Annotation of src/sys/kern/subr_pool.c, Revision 1.112.2.2
1.112.2.2! yamt 1: /* $NetBSD: subr_pool.c,v 1.112.2.1 2006/04/01 12:07:40 yamt Exp $ */
1.1 pk 2:
3: /*-
1.43 thorpej 4: * Copyright (c) 1997, 1999, 2000 The NetBSD Foundation, Inc.
1.1 pk 5: * All rights reserved.
6: *
7: * This code is derived from software contributed to The NetBSD Foundation
1.20 thorpej 8: * by Paul Kranenburg; by Jason R. Thorpe of the Numerical Aerospace
9: * Simulation Facility, NASA Ames Research Center.
1.1 pk 10: *
11: * Redistribution and use in source and binary forms, with or without
12: * modification, are permitted provided that the following conditions
13: * are met:
14: * 1. Redistributions of source code must retain the above copyright
15: * notice, this list of conditions and the following disclaimer.
16: * 2. Redistributions in binary form must reproduce the above copyright
17: * notice, this list of conditions and the following disclaimer in the
18: * documentation and/or other materials provided with the distribution.
19: * 3. All advertising materials mentioning features or use of this software
20: * must display the following acknowledgement:
1.13 christos 21: * This product includes software developed by the NetBSD
22: * Foundation, Inc. and its contributors.
1.1 pk 23: * 4. Neither the name of The NetBSD Foundation nor the names of its
24: * contributors may be used to endorse or promote products derived
25: * from this software without specific prior written permission.
26: *
27: * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28: * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29: * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30: * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31: * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32: * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33: * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34: * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35: * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36: * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37: * POSSIBILITY OF SUCH DAMAGE.
38: */
1.64 lukem 39:
40: #include <sys/cdefs.h>
1.112.2.2! yamt 41: __KERNEL_RCSID(0, "$NetBSD: subr_pool.c,v 1.112.2.1 2006/04/01 12:07:40 yamt Exp $");
1.24 scottr 42:
1.25 thorpej 43: #include "opt_pool.h"
1.24 scottr 44: #include "opt_poollog.h"
1.28 thorpej 45: #include "opt_lockdebug.h"
1.1 pk 46:
47: #include <sys/param.h>
48: #include <sys/systm.h>
49: #include <sys/proc.h>
50: #include <sys/errno.h>
51: #include <sys/kernel.h>
52: #include <sys/malloc.h>
53: #include <sys/lock.h>
54: #include <sys/pool.h>
1.20 thorpej 55: #include <sys/syslog.h>
1.3 pk 56:
57: #include <uvm/uvm.h>
58:
1.1 pk 59: /*
60: * Pool resource management utility.
1.3 pk 61: *
1.88 chs 62: * Memory is allocated in pages which are split into pieces according to
63: * the pool item size. Each page is kept on one of three lists in the
64: * pool structure: `pr_emptypages', `pr_fullpages' and `pr_partpages',
65: * for empty, full and partially-full pages respectively. The individual
66: * pool items are on a linked list headed by `ph_itemlist' in each page
67: * header. The memory for building the page list is either taken from
68: * the allocated pages themselves (for small pool items) or taken from
69: * an internal pool of page headers (`phpool').
1.1 pk 70: */
71:
1.3 pk 72: /* List of all pools */
1.102 chs 73: LIST_HEAD(,pool) pool_head = LIST_HEAD_INITIALIZER(pool_head);
1.3 pk 74:
75: /* Private pool for page header structures */
1.97 yamt 76: #define PHPOOL_MAX 8
77: static struct pool phpool[PHPOOL_MAX];
78: #define PHPOOL_FREELIST_NELEM(idx) (((idx) == 0) ? 0 : (1 << (idx)))
1.3 pk 79:
1.62 bjh21 80: #ifdef POOL_SUBPAGE
81: /* Pool of subpages for use by normal pools. */
82: static struct pool psppool;
83: #endif
84:
1.98 yamt 85: static void *pool_page_alloc_meta(struct pool *, int);
86: static void pool_page_free_meta(struct pool *, void *);
87:
88: /* allocator for pool metadata */
89: static struct pool_allocator pool_allocator_meta = {
90: pool_page_alloc_meta, pool_page_free_meta
91: };
92:
1.3 pk 93: /* # of seconds to retain page after last use */
94: int pool_inactive_time = 10;
95:
96: /* Next candidate for drainage (see pool_drain()) */
1.23 thorpej 97: static struct pool *drainpp;
98:
99: /* This spin lock protects both pool_head and drainpp. */
100: struct simplelock pool_head_slock = SIMPLELOCK_INITIALIZER;
1.3 pk 101:
1.99 yamt 102: typedef uint8_t pool_item_freelist_t;
103:
1.3 pk 104: struct pool_item_header {
105: /* Page headers */
1.88 chs 106: LIST_ENTRY(pool_item_header)
1.3 pk 107: ph_pagelist; /* pool page list */
1.88 chs 108: SPLAY_ENTRY(pool_item_header)
109: ph_node; /* Off-page page headers */
1.3 pk 110: caddr_t ph_page; /* this page's address */
111: struct timeval ph_time; /* last referenced */
1.97 yamt 112: union {
113: /* !PR_NOTOUCH */
114: struct {
1.102 chs 115: LIST_HEAD(, pool_item)
1.97 yamt 116: phu_itemlist; /* chunk list for this page */
117: } phu_normal;
118: /* PR_NOTOUCH */
119: struct {
120: uint16_t
121: phu_off; /* start offset in page */
1.99 yamt 122: pool_item_freelist_t
1.97 yamt 123: phu_firstfree; /* first free item */
1.99 yamt 124: /*
125: * XXX it might be better to use
126: * a simple bitmap and ffs(3)
127: */
1.97 yamt 128: } phu_notouch;
129: } ph_u;
130: uint16_t ph_nmissing; /* # of chunks in use */
1.3 pk 131: };
1.97 yamt 132: #define ph_itemlist ph_u.phu_normal.phu_itemlist
133: #define ph_off ph_u.phu_notouch.phu_off
134: #define ph_firstfree ph_u.phu_notouch.phu_firstfree
1.3 pk 135:
1.1 pk 136: struct pool_item {
1.3 pk 137: #ifdef DIAGNOSTIC
1.82 thorpej 138: u_int pi_magic;
1.33 chs 139: #endif
1.82 thorpej 140: #define PI_MAGIC 0xdeadbeefU
1.3 pk 141: /* Other entries use only this list entry */
1.102 chs 142: LIST_ENTRY(pool_item) pi_list;
1.3 pk 143: };
144:
1.53 thorpej 145: #define POOL_NEEDS_CATCHUP(pp) \
146: ((pp)->pr_nitems < (pp)->pr_minitems)
147:
1.43 thorpej 148: /*
149: * Pool cache management.
150: *
151: * Pool caches provide a way for constructed objects to be cached by the
152: * pool subsystem. This can lead to performance improvements by avoiding
153: * needless object construction/destruction; it is deferred until absolutely
154: * necessary.
155: *
156: * Caches are grouped into cache groups. Each cache group references
157: * up to 16 constructed objects. When a cache allocates an object
158: * from the pool, it calls the object's constructor and places it into
159: * a cache group. When a cache group frees an object back to the pool,
160: * it first calls the object's destructor. This allows the object to
161: * persist in constructed form while freed to the cache.
162: *
163: * Multiple caches may exist for each pool. This allows a single
164: * object type to have multiple constructed forms. The pool references
165: * each cache, so that when a pool is drained by the pagedaemon, it can
166: * drain each individual cache as well. Each time a cache is drained,
167: * the most idle cache group is freed to the pool in its entirety.
168: *
169: * Pool caches are layed on top of pools. By layering them, we can avoid
170: * the complexity of cache management for pools which would not benefit
171: * from it.
172: */
173:
174: /* The cache group pool. */
175: static struct pool pcgpool;
1.3 pk 176:
1.102 chs 177: static void pool_cache_reclaim(struct pool_cache *, struct pool_pagelist *,
178: struct pool_cache_grouplist *);
179: static void pcg_grouplist_free(struct pool_cache_grouplist *);
1.3 pk 180:
1.42 thorpej 181: static int pool_catchup(struct pool *);
1.55 thorpej 182: static void pool_prime_page(struct pool *, caddr_t,
183: struct pool_item_header *);
1.88 chs 184: static void pool_update_curpage(struct pool *);
1.66 thorpej 185:
1.112.2.1 yamt 186: static int pool_grow(struct pool *, int);
1.66 thorpej 187: void *pool_allocator_alloc(struct pool *, int);
188: void pool_allocator_free(struct pool *, void *);
1.3 pk 189:
1.97 yamt 190: static void pool_print_pagelist(struct pool *, struct pool_pagelist *,
1.88 chs 191: void (*)(const char *, ...));
1.42 thorpej 192: static void pool_print1(struct pool *, const char *,
193: void (*)(const char *, ...));
1.3 pk 194:
1.88 chs 195: static int pool_chk_page(struct pool *, const char *,
196: struct pool_item_header *);
197:
1.3 pk 198: /*
1.52 thorpej 199: * Pool log entry. An array of these is allocated in pool_init().
1.3 pk 200: */
201: struct pool_log {
202: const char *pl_file;
203: long pl_line;
204: int pl_action;
1.25 thorpej 205: #define PRLOG_GET 1
206: #define PRLOG_PUT 2
1.3 pk 207: void *pl_addr;
1.1 pk 208: };
209:
1.86 matt 210: #ifdef POOL_DIAGNOSTIC
1.3 pk 211: /* Number of entries in pool log buffers */
1.17 thorpej 212: #ifndef POOL_LOGSIZE
213: #define POOL_LOGSIZE 10
214: #endif
215:
216: int pool_logsize = POOL_LOGSIZE;
1.1 pk 217:
1.110 perry 218: static inline void
1.42 thorpej 219: pr_log(struct pool *pp, void *v, int action, const char *file, long line)
1.3 pk 220: {
221: int n = pp->pr_curlogentry;
222: struct pool_log *pl;
223:
1.20 thorpej 224: if ((pp->pr_roflags & PR_LOGGING) == 0)
1.3 pk 225: return;
226:
227: /*
228: * Fill in the current entry. Wrap around and overwrite
229: * the oldest entry if necessary.
230: */
231: pl = &pp->pr_log[n];
232: pl->pl_file = file;
233: pl->pl_line = line;
234: pl->pl_action = action;
235: pl->pl_addr = v;
236: if (++n >= pp->pr_logsize)
237: n = 0;
238: pp->pr_curlogentry = n;
239: }
240:
241: static void
1.42 thorpej 242: pr_printlog(struct pool *pp, struct pool_item *pi,
243: void (*pr)(const char *, ...))
1.3 pk 244: {
245: int i = pp->pr_logsize;
246: int n = pp->pr_curlogentry;
247:
1.20 thorpej 248: if ((pp->pr_roflags & PR_LOGGING) == 0)
1.3 pk 249: return;
250:
251: /*
252: * Print all entries in this pool's log.
253: */
254: while (i-- > 0) {
255: struct pool_log *pl = &pp->pr_log[n];
256: if (pl->pl_action != 0) {
1.25 thorpej 257: if (pi == NULL || pi == pl->pl_addr) {
258: (*pr)("\tlog entry %d:\n", i);
259: (*pr)("\t\taction = %s, addr = %p\n",
260: pl->pl_action == PRLOG_GET ? "get" : "put",
261: pl->pl_addr);
262: (*pr)("\t\tfile: %s at line %lu\n",
263: pl->pl_file, pl->pl_line);
264: }
1.3 pk 265: }
266: if (++n >= pp->pr_logsize)
267: n = 0;
268: }
269: }
1.25 thorpej 270:
1.110 perry 271: static inline void
1.42 thorpej 272: pr_enter(struct pool *pp, const char *file, long line)
1.25 thorpej 273: {
274:
1.34 thorpej 275: if (__predict_false(pp->pr_entered_file != NULL)) {
1.25 thorpej 276: printf("pool %s: reentrancy at file %s line %ld\n",
277: pp->pr_wchan, file, line);
278: printf(" previous entry at file %s line %ld\n",
279: pp->pr_entered_file, pp->pr_entered_line);
280: panic("pr_enter");
281: }
282:
283: pp->pr_entered_file = file;
284: pp->pr_entered_line = line;
285: }
286:
1.110 perry 287: static inline void
1.42 thorpej 288: pr_leave(struct pool *pp)
1.25 thorpej 289: {
290:
1.34 thorpej 291: if (__predict_false(pp->pr_entered_file == NULL)) {
1.25 thorpej 292: printf("pool %s not entered?\n", pp->pr_wchan);
293: panic("pr_leave");
294: }
295:
296: pp->pr_entered_file = NULL;
297: pp->pr_entered_line = 0;
298: }
299:
1.110 perry 300: static inline void
1.42 thorpej 301: pr_enter_check(struct pool *pp, void (*pr)(const char *, ...))
1.25 thorpej 302: {
303:
304: if (pp->pr_entered_file != NULL)
305: (*pr)("\n\tcurrently entered from file %s line %ld\n",
306: pp->pr_entered_file, pp->pr_entered_line);
307: }
1.3 pk 308: #else
1.25 thorpej 309: #define pr_log(pp, v, action, file, line)
310: #define pr_printlog(pp, pi, pr)
311: #define pr_enter(pp, file, line)
312: #define pr_leave(pp)
313: #define pr_enter_check(pp, pr)
1.59 thorpej 314: #endif /* POOL_DIAGNOSTIC */
1.3 pk 315:
1.110 perry 316: static inline int
1.97 yamt 317: pr_item_notouch_index(const struct pool *pp, const struct pool_item_header *ph,
318: const void *v)
319: {
320: const char *cp = v;
321: int idx;
322:
323: KASSERT(pp->pr_roflags & PR_NOTOUCH);
324: idx = (cp - ph->ph_page - ph->ph_off) / pp->pr_size;
325: KASSERT(idx < pp->pr_itemsperpage);
326: return idx;
327: }
328:
1.99 yamt 329: #define PR_FREELIST_ALIGN(p) \
330: roundup((uintptr_t)(p), sizeof(pool_item_freelist_t))
331: #define PR_FREELIST(ph) ((pool_item_freelist_t *)PR_FREELIST_ALIGN((ph) + 1))
332: #define PR_INDEX_USED ((pool_item_freelist_t)-1)
333: #define PR_INDEX_EOL ((pool_item_freelist_t)-2)
1.97 yamt 334:
1.110 perry 335: static inline void
1.97 yamt 336: pr_item_notouch_put(const struct pool *pp, struct pool_item_header *ph,
337: void *obj)
338: {
339: int idx = pr_item_notouch_index(pp, ph, obj);
1.99 yamt 340: pool_item_freelist_t *freelist = PR_FREELIST(ph);
1.97 yamt 341:
342: KASSERT(freelist[idx] == PR_INDEX_USED);
343: freelist[idx] = ph->ph_firstfree;
344: ph->ph_firstfree = idx;
345: }
346:
1.110 perry 347: static inline void *
1.97 yamt 348: pr_item_notouch_get(const struct pool *pp, struct pool_item_header *ph)
349: {
350: int idx = ph->ph_firstfree;
1.99 yamt 351: pool_item_freelist_t *freelist = PR_FREELIST(ph);
1.97 yamt 352:
353: KASSERT(freelist[idx] != PR_INDEX_USED);
354: ph->ph_firstfree = freelist[idx];
355: freelist[idx] = PR_INDEX_USED;
356:
357: return ph->ph_page + ph->ph_off + idx * pp->pr_size;
358: }
359:
1.110 perry 360: static inline int
1.88 chs 361: phtree_compare(struct pool_item_header *a, struct pool_item_header *b)
362: {
363: if (a->ph_page < b->ph_page)
364: return (-1);
365: else if (a->ph_page > b->ph_page)
366: return (1);
367: else
368: return (0);
369: }
370:
371: SPLAY_PROTOTYPE(phtree, pool_item_header, ph_node, phtree_compare);
372: SPLAY_GENERATE(phtree, pool_item_header, ph_node, phtree_compare);
373:
1.3 pk 374: /*
375: * Return the pool page header based on page address.
376: */
1.110 perry 377: static inline struct pool_item_header *
1.42 thorpej 378: pr_find_pagehead(struct pool *pp, caddr_t page)
1.3 pk 379: {
1.88 chs 380: struct pool_item_header *ph, tmp;
1.3 pk 381:
1.20 thorpej 382: if ((pp->pr_roflags & PR_PHINPAGE) != 0)
1.3 pk 383: return ((struct pool_item_header *)(page + pp->pr_phoffset));
384:
1.88 chs 385: tmp.ph_page = page;
386: ph = SPLAY_FIND(phtree, &pp->pr_phtree, &tmp);
387: return ph;
1.3 pk 388: }
389:
1.101 thorpej 390: static void
391: pr_pagelist_free(struct pool *pp, struct pool_pagelist *pq)
392: {
393: struct pool_item_header *ph;
394: int s;
395:
396: while ((ph = LIST_FIRST(pq)) != NULL) {
397: LIST_REMOVE(ph, ph_pagelist);
398: pool_allocator_free(pp, ph->ph_page);
399: if ((pp->pr_roflags & PR_PHINPAGE) == 0) {
400: s = splvm();
401: pool_put(pp->pr_phpool, ph);
402: splx(s);
403: }
404: }
405: }
406:
1.3 pk 407: /*
408: * Remove a page from the pool.
409: */
1.110 perry 410: static inline void
1.61 chs 411: pr_rmpage(struct pool *pp, struct pool_item_header *ph,
412: struct pool_pagelist *pq)
1.3 pk 413: {
414:
1.101 thorpej 415: LOCK_ASSERT(simple_lock_held(&pp->pr_slock));
1.91 yamt 416:
1.3 pk 417: /*
1.7 thorpej 418: * If the page was idle, decrement the idle page count.
1.3 pk 419: */
1.6 thorpej 420: if (ph->ph_nmissing == 0) {
421: #ifdef DIAGNOSTIC
422: if (pp->pr_nidle == 0)
423: panic("pr_rmpage: nidle inconsistent");
1.20 thorpej 424: if (pp->pr_nitems < pp->pr_itemsperpage)
425: panic("pr_rmpage: nitems inconsistent");
1.6 thorpej 426: #endif
427: pp->pr_nidle--;
428: }
1.7 thorpej 429:
1.20 thorpej 430: pp->pr_nitems -= pp->pr_itemsperpage;
431:
1.7 thorpej 432: /*
1.101 thorpej 433: * Unlink the page from the pool and queue it for release.
1.7 thorpej 434: */
1.88 chs 435: LIST_REMOVE(ph, ph_pagelist);
1.91 yamt 436: if ((pp->pr_roflags & PR_PHINPAGE) == 0)
437: SPLAY_REMOVE(phtree, &pp->pr_phtree, ph);
1.101 thorpej 438: LIST_INSERT_HEAD(pq, ph, ph_pagelist);
439:
1.7 thorpej 440: pp->pr_npages--;
441: pp->pr_npagefree++;
1.6 thorpej 442:
1.88 chs 443: pool_update_curpage(pp);
1.3 pk 444: }
445:
446: /*
1.94 simonb 447: * Initialize all the pools listed in the "pools" link set.
448: */
449: void
450: link_pool_init(void)
451: {
452: __link_set_decl(pools, struct link_pool_init);
453: struct link_pool_init * const *pi;
454:
455: __link_set_foreach(pi, pools)
456: pool_init((*pi)->pp, (*pi)->size, (*pi)->align,
457: (*pi)->align_offset, (*pi)->flags, (*pi)->wchan,
458: (*pi)->palloc);
459: }
460:
461: /*
1.3 pk 462: * Initialize the given pool resource structure.
463: *
464: * We export this routine to allow other kernel parts to declare
465: * static pools that must be initialized before malloc() is available.
466: */
467: void
1.42 thorpej 468: pool_init(struct pool *pp, size_t size, u_int align, u_int ioff, int flags,
1.66 thorpej 469: const char *wchan, struct pool_allocator *palloc)
1.3 pk 470: {
1.88 chs 471: int off, slack;
1.92 enami 472: size_t trysize, phsize;
1.93 dbj 473: int s;
1.3 pk 474:
1.99 yamt 475: KASSERT((1UL << (CHAR_BIT * sizeof(pool_item_freelist_t))) - 2 >=
476: PHPOOL_FREELIST_NELEM(PHPOOL_MAX - 1));
477:
1.25 thorpej 478: #ifdef POOL_DIAGNOSTIC
479: /*
480: * Always log if POOL_DIAGNOSTIC is defined.
481: */
482: if (pool_logsize != 0)
483: flags |= PR_LOGGING;
484: #endif
485:
1.66 thorpej 486: if (palloc == NULL)
487: palloc = &pool_allocator_kmem;
1.112 bjh21 488: #ifdef POOL_SUBPAGE
489: if (size > palloc->pa_pagesz) {
490: if (palloc == &pool_allocator_kmem)
491: palloc = &pool_allocator_kmem_fullpage;
492: else if (palloc == &pool_allocator_nointr)
493: palloc = &pool_allocator_nointr_fullpage;
494: }
1.66 thorpej 495: #endif /* POOL_SUBPAGE */
496: if ((palloc->pa_flags & PA_INITIALIZED) == 0) {
1.112 bjh21 497: if (palloc->pa_pagesz == 0)
1.66 thorpej 498: palloc->pa_pagesz = PAGE_SIZE;
499:
500: TAILQ_INIT(&palloc->pa_list);
501:
502: simple_lock_init(&palloc->pa_slock);
503: palloc->pa_pagemask = ~(palloc->pa_pagesz - 1);
504: palloc->pa_pageshift = ffs(palloc->pa_pagesz) - 1;
505: palloc->pa_flags |= PA_INITIALIZED;
1.4 thorpej 506: }
1.3 pk 507:
508: if (align == 0)
509: align = ALIGN(1);
1.14 thorpej 510:
511: if (size < sizeof(struct pool_item))
512: size = sizeof(struct pool_item);
1.3 pk 513:
1.78 thorpej 514: size = roundup(size, align);
1.66 thorpej 515: #ifdef DIAGNOSTIC
516: if (size > palloc->pa_pagesz)
1.35 pk 517: panic("pool_init: pool item size (%lu) too large",
518: (u_long)size);
1.66 thorpej 519: #endif
1.35 pk 520:
1.3 pk 521: /*
522: * Initialize the pool structure.
523: */
1.88 chs 524: LIST_INIT(&pp->pr_emptypages);
525: LIST_INIT(&pp->pr_fullpages);
526: LIST_INIT(&pp->pr_partpages);
1.102 chs 527: LIST_INIT(&pp->pr_cachelist);
1.3 pk 528: pp->pr_curpage = NULL;
529: pp->pr_npages = 0;
530: pp->pr_minitems = 0;
531: pp->pr_minpages = 0;
532: pp->pr_maxpages = UINT_MAX;
1.20 thorpej 533: pp->pr_roflags = flags;
534: pp->pr_flags = 0;
1.35 pk 535: pp->pr_size = size;
1.3 pk 536: pp->pr_align = align;
537: pp->pr_wchan = wchan;
1.66 thorpej 538: pp->pr_alloc = palloc;
1.20 thorpej 539: pp->pr_nitems = 0;
540: pp->pr_nout = 0;
541: pp->pr_hardlimit = UINT_MAX;
542: pp->pr_hardlimit_warning = NULL;
1.31 thorpej 543: pp->pr_hardlimit_ratecap.tv_sec = 0;
544: pp->pr_hardlimit_ratecap.tv_usec = 0;
545: pp->pr_hardlimit_warning_last.tv_sec = 0;
546: pp->pr_hardlimit_warning_last.tv_usec = 0;
1.68 thorpej 547: pp->pr_drain_hook = NULL;
548: pp->pr_drain_hook_arg = NULL;
1.3 pk 549:
550: /*
551: * Decide whether to put the page header off page to avoid
1.92 enami 552: * wasting too large a part of the page or too big item.
553: * Off-page page headers go on a hash table, so we can match
554: * a returned item with its header based on the page address.
555: * We use 1/16 of the page size and about 8 times of the item
556: * size as the threshold (XXX: tune)
557: *
558: * However, we'll put the header into the page if we can put
559: * it without wasting any items.
560: *
561: * Silently enforce `0 <= ioff < align'.
1.3 pk 562: */
1.92 enami 563: pp->pr_itemoffset = ioff %= align;
564: /* See the comment below about reserved bytes. */
565: trysize = palloc->pa_pagesz - ((align - ioff) % align);
566: phsize = ALIGN(sizeof(struct pool_item_header));
1.97 yamt 567: if ((pp->pr_roflags & PR_NOTOUCH) == 0 &&
568: (pp->pr_size < MIN(palloc->pa_pagesz / 16, phsize << 3) ||
569: trysize / pp->pr_size == (trysize - phsize) / pp->pr_size)) {
1.3 pk 570: /* Use the end of the page for the page header */
1.20 thorpej 571: pp->pr_roflags |= PR_PHINPAGE;
1.92 enami 572: pp->pr_phoffset = off = palloc->pa_pagesz - phsize;
1.2 pk 573: } else {
1.3 pk 574: /* The page header will be taken from our page header pool */
575: pp->pr_phoffset = 0;
1.66 thorpej 576: off = palloc->pa_pagesz;
1.88 chs 577: SPLAY_INIT(&pp->pr_phtree);
1.2 pk 578: }
1.1 pk 579:
1.3 pk 580: /*
581: * Alignment is to take place at `ioff' within the item. This means
582: * we must reserve up to `align - 1' bytes on the page to allow
583: * appropriate positioning of each item.
584: */
585: pp->pr_itemsperpage = (off - ((align - ioff) % align)) / pp->pr_size;
1.43 thorpej 586: KASSERT(pp->pr_itemsperpage != 0);
1.97 yamt 587: if ((pp->pr_roflags & PR_NOTOUCH)) {
588: int idx;
589:
590: for (idx = 0; pp->pr_itemsperpage > PHPOOL_FREELIST_NELEM(idx);
591: idx++) {
592: /* nothing */
593: }
594: if (idx >= PHPOOL_MAX) {
595: /*
596: * if you see this panic, consider to tweak
597: * PHPOOL_MAX and PHPOOL_FREELIST_NELEM.
598: */
599: panic("%s: too large itemsperpage(%d) for PR_NOTOUCH",
600: pp->pr_wchan, pp->pr_itemsperpage);
601: }
602: pp->pr_phpool = &phpool[idx];
603: } else if ((pp->pr_roflags & PR_PHINPAGE) == 0) {
604: pp->pr_phpool = &phpool[0];
605: }
606: #if defined(DIAGNOSTIC)
607: else {
608: pp->pr_phpool = NULL;
609: }
610: #endif
1.3 pk 611:
612: /*
613: * Use the slack between the chunks and the page header
614: * for "cache coloring".
615: */
616: slack = off - pp->pr_itemsperpage * pp->pr_size;
617: pp->pr_maxcolor = (slack / align) * align;
618: pp->pr_curcolor = 0;
619:
620: pp->pr_nget = 0;
621: pp->pr_nfail = 0;
622: pp->pr_nput = 0;
623: pp->pr_npagealloc = 0;
624: pp->pr_npagefree = 0;
1.1 pk 625: pp->pr_hiwat = 0;
1.8 thorpej 626: pp->pr_nidle = 0;
1.3 pk 627:
1.59 thorpej 628: #ifdef POOL_DIAGNOSTIC
1.25 thorpej 629: if (flags & PR_LOGGING) {
630: if (kmem_map == NULL ||
631: (pp->pr_log = malloc(pool_logsize * sizeof(struct pool_log),
632: M_TEMP, M_NOWAIT)) == NULL)
1.20 thorpej 633: pp->pr_roflags &= ~PR_LOGGING;
1.3 pk 634: pp->pr_curlogentry = 0;
635: pp->pr_logsize = pool_logsize;
636: }
1.59 thorpej 637: #endif
1.25 thorpej 638:
639: pp->pr_entered_file = NULL;
640: pp->pr_entered_line = 0;
1.3 pk 641:
1.21 thorpej 642: simple_lock_init(&pp->pr_slock);
1.1 pk 643:
1.3 pk 644: /*
1.43 thorpej 645: * Initialize private page header pool and cache magazine pool if we
646: * haven't done so yet.
1.23 thorpej 647: * XXX LOCKING.
1.3 pk 648: */
1.97 yamt 649: if (phpool[0].pr_size == 0) {
650: int idx;
651: for (idx = 0; idx < PHPOOL_MAX; idx++) {
652: static char phpool_names[PHPOOL_MAX][6+1+6+1];
653: int nelem;
654: size_t sz;
655:
656: nelem = PHPOOL_FREELIST_NELEM(idx);
657: snprintf(phpool_names[idx], sizeof(phpool_names[idx]),
658: "phpool-%d", nelem);
659: sz = sizeof(struct pool_item_header);
660: if (nelem) {
661: sz = PR_FREELIST_ALIGN(sz)
1.99 yamt 662: + nelem * sizeof(pool_item_freelist_t);
1.97 yamt 663: }
664: pool_init(&phpool[idx], sz, 0, 0, 0,
1.98 yamt 665: phpool_names[idx], &pool_allocator_meta);
1.97 yamt 666: }
1.62 bjh21 667: #ifdef POOL_SUBPAGE
668: pool_init(&psppool, POOL_SUBPAGE, POOL_SUBPAGE, 0,
1.98 yamt 669: PR_RECURSIVE, "psppool", &pool_allocator_meta);
1.62 bjh21 670: #endif
1.43 thorpej 671: pool_init(&pcgpool, sizeof(struct pool_cache_group), 0, 0,
1.98 yamt 672: 0, "pcgpool", &pool_allocator_meta);
1.1 pk 673: }
674:
1.23 thorpej 675: /* Insert into the list of all pools. */
676: simple_lock(&pool_head_slock);
1.102 chs 677: LIST_INSERT_HEAD(&pool_head, pp, pr_poollist);
1.23 thorpej 678: simple_unlock(&pool_head_slock);
1.66 thorpej 679:
680: /* Insert this into the list of pools using this allocator. */
1.93 dbj 681: s = splvm();
1.66 thorpej 682: simple_lock(&palloc->pa_slock);
683: TAILQ_INSERT_TAIL(&palloc->pa_list, pp, pr_alloc_list);
684: simple_unlock(&palloc->pa_slock);
1.93 dbj 685: splx(s);
1.1 pk 686: }
687:
688: /*
689: * De-commision a pool resource.
690: */
691: void
1.42 thorpej 692: pool_destroy(struct pool *pp)
1.1 pk 693: {
1.101 thorpej 694: struct pool_pagelist pq;
1.3 pk 695: struct pool_item_header *ph;
1.93 dbj 696: int s;
1.43 thorpej 697:
1.101 thorpej 698: /* Remove from global pool list */
699: simple_lock(&pool_head_slock);
1.102 chs 700: LIST_REMOVE(pp, pr_poollist);
1.101 thorpej 701: if (drainpp == pp)
702: drainpp = NULL;
703: simple_unlock(&pool_head_slock);
704:
705: /* Remove this pool from its allocator's list of pools. */
1.93 dbj 706: s = splvm();
1.66 thorpej 707: simple_lock(&pp->pr_alloc->pa_slock);
708: TAILQ_REMOVE(&pp->pr_alloc->pa_list, pp, pr_alloc_list);
709: simple_unlock(&pp->pr_alloc->pa_slock);
1.93 dbj 710: splx(s);
1.66 thorpej 711:
1.101 thorpej 712: s = splvm();
713: simple_lock(&pp->pr_slock);
714:
1.102 chs 715: KASSERT(LIST_EMPTY(&pp->pr_cachelist));
1.3 pk 716:
717: #ifdef DIAGNOSTIC
1.20 thorpej 718: if (pp->pr_nout != 0) {
1.25 thorpej 719: pr_printlog(pp, NULL, printf);
1.80 provos 720: panic("pool_destroy: pool busy: still out: %u",
1.20 thorpej 721: pp->pr_nout);
1.3 pk 722: }
723: #endif
1.1 pk 724:
1.101 thorpej 725: KASSERT(LIST_EMPTY(&pp->pr_fullpages));
726: KASSERT(LIST_EMPTY(&pp->pr_partpages));
727:
1.3 pk 728: /* Remove all pages */
1.101 thorpej 729: LIST_INIT(&pq);
1.88 chs 730: while ((ph = LIST_FIRST(&pp->pr_emptypages)) != NULL)
1.101 thorpej 731: pr_rmpage(pp, ph, &pq);
732:
733: simple_unlock(&pp->pr_slock);
734: splx(s);
1.3 pk 735:
1.101 thorpej 736: pr_pagelist_free(pp, &pq);
1.3 pk 737:
1.59 thorpej 738: #ifdef POOL_DIAGNOSTIC
1.20 thorpej 739: if ((pp->pr_roflags & PR_LOGGING) != 0)
1.3 pk 740: free(pp->pr_log, M_TEMP);
1.59 thorpej 741: #endif
1.1 pk 742: }
743:
1.68 thorpej 744: void
745: pool_set_drain_hook(struct pool *pp, void (*fn)(void *, int), void *arg)
746: {
747:
748: /* XXX no locking -- must be used just after pool_init() */
749: #ifdef DIAGNOSTIC
750: if (pp->pr_drain_hook != NULL)
751: panic("pool_set_drain_hook(%s): already set", pp->pr_wchan);
752: #endif
753: pp->pr_drain_hook = fn;
754: pp->pr_drain_hook_arg = arg;
755: }
756:
1.88 chs 757: static struct pool_item_header *
1.55 thorpej 758: pool_alloc_item_header(struct pool *pp, caddr_t storage, int flags)
759: {
760: struct pool_item_header *ph;
761: int s;
762:
763: LOCK_ASSERT(simple_lock_held(&pp->pr_slock) == 0);
764:
765: if ((pp->pr_roflags & PR_PHINPAGE) != 0)
766: ph = (struct pool_item_header *) (storage + pp->pr_phoffset);
767: else {
1.85 pk 768: s = splvm();
1.97 yamt 769: ph = pool_get(pp->pr_phpool, flags);
1.55 thorpej 770: splx(s);
771: }
772:
773: return (ph);
774: }
1.1 pk 775:
776: /*
1.3 pk 777: * Grab an item from the pool; must be called at appropriate spl level
1.1 pk 778: */
1.3 pk 779: void *
1.59 thorpej 780: #ifdef POOL_DIAGNOSTIC
1.42 thorpej 781: _pool_get(struct pool *pp, int flags, const char *file, long line)
1.56 sommerfe 782: #else
783: pool_get(struct pool *pp, int flags)
784: #endif
1.1 pk 785: {
786: struct pool_item *pi;
1.3 pk 787: struct pool_item_header *ph;
1.55 thorpej 788: void *v;
1.1 pk 789:
1.2 pk 790: #ifdef DIAGNOSTIC
1.95 atatat 791: if (__predict_false(pp->pr_itemsperpage == 0))
792: panic("pool_get: pool %p: pr_itemsperpage is zero, "
793: "pool not initialized?", pp);
1.84 thorpej 794: if (__predict_false(curlwp == NULL && doing_shutdown == 0 &&
1.37 sommerfe 795: (flags & PR_WAITOK) != 0))
1.77 matt 796: panic("pool_get: %s: must have NOWAIT", pp->pr_wchan);
1.58 thorpej 797:
1.102 chs 798: #endif /* DIAGNOSTIC */
1.58 thorpej 799: #ifdef LOCKDEBUG
800: if (flags & PR_WAITOK)
801: simple_lock_only_held(NULL, "pool_get(PR_WAITOK)");
1.102 chs 802: SCHED_ASSERT_UNLOCKED();
1.56 sommerfe 803: #endif
1.1 pk 804:
1.21 thorpej 805: simple_lock(&pp->pr_slock);
1.25 thorpej 806: pr_enter(pp, file, line);
1.20 thorpej 807:
808: startover:
809: /*
810: * Check to see if we've reached the hard limit. If we have,
811: * and we can wait, then wait until an item has been returned to
812: * the pool.
813: */
814: #ifdef DIAGNOSTIC
1.34 thorpej 815: if (__predict_false(pp->pr_nout > pp->pr_hardlimit)) {
1.25 thorpej 816: pr_leave(pp);
1.21 thorpej 817: simple_unlock(&pp->pr_slock);
1.20 thorpej 818: panic("pool_get: %s: crossed hard limit", pp->pr_wchan);
819: }
820: #endif
1.34 thorpej 821: if (__predict_false(pp->pr_nout == pp->pr_hardlimit)) {
1.68 thorpej 822: if (pp->pr_drain_hook != NULL) {
823: /*
824: * Since the drain hook is going to free things
825: * back to the pool, unlock, call the hook, re-lock,
826: * and check the hardlimit condition again.
827: */
828: pr_leave(pp);
829: simple_unlock(&pp->pr_slock);
830: (*pp->pr_drain_hook)(pp->pr_drain_hook_arg, flags);
831: simple_lock(&pp->pr_slock);
832: pr_enter(pp, file, line);
833: if (pp->pr_nout < pp->pr_hardlimit)
834: goto startover;
835: }
836:
1.29 sommerfe 837: if ((flags & PR_WAITOK) && !(flags & PR_LIMITFAIL)) {
1.20 thorpej 838: /*
839: * XXX: A warning isn't logged in this case. Should
840: * it be?
841: */
842: pp->pr_flags |= PR_WANTED;
1.25 thorpej 843: pr_leave(pp);
1.40 sommerfe 844: ltsleep(pp, PSWP, pp->pr_wchan, 0, &pp->pr_slock);
1.25 thorpej 845: pr_enter(pp, file, line);
1.20 thorpej 846: goto startover;
847: }
1.31 thorpej 848:
849: /*
850: * Log a message that the hard limit has been hit.
851: */
852: if (pp->pr_hardlimit_warning != NULL &&
853: ratecheck(&pp->pr_hardlimit_warning_last,
854: &pp->pr_hardlimit_ratecap))
855: log(LOG_ERR, "%s\n", pp->pr_hardlimit_warning);
1.21 thorpej 856:
857: pp->pr_nfail++;
858:
1.25 thorpej 859: pr_leave(pp);
1.21 thorpej 860: simple_unlock(&pp->pr_slock);
1.20 thorpej 861: return (NULL);
862: }
863:
1.3 pk 864: /*
865: * The convention we use is that if `curpage' is not NULL, then
866: * it points at a non-empty bucket. In particular, `curpage'
867: * never points at a page header which has PR_PHINPAGE set and
868: * has no items in its bucket.
869: */
1.20 thorpej 870: if ((ph = pp->pr_curpage) == NULL) {
1.112.2.1 yamt 871: int error;
872:
1.20 thorpej 873: #ifdef DIAGNOSTIC
874: if (pp->pr_nitems != 0) {
1.21 thorpej 875: simple_unlock(&pp->pr_slock);
1.20 thorpej 876: printf("pool_get: %s: curpage NULL, nitems %u\n",
877: pp->pr_wchan, pp->pr_nitems);
1.80 provos 878: panic("pool_get: nitems inconsistent");
1.20 thorpej 879: }
880: #endif
881:
1.21 thorpej 882: /*
883: * Call the back-end page allocator for more memory.
884: * Release the pool lock, as the back-end page allocator
885: * may block.
886: */
1.25 thorpej 887: pr_leave(pp);
1.112.2.1 yamt 888: error = pool_grow(pp, flags);
889: pr_enter(pp, file, line);
890: if (error != 0) {
1.21 thorpej 891: /*
1.55 thorpej 892: * We were unable to allocate a page or item
893: * header, but we released the lock during
894: * allocation, so perhaps items were freed
895: * back to the pool. Check for this case.
1.21 thorpej 896: */
897: if (pp->pr_curpage != NULL)
898: goto startover;
1.15 pk 899:
1.3 pk 900: if ((flags & PR_WAITOK) == 0) {
901: pp->pr_nfail++;
1.25 thorpej 902: pr_leave(pp);
1.21 thorpej 903: simple_unlock(&pp->pr_slock);
1.1 pk 904: return (NULL);
1.3 pk 905: }
906:
1.15 pk 907: /*
908: * Wait for items to be returned to this pool.
1.21 thorpej 909: *
1.109 christos 910: * wake up once a second and try again,
911: * as the check in pool_cache_put_paddr() is racy.
1.15 pk 912: */
1.1 pk 913: pp->pr_flags |= PR_WANTED;
1.66 thorpej 914: /* PA_WANTED is already set on the allocator. */
1.25 thorpej 915: pr_leave(pp);
1.109 christos 916: ltsleep(pp, PSWP, pp->pr_wchan, hz, &pp->pr_slock);
1.25 thorpej 917: pr_enter(pp, file, line);
1.1 pk 918: }
1.3 pk 919:
1.20 thorpej 920: /* Start the allocation process over. */
921: goto startover;
1.3 pk 922: }
1.97 yamt 923: if (pp->pr_roflags & PR_NOTOUCH) {
924: #ifdef DIAGNOSTIC
925: if (__predict_false(ph->ph_nmissing == pp->pr_itemsperpage)) {
926: pr_leave(pp);
927: simple_unlock(&pp->pr_slock);
928: panic("pool_get: %s: page empty", pp->pr_wchan);
929: }
930: #endif
931: v = pr_item_notouch_get(pp, ph);
932: #ifdef POOL_DIAGNOSTIC
933: pr_log(pp, v, PRLOG_GET, file, line);
934: #endif
935: } else {
1.102 chs 936: v = pi = LIST_FIRST(&ph->ph_itemlist);
1.97 yamt 937: if (__predict_false(v == NULL)) {
938: pr_leave(pp);
939: simple_unlock(&pp->pr_slock);
940: panic("pool_get: %s: page empty", pp->pr_wchan);
941: }
1.20 thorpej 942: #ifdef DIAGNOSTIC
1.97 yamt 943: if (__predict_false(pp->pr_nitems == 0)) {
944: pr_leave(pp);
945: simple_unlock(&pp->pr_slock);
946: printf("pool_get: %s: items on itemlist, nitems %u\n",
947: pp->pr_wchan, pp->pr_nitems);
948: panic("pool_get: nitems inconsistent");
949: }
1.65 enami 950: #endif
1.56 sommerfe 951:
1.65 enami 952: #ifdef POOL_DIAGNOSTIC
1.97 yamt 953: pr_log(pp, v, PRLOG_GET, file, line);
1.65 enami 954: #endif
1.3 pk 955:
1.65 enami 956: #ifdef DIAGNOSTIC
1.97 yamt 957: if (__predict_false(pi->pi_magic != PI_MAGIC)) {
958: pr_printlog(pp, pi, printf);
959: panic("pool_get(%s): free list modified: "
960: "magic=%x; page %p; item addr %p\n",
961: pp->pr_wchan, pi->pi_magic, ph->ph_page, pi);
962: }
1.3 pk 963: #endif
964:
1.97 yamt 965: /*
966: * Remove from item list.
967: */
1.102 chs 968: LIST_REMOVE(pi, pi_list);
1.97 yamt 969: }
1.20 thorpej 970: pp->pr_nitems--;
971: pp->pr_nout++;
1.6 thorpej 972: if (ph->ph_nmissing == 0) {
973: #ifdef DIAGNOSTIC
1.34 thorpej 974: if (__predict_false(pp->pr_nidle == 0))
1.6 thorpej 975: panic("pool_get: nidle inconsistent");
976: #endif
977: pp->pr_nidle--;
1.88 chs 978:
979: /*
980: * This page was previously empty. Move it to the list of
981: * partially-full pages. This page is already curpage.
982: */
983: LIST_REMOVE(ph, ph_pagelist);
984: LIST_INSERT_HEAD(&pp->pr_partpages, ph, ph_pagelist);
1.6 thorpej 985: }
1.3 pk 986: ph->ph_nmissing++;
1.97 yamt 987: if (ph->ph_nmissing == pp->pr_itemsperpage) {
1.21 thorpej 988: #ifdef DIAGNOSTIC
1.97 yamt 989: if (__predict_false((pp->pr_roflags & PR_NOTOUCH) == 0 &&
1.102 chs 990: !LIST_EMPTY(&ph->ph_itemlist))) {
1.25 thorpej 991: pr_leave(pp);
1.21 thorpej 992: simple_unlock(&pp->pr_slock);
993: panic("pool_get: %s: nmissing inconsistent",
994: pp->pr_wchan);
995: }
996: #endif
1.3 pk 997: /*
1.88 chs 998: * This page is now full. Move it to the full list
999: * and select a new current page.
1.3 pk 1000: */
1.88 chs 1001: LIST_REMOVE(ph, ph_pagelist);
1002: LIST_INSERT_HEAD(&pp->pr_fullpages, ph, ph_pagelist);
1003: pool_update_curpage(pp);
1.1 pk 1004: }
1.3 pk 1005:
1006: pp->pr_nget++;
1.111 christos 1007: pr_leave(pp);
1.20 thorpej 1008:
1009: /*
1010: * If we have a low water mark and we are now below that low
1011: * water mark, add more items to the pool.
1012: */
1.53 thorpej 1013: if (POOL_NEEDS_CATCHUP(pp) && pool_catchup(pp) != 0) {
1.20 thorpej 1014: /*
1015: * XXX: Should we log a warning? Should we set up a timeout
1016: * to try again in a second or so? The latter could break
1017: * a caller's assumptions about interrupt protection, etc.
1018: */
1019: }
1020:
1.21 thorpej 1021: simple_unlock(&pp->pr_slock);
1.1 pk 1022: return (v);
1023: }
1024:
1025: /*
1.43 thorpej 1026: * Internal version of pool_put(). Pool is already locked/entered.
1.1 pk 1027: */
1.43 thorpej 1028: static void
1.101 thorpej 1029: pool_do_put(struct pool *pp, void *v, struct pool_pagelist *pq)
1.1 pk 1030: {
1031: struct pool_item *pi = v;
1.3 pk 1032: struct pool_item_header *ph;
1033: caddr_t page;
1.21 thorpej 1034: int s;
1.3 pk 1035:
1.61 chs 1036: LOCK_ASSERT(simple_lock_held(&pp->pr_slock));
1.102 chs 1037: SCHED_ASSERT_UNLOCKED();
1.61 chs 1038:
1.66 thorpej 1039: page = (caddr_t)((u_long)v & pp->pr_alloc->pa_pagemask);
1.1 pk 1040:
1.30 thorpej 1041: #ifdef DIAGNOSTIC
1.34 thorpej 1042: if (__predict_false(pp->pr_nout == 0)) {
1.30 thorpej 1043: printf("pool %s: putting with none out\n",
1044: pp->pr_wchan);
1045: panic("pool_put");
1046: }
1047: #endif
1.3 pk 1048:
1.34 thorpej 1049: if (__predict_false((ph = pr_find_pagehead(pp, page)) == NULL)) {
1.25 thorpej 1050: pr_printlog(pp, NULL, printf);
1.3 pk 1051: panic("pool_put: %s: page header missing", pp->pr_wchan);
1052: }
1.28 thorpej 1053:
1054: #ifdef LOCKDEBUG
1055: /*
1056: * Check if we're freeing a locked simple lock.
1057: */
1058: simple_lock_freecheck((caddr_t)pi, ((caddr_t)pi) + pp->pr_size);
1059: #endif
1.3 pk 1060:
1061: /*
1062: * Return to item list.
1063: */
1.97 yamt 1064: if (pp->pr_roflags & PR_NOTOUCH) {
1065: pr_item_notouch_put(pp, ph, v);
1066: } else {
1.2 pk 1067: #ifdef DIAGNOSTIC
1.97 yamt 1068: pi->pi_magic = PI_MAGIC;
1.3 pk 1069: #endif
1.32 chs 1070: #ifdef DEBUG
1.97 yamt 1071: {
1072: int i, *ip = v;
1.32 chs 1073:
1.97 yamt 1074: for (i = 0; i < pp->pr_size / sizeof(int); i++) {
1075: *ip++ = PI_MAGIC;
1076: }
1.32 chs 1077: }
1078: #endif
1079:
1.102 chs 1080: LIST_INSERT_HEAD(&ph->ph_itemlist, pi, pi_list);
1.97 yamt 1081: }
1.79 thorpej 1082: KDASSERT(ph->ph_nmissing != 0);
1.3 pk 1083: ph->ph_nmissing--;
1084: pp->pr_nput++;
1.20 thorpej 1085: pp->pr_nitems++;
1086: pp->pr_nout--;
1.3 pk 1087:
1088: /* Cancel "pool empty" condition if it exists */
1089: if (pp->pr_curpage == NULL)
1090: pp->pr_curpage = ph;
1091:
1092: if (pp->pr_flags & PR_WANTED) {
1093: pp->pr_flags &= ~PR_WANTED;
1.15 pk 1094: if (ph->ph_nmissing == 0)
1095: pp->pr_nidle++;
1.3 pk 1096: wakeup((caddr_t)pp);
1097: return;
1098: }
1099:
1100: /*
1.88 chs 1101: * If this page is now empty, do one of two things:
1.21 thorpej 1102: *
1.88 chs 1103: * (1) If we have more pages than the page high water mark,
1.96 thorpej 1104: * free the page back to the system. ONLY CONSIDER
1.90 thorpej 1105: * FREEING BACK A PAGE IF WE HAVE MORE THAN OUR MINIMUM PAGE
1106: * CLAIM.
1.21 thorpej 1107: *
1.88 chs 1108: * (2) Otherwise, move the page to the empty page list.
1109: *
1110: * Either way, select a new current page (so we use a partially-full
1111: * page if one is available).
1.3 pk 1112: */
1113: if (ph->ph_nmissing == 0) {
1.6 thorpej 1114: pp->pr_nidle++;
1.90 thorpej 1115: if (pp->pr_npages > pp->pr_minpages &&
1116: (pp->pr_npages > pp->pr_maxpages ||
1117: (pp->pr_alloc->pa_flags & PA_WANT) != 0)) {
1.101 thorpej 1118: pr_rmpage(pp, ph, pq);
1.3 pk 1119: } else {
1.88 chs 1120: LIST_REMOVE(ph, ph_pagelist);
1121: LIST_INSERT_HEAD(&pp->pr_emptypages, ph, ph_pagelist);
1.3 pk 1122:
1.21 thorpej 1123: /*
1124: * Update the timestamp on the page. A page must
1125: * be idle for some period of time before it can
1126: * be reclaimed by the pagedaemon. This minimizes
1127: * ping-pong'ing for memory.
1128: */
1129: s = splclock();
1130: ph->ph_time = mono_time;
1131: splx(s);
1.1 pk 1132: }
1.88 chs 1133: pool_update_curpage(pp);
1.1 pk 1134: }
1.88 chs 1135:
1.21 thorpej 1136: /*
1.88 chs 1137: * If the page was previously completely full, move it to the
1138: * partially-full list and make it the current page. The next
1139: * allocation will get the item from this page, instead of
1140: * further fragmenting the pool.
1.21 thorpej 1141: */
1142: else if (ph->ph_nmissing == (pp->pr_itemsperpage - 1)) {
1.88 chs 1143: LIST_REMOVE(ph, ph_pagelist);
1144: LIST_INSERT_HEAD(&pp->pr_partpages, ph, ph_pagelist);
1.21 thorpej 1145: pp->pr_curpage = ph;
1146: }
1.43 thorpej 1147: }
1148:
1149: /*
1150: * Return resource to the pool; must be called at appropriate spl level
1151: */
1.59 thorpej 1152: #ifdef POOL_DIAGNOSTIC
1.43 thorpej 1153: void
1154: _pool_put(struct pool *pp, void *v, const char *file, long line)
1155: {
1.101 thorpej 1156: struct pool_pagelist pq;
1157:
1158: LIST_INIT(&pq);
1.43 thorpej 1159:
1160: simple_lock(&pp->pr_slock);
1161: pr_enter(pp, file, line);
1162:
1.56 sommerfe 1163: pr_log(pp, v, PRLOG_PUT, file, line);
1164:
1.101 thorpej 1165: pool_do_put(pp, v, &pq);
1.21 thorpej 1166:
1.25 thorpej 1167: pr_leave(pp);
1.21 thorpej 1168: simple_unlock(&pp->pr_slock);
1.101 thorpej 1169:
1.102 chs 1170: pr_pagelist_free(pp, &pq);
1.1 pk 1171: }
1.57 sommerfe 1172: #undef pool_put
1.59 thorpej 1173: #endif /* POOL_DIAGNOSTIC */
1.1 pk 1174:
1.56 sommerfe 1175: void
1176: pool_put(struct pool *pp, void *v)
1177: {
1.101 thorpej 1178: struct pool_pagelist pq;
1179:
1180: LIST_INIT(&pq);
1.56 sommerfe 1181:
1182: simple_lock(&pp->pr_slock);
1.101 thorpej 1183: pool_do_put(pp, v, &pq);
1184: simple_unlock(&pp->pr_slock);
1.56 sommerfe 1185:
1.102 chs 1186: pr_pagelist_free(pp, &pq);
1.56 sommerfe 1187: }
1.57 sommerfe 1188:
1.59 thorpej 1189: #ifdef POOL_DIAGNOSTIC
1.57 sommerfe 1190: #define pool_put(h, v) _pool_put((h), (v), __FILE__, __LINE__)
1.56 sommerfe 1191: #endif
1.74 thorpej 1192:
1193: /*
1.112.2.1 yamt 1194: * pool_grow: grow a pool by a page.
1195: *
1196: * => called with pool locked.
1197: * => unlock and relock the pool.
1198: * => return with pool locked.
1199: */
1200:
1201: static int
1202: pool_grow(struct pool *pp, int flags)
1203: {
1204: struct pool_item_header *ph = NULL;
1205: char *cp;
1206:
1207: simple_unlock(&pp->pr_slock);
1208: cp = pool_allocator_alloc(pp, flags);
1209: if (__predict_true(cp != NULL)) {
1210: ph = pool_alloc_item_header(pp, cp, flags);
1211: }
1212: if (__predict_false(cp == NULL || ph == NULL)) {
1213: if (cp != NULL) {
1214: pool_allocator_free(pp, cp);
1215: }
1216: simple_lock(&pp->pr_slock);
1217: return ENOMEM;
1218: }
1219:
1220: simple_lock(&pp->pr_slock);
1221: pool_prime_page(pp, cp, ph);
1222: pp->pr_npagealloc++;
1223: return 0;
1224: }
1225:
1226: /*
1.74 thorpej 1227: * Add N items to the pool.
1228: */
1229: int
1230: pool_prime(struct pool *pp, int n)
1231: {
1.75 simonb 1232: int newpages;
1.112.2.1 yamt 1233: int error = 0;
1.74 thorpej 1234:
1235: simple_lock(&pp->pr_slock);
1236:
1237: newpages = roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1238:
1239: while (newpages-- > 0) {
1.112.2.1 yamt 1240: error = pool_grow(pp, PR_NOWAIT);
1241: if (error) {
1.74 thorpej 1242: break;
1243: }
1244: pp->pr_minpages++;
1245: }
1246:
1247: if (pp->pr_minpages >= pp->pr_maxpages)
1248: pp->pr_maxpages = pp->pr_minpages + 1; /* XXX */
1249:
1250: simple_unlock(&pp->pr_slock);
1.112.2.1 yamt 1251: return error;
1.74 thorpej 1252: }
1.55 thorpej 1253:
1254: /*
1.3 pk 1255: * Add a page worth of items to the pool.
1.21 thorpej 1256: *
1257: * Note, we must be called with the pool descriptor LOCKED.
1.3 pk 1258: */
1.55 thorpej 1259: static void
1260: pool_prime_page(struct pool *pp, caddr_t storage, struct pool_item_header *ph)
1.3 pk 1261: {
1262: struct pool_item *pi;
1263: caddr_t cp = storage;
1264: unsigned int align = pp->pr_align;
1265: unsigned int ioff = pp->pr_itemoffset;
1.55 thorpej 1266: int n;
1.89 yamt 1267: int s;
1.36 pk 1268:
1.91 yamt 1269: LOCK_ASSERT(simple_lock_held(&pp->pr_slock));
1270:
1.66 thorpej 1271: #ifdef DIAGNOSTIC
1272: if (((u_long)cp & (pp->pr_alloc->pa_pagesz - 1)) != 0)
1.36 pk 1273: panic("pool_prime_page: %s: unaligned page", pp->pr_wchan);
1.66 thorpej 1274: #endif
1.3 pk 1275:
1276: /*
1277: * Insert page header.
1278: */
1.88 chs 1279: LIST_INSERT_HEAD(&pp->pr_emptypages, ph, ph_pagelist);
1.102 chs 1280: LIST_INIT(&ph->ph_itemlist);
1.3 pk 1281: ph->ph_page = storage;
1282: ph->ph_nmissing = 0;
1.89 yamt 1283: s = splclock();
1284: ph->ph_time = mono_time;
1285: splx(s);
1.88 chs 1286: if ((pp->pr_roflags & PR_PHINPAGE) == 0)
1287: SPLAY_INSERT(phtree, &pp->pr_phtree, ph);
1.3 pk 1288:
1.6 thorpej 1289: pp->pr_nidle++;
1290:
1.3 pk 1291: /*
1292: * Color this page.
1293: */
1294: cp = (caddr_t)(cp + pp->pr_curcolor);
1295: if ((pp->pr_curcolor += align) > pp->pr_maxcolor)
1296: pp->pr_curcolor = 0;
1297:
1298: /*
1299: * Adjust storage to apply aligment to `pr_itemoffset' in each item.
1300: */
1301: if (ioff != 0)
1302: cp = (caddr_t)(cp + (align - ioff));
1303:
1304: /*
1305: * Insert remaining chunks on the bucket list.
1306: */
1307: n = pp->pr_itemsperpage;
1.20 thorpej 1308: pp->pr_nitems += n;
1.3 pk 1309:
1.97 yamt 1310: if (pp->pr_roflags & PR_NOTOUCH) {
1.99 yamt 1311: pool_item_freelist_t *freelist = PR_FREELIST(ph);
1.97 yamt 1312: int i;
1313:
1.99 yamt 1314: ph->ph_off = cp - storage;
1.97 yamt 1315: ph->ph_firstfree = 0;
1316: for (i = 0; i < n - 1; i++)
1317: freelist[i] = i + 1;
1318: freelist[n - 1] = PR_INDEX_EOL;
1319: } else {
1320: while (n--) {
1321: pi = (struct pool_item *)cp;
1.78 thorpej 1322:
1.97 yamt 1323: KASSERT(((((vaddr_t)pi) + ioff) & (align - 1)) == 0);
1.3 pk 1324:
1.97 yamt 1325: /* Insert on page list */
1.102 chs 1326: LIST_INSERT_HEAD(&ph->ph_itemlist, pi, pi_list);
1.3 pk 1327: #ifdef DIAGNOSTIC
1.97 yamt 1328: pi->pi_magic = PI_MAGIC;
1.3 pk 1329: #endif
1.97 yamt 1330: cp = (caddr_t)(cp + pp->pr_size);
1331: }
1.3 pk 1332: }
1333:
1334: /*
1335: * If the pool was depleted, point at the new page.
1336: */
1337: if (pp->pr_curpage == NULL)
1338: pp->pr_curpage = ph;
1339:
1340: if (++pp->pr_npages > pp->pr_hiwat)
1341: pp->pr_hiwat = pp->pr_npages;
1342: }
1343:
1.20 thorpej 1344: /*
1.52 thorpej 1345: * Used by pool_get() when nitems drops below the low water mark. This
1.88 chs 1346: * is used to catch up pr_nitems with the low water mark.
1.20 thorpej 1347: *
1.21 thorpej 1348: * Note 1, we never wait for memory here, we let the caller decide what to do.
1.20 thorpej 1349: *
1.73 thorpej 1350: * Note 2, we must be called with the pool already locked, and we return
1.20 thorpej 1351: * with it locked.
1352: */
1353: static int
1.42 thorpej 1354: pool_catchup(struct pool *pp)
1.20 thorpej 1355: {
1356: int error = 0;
1357:
1.54 thorpej 1358: while (POOL_NEEDS_CATCHUP(pp)) {
1.112.2.1 yamt 1359: error = pool_grow(pp, PR_NOWAIT);
1360: if (error) {
1.20 thorpej 1361: break;
1362: }
1363: }
1.112.2.1 yamt 1364: return error;
1.20 thorpej 1365: }
1366:
1.88 chs 1367: static void
1368: pool_update_curpage(struct pool *pp)
1369: {
1370:
1371: pp->pr_curpage = LIST_FIRST(&pp->pr_partpages);
1372: if (pp->pr_curpage == NULL) {
1373: pp->pr_curpage = LIST_FIRST(&pp->pr_emptypages);
1374: }
1375: }
1376:
1.3 pk 1377: void
1.42 thorpej 1378: pool_setlowat(struct pool *pp, int n)
1.3 pk 1379: {
1.15 pk 1380:
1.21 thorpej 1381: simple_lock(&pp->pr_slock);
1382:
1.3 pk 1383: pp->pr_minitems = n;
1.15 pk 1384: pp->pr_minpages = (n == 0)
1385: ? 0
1.18 thorpej 1386: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.20 thorpej 1387:
1388: /* Make sure we're caught up with the newly-set low water mark. */
1.75 simonb 1389: if (POOL_NEEDS_CATCHUP(pp) && pool_catchup(pp) != 0) {
1.20 thorpej 1390: /*
1391: * XXX: Should we log a warning? Should we set up a timeout
1392: * to try again in a second or so? The latter could break
1393: * a caller's assumptions about interrupt protection, etc.
1394: */
1395: }
1.21 thorpej 1396:
1397: simple_unlock(&pp->pr_slock);
1.3 pk 1398: }
1399:
1400: void
1.42 thorpej 1401: pool_sethiwat(struct pool *pp, int n)
1.3 pk 1402: {
1.15 pk 1403:
1.21 thorpej 1404: simple_lock(&pp->pr_slock);
1405:
1.15 pk 1406: pp->pr_maxpages = (n == 0)
1407: ? 0
1.18 thorpej 1408: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.21 thorpej 1409:
1410: simple_unlock(&pp->pr_slock);
1.3 pk 1411: }
1412:
1.20 thorpej 1413: void
1.42 thorpej 1414: pool_sethardlimit(struct pool *pp, int n, const char *warnmess, int ratecap)
1.20 thorpej 1415: {
1416:
1.21 thorpej 1417: simple_lock(&pp->pr_slock);
1.20 thorpej 1418:
1419: pp->pr_hardlimit = n;
1420: pp->pr_hardlimit_warning = warnmess;
1.31 thorpej 1421: pp->pr_hardlimit_ratecap.tv_sec = ratecap;
1422: pp->pr_hardlimit_warning_last.tv_sec = 0;
1423: pp->pr_hardlimit_warning_last.tv_usec = 0;
1.20 thorpej 1424:
1425: /*
1.21 thorpej 1426: * In-line version of pool_sethiwat(), because we don't want to
1427: * release the lock.
1.20 thorpej 1428: */
1429: pp->pr_maxpages = (n == 0)
1430: ? 0
1431: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.21 thorpej 1432:
1433: simple_unlock(&pp->pr_slock);
1.20 thorpej 1434: }
1.3 pk 1435:
1436: /*
1437: * Release all complete pages that have not been used recently.
1438: */
1.66 thorpej 1439: int
1.59 thorpej 1440: #ifdef POOL_DIAGNOSTIC
1.42 thorpej 1441: _pool_reclaim(struct pool *pp, const char *file, long line)
1.56 sommerfe 1442: #else
1443: pool_reclaim(struct pool *pp)
1444: #endif
1.3 pk 1445: {
1446: struct pool_item_header *ph, *phnext;
1.43 thorpej 1447: struct pool_cache *pc;
1.61 chs 1448: struct pool_pagelist pq;
1.102 chs 1449: struct pool_cache_grouplist pcgl;
1450: struct timeval curtime, diff;
1.21 thorpej 1451: int s;
1.3 pk 1452:
1.68 thorpej 1453: if (pp->pr_drain_hook != NULL) {
1454: /*
1455: * The drain hook must be called with the pool unlocked.
1456: */
1457: (*pp->pr_drain_hook)(pp->pr_drain_hook_arg, PR_NOWAIT);
1458: }
1459:
1.21 thorpej 1460: if (simple_lock_try(&pp->pr_slock) == 0)
1.66 thorpej 1461: return (0);
1.25 thorpej 1462: pr_enter(pp, file, line);
1.68 thorpej 1463:
1.88 chs 1464: LIST_INIT(&pq);
1.102 chs 1465: LIST_INIT(&pcgl);
1.3 pk 1466:
1.43 thorpej 1467: /*
1468: * Reclaim items from the pool's caches.
1469: */
1.102 chs 1470: LIST_FOREACH(pc, &pp->pr_cachelist, pc_poollist)
1471: pool_cache_reclaim(pc, &pq, &pcgl);
1.43 thorpej 1472:
1.21 thorpej 1473: s = splclock();
1474: curtime = mono_time;
1475: splx(s);
1476:
1.88 chs 1477: for (ph = LIST_FIRST(&pp->pr_emptypages); ph != NULL; ph = phnext) {
1478: phnext = LIST_NEXT(ph, ph_pagelist);
1.3 pk 1479:
1480: /* Check our minimum page claim */
1481: if (pp->pr_npages <= pp->pr_minpages)
1482: break;
1483:
1.88 chs 1484: KASSERT(ph->ph_nmissing == 0);
1485: timersub(&curtime, &ph->ph_time, &diff);
1486: if (diff.tv_sec < pool_inactive_time)
1487: continue;
1.21 thorpej 1488:
1.88 chs 1489: /*
1490: * If freeing this page would put us below
1491: * the low water mark, stop now.
1492: */
1493: if ((pp->pr_nitems - pp->pr_itemsperpage) <
1494: pp->pr_minitems)
1495: break;
1.21 thorpej 1496:
1.88 chs 1497: pr_rmpage(pp, ph, &pq);
1.3 pk 1498: }
1499:
1.25 thorpej 1500: pr_leave(pp);
1.21 thorpej 1501: simple_unlock(&pp->pr_slock);
1.102 chs 1502: if (LIST_EMPTY(&pq) && LIST_EMPTY(&pcgl))
1503: return 0;
1.66 thorpej 1504:
1.101 thorpej 1505: pr_pagelist_free(pp, &pq);
1.102 chs 1506: pcg_grouplist_free(&pcgl);
1.66 thorpej 1507: return (1);
1.3 pk 1508: }
1509:
1510: /*
1511: * Drain pools, one at a time.
1.21 thorpej 1512: *
1513: * Note, we must never be called from an interrupt context.
1.3 pk 1514: */
1515: void
1.42 thorpej 1516: pool_drain(void *arg)
1.3 pk 1517: {
1518: struct pool *pp;
1.23 thorpej 1519: int s;
1.3 pk 1520:
1.61 chs 1521: pp = NULL;
1.49 thorpej 1522: s = splvm();
1.23 thorpej 1523: simple_lock(&pool_head_slock);
1.61 chs 1524: if (drainpp == NULL) {
1.102 chs 1525: drainpp = LIST_FIRST(&pool_head);
1.61 chs 1526: }
1527: if (drainpp) {
1528: pp = drainpp;
1.102 chs 1529: drainpp = LIST_NEXT(pp, pr_poollist);
1.61 chs 1530: }
1531: simple_unlock(&pool_head_slock);
1.63 chs 1532: pool_reclaim(pp);
1.61 chs 1533: splx(s);
1.3 pk 1534: }
1535:
1536: /*
1537: * Diagnostic helpers.
1538: */
1539: void
1.42 thorpej 1540: pool_print(struct pool *pp, const char *modif)
1.21 thorpej 1541: {
1542: int s;
1543:
1.49 thorpej 1544: s = splvm();
1.25 thorpej 1545: if (simple_lock_try(&pp->pr_slock) == 0) {
1546: printf("pool %s is locked; try again later\n",
1547: pp->pr_wchan);
1548: splx(s);
1549: return;
1550: }
1551: pool_print1(pp, modif, printf);
1.21 thorpej 1552: simple_unlock(&pp->pr_slock);
1553: splx(s);
1554: }
1555:
1.25 thorpej 1556: void
1.108 yamt 1557: pool_printall(const char *modif, void (*pr)(const char *, ...))
1558: {
1559: struct pool *pp;
1560:
1561: if (simple_lock_try(&pool_head_slock) == 0) {
1562: (*pr)("WARNING: pool_head_slock is locked\n");
1563: } else {
1564: simple_unlock(&pool_head_slock);
1565: }
1566:
1567: LIST_FOREACH(pp, &pool_head, pr_poollist) {
1568: pool_printit(pp, modif, pr);
1569: }
1570: }
1571:
1572: void
1.42 thorpej 1573: pool_printit(struct pool *pp, const char *modif, void (*pr)(const char *, ...))
1.25 thorpej 1574: {
1575:
1576: if (pp == NULL) {
1577: (*pr)("Must specify a pool to print.\n");
1578: return;
1579: }
1580:
1581: /*
1582: * Called from DDB; interrupts should be blocked, and all
1583: * other processors should be paused. We can skip locking
1584: * the pool in this case.
1585: *
1586: * We do a simple_lock_try() just to print the lock
1587: * status, however.
1588: */
1589:
1590: if (simple_lock_try(&pp->pr_slock) == 0)
1591: (*pr)("WARNING: pool %s is locked\n", pp->pr_wchan);
1592: else
1.107 yamt 1593: simple_unlock(&pp->pr_slock);
1.25 thorpej 1594:
1595: pool_print1(pp, modif, pr);
1596: }
1597:
1.21 thorpej 1598: static void
1.97 yamt 1599: pool_print_pagelist(struct pool *pp, struct pool_pagelist *pl,
1600: void (*pr)(const char *, ...))
1.88 chs 1601: {
1602: struct pool_item_header *ph;
1603: #ifdef DIAGNOSTIC
1604: struct pool_item *pi;
1605: #endif
1606:
1607: LIST_FOREACH(ph, pl, ph_pagelist) {
1608: (*pr)("\t\tpage %p, nmissing %d, time %lu,%lu\n",
1609: ph->ph_page, ph->ph_nmissing,
1610: (u_long)ph->ph_time.tv_sec,
1611: (u_long)ph->ph_time.tv_usec);
1612: #ifdef DIAGNOSTIC
1.97 yamt 1613: if (!(pp->pr_roflags & PR_NOTOUCH)) {
1.102 chs 1614: LIST_FOREACH(pi, &ph->ph_itemlist, pi_list) {
1.97 yamt 1615: if (pi->pi_magic != PI_MAGIC) {
1616: (*pr)("\t\t\titem %p, magic 0x%x\n",
1617: pi, pi->pi_magic);
1618: }
1.88 chs 1619: }
1620: }
1621: #endif
1622: }
1623: }
1624:
1625: static void
1.42 thorpej 1626: pool_print1(struct pool *pp, const char *modif, void (*pr)(const char *, ...))
1.3 pk 1627: {
1.25 thorpej 1628: struct pool_item_header *ph;
1.44 thorpej 1629: struct pool_cache *pc;
1630: struct pool_cache_group *pcg;
1631: int i, print_log = 0, print_pagelist = 0, print_cache = 0;
1.25 thorpej 1632: char c;
1633:
1634: while ((c = *modif++) != '\0') {
1635: if (c == 'l')
1636: print_log = 1;
1637: if (c == 'p')
1638: print_pagelist = 1;
1.44 thorpej 1639: if (c == 'c')
1640: print_cache = 1;
1.25 thorpej 1641: }
1642:
1643: (*pr)("POOL %s: size %u, align %u, ioff %u, roflags 0x%08x\n",
1644: pp->pr_wchan, pp->pr_size, pp->pr_align, pp->pr_itemoffset,
1645: pp->pr_roflags);
1.66 thorpej 1646: (*pr)("\talloc %p\n", pp->pr_alloc);
1.25 thorpej 1647: (*pr)("\tminitems %u, minpages %u, maxpages %u, npages %u\n",
1648: pp->pr_minitems, pp->pr_minpages, pp->pr_maxpages, pp->pr_npages);
1649: (*pr)("\titemsperpage %u, nitems %u, nout %u, hardlimit %u\n",
1650: pp->pr_itemsperpage, pp->pr_nitems, pp->pr_nout, pp->pr_hardlimit);
1651:
1652: (*pr)("\n\tnget %lu, nfail %lu, nput %lu\n",
1653: pp->pr_nget, pp->pr_nfail, pp->pr_nput);
1654: (*pr)("\tnpagealloc %lu, npagefree %lu, hiwat %u, nidle %lu\n",
1655: pp->pr_npagealloc, pp->pr_npagefree, pp->pr_hiwat, pp->pr_nidle);
1656:
1657: if (print_pagelist == 0)
1658: goto skip_pagelist;
1659:
1.88 chs 1660: if ((ph = LIST_FIRST(&pp->pr_emptypages)) != NULL)
1661: (*pr)("\n\tempty page list:\n");
1.97 yamt 1662: pool_print_pagelist(pp, &pp->pr_emptypages, pr);
1.88 chs 1663: if ((ph = LIST_FIRST(&pp->pr_fullpages)) != NULL)
1664: (*pr)("\n\tfull page list:\n");
1.97 yamt 1665: pool_print_pagelist(pp, &pp->pr_fullpages, pr);
1.88 chs 1666: if ((ph = LIST_FIRST(&pp->pr_partpages)) != NULL)
1667: (*pr)("\n\tpartial-page list:\n");
1.97 yamt 1668: pool_print_pagelist(pp, &pp->pr_partpages, pr);
1.88 chs 1669:
1.25 thorpej 1670: if (pp->pr_curpage == NULL)
1671: (*pr)("\tno current page\n");
1672: else
1673: (*pr)("\tcurpage %p\n", pp->pr_curpage->ph_page);
1674:
1675: skip_pagelist:
1676: if (print_log == 0)
1677: goto skip_log;
1678:
1679: (*pr)("\n");
1680: if ((pp->pr_roflags & PR_LOGGING) == 0)
1681: (*pr)("\tno log\n");
1682: else
1683: pr_printlog(pp, NULL, pr);
1.3 pk 1684:
1.25 thorpej 1685: skip_log:
1.44 thorpej 1686: if (print_cache == 0)
1687: goto skip_cache;
1688:
1.102 chs 1689: #define PR_GROUPLIST(pcg) \
1690: (*pr)("\t\tgroup %p: avail %d\n", pcg, pcg->pcg_avail); \
1691: for (i = 0; i < PCG_NOBJECTS; i++) { \
1692: if (pcg->pcg_objects[i].pcgo_pa != \
1693: POOL_PADDR_INVALID) { \
1694: (*pr)("\t\t\t%p, 0x%llx\n", \
1695: pcg->pcg_objects[i].pcgo_va, \
1696: (unsigned long long) \
1697: pcg->pcg_objects[i].pcgo_pa); \
1698: } else { \
1699: (*pr)("\t\t\t%p\n", \
1700: pcg->pcg_objects[i].pcgo_va); \
1701: } \
1702: }
1703:
1704: LIST_FOREACH(pc, &pp->pr_cachelist, pc_poollist) {
1.103 chs 1705: (*pr)("\tcache %p\n", pc);
1.48 thorpej 1706: (*pr)("\t hits %lu misses %lu ngroups %lu nitems %lu\n",
1707: pc->pc_hits, pc->pc_misses, pc->pc_ngroups, pc->pc_nitems);
1.102 chs 1708: (*pr)("\t full groups:\n");
1.103 chs 1709: LIST_FOREACH(pcg, &pc->pc_fullgroups, pcg_list) {
1.102 chs 1710: PR_GROUPLIST(pcg);
1.103 chs 1711: }
1.102 chs 1712: (*pr)("\t partial groups:\n");
1.103 chs 1713: LIST_FOREACH(pcg, &pc->pc_partgroups, pcg_list) {
1.102 chs 1714: PR_GROUPLIST(pcg);
1.103 chs 1715: }
1.102 chs 1716: (*pr)("\t empty groups:\n");
1.103 chs 1717: LIST_FOREACH(pcg, &pc->pc_emptygroups, pcg_list) {
1.102 chs 1718: PR_GROUPLIST(pcg);
1.103 chs 1719: }
1.44 thorpej 1720: }
1.102 chs 1721: #undef PR_GROUPLIST
1.44 thorpej 1722:
1723: skip_cache:
1.88 chs 1724: pr_enter_check(pp, pr);
1725: }
1726:
1727: static int
1728: pool_chk_page(struct pool *pp, const char *label, struct pool_item_header *ph)
1729: {
1730: struct pool_item *pi;
1731: caddr_t page;
1732: int n;
1733:
1734: page = (caddr_t)((u_long)ph & pp->pr_alloc->pa_pagemask);
1735: if (page != ph->ph_page &&
1736: (pp->pr_roflags & PR_PHINPAGE) != 0) {
1737: if (label != NULL)
1738: printf("%s: ", label);
1739: printf("pool(%p:%s): page inconsistency: page %p;"
1740: " at page head addr %p (p %p)\n", pp,
1741: pp->pr_wchan, ph->ph_page,
1742: ph, page);
1743: return 1;
1744: }
1.3 pk 1745:
1.97 yamt 1746: if ((pp->pr_roflags & PR_NOTOUCH) != 0)
1747: return 0;
1748:
1.102 chs 1749: for (pi = LIST_FIRST(&ph->ph_itemlist), n = 0;
1.88 chs 1750: pi != NULL;
1.102 chs 1751: pi = LIST_NEXT(pi,pi_list), n++) {
1.88 chs 1752:
1753: #ifdef DIAGNOSTIC
1754: if (pi->pi_magic != PI_MAGIC) {
1755: if (label != NULL)
1756: printf("%s: ", label);
1757: printf("pool(%s): free list modified: magic=%x;"
1758: " page %p; item ordinal %d;"
1759: " addr %p (p %p)\n",
1760: pp->pr_wchan, pi->pi_magic, ph->ph_page,
1761: n, pi, page);
1762: panic("pool");
1763: }
1764: #endif
1765: page =
1766: (caddr_t)((u_long)pi & pp->pr_alloc->pa_pagemask);
1767: if (page == ph->ph_page)
1768: continue;
1769:
1770: if (label != NULL)
1771: printf("%s: ", label);
1772: printf("pool(%p:%s): page inconsistency: page %p;"
1773: " item ordinal %d; addr %p (p %p)\n", pp,
1774: pp->pr_wchan, ph->ph_page,
1775: n, pi, page);
1776: return 1;
1777: }
1778: return 0;
1.3 pk 1779: }
1780:
1.88 chs 1781:
1.3 pk 1782: int
1.42 thorpej 1783: pool_chk(struct pool *pp, const char *label)
1.3 pk 1784: {
1785: struct pool_item_header *ph;
1786: int r = 0;
1787:
1.21 thorpej 1788: simple_lock(&pp->pr_slock);
1.88 chs 1789: LIST_FOREACH(ph, &pp->pr_emptypages, ph_pagelist) {
1790: r = pool_chk_page(pp, label, ph);
1791: if (r) {
1792: goto out;
1793: }
1794: }
1795: LIST_FOREACH(ph, &pp->pr_fullpages, ph_pagelist) {
1796: r = pool_chk_page(pp, label, ph);
1797: if (r) {
1.3 pk 1798: goto out;
1799: }
1.88 chs 1800: }
1801: LIST_FOREACH(ph, &pp->pr_partpages, ph_pagelist) {
1802: r = pool_chk_page(pp, label, ph);
1803: if (r) {
1.3 pk 1804: goto out;
1805: }
1806: }
1.88 chs 1807:
1.3 pk 1808: out:
1.21 thorpej 1809: simple_unlock(&pp->pr_slock);
1.3 pk 1810: return (r);
1.43 thorpej 1811: }
1812:
1813: /*
1814: * pool_cache_init:
1815: *
1816: * Initialize a pool cache.
1817: *
1818: * NOTE: If the pool must be protected from interrupts, we expect
1819: * to be called at the appropriate interrupt priority level.
1820: */
1821: void
1822: pool_cache_init(struct pool_cache *pc, struct pool *pp,
1823: int (*ctor)(void *, void *, int),
1824: void (*dtor)(void *, void *),
1825: void *arg)
1826: {
1827:
1.102 chs 1828: LIST_INIT(&pc->pc_emptygroups);
1829: LIST_INIT(&pc->pc_fullgroups);
1830: LIST_INIT(&pc->pc_partgroups);
1.43 thorpej 1831: simple_lock_init(&pc->pc_slock);
1832:
1833: pc->pc_pool = pp;
1834:
1835: pc->pc_ctor = ctor;
1836: pc->pc_dtor = dtor;
1837: pc->pc_arg = arg;
1838:
1.48 thorpej 1839: pc->pc_hits = 0;
1840: pc->pc_misses = 0;
1841:
1842: pc->pc_ngroups = 0;
1843:
1844: pc->pc_nitems = 0;
1845:
1.43 thorpej 1846: simple_lock(&pp->pr_slock);
1.102 chs 1847: LIST_INSERT_HEAD(&pp->pr_cachelist, pc, pc_poollist);
1.43 thorpej 1848: simple_unlock(&pp->pr_slock);
1849: }
1850:
1851: /*
1852: * pool_cache_destroy:
1853: *
1854: * Destroy a pool cache.
1855: */
1856: void
1857: pool_cache_destroy(struct pool_cache *pc)
1858: {
1859: struct pool *pp = pc->pc_pool;
1860:
1861: /* First, invalidate the entire cache. */
1862: pool_cache_invalidate(pc);
1863:
1864: /* ...and remove it from the pool's cache list. */
1865: simple_lock(&pp->pr_slock);
1.102 chs 1866: LIST_REMOVE(pc, pc_poollist);
1.43 thorpej 1867: simple_unlock(&pp->pr_slock);
1868: }
1869:
1.110 perry 1870: static inline void *
1.87 thorpej 1871: pcg_get(struct pool_cache_group *pcg, paddr_t *pap)
1.43 thorpej 1872: {
1873: void *object;
1874: u_int idx;
1875:
1876: KASSERT(pcg->pcg_avail <= PCG_NOBJECTS);
1.45 thorpej 1877: KASSERT(pcg->pcg_avail != 0);
1.43 thorpej 1878: idx = --pcg->pcg_avail;
1879:
1.87 thorpej 1880: KASSERT(pcg->pcg_objects[idx].pcgo_va != NULL);
1881: object = pcg->pcg_objects[idx].pcgo_va;
1882: if (pap != NULL)
1883: *pap = pcg->pcg_objects[idx].pcgo_pa;
1884: pcg->pcg_objects[idx].pcgo_va = NULL;
1.43 thorpej 1885:
1886: return (object);
1887: }
1888:
1.110 perry 1889: static inline void
1.87 thorpej 1890: pcg_put(struct pool_cache_group *pcg, void *object, paddr_t pa)
1.43 thorpej 1891: {
1892: u_int idx;
1893:
1894: KASSERT(pcg->pcg_avail < PCG_NOBJECTS);
1895: idx = pcg->pcg_avail++;
1896:
1.87 thorpej 1897: KASSERT(pcg->pcg_objects[idx].pcgo_va == NULL);
1898: pcg->pcg_objects[idx].pcgo_va = object;
1899: pcg->pcg_objects[idx].pcgo_pa = pa;
1.43 thorpej 1900: }
1901:
1.102 chs 1902: static void
1903: pcg_grouplist_free(struct pool_cache_grouplist *pcgl)
1904: {
1905: struct pool_cache_group *pcg;
1906: int s;
1907:
1908: s = splvm();
1909: while ((pcg = LIST_FIRST(pcgl)) != NULL) {
1910: LIST_REMOVE(pcg, pcg_list);
1911: pool_put(&pcgpool, pcg);
1912: }
1913: splx(s);
1914: }
1915:
1.43 thorpej 1916: /*
1.87 thorpej 1917: * pool_cache_get{,_paddr}:
1.43 thorpej 1918: *
1.87 thorpej 1919: * Get an object from a pool cache (optionally returning
1920: * the physical address of the object).
1.43 thorpej 1921: */
1922: void *
1.87 thorpej 1923: pool_cache_get_paddr(struct pool_cache *pc, int flags, paddr_t *pap)
1.43 thorpej 1924: {
1925: struct pool_cache_group *pcg;
1926: void *object;
1.58 thorpej 1927:
1928: #ifdef LOCKDEBUG
1929: if (flags & PR_WAITOK)
1930: simple_lock_only_held(NULL, "pool_cache_get(PR_WAITOK)");
1931: #endif
1.43 thorpej 1932:
1933: simple_lock(&pc->pc_slock);
1934:
1.102 chs 1935: pcg = LIST_FIRST(&pc->pc_partgroups);
1936: if (pcg == NULL) {
1937: pcg = LIST_FIRST(&pc->pc_fullgroups);
1938: if (pcg != NULL) {
1939: LIST_REMOVE(pcg, pcg_list);
1940: LIST_INSERT_HEAD(&pc->pc_partgroups, pcg, pcg_list);
1.43 thorpej 1941: }
1.102 chs 1942: }
1943: if (pcg == NULL) {
1.43 thorpej 1944:
1945: /*
1946: * No groups with any available objects. Allocate
1947: * a new object, construct it, and return it to
1948: * the caller. We will allocate a group, if necessary,
1949: * when the object is freed back to the cache.
1950: */
1.48 thorpej 1951: pc->pc_misses++;
1.43 thorpej 1952: simple_unlock(&pc->pc_slock);
1953: object = pool_get(pc->pc_pool, flags);
1954: if (object != NULL && pc->pc_ctor != NULL) {
1955: if ((*pc->pc_ctor)(pc->pc_arg, object, flags) != 0) {
1956: pool_put(pc->pc_pool, object);
1957: return (NULL);
1958: }
1959: }
1.87 thorpej 1960: if (object != NULL && pap != NULL) {
1961: #ifdef POOL_VTOPHYS
1962: *pap = POOL_VTOPHYS(object);
1963: #else
1964: *pap = POOL_PADDR_INVALID;
1965: #endif
1966: }
1.43 thorpej 1967: return (object);
1968: }
1969:
1.48 thorpej 1970: pc->pc_hits++;
1971: pc->pc_nitems--;
1.87 thorpej 1972: object = pcg_get(pcg, pap);
1.43 thorpej 1973:
1.102 chs 1974: if (pcg->pcg_avail == 0) {
1975: LIST_REMOVE(pcg, pcg_list);
1976: LIST_INSERT_HEAD(&pc->pc_emptygroups, pcg, pcg_list);
1977: }
1.43 thorpej 1978: simple_unlock(&pc->pc_slock);
1979:
1980: return (object);
1981: }
1982:
1983: /*
1.87 thorpej 1984: * pool_cache_put{,_paddr}:
1.43 thorpej 1985: *
1.87 thorpej 1986: * Put an object back to the pool cache (optionally caching the
1987: * physical address of the object).
1.43 thorpej 1988: */
1989: void
1.87 thorpej 1990: pool_cache_put_paddr(struct pool_cache *pc, void *object, paddr_t pa)
1.43 thorpej 1991: {
1992: struct pool_cache_group *pcg;
1.60 thorpej 1993: int s;
1.43 thorpej 1994:
1.109 christos 1995: if (__predict_false((pc->pc_pool->pr_flags & PR_WANTED) != 0)) {
1996: goto destruct;
1997: }
1998:
1.43 thorpej 1999: simple_lock(&pc->pc_slock);
2000:
1.102 chs 2001: pcg = LIST_FIRST(&pc->pc_partgroups);
2002: if (pcg == NULL) {
2003: pcg = LIST_FIRST(&pc->pc_emptygroups);
2004: if (pcg != NULL) {
2005: LIST_REMOVE(pcg, pcg_list);
2006: LIST_INSERT_HEAD(&pc->pc_partgroups, pcg, pcg_list);
1.43 thorpej 2007: }
1.102 chs 2008: }
2009: if (pcg == NULL) {
1.43 thorpej 2010:
2011: /*
2012: * No empty groups to free the object to. Attempt to
1.47 thorpej 2013: * allocate one.
1.43 thorpej 2014: */
1.47 thorpej 2015: simple_unlock(&pc->pc_slock);
1.60 thorpej 2016: s = splvm();
1.43 thorpej 2017: pcg = pool_get(&pcgpool, PR_NOWAIT);
1.60 thorpej 2018: splx(s);
1.102 chs 2019: if (pcg == NULL) {
1.109 christos 2020: destruct:
1.102 chs 2021:
2022: /*
2023: * Unable to allocate a cache group; destruct the object
2024: * and free it back to the pool.
2025: */
2026: pool_cache_destruct_object(pc, object);
2027: return;
1.43 thorpej 2028: }
1.102 chs 2029: memset(pcg, 0, sizeof(*pcg));
2030: simple_lock(&pc->pc_slock);
2031: pc->pc_ngroups++;
2032: LIST_INSERT_HEAD(&pc->pc_partgroups, pcg, pcg_list);
1.43 thorpej 2033: }
2034:
1.48 thorpej 2035: pc->pc_nitems++;
1.87 thorpej 2036: pcg_put(pcg, object, pa);
1.43 thorpej 2037:
1.102 chs 2038: if (pcg->pcg_avail == PCG_NOBJECTS) {
2039: LIST_REMOVE(pcg, pcg_list);
2040: LIST_INSERT_HEAD(&pc->pc_fullgroups, pcg, pcg_list);
2041: }
1.43 thorpej 2042: simple_unlock(&pc->pc_slock);
1.51 thorpej 2043: }
2044:
2045: /*
2046: * pool_cache_destruct_object:
2047: *
2048: * Force destruction of an object and its release back into
2049: * the pool.
2050: */
2051: void
2052: pool_cache_destruct_object(struct pool_cache *pc, void *object)
2053: {
2054:
2055: if (pc->pc_dtor != NULL)
2056: (*pc->pc_dtor)(pc->pc_arg, object);
2057: pool_put(pc->pc_pool, object);
1.43 thorpej 2058: }
2059:
1.102 chs 2060: static void
1.106 christos 2061: pool_do_cache_invalidate_grouplist(struct pool_cache_grouplist *pcgsl,
1.105 christos 2062: struct pool_cache *pc, struct pool_pagelist *pq,
1.106 christos 2063: struct pool_cache_grouplist *pcgdl)
1.102 chs 2064: {
1.106 christos 2065: struct pool_cache_group *pcg, *npcg;
1.102 chs 2066: void *object;
2067:
1.106 christos 2068: for (pcg = LIST_FIRST(pcgsl); pcg != NULL; pcg = npcg) {
1.102 chs 2069: npcg = LIST_NEXT(pcg, pcg_list);
2070: while (pcg->pcg_avail != 0) {
2071: pc->pc_nitems--;
2072: object = pcg_get(pcg, NULL);
2073: if (pc->pc_dtor != NULL)
2074: (*pc->pc_dtor)(pc->pc_arg, object);
2075: pool_do_put(pc->pc_pool, object, pq);
2076: }
1.103 chs 2077: pc->pc_ngroups--;
1.102 chs 2078: LIST_REMOVE(pcg, pcg_list);
1.106 christos 2079: LIST_INSERT_HEAD(pcgdl, pcg, pcg_list);
1.102 chs 2080: }
1.105 christos 2081: }
2082:
2083: static void
2084: pool_do_cache_invalidate(struct pool_cache *pc, struct pool_pagelist *pq,
2085: struct pool_cache_grouplist *pcgl)
2086: {
2087:
2088: LOCK_ASSERT(simple_lock_held(&pc->pc_slock));
2089: LOCK_ASSERT(simple_lock_held(&pc->pc_pool->pr_slock));
2090:
1.106 christos 2091: pool_do_cache_invalidate_grouplist(&pc->pc_fullgroups, pc, pq, pcgl);
2092: pool_do_cache_invalidate_grouplist(&pc->pc_partgroups, pc, pq, pcgl);
1.103 chs 2093:
2094: KASSERT(LIST_EMPTY(&pc->pc_partgroups));
2095: KASSERT(LIST_EMPTY(&pc->pc_fullgroups));
2096: KASSERT(pc->pc_nitems == 0);
1.102 chs 2097: }
2098:
1.43 thorpej 2099: /*
1.101 thorpej 2100: * pool_cache_invalidate:
1.43 thorpej 2101: *
1.101 thorpej 2102: * Invalidate a pool cache (destruct and release all of the
2103: * cached objects).
1.43 thorpej 2104: */
1.101 thorpej 2105: void
2106: pool_cache_invalidate(struct pool_cache *pc)
1.43 thorpej 2107: {
1.101 thorpej 2108: struct pool_pagelist pq;
1.102 chs 2109: struct pool_cache_grouplist pcgl;
1.101 thorpej 2110:
2111: LIST_INIT(&pq);
1.102 chs 2112: LIST_INIT(&pcgl);
1.101 thorpej 2113:
2114: simple_lock(&pc->pc_slock);
2115: simple_lock(&pc->pc_pool->pr_slock);
1.43 thorpej 2116:
1.102 chs 2117: pool_do_cache_invalidate(pc, &pq, &pcgl);
1.43 thorpej 2118:
1.101 thorpej 2119: simple_unlock(&pc->pc_pool->pr_slock);
2120: simple_unlock(&pc->pc_slock);
1.43 thorpej 2121:
1.102 chs 2122: pr_pagelist_free(pc->pc_pool, &pq);
2123: pcg_grouplist_free(&pcgl);
1.43 thorpej 2124: }
2125:
2126: /*
2127: * pool_cache_reclaim:
2128: *
2129: * Reclaim a pool cache for pool_reclaim().
2130: */
2131: static void
1.102 chs 2132: pool_cache_reclaim(struct pool_cache *pc, struct pool_pagelist *pq,
2133: struct pool_cache_grouplist *pcgl)
1.43 thorpej 2134: {
1.101 thorpej 2135:
2136: /*
2137: * We're locking in the wrong order (normally pool_cache -> pool,
2138: * but the pool is already locked when we get here), so we have
2139: * to use trylock. If we can't lock the pool_cache, it's not really
2140: * a big deal here.
2141: */
2142: if (simple_lock_try(&pc->pc_slock) == 0)
2143: return;
2144:
1.102 chs 2145: pool_do_cache_invalidate(pc, pq, pcgl);
1.43 thorpej 2146:
2147: simple_unlock(&pc->pc_slock);
1.3 pk 2148: }
1.66 thorpej 2149:
2150: /*
2151: * Pool backend allocators.
2152: *
2153: * Each pool has a backend allocator that handles allocation, deallocation,
2154: * and any additional draining that might be needed.
2155: *
2156: * We provide two standard allocators:
2157: *
2158: * pool_allocator_kmem - the default when no allocator is specified
2159: *
2160: * pool_allocator_nointr - used for pools that will not be accessed
2161: * in interrupt context.
2162: */
2163: void *pool_page_alloc(struct pool *, int);
2164: void pool_page_free(struct pool *, void *);
2165:
1.112 bjh21 2166: #ifdef POOL_SUBPAGE
2167: struct pool_allocator pool_allocator_kmem_fullpage = {
2168: pool_page_alloc, pool_page_free, 0,
2169: };
2170: #else
1.66 thorpej 2171: struct pool_allocator pool_allocator_kmem = {
2172: pool_page_alloc, pool_page_free, 0,
2173: };
1.112 bjh21 2174: #endif
1.66 thorpej 2175:
2176: void *pool_page_alloc_nointr(struct pool *, int);
2177: void pool_page_free_nointr(struct pool *, void *);
2178:
1.112 bjh21 2179: #ifdef POOL_SUBPAGE
2180: struct pool_allocator pool_allocator_nointr_fullpage = {
2181: pool_page_alloc_nointr, pool_page_free_nointr, 0,
2182: };
2183: #else
1.66 thorpej 2184: struct pool_allocator pool_allocator_nointr = {
2185: pool_page_alloc_nointr, pool_page_free_nointr, 0,
2186: };
1.112 bjh21 2187: #endif
1.66 thorpej 2188:
2189: #ifdef POOL_SUBPAGE
2190: void *pool_subpage_alloc(struct pool *, int);
2191: void pool_subpage_free(struct pool *, void *);
2192:
1.112 bjh21 2193: struct pool_allocator pool_allocator_kmem = {
2194: pool_subpage_alloc, pool_subpage_free, POOL_SUBPAGE,
2195: };
2196:
2197: void *pool_subpage_alloc_nointr(struct pool *, int);
2198: void pool_subpage_free_nointr(struct pool *, void *);
2199:
2200: struct pool_allocator pool_allocator_nointr = {
2201: pool_subpage_alloc, pool_subpage_free, POOL_SUBPAGE,
1.66 thorpej 2202: };
2203: #endif /* POOL_SUBPAGE */
2204:
2205: /*
2206: * We have at least three different resources for the same allocation and
2207: * each resource can be depleted. First, we have the ready elements in the
2208: * pool. Then we have the resource (typically a vm_map) for this allocator.
2209: * Finally, we have physical memory. Waiting for any of these can be
2210: * unnecessary when any other is freed, but the kernel doesn't support
2211: * sleeping on multiple wait channels, so we have to employ another strategy.
2212: *
2213: * The caller sleeps on the pool (so that it can be awakened when an item
2214: * is returned to the pool), but we set PA_WANT on the allocator. When a
2215: * page is returned to the allocator and PA_WANT is set, pool_allocator_free
2216: * will wake up all sleeping pools belonging to this allocator.
2217: *
2218: * XXX Thundering herd.
2219: */
2220: void *
2221: pool_allocator_alloc(struct pool *org, int flags)
2222: {
2223: struct pool_allocator *pa = org->pr_alloc;
2224: struct pool *pp, *start;
2225: int s, freed;
2226: void *res;
2227:
1.91 yamt 2228: LOCK_ASSERT(!simple_lock_held(&org->pr_slock));
2229:
1.66 thorpej 2230: do {
2231: if ((res = (*pa->pa_alloc)(org, flags)) != NULL)
2232: return (res);
1.68 thorpej 2233: if ((flags & PR_WAITOK) == 0) {
2234: /*
2235: * We only run the drain hookhere if PR_NOWAIT.
2236: * In other cases, the hook will be run in
2237: * pool_reclaim().
2238: */
2239: if (org->pr_drain_hook != NULL) {
2240: (*org->pr_drain_hook)(org->pr_drain_hook_arg,
2241: flags);
2242: if ((res = (*pa->pa_alloc)(org, flags)) != NULL)
2243: return (res);
2244: }
1.66 thorpej 2245: break;
1.68 thorpej 2246: }
1.66 thorpej 2247:
2248: /*
1.109 christos 2249: * Drain all pools, that use this allocator.
2250: * We do this to reclaim VA space.
1.66 thorpej 2251: * pa_alloc is responsible for waiting for
2252: * physical memory.
2253: *
2254: * XXX We risk looping forever if start if someone
2255: * calls pool_destroy on "start". But there is no
2256: * other way to have potentially sleeping pool_reclaim,
2257: * non-sleeping locks on pool_allocator, and some
2258: * stirring of drained pools in the allocator.
1.68 thorpej 2259: *
2260: * XXX Maybe we should use pool_head_slock for locking
2261: * the allocators?
1.66 thorpej 2262: */
2263: freed = 0;
2264:
2265: s = splvm();
2266: simple_lock(&pa->pa_slock);
2267: pp = start = TAILQ_FIRST(&pa->pa_list);
2268: do {
2269: TAILQ_REMOVE(&pa->pa_list, pp, pr_alloc_list);
2270: TAILQ_INSERT_TAIL(&pa->pa_list, pp, pr_alloc_list);
1.73 thorpej 2271: simple_unlock(&pa->pa_slock);
1.66 thorpej 2272: freed = pool_reclaim(pp);
1.73 thorpej 2273: simple_lock(&pa->pa_slock);
1.66 thorpej 2274: } while ((pp = TAILQ_FIRST(&pa->pa_list)) != start &&
2275: freed == 0);
2276:
2277: if (freed == 0) {
2278: /*
2279: * We set PA_WANT here, the caller will most likely
2280: * sleep waiting for pages (if not, this won't hurt
2281: * that much), and there is no way to set this in
2282: * the caller without violating locking order.
2283: */
2284: pa->pa_flags |= PA_WANT;
2285: }
2286: simple_unlock(&pa->pa_slock);
2287: splx(s);
2288: } while (freed);
2289: return (NULL);
2290: }
2291:
2292: void
2293: pool_allocator_free(struct pool *pp, void *v)
2294: {
2295: struct pool_allocator *pa = pp->pr_alloc;
2296: int s;
2297:
1.91 yamt 2298: LOCK_ASSERT(!simple_lock_held(&pp->pr_slock));
2299:
1.66 thorpej 2300: (*pa->pa_free)(pp, v);
2301:
2302: s = splvm();
2303: simple_lock(&pa->pa_slock);
2304: if ((pa->pa_flags & PA_WANT) == 0) {
2305: simple_unlock(&pa->pa_slock);
2306: splx(s);
2307: return;
2308: }
2309:
2310: TAILQ_FOREACH(pp, &pa->pa_list, pr_alloc_list) {
2311: simple_lock(&pp->pr_slock);
2312: if ((pp->pr_flags & PR_WANTED) != 0) {
2313: pp->pr_flags &= ~PR_WANTED;
2314: wakeup(pp);
2315: }
1.69 thorpej 2316: simple_unlock(&pp->pr_slock);
1.66 thorpej 2317: }
2318: pa->pa_flags &= ~PA_WANT;
2319: simple_unlock(&pa->pa_slock);
2320: splx(s);
2321: }
2322:
2323: void *
2324: pool_page_alloc(struct pool *pp, int flags)
2325: {
2326: boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
2327:
1.100 yamt 2328: return ((void *) uvm_km_alloc_poolpage_cache(kmem_map, waitok));
1.66 thorpej 2329: }
2330:
2331: void
2332: pool_page_free(struct pool *pp, void *v)
2333: {
2334:
1.98 yamt 2335: uvm_km_free_poolpage_cache(kmem_map, (vaddr_t) v);
2336: }
2337:
2338: static void *
2339: pool_page_alloc_meta(struct pool *pp, int flags)
2340: {
2341: boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
2342:
1.100 yamt 2343: return ((void *) uvm_km_alloc_poolpage(kmem_map, waitok));
1.98 yamt 2344: }
2345:
2346: static void
2347: pool_page_free_meta(struct pool *pp, void *v)
2348: {
2349:
1.100 yamt 2350: uvm_km_free_poolpage(kmem_map, (vaddr_t) v);
1.66 thorpej 2351: }
2352:
2353: #ifdef POOL_SUBPAGE
2354: /* Sub-page allocator, for machines with large hardware pages. */
2355: void *
2356: pool_subpage_alloc(struct pool *pp, int flags)
2357: {
1.93 dbj 2358: void *v;
2359: int s;
2360: s = splvm();
2361: v = pool_get(&psppool, flags);
2362: splx(s);
2363: return v;
1.66 thorpej 2364: }
2365:
2366: void
2367: pool_subpage_free(struct pool *pp, void *v)
2368: {
1.93 dbj 2369: int s;
2370: s = splvm();
1.66 thorpej 2371: pool_put(&psppool, v);
1.93 dbj 2372: splx(s);
1.66 thorpej 2373: }
2374:
2375: /* We don't provide a real nointr allocator. Maybe later. */
2376: void *
1.112 bjh21 2377: pool_subpage_alloc_nointr(struct pool *pp, int flags)
1.66 thorpej 2378: {
2379:
2380: return (pool_subpage_alloc(pp, flags));
2381: }
2382:
2383: void
1.112 bjh21 2384: pool_subpage_free_nointr(struct pool *pp, void *v)
1.66 thorpej 2385: {
2386:
2387: pool_subpage_free(pp, v);
2388: }
1.112 bjh21 2389: #endif /* POOL_SUBPAGE */
1.66 thorpej 2390: void *
2391: pool_page_alloc_nointr(struct pool *pp, int flags)
2392: {
2393: boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
2394:
1.100 yamt 2395: return ((void *) uvm_km_alloc_poolpage_cache(kernel_map, waitok));
1.66 thorpej 2396: }
2397:
2398: void
2399: pool_page_free_nointr(struct pool *pp, void *v)
2400: {
2401:
1.98 yamt 2402: uvm_km_free_poolpage_cache(kernel_map, (vaddr_t) v);
1.66 thorpej 2403: }
CVSweb <webmaster@jp.NetBSD.org>