Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly into...
[dragonfly.git] / sys / vfs / hammer / hammer.h
CommitLineData
8750964d 1/*
1f07f686 2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
8750964d
MD
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:
10 *
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.
20 *
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.
33 *
44a83111 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.130 2008/11/13 02:18:43 dillon Exp $
8750964d
MD
35 */
36/*
37 * This header file contains structures used internally by the HAMMERFS
c60bb2c5 38 * implementation. See hammer_disk.h for on-disk structures.
8750964d
MD
39 */
40
427e5fc6
MD
41#include <sys/param.h>
42#include <sys/types.h>
43#include <sys/kernel.h>
42c7d26b 44#include <sys/conf.h>
427e5fc6 45#include <sys/systm.h>
8750964d
MD
46#include <sys/tree.h>
47#include <sys/malloc.h>
427e5fc6 48#include <sys/mount.h>
513ca7d7 49#include <sys/mountctl.h>
427e5fc6 50#include <sys/vnode.h>
42c7d26b 51#include <sys/proc.h>
e63644f0 52#include <sys/stat.h>
427e5fc6 53#include <sys/globaldata.h>
66325755
MD
54#include <sys/lockf.h>
55#include <sys/buf.h>
8cd0a023 56#include <sys/queue.h>
bcac4bbb 57#include <sys/ktr.h>
66325755
MD
58#include <sys/globaldata.h>
59
60#include <sys/buf2.h>
855942b6 61#include <sys/signal2.h>
427e5fc6 62#include "hammer_disk.h"
8750964d 63#include "hammer_mount.h"
7dc57964 64#include "hammer_ioctl.h"
8750964d
MD
65
66#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
67
68MALLOC_DECLARE(M_HAMMER);
69
bcac4bbb
MD
70/*
71 * Kernel trace
72 */
73#if !defined(KTR_HAMMER)
74#define KTR_HAMMER KTR_ALL
75#endif
76KTR_INFO_MASTER_EXTERN(hammer);
77
78/*
79 * Misc structures
80 */
66325755
MD
81struct hammer_mount;
82
8750964d
MD
83/*
84 * Key structure used for custom RB tree inode lookups. This prototypes
85 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
86 */
87typedef struct hammer_inode_info {
513ca7d7 88 int64_t obj_id; /* (key) object identifier */
8750964d 89 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 90 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
91 union {
92 struct hammer_btree_leaf_elm *leaf;
93 } u;
8750964d
MD
94} *hammer_inode_info_t;
95
b84de5af
MD
96typedef enum hammer_transaction_type {
97 HAMMER_TRANS_RO,
98 HAMMER_TRANS_STD,
99 HAMMER_TRANS_FLS
100} hammer_transaction_type_t;
101
66325755
MD
102/*
103 * HAMMER Transaction tracking
104 */
105struct hammer_transaction {
b84de5af 106 hammer_transaction_type_t type;
66325755
MD
107 struct hammer_mount *hmp;
108 hammer_tid_t tid;
ddfdf542 109 u_int64_t time;
dd94f1b1 110 u_int32_t time32;
2f85fa4d 111 int sync_lock_refs;
21fde338 112 int flags;
a89aec1b 113 struct hammer_volume *rootvol;
66325755
MD
114};
115
8cd0a023
MD
116typedef struct hammer_transaction *hammer_transaction_t;
117
21fde338
MD
118#define HAMMER_TRANSF_NEWINODE 0x0001
119
66325755
MD
120/*
121 * HAMMER locks
122 */
427e5fc6 123struct hammer_lock {
0b075555 124 int refs; /* active references delay writes */
0b075555 125 int lockcount; /* lock count for exclusive/shared access */
427e5fc6 126 int wanted;
4a2796f3 127 int exwanted; /* number of threads waiting for ex lock */
427e5fc6 128 struct thread *locktd;
8750964d
MD
129};
130
427e5fc6
MD
131static __inline int
132hammer_islocked(struct hammer_lock *lock)
133{
8cd0a023 134 return(lock->lockcount != 0);
427e5fc6
MD
135}
136
0b075555
MD
137static __inline int
138hammer_isactive(struct hammer_lock *lock)
139{
140 return(lock->refs != 0);
141}
142
427e5fc6
MD
143static __inline int
144hammer_islastref(struct hammer_lock *lock)
145{
146 return(lock->refs == 1);
147}
c60bb2c5 148
6a37e7e4 149/*
7aa3b8a6 150 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
151 */
152static __inline int
7aa3b8a6 153hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 154{
7aa3b8a6
MD
155 if (lock->lockcount > 0 && lock->locktd == td)
156 return(1);
157 return(0);
6a37e7e4
MD
158}
159
ec4e8497 160/*
1f07f686 161 * Flush state, used by various structures
ec4e8497 162 */
1f07f686
MD
163typedef enum hammer_inode_state {
164 HAMMER_FST_IDLE,
165 HAMMER_FST_SETUP,
166 HAMMER_FST_FLUSH
167} hammer_inode_state_t;
ec4e8497 168
1f07f686 169TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 170
5fa5c92f
MD
171/*
172 * Pseudo-filesystem extended data tracking
173 */
174struct hammer_pfs_rb_tree;
175struct hammer_pseudofs_inmem;
176RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
177RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
178 hammer_pfs_rb_compare, u_int32_t);
179
180struct hammer_pseudofs_inmem {
181 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
182 struct hammer_lock lock;
183 u_int32_t localization;
ea434b6f 184 hammer_tid_t create_tid;
842e7a70 185 int flags;
a56cb012 186 udev_t fsid_udev;
5fa5c92f
MD
187 struct hammer_pseudofs_data pfsd;
188};
189
190typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
191
842e7a70
MD
192#define HAMMER_PFSM_DELETED 0x0001
193
0729c8c8
MD
194/*
195 * Cache object ids. A fixed number of objid cache structures are
196 * created to reserve object id's for newly created files in multiples
197 * of 100,000, localized to a particular directory, and recycled as
198 * needed. This allows parallel create operations in different
199 * directories to retain fairly localized object ids which in turn
200 * improves reblocking performance and layout.
201 */
bf3b416b 202#define OBJID_CACHE_SIZE 1024
0729c8c8
MD
203#define OBJID_CACHE_BULK 100000
204
205typedef struct hammer_objid_cache {
206 TAILQ_ENTRY(hammer_objid_cache) entry;
207 struct hammer_inode *dip;
208 hammer_tid_t next_tid;
209 int count;
210} *hammer_objid_cache_t;
211
bcac4bbb
MD
212/*
213 * Associate an inode with a B-Tree node to cache search start positions
214 */
215typedef struct hammer_node_cache {
7a61b85d 216 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
217 struct hammer_node *node;
218 struct hammer_inode *ip;
219} *hammer_node_cache_t;
220
221TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
222
7a61b85d
MD
223/*
224 * Structure used to organize flush groups. Flush groups must be
225 * organized into chunks in order to avoid blowing out the UNDO FIFO.
226 * Without this a 'sync' could end up flushing 50,000 inodes in a single
227 * transaction.
228 */
229struct hammer_flush_group {
230 TAILQ_ENTRY(hammer_flush_group) flush_entry;
231 TAILQ_HEAD(, hammer_inode) flush_list;
232 int inode_count; /* inode load */
233 int total_count; /* record load */
234 int running; /* group is running */
235 int closed;
236 int refs;
237};
238
239typedef struct hammer_flush_group *hammer_flush_group_t;
240
241TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
242
8750964d 243/*
8cd0a023
MD
244 * Structure used to represent an inode in-memory.
245 *
246 * The record and data associated with an inode may be out of sync with
247 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
248 * clear).
249 *
250 * An inode may also hold a cache of unsynchronized records, used for
251 * database and directories only. Unsynchronized regular file data is
252 * stored in the buffer cache.
253 *
254 * NOTE: A file which is created and destroyed within the initial
255 * synchronization period can wind up not doing any disk I/O at all.
256 *
257 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
258 */
259struct hammer_ino_rb_tree;
260struct hammer_inode;
261RB_HEAD(hammer_ino_rb_tree, hammer_inode);
262RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
263 hammer_ino_rb_compare, hammer_inode_info_t);
264
265struct hammer_rec_rb_tree;
266struct hammer_record;
267RB_HEAD(hammer_rec_rb_tree, hammer_record);
268RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 269 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
270
271TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
272
273struct hammer_inode {
1f07f686
MD
274 RB_ENTRY(hammer_inode) rb_node;
275 hammer_inode_state_t flush_state;
7a61b85d 276 hammer_flush_group_t flush_group;
059819e3 277 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 278 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 279 int64_t obj_id; /* (key) object identifier */
b84de5af 280 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 281 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 282 struct hammer_mount *hmp;
0729c8c8 283 hammer_objid_cache_t objid_cache;
b84de5af
MD
284 int flags;
285 int error; /* flush error */
286 int cursor_ip_refs; /* sanity */
47637bff 287 int rsv_recs;
b84de5af 288 struct vnode *vp;
5fa5c92f 289 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
290 struct lockf advlock;
291 struct hammer_lock lock; /* sync copy interlock */
b84de5af 292 off_t trunc_off;
11ad5ade 293 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
294 struct hammer_inode_data ino_data; /* in-memory cache */
295 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
bcac4bbb 296 struct hammer_node_cache cache[2]; /* search initiate cache */
b84de5af
MD
297
298 /*
299 * When a demark is created to synchronize an inode to
300 * disk, certain fields are copied so the front-end VOPs
301 * can continue to run in parallel with the synchronization
302 * occuring in the background.
303 */
304 int sync_flags; /* to-sync flags cache */
305 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 306 off_t save_trunc_off; /* write optimization */
11ad5ade 307 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 308 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
309};
310
8cd0a023
MD
311typedef struct hammer_inode *hammer_inode_t;
312
66325755
MD
313#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
314
8cd0a023 315#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
ddfdf542 316 /* (not including atime/mtime) */
e63644f0 317#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
7b6ccb11 318#define HAMMER_INODE_CONN_DOWN 0x0004 /* include in downward recursion */
1f07f686 319#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 320#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 321#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 322#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 323#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 324#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 325#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 326#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 327#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
f153644d 328#define HAMMER_INODE_REFLUSH 0x2000 /* flush on dependancy / reflush */
9f5097dc 329#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 330#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 331
b84de5af 332#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 333#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 334#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 335#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
336#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
337#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
b84de5af 338
11ad5ade
MD
339#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \
340 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
341 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
342 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686
MD
343
344#define HAMMER_INODE_MODMASK_NOXDIRTY \
345 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 346
cb51be26 347#define HAMMER_FLUSH_GROUP_SIZE 64
66325755 348
f90dde4c 349#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 350#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 351
7bc5b8c2
MD
352/*
353 * Used by the inode reclaim code to pipeline reclaims and avoid
354 * blowing out kernel memory or letting the flusher get too far
355 * behind.
356 */
357struct hammer_reclaim {
358 TAILQ_ENTRY(hammer_reclaim) entry;
359 int okydoky;
360};
361
4a2796f3
MD
362#define HAMMER_RECLAIM_FLUSH 2000
363#define HAMMER_RECLAIM_WAIT 4000
9f5097dc 364
8750964d 365/*
1f07f686
MD
366 * Structure used to represent an unsynchronized record in-memory. These
367 * records typically represent directory entries. Only non-historical
368 * records are kept in-memory.
369 *
370 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
371 * on disk then neither are any records and the in-memory record tree
372 * represents the entire contents of the inode. If the inode is on disk
373 * then the on-disk B-Tree is scanned in parallel with the in-memory
374 * RB-Tree to synthesize the current state of the file.
375 *
1f07f686
MD
376 * Records are also used to enforce the ordering of directory create/delete
377 * operations. A new inode will not be flushed to disk unless its related
378 * directory entry is also being flushed at the same time. A directory entry
379 * will not be removed unless its related inode is also being removed at the
380 * same time.
8750964d 381 */
1f07f686 382typedef enum hammer_record_type {
e8599db1 383 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 384 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 385 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
386 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
387 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
388} hammer_record_type_t;
389
8cd0a023
MD
390struct hammer_record {
391 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
392 TAILQ_ENTRY(hammer_record) target_entry;
393 hammer_inode_state_t flush_state;
7a61b85d 394 hammer_flush_group_t flush_group;
1f07f686 395 hammer_record_type_t type;
a89aec1b 396 struct hammer_lock lock;
0832c9bb 397 struct hammer_reserve *resv;
8cd0a023 398 struct hammer_inode *ip;
1f07f686 399 struct hammer_inode *target_ip;
11ad5ade 400 struct hammer_btree_leaf_elm leaf;
8cd0a023 401 union hammer_data_ondisk *data;
8cd0a023 402 int flags;
e469566b 403 hammer_off_t zone2_offset; /* direct-write only */
8cd0a023 404};
8750964d 405
8cd0a023
MD
406typedef struct hammer_record *hammer_record_t;
407
d36ec43b
MD
408/*
409 * Record flags. Note that FE can only be set by the frontend if the
410 * record has not been interlocked by the backend w/ BE.
411 */
8cd0a023
MD
412#define HAMMER_RECF_ALLOCDATA 0x0001
413#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 414#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 415#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
1b0ab2c3 416#define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */
d36ec43b 417#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 418#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
1b0ab2c3
MD
419#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
420#define HAMMER_RECF_DIRECT_IO 0x0200 /* related direct I/O running*/
421#define HAMMER_RECF_DIRECT_WAIT 0x0400 /* related direct I/O running*/
e469566b 422#define HAMMER_RECF_DIRECT_INVAL 0x0800 /* buffer alias invalidation */
8cd0a023 423
602c6cb8
MD
424/*
425 * hammer_delete_at_cursor() flags
426 */
427#define HAMMER_DELETE_ADJUST 0x0001
428#define HAMMER_DELETE_DESTROY 0x0002
429
8cd0a023 430/*
47197d71 431 * In-memory structures representing on-disk structures.
8cd0a023 432 */
8750964d 433struct hammer_volume;
427e5fc6 434struct hammer_buffer;
8cd0a023 435struct hammer_node;
e8599db1 436struct hammer_undo;
0832c9bb
MD
437struct hammer_reserve;
438
8750964d 439RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 440RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 441RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 442RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 443RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
444
445RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
446 hammer_vol_rb_compare, int32_t);
427e5fc6 447RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 448 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 449RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 450 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
451RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
452 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
453RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
454 hammer_res_rb_compare, hammer_off_t);
8750964d 455
66325755
MD
456/*
457 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
458 *
459 * VOLUME - hammer_volume containing meta-data
460 * META_BUFFER - hammer_buffer containing meta-data
461 * DATA_BUFFER - hammer_buffer containing pure-data
462 *
463 * Dirty volume headers and dirty meta-data buffers are locked until the
464 * flusher can sequence them out. Dirty pure-data buffers can be written.
465 * Clean buffers can be passively released.
66325755 466 */
10a5d1ba
MD
467typedef enum hammer_io_type {
468 HAMMER_STRUCTURE_VOLUME,
469 HAMMER_STRUCTURE_META_BUFFER,
470 HAMMER_STRUCTURE_UNDO_BUFFER,
471 HAMMER_STRUCTURE_DATA_BUFFER
472} hammer_io_type_t;
66325755
MD
473
474union hammer_io_structure;
055f5ff8 475struct hammer_io;
66325755
MD
476
477struct worklist {
478 LIST_ENTRY(worklist) node;
479};
480
10a5d1ba
MD
481TAILQ_HEAD(hammer_io_list, hammer_io);
482typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 483
66325755 484struct hammer_io {
10a5d1ba
MD
485 struct worklist worklist;
486 struct hammer_lock lock;
487 enum hammer_io_type type;
488 struct hammer_mount *hmp;
748efb59 489 struct hammer_volume *volume;
10a5d1ba
MD
490 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
491 hammer_io_list_t mod_list;
492 struct buf *bp;
4a2796f3
MD
493 int64_t offset; /* zone-2 offset */
494 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
495 int loading; /* loading/unloading interlock */
496 int modify_refs;
497
66325755
MD
498 u_int modified : 1; /* bp's data was modified */
499 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
500 u_int running : 1; /* bp write IO in progress */
501 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 502 u_int validated : 1; /* ondisk has been validated */
b58c6388 503 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 504 u_int recovered : 1; /* has recovery ref */
9f5097dc 505 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 506 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 507 u_int gencrc : 1; /* crc needs to be generated */
cdb6e4e6 508 u_int ioerror : 1; /* abort on io-error */
66325755
MD
509};
510
8cd0a023
MD
511typedef struct hammer_io *hammer_io_t;
512
af209b0f
MD
513#define HAMMER_CLUSTER_SIZE (64 * 1024)
514#if HAMMER_CLUSTER_SIZE > MAXBSIZE
515#undef HAMMER_CLUSTER_SIZE
516#define HAMMER_CLUSTER_SIZE MAXBSIZE
517#endif
518#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
519
66325755 520/*
8cd0a023 521 * In-memory volume representing on-disk buffer
66325755 522 */
8750964d 523struct hammer_volume {
66325755 524 struct hammer_io io;
8750964d 525 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
526 struct hammer_volume_ondisk *ondisk;
527 int32_t vol_no;
fbc6e32a 528 int64_t nblocks; /* note: special calculation for statfs */
47197d71 529 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
530 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
531 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 532 char *vol_name;
8750964d 533 struct vnode *devvp;
427e5fc6 534 int vol_flags;
8750964d
MD
535};
536
8cd0a023
MD
537typedef struct hammer_volume *hammer_volume_t;
538
66325755 539/*
8cd0a023
MD
540 * In-memory buffer (other then volume, super-cluster, or cluster),
541 * representing an on-disk buffer.
66325755 542 */
427e5fc6 543struct hammer_buffer {
66325755 544 struct hammer_io io;
427e5fc6 545 RB_ENTRY(hammer_buffer) rb_node;
47197d71 546 void *ondisk;
34d829f7 547 hammer_off_t zoneX_offset;
0832c9bb 548 hammer_off_t zone2_offset;
cebe9493 549 struct hammer_reserve *resv;
8cd0a023 550 struct hammer_node_list clist;
66325755
MD
551};
552
8cd0a023
MD
553typedef struct hammer_buffer *hammer_buffer_t;
554
555/*
556 * In-memory B-Tree node, representing an on-disk B-Tree node.
557 *
558 * This is a hang-on structure which is backed by a hammer_buffer,
559 * indexed by a hammer_cluster, and used for fine-grained locking of
560 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
561 * can contain multiple nodes representing wildly disassociated portions
562 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
563 *
564 * This structure uses a cluster-relative index to reduce the number
565 * of layers required to access it, and also because all on-disk B-Tree
566 * references are cluster-relative offsets.
567 */
568struct hammer_node {
569 struct hammer_lock lock; /* node-by-node lock */
570 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
571 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 572 hammer_off_t node_offset; /* full offset spec */
40043e7f 573 struct hammer_mount *hmp;
8cd0a023
MD
574 struct hammer_buffer *buffer; /* backing buffer */
575 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 576 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 577 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 578 int flags;
740d8317 579 int loading; /* load interlock */
8cd0a023
MD
580};
581
b3deaf57
MD
582#define HAMMER_NODE_DELETED 0x0001
583#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
584#define HAMMER_NODE_CRCGOOD 0x0004
585#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 586#define HAMMER_NODE_NEEDSMIRROR 0x0010
b3deaf57 587
8cd0a023
MD
588typedef struct hammer_node *hammer_node_t;
589
b33e2cc0
MD
590/*
591 * List of locked nodes.
592 */
593struct hammer_node_locklist {
594 struct hammer_node_locklist *next;
595 hammer_node_t node;
596};
597
598typedef struct hammer_node_locklist *hammer_node_locklist_t;
599
600
8cd0a023
MD
601/*
602 * Common I/O management structure - embedded in in-memory structures
603 * which are backed by filesystem buffers.
604 */
66325755
MD
605union hammer_io_structure {
606 struct hammer_io io;
607 struct hammer_volume volume;
66325755 608 struct hammer_buffer buffer;
8750964d
MD
609};
610
055f5ff8
MD
611typedef union hammer_io_structure *hammer_io_structure_t;
612
cebe9493
MD
613/*
614 * The reserve structure prevents the blockmap from allocating
615 * out of a reserved bigblock. Such reservations are used by
616 * the direct-write mechanism.
617 *
618 * The structure is also used to hold off on reallocations of
619 * big blocks from the freemap until flush dependancies have
620 * been dealt with.
621 */
0832c9bb
MD
622struct hammer_reserve {
623 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
624 TAILQ_ENTRY(hammer_reserve) delay_entry;
625 int flush_group;
cb51be26 626 int flags;
0832c9bb 627 int refs;
cb51be26 628 int zone;
df301614 629 int append_off;
cebe9493 630 hammer_off_t zone_offset;
0832c9bb
MD
631};
632
633typedef struct hammer_reserve *hammer_reserve_t;
634
cb51be26 635#define HAMMER_RESF_ONDELAY 0x0001
5e435c92 636#define HAMMER_RESF_LAYER2FREE 0x0002
cb51be26 637
8cd0a023
MD
638#include "hammer_cursor.h"
639
e8599db1 640/*
cebe9493
MD
641 * The undo structure tracks recent undos to avoid laying down duplicate
642 * undos within a flush group, saving us a significant amount of overhead.
643 *
644 * This is strictly a heuristic.
e8599db1 645 */
7a61b85d
MD
646#define HAMMER_MAX_UNDOS 1024
647#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
648
649struct hammer_undo {
650 RB_ENTRY(hammer_undo) rb_node;
651 TAILQ_ENTRY(hammer_undo) lru_entry;
652 hammer_off_t offset;
653 int bytes;
654};
655
656typedef struct hammer_undo *hammer_undo_t;
657
af209b0f 658struct hammer_flusher_info;
7a61b85d 659TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
660
661struct hammer_flusher {
662 int signal; /* flusher thread sequencer */
663 int act; /* currently active flush group */
664 int done; /* set to act when complete */
665 int next; /* next flush group */
666 int group_lock; /* lock sequencing of the next flush */
667 int exiting; /* request master exit */
da2da375
MD
668 thread_t td; /* master flusher thread */
669 hammer_tid_t tid; /* last flushed transaction id */
670 int finalize_want; /* serialize finalization */
671 struct hammer_lock finalize_lock; /* serialize finalization */
672 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
673 struct hammer_flusher_info_list run_list;
674 struct hammer_flusher_info_list ready_list;
da2da375
MD
675};
676
8750964d
MD
677/*
678 * Internal hammer mount data structure
679 */
680struct hammer_mount {
681 struct mount *mp;
27ea2398 682 /*struct vnode *rootvp;*/
8750964d
MD
683 struct hammer_ino_rb_tree rb_inos_root;
684 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 685 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 686 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
687 struct hammer_res_rb_tree rb_resv_root;
688 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 689 struct hammer_pfs_rb_tree rb_pfsm_root;
8750964d 690 struct hammer_volume *rootvol;
47197d71
MD
691 struct hammer_base_elm root_btree_beg;
692 struct hammer_base_elm root_btree_end;
bac808fe
MD
693
694 struct malloc_type *m_misc;
695 struct malloc_type *m_inodes;
696
cdb6e4e6 697 int flags; /* HAMMER_MOUNT_xxx flags */
195c19a1
MD
698 int hflags;
699 int ronly;
700 int nvolumes;
d26d0ae9 701 int volume_iterator;
732a1697 702 int master_id; /* -1 or 0-15 - clustering and mirroring */
44a83111 703 int version; /* hammer filesystem version to use */
e63644f0 704 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 705 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 706 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 707 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 708 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
709 int last_newrecords;
710 int count_newrecords;
da2da375 711
9f5097dc
MD
712 int inode_reclaims; /* inodes pending reclaim by flusher */
713 int count_inodes; /* total number of inodes */
af209b0f 714 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
715
716 struct hammer_flusher flusher;
717
855942b6 718 u_int check_interrupt;
8750964d 719 uuid_t fsid;
10a5d1ba
MD
720 struct hammer_io_list volu_list; /* dirty undo buffers */
721 struct hammer_io_list undo_list; /* dirty undo buffers */
722 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 723 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
724 struct hammer_io_list meta_list; /* dirty meta bufs */
725 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a
MD
726 int locked_dirty_space; /* meta/volu count */
727 int io_running_space;
0729c8c8 728 int objid_cache_count;
cdb6e4e6
MD
729 int error; /* critical I/O error */
730 struct krate krate; /* rate limited kprintf */
ddfdf542 731 hammer_tid_t asof; /* snapshot mount */
4889cbd4
MD
732 hammer_tid_t next_tid;
733 hammer_tid_t flush_tid1; /* flusher tid sequencing */
734 hammer_tid_t flush_tid2; /* flusher tid sequencing */
e63644f0
MD
735 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
736
513ca7d7 737 struct netexport export;
9480ff55 738 struct hammer_lock sync_lock;
c9b9e29d 739 struct hammer_lock free_lock;
d99d6bf5
MD
740 struct hammer_lock undo_lock;
741 struct hammer_lock blkmap_lock;
0729c8c8 742 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
743 struct hammer_undo undos[HAMMER_MAX_UNDOS];
744 int undo_alloc;
745 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 746 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 747 struct hammer_flush_group_list flush_group_list;
7b6ccb11 748 hammer_flush_group_t next_flush_group;
0729c8c8 749 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
7bc5b8c2 750 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
8750964d
MD
751};
752
8cd0a023 753typedef struct hammer_mount *hammer_mount_t;
8750964d 754
cdb6e4e6 755#define HAMMER_MOUNT_CRITICAL_ERROR 0x0001
9f5097dc 756
fbc6e32a
MD
757struct hammer_sync_info {
758 int error;
759 int waitfor;
760};
761
8750964d
MD
762#endif
763
a7e9bef1
MD
764/*
765 * checkspace slop (8MB chunks), higher numbers are more conservative.
766 */
93291532
MD
767#define HAMMER_CHKSPC_REBLOCK 25
768#define HAMMER_CHKSPC_MIRROR 20
769#define HAMMER_CHKSPC_WRITE 20
770#define HAMMER_CHKSPC_CREATE 20
771#define HAMMER_CHKSPC_REMOVE 10
7b6ccb11 772#define HAMMER_CHKSPC_EMERGENCY 0
a7e9bef1 773
8750964d
MD
774#if defined(_KERNEL)
775
776extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
777extern struct vop_ops hammer_spec_vops;
778extern struct vop_ops hammer_fifo_vops;
66325755 779extern struct bio_ops hammer_bioops;
427e5fc6 780
2f85fa4d 781extern int hammer_debug_io;
d5ef456e 782extern int hammer_debug_general;
77062c8a 783extern int hammer_debug_debug;
e8599db1 784extern int hammer_debug_inode;
7d683b0f 785extern int hammer_debug_locks;
b3deaf57 786extern int hammer_debug_btree;
d113fda1 787extern int hammer_debug_tid;
b33e2cc0 788extern int hammer_debug_recover;
46fe7ae1 789extern int hammer_debug_recover_faults;
1b0ab2c3 790extern int hammer_cluster_enable;
7a61b85d 791extern int hammer_count_fsyncs;
b3deaf57 792extern int hammer_count_inodes;
af209b0f 793extern int hammer_count_iqueued;
9f5097dc 794extern int hammer_count_reclaiming;
b3deaf57
MD
795extern int hammer_count_records;
796extern int hammer_count_record_datas;
797extern int hammer_count_volumes;
b3deaf57
MD
798extern int hammer_count_buffers;
799extern int hammer_count_nodes;
a7e9bef1 800extern int64_t hammer_count_extra_space_used;
cb51be26
MD
801extern int64_t hammer_stats_btree_lookups;
802extern int64_t hammer_stats_btree_searches;
803extern int64_t hammer_stats_btree_inserts;
804extern int64_t hammer_stats_btree_deletes;
805extern int64_t hammer_stats_btree_elements;
806extern int64_t hammer_stats_btree_splits;
807extern int64_t hammer_stats_btree_iterations;
808extern int64_t hammer_stats_record_iterations;
ce0138a6
MD
809extern int64_t hammer_stats_file_read;
810extern int64_t hammer_stats_file_write;
811extern int64_t hammer_stats_file_iopsr;
812extern int64_t hammer_stats_file_iopsw;
813extern int64_t hammer_stats_disk_read;
814extern int64_t hammer_stats_disk_write;
815extern int64_t hammer_stats_inode_flushes;
816extern int64_t hammer_stats_commits;
f5a07a7a 817extern int hammer_count_dirtybufspace;
a99b9ea2 818extern int hammer_count_refedbufs;
0832c9bb 819extern int hammer_count_reservations;
a99b9ea2
MD
820extern int hammer_count_io_running_read;
821extern int hammer_count_io_running_write;
822extern int hammer_count_io_locked;
f5a07a7a 823extern int hammer_limit_dirtybufspace;
af209b0f 824extern int hammer_limit_iqueued;
47637bff 825extern int hammer_limit_recs;
1f07f686 826extern int hammer_bio_count;
cb51be26 827extern int hammer_verify_zone;
1b0ab2c3 828extern int hammer_verify_data;
cb51be26 829extern int hammer_write_mode;
21fde338 830extern int hammer_autoflush;
7d683b0f 831extern int64_t hammer_contention_count;
b3deaf57 832
cdb6e4e6
MD
833void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
834 int error, const char *msg);
8750964d
MD
835int hammer_vop_inactive(struct vop_inactive_args *);
836int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 837int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 838struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 839 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
840 hammer_tid_t asof, u_int32_t localization,
841 int flags, int *errorp);
43c665ae
MD
842void hammer_scan_inode_snapshots(hammer_mount_t hmp,
843 hammer_inode_info_t iinfo,
844 int (*callback)(hammer_inode_t ip, void *data),
845 void *data);
66325755
MD
846void hammer_put_inode(struct hammer_inode *ip);
847void hammer_put_inode_ref(struct hammer_inode *ip);
4a2796f3 848void hammer_inode_waitreclaims(hammer_mount_t hmp);
66325755 849
8cd0a023 850int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
851int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
852
a89aec1b 853int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
7c19b529
MN
854int hammer_install_volume(hammer_mount_t hmp, const char *volname,
855 struct vnode *devvp);
1b0ab2c3 856int hammer_mountcheck_volumes(hammer_mount_t hmp);
8cd0a023 857
e469566b 858int hammer_mem_add(hammer_record_t record);
45a014dc 859int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 860int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 861int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 862int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
863int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
864 hammer_tid_t tid);
602c6cb8 865int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
866 hammer_tid_t delete_tid, u_int32_t delete_ts,
867 int track, int64_t *stat_bytes);
b3deaf57 868int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 869 hammer_inode_t ip);
fbc6e32a 870int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
871int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
872
8cd0a023 873hammer_record_t
11ad5ade 874 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 875void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 876void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 877void hammer_rel_mem_record(hammer_record_t record);
b84de5af 878
6a37e7e4 879int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 880int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 881int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 882int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 883int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 884void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
885int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
886 int index);
af209b0f 887void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
888int hammer_lock_ex_try(struct hammer_lock *lock);
889void hammer_lock_sh(struct hammer_lock *lock);
47637bff 890int hammer_lock_sh_try(struct hammer_lock *lock);
6a37e7e4
MD
891int hammer_lock_upgrade(struct hammer_lock *lock);
892void hammer_lock_downgrade(struct hammer_lock *lock);
b3bad96f 893int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 894void hammer_unlock(struct hammer_lock *lock);
66325755
MD
895void hammer_ref(struct hammer_lock *lock);
896void hammer_unref(struct hammer_lock *lock);
8cd0a023 897
2f85fa4d
MD
898void hammer_sync_lock_ex(hammer_transaction_t trans);
899void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 900int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
901void hammer_sync_unlock(hammer_transaction_t trans);
902
66325755 903u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 904void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
905void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
906u_int64_t hammer_timespec_to_time(struct timespec *ts);
bc6c1f13
MD
907int hammer_str_to_tid(const char *str, int *ispfsp,
908 hammer_tid_t *tidp, u_int32_t *localizationp);
909int hammer_is_atatext(const char *name, int len);
5de0c0e5 910hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip);
0729c8c8
MD
911void hammer_clear_objid(hammer_inode_t dip);
912void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 913
e8599db1 914int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 915 int bytes);
e8599db1 916void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 917enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 918int hammer_get_dtype(u_int8_t obj_type);
66325755 919u_int8_t hammer_get_obj_type(enum vtype vtype);
5e435c92
MD
920int64_t hammer_directory_namekey(hammer_inode_t dip, const void *name, int len,
921 u_int32_t *max_iterationsp);
e63644f0 922int hammer_nohistory(hammer_inode_t ip);
427e5fc6 923
36f82b23 924int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 925 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 926void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 927void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 928int hammer_recover_cursor(hammer_cursor_t cursor);
3f43fb33
MD
929void hammer_unlock_cursor(hammer_cursor_t cursor, int also_ip);
930int hammer_lock_cursor(hammer_cursor_t cursor, int also_ip);
931hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
932void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
933
934void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
935void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
936 int index);
937void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
938 int index);
939void hammer_cursor_inserted_element(hammer_node_t node, int index);
940void hammer_cursor_deleted_element(hammer_node_t node, int index);
8cd0a023
MD
941
942int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 943int hammer_btree_first(hammer_cursor_t cursor);
32c90105 944int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
945int hammer_btree_extract(hammer_cursor_t cursor, int flags);
946int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 947int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 948int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 949 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 950int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
951void hammer_btree_do_propagation(hammer_cursor_t cursor,
952 hammer_pseudofs_inmem_t pfsm,
602c6cb8 953 hammer_btree_leaf_elm_t leaf);
8cd0a023 954int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 955int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
956int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
957int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
958
2f85fa4d
MD
959int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
960 hammer_btree_elm_t elm);
b33e2cc0
MD
961int hammer_btree_lock_children(hammer_cursor_t cursor,
962 struct hammer_node_locklist **locklistp);
bac808fe
MD
963void hammer_btree_unlock_children(hammer_cursor_t cursor,
964 struct hammer_node_locklist **locklistp);
bcac4bbb 965int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
c82af904
MD
966hammer_node_t hammer_btree_get_parent(hammer_node_t node, int *parent_indexp,
967 int *errorp, int try_exclusive);
b33e2cc0 968
c0ade690
MD
969void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
970void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 971
47197d71
MD
972void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
973 int *errorp, struct hammer_buffer **bufferp);
974void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
975 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
976void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
977 int *errorp, struct hammer_buffer **bufferp);
978void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
979 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
980
981hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 982
8cd0a023 983hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 984 int32_t vol_no, int *errorp);
4a2796f3
MD
985hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
986 int bytes, int isnew, int *errorp);
1b0ab2c3
MD
987void hammer_sync_buffers(hammer_mount_t hmp,
988 hammer_off_t base_offset, int bytes);
989void hammer_del_buffers(hammer_mount_t hmp,
990 hammer_off_t base_offset,
0832c9bb 991 hammer_off_t zone2_offset, int bytes);
8cd0a023 992
fbc6e32a 993int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
994int hammer_ref_buffer(hammer_buffer_t buffer);
995void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
996
8cd0a023 997void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
998void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
999
513ca7d7
MD
1000int hammer_vfs_export(struct mount *mp, int op,
1001 const struct export_args *export);
19619882
MD
1002hammer_node_t hammer_get_node(hammer_mount_t hmp, hammer_off_t node_offset,
1003 int isnew, int *errorp);
740d8317 1004void hammer_ref_node(hammer_node_t node);
055f5ff8 1005hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
bcac4bbb 1006 hammer_node_cache_t cache, int *errorp);
8cd0a023 1007void hammer_rel_node(hammer_node_t node);
36f82b23
MD
1008void hammer_delete_node(hammer_transaction_t trans,
1009 hammer_node_t node);
bcac4bbb
MD
1010void hammer_cache_node(hammer_node_cache_t cache,
1011 hammer_node_t node);
1012void hammer_uncache_node(hammer_node_cache_t cache);
8cd0a023
MD
1013void hammer_flush_node(hammer_node_t node);
1014
427e5fc6
MD
1015void hammer_dup_buffer(struct hammer_buffer **bufferp,
1016 struct hammer_buffer *buffer);
36f82b23 1017hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
36f82b23 1018void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 1019 u_int16_t rec_type, hammer_off_t *data_offsetp,
bf686dbe
MD
1020 struct hammer_buffer **data_bufferp, int *errorp);
1021
059819e3
MD
1022int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
1023 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 1024
66325755 1025void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 1026void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 1027
36f82b23
MD
1028hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
1029 hammer_off_t owner, int *errorp);
1030void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 1031 hammer_off_t owner, int *errorp);
a7e9bef1 1032int hammer_checkspace(hammer_mount_t hmp, int slop);
36f82b23 1033hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 1034 int bytes, int *errorp);
0832c9bb
MD
1035hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
1036 int bytes, hammer_off_t *zone_offp, int *errorp);
5e435c92 1037void hammer_blockmap_reserve_undo(hammer_mount_t hmp, hammer_reserve_t resv,
1b0ab2c3 1038 hammer_off_t zone_offset, int bytes);
0832c9bb
MD
1039void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
1040 hammer_reserve_t resv);
cb51be26 1041void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1042void hammer_blockmap_free(hammer_transaction_t trans,
1043 hammer_off_t bmap_off, int bytes);
cdb6e4e6 1044int hammer_blockmap_finalize(hammer_transaction_t trans,
5e435c92 1045 hammer_reserve_t resv,
4a2796f3 1046 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1047int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1048 int *curp, int *errorp);
40043e7f
MD
1049hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1050 int *errorp);
bf686dbe
MD
1051hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1052 int *errorp);
06ad81ff
MD
1053int64_t hammer_undo_used(hammer_transaction_t trans);
1054int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686
MD
1055int64_t hammer_undo_max(hammer_mount_t hmp);
1056
8cd0a023
MD
1057void hammer_start_transaction(struct hammer_transaction *trans,
1058 struct hammer_mount *hmp);
36f82b23
MD
1059void hammer_simple_transaction(struct hammer_transaction *trans,
1060 struct hammer_mount *hmp);
b84de5af
MD
1061void hammer_start_transaction_fls(struct hammer_transaction *trans,
1062 struct hammer_mount *hmp);
1063void hammer_done_transaction(struct hammer_transaction *trans);
66325755 1064
47637bff 1065void hammer_modify_inode(hammer_inode_t ip, int flags);
f90dde4c 1066void hammer_flush_inode(hammer_inode_t ip, int flags);
cdb6e4e6 1067void hammer_flush_inode_done(hammer_inode_t ip, int error);
b84de5af
MD
1068void hammer_wait_inode(hammer_inode_t ip);
1069
8cd0a023
MD
1070int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1071 struct ucred *cred, struct hammer_inode *dip,
ea434b6f
MD
1072 hammer_pseudofs_inmem_t pfsm,
1073 struct hammer_inode **ipp);
a89aec1b 1074void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1075int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1076int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
cdb6e4e6 1077int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused);
51c35492 1078
02325004 1079int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip);
7b6ccb11 1080void hammer_test_inode(hammer_inode_t dip);
e8599db1 1081void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 1082
a89aec1b 1083int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1084 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1085 hammer_inode_t nip);
a89aec1b 1086int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1087 hammer_cursor_t cursor, hammer_inode_t dip,
1088 hammer_inode_t ip);
47637bff 1089hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1090 void *data, int bytes, int *errorp);
47637bff 1091int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1092int hammer_ip_add_record(struct hammer_transaction *trans,
1093 hammer_record_t record);
4e17f465 1094int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1095 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1096int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1097 int *countp);
1098int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1099 int64_t offset, void *data, int bytes);
36f82b23 1100int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1101int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1102hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1103 u_int32_t localization, int *errorp);
1104int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1105 hammer_pseudofs_inmem_t pfsm);
1106int hammer_save_pseudofs(hammer_transaction_t trans,
1107 hammer_pseudofs_inmem_t pfsm);
842e7a70 1108int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1109void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1110int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1111 struct ucred *cred);
1112
748efb59 1113void hammer_io_init(hammer_io_t io, hammer_volume_t volume,
10a5d1ba 1114 enum hammer_io_type type);
2f85fa4d
MD
1115int hammer_io_read(struct vnode *devvp, struct hammer_io *io,
1116 hammer_off_t limit);
66325755 1117int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
47637bff 1118void hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1119struct buf *hammer_io_release(struct hammer_io *io, int flush);
055f5ff8 1120void hammer_io_flush(struct hammer_io *io);
1b0ab2c3 1121void hammer_io_wait(struct hammer_io *io);
055f5ff8 1122void hammer_io_waitdep(struct hammer_io *io);
af209b0f 1123void hammer_io_wait_all(hammer_mount_t hmp, const char *ident);
1b0ab2c3
MD
1124int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio,
1125 hammer_btree_leaf_elm_t leaf);
1126int hammer_io_direct_write(hammer_mount_t hmp, hammer_record_t record,
1127 struct bio *bio);
1128void hammer_io_direct_wait(hammer_record_t record);
43c665ae 1129void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1130void hammer_io_write_interlock(hammer_io_t io);
1131void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1132void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1133void hammer_io_clear_modlist(struct hammer_io *io);
748efb59
MD
1134void hammer_io_flush_sync(hammer_mount_t hmp);
1135
36f82b23
MD
1136void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1137 void *base, int len);
1138void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1139 void *base, int len);
10a5d1ba
MD
1140void hammer_modify_volume_done(hammer_volume_t volume);
1141void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1142
36f82b23
MD
1143int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1144 struct hammer_ioc_reblock *reblock);
11ad5ade
MD
1145int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1146 struct hammer_ioc_prune *prune);
c82af904
MD
1147int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1148 struct hammer_ioc_mirror_rw *mirror);
1149int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1150 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1151int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1152 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1153int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1154 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1155int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1156 struct hammer_ioc_pseudofs_rw *pfs);
1157int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1158 struct hammer_ioc_pseudofs_rw *pfs);
1159int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1160 struct hammer_ioc_pseudofs_rw *pfs);
4889cbd4
MD
1161int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1162 struct hammer_ioc_pseudofs_rw *pfs);
bf686dbe 1163
855942b6 1164int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1165
059819e3
MD
1166void hammer_flusher_create(hammer_mount_t hmp);
1167void hammer_flusher_destroy(hammer_mount_t hmp);
1168void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1169int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1170int hammer_flusher_async_one(hammer_mount_t hmp);
93291532 1171void hammer_flusher_wait(hammer_mount_t hmp, int seq);
06ad81ff 1172int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1173int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1174int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1175void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1176void hammer_flusher_finalize(hammer_transaction_t trans, int final);
1b0ab2c3
MD
1177int hammer_flusher_haswork(hammer_mount_t hmp);
1178
059819e3 1179
f90dde4c 1180int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1181void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1182 hammer_volume_t root_volume, int final);
f90dde4c 1183
19619882
MD
1184void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1185void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1186void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1187
1188int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1189int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1190int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1191int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
77062c8a 1192void hkprintf(const char *ctl, ...);
a56cb012
MD
1193udev_t hammer_fsid_to_udev(uuid_t *uuid);
1194
19619882 1195
4a2796f3
MD
1196int hammer_blocksize(int64_t file_offset);
1197int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1198
055f5ff8 1199#endif
427e5fc6 1200
af209b0f
MD
1201static __inline void
1202hammer_wait_mem_record(hammer_record_t record)
1203{
1204 hammer_wait_mem_record_ident(record, "hmmwai");
1205}
1206
1207static __inline void
1208hammer_lock_ex(struct hammer_lock *lock)
1209{
1210 hammer_lock_ex_ident(lock, "hmrlck");
1211}
1212
bcac4bbb
MD
1213/*
1214 * Indicate that a B-Tree node is being modified.
1215 */
8cd0a023 1216static __inline void
36f82b23 1217hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1218{
36f82b23 1219 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1220}
1221
1222static __inline void
36f82b23 1223hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1224{
36f82b23
MD
1225 hammer_modify_buffer(trans, node->buffer,
1226 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1227}
1228
bf686dbe 1229static __inline void
36f82b23
MD
1230hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1231 void *base, int len)
bf686dbe 1232{
19619882
MD
1233 hammer_crc_t *crcptr;
1234
bf686dbe
MD
1235 KKASSERT((char *)base >= (char *)node->ondisk &&
1236 (char *)base + len <=
1237 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 1238 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1239 crcptr = &node->ondisk->crc;
1240 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1241 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1242}
1243
bcac4bbb
MD
1244/*
1245 * Indicate that the specified modifications have been completed.
1246 *
1247 * Do not try to generate the crc here, it's very expensive to do and a
1248 * sequence of insertions or deletions can result in many calls to this
1249 * function on the same node.
1250 */
10a5d1ba
MD
1251static __inline void
1252hammer_modify_node_done(hammer_node_t node)
1253{
bcac4bbb
MD
1254 node->flags |= HAMMER_NODE_CRCGOOD;
1255 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1256 node->flags |= HAMMER_NODE_NEEDSCRC;
1257 node->buffer->io.gencrc = 1;
1258 hammer_ref_node(node);
1259 }
10a5d1ba
MD
1260 hammer_modify_buffer_done(node->buffer);
1261}
1262
e8599db1
MD
1263#define hammer_modify_volume_field(trans, vol, field) \
1264 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1265 sizeof((vol)->ondisk->field))
1266
c9b9e29d
MD
1267#define hammer_modify_node_field(trans, node, field) \
1268 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1269 sizeof((node)->ondisk->field))
1270