Annotation of src/sys/kern/subr_pool.c, Revision 1.77
1.77 ! matt 1: /* $NetBSD: subr_pool.c,v 1.76 2002/03/13 10:57:18 simonb 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>
1.77 ! matt 41: __KERNEL_RCSID(0, "$NetBSD: subr_pool.c,v 1.76 2002/03/13 10:57:18 simonb 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: *
62: * Memory is allocated in pages which are split into pieces according
63: * to the pool item size. Each page is kept on a list headed by `pr_pagelist'
64: * in the pool structure and the individual pool items are on a linked list
65: * headed by `ph_itemlist' in each page header. The memory for building
66: * the page list is either taken from the allocated pages themselves (for
67: * small pool items) or taken from an internal pool of page headers (`phpool').
1.1 pk 68: */
69:
1.3 pk 70: /* List of all pools */
1.5 thorpej 71: TAILQ_HEAD(,pool) pool_head = TAILQ_HEAD_INITIALIZER(pool_head);
1.3 pk 72:
73: /* Private pool for page header structures */
74: static struct pool phpool;
75:
1.62 bjh21 76: #ifdef POOL_SUBPAGE
77: /* Pool of subpages for use by normal pools. */
78: static struct pool psppool;
79: #endif
80:
1.3 pk 81: /* # of seconds to retain page after last use */
82: int pool_inactive_time = 10;
83:
84: /* Next candidate for drainage (see pool_drain()) */
1.23 thorpej 85: static struct pool *drainpp;
86:
87: /* This spin lock protects both pool_head and drainpp. */
88: struct simplelock pool_head_slock = SIMPLELOCK_INITIALIZER;
1.3 pk 89:
90: struct pool_item_header {
91: /* Page headers */
92: TAILQ_ENTRY(pool_item_header)
93: ph_pagelist; /* pool page list */
94: TAILQ_HEAD(,pool_item) ph_itemlist; /* chunk list for this page */
95: LIST_ENTRY(pool_item_header)
96: ph_hashlist; /* Off-page page headers */
97: int ph_nmissing; /* # of chunks in use */
98: caddr_t ph_page; /* this page's address */
99: struct timeval ph_time; /* last referenced */
100: };
1.61 chs 101: TAILQ_HEAD(pool_pagelist,pool_item_header);
1.3 pk 102:
1.1 pk 103: struct pool_item {
1.3 pk 104: #ifdef DIAGNOSTIC
105: int pi_magic;
1.33 chs 106: #endif
1.25 thorpej 107: #define PI_MAGIC 0xdeadbeef
1.3 pk 108: /* Other entries use only this list entry */
109: TAILQ_ENTRY(pool_item) pi_list;
110: };
111:
1.25 thorpej 112: #define PR_HASH_INDEX(pp,addr) \
1.66 thorpej 113: (((u_long)(addr) >> (pp)->pr_alloc->pa_pageshift) & \
114: (PR_HASHTABSIZE - 1))
1.3 pk 115:
1.53 thorpej 116: #define POOL_NEEDS_CATCHUP(pp) \
117: ((pp)->pr_nitems < (pp)->pr_minitems)
118:
1.43 thorpej 119: /*
120: * Pool cache management.
121: *
122: * Pool caches provide a way for constructed objects to be cached by the
123: * pool subsystem. This can lead to performance improvements by avoiding
124: * needless object construction/destruction; it is deferred until absolutely
125: * necessary.
126: *
127: * Caches are grouped into cache groups. Each cache group references
128: * up to 16 constructed objects. When a cache allocates an object
129: * from the pool, it calls the object's constructor and places it into
130: * a cache group. When a cache group frees an object back to the pool,
131: * it first calls the object's destructor. This allows the object to
132: * persist in constructed form while freed to the cache.
133: *
134: * Multiple caches may exist for each pool. This allows a single
135: * object type to have multiple constructed forms. The pool references
136: * each cache, so that when a pool is drained by the pagedaemon, it can
137: * drain each individual cache as well. Each time a cache is drained,
138: * the most idle cache group is freed to the pool in its entirety.
139: *
140: * Pool caches are layed on top of pools. By layering them, we can avoid
141: * the complexity of cache management for pools which would not benefit
142: * from it.
143: */
144:
145: /* The cache group pool. */
146: static struct pool pcgpool;
1.3 pk 147:
1.43 thorpej 148: static void pool_cache_reclaim(struct pool_cache *);
1.3 pk 149:
1.42 thorpej 150: static int pool_catchup(struct pool *);
1.55 thorpej 151: static void pool_prime_page(struct pool *, caddr_t,
152: struct pool_item_header *);
1.66 thorpej 153:
154: void *pool_allocator_alloc(struct pool *, int);
155: void pool_allocator_free(struct pool *, void *);
1.3 pk 156:
1.42 thorpej 157: static void pool_print1(struct pool *, const char *,
158: void (*)(const char *, ...));
1.3 pk 159:
160: /*
1.52 thorpej 161: * Pool log entry. An array of these is allocated in pool_init().
1.3 pk 162: */
163: struct pool_log {
164: const char *pl_file;
165: long pl_line;
166: int pl_action;
1.25 thorpej 167: #define PRLOG_GET 1
168: #define PRLOG_PUT 2
1.3 pk 169: void *pl_addr;
1.1 pk 170: };
171:
1.3 pk 172: /* Number of entries in pool log buffers */
1.17 thorpej 173: #ifndef POOL_LOGSIZE
174: #define POOL_LOGSIZE 10
175: #endif
176:
177: int pool_logsize = POOL_LOGSIZE;
1.1 pk 178:
1.59 thorpej 179: #ifdef POOL_DIAGNOSTIC
1.42 thorpej 180: static __inline void
181: pr_log(struct pool *pp, void *v, int action, const char *file, long line)
1.3 pk 182: {
183: int n = pp->pr_curlogentry;
184: struct pool_log *pl;
185:
1.20 thorpej 186: if ((pp->pr_roflags & PR_LOGGING) == 0)
1.3 pk 187: return;
188:
189: /*
190: * Fill in the current entry. Wrap around and overwrite
191: * the oldest entry if necessary.
192: */
193: pl = &pp->pr_log[n];
194: pl->pl_file = file;
195: pl->pl_line = line;
196: pl->pl_action = action;
197: pl->pl_addr = v;
198: if (++n >= pp->pr_logsize)
199: n = 0;
200: pp->pr_curlogentry = n;
201: }
202:
203: static void
1.42 thorpej 204: pr_printlog(struct pool *pp, struct pool_item *pi,
205: void (*pr)(const char *, ...))
1.3 pk 206: {
207: int i = pp->pr_logsize;
208: int n = pp->pr_curlogentry;
209:
1.20 thorpej 210: if ((pp->pr_roflags & PR_LOGGING) == 0)
1.3 pk 211: return;
212:
213: /*
214: * Print all entries in this pool's log.
215: */
216: while (i-- > 0) {
217: struct pool_log *pl = &pp->pr_log[n];
218: if (pl->pl_action != 0) {
1.25 thorpej 219: if (pi == NULL || pi == pl->pl_addr) {
220: (*pr)("\tlog entry %d:\n", i);
221: (*pr)("\t\taction = %s, addr = %p\n",
222: pl->pl_action == PRLOG_GET ? "get" : "put",
223: pl->pl_addr);
224: (*pr)("\t\tfile: %s at line %lu\n",
225: pl->pl_file, pl->pl_line);
226: }
1.3 pk 227: }
228: if (++n >= pp->pr_logsize)
229: n = 0;
230: }
231: }
1.25 thorpej 232:
1.42 thorpej 233: static __inline void
234: pr_enter(struct pool *pp, const char *file, long line)
1.25 thorpej 235: {
236:
1.34 thorpej 237: if (__predict_false(pp->pr_entered_file != NULL)) {
1.25 thorpej 238: printf("pool %s: reentrancy at file %s line %ld\n",
239: pp->pr_wchan, file, line);
240: printf(" previous entry at file %s line %ld\n",
241: pp->pr_entered_file, pp->pr_entered_line);
242: panic("pr_enter");
243: }
244:
245: pp->pr_entered_file = file;
246: pp->pr_entered_line = line;
247: }
248:
1.42 thorpej 249: static __inline void
250: pr_leave(struct pool *pp)
1.25 thorpej 251: {
252:
1.34 thorpej 253: if (__predict_false(pp->pr_entered_file == NULL)) {
1.25 thorpej 254: printf("pool %s not entered?\n", pp->pr_wchan);
255: panic("pr_leave");
256: }
257:
258: pp->pr_entered_file = NULL;
259: pp->pr_entered_line = 0;
260: }
261:
1.42 thorpej 262: static __inline void
263: pr_enter_check(struct pool *pp, void (*pr)(const char *, ...))
1.25 thorpej 264: {
265:
266: if (pp->pr_entered_file != NULL)
267: (*pr)("\n\tcurrently entered from file %s line %ld\n",
268: pp->pr_entered_file, pp->pr_entered_line);
269: }
1.3 pk 270: #else
1.25 thorpej 271: #define pr_log(pp, v, action, file, line)
272: #define pr_printlog(pp, pi, pr)
273: #define pr_enter(pp, file, line)
274: #define pr_leave(pp)
275: #define pr_enter_check(pp, pr)
1.59 thorpej 276: #endif /* POOL_DIAGNOSTIC */
1.3 pk 277:
278: /*
279: * Return the pool page header based on page address.
280: */
1.42 thorpej 281: static __inline struct pool_item_header *
282: pr_find_pagehead(struct pool *pp, caddr_t page)
1.3 pk 283: {
284: struct pool_item_header *ph;
285:
1.20 thorpej 286: if ((pp->pr_roflags & PR_PHINPAGE) != 0)
1.3 pk 287: return ((struct pool_item_header *)(page + pp->pr_phoffset));
288:
289: for (ph = LIST_FIRST(&pp->pr_hashtab[PR_HASH_INDEX(pp, page)]);
290: ph != NULL;
291: ph = LIST_NEXT(ph, ph_hashlist)) {
292: if (ph->ph_page == page)
293: return (ph);
294: }
295: return (NULL);
296: }
297:
298: /*
299: * Remove a page from the pool.
300: */
1.42 thorpej 301: static __inline void
1.61 chs 302: pr_rmpage(struct pool *pp, struct pool_item_header *ph,
303: struct pool_pagelist *pq)
1.3 pk 304: {
1.61 chs 305: int s;
1.3 pk 306:
307: /*
1.7 thorpej 308: * If the page was idle, decrement the idle page count.
1.3 pk 309: */
1.6 thorpej 310: if (ph->ph_nmissing == 0) {
311: #ifdef DIAGNOSTIC
312: if (pp->pr_nidle == 0)
313: panic("pr_rmpage: nidle inconsistent");
1.20 thorpej 314: if (pp->pr_nitems < pp->pr_itemsperpage)
315: panic("pr_rmpage: nitems inconsistent");
1.6 thorpej 316: #endif
317: pp->pr_nidle--;
318: }
1.7 thorpej 319:
1.20 thorpej 320: pp->pr_nitems -= pp->pr_itemsperpage;
321:
1.7 thorpej 322: /*
1.61 chs 323: * Unlink a page from the pool and release it (or queue it for release).
1.7 thorpej 324: */
325: TAILQ_REMOVE(&pp->pr_pagelist, ph, ph_pagelist);
1.61 chs 326: if (pq) {
327: TAILQ_INSERT_HEAD(pq, ph, ph_pagelist);
328: } else {
1.66 thorpej 329: pool_allocator_free(pp, ph->ph_page);
1.61 chs 330: if ((pp->pr_roflags & PR_PHINPAGE) == 0) {
331: LIST_REMOVE(ph, ph_hashlist);
332: s = splhigh();
333: pool_put(&phpool, ph);
334: splx(s);
335: }
336: }
1.7 thorpej 337: pp->pr_npages--;
338: pp->pr_npagefree++;
1.6 thorpej 339:
1.3 pk 340: if (pp->pr_curpage == ph) {
341: /*
342: * Find a new non-empty page header, if any.
343: * Start search from the page head, to increase the
344: * chance for "high water" pages to be freed.
345: */
1.61 chs 346: TAILQ_FOREACH(ph, &pp->pr_pagelist, ph_pagelist)
1.3 pk 347: if (TAILQ_FIRST(&ph->ph_itemlist) != NULL)
348: break;
349:
350: pp->pr_curpage = ph;
1.21 thorpej 351: }
1.3 pk 352: }
353:
354: /*
355: * Initialize the given pool resource structure.
356: *
357: * We export this routine to allow other kernel parts to declare
358: * static pools that must be initialized before malloc() is available.
359: */
360: void
1.42 thorpej 361: pool_init(struct pool *pp, size_t size, u_int align, u_int ioff, int flags,
1.66 thorpej 362: const char *wchan, struct pool_allocator *palloc)
1.3 pk 363: {
1.16 briggs 364: int off, slack, i;
1.3 pk 365:
1.25 thorpej 366: #ifdef POOL_DIAGNOSTIC
367: /*
368: * Always log if POOL_DIAGNOSTIC is defined.
369: */
370: if (pool_logsize != 0)
371: flags |= PR_LOGGING;
372: #endif
373:
1.66 thorpej 374: #ifdef POOL_SUBPAGE
375: /*
376: * XXX We don't provide a real `nointr' back-end
377: * yet; all sub-pages come from a kmem back-end.
378: * maybe some day...
379: */
380: if (palloc == NULL) {
381: extern struct pool_allocator pool_allocator_kmem_subpage;
382: palloc = &pool_allocator_kmem_subpage;
383: }
1.3 pk 384: /*
1.66 thorpej 385: * We'll assume any user-specified back-end allocator
386: * will deal with sub-pages, or simply don't care.
1.3 pk 387: */
1.66 thorpej 388: #else
389: if (palloc == NULL)
390: palloc = &pool_allocator_kmem;
391: #endif /* POOL_SUBPAGE */
392: if ((palloc->pa_flags & PA_INITIALIZED) == 0) {
393: if (palloc->pa_pagesz == 0) {
1.62 bjh21 394: #ifdef POOL_SUBPAGE
1.66 thorpej 395: if (palloc == &pool_allocator_kmem)
396: palloc->pa_pagesz = PAGE_SIZE;
397: else
398: palloc->pa_pagesz = POOL_SUBPAGE;
1.62 bjh21 399: #else
1.66 thorpej 400: palloc->pa_pagesz = PAGE_SIZE;
401: #endif /* POOL_SUBPAGE */
402: }
403:
404: TAILQ_INIT(&palloc->pa_list);
405:
406: simple_lock_init(&palloc->pa_slock);
407: palloc->pa_pagemask = ~(palloc->pa_pagesz - 1);
408: palloc->pa_pageshift = ffs(palloc->pa_pagesz) - 1;
409: palloc->pa_flags |= PA_INITIALIZED;
1.4 thorpej 410: }
1.3 pk 411:
412: if (align == 0)
413: align = ALIGN(1);
1.14 thorpej 414:
415: if (size < sizeof(struct pool_item))
416: size = sizeof(struct pool_item);
1.3 pk 417:
1.35 pk 418: size = ALIGN(size);
1.66 thorpej 419: #ifdef DIAGNOSTIC
420: if (size > palloc->pa_pagesz)
1.35 pk 421: panic("pool_init: pool item size (%lu) too large",
422: (u_long)size);
1.66 thorpej 423: #endif
1.35 pk 424:
1.3 pk 425: /*
426: * Initialize the pool structure.
427: */
428: TAILQ_INIT(&pp->pr_pagelist);
1.43 thorpej 429: TAILQ_INIT(&pp->pr_cachelist);
1.3 pk 430: pp->pr_curpage = NULL;
431: pp->pr_npages = 0;
432: pp->pr_minitems = 0;
433: pp->pr_minpages = 0;
434: pp->pr_maxpages = UINT_MAX;
1.20 thorpej 435: pp->pr_roflags = flags;
436: pp->pr_flags = 0;
1.35 pk 437: pp->pr_size = size;
1.3 pk 438: pp->pr_align = align;
439: pp->pr_wchan = wchan;
1.66 thorpej 440: pp->pr_alloc = palloc;
1.20 thorpej 441: pp->pr_nitems = 0;
442: pp->pr_nout = 0;
443: pp->pr_hardlimit = UINT_MAX;
444: pp->pr_hardlimit_warning = NULL;
1.31 thorpej 445: pp->pr_hardlimit_ratecap.tv_sec = 0;
446: pp->pr_hardlimit_ratecap.tv_usec = 0;
447: pp->pr_hardlimit_warning_last.tv_sec = 0;
448: pp->pr_hardlimit_warning_last.tv_usec = 0;
1.68 thorpej 449: pp->pr_drain_hook = NULL;
450: pp->pr_drain_hook_arg = NULL;
1.3 pk 451:
452: /*
453: * Decide whether to put the page header off page to avoid
454: * wasting too large a part of the page. Off-page page headers
455: * go on a hash table, so we can match a returned item
456: * with its header based on the page address.
457: * We use 1/16 of the page size as the threshold (XXX: tune)
458: */
1.66 thorpej 459: if (pp->pr_size < palloc->pa_pagesz/16) {
1.3 pk 460: /* Use the end of the page for the page header */
1.20 thorpej 461: pp->pr_roflags |= PR_PHINPAGE;
1.66 thorpej 462: pp->pr_phoffset = off = palloc->pa_pagesz -
463: ALIGN(sizeof(struct pool_item_header));
1.2 pk 464: } else {
1.3 pk 465: /* The page header will be taken from our page header pool */
466: pp->pr_phoffset = 0;
1.66 thorpej 467: off = palloc->pa_pagesz;
1.16 briggs 468: for (i = 0; i < PR_HASHTABSIZE; i++) {
469: LIST_INIT(&pp->pr_hashtab[i]);
470: }
1.2 pk 471: }
1.1 pk 472:
1.3 pk 473: /*
474: * Alignment is to take place at `ioff' within the item. This means
475: * we must reserve up to `align - 1' bytes on the page to allow
476: * appropriate positioning of each item.
477: *
478: * Silently enforce `0 <= ioff < align'.
479: */
480: pp->pr_itemoffset = ioff = ioff % align;
481: pp->pr_itemsperpage = (off - ((align - ioff) % align)) / pp->pr_size;
1.43 thorpej 482: KASSERT(pp->pr_itemsperpage != 0);
1.3 pk 483:
484: /*
485: * Use the slack between the chunks and the page header
486: * for "cache coloring".
487: */
488: slack = off - pp->pr_itemsperpage * pp->pr_size;
489: pp->pr_maxcolor = (slack / align) * align;
490: pp->pr_curcolor = 0;
491:
492: pp->pr_nget = 0;
493: pp->pr_nfail = 0;
494: pp->pr_nput = 0;
495: pp->pr_npagealloc = 0;
496: pp->pr_npagefree = 0;
1.1 pk 497: pp->pr_hiwat = 0;
1.8 thorpej 498: pp->pr_nidle = 0;
1.3 pk 499:
1.59 thorpej 500: #ifdef POOL_DIAGNOSTIC
1.25 thorpej 501: if (flags & PR_LOGGING) {
502: if (kmem_map == NULL ||
503: (pp->pr_log = malloc(pool_logsize * sizeof(struct pool_log),
504: M_TEMP, M_NOWAIT)) == NULL)
1.20 thorpej 505: pp->pr_roflags &= ~PR_LOGGING;
1.3 pk 506: pp->pr_curlogentry = 0;
507: pp->pr_logsize = pool_logsize;
508: }
1.59 thorpej 509: #endif
1.25 thorpej 510:
511: pp->pr_entered_file = NULL;
512: pp->pr_entered_line = 0;
1.3 pk 513:
1.21 thorpej 514: simple_lock_init(&pp->pr_slock);
1.1 pk 515:
1.3 pk 516: /*
1.43 thorpej 517: * Initialize private page header pool and cache magazine pool if we
518: * haven't done so yet.
1.23 thorpej 519: * XXX LOCKING.
1.3 pk 520: */
521: if (phpool.pr_size == 0) {
1.62 bjh21 522: #ifdef POOL_SUBPAGE
523: pool_init(&phpool, sizeof(struct pool_item_header), 0, 0, 0,
1.66 thorpej 524: "phpool", &pool_allocator_kmem);
1.62 bjh21 525: pool_init(&psppool, POOL_SUBPAGE, POOL_SUBPAGE, 0,
1.66 thorpej 526: PR_RECURSIVE, "psppool", &pool_allocator_kmem);
1.62 bjh21 527: #else
1.3 pk 528: pool_init(&phpool, sizeof(struct pool_item_header), 0, 0,
1.66 thorpej 529: 0, "phpool", NULL);
1.62 bjh21 530: #endif
1.43 thorpej 531: pool_init(&pcgpool, sizeof(struct pool_cache_group), 0, 0,
1.66 thorpej 532: 0, "pcgpool", NULL);
1.1 pk 533: }
534:
1.23 thorpej 535: /* Insert into the list of all pools. */
536: simple_lock(&pool_head_slock);
537: TAILQ_INSERT_TAIL(&pool_head, pp, pr_poollist);
538: simple_unlock(&pool_head_slock);
1.66 thorpej 539:
540: /* Insert this into the list of pools using this allocator. */
541: simple_lock(&palloc->pa_slock);
542: TAILQ_INSERT_TAIL(&palloc->pa_list, pp, pr_alloc_list);
543: simple_unlock(&palloc->pa_slock);
1.1 pk 544: }
545:
546: /*
547: * De-commision a pool resource.
548: */
549: void
1.42 thorpej 550: pool_destroy(struct pool *pp)
1.1 pk 551: {
1.3 pk 552: struct pool_item_header *ph;
1.43 thorpej 553: struct pool_cache *pc;
554:
1.66 thorpej 555: /* Locking order: pool_allocator -> pool */
556: simple_lock(&pp->pr_alloc->pa_slock);
557: TAILQ_REMOVE(&pp->pr_alloc->pa_list, pp, pr_alloc_list);
558: simple_unlock(&pp->pr_alloc->pa_slock);
559:
1.43 thorpej 560: /* Destroy all caches for this pool. */
561: while ((pc = TAILQ_FIRST(&pp->pr_cachelist)) != NULL)
562: pool_cache_destroy(pc);
1.3 pk 563:
564: #ifdef DIAGNOSTIC
1.20 thorpej 565: if (pp->pr_nout != 0) {
1.25 thorpej 566: pr_printlog(pp, NULL, printf);
1.20 thorpej 567: panic("pool_destroy: pool busy: still out: %u\n",
568: pp->pr_nout);
1.3 pk 569: }
570: #endif
1.1 pk 571:
1.3 pk 572: /* Remove all pages */
1.70 thorpej 573: while ((ph = TAILQ_FIRST(&pp->pr_pagelist)) != NULL)
574: pr_rmpage(pp, ph, NULL);
1.3 pk 575:
576: /* Remove from global pool list */
1.23 thorpej 577: simple_lock(&pool_head_slock);
1.3 pk 578: TAILQ_REMOVE(&pool_head, pp, pr_poollist);
1.61 chs 579: if (drainpp == pp) {
580: drainpp = NULL;
581: }
1.23 thorpej 582: simple_unlock(&pool_head_slock);
1.3 pk 583:
1.59 thorpej 584: #ifdef POOL_DIAGNOSTIC
1.20 thorpej 585: if ((pp->pr_roflags & PR_LOGGING) != 0)
1.3 pk 586: free(pp->pr_log, M_TEMP);
1.59 thorpej 587: #endif
1.1 pk 588: }
589:
1.68 thorpej 590: void
591: pool_set_drain_hook(struct pool *pp, void (*fn)(void *, int), void *arg)
592: {
593:
594: /* XXX no locking -- must be used just after pool_init() */
595: #ifdef DIAGNOSTIC
596: if (pp->pr_drain_hook != NULL)
597: panic("pool_set_drain_hook(%s): already set", pp->pr_wchan);
598: #endif
599: pp->pr_drain_hook = fn;
600: pp->pr_drain_hook_arg = arg;
601: }
602:
1.55 thorpej 603: static __inline struct pool_item_header *
604: pool_alloc_item_header(struct pool *pp, caddr_t storage, int flags)
605: {
606: struct pool_item_header *ph;
607: int s;
608:
609: LOCK_ASSERT(simple_lock_held(&pp->pr_slock) == 0);
610:
611: if ((pp->pr_roflags & PR_PHINPAGE) != 0)
612: ph = (struct pool_item_header *) (storage + pp->pr_phoffset);
613: else {
614: s = splhigh();
615: ph = pool_get(&phpool, flags);
616: splx(s);
617: }
618:
619: return (ph);
620: }
1.1 pk 621:
622: /*
1.3 pk 623: * Grab an item from the pool; must be called at appropriate spl level
1.1 pk 624: */
1.3 pk 625: void *
1.59 thorpej 626: #ifdef POOL_DIAGNOSTIC
1.42 thorpej 627: _pool_get(struct pool *pp, int flags, const char *file, long line)
1.56 sommerfe 628: #else
629: pool_get(struct pool *pp, int flags)
630: #endif
1.1 pk 631: {
632: struct pool_item *pi;
1.3 pk 633: struct pool_item_header *ph;
1.55 thorpej 634: void *v;
1.1 pk 635:
1.2 pk 636: #ifdef DIAGNOSTIC
1.37 sommerfe 637: if (__predict_false(curproc == NULL && doing_shutdown == 0 &&
638: (flags & PR_WAITOK) != 0))
1.77 ! matt 639: panic("pool_get: %s: must have NOWAIT", pp->pr_wchan);
1.58 thorpej 640:
641: #ifdef LOCKDEBUG
642: if (flags & PR_WAITOK)
643: simple_lock_only_held(NULL, "pool_get(PR_WAITOK)");
1.56 sommerfe 644: #endif
1.58 thorpej 645: #endif /* DIAGNOSTIC */
1.1 pk 646:
1.21 thorpej 647: simple_lock(&pp->pr_slock);
1.25 thorpej 648: pr_enter(pp, file, line);
1.20 thorpej 649:
650: startover:
651: /*
652: * Check to see if we've reached the hard limit. If we have,
653: * and we can wait, then wait until an item has been returned to
654: * the pool.
655: */
656: #ifdef DIAGNOSTIC
1.34 thorpej 657: if (__predict_false(pp->pr_nout > pp->pr_hardlimit)) {
1.25 thorpej 658: pr_leave(pp);
1.21 thorpej 659: simple_unlock(&pp->pr_slock);
1.20 thorpej 660: panic("pool_get: %s: crossed hard limit", pp->pr_wchan);
661: }
662: #endif
1.34 thorpej 663: if (__predict_false(pp->pr_nout == pp->pr_hardlimit)) {
1.68 thorpej 664: if (pp->pr_drain_hook != NULL) {
665: /*
666: * Since the drain hook is going to free things
667: * back to the pool, unlock, call the hook, re-lock,
668: * and check the hardlimit condition again.
669: */
670: pr_leave(pp);
671: simple_unlock(&pp->pr_slock);
672: (*pp->pr_drain_hook)(pp->pr_drain_hook_arg, flags);
673: simple_lock(&pp->pr_slock);
674: pr_enter(pp, file, line);
675: if (pp->pr_nout < pp->pr_hardlimit)
676: goto startover;
677: }
678:
1.29 sommerfe 679: if ((flags & PR_WAITOK) && !(flags & PR_LIMITFAIL)) {
1.20 thorpej 680: /*
681: * XXX: A warning isn't logged in this case. Should
682: * it be?
683: */
684: pp->pr_flags |= PR_WANTED;
1.25 thorpej 685: pr_leave(pp);
1.40 sommerfe 686: ltsleep(pp, PSWP, pp->pr_wchan, 0, &pp->pr_slock);
1.25 thorpej 687: pr_enter(pp, file, line);
1.20 thorpej 688: goto startover;
689: }
1.31 thorpej 690:
691: /*
692: * Log a message that the hard limit has been hit.
693: */
694: if (pp->pr_hardlimit_warning != NULL &&
695: ratecheck(&pp->pr_hardlimit_warning_last,
696: &pp->pr_hardlimit_ratecap))
697: log(LOG_ERR, "%s\n", pp->pr_hardlimit_warning);
1.21 thorpej 698:
699: pp->pr_nfail++;
700:
1.25 thorpej 701: pr_leave(pp);
1.21 thorpej 702: simple_unlock(&pp->pr_slock);
1.20 thorpej 703: return (NULL);
704: }
705:
1.3 pk 706: /*
707: * The convention we use is that if `curpage' is not NULL, then
708: * it points at a non-empty bucket. In particular, `curpage'
709: * never points at a page header which has PR_PHINPAGE set and
710: * has no items in its bucket.
711: */
1.20 thorpej 712: if ((ph = pp->pr_curpage) == NULL) {
713: #ifdef DIAGNOSTIC
714: if (pp->pr_nitems != 0) {
1.21 thorpej 715: simple_unlock(&pp->pr_slock);
1.20 thorpej 716: printf("pool_get: %s: curpage NULL, nitems %u\n",
717: pp->pr_wchan, pp->pr_nitems);
718: panic("pool_get: nitems inconsistent\n");
719: }
720: #endif
721:
1.21 thorpej 722: /*
723: * Call the back-end page allocator for more memory.
724: * Release the pool lock, as the back-end page allocator
725: * may block.
726: */
1.25 thorpej 727: pr_leave(pp);
1.21 thorpej 728: simple_unlock(&pp->pr_slock);
1.66 thorpej 729: v = pool_allocator_alloc(pp, flags);
1.55 thorpej 730: if (__predict_true(v != NULL))
731: ph = pool_alloc_item_header(pp, v, flags);
1.21 thorpej 732: simple_lock(&pp->pr_slock);
1.25 thorpej 733: pr_enter(pp, file, line);
1.15 pk 734:
1.55 thorpej 735: if (__predict_false(v == NULL || ph == NULL)) {
736: if (v != NULL)
1.66 thorpej 737: pool_allocator_free(pp, v);
1.55 thorpej 738:
1.21 thorpej 739: /*
1.55 thorpej 740: * We were unable to allocate a page or item
741: * header, but we released the lock during
742: * allocation, so perhaps items were freed
743: * back to the pool. Check for this case.
1.21 thorpej 744: */
745: if (pp->pr_curpage != NULL)
746: goto startover;
1.15 pk 747:
1.3 pk 748: if ((flags & PR_WAITOK) == 0) {
749: pp->pr_nfail++;
1.25 thorpej 750: pr_leave(pp);
1.21 thorpej 751: simple_unlock(&pp->pr_slock);
1.1 pk 752: return (NULL);
1.3 pk 753: }
754:
1.15 pk 755: /*
756: * Wait for items to be returned to this pool.
1.21 thorpej 757: *
1.20 thorpej 758: * XXX: maybe we should wake up once a second and
759: * try again?
1.15 pk 760: */
1.1 pk 761: pp->pr_flags |= PR_WANTED;
1.66 thorpej 762: /* PA_WANTED is already set on the allocator. */
1.25 thorpej 763: pr_leave(pp);
1.40 sommerfe 764: ltsleep(pp, PSWP, pp->pr_wchan, 0, &pp->pr_slock);
1.25 thorpej 765: pr_enter(pp, file, line);
1.20 thorpej 766: goto startover;
1.1 pk 767: }
1.3 pk 768:
1.15 pk 769: /* We have more memory; add it to the pool */
1.55 thorpej 770: pool_prime_page(pp, v, ph);
1.15 pk 771: pp->pr_npagealloc++;
772:
1.20 thorpej 773: /* Start the allocation process over. */
774: goto startover;
1.3 pk 775: }
776:
1.34 thorpej 777: if (__predict_false((v = pi = TAILQ_FIRST(&ph->ph_itemlist)) == NULL)) {
1.25 thorpej 778: pr_leave(pp);
1.21 thorpej 779: simple_unlock(&pp->pr_slock);
1.3 pk 780: panic("pool_get: %s: page empty", pp->pr_wchan);
1.21 thorpej 781: }
1.20 thorpej 782: #ifdef DIAGNOSTIC
1.34 thorpej 783: if (__predict_false(pp->pr_nitems == 0)) {
1.25 thorpej 784: pr_leave(pp);
1.21 thorpej 785: simple_unlock(&pp->pr_slock);
1.20 thorpej 786: printf("pool_get: %s: items on itemlist, nitems %u\n",
787: pp->pr_wchan, pp->pr_nitems);
788: panic("pool_get: nitems inconsistent\n");
789: }
1.65 enami 790: #endif
1.56 sommerfe 791:
1.65 enami 792: #ifdef POOL_DIAGNOSTIC
1.3 pk 793: pr_log(pp, v, PRLOG_GET, file, line);
1.65 enami 794: #endif
1.3 pk 795:
1.65 enami 796: #ifdef DIAGNOSTIC
1.34 thorpej 797: if (__predict_false(pi->pi_magic != PI_MAGIC)) {
1.25 thorpej 798: pr_printlog(pp, pi, printf);
1.3 pk 799: panic("pool_get(%s): free list modified: magic=%x; page %p;"
800: " item addr %p\n",
801: pp->pr_wchan, pi->pi_magic, ph->ph_page, pi);
802: }
803: #endif
804:
805: /*
806: * Remove from item list.
807: */
808: TAILQ_REMOVE(&ph->ph_itemlist, pi, pi_list);
1.20 thorpej 809: pp->pr_nitems--;
810: pp->pr_nout++;
1.6 thorpej 811: if (ph->ph_nmissing == 0) {
812: #ifdef DIAGNOSTIC
1.34 thorpej 813: if (__predict_false(pp->pr_nidle == 0))
1.6 thorpej 814: panic("pool_get: nidle inconsistent");
815: #endif
816: pp->pr_nidle--;
817: }
1.3 pk 818: ph->ph_nmissing++;
819: if (TAILQ_FIRST(&ph->ph_itemlist) == NULL) {
1.21 thorpej 820: #ifdef DIAGNOSTIC
1.34 thorpej 821: if (__predict_false(ph->ph_nmissing != pp->pr_itemsperpage)) {
1.25 thorpej 822: pr_leave(pp);
1.21 thorpej 823: simple_unlock(&pp->pr_slock);
824: panic("pool_get: %s: nmissing inconsistent",
825: pp->pr_wchan);
826: }
827: #endif
1.3 pk 828: /*
829: * Find a new non-empty page header, if any.
830: * Start search from the page head, to increase
831: * the chance for "high water" pages to be freed.
832: *
1.21 thorpej 833: * Migrate empty pages to the end of the list. This
834: * will speed the update of curpage as pages become
835: * idle. Empty pages intermingled with idle pages
836: * is no big deal. As soon as a page becomes un-empty,
837: * it will move back to the head of the list.
1.3 pk 838: */
839: TAILQ_REMOVE(&pp->pr_pagelist, ph, ph_pagelist);
1.21 thorpej 840: TAILQ_INSERT_TAIL(&pp->pr_pagelist, ph, ph_pagelist);
1.61 chs 841: TAILQ_FOREACH(ph, &pp->pr_pagelist, ph_pagelist)
1.3 pk 842: if (TAILQ_FIRST(&ph->ph_itemlist) != NULL)
843: break;
844:
845: pp->pr_curpage = ph;
1.1 pk 846: }
1.3 pk 847:
848: pp->pr_nget++;
1.20 thorpej 849:
850: /*
851: * If we have a low water mark and we are now below that low
852: * water mark, add more items to the pool.
853: */
1.53 thorpej 854: if (POOL_NEEDS_CATCHUP(pp) && pool_catchup(pp) != 0) {
1.20 thorpej 855: /*
856: * XXX: Should we log a warning? Should we set up a timeout
857: * to try again in a second or so? The latter could break
858: * a caller's assumptions about interrupt protection, etc.
859: */
860: }
861:
1.25 thorpej 862: pr_leave(pp);
1.21 thorpej 863: simple_unlock(&pp->pr_slock);
1.1 pk 864: return (v);
865: }
866:
867: /*
1.43 thorpej 868: * Internal version of pool_put(). Pool is already locked/entered.
1.1 pk 869: */
1.43 thorpej 870: static void
1.56 sommerfe 871: pool_do_put(struct pool *pp, void *v)
1.1 pk 872: {
873: struct pool_item *pi = v;
1.3 pk 874: struct pool_item_header *ph;
875: caddr_t page;
1.21 thorpej 876: int s;
1.3 pk 877:
1.61 chs 878: LOCK_ASSERT(simple_lock_held(&pp->pr_slock));
879:
1.66 thorpej 880: page = (caddr_t)((u_long)v & pp->pr_alloc->pa_pagemask);
1.1 pk 881:
1.30 thorpej 882: #ifdef DIAGNOSTIC
1.34 thorpej 883: if (__predict_false(pp->pr_nout == 0)) {
1.30 thorpej 884: printf("pool %s: putting with none out\n",
885: pp->pr_wchan);
886: panic("pool_put");
887: }
888: #endif
1.3 pk 889:
1.34 thorpej 890: if (__predict_false((ph = pr_find_pagehead(pp, page)) == NULL)) {
1.25 thorpej 891: pr_printlog(pp, NULL, printf);
1.3 pk 892: panic("pool_put: %s: page header missing", pp->pr_wchan);
893: }
1.28 thorpej 894:
895: #ifdef LOCKDEBUG
896: /*
897: * Check if we're freeing a locked simple lock.
898: */
899: simple_lock_freecheck((caddr_t)pi, ((caddr_t)pi) + pp->pr_size);
900: #endif
1.3 pk 901:
902: /*
903: * Return to item list.
904: */
1.2 pk 905: #ifdef DIAGNOSTIC
1.3 pk 906: pi->pi_magic = PI_MAGIC;
907: #endif
1.32 chs 908: #ifdef DEBUG
909: {
910: int i, *ip = v;
911:
912: for (i = 0; i < pp->pr_size / sizeof(int); i++) {
913: *ip++ = PI_MAGIC;
914: }
915: }
916: #endif
917:
1.3 pk 918: TAILQ_INSERT_HEAD(&ph->ph_itemlist, pi, pi_list);
919: ph->ph_nmissing--;
920: pp->pr_nput++;
1.20 thorpej 921: pp->pr_nitems++;
922: pp->pr_nout--;
1.3 pk 923:
924: /* Cancel "pool empty" condition if it exists */
925: if (pp->pr_curpage == NULL)
926: pp->pr_curpage = ph;
927:
928: if (pp->pr_flags & PR_WANTED) {
929: pp->pr_flags &= ~PR_WANTED;
1.15 pk 930: if (ph->ph_nmissing == 0)
931: pp->pr_nidle++;
1.3 pk 932: wakeup((caddr_t)pp);
933: return;
934: }
935:
936: /*
1.21 thorpej 937: * If this page is now complete, do one of two things:
938: *
939: * (1) If we have more pages than the page high water
940: * mark, free the page back to the system.
941: *
942: * (2) Move it to the end of the page list, so that
943: * we minimize our chances of fragmenting the
944: * pool. Idle pages migrate to the end (along with
945: * completely empty pages, so that we find un-empty
946: * pages more quickly when we update curpage) of the
947: * list so they can be more easily swept up by
948: * the pagedaemon when pages are scarce.
1.3 pk 949: */
950: if (ph->ph_nmissing == 0) {
1.6 thorpej 951: pp->pr_nidle++;
1.71 thorpej 952: if (pp->pr_npages > pp->pr_maxpages ||
953: (pp->pr_alloc->pa_flags & PA_WANT) != 0) {
1.61 chs 954: pr_rmpage(pp, ph, NULL);
1.3 pk 955: } else {
956: TAILQ_REMOVE(&pp->pr_pagelist, ph, ph_pagelist);
957: TAILQ_INSERT_TAIL(&pp->pr_pagelist, ph, ph_pagelist);
958:
1.21 thorpej 959: /*
960: * Update the timestamp on the page. A page must
961: * be idle for some period of time before it can
962: * be reclaimed by the pagedaemon. This minimizes
963: * ping-pong'ing for memory.
964: */
965: s = splclock();
966: ph->ph_time = mono_time;
967: splx(s);
968:
969: /*
970: * Update the current page pointer. Just look for
971: * the first page with any free items.
972: *
973: * XXX: Maybe we want an option to look for the
974: * page with the fewest available items, to minimize
975: * fragmentation?
976: */
1.61 chs 977: TAILQ_FOREACH(ph, &pp->pr_pagelist, ph_pagelist)
1.3 pk 978: if (TAILQ_FIRST(&ph->ph_itemlist) != NULL)
979: break;
1.1 pk 980:
1.3 pk 981: pp->pr_curpage = ph;
1.1 pk 982: }
983: }
1.21 thorpej 984: /*
985: * If the page has just become un-empty, move it to the head of
986: * the list, and make it the current page. The next allocation
987: * will get the item from this page, instead of further fragmenting
988: * the pool.
989: */
990: else if (ph->ph_nmissing == (pp->pr_itemsperpage - 1)) {
991: TAILQ_REMOVE(&pp->pr_pagelist, ph, ph_pagelist);
992: TAILQ_INSERT_HEAD(&pp->pr_pagelist, ph, ph_pagelist);
993: pp->pr_curpage = ph;
994: }
1.43 thorpej 995: }
996:
997: /*
998: * Return resource to the pool; must be called at appropriate spl level
999: */
1.59 thorpej 1000: #ifdef POOL_DIAGNOSTIC
1.43 thorpej 1001: void
1002: _pool_put(struct pool *pp, void *v, const char *file, long line)
1003: {
1004:
1005: simple_lock(&pp->pr_slock);
1006: pr_enter(pp, file, line);
1007:
1.56 sommerfe 1008: pr_log(pp, v, PRLOG_PUT, file, line);
1009:
1010: pool_do_put(pp, v);
1.21 thorpej 1011:
1.25 thorpej 1012: pr_leave(pp);
1.21 thorpej 1013: simple_unlock(&pp->pr_slock);
1.1 pk 1014: }
1.57 sommerfe 1015: #undef pool_put
1.59 thorpej 1016: #endif /* POOL_DIAGNOSTIC */
1.1 pk 1017:
1.56 sommerfe 1018: void
1019: pool_put(struct pool *pp, void *v)
1020: {
1021:
1022: simple_lock(&pp->pr_slock);
1023:
1024: pool_do_put(pp, v);
1025:
1026: simple_unlock(&pp->pr_slock);
1027: }
1.57 sommerfe 1028:
1.59 thorpej 1029: #ifdef POOL_DIAGNOSTIC
1.57 sommerfe 1030: #define pool_put(h, v) _pool_put((h), (v), __FILE__, __LINE__)
1.56 sommerfe 1031: #endif
1.74 thorpej 1032:
1033: /*
1034: * Add N items to the pool.
1035: */
1036: int
1037: pool_prime(struct pool *pp, int n)
1038: {
1039: struct pool_item_header *ph;
1040: caddr_t cp;
1.75 simonb 1041: int newpages;
1.74 thorpej 1042:
1043: simple_lock(&pp->pr_slock);
1044:
1045: newpages = roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1046:
1047: while (newpages-- > 0) {
1048: simple_unlock(&pp->pr_slock);
1049: cp = pool_allocator_alloc(pp, PR_NOWAIT);
1050: if (__predict_true(cp != NULL))
1051: ph = pool_alloc_item_header(pp, cp, PR_NOWAIT);
1052: simple_lock(&pp->pr_slock);
1053:
1054: if (__predict_false(cp == NULL || ph == NULL)) {
1055: if (cp != NULL)
1056: pool_allocator_free(pp, cp);
1057: break;
1058: }
1059:
1060: pool_prime_page(pp, cp, ph);
1061: pp->pr_npagealloc++;
1062: pp->pr_minpages++;
1063: }
1064:
1065: if (pp->pr_minpages >= pp->pr_maxpages)
1066: pp->pr_maxpages = pp->pr_minpages + 1; /* XXX */
1067:
1068: simple_unlock(&pp->pr_slock);
1069: return (0);
1070: }
1.55 thorpej 1071:
1072: /*
1.3 pk 1073: * Add a page worth of items to the pool.
1.21 thorpej 1074: *
1075: * Note, we must be called with the pool descriptor LOCKED.
1.3 pk 1076: */
1.55 thorpej 1077: static void
1078: pool_prime_page(struct pool *pp, caddr_t storage, struct pool_item_header *ph)
1.3 pk 1079: {
1080: struct pool_item *pi;
1081: caddr_t cp = storage;
1082: unsigned int align = pp->pr_align;
1083: unsigned int ioff = pp->pr_itemoffset;
1.55 thorpej 1084: int n;
1.36 pk 1085:
1.66 thorpej 1086: #ifdef DIAGNOSTIC
1087: if (((u_long)cp & (pp->pr_alloc->pa_pagesz - 1)) != 0)
1.36 pk 1088: panic("pool_prime_page: %s: unaligned page", pp->pr_wchan);
1.66 thorpej 1089: #endif
1.3 pk 1090:
1.55 thorpej 1091: if ((pp->pr_roflags & PR_PHINPAGE) == 0)
1.3 pk 1092: LIST_INSERT_HEAD(&pp->pr_hashtab[PR_HASH_INDEX(pp, cp)],
1.55 thorpej 1093: ph, ph_hashlist);
1.3 pk 1094:
1095: /*
1096: * Insert page header.
1097: */
1098: TAILQ_INSERT_HEAD(&pp->pr_pagelist, ph, ph_pagelist);
1099: TAILQ_INIT(&ph->ph_itemlist);
1100: ph->ph_page = storage;
1101: ph->ph_nmissing = 0;
1.21 thorpej 1102: memset(&ph->ph_time, 0, sizeof(ph->ph_time));
1.3 pk 1103:
1.6 thorpej 1104: pp->pr_nidle++;
1105:
1.3 pk 1106: /*
1107: * Color this page.
1108: */
1109: cp = (caddr_t)(cp + pp->pr_curcolor);
1110: if ((pp->pr_curcolor += align) > pp->pr_maxcolor)
1111: pp->pr_curcolor = 0;
1112:
1113: /*
1114: * Adjust storage to apply aligment to `pr_itemoffset' in each item.
1115: */
1116: if (ioff != 0)
1117: cp = (caddr_t)(cp + (align - ioff));
1118:
1119: /*
1120: * Insert remaining chunks on the bucket list.
1121: */
1122: n = pp->pr_itemsperpage;
1.20 thorpej 1123: pp->pr_nitems += n;
1.3 pk 1124:
1125: while (n--) {
1126: pi = (struct pool_item *)cp;
1127:
1128: /* Insert on page list */
1129: TAILQ_INSERT_TAIL(&ph->ph_itemlist, pi, pi_list);
1130: #ifdef DIAGNOSTIC
1131: pi->pi_magic = PI_MAGIC;
1132: #endif
1133: cp = (caddr_t)(cp + pp->pr_size);
1134: }
1135:
1136: /*
1137: * If the pool was depleted, point at the new page.
1138: */
1139: if (pp->pr_curpage == NULL)
1140: pp->pr_curpage = ph;
1141:
1142: if (++pp->pr_npages > pp->pr_hiwat)
1143: pp->pr_hiwat = pp->pr_npages;
1144: }
1145:
1.20 thorpej 1146: /*
1.52 thorpej 1147: * Used by pool_get() when nitems drops below the low water mark. This
1148: * is used to catch up nitmes with the low water mark.
1.20 thorpej 1149: *
1.21 thorpej 1150: * Note 1, we never wait for memory here, we let the caller decide what to do.
1.20 thorpej 1151: *
1.73 thorpej 1152: * Note 2, we must be called with the pool already locked, and we return
1.20 thorpej 1153: * with it locked.
1154: */
1155: static int
1.42 thorpej 1156: pool_catchup(struct pool *pp)
1.20 thorpej 1157: {
1.55 thorpej 1158: struct pool_item_header *ph;
1.20 thorpej 1159: caddr_t cp;
1160: int error = 0;
1161:
1.54 thorpej 1162: while (POOL_NEEDS_CATCHUP(pp)) {
1.20 thorpej 1163: /*
1.21 thorpej 1164: * Call the page back-end allocator for more memory.
1165: *
1166: * XXX: We never wait, so should we bother unlocking
1167: * the pool descriptor?
1.20 thorpej 1168: */
1.21 thorpej 1169: simple_unlock(&pp->pr_slock);
1.66 thorpej 1170: cp = pool_allocator_alloc(pp, PR_NOWAIT);
1.55 thorpej 1171: if (__predict_true(cp != NULL))
1172: ph = pool_alloc_item_header(pp, cp, PR_NOWAIT);
1.21 thorpej 1173: simple_lock(&pp->pr_slock);
1.55 thorpej 1174: if (__predict_false(cp == NULL || ph == NULL)) {
1175: if (cp != NULL)
1.66 thorpej 1176: pool_allocator_free(pp, cp);
1.20 thorpej 1177: error = ENOMEM;
1178: break;
1179: }
1.55 thorpej 1180: pool_prime_page(pp, cp, ph);
1.26 thorpej 1181: pp->pr_npagealloc++;
1.20 thorpej 1182: }
1183:
1184: return (error);
1185: }
1186:
1.3 pk 1187: void
1.42 thorpej 1188: pool_setlowat(struct pool *pp, int n)
1.3 pk 1189: {
1.15 pk 1190:
1.21 thorpej 1191: simple_lock(&pp->pr_slock);
1192:
1.3 pk 1193: pp->pr_minitems = n;
1.15 pk 1194: pp->pr_minpages = (n == 0)
1195: ? 0
1.18 thorpej 1196: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.20 thorpej 1197:
1198: /* Make sure we're caught up with the newly-set low water mark. */
1.75 simonb 1199: if (POOL_NEEDS_CATCHUP(pp) && pool_catchup(pp) != 0) {
1.20 thorpej 1200: /*
1201: * XXX: Should we log a warning? Should we set up a timeout
1202: * to try again in a second or so? The latter could break
1203: * a caller's assumptions about interrupt protection, etc.
1204: */
1205: }
1.21 thorpej 1206:
1207: simple_unlock(&pp->pr_slock);
1.3 pk 1208: }
1209:
1210: void
1.42 thorpej 1211: pool_sethiwat(struct pool *pp, int n)
1.3 pk 1212: {
1.15 pk 1213:
1.21 thorpej 1214: simple_lock(&pp->pr_slock);
1215:
1.15 pk 1216: pp->pr_maxpages = (n == 0)
1217: ? 0
1.18 thorpej 1218: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.21 thorpej 1219:
1220: simple_unlock(&pp->pr_slock);
1.3 pk 1221: }
1222:
1.20 thorpej 1223: void
1.42 thorpej 1224: pool_sethardlimit(struct pool *pp, int n, const char *warnmess, int ratecap)
1.20 thorpej 1225: {
1226:
1.21 thorpej 1227: simple_lock(&pp->pr_slock);
1.20 thorpej 1228:
1229: pp->pr_hardlimit = n;
1230: pp->pr_hardlimit_warning = warnmess;
1.31 thorpej 1231: pp->pr_hardlimit_ratecap.tv_sec = ratecap;
1232: pp->pr_hardlimit_warning_last.tv_sec = 0;
1233: pp->pr_hardlimit_warning_last.tv_usec = 0;
1.20 thorpej 1234:
1235: /*
1.21 thorpej 1236: * In-line version of pool_sethiwat(), because we don't want to
1237: * release the lock.
1.20 thorpej 1238: */
1239: pp->pr_maxpages = (n == 0)
1240: ? 0
1241: : roundup(n, pp->pr_itemsperpage) / pp->pr_itemsperpage;
1.21 thorpej 1242:
1243: simple_unlock(&pp->pr_slock);
1.20 thorpej 1244: }
1.3 pk 1245:
1246: /*
1247: * Release all complete pages that have not been used recently.
1248: */
1.66 thorpej 1249: int
1.59 thorpej 1250: #ifdef POOL_DIAGNOSTIC
1.42 thorpej 1251: _pool_reclaim(struct pool *pp, const char *file, long line)
1.56 sommerfe 1252: #else
1253: pool_reclaim(struct pool *pp)
1254: #endif
1.3 pk 1255: {
1256: struct pool_item_header *ph, *phnext;
1.43 thorpej 1257: struct pool_cache *pc;
1.21 thorpej 1258: struct timeval curtime;
1.61 chs 1259: struct pool_pagelist pq;
1.21 thorpej 1260: int s;
1.3 pk 1261:
1.68 thorpej 1262: if (pp->pr_drain_hook != NULL) {
1263: /*
1264: * The drain hook must be called with the pool unlocked.
1265: */
1266: (*pp->pr_drain_hook)(pp->pr_drain_hook_arg, PR_NOWAIT);
1267: }
1268:
1.21 thorpej 1269: if (simple_lock_try(&pp->pr_slock) == 0)
1.66 thorpej 1270: return (0);
1.25 thorpej 1271: pr_enter(pp, file, line);
1.68 thorpej 1272:
1.61 chs 1273: TAILQ_INIT(&pq);
1.3 pk 1274:
1.43 thorpej 1275: /*
1276: * Reclaim items from the pool's caches.
1277: */
1.61 chs 1278: TAILQ_FOREACH(pc, &pp->pr_cachelist, pc_poollist)
1.43 thorpej 1279: pool_cache_reclaim(pc);
1280:
1.21 thorpej 1281: s = splclock();
1282: curtime = mono_time;
1283: splx(s);
1284:
1.3 pk 1285: for (ph = TAILQ_FIRST(&pp->pr_pagelist); ph != NULL; ph = phnext) {
1286: phnext = TAILQ_NEXT(ph, ph_pagelist);
1287:
1288: /* Check our minimum page claim */
1289: if (pp->pr_npages <= pp->pr_minpages)
1290: break;
1291:
1292: if (ph->ph_nmissing == 0) {
1293: struct timeval diff;
1294: timersub(&curtime, &ph->ph_time, &diff);
1295: if (diff.tv_sec < pool_inactive_time)
1296: continue;
1.21 thorpej 1297:
1298: /*
1299: * If freeing this page would put us below
1300: * the low water mark, stop now.
1301: */
1302: if ((pp->pr_nitems - pp->pr_itemsperpage) <
1303: pp->pr_minitems)
1304: break;
1305:
1.61 chs 1306: pr_rmpage(pp, ph, &pq);
1.3 pk 1307: }
1308: }
1309:
1.25 thorpej 1310: pr_leave(pp);
1.21 thorpej 1311: simple_unlock(&pp->pr_slock);
1.66 thorpej 1312: if (TAILQ_EMPTY(&pq))
1313: return (0);
1314:
1.61 chs 1315: while ((ph = TAILQ_FIRST(&pq)) != NULL) {
1316: TAILQ_REMOVE(&pq, ph, ph_pagelist);
1.66 thorpej 1317: pool_allocator_free(pp, ph->ph_page);
1.61 chs 1318: if (pp->pr_roflags & PR_PHINPAGE) {
1319: continue;
1320: }
1321: LIST_REMOVE(ph, ph_hashlist);
1322: s = splhigh();
1323: pool_put(&phpool, ph);
1324: splx(s);
1325: }
1.66 thorpej 1326:
1327: return (1);
1.3 pk 1328: }
1329:
1330: /*
1331: * Drain pools, one at a time.
1.21 thorpej 1332: *
1333: * Note, we must never be called from an interrupt context.
1.3 pk 1334: */
1335: void
1.42 thorpej 1336: pool_drain(void *arg)
1.3 pk 1337: {
1338: struct pool *pp;
1.23 thorpej 1339: int s;
1.3 pk 1340:
1.61 chs 1341: pp = NULL;
1.49 thorpej 1342: s = splvm();
1.23 thorpej 1343: simple_lock(&pool_head_slock);
1.61 chs 1344: if (drainpp == NULL) {
1345: drainpp = TAILQ_FIRST(&pool_head);
1346: }
1347: if (drainpp) {
1348: pp = drainpp;
1349: drainpp = TAILQ_NEXT(pp, pr_poollist);
1350: }
1351: simple_unlock(&pool_head_slock);
1.63 chs 1352: pool_reclaim(pp);
1.61 chs 1353: splx(s);
1.3 pk 1354: }
1355:
1356: /*
1357: * Diagnostic helpers.
1358: */
1359: void
1.42 thorpej 1360: pool_print(struct pool *pp, const char *modif)
1.21 thorpej 1361: {
1362: int s;
1363:
1.49 thorpej 1364: s = splvm();
1.25 thorpej 1365: if (simple_lock_try(&pp->pr_slock) == 0) {
1366: printf("pool %s is locked; try again later\n",
1367: pp->pr_wchan);
1368: splx(s);
1369: return;
1370: }
1371: pool_print1(pp, modif, printf);
1.21 thorpej 1372: simple_unlock(&pp->pr_slock);
1373: splx(s);
1374: }
1375:
1.25 thorpej 1376: void
1.42 thorpej 1377: pool_printit(struct pool *pp, const char *modif, void (*pr)(const char *, ...))
1.25 thorpej 1378: {
1379: int didlock = 0;
1380:
1381: if (pp == NULL) {
1382: (*pr)("Must specify a pool to print.\n");
1383: return;
1384: }
1385:
1386: /*
1387: * Called from DDB; interrupts should be blocked, and all
1388: * other processors should be paused. We can skip locking
1389: * the pool in this case.
1390: *
1391: * We do a simple_lock_try() just to print the lock
1392: * status, however.
1393: */
1394:
1395: if (simple_lock_try(&pp->pr_slock) == 0)
1396: (*pr)("WARNING: pool %s is locked\n", pp->pr_wchan);
1397: else
1398: didlock = 1;
1399:
1400: pool_print1(pp, modif, pr);
1401:
1402: if (didlock)
1403: simple_unlock(&pp->pr_slock);
1404: }
1405:
1.21 thorpej 1406: static void
1.42 thorpej 1407: pool_print1(struct pool *pp, const char *modif, void (*pr)(const char *, ...))
1.3 pk 1408: {
1.25 thorpej 1409: struct pool_item_header *ph;
1.44 thorpej 1410: struct pool_cache *pc;
1411: struct pool_cache_group *pcg;
1.25 thorpej 1412: #ifdef DIAGNOSTIC
1413: struct pool_item *pi;
1414: #endif
1.44 thorpej 1415: int i, print_log = 0, print_pagelist = 0, print_cache = 0;
1.25 thorpej 1416: char c;
1417:
1418: while ((c = *modif++) != '\0') {
1419: if (c == 'l')
1420: print_log = 1;
1421: if (c == 'p')
1422: print_pagelist = 1;
1.44 thorpej 1423: if (c == 'c')
1424: print_cache = 1;
1.25 thorpej 1425: modif++;
1426: }
1427:
1428: (*pr)("POOL %s: size %u, align %u, ioff %u, roflags 0x%08x\n",
1429: pp->pr_wchan, pp->pr_size, pp->pr_align, pp->pr_itemoffset,
1430: pp->pr_roflags);
1.66 thorpej 1431: (*pr)("\talloc %p\n", pp->pr_alloc);
1.25 thorpej 1432: (*pr)("\tminitems %u, minpages %u, maxpages %u, npages %u\n",
1433: pp->pr_minitems, pp->pr_minpages, pp->pr_maxpages, pp->pr_npages);
1434: (*pr)("\titemsperpage %u, nitems %u, nout %u, hardlimit %u\n",
1435: pp->pr_itemsperpage, pp->pr_nitems, pp->pr_nout, pp->pr_hardlimit);
1436:
1437: (*pr)("\n\tnget %lu, nfail %lu, nput %lu\n",
1438: pp->pr_nget, pp->pr_nfail, pp->pr_nput);
1439: (*pr)("\tnpagealloc %lu, npagefree %lu, hiwat %u, nidle %lu\n",
1440: pp->pr_npagealloc, pp->pr_npagefree, pp->pr_hiwat, pp->pr_nidle);
1441:
1442: if (print_pagelist == 0)
1443: goto skip_pagelist;
1444:
1445: if ((ph = TAILQ_FIRST(&pp->pr_pagelist)) != NULL)
1446: (*pr)("\n\tpage list:\n");
1447: for (; ph != NULL; ph = TAILQ_NEXT(ph, ph_pagelist)) {
1448: (*pr)("\t\tpage %p, nmissing %d, time %lu,%lu\n",
1449: ph->ph_page, ph->ph_nmissing,
1450: (u_long)ph->ph_time.tv_sec,
1451: (u_long)ph->ph_time.tv_usec);
1452: #ifdef DIAGNOSTIC
1.61 chs 1453: TAILQ_FOREACH(pi, &ph->ph_itemlist, pi_list) {
1.25 thorpej 1454: if (pi->pi_magic != PI_MAGIC) {
1455: (*pr)("\t\t\titem %p, magic 0x%x\n",
1456: pi, pi->pi_magic);
1457: }
1458: }
1459: #endif
1460: }
1461: if (pp->pr_curpage == NULL)
1462: (*pr)("\tno current page\n");
1463: else
1464: (*pr)("\tcurpage %p\n", pp->pr_curpage->ph_page);
1465:
1466: skip_pagelist:
1467:
1468: if (print_log == 0)
1469: goto skip_log;
1470:
1471: (*pr)("\n");
1472: if ((pp->pr_roflags & PR_LOGGING) == 0)
1473: (*pr)("\tno log\n");
1474: else
1475: pr_printlog(pp, NULL, pr);
1.3 pk 1476:
1.25 thorpej 1477: skip_log:
1.44 thorpej 1478:
1479: if (print_cache == 0)
1480: goto skip_cache;
1481:
1.61 chs 1482: TAILQ_FOREACH(pc, &pp->pr_cachelist, pc_poollist) {
1.44 thorpej 1483: (*pr)("\tcache %p: allocfrom %p freeto %p\n", pc,
1484: pc->pc_allocfrom, pc->pc_freeto);
1.48 thorpej 1485: (*pr)("\t hits %lu misses %lu ngroups %lu nitems %lu\n",
1486: pc->pc_hits, pc->pc_misses, pc->pc_ngroups, pc->pc_nitems);
1.61 chs 1487: TAILQ_FOREACH(pcg, &pc->pc_grouplist, pcg_list) {
1.44 thorpej 1488: (*pr)("\t\tgroup %p: avail %d\n", pcg, pcg->pcg_avail);
1489: for (i = 0; i < PCG_NOBJECTS; i++)
1490: (*pr)("\t\t\t%p\n", pcg->pcg_objects[i]);
1491: }
1492: }
1493:
1494: skip_cache:
1.3 pk 1495:
1.25 thorpej 1496: pr_enter_check(pp, pr);
1.3 pk 1497: }
1498:
1499: int
1.42 thorpej 1500: pool_chk(struct pool *pp, const char *label)
1.3 pk 1501: {
1502: struct pool_item_header *ph;
1503: int r = 0;
1504:
1.21 thorpej 1505: simple_lock(&pp->pr_slock);
1.3 pk 1506:
1.61 chs 1507: TAILQ_FOREACH(ph, &pp->pr_pagelist, ph_pagelist) {
1.3 pk 1508: struct pool_item *pi;
1509: int n;
1510: caddr_t page;
1511:
1.66 thorpej 1512: page = (caddr_t)((u_long)ph & pp->pr_alloc->pa_pagemask);
1.20 thorpej 1513: if (page != ph->ph_page &&
1514: (pp->pr_roflags & PR_PHINPAGE) != 0) {
1.3 pk 1515: if (label != NULL)
1516: printf("%s: ", label);
1.16 briggs 1517: printf("pool(%p:%s): page inconsistency: page %p;"
1518: " at page head addr %p (p %p)\n", pp,
1.3 pk 1519: pp->pr_wchan, ph->ph_page,
1520: ph, page);
1521: r++;
1522: goto out;
1523: }
1524:
1525: for (pi = TAILQ_FIRST(&ph->ph_itemlist), n = 0;
1526: pi != NULL;
1527: pi = TAILQ_NEXT(pi,pi_list), n++) {
1528:
1529: #ifdef DIAGNOSTIC
1530: if (pi->pi_magic != PI_MAGIC) {
1531: if (label != NULL)
1532: printf("%s: ", label);
1533: printf("pool(%s): free list modified: magic=%x;"
1534: " page %p; item ordinal %d;"
1535: " addr %p (p %p)\n",
1536: pp->pr_wchan, pi->pi_magic, ph->ph_page,
1537: n, pi, page);
1538: panic("pool");
1539: }
1540: #endif
1.66 thorpej 1541: page =
1542: (caddr_t)((u_long)pi & pp->pr_alloc->pa_pagemask);
1.3 pk 1543: if (page == ph->ph_page)
1544: continue;
1545:
1546: if (label != NULL)
1547: printf("%s: ", label);
1.16 briggs 1548: printf("pool(%p:%s): page inconsistency: page %p;"
1549: " item ordinal %d; addr %p (p %p)\n", pp,
1.3 pk 1550: pp->pr_wchan, ph->ph_page,
1551: n, pi, page);
1552: r++;
1553: goto out;
1554: }
1555: }
1556: out:
1.21 thorpej 1557: simple_unlock(&pp->pr_slock);
1.3 pk 1558: return (r);
1.43 thorpej 1559: }
1560:
1561: /*
1562: * pool_cache_init:
1563: *
1564: * Initialize a pool cache.
1565: *
1566: * NOTE: If the pool must be protected from interrupts, we expect
1567: * to be called at the appropriate interrupt priority level.
1568: */
1569: void
1570: pool_cache_init(struct pool_cache *pc, struct pool *pp,
1571: int (*ctor)(void *, void *, int),
1572: void (*dtor)(void *, void *),
1573: void *arg)
1574: {
1575:
1576: TAILQ_INIT(&pc->pc_grouplist);
1577: simple_lock_init(&pc->pc_slock);
1578:
1579: pc->pc_allocfrom = NULL;
1580: pc->pc_freeto = NULL;
1581: pc->pc_pool = pp;
1582:
1583: pc->pc_ctor = ctor;
1584: pc->pc_dtor = dtor;
1585: pc->pc_arg = arg;
1586:
1.48 thorpej 1587: pc->pc_hits = 0;
1588: pc->pc_misses = 0;
1589:
1590: pc->pc_ngroups = 0;
1591:
1592: pc->pc_nitems = 0;
1593:
1.43 thorpej 1594: simple_lock(&pp->pr_slock);
1595: TAILQ_INSERT_TAIL(&pp->pr_cachelist, pc, pc_poollist);
1596: simple_unlock(&pp->pr_slock);
1597: }
1598:
1599: /*
1600: * pool_cache_destroy:
1601: *
1602: * Destroy a pool cache.
1603: */
1604: void
1605: pool_cache_destroy(struct pool_cache *pc)
1606: {
1607: struct pool *pp = pc->pc_pool;
1608:
1609: /* First, invalidate the entire cache. */
1610: pool_cache_invalidate(pc);
1611:
1612: /* ...and remove it from the pool's cache list. */
1613: simple_lock(&pp->pr_slock);
1614: TAILQ_REMOVE(&pp->pr_cachelist, pc, pc_poollist);
1615: simple_unlock(&pp->pr_slock);
1616: }
1617:
1618: static __inline void *
1619: pcg_get(struct pool_cache_group *pcg)
1620: {
1621: void *object;
1622: u_int idx;
1623:
1624: KASSERT(pcg->pcg_avail <= PCG_NOBJECTS);
1.45 thorpej 1625: KASSERT(pcg->pcg_avail != 0);
1.43 thorpej 1626: idx = --pcg->pcg_avail;
1627:
1628: KASSERT(pcg->pcg_objects[idx] != NULL);
1629: object = pcg->pcg_objects[idx];
1630: pcg->pcg_objects[idx] = NULL;
1631:
1632: return (object);
1633: }
1634:
1635: static __inline void
1636: pcg_put(struct pool_cache_group *pcg, void *object)
1637: {
1638: u_int idx;
1639:
1640: KASSERT(pcg->pcg_avail < PCG_NOBJECTS);
1641: idx = pcg->pcg_avail++;
1642:
1643: KASSERT(pcg->pcg_objects[idx] == NULL);
1644: pcg->pcg_objects[idx] = object;
1645: }
1646:
1647: /*
1648: * pool_cache_get:
1649: *
1650: * Get an object from a pool cache.
1651: */
1652: void *
1653: pool_cache_get(struct pool_cache *pc, int flags)
1654: {
1655: struct pool_cache_group *pcg;
1656: void *object;
1.58 thorpej 1657:
1658: #ifdef LOCKDEBUG
1659: if (flags & PR_WAITOK)
1660: simple_lock_only_held(NULL, "pool_cache_get(PR_WAITOK)");
1661: #endif
1.43 thorpej 1662:
1663: simple_lock(&pc->pc_slock);
1664:
1665: if ((pcg = pc->pc_allocfrom) == NULL) {
1.61 chs 1666: TAILQ_FOREACH(pcg, &pc->pc_grouplist, pcg_list) {
1.43 thorpej 1667: if (pcg->pcg_avail != 0) {
1668: pc->pc_allocfrom = pcg;
1669: goto have_group;
1670: }
1671: }
1672:
1673: /*
1674: * No groups with any available objects. Allocate
1675: * a new object, construct it, and return it to
1676: * the caller. We will allocate a group, if necessary,
1677: * when the object is freed back to the cache.
1678: */
1.48 thorpej 1679: pc->pc_misses++;
1.43 thorpej 1680: simple_unlock(&pc->pc_slock);
1681: object = pool_get(pc->pc_pool, flags);
1682: if (object != NULL && pc->pc_ctor != NULL) {
1683: if ((*pc->pc_ctor)(pc->pc_arg, object, flags) != 0) {
1684: pool_put(pc->pc_pool, object);
1685: return (NULL);
1686: }
1687: }
1688: return (object);
1689: }
1690:
1691: have_group:
1.48 thorpej 1692: pc->pc_hits++;
1693: pc->pc_nitems--;
1.43 thorpej 1694: object = pcg_get(pcg);
1695:
1696: if (pcg->pcg_avail == 0)
1697: pc->pc_allocfrom = NULL;
1.45 thorpej 1698:
1.43 thorpej 1699: simple_unlock(&pc->pc_slock);
1700:
1701: return (object);
1702: }
1703:
1704: /*
1705: * pool_cache_put:
1706: *
1707: * Put an object back to the pool cache.
1708: */
1709: void
1710: pool_cache_put(struct pool_cache *pc, void *object)
1711: {
1712: struct pool_cache_group *pcg;
1.60 thorpej 1713: int s;
1.43 thorpej 1714:
1715: simple_lock(&pc->pc_slock);
1716:
1717: if ((pcg = pc->pc_freeto) == NULL) {
1.61 chs 1718: TAILQ_FOREACH(pcg, &pc->pc_grouplist, pcg_list) {
1.43 thorpej 1719: if (pcg->pcg_avail != PCG_NOBJECTS) {
1720: pc->pc_freeto = pcg;
1721: goto have_group;
1722: }
1723: }
1724:
1725: /*
1726: * No empty groups to free the object to. Attempt to
1.47 thorpej 1727: * allocate one.
1.43 thorpej 1728: */
1.47 thorpej 1729: simple_unlock(&pc->pc_slock);
1.60 thorpej 1730: s = splvm();
1.43 thorpej 1731: pcg = pool_get(&pcgpool, PR_NOWAIT);
1.60 thorpej 1732: splx(s);
1.43 thorpej 1733: if (pcg != NULL) {
1734: memset(pcg, 0, sizeof(*pcg));
1.47 thorpej 1735: simple_lock(&pc->pc_slock);
1.48 thorpej 1736: pc->pc_ngroups++;
1.43 thorpej 1737: TAILQ_INSERT_TAIL(&pc->pc_grouplist, pcg, pcg_list);
1.47 thorpej 1738: if (pc->pc_freeto == NULL)
1739: pc->pc_freeto = pcg;
1.43 thorpej 1740: goto have_group;
1741: }
1742:
1743: /*
1744: * Unable to allocate a cache group; destruct the object
1745: * and free it back to the pool.
1746: */
1.51 thorpej 1747: pool_cache_destruct_object(pc, object);
1.43 thorpej 1748: return;
1749: }
1750:
1751: have_group:
1.48 thorpej 1752: pc->pc_nitems++;
1.43 thorpej 1753: pcg_put(pcg, object);
1754:
1755: if (pcg->pcg_avail == PCG_NOBJECTS)
1756: pc->pc_freeto = NULL;
1757:
1758: simple_unlock(&pc->pc_slock);
1.51 thorpej 1759: }
1760:
1761: /*
1762: * pool_cache_destruct_object:
1763: *
1764: * Force destruction of an object and its release back into
1765: * the pool.
1766: */
1767: void
1768: pool_cache_destruct_object(struct pool_cache *pc, void *object)
1769: {
1770:
1771: if (pc->pc_dtor != NULL)
1772: (*pc->pc_dtor)(pc->pc_arg, object);
1773: pool_put(pc->pc_pool, object);
1.43 thorpej 1774: }
1775:
1776: /*
1777: * pool_cache_do_invalidate:
1778: *
1779: * This internal function implements pool_cache_invalidate() and
1780: * pool_cache_reclaim().
1781: */
1782: static void
1783: pool_cache_do_invalidate(struct pool_cache *pc, int free_groups,
1.56 sommerfe 1784: void (*putit)(struct pool *, void *))
1.43 thorpej 1785: {
1786: struct pool_cache_group *pcg, *npcg;
1787: void *object;
1.60 thorpej 1788: int s;
1.43 thorpej 1789:
1790: for (pcg = TAILQ_FIRST(&pc->pc_grouplist); pcg != NULL;
1791: pcg = npcg) {
1792: npcg = TAILQ_NEXT(pcg, pcg_list);
1793: while (pcg->pcg_avail != 0) {
1.48 thorpej 1794: pc->pc_nitems--;
1.43 thorpej 1795: object = pcg_get(pcg);
1.45 thorpej 1796: if (pcg->pcg_avail == 0 && pc->pc_allocfrom == pcg)
1797: pc->pc_allocfrom = NULL;
1.43 thorpej 1798: if (pc->pc_dtor != NULL)
1799: (*pc->pc_dtor)(pc->pc_arg, object);
1.56 sommerfe 1800: (*putit)(pc->pc_pool, object);
1.43 thorpej 1801: }
1802: if (free_groups) {
1.48 thorpej 1803: pc->pc_ngroups--;
1.43 thorpej 1804: TAILQ_REMOVE(&pc->pc_grouplist, pcg, pcg_list);
1.46 thorpej 1805: if (pc->pc_freeto == pcg)
1806: pc->pc_freeto = NULL;
1.60 thorpej 1807: s = splvm();
1.43 thorpej 1808: pool_put(&pcgpool, pcg);
1.60 thorpej 1809: splx(s);
1.43 thorpej 1810: }
1811: }
1812: }
1813:
1814: /*
1815: * pool_cache_invalidate:
1816: *
1817: * Invalidate a pool cache (destruct and release all of the
1818: * cached objects).
1819: */
1820: void
1821: pool_cache_invalidate(struct pool_cache *pc)
1822: {
1823:
1824: simple_lock(&pc->pc_slock);
1.56 sommerfe 1825: pool_cache_do_invalidate(pc, 0, pool_put);
1.43 thorpej 1826: simple_unlock(&pc->pc_slock);
1827: }
1828:
1829: /*
1830: * pool_cache_reclaim:
1831: *
1832: * Reclaim a pool cache for pool_reclaim().
1833: */
1834: static void
1835: pool_cache_reclaim(struct pool_cache *pc)
1836: {
1837:
1.47 thorpej 1838: simple_lock(&pc->pc_slock);
1.43 thorpej 1839: pool_cache_do_invalidate(pc, 1, pool_do_put);
1840: simple_unlock(&pc->pc_slock);
1.3 pk 1841: }
1.66 thorpej 1842:
1843: /*
1844: * Pool backend allocators.
1845: *
1846: * Each pool has a backend allocator that handles allocation, deallocation,
1847: * and any additional draining that might be needed.
1848: *
1849: * We provide two standard allocators:
1850: *
1851: * pool_allocator_kmem - the default when no allocator is specified
1852: *
1853: * pool_allocator_nointr - used for pools that will not be accessed
1854: * in interrupt context.
1855: */
1856: void *pool_page_alloc(struct pool *, int);
1857: void pool_page_free(struct pool *, void *);
1858:
1859: struct pool_allocator pool_allocator_kmem = {
1860: pool_page_alloc, pool_page_free, 0,
1861: };
1862:
1863: void *pool_page_alloc_nointr(struct pool *, int);
1864: void pool_page_free_nointr(struct pool *, void *);
1865:
1866: struct pool_allocator pool_allocator_nointr = {
1867: pool_page_alloc_nointr, pool_page_free_nointr, 0,
1868: };
1869:
1870: #ifdef POOL_SUBPAGE
1871: void *pool_subpage_alloc(struct pool *, int);
1872: void pool_subpage_free(struct pool *, void *);
1873:
1874: struct pool_allocator pool_allocator_kmem_subpage = {
1875: pool_subpage_alloc, pool_subpage_free, 0,
1876: };
1877: #endif /* POOL_SUBPAGE */
1878:
1879: /*
1880: * We have at least three different resources for the same allocation and
1881: * each resource can be depleted. First, we have the ready elements in the
1882: * pool. Then we have the resource (typically a vm_map) for this allocator.
1883: * Finally, we have physical memory. Waiting for any of these can be
1884: * unnecessary when any other is freed, but the kernel doesn't support
1885: * sleeping on multiple wait channels, so we have to employ another strategy.
1886: *
1887: * The caller sleeps on the pool (so that it can be awakened when an item
1888: * is returned to the pool), but we set PA_WANT on the allocator. When a
1889: * page is returned to the allocator and PA_WANT is set, pool_allocator_free
1890: * will wake up all sleeping pools belonging to this allocator.
1891: *
1892: * XXX Thundering herd.
1893: */
1894: void *
1895: pool_allocator_alloc(struct pool *org, int flags)
1896: {
1897: struct pool_allocator *pa = org->pr_alloc;
1898: struct pool *pp, *start;
1899: int s, freed;
1900: void *res;
1901:
1902: do {
1903: if ((res = (*pa->pa_alloc)(org, flags)) != NULL)
1904: return (res);
1.68 thorpej 1905: if ((flags & PR_WAITOK) == 0) {
1906: /*
1907: * We only run the drain hookhere if PR_NOWAIT.
1908: * In other cases, the hook will be run in
1909: * pool_reclaim().
1910: */
1911: if (org->pr_drain_hook != NULL) {
1912: (*org->pr_drain_hook)(org->pr_drain_hook_arg,
1913: flags);
1914: if ((res = (*pa->pa_alloc)(org, flags)) != NULL)
1915: return (res);
1916: }
1.66 thorpej 1917: break;
1.68 thorpej 1918: }
1.66 thorpej 1919:
1920: /*
1921: * Drain all pools, except "org", that use this
1922: * allocator. We do this to reclaim VA space.
1923: * pa_alloc is responsible for waiting for
1924: * physical memory.
1925: *
1926: * XXX We risk looping forever if start if someone
1927: * calls pool_destroy on "start". But there is no
1928: * other way to have potentially sleeping pool_reclaim,
1929: * non-sleeping locks on pool_allocator, and some
1930: * stirring of drained pools in the allocator.
1.68 thorpej 1931: *
1932: * XXX Maybe we should use pool_head_slock for locking
1933: * the allocators?
1.66 thorpej 1934: */
1935: freed = 0;
1936:
1937: s = splvm();
1938: simple_lock(&pa->pa_slock);
1939: pp = start = TAILQ_FIRST(&pa->pa_list);
1940: do {
1941: TAILQ_REMOVE(&pa->pa_list, pp, pr_alloc_list);
1942: TAILQ_INSERT_TAIL(&pa->pa_list, pp, pr_alloc_list);
1943: if (pp == org)
1944: continue;
1.73 thorpej 1945: simple_unlock(&pa->pa_slock);
1.66 thorpej 1946: freed = pool_reclaim(pp);
1.73 thorpej 1947: simple_lock(&pa->pa_slock);
1.66 thorpej 1948: } while ((pp = TAILQ_FIRST(&pa->pa_list)) != start &&
1949: freed == 0);
1950:
1951: if (freed == 0) {
1952: /*
1953: * We set PA_WANT here, the caller will most likely
1954: * sleep waiting for pages (if not, this won't hurt
1955: * that much), and there is no way to set this in
1956: * the caller without violating locking order.
1957: */
1958: pa->pa_flags |= PA_WANT;
1959: }
1960: simple_unlock(&pa->pa_slock);
1961: splx(s);
1962: } while (freed);
1963: return (NULL);
1964: }
1965:
1966: void
1967: pool_allocator_free(struct pool *pp, void *v)
1968: {
1969: struct pool_allocator *pa = pp->pr_alloc;
1970: int s;
1971:
1972: (*pa->pa_free)(pp, v);
1973:
1974: s = splvm();
1975: simple_lock(&pa->pa_slock);
1976: if ((pa->pa_flags & PA_WANT) == 0) {
1977: simple_unlock(&pa->pa_slock);
1978: splx(s);
1979: return;
1980: }
1981:
1982: TAILQ_FOREACH(pp, &pa->pa_list, pr_alloc_list) {
1983: simple_lock(&pp->pr_slock);
1984: if ((pp->pr_flags & PR_WANTED) != 0) {
1985: pp->pr_flags &= ~PR_WANTED;
1986: wakeup(pp);
1987: }
1.69 thorpej 1988: simple_unlock(&pp->pr_slock);
1.66 thorpej 1989: }
1990: pa->pa_flags &= ~PA_WANT;
1991: simple_unlock(&pa->pa_slock);
1992: splx(s);
1993: }
1994:
1995: void *
1996: pool_page_alloc(struct pool *pp, int flags)
1997: {
1998: boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
1999:
2000: return ((void *) uvm_km_alloc_poolpage(waitok));
2001: }
2002:
2003: void
2004: pool_page_free(struct pool *pp, void *v)
2005: {
2006:
2007: uvm_km_free_poolpage((vaddr_t) v);
2008: }
2009:
2010: #ifdef POOL_SUBPAGE
2011: /* Sub-page allocator, for machines with large hardware pages. */
2012: void *
2013: pool_subpage_alloc(struct pool *pp, int flags)
2014: {
2015:
2016: return (pool_get(&psppool, flags));
2017: }
2018:
2019: void
2020: pool_subpage_free(struct pool *pp, void *v)
2021: {
2022:
2023: pool_put(&psppool, v);
2024: }
2025:
2026: /* We don't provide a real nointr allocator. Maybe later. */
2027: void *
2028: pool_page_alloc_nointr(struct pool *pp, int flags)
2029: {
2030:
2031: return (pool_subpage_alloc(pp, flags));
2032: }
2033:
2034: void
2035: pool_page_free_nointr(struct pool *pp, void *v)
2036: {
2037:
2038: pool_subpage_free(pp, v);
2039: }
2040: #else
2041: void *
2042: pool_page_alloc_nointr(struct pool *pp, int flags)
2043: {
2044: boolean_t waitok = (flags & PR_WAITOK) ? TRUE : FALSE;
2045:
2046: return ((void *) uvm_km_alloc_poolpage1(kernel_map,
2047: uvm.kernel_object, waitok));
2048: }
2049:
2050: void
2051: pool_page_free_nointr(struct pool *pp, void *v)
2052: {
2053:
2054: uvm_km_free_poolpage1(kernel_map, (vaddr_t) v);
2055: }
2056: #endif /* POOL_SUBPAGE */
CVSweb <webmaster@jp.NetBSD.org>