Home | History | Annotate | Line # | Download | only in kern
vfs_dirhash.c revision 1.12
      1 /* $NetBSD: vfs_dirhash.c,v 1.12 2014/09/05 05:57:21 matt Exp $ */
      2 
      3 /*
      4  * Copyright (c) 2008 Reinoud Zandijk
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  *
     16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
     17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
     18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
     19  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
     20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
     21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
     22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
     23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
     24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
     25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     26  *
     27  */
     28 
     29 
     30 #include <sys/cdefs.h>
     31 __KERNEL_RCSID(0, "$NetBSD: vfs_dirhash.c,v 1.12 2014/09/05 05:57:21 matt Exp $");
     32 
     33 /* CLEAN UP! */
     34 #include <sys/param.h>
     35 #include <sys/kernel.h>
     36 #include <sys/buf.h>
     37 #include <sys/dirent.h>
     38 #include <sys/hash.h>
     39 #include <sys/mutex.h>
     40 #include <sys/pool.h>
     41 #include <sys/queue.h>
     42 #include <sys/vnode.h>
     43 #include <sys/sysctl.h>
     44 
     45 #include <sys/dirhash.h>
     46 
     47 #if 1
     48 #	define DPRINTF(a) ;
     49 #else
     50 #	define DPRINTF(a) printf a;
     51 #endif
     52 
     53 /*
     54  * The locking protocol of the dirhash structures is fairly simple:
     55  *
     56  * The global dirhash_queue is protected by the dirhashmutex. This lock is
     57  * internal only and is FS/mountpoint/vnode independent. On exit of the
     58  * exported functions this mutex is not helt.
     59  *
     60  * The dirhash structure is considered part of the vnode/inode/udf_node
     61  * structure and will thus use the lock that protects that vnode/inode.
     62  *
     63  * The dirhash entries are considered part of the dirhash structure and thus
     64  * are on the same lock.
     65  */
     66 
     67 static struct sysctllog *sysctl_log;
     68 static struct pool dirhash_pool;
     69 static struct pool dirhash_entry_pool;
     70 
     71 static kmutex_t dirhashmutex;
     72 static uint32_t maxdirhashsize = DIRHASH_SIZE;
     73 static uint32_t dirhashsize    = 0;
     74 static TAILQ_HEAD(_dirhash, dirhash) dirhash_queue;
     75 
     76 
     77 void
     78 dirhash_init(void)
     79 {
     80 	const struct sysctlnode *rnode, *cnode;
     81 	size_t sz;
     82 	uint32_t max_entries;
     83 
     84 	/* initialise dirhash queue */
     85 	TAILQ_INIT(&dirhash_queue);
     86 
     87 	/* init dirhash pools */
     88 	sz = sizeof(struct dirhash);
     89 	pool_init(&dirhash_pool, sz, 0, 0, 0,
     90 		"dirhpl", NULL, IPL_NONE);
     91 
     92 	sz = sizeof(struct dirhash_entry);
     93 	pool_init(&dirhash_entry_pool, sz, 0, 0, 0,
     94 		"dirhepl", NULL, IPL_NONE);
     95 
     96 	mutex_init(&dirhashmutex, MUTEX_DEFAULT, IPL_NONE);
     97 	max_entries = maxdirhashsize / sz;
     98 	pool_sethiwat(&dirhash_entry_pool, max_entries);
     99 	dirhashsize = 0;
    100 
    101 	/* create sysctl knobs and dials */
    102 	sysctl_log = NULL;
    103 	sysctl_createv(&sysctl_log, 0, NULL, &rnode,
    104 		       CTLFLAG_PERMANENT,
    105 		       CTLTYPE_NODE, "dirhash", NULL,
    106 		       NULL, 0, NULL, 0,
    107 		       CTL_VFS, VFS_GENERIC, CTL_CREATE, CTL_EOL);
    108 	sysctl_createv(&sysctl_log, 0, &rnode, &cnode,
    109 		       CTLFLAG_PERMANENT,
    110 		       CTLTYPE_INT, "memused",
    111 		       SYSCTL_DESCR("current dirhash memory usage"),
    112 		       NULL, 0, &dirhashsize, 0,
    113 		       CTL_CREATE, CTL_EOL);
    114 	sysctl_createv(&sysctl_log, 0, &rnode, &cnode,
    115 		       CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
    116 		       CTLTYPE_INT, "maxmem",
    117 		       SYSCTL_DESCR("maximum dirhash memory usage"),
    118 		       NULL, 0, &maxdirhashsize, 0,
    119 		       CTL_CREATE, CTL_EOL);
    120 }
    121 
    122 
    123 #if 0
    124 void
    125 dirhash_finish(void)
    126 {
    127 	pool_destroy(&dirhash_pool);
    128 	pool_destroy(&dirhash_entry_pool);
    129 
    130 	mutex_destroy(&dirhashmutex);
    131 
    132 	/* sysctl_teardown(&sysctl_log); */
    133 }
    134 #endif
    135 
    136 
    137 /*
    138  * generic dirhash implementation
    139  */
    140 
    141 void
    142 dirhash_purge_entries(struct dirhash *dirh)
    143 {
    144 	struct dirhash_entry *dirh_e;
    145 	uint32_t hashline;
    146 
    147 	if (dirh == NULL)
    148 		return;
    149 
    150 	if (dirh->size == 0)
    151 		return;
    152 
    153 	for (hashline = 0; hashline < DIRHASH_HASHSIZE; hashline++) {
    154 		while ((dirh_e =
    155 		    LIST_FIRST(&dirh->entries[hashline])) != NULL) {
    156 			LIST_REMOVE(dirh_e, next);
    157 			pool_put(&dirhash_entry_pool, dirh_e);
    158 		}
    159 	}
    160 
    161 	while ((dirh_e = LIST_FIRST(&dirh->free_entries)) != NULL) {
    162 		LIST_REMOVE(dirh_e, next);
    163 		pool_put(&dirhash_entry_pool, dirh_e);
    164 	}
    165 
    166 	dirh->flags &= ~DIRH_COMPLETE;
    167 	dirh->flags |=  DIRH_PURGED;
    168 	dirh->num_files = 0;
    169 
    170 	dirhashsize -= dirh->size;
    171 	dirh->size = 0;
    172 }
    173 
    174 
    175 void
    176 dirhash_purge(struct dirhash **dirhp)
    177 {
    178 	struct dirhash *dirh = *dirhp;
    179 
    180 	if (dirh == NULL)
    181 		return;
    182 
    183 	/* purge its entries */
    184 	dirhash_purge_entries(dirh);
    185 
    186 	/* recycle */
    187 	mutex_enter(&dirhashmutex);
    188 	TAILQ_REMOVE(&dirhash_queue, dirh, next);
    189 	mutex_exit(&dirhashmutex);
    190 
    191 	pool_put(&dirhash_pool, dirh);
    192 	*dirhp = NULL;
    193 }
    194 
    195 
    196 void
    197 dirhash_get(struct dirhash **dirhp)
    198 {
    199 	struct dirhash *dirh;
    200 	uint32_t hashline;
    201 
    202 	/* if no dirhash was given, allocate one */
    203 	dirh = *dirhp;
    204 	if (dirh == NULL) {
    205 		dirh = pool_get(&dirhash_pool, PR_WAITOK);
    206 		memset(dirh, 0, sizeof(struct dirhash));
    207 		for (hashline = 0; hashline < DIRHASH_HASHSIZE; hashline++) {
    208 			LIST_INIT(&dirh->entries[hashline]);
    209 		}
    210 	}
    211 
    212 	/* implement LRU on the dirhash queue */
    213 	mutex_enter(&dirhashmutex);
    214 	if (*dirhp) {
    215 		/* remove from queue to be requeued */
    216 		TAILQ_REMOVE(&dirhash_queue, dirh, next);
    217 	}
    218 	dirh->refcnt++;
    219 	TAILQ_INSERT_HEAD(&dirhash_queue, dirh, next);
    220 	mutex_exit(&dirhashmutex);
    221 
    222 	*dirhp = dirh;
    223 }
    224 
    225 
    226 void
    227 dirhash_put(struct dirhash *dirh)
    228 {
    229 
    230 	mutex_enter(&dirhashmutex);
    231 	dirh->refcnt--;
    232 	mutex_exit(&dirhashmutex);
    233 }
    234 
    235 
    236 void
    237 dirhash_enter(struct dirhash *dirh,
    238 	struct dirent *dirent, uint64_t offset, uint32_t entry_size, int new_p)
    239 {
    240 	struct dirhash *del_dirh, *prev_dirh;
    241 	struct dirhash_entry *dirh_e;
    242 	uint32_t hashvalue, hashline;
    243 	int entrysize;
    244 
    245 	/* make sure we have a dirhash to work on */
    246 	KASSERT(dirh);
    247 	KASSERT(dirh->refcnt > 0);
    248 
    249 	/* are we trying to re-enter an entry? */
    250 	if (!new_p && (dirh->flags & DIRH_COMPLETE))
    251 		return;
    252 
    253 	/* calculate our hash */
    254 	hashvalue = hash32_strn(dirent->d_name, dirent->d_namlen, HASH32_STR_INIT);
    255 	hashline  = hashvalue & DIRHASH_HASHMASK;
    256 
    257 	/* lookup and insert entry if not there yet */
    258 	LIST_FOREACH(dirh_e, &dirh->entries[hashline], next) {
    259 		/* check for hash collision */
    260 		if (dirh_e->hashvalue != hashvalue)
    261 			continue;
    262 		if (dirh_e->offset != offset)
    263 			continue;
    264 		/* got it already */
    265 		KASSERT(dirh_e->d_namlen == dirent->d_namlen);
    266 		KASSERT(dirh_e->entry_size == entry_size);
    267 		return;
    268 	}
    269 
    270 	DPRINTF(("dirhash enter %"PRIu64", %d, %d for `%*.*s`\n",
    271 		offset, entry_size, dirent->d_namlen,
    272 		dirent->d_namlen, dirent->d_namlen, dirent->d_name));
    273 
    274 	/* check if entry is in free space list */
    275 	LIST_FOREACH(dirh_e, &dirh->free_entries, next) {
    276 		if (dirh_e->offset == offset) {
    277 			DPRINTF(("\tremoving free entry\n"));
    278 			LIST_REMOVE(dirh_e, next);
    279 			pool_put(&dirhash_entry_pool, dirh_e);
    280 			break;
    281 		}
    282 	}
    283 
    284 	/* ensure we are not passing the dirhash limit */
    285 	entrysize = sizeof(struct dirhash_entry);
    286 	if (dirhashsize + entrysize > maxdirhashsize) {
    287 		/* we walk the dirhash_queue, so need to lock it */
    288 		mutex_enter(&dirhashmutex);
    289 		del_dirh = TAILQ_LAST(&dirhash_queue, _dirhash);
    290 		KASSERT(del_dirh);
    291 		while (dirhashsize + entrysize > maxdirhashsize) {
    292 			/* no use trying to delete myself */
    293 			if (del_dirh == dirh)
    294 				break;
    295 			prev_dirh = TAILQ_PREV(del_dirh, _dirhash, next);
    296 			if (del_dirh->refcnt == 0)
    297 				dirhash_purge_entries(del_dirh);
    298 			del_dirh = prev_dirh;
    299 		}
    300 		mutex_exit(&dirhashmutex);
    301 	}
    302 
    303 	/* add to the hashline */
    304 	dirh_e = pool_get(&dirhash_entry_pool, PR_WAITOK);
    305 	memset(dirh_e, 0, sizeof(struct dirhash_entry));
    306 
    307 	dirh_e->hashvalue = hashvalue;
    308 	dirh_e->offset    = offset;
    309 	dirh_e->d_namlen  = dirent->d_namlen;
    310 	dirh_e->entry_size  = entry_size;
    311 
    312 	dirh->size  += sizeof(struct dirhash_entry);
    313 	dirh->num_files++;
    314 	dirhashsize += sizeof(struct dirhash_entry);
    315 	LIST_INSERT_HEAD(&dirh->entries[hashline], dirh_e, next);
    316 }
    317 
    318 
    319 void
    320 dirhash_enter_freed(struct dirhash *dirh, uint64_t offset,
    321 	uint32_t entry_size)
    322 {
    323 	struct dirhash_entry *dirh_e;
    324 
    325 	/* make sure we have a dirhash to work on */
    326 	KASSERT(dirh);
    327 	KASSERT(dirh->refcnt > 0);
    328 
    329 	/* check for double entry of free space */
    330 	LIST_FOREACH(dirh_e, &dirh->free_entries, next) {
    331 		KASSERT(dirh_e->offset != offset);
    332 	}
    333 
    334 	DPRINTF(("dirhash enter FREED %"PRIu64", %d\n",
    335 		offset, entry_size));
    336 	dirh_e = pool_get(&dirhash_entry_pool, PR_WAITOK);
    337 	memset(dirh_e, 0, sizeof(struct dirhash_entry));
    338 
    339 	dirh_e->hashvalue = 0;		/* not relevant */
    340 	dirh_e->offset    = offset;
    341 	dirh_e->d_namlen  = 0;		/* not relevant */
    342 	dirh_e->entry_size  = entry_size;
    343 
    344 	/* XXX it might be preferable to append them at the tail */
    345 	LIST_INSERT_HEAD(&dirh->free_entries, dirh_e, next);
    346 	dirh->size  += sizeof(struct dirhash_entry);
    347 	dirhashsize += sizeof(struct dirhash_entry);
    348 }
    349 
    350 
    351 void
    352 dirhash_remove(struct dirhash *dirh, struct dirent *dirent,
    353 	uint64_t offset, uint32_t entry_size)
    354 {
    355 	struct dirhash_entry *dirh_e;
    356 	uint32_t hashvalue, hashline;
    357 
    358 	DPRINTF(("dirhash remove %"PRIu64", %d for `%*.*s`\n",
    359 		offset, entry_size,
    360 		dirent->d_namlen, dirent->d_namlen, dirent->d_name));
    361 
    362 	/* make sure we have a dirhash to work on */
    363 	KASSERT(dirh);
    364 	KASSERT(dirh->refcnt > 0);
    365 
    366 	/* calculate our hash */
    367 	hashvalue = hash32_strn(dirent->d_name, dirent->d_namlen, HASH32_STR_INIT);
    368 	hashline  = hashvalue & DIRHASH_HASHMASK;
    369 
    370 	/* lookup entry */
    371 	LIST_FOREACH(dirh_e, &dirh->entries[hashline], next) {
    372 		/* check for hash collision */
    373 		if (dirh_e->hashvalue != hashvalue)
    374 			continue;
    375 		if (dirh_e->offset != offset)
    376 			continue;
    377 
    378 		/* got it! */
    379 		KASSERT(dirh_e->d_namlen == dirent->d_namlen);
    380 		KASSERT(dirh_e->entry_size == entry_size);
    381 		LIST_REMOVE(dirh_e, next);
    382 		dirh->size -= sizeof(struct dirhash_entry);
    383 		KASSERT(dirh->num_files > 0);
    384 		dirh->num_files--;
    385 		dirhashsize -= sizeof(struct dirhash_entry);
    386 
    387 		dirhash_enter_freed(dirh, offset, entry_size);
    388 		return;
    389 	}
    390 
    391 	/* not found! */
    392 	panic("dirhash_remove couldn't find entry in hash table\n");
    393 }
    394 
    395 
    396 /*
    397  * BUGALERT: don't use result longer than needed, never past the node lock.
    398  * Call with NULL *result initially and it will return nonzero if again.
    399  */
    400 int
    401 dirhash_lookup(struct dirhash *dirh, const char *d_name, int d_namlen,
    402 	struct dirhash_entry **result)
    403 {
    404 	struct dirhash_entry *dirh_e;
    405 	uint32_t hashvalue, hashline;
    406 
    407 	/* make sure we have a dirhash to work on */
    408 	KASSERT(dirh);
    409 	KASSERT(dirh->refcnt > 0);
    410 
    411 	/* start where we were */
    412 	if (*result) {
    413 		dirh_e = *result;
    414 
    415 		/* retrieve information to avoid recalculation and advance */
    416 		hashvalue = dirh_e->hashvalue;
    417 		dirh_e = LIST_NEXT(*result, next);
    418 	} else {
    419 		/* calculate our hash and lookup all entries in hashline */
    420 		hashvalue = hash32_strn(d_name, d_namlen, HASH32_STR_INIT);
    421 		hashline  = hashvalue & DIRHASH_HASHMASK;
    422 		dirh_e = LIST_FIRST(&dirh->entries[hashline]);
    423 	}
    424 
    425 	for (; dirh_e; dirh_e = LIST_NEXT(dirh_e, next)) {
    426 		/* check for hash collision */
    427 		if (dirh_e->hashvalue != hashvalue)
    428 			continue;
    429 		if (dirh_e->d_namlen != d_namlen)
    430 			continue;
    431 		/* might have an entry in the cache */
    432 		*result = dirh_e;
    433 		return 1;
    434 	}
    435 
    436 	*result = NULL;
    437 	return 0;
    438 }
    439 
    440 
    441 /*
    442  * BUGALERT: don't use result longer than needed, never past the node lock.
    443  * Call with NULL *result initially and it will return nonzero if again.
    444  */
    445 
    446 int
    447 dirhash_lookup_freed(struct dirhash *dirh, uint32_t min_entrysize,
    448 	struct dirhash_entry **result)
    449 {
    450 	struct dirhash_entry *dirh_e;
    451 
    452 	/* make sure we have a dirhash to work on */
    453 	KASSERT(dirh);
    454 	KASSERT(dirh->refcnt > 0);
    455 
    456 	/* start where we were */
    457 	if (*result) {
    458 		dirh_e = LIST_NEXT(*result, next);
    459 	} else {
    460 		/* lookup all entries that match */
    461 		dirh_e = LIST_FIRST(&dirh->free_entries);
    462 	}
    463 
    464 	for (; dirh_e; dirh_e = LIST_NEXT(dirh_e, next)) {
    465 		/* check for minimum size */
    466 		if (dirh_e->entry_size < min_entrysize)
    467 			continue;
    468 		/* might be a candidate */
    469 		*result = dirh_e;
    470 		return 1;
    471 	}
    472 
    473 	*result = NULL;
    474 	return 0;
    475 }
    476 
    477 
    478 bool
    479 dirhash_dir_isempty(struct dirhash *dirh)
    480 {
    481 #ifdef DEBUG
    482 	struct dirhash_entry *dirh_e;
    483 	int hashline, num;
    484 
    485 	num = 0;
    486 	for (hashline = 0; hashline < DIRHASH_HASHSIZE; hashline++) {
    487 		LIST_FOREACH(dirh_e, &dirh->entries[hashline], next) {
    488 			num++;
    489 		}
    490 	}
    491 
    492 	if (dirh->num_files != num) {
    493 		printf("dirhash_dir_isempy: dirhash_counter failed: "
    494 			"dirh->num_files = %d, counted %d\n",
    495 			dirh->num_files, num);
    496 		assert(dirh->num_files == num);
    497 	}
    498 #endif
    499 	/* assert the directory hash info is valid */
    500 	KASSERT(dirh->flags & DIRH_COMPLETE);
    501 
    502 	/* the directory is empty when only '..' lifes in it or is absent */
    503 	return (dirh->num_files <= 1);
    504 }
    505 
    506