size_t size, int wait);
340 void uma_large_free(uma_slab_t slab);
341 
342 /* Lock Macros */
343 
344 #define	KEG_LOCK_INIT(k, lc)					\
345 	do {							\
346 		if ((lc))					\
347 			mtx_init(&(k)->uk_lock, (k)->uk_name,	\
348 			    (k)->uk_name, MTX_DEF | MTX_DUPOK);	\
349 		else						\
350 			mtx_init(&(k)->uk_lock, (k)->uk_name,	\
351 			    "UMA zone", MTX_DEF | MTX_DUPOK);	\
352 	} while (0)
353 
354 #define	KEG_LOCK_FINI(k)	mtx_destroy(&(k)->uk_lock)
355 #define	KEG_LOCK(k)	mtx_lock(&(k)->uk_lock)
356 #define	KEG_UNLOCK(k)	mtx_unlock(&(k)->uk_lock)
357 
358 #define	ZONE_LOCK_INIT(z, lc)					\
359 	do {							\
360 		if ((lc))					\
361 			mtx_init(&(z)->uz_lock, (z)->uz_name,	\
362 			    (z)->uz_name, MTX_DEF | MTX_DUPOK);	\
363 		else						\
364 			mtx_init(&(z)->uz_lock, (z)->uz_name,	\
365 			    "UMA zone", MTX_DEF | MTX_DUPOK);	\
366 	} while (0)
367 
368 #define	ZONE_LOCK(z)	mtx_lock((z)->uz_lockptr)
369 #define	ZONE_TRYLOCK(z)	mtx_trylock((z)->uz_lockptr)
370 #define	ZONE_UNLOCK(z)	mtx_unlock((z)->uz_lockptr)
371 #define	ZONE_LOCK_FINI(z)	mtx_destroy(&(z)->uz_lock)
372 
373 /*
374  * Find a slab within a hash table.  This is used for OFFPAGE zones to lookup
375  * the slab structure.
376  *
377  * Arguments:
378  *	hash  The hash table to search.
379  *	data  The base page of the item.
380  *
381  * Returns:
382  *	A pointer to a slab if successful, else NULL.
383  */
384 static __inline uma_slab_t
385 hash_sfind(struct uma_hash *hash, uint8_t *data)
386 {
387         uma_slab_t slab;
388         int hval;
389 
390         hval = UMA_HASH(hash, data);
391 
392         SLIST_FOREACH(slab, &hash->uh_slab_hash[hval], us_hlink) {
393                 if ((uint8_t *)slab->us_data == data)
394                         return (slab);
395         }
396         return (NULL);
397 }
398 
399 static __inline uma_slab_t
400 vtoslab(vm_offset_t va)
401 {
402 	vm_page_t p;
403 
404 	p = PHYS_TO_VM_PAGE(pmap_kextract(va));
405 	return ((uma_slab_t)p->plinks.s.pv);
406 }
407 
408 static __inline void
409 vsetslab(vm_offset_t va, uma_slab_t slab)
410 {
411 	vm_page_t p;
412 
413 	p = PHYS_TO_VM_PAGE(pmap_kextract(va));
414 	p->plinks.s.pv = slab;
415 }
416 
417 /*
418  * The following two functions may be defined by architecture specific code
419  * if they can provide more efficient allocation functions.  This is useful
420  * for using direct mapped addresses.
421  */
422 void *uma_small_alloc(uma_zone_t zone, vm_size_t bytes, uint8_t *pflag,
423     int wait);
424 void uma_small_free(void *mem, vm_size_t size, uint8_t flags);
425 #endif /* _KERNEL */
426 
427 #endif /* VM_UMA_INT_H */
428