2 * Copyright (c) 2011-2013 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 * by Venkatesh Srinivas <vsrinivas@dragonflybsd.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 #include <sys/cdefs.h>
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/types.h>
41 #include <sys/dirent.h>
48 * HAMMER2 offers shared locks and exclusive locks on inodes.
50 * An inode's ip->chain pointer is resolved and stable while an inode is
51 * locked, and can be cleaned out at any time (become NULL) when an inode
54 * The underlying chain is also locked and returned.
56 * NOTE: We don't combine the inode/chain lock because putting away an
57 * inode would otherwise confuse multiple lock holders of the inode.
60 hammer2_inode_lock_ex(hammer2_inode_t *ip)
62 hammer2_chain_t *chain;
64 hammer2_inode_ref(ip);
65 ccms_thread_lock(&ip->topo_cst, CCMS_STATE_EXCLUSIVE);
68 * ip->chain fixup. Certain duplications used to move inodes
69 * into indirect blocks (for example) can cause ip->chain to
74 while (chain->duplink && (chain->flags & HAMMER2_CHAIN_DELETED))
75 chain = chain->duplink;
76 if (ip->chain != chain) {
77 hammer2_chain_ref(chain);
78 hammer2_chain_drop(ip->chain);
82 KKASSERT(chain != NULL); /* for now */
83 hammer2_chain_lock(chain, HAMMER2_RESOLVE_ALWAYS);
86 * Resolve duplication races
88 if (chain->duplink && (chain->flags & HAMMER2_CHAIN_DELETED)) {
89 hammer2_chain_unlock(chain);
96 hammer2_inode_unlock_ex(hammer2_inode_t *ip, hammer2_chain_t *chain)
99 * XXX this will catch parent directories too which we don't
103 hammer2_chain_unlock(chain);
106 * Recalculate ip->chain on exclusive unlock too, it may
107 * allow us to free stale chains more quickly.
109 if ((chain = ip->chain) != NULL) {
110 while (chain->duplink && (chain->flags & HAMMER2_CHAIN_DELETED))
111 chain = chain->duplink;
112 if (ip->chain != chain) {
113 hammer2_chain_ref(chain);
114 hammer2_chain_drop(ip->chain);
119 ccms_thread_unlock(&ip->topo_cst);
120 hammer2_inode_drop(ip);
124 * NOTE: We don't combine the inode/chain lock because putting away an
125 * inode would otherwise confuse multiple lock holders of the inode.
127 * Shared locks are especially sensitive to having too many shared
128 * lock counts (from the same thread) on certain paths which might
129 * need to upgrade them. Only one count of a shared lock can be
133 hammer2_inode_lock_sh(hammer2_inode_t *ip)
135 hammer2_chain_t *chain;
137 hammer2_inode_ref(ip);
139 ccms_thread_lock(&ip->topo_cst, CCMS_STATE_SHARED);
142 KKASSERT(chain != NULL); /* for now */
143 hammer2_chain_lock(chain, HAMMER2_RESOLVE_ALWAYS |
144 HAMMER2_RESOLVE_SHARED);
147 * Resolve duplication races
149 if (chain->duplink && (chain->flags & HAMMER2_CHAIN_DELETED)) {
150 hammer2_chain_unlock(chain);
151 ccms_thread_unlock(&ip->topo_cst);
152 chain = hammer2_inode_lock_ex(ip);
153 hammer2_inode_unlock_ex(ip, chain);
160 hammer2_inode_unlock_sh(hammer2_inode_t *ip, hammer2_chain_t *chain)
163 hammer2_chain_unlock(chain);
164 ccms_thread_unlock(&ip->topo_cst);
165 hammer2_inode_drop(ip);
169 hammer2_inode_lock_temp_release(hammer2_inode_t *ip)
171 return(ccms_thread_lock_temp_release(&ip->topo_cst));
175 hammer2_inode_lock_temp_restore(hammer2_inode_t *ip, ccms_state_t ostate)
177 ccms_thread_lock_temp_restore(&ip->topo_cst, ostate);
181 hammer2_inode_lock_upgrade(hammer2_inode_t *ip)
183 return(ccms_thread_lock_upgrade(&ip->topo_cst));
187 hammer2_inode_lock_downgrade(hammer2_inode_t *ip, ccms_state_t ostate)
189 ccms_thread_lock_downgrade(&ip->topo_cst, ostate);
197 hammer2_mount_exlock(hammer2_mount_t *hmp)
199 ccms_thread_lock(&hmp->vchain.core->cst, CCMS_STATE_EXCLUSIVE);
203 hammer2_mount_shlock(hammer2_mount_t *hmp)
205 ccms_thread_lock(&hmp->vchain.core->cst, CCMS_STATE_SHARED);
209 hammer2_mount_unlock(hammer2_mount_t *hmp)
211 ccms_thread_unlock(&hmp->vchain.core->cst);
215 hammer2_voldata_lock(hammer2_mount_t *hmp)
217 lockmgr(&hmp->voldatalk, LK_EXCLUSIVE);
221 hammer2_voldata_unlock(hammer2_mount_t *hmp, int modify)
224 (hmp->vchain.flags & HAMMER2_CHAIN_MODIFIED) == 0) {
225 atomic_set_int(&hmp->vchain.flags, HAMMER2_CHAIN_MODIFIED);
226 hammer2_chain_ref(&hmp->vchain);
228 lockmgr(&hmp->voldatalk, LK_RELEASE);
232 * Return the directory entry type for an inode.
234 * ip must be locked sh/ex.
237 hammer2_get_dtype(hammer2_chain_t *chain)
241 KKASSERT(chain->bref.type == HAMMER2_BREF_TYPE_INODE);
243 if ((type = chain->data->ipdata.type) == HAMMER2_OBJTYPE_HARDLINK)
244 type = chain->data->ipdata.target_type;
247 case HAMMER2_OBJTYPE_UNKNOWN:
249 case HAMMER2_OBJTYPE_DIRECTORY:
251 case HAMMER2_OBJTYPE_REGFILE:
253 case HAMMER2_OBJTYPE_FIFO:
255 case HAMMER2_OBJTYPE_CDEV: /* not supported */
257 case HAMMER2_OBJTYPE_BDEV: /* not supported */
259 case HAMMER2_OBJTYPE_SOFTLINK:
261 case HAMMER2_OBJTYPE_HARDLINK: /* (never directly associated w/vp) */
263 case HAMMER2_OBJTYPE_SOCKET:
265 case HAMMER2_OBJTYPE_WHITEOUT: /* not supported */
274 * Return the directory entry type for an inode
277 hammer2_get_vtype(hammer2_chain_t *chain)
279 KKASSERT(chain->bref.type == HAMMER2_BREF_TYPE_INODE);
281 switch(chain->data->ipdata.type) {
282 case HAMMER2_OBJTYPE_UNKNOWN:
284 case HAMMER2_OBJTYPE_DIRECTORY:
286 case HAMMER2_OBJTYPE_REGFILE:
288 case HAMMER2_OBJTYPE_FIFO:
290 case HAMMER2_OBJTYPE_CDEV: /* not supported */
292 case HAMMER2_OBJTYPE_BDEV: /* not supported */
294 case HAMMER2_OBJTYPE_SOFTLINK:
296 case HAMMER2_OBJTYPE_HARDLINK: /* XXX */
298 case HAMMER2_OBJTYPE_SOCKET:
300 case HAMMER2_OBJTYPE_WHITEOUT: /* not supported */
309 hammer2_get_obj_type(enum vtype vtype)
313 return(HAMMER2_OBJTYPE_DIRECTORY);
315 return(HAMMER2_OBJTYPE_REGFILE);
317 return(HAMMER2_OBJTYPE_FIFO);
319 return(HAMMER2_OBJTYPE_SOCKET);
321 return(HAMMER2_OBJTYPE_CDEV);
323 return(HAMMER2_OBJTYPE_BDEV);
325 return(HAMMER2_OBJTYPE_SOFTLINK);
327 return(HAMMER2_OBJTYPE_UNKNOWN);
333 * Convert a hammer2 64-bit time to a timespec.
336 hammer2_time_to_timespec(u_int64_t xtime, struct timespec *ts)
338 ts->tv_sec = (unsigned long)(xtime / 1000000);
339 ts->tv_nsec = (unsigned int)(xtime % 1000000) * 1000L;
343 hammer2_timespec_to_time(struct timespec *ts)
347 xtime = (unsigned)(ts->tv_nsec / 1000) +
348 (unsigned long)ts->tv_sec * 1000000ULL;
353 * Convert a uuid to a unix uid or gid
356 hammer2_to_unix_xid(uuid_t *uuid)
358 return(*(u_int32_t *)&uuid->node[2]);
362 hammer2_guid_to_uuid(uuid_t *uuid, u_int32_t guid)
364 bzero(uuid, sizeof(*uuid));
365 *(u_int32_t *)&uuid->node[2] = guid;
369 * Borrow HAMMER1's directory hash algorithm #1 with a few modifications.
370 * The filename is split into fields which are hashed separately and then
373 * Differences include: bit 63 must be set to 1 for HAMMER2 (HAMMER1 sets
374 * it to 0), this is because bit63=0 is used for hidden hardlinked inodes.
375 * (This means we do not need to do a 0-check/or-with-0x100000000 either).
377 * Also, the iscsi crc code is used instead of the old crc32 code.
380 hammer2_dirhash(const unsigned char *name, size_t len)
382 const unsigned char *aname = name;
394 for (i = j = 0; i < len; ++i) {
395 if (aname[i] == '.' ||
400 crcx += hammer2_icrc32(aname + j, i - j);
405 crcx += hammer2_icrc32(aname + j, i - j);
408 * The directory hash utilizes the top 32 bits of the 64-bit key.
409 * Bit 63 must be set to 1.
412 key |= (uint64_t)crcx << 32;
415 * l16 - crc of entire filename
417 * This crc reduces degenerate hash collision conditions
419 crcx = hammer2_icrc32(aname, len);
420 crcx = crcx ^ (crcx << 16);
421 key |= crcx & 0xFFFF0000U;
424 * Set bit 15. This allows readdir to strip bit 63 so a positive
425 * 64-bit cookie/offset can always be returned, and still guarantee
426 * that the values 0x0000-0x7FFF are available for artificial entries.
435 * Return the power-of-2 radix greater or equal to
436 * the specified number of bytes.
438 * Always returns at least the minimum media allocation
439 * size radix, HAMMER2_MIN_RADIX (10), which is 1KB.
442 hammer2_allocsize(size_t bytes)
446 if (bytes < HAMMER2_MIN_ALLOC)
447 bytes = HAMMER2_MIN_ALLOC;
448 if (bytes == HAMMER2_PBUFSIZE)
449 radix = HAMMER2_PBUFRADIX;
450 else if (bytes >= 16384)
452 else if (bytes >= 1024)
455 radix = HAMMER2_MIN_RADIX;
457 while (((size_t)1 << radix) < bytes)
463 * ip must be locked sh/ex
466 hammer2_calc_logical(hammer2_inode_t *ip, hammer2_off_t uoff,
467 hammer2_key_t *lbasep, hammer2_key_t *leofp)
469 hammer2_inode_data_t *ipdata = &ip->chain->data->ipdata;
472 *lbasep = uoff & ~HAMMER2_PBUFMASK64;
473 *leofp = ipdata->size & ~HAMMER2_PBUFMASK64;
474 KKASSERT(*lbasep <= *leofp);
475 if (*lbasep == *leofp /*&& *leofp < 1024 * 1024*/) {
476 radix = hammer2_allocsize((size_t)(ipdata->size - *leofp));
477 if (radix < HAMMER2_MINALLOCRADIX)
478 radix = HAMMER2_MINALLOCRADIX;
479 *leofp += 1U << radix;
480 return (1U << radix);
482 return (HAMMER2_PBUFSIZE);
487 hammer2_update_time(uint64_t *timep)
492 *timep = (unsigned long)tv.tv_sec * 1000000 + tv.tv_usec;