2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 static u_int32_t ocp_allocbit(hammer_objid_cache_t ocp, u_int32_t n);
41 * Start a standard transaction.
44 hammer_start_transaction(struct hammer_transaction *trans,
45 struct hammer_mount *hmp)
50 trans->type = HAMMER_TRANS_STD;
52 trans->rootvol = hammer_get_root_volume(hmp, &error);
55 trans->sync_lock_refs = 0;
59 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
60 trans->time32 = (u_int32_t)tv.tv_sec;
64 * Start a simple read-only transaction. This will not stall.
67 hammer_simple_transaction(struct hammer_transaction *trans,
68 struct hammer_mount *hmp)
73 trans->type = HAMMER_TRANS_RO;
75 trans->rootvol = hammer_get_root_volume(hmp, &error);
78 trans->sync_lock_refs = 0;
82 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
83 trans->time32 = (u_int32_t)tv.tv_sec;
87 * Start a transaction using a particular TID. Used by the sync code.
88 * This does not stall.
90 * This routine may only be called from the flusher thread. We predispose
91 * sync_lock_refs, implying serialization against the synchronization stage
92 * (which the flusher is responsible for).
95 hammer_start_transaction_fls(struct hammer_transaction *trans,
96 struct hammer_mount *hmp)
101 bzero(trans, sizeof(*trans));
103 trans->type = HAMMER_TRANS_FLS;
105 trans->rootvol = hammer_get_root_volume(hmp, &error);
106 KKASSERT(error == 0);
107 trans->tid = hammer_alloc_tid(hmp, 1);
108 trans->sync_lock_refs = 1;
112 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
113 trans->time32 = (u_int32_t)tv.tv_sec;
117 hammer_done_transaction(struct hammer_transaction *trans)
119 int expected_lock_refs __debugvar;
121 hammer_rel_volume(trans->rootvol, 0);
122 trans->rootvol = NULL;
123 expected_lock_refs = (trans->type == HAMMER_TRANS_FLS) ? 1 : 0;
124 KKASSERT(trans->sync_lock_refs == expected_lock_refs);
125 trans->sync_lock_refs = 0;
126 if (trans->type != HAMMER_TRANS_FLS) {
127 if (trans->flags & HAMMER_TRANSF_NEWINODE)
128 hammer_inode_waitreclaims(trans);
130 else if (trans->flags & HAMMER_TRANSF_DIDIO)
131 hammer_inode_waitreclaims(trans);
137 * Allocate (count) TIDs. If running in multi-master mode the returned
138 * base will be aligned to a 16-count plus the master id (0-15).
139 * Multi-master mode allows non-conflicting to run and new objects to be
140 * created on multiple masters in parallel. The transaction id identifies
141 * the original master. The object_id is also subject to this rule in
142 * order to allow objects to be created on multiple masters in parallel.
144 * Directories may pre-allocate a large number of object ids (100,000).
146 * NOTE: There is no longer a requirement that successive transaction
147 * ids be 2 apart for separator generation.
149 * NOTE: When called by pseudo-backends such as ioctls the allocated
150 * TID will be larger then the current flush TID, if a flush is running,
151 * so any mirroring will pick the records up on a later flush.
154 hammer_alloc_tid(hammer_mount_t hmp, int count)
158 if (hmp->master_id < 0) {
159 tid = hmp->next_tid + 1;
160 hmp->next_tid = tid + count;
162 tid = (hmp->next_tid + HAMMER_MAX_MASTERS) &
163 ~(hammer_tid_t)(HAMMER_MAX_MASTERS - 1);
164 hmp->next_tid = tid + count * HAMMER_MAX_MASTERS;
165 tid |= hmp->master_id;
167 if (tid >= 0xFFFFFFFFFF000000ULL)
168 panic("hammer_start_transaction: Ran out of TIDs!");
169 if (hammer_debug_tid)
170 kprintf("alloc_tid %016llx\n", (long long)tid);
175 * Allocate an object id.
177 * We use the upper OBJID_CACHE_BITS bits of the namekey to try to match
178 * the low bits of the objid we allocate.
181 hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip, int64_t namekey)
183 hammer_objid_cache_t ocp;
187 while ((ocp = dip->objid_cache) == NULL) {
188 if (hmp->objid_cache_count < OBJID_CACHE_SIZE) {
189 ocp = kmalloc(sizeof(*ocp), hmp->m_misc,
191 ocp->base_tid = hammer_alloc_tid(hmp,
192 OBJID_CACHE_BULK * 2);
193 ocp->base_tid += OBJID_CACHE_BULK_MASK64;
194 ocp->base_tid &= ~OBJID_CACHE_BULK_MASK64;
195 /* may have blocked, recheck */
196 if (dip->objid_cache == NULL) {
197 TAILQ_INSERT_TAIL(&hmp->objid_cache_list,
199 ++hmp->objid_cache_count;
200 dip->objid_cache = ocp;
203 kfree(ocp, hmp->m_misc);
207 * Steal one from another directory?
209 * Throw away ocp's that are more then half full, they
210 * aren't worth stealing.
212 ocp = TAILQ_FIRST(&hmp->objid_cache_list);
214 ocp->dip->objid_cache = NULL;
215 if (ocp->count >= OBJID_CACHE_BULK / 2) {
216 TAILQ_REMOVE(&hmp->objid_cache_list,
218 --hmp->objid_cache_count;
219 kfree(ocp, hmp->m_misc);
221 dip->objid_cache = ocp;
226 TAILQ_REMOVE(&hmp->objid_cache_list, ocp, entry);
229 * Allocate inode numbers uniformly.
232 n = (namekey >> (63 - OBJID_CACHE_BULK_BITS)) & OBJID_CACHE_BULK_MASK;
233 n = ocp_allocbit(ocp, n);
234 tid = ocp->base_tid + n;
238 * The TID is incremented by 1 or by 16 depending what mode the
239 * mount is operating in.
241 ocp->next_tid += (hmp->master_id < 0) ? 1 : HAMMER_MAX_MASTERS;
243 if (ocp->count >= OBJID_CACHE_BULK * 3 / 4) {
244 dip->objid_cache = NULL;
245 --hmp->objid_cache_count;
247 kfree(ocp, hmp->m_misc);
249 TAILQ_INSERT_TAIL(&hmp->objid_cache_list, ocp, entry);
255 * Allocate a bit starting with bit n. Wrap if necessary.
257 * This routine is only ever called if a bit is available somewhere
261 ocp_allocbit(hammer_objid_cache_t ocp, u_int32_t n)
268 while (ocp->bm1[n0] & (1 << n)) {
269 if (ocp->bm0 & (1 << n0)) {
272 } else if (++n == 32) {
278 ocp->bm1[n0] |= 1 << n;
279 if (ocp->bm1[n0] == 0xFFFFFFFFU)
281 return((n0 << 5) + n);
285 hammer_clear_objid(hammer_inode_t dip)
287 hammer_objid_cache_t ocp;
289 if ((ocp = dip->objid_cache) != NULL) {
290 dip->objid_cache = NULL;
292 TAILQ_REMOVE(&dip->hmp->objid_cache_list, ocp, entry);
293 TAILQ_INSERT_HEAD(&dip->hmp->objid_cache_list, ocp, entry);
298 hammer_destroy_objid_cache(hammer_mount_t hmp)
300 hammer_objid_cache_t ocp;
302 while ((ocp = TAILQ_FIRST(&hmp->objid_cache_list)) != NULL) {
303 TAILQ_REMOVE(&hmp->objid_cache_list, ocp, entry);
305 ocp->dip->objid_cache = NULL;
306 kfree(ocp, hmp->m_misc);
307 --hmp->objid_cache_count;
309 KKASSERT(hmp->objid_cache_count == 0);