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 uint32_t ocp_allocbit(hammer_objid_cache_t ocp, uint32_t n);
41 * Start a standard transaction.
43 * May be called without fs_token
46 hammer_start_transaction(hammer_transaction_t trans, hammer_mount_t hmp)
51 trans->type = HAMMER_TRANS_STD;
53 trans->rootvol = hammer_get_root_volume(hmp, &error);
56 trans->sync_lock_refs = 0;
60 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
61 trans->time32 = (uint32_t)tv.tv_sec;
65 * Start a simple read-only transaction. This will not stall.
67 * May be called without fs_token
70 hammer_simple_transaction(hammer_transaction_t trans, hammer_mount_t hmp)
75 trans->type = HAMMER_TRANS_RO;
77 trans->rootvol = hammer_get_root_volume(hmp, &error);
80 trans->sync_lock_refs = 0;
84 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
85 trans->time32 = (uint32_t)tv.tv_sec;
89 * Start a transaction using a particular TID. Used by the sync code.
90 * This does not stall.
92 * This routine may only be called from the flusher thread. We predispose
93 * sync_lock_refs, implying serialization against the synchronization stage
94 * (which the flusher is responsible for).
97 hammer_start_transaction_fls(hammer_transaction_t trans, hammer_mount_t hmp)
102 bzero(trans, sizeof(*trans));
104 trans->type = HAMMER_TRANS_FLS;
106 trans->rootvol = hammer_get_root_volume(hmp, &error);
107 KKASSERT(error == 0);
108 trans->tid = hammer_alloc_tid(hmp, 1);
109 trans->sync_lock_refs = 1;
113 trans->time = (unsigned long)tv.tv_sec * 1000000ULL + tv.tv_usec;
114 trans->time32 = (uint32_t)tv.tv_sec;
118 * May be called without fs_token
121 hammer_done_transaction(hammer_transaction_t trans)
123 int expected_lock_refs __debugvar;
125 hammer_rel_volume(trans->rootvol, 0);
126 trans->rootvol = NULL;
127 expected_lock_refs = (trans->type == HAMMER_TRANS_FLS) ? 1 : 0;
128 KKASSERT(trans->sync_lock_refs == expected_lock_refs);
129 trans->sync_lock_refs = 0;
130 if (trans->type != HAMMER_TRANS_FLS) {
131 if (trans->flags & HAMMER_TRANSF_NEWINODE) {
132 lwkt_gettoken(&trans->hmp->fs_token);
133 hammer_inode_waitreclaims(trans);
134 lwkt_reltoken(&trans->hmp->fs_token);
140 * Allocate (count) TIDs. If running in multi-master mode the returned
141 * base will be aligned to a 16-count plus the master id (0-15).
142 * Multi-master mode allows non-conflicting to run and new objects to be
143 * created on multiple masters in parallel. The transaction id identifies
144 * the original master. The object_id is also subject to this rule in
145 * order to allow objects to be created on multiple masters in parallel.
147 * Directories may pre-allocate a large number of object ids (100,000).
149 * NOTE: There is no longer a requirement that successive transaction
150 * ids be 2 apart for separator generation.
152 * NOTE: When called by pseudo-backends such as ioctls the allocated
153 * TID will be larger then the current flush TID, if a flush is running,
154 * so any mirroring will pick the records up on a later flush.
156 * NOTE: HAMMER1 does not support multi-master clustering as of 2015.
159 hammer_alloc_tid(hammer_mount_t hmp, int count)
163 if (hmp->master_id < 0) {
164 tid = hmp->next_tid + 1;
165 hmp->next_tid = tid + count;
167 tid = (hmp->next_tid + HAMMER_MAX_MASTERS) &
168 ~(hammer_tid_t)(HAMMER_MAX_MASTERS - 1);
169 hmp->next_tid = tid + count * HAMMER_MAX_MASTERS;
170 tid |= hmp->master_id;
172 if (tid >= 0xFFFFFFFFFF000000ULL)
173 hpanic("Ran out of TIDs!");
174 if (hammer_debug_tid)
175 hdkprintf("%016jx\n", (intmax_t)tid);
180 * Allocate an object id.
182 * We use the upper OBJID_CACHE_BITS bits of the namekey to try to match
183 * the low bits of the objid we allocate.
186 hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip, int64_t namekey)
188 hammer_objid_cache_t ocp;
192 while ((ocp = dip->objid_cache) == NULL) {
193 if (hmp->objid_cache_count < OBJID_CACHE_SIZE) {
194 ocp = kmalloc(sizeof(*ocp), hmp->m_misc,
196 ocp->base_tid = hammer_alloc_tid(hmp,
197 OBJID_CACHE_BULK * 2);
198 ocp->base_tid += OBJID_CACHE_BULK_MASK64;
199 ocp->base_tid &= ~OBJID_CACHE_BULK_MASK64;
200 /* may have blocked, recheck */
201 if (dip->objid_cache == NULL) {
202 TAILQ_INSERT_TAIL(&hmp->objid_cache_list,
204 ++hmp->objid_cache_count;
205 dip->objid_cache = ocp;
208 kfree(ocp, hmp->m_misc);
212 * Steal one from another directory?
214 * Throw away ocp's that are more then half full, they
215 * aren't worth stealing.
217 ocp = TAILQ_FIRST(&hmp->objid_cache_list);
219 ocp->dip->objid_cache = NULL;
220 if (ocp->count >= OBJID_CACHE_BULK / 2) {
221 TAILQ_REMOVE(&hmp->objid_cache_list,
223 --hmp->objid_cache_count;
224 kfree(ocp, hmp->m_misc);
226 dip->objid_cache = ocp;
231 TAILQ_REMOVE(&hmp->objid_cache_list, ocp, entry);
234 * Allocate inode numbers uniformly.
237 n = (namekey >> (63 - OBJID_CACHE_BULK_BITS)) & OBJID_CACHE_BULK_MASK;
238 n = ocp_allocbit(ocp, n);
239 tid = ocp->base_tid + n;
243 * The TID is incremented by 1 or by 16 depending what mode the
244 * mount is operating in.
246 ocp->next_tid += (hmp->master_id < 0) ? 1 : HAMMER_MAX_MASTERS;
248 if (ocp->count >= OBJID_CACHE_BULK * 3 / 4) {
249 dip->objid_cache = NULL;
250 --hmp->objid_cache_count;
252 kfree(ocp, hmp->m_misc);
254 TAILQ_INSERT_TAIL(&hmp->objid_cache_list, ocp, entry);
260 * Allocate a bit starting with bit n. Wrap if necessary.
262 * This routine is only ever called if a bit is available somewhere
266 ocp_allocbit(hammer_objid_cache_t ocp, uint32_t n)
273 while (ocp->bm1[n0] & (1 << n)) {
274 if (ocp->bm0 & (1 << n0)) {
277 } else if (++n == 32) {
283 ocp->bm1[n0] |= 1 << n;
284 if (ocp->bm1[n0] == 0xFFFFFFFFU)
286 return((n0 << 5) + n);
290 hammer_clear_objid(hammer_inode_t dip)
292 hammer_objid_cache_t ocp;
294 if ((ocp = dip->objid_cache) != NULL) {
295 dip->objid_cache = NULL;
297 TAILQ_REMOVE(&dip->hmp->objid_cache_list, ocp, entry);
298 TAILQ_INSERT_HEAD(&dip->hmp->objid_cache_list, ocp, entry);
303 hammer_destroy_objid_cache(hammer_mount_t hmp)
305 hammer_objid_cache_t ocp;
307 while ((ocp = TAILQ_FIRST(&hmp->objid_cache_list)) != NULL) {
308 TAILQ_REMOVE(&hmp->objid_cache_list, ocp, entry);
310 ocp->dip->objid_cache = NULL;
311 kfree(ocp, hmp->m_misc);
312 --hmp->objid_cache_count;
314 KKASSERT(hmp->objid_cache_count == 0);