HAMMER 56C/Many: Performance tuning - MEDIA STRUCTURES CHANGED!
[dragonfly.git] / sys / vfs / hammer / hammer_subs.c
blobb57429b1e16673436e5bd8ff59e6f386686b444d
1 /*
2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
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:
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
16 * distribution.
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
32 * SUCH DAMAGE.
34 * $DragonFly: src/sys/vfs/hammer/hammer_subs.c,v 1.25 2008/06/20 05:38:26 dillon Exp $
37 * HAMMER structural locking
40 #include "hammer.h"
41 #include <sys/dirent.h>
43 void
44 hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident)
46 thread_t td = curthread;
48 KKASSERT(lock->refs > 0);
49 crit_enter();
50 if (lock->locktd != td) {
51 while (lock->locktd != NULL || lock->lockcount) {
52 ++lock->exwanted;
53 lock->wanted = 1;
54 if (hammer_debug_locks) {
55 kprintf("hammer_lock_ex: held by %p\n",
56 lock->locktd);
58 ++hammer_contention_count;
59 tsleep(lock, 0, ident, 0);
60 if (hammer_debug_locks)
61 kprintf("hammer_lock_ex: try again\n");
62 --lock->exwanted;
64 lock->locktd = td;
66 KKASSERT(lock->lockcount >= 0);
67 ++lock->lockcount;
68 crit_exit();
72 * Try to obtain an exclusive lock
74 int
75 hammer_lock_ex_try(struct hammer_lock *lock)
77 thread_t td = curthread;
79 KKASSERT(lock->refs > 0);
80 crit_enter();
81 if (lock->locktd != td) {
82 if (lock->locktd != NULL || lock->lockcount) {
83 crit_exit();
84 return(EAGAIN);
86 lock->locktd = td;
88 KKASSERT(lock->lockcount >= 0);
89 ++lock->lockcount;
90 crit_exit();
91 return(0);
95 * Obtain a shared lock
97 void
98 hammer_lock_sh(struct hammer_lock *lock)
100 KKASSERT(lock->refs > 0);
101 crit_enter();
102 while (lock->locktd != NULL) {
103 if (lock->locktd == curthread) {
104 Debugger("hammer_lock_sh: lock_sh on exclusive");
105 ++lock->lockcount;
106 crit_exit();
107 return;
109 lock->wanted = 1;
110 tsleep(lock, 0, "hmrlck", 0);
112 KKASSERT(lock->lockcount <= 0);
113 --lock->lockcount;
114 crit_exit();
118 * Obtain a shared lock at a lower priority then thread waiting for an
119 * exclusive lock. To avoid a deadlock this may only be done if no other
120 * shared locks are being held by the caller.
122 void
123 hammer_lock_sh_lowpri(struct hammer_lock *lock)
125 KKASSERT(lock->refs > 0);
126 crit_enter();
127 while (lock->locktd != NULL || lock->exwanted) {
128 if (lock->locktd == curthread) {
129 Debugger("hammer_lock_sh: lock_sh on exclusive");
130 ++lock->lockcount;
131 crit_exit();
132 return;
134 lock->wanted = 1;
135 tsleep(lock, 0, "hmrlck", 0);
137 KKASSERT(lock->lockcount <= 0);
138 --lock->lockcount;
139 crit_exit();
143 hammer_lock_sh_try(struct hammer_lock *lock)
145 KKASSERT(lock->refs > 0);
146 crit_enter();
147 if (lock->locktd) {
148 crit_exit();
149 return(EAGAIN);
151 KKASSERT(lock->lockcount <= 0);
152 --lock->lockcount;
153 crit_exit();
154 return(0);
158 * Upgrade a shared lock to an exclusively held lock. This function will
159 * return EDEADLK If there is more then one shared holder.
161 * No error occurs and no action is taken if the lock is already exclusively
162 * held by the caller. If the lock is not held at all or held exclusively
163 * by someone else, this function will panic.
166 hammer_lock_upgrade(struct hammer_lock *lock)
168 int error;
170 crit_enter();
171 if (lock->lockcount > 0) {
172 if (lock->locktd != curthread)
173 panic("hammer_lock_upgrade: illegal lock state");
174 error = 0;
175 } else if (lock->lockcount == -1) {
176 lock->lockcount = 1;
177 lock->locktd = curthread;
178 error = 0;
179 } else if (lock->lockcount != 0) {
180 error = EDEADLK;
181 } else {
182 panic("hammer_lock_upgrade: lock is not held");
183 /* NOT REACHED */
184 error = 0;
186 crit_exit();
187 return(error);
191 * Downgrade an exclusively held lock to a shared lock.
193 void
194 hammer_lock_downgrade(struct hammer_lock *lock)
196 KKASSERT(lock->lockcount == 1 && lock->locktd == curthread);
197 crit_enter();
198 lock->lockcount = -1;
199 lock->locktd = NULL;
200 if (lock->wanted) {
201 lock->wanted = 0;
202 wakeup(lock);
204 crit_exit();
205 /* XXX memory barrier */
208 void
209 hammer_unlock(struct hammer_lock *lock)
211 crit_enter();
212 KKASSERT(lock->lockcount != 0);
213 if (lock->lockcount < 0) {
214 if (++lock->lockcount == 0 && lock->wanted) {
215 lock->wanted = 0;
216 wakeup(lock);
218 } else {
219 KKASSERT(lock->locktd == curthread);
220 if (--lock->lockcount == 0) {
221 lock->locktd = NULL;
222 if (lock->wanted) {
223 lock->wanted = 0;
224 wakeup(lock);
229 crit_exit();
232 void
233 hammer_ref(struct hammer_lock *lock)
235 KKASSERT(lock->refs >= 0);
236 crit_enter();
237 ++lock->refs;
238 crit_exit();
241 void
242 hammer_unref(struct hammer_lock *lock)
244 KKASSERT(lock->refs > 0);
245 crit_enter();
246 --lock->refs;
247 crit_exit();
251 * The sync_lock must be held when doing any modifying operations on
252 * meta-data. The flusher holds the lock exclusively while the reblocker
253 * and pruner use a shared lock.
255 * Modifying operations can run in parallel until the flusher needs to
256 * sync the disk media.
258 void
259 hammer_sync_lock_ex(hammer_transaction_t trans)
261 ++trans->sync_lock_refs;
262 hammer_lock_ex(&trans->hmp->sync_lock);
265 void
266 hammer_sync_lock_sh(hammer_transaction_t trans)
268 ++trans->sync_lock_refs;
269 hammer_lock_sh(&trans->hmp->sync_lock);
273 hammer_sync_lock_sh_try(hammer_transaction_t trans)
275 int error;
277 ++trans->sync_lock_refs;
278 if ((error = hammer_lock_sh_try(&trans->hmp->sync_lock)) != 0)
279 --trans->sync_lock_refs;
280 return (error);
283 void
284 hammer_sync_unlock(hammer_transaction_t trans)
286 --trans->sync_lock_refs;
287 hammer_unlock(&trans->hmp->sync_lock);
291 * Misc
293 u_int32_t
294 hammer_to_unix_xid(uuid_t *uuid)
296 return(*(u_int32_t *)&uuid->node[2]);
299 void
300 hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid)
302 bzero(uuid, sizeof(*uuid));
303 *(u_int32_t *)&uuid->node[2] = guid;
306 void
307 hammer_to_timespec(hammer_tid_t tid, struct timespec *ts)
309 ts->tv_sec = tid / 1000000000;
310 ts->tv_nsec = tid % 1000000000;
313 hammer_tid_t
314 hammer_timespec_to_transid(struct timespec *ts)
316 hammer_tid_t tid;
318 tid = ts->tv_nsec + (unsigned long)ts->tv_sec * 1000000000LL;
319 return(tid);
324 * Convert a HAMMER filesystem object type to a vnode type
326 enum vtype
327 hammer_get_vnode_type(u_int8_t obj_type)
329 switch(obj_type) {
330 case HAMMER_OBJTYPE_DIRECTORY:
331 return(VDIR);
332 case HAMMER_OBJTYPE_REGFILE:
333 return(VREG);
334 case HAMMER_OBJTYPE_DBFILE:
335 return(VDATABASE);
336 case HAMMER_OBJTYPE_FIFO:
337 return(VFIFO);
338 case HAMMER_OBJTYPE_CDEV:
339 return(VCHR);
340 case HAMMER_OBJTYPE_BDEV:
341 return(VBLK);
342 case HAMMER_OBJTYPE_SOFTLINK:
343 return(VLNK);
344 default:
345 return(VBAD);
347 /* not reached */
351 hammer_get_dtype(u_int8_t obj_type)
353 switch(obj_type) {
354 case HAMMER_OBJTYPE_DIRECTORY:
355 return(DT_DIR);
356 case HAMMER_OBJTYPE_REGFILE:
357 return(DT_REG);
358 case HAMMER_OBJTYPE_DBFILE:
359 return(DT_DBF);
360 case HAMMER_OBJTYPE_FIFO:
361 return(DT_FIFO);
362 case HAMMER_OBJTYPE_CDEV:
363 return(DT_CHR);
364 case HAMMER_OBJTYPE_BDEV:
365 return(DT_BLK);
366 case HAMMER_OBJTYPE_SOFTLINK:
367 return(DT_LNK);
368 default:
369 return(DT_UNKNOWN);
371 /* not reached */
374 u_int8_t
375 hammer_get_obj_type(enum vtype vtype)
377 switch(vtype) {
378 case VDIR:
379 return(HAMMER_OBJTYPE_DIRECTORY);
380 case VREG:
381 return(HAMMER_OBJTYPE_REGFILE);
382 case VDATABASE:
383 return(HAMMER_OBJTYPE_DBFILE);
384 case VFIFO:
385 return(HAMMER_OBJTYPE_FIFO);
386 case VCHR:
387 return(HAMMER_OBJTYPE_CDEV);
388 case VBLK:
389 return(HAMMER_OBJTYPE_BDEV);
390 case VLNK:
391 return(HAMMER_OBJTYPE_SOFTLINK);
392 default:
393 return(HAMMER_OBJTYPE_UNKNOWN);
395 /* not reached */
399 hammer_nohistory(hammer_inode_t ip)
401 if (ip->hmp->hflags & HMNT_NOHISTORY)
402 return(1);
403 if (ip->ino_data.uflags & (SF_NOHISTORY|UF_NOHISTORY))
404 return(1);
405 return(0);
409 * Return a namekey hash. The 64 bit namekey hash consists of a 32 bit
410 * crc in the MSB and 0 in the LSB. The caller will use the low bits to
411 * generate a unique key and will scan all entries with the same upper
412 * 32 bits when issuing a lookup.
414 * We strip bit 63 in order to provide a positive key, this way a seek
415 * offset of 0 will represent the base of the directory.
417 * This function can never return 0. We use the MSB-0 space to synthesize
418 * artificial directory entries such as "." and "..".
420 int64_t
421 hammer_directory_namekey(void *name, int len)
423 int64_t key;
425 key = (int64_t)(crc32(name, len) & 0x7FFFFFFF) << 32;
426 if (key == 0)
427 key |= 0x100000000LL;
428 return(key);
431 hammer_tid_t
432 hammer_now_tid(void)
434 struct timespec ts;
435 hammer_tid_t tid;
437 getnanotime(&ts);
438 tid = ts.tv_sec * 1000000000LL + ts.tv_nsec;
439 return(tid);
442 hammer_tid_t
443 hammer_str_to_tid(const char *str)
445 hammer_tid_t tid;
446 int len = strlen(str);
448 if (len > 10)
449 tid = strtouq(str, NULL, 0); /* full TID */
450 else
451 tid = strtouq(str, NULL, 0) * 1000000000LL; /* time_t */
452 return(tid);
455 void
456 hammer_crc_set_blockmap(hammer_blockmap_t blockmap)
458 blockmap->entry_crc = crc32(blockmap, HAMMER_BLOCKMAP_CRCSIZE);
461 void
462 hammer_crc_set_volume(hammer_volume_ondisk_t ondisk)
464 ondisk->vol_crc = crc32(ondisk, HAMMER_VOL_CRCSIZE1) ^
465 crc32(&ondisk->vol_crc + 1, HAMMER_VOL_CRCSIZE2);
469 hammer_crc_test_blockmap(hammer_blockmap_t blockmap)
471 hammer_crc_t crc;
473 crc = crc32(blockmap, HAMMER_BLOCKMAP_CRCSIZE);
474 return (blockmap->entry_crc == crc);
478 hammer_crc_test_volume(hammer_volume_ondisk_t ondisk)
480 hammer_crc_t crc;
482 crc = crc32(ondisk, HAMMER_VOL_CRCSIZE1) ^
483 crc32(&ondisk->vol_crc + 1, HAMMER_VOL_CRCSIZE2);
484 return (ondisk->vol_crc == crc);
488 hammer_crc_test_btree(hammer_node_ondisk_t ondisk)
490 hammer_crc_t crc;
492 crc = crc32(&ondisk->crc + 1, HAMMER_BTREE_CRCSIZE);
493 return (ondisk->crc == crc);
496 void
497 hkprintf(const char *ctl, ...)
499 __va_list va;
501 if (hammer_debug_debug) {
502 __va_start(va, ctl);
503 kvprintf(ctl, va);
504 __va_end(va);
509 * Return the block size at the specified file offset.
512 hammer_blocksize(int64_t file_offset)
514 if (file_offset < HAMMER_XDEMARC)
515 return(HAMMER_BUFSIZE);
516 else
517 return(HAMMER_XBUFSIZE);
521 * Return the demarkation point between the two offsets where
522 * the block size changes.
524 int64_t
525 hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2)
527 if (file_offset1 < HAMMER_XDEMARC) {
528 if (file_offset2 <= HAMMER_XDEMARC)
529 return(file_offset2);
530 return(HAMMER_XDEMARC);
532 panic("hammer_blockdemarc: illegal range %lld %lld\n",
533 file_offset1, file_offset2);