HAMMER VFS - Remove the unused also_ip argument from the cursor API
[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>
895c1f85 52#include <sys/priv.h>
e63644f0 53#include <sys/stat.h>
427e5fc6 54#include <sys/globaldata.h>
66325755
MD
55#include <sys/lockf.h>
56#include <sys/buf.h>
8cd0a023 57#include <sys/queue.h>
bcac4bbb 58#include <sys/ktr.h>
66325755
MD
59#include <sys/globaldata.h>
60
61#include <sys/buf2.h>
855942b6 62#include <sys/signal2.h>
427e5fc6 63#include "hammer_disk.h"
8750964d 64#include "hammer_mount.h"
7dc57964 65#include "hammer_ioctl.h"
8750964d
MD
66
67#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
68
69MALLOC_DECLARE(M_HAMMER);
70
bcac4bbb
MD
71/*
72 * Kernel trace
73 */
74#if !defined(KTR_HAMMER)
75#define KTR_HAMMER KTR_ALL
76#endif
77KTR_INFO_MASTER_EXTERN(hammer);
78
79/*
80 * Misc structures
81 */
66325755
MD
82struct hammer_mount;
83
8750964d
MD
84/*
85 * Key structure used for custom RB tree inode lookups. This prototypes
86 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
87 */
88typedef struct hammer_inode_info {
513ca7d7 89 int64_t obj_id; /* (key) object identifier */
8750964d 90 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 91 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
92 union {
93 struct hammer_btree_leaf_elm *leaf;
94 } u;
8750964d
MD
95} *hammer_inode_info_t;
96
b84de5af
MD
97typedef enum hammer_transaction_type {
98 HAMMER_TRANS_RO,
99 HAMMER_TRANS_STD,
100 HAMMER_TRANS_FLS
101} hammer_transaction_type_t;
102
66325755
MD
103/*
104 * HAMMER Transaction tracking
105 */
106struct hammer_transaction {
b84de5af 107 hammer_transaction_type_t type;
66325755
MD
108 struct hammer_mount *hmp;
109 hammer_tid_t tid;
ddfdf542 110 u_int64_t time;
dd94f1b1 111 u_int32_t time32;
2f85fa4d 112 int sync_lock_refs;
21fde338 113 int flags;
a89aec1b 114 struct hammer_volume *rootvol;
66325755
MD
115};
116
8cd0a023
MD
117typedef struct hammer_transaction *hammer_transaction_t;
118
21fde338 119#define HAMMER_TRANSF_NEWINODE 0x0001
82010f9f 120#define HAMMER_TRANSF_DIDIO 0x0002
21fde338 121
66325755
MD
122/*
123 * HAMMER locks
124 */
427e5fc6 125struct hammer_lock {
0b075555 126 int refs; /* active references delay writes */
0b075555 127 int lockcount; /* lock count for exclusive/shared access */
427e5fc6 128 int wanted;
4a2796f3 129 int exwanted; /* number of threads waiting for ex lock */
427e5fc6 130 struct thread *locktd;
8750964d
MD
131};
132
427e5fc6
MD
133static __inline int
134hammer_islocked(struct hammer_lock *lock)
135{
8cd0a023 136 return(lock->lockcount != 0);
427e5fc6
MD
137}
138
0b075555
MD
139static __inline int
140hammer_isactive(struct hammer_lock *lock)
141{
142 return(lock->refs != 0);
143}
144
427e5fc6
MD
145static __inline int
146hammer_islastref(struct hammer_lock *lock)
147{
148 return(lock->refs == 1);
149}
c60bb2c5 150
6a37e7e4 151/*
7aa3b8a6 152 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
153 */
154static __inline int
7aa3b8a6 155hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 156{
7aa3b8a6
MD
157 if (lock->lockcount > 0 && lock->locktd == td)
158 return(1);
159 return(0);
6a37e7e4
MD
160}
161
ec4e8497 162/*
1f07f686 163 * Flush state, used by various structures
ec4e8497 164 */
1f07f686
MD
165typedef enum hammer_inode_state {
166 HAMMER_FST_IDLE,
167 HAMMER_FST_SETUP,
168 HAMMER_FST_FLUSH
169} hammer_inode_state_t;
ec4e8497 170
1f07f686 171TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 172
5fa5c92f
MD
173/*
174 * Pseudo-filesystem extended data tracking
175 */
176struct hammer_pfs_rb_tree;
177struct hammer_pseudofs_inmem;
178RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
179RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
180 hammer_pfs_rb_compare, u_int32_t);
181
182struct hammer_pseudofs_inmem {
183 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
184 struct hammer_lock lock;
185 u_int32_t localization;
ea434b6f 186 hammer_tid_t create_tid;
842e7a70 187 int flags;
a56cb012 188 udev_t fsid_udev;
5fa5c92f
MD
189 struct hammer_pseudofs_data pfsd;
190};
191
192typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
193
842e7a70
MD
194#define HAMMER_PFSM_DELETED 0x0001
195
0729c8c8
MD
196/*
197 * Cache object ids. A fixed number of objid cache structures are
198 * created to reserve object id's for newly created files in multiples
199 * of 100,000, localized to a particular directory, and recycled as
200 * needed. This allows parallel create operations in different
201 * directories to retain fairly localized object ids which in turn
202 * improves reblocking performance and layout.
203 */
bf3b416b 204#define OBJID_CACHE_SIZE 1024
0729c8c8
MD
205#define OBJID_CACHE_BULK 100000
206
207typedef struct hammer_objid_cache {
208 TAILQ_ENTRY(hammer_objid_cache) entry;
209 struct hammer_inode *dip;
210 hammer_tid_t next_tid;
211 int count;
212} *hammer_objid_cache_t;
213
bcac4bbb
MD
214/*
215 * Associate an inode with a B-Tree node to cache search start positions
216 */
217typedef struct hammer_node_cache {
7a61b85d 218 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
219 struct hammer_node *node;
220 struct hammer_inode *ip;
221} *hammer_node_cache_t;
222
223TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
224
7a61b85d
MD
225/*
226 * Structure used to organize flush groups. Flush groups must be
227 * organized into chunks in order to avoid blowing out the UNDO FIFO.
228 * Without this a 'sync' could end up flushing 50,000 inodes in a single
229 * transaction.
230 */
231struct hammer_flush_group {
232 TAILQ_ENTRY(hammer_flush_group) flush_entry;
233 TAILQ_HEAD(, hammer_inode) flush_list;
38b9080b 234 int unused01; /* inode load */
7a61b85d
MD
235 int total_count; /* record load */
236 int running; /* group is running */
237 int closed;
238 int refs;
239};
240
241typedef struct hammer_flush_group *hammer_flush_group_t;
242
243TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
244
8750964d 245/*
8cd0a023
MD
246 * Structure used to represent an inode in-memory.
247 *
248 * The record and data associated with an inode may be out of sync with
249 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
250 * clear).
251 *
252 * An inode may also hold a cache of unsynchronized records, used for
253 * database and directories only. Unsynchronized regular file data is
254 * stored in the buffer cache.
255 *
256 * NOTE: A file which is created and destroyed within the initial
257 * synchronization period can wind up not doing any disk I/O at all.
258 *
259 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
260 */
261struct hammer_ino_rb_tree;
262struct hammer_inode;
263RB_HEAD(hammer_ino_rb_tree, hammer_inode);
264RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
265 hammer_ino_rb_compare, hammer_inode_info_t);
266
267struct hammer_rec_rb_tree;
268struct hammer_record;
269RB_HEAD(hammer_rec_rb_tree, hammer_record);
270RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 271 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
272
273TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
274
275struct hammer_inode {
1f07f686
MD
276 RB_ENTRY(hammer_inode) rb_node;
277 hammer_inode_state_t flush_state;
7a61b85d 278 hammer_flush_group_t flush_group;
059819e3 279 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 280 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 281 int64_t obj_id; /* (key) object identifier */
b84de5af 282 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 283 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 284 struct hammer_mount *hmp;
0729c8c8 285 hammer_objid_cache_t objid_cache;
b84de5af
MD
286 int flags;
287 int error; /* flush error */
288 int cursor_ip_refs; /* sanity */
47637bff 289 int rsv_recs;
b84de5af 290 struct vnode *vp;
5fa5c92f 291 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
292 struct lockf advlock;
293 struct hammer_lock lock; /* sync copy interlock */
b84de5af 294 off_t trunc_off;
11ad5ade 295 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
296 struct hammer_inode_data ino_data; /* in-memory cache */
297 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
bcac4bbb 298 struct hammer_node_cache cache[2]; /* search initiate cache */
b84de5af
MD
299
300 /*
301 * When a demark is created to synchronize an inode to
302 * disk, certain fields are copied so the front-end VOPs
303 * can continue to run in parallel with the synchronization
304 * occuring in the background.
305 */
306 int sync_flags; /* to-sync flags cache */
307 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 308 off_t save_trunc_off; /* write optimization */
11ad5ade 309 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 310 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
311};
312
8cd0a023
MD
313typedef struct hammer_inode *hammer_inode_t;
314
66325755
MD
315#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
316
8cd0a023 317#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
ddfdf542 318 /* (not including atime/mtime) */
e63644f0 319#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
7b6ccb11 320#define HAMMER_INODE_CONN_DOWN 0x0004 /* include in downward recursion */
1f07f686 321#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 322#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 323#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 324#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 325#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 326#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 327#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 328#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 329#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
f153644d 330#define HAMMER_INODE_REFLUSH 0x2000 /* flush on dependancy / reflush */
9f5097dc 331#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 332#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 333
b84de5af 334#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 335#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 336#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 337#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
338#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
339#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
b84de5af 340
11ad5ade
MD
341#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \
342 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
343 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
344 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686
MD
345
346#define HAMMER_INODE_MODMASK_NOXDIRTY \
347 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 348
cb51be26 349#define HAMMER_FLUSH_GROUP_SIZE 64
66325755 350
f90dde4c 351#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 352#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 353
7bc5b8c2
MD
354/*
355 * Used by the inode reclaim code to pipeline reclaims and avoid
356 * blowing out kernel memory or letting the flusher get too far
82010f9f
MD
357 * behind. The reclaim wakes up when count reaches 0 or the
358 * timer expires.
7bc5b8c2
MD
359 */
360struct hammer_reclaim {
361 TAILQ_ENTRY(hammer_reclaim) entry;
82010f9f 362 int count;
7bc5b8c2
MD
363};
364
4a2796f3
MD
365#define HAMMER_RECLAIM_FLUSH 2000
366#define HAMMER_RECLAIM_WAIT 4000
9f5097dc 367
8750964d 368/*
1f07f686
MD
369 * Structure used to represent an unsynchronized record in-memory. These
370 * records typically represent directory entries. Only non-historical
371 * records are kept in-memory.
372 *
373 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
374 * on disk then neither are any records and the in-memory record tree
375 * represents the entire contents of the inode. If the inode is on disk
376 * then the on-disk B-Tree is scanned in parallel with the in-memory
377 * RB-Tree to synthesize the current state of the file.
378 *
1f07f686
MD
379 * Records are also used to enforce the ordering of directory create/delete
380 * operations. A new inode will not be flushed to disk unless its related
381 * directory entry is also being flushed at the same time. A directory entry
382 * will not be removed unless its related inode is also being removed at the
383 * same time.
8750964d 384 */
1f07f686 385typedef enum hammer_record_type {
e8599db1 386 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 387 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 388 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
389 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
390 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
391} hammer_record_type_t;
392
8cd0a023
MD
393struct hammer_record {
394 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
395 TAILQ_ENTRY(hammer_record) target_entry;
396 hammer_inode_state_t flush_state;
7a61b85d 397 hammer_flush_group_t flush_group;
1f07f686 398 hammer_record_type_t type;
a89aec1b 399 struct hammer_lock lock;
0832c9bb 400 struct hammer_reserve *resv;
8cd0a023 401 struct hammer_inode *ip;
1f07f686 402 struct hammer_inode *target_ip;
11ad5ade 403 struct hammer_btree_leaf_elm leaf;
8cd0a023 404 union hammer_data_ondisk *data;
8cd0a023 405 int flags;
e469566b 406 hammer_off_t zone2_offset; /* direct-write only */
8cd0a023 407};
8750964d 408
8cd0a023
MD
409typedef struct hammer_record *hammer_record_t;
410
d36ec43b
MD
411/*
412 * Record flags. Note that FE can only be set by the frontend if the
413 * record has not been interlocked by the backend w/ BE.
414 */
8cd0a023
MD
415#define HAMMER_RECF_ALLOCDATA 0x0001
416#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 417#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 418#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
1b0ab2c3 419#define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */
d36ec43b 420#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 421#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
1b0ab2c3
MD
422#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
423#define HAMMER_RECF_DIRECT_IO 0x0200 /* related direct I/O running*/
424#define HAMMER_RECF_DIRECT_WAIT 0x0400 /* related direct I/O running*/
e469566b 425#define HAMMER_RECF_DIRECT_INVAL 0x0800 /* buffer alias invalidation */
8cd0a023 426
602c6cb8
MD
427/*
428 * hammer_delete_at_cursor() flags
429 */
430#define HAMMER_DELETE_ADJUST 0x0001
431#define HAMMER_DELETE_DESTROY 0x0002
432
8cd0a023 433/*
47197d71 434 * In-memory structures representing on-disk structures.
8cd0a023 435 */
8750964d 436struct hammer_volume;
427e5fc6 437struct hammer_buffer;
8cd0a023 438struct hammer_node;
e8599db1 439struct hammer_undo;
0832c9bb
MD
440struct hammer_reserve;
441
8750964d 442RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 443RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 444RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 445RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 446RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
447
448RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
449 hammer_vol_rb_compare, int32_t);
427e5fc6 450RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 451 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 452RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 453 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
454RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
455 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
456RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
457 hammer_res_rb_compare, hammer_off_t);
8750964d 458
66325755
MD
459/*
460 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
461 *
462 * VOLUME - hammer_volume containing meta-data
463 * META_BUFFER - hammer_buffer containing meta-data
464 * DATA_BUFFER - hammer_buffer containing pure-data
465 *
466 * Dirty volume headers and dirty meta-data buffers are locked until the
467 * flusher can sequence them out. Dirty pure-data buffers can be written.
468 * Clean buffers can be passively released.
66325755 469 */
10a5d1ba
MD
470typedef enum hammer_io_type {
471 HAMMER_STRUCTURE_VOLUME,
472 HAMMER_STRUCTURE_META_BUFFER,
473 HAMMER_STRUCTURE_UNDO_BUFFER,
474 HAMMER_STRUCTURE_DATA_BUFFER
475} hammer_io_type_t;
66325755
MD
476
477union hammer_io_structure;
055f5ff8 478struct hammer_io;
66325755
MD
479
480struct worklist {
481 LIST_ENTRY(worklist) node;
482};
483
10a5d1ba
MD
484TAILQ_HEAD(hammer_io_list, hammer_io);
485typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 486
66325755 487struct hammer_io {
10a5d1ba
MD
488 struct worklist worklist;
489 struct hammer_lock lock;
490 enum hammer_io_type type;
491 struct hammer_mount *hmp;
748efb59 492 struct hammer_volume *volume;
10a5d1ba
MD
493 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
494 hammer_io_list_t mod_list;
495 struct buf *bp;
4a2796f3
MD
496 int64_t offset; /* zone-2 offset */
497 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
498 int loading; /* loading/unloading interlock */
499 int modify_refs;
500
66325755
MD
501 u_int modified : 1; /* bp's data was modified */
502 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
503 u_int running : 1; /* bp write IO in progress */
504 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 505 u_int validated : 1; /* ondisk has been validated */
b58c6388 506 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 507 u_int recovered : 1; /* has recovery ref */
9f5097dc 508 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 509 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 510 u_int gencrc : 1; /* crc needs to be generated */
cdb6e4e6 511 u_int ioerror : 1; /* abort on io-error */
66325755
MD
512};
513
8cd0a023
MD
514typedef struct hammer_io *hammer_io_t;
515
af209b0f
MD
516#define HAMMER_CLUSTER_SIZE (64 * 1024)
517#if HAMMER_CLUSTER_SIZE > MAXBSIZE
518#undef HAMMER_CLUSTER_SIZE
519#define HAMMER_CLUSTER_SIZE MAXBSIZE
520#endif
521#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
522
66325755 523/*
8cd0a023 524 * In-memory volume representing on-disk buffer
66325755 525 */
8750964d 526struct hammer_volume {
66325755 527 struct hammer_io io;
8750964d 528 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
529 struct hammer_volume_ondisk *ondisk;
530 int32_t vol_no;
fbc6e32a 531 int64_t nblocks; /* note: special calculation for statfs */
47197d71 532 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
533 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
534 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 535 char *vol_name;
8750964d 536 struct vnode *devvp;
427e5fc6 537 int vol_flags;
8750964d
MD
538};
539
8cd0a023
MD
540typedef struct hammer_volume *hammer_volume_t;
541
66325755 542/*
8cd0a023
MD
543 * In-memory buffer (other then volume, super-cluster, or cluster),
544 * representing an on-disk buffer.
66325755 545 */
427e5fc6 546struct hammer_buffer {
66325755 547 struct hammer_io io;
427e5fc6 548 RB_ENTRY(hammer_buffer) rb_node;
47197d71 549 void *ondisk;
34d829f7 550 hammer_off_t zoneX_offset;
0832c9bb 551 hammer_off_t zone2_offset;
cebe9493 552 struct hammer_reserve *resv;
8cd0a023 553 struct hammer_node_list clist;
66325755
MD
554};
555
8cd0a023
MD
556typedef struct hammer_buffer *hammer_buffer_t;
557
558/*
559 * In-memory B-Tree node, representing an on-disk B-Tree node.
560 *
561 * This is a hang-on structure which is backed by a hammer_buffer,
562 * indexed by a hammer_cluster, and used for fine-grained locking of
563 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
564 * can contain multiple nodes representing wildly disassociated portions
565 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
566 *
567 * This structure uses a cluster-relative index to reduce the number
568 * of layers required to access it, and also because all on-disk B-Tree
569 * references are cluster-relative offsets.
570 */
571struct hammer_node {
572 struct hammer_lock lock; /* node-by-node lock */
573 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
574 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 575 hammer_off_t node_offset; /* full offset spec */
40043e7f 576 struct hammer_mount *hmp;
8cd0a023
MD
577 struct hammer_buffer *buffer; /* backing buffer */
578 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 579 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 580 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 581 int flags;
740d8317 582 int loading; /* load interlock */
8cd0a023
MD
583};
584
b3deaf57
MD
585#define HAMMER_NODE_DELETED 0x0001
586#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
587#define HAMMER_NODE_CRCGOOD 0x0004
588#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 589#define HAMMER_NODE_NEEDSMIRROR 0x0010
b3deaf57 590
8cd0a023
MD
591typedef struct hammer_node *hammer_node_t;
592
b33e2cc0
MD
593/*
594 * List of locked nodes.
595 */
596struct hammer_node_locklist {
597 struct hammer_node_locklist *next;
598 hammer_node_t node;
599};
600
601typedef struct hammer_node_locklist *hammer_node_locklist_t;
602
603
8cd0a023
MD
604/*
605 * Common I/O management structure - embedded in in-memory structures
606 * which are backed by filesystem buffers.
607 */
66325755
MD
608union hammer_io_structure {
609 struct hammer_io io;
610 struct hammer_volume volume;
66325755 611 struct hammer_buffer buffer;
8750964d
MD
612};
613
055f5ff8
MD
614typedef union hammer_io_structure *hammer_io_structure_t;
615
cebe9493
MD
616/*
617 * The reserve structure prevents the blockmap from allocating
618 * out of a reserved bigblock. Such reservations are used by
619 * the direct-write mechanism.
620 *
621 * The structure is also used to hold off on reallocations of
622 * big blocks from the freemap until flush dependancies have
623 * been dealt with.
624 */
0832c9bb
MD
625struct hammer_reserve {
626 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
627 TAILQ_ENTRY(hammer_reserve) delay_entry;
628 int flush_group;
cb51be26 629 int flags;
0832c9bb 630 int refs;
cb51be26 631 int zone;
df301614 632 int append_off;
cebe9493 633 hammer_off_t zone_offset;
0832c9bb
MD
634};
635
636typedef struct hammer_reserve *hammer_reserve_t;
637
cb51be26 638#define HAMMER_RESF_ONDELAY 0x0001
5e435c92 639#define HAMMER_RESF_LAYER2FREE 0x0002
cb51be26 640
8cd0a023
MD
641#include "hammer_cursor.h"
642
e8599db1 643/*
cebe9493
MD
644 * The undo structure tracks recent undos to avoid laying down duplicate
645 * undos within a flush group, saving us a significant amount of overhead.
646 *
647 * This is strictly a heuristic.
e8599db1 648 */
7a61b85d
MD
649#define HAMMER_MAX_UNDOS 1024
650#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
651
652struct hammer_undo {
653 RB_ENTRY(hammer_undo) rb_node;
654 TAILQ_ENTRY(hammer_undo) lru_entry;
655 hammer_off_t offset;
656 int bytes;
657};
658
659typedef struct hammer_undo *hammer_undo_t;
660
af209b0f 661struct hammer_flusher_info;
7a61b85d 662TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
663
664struct hammer_flusher {
665 int signal; /* flusher thread sequencer */
666 int act; /* currently active flush group */
667 int done; /* set to act when complete */
668 int next; /* next flush group */
669 int group_lock; /* lock sequencing of the next flush */
670 int exiting; /* request master exit */
da2da375
MD
671 thread_t td; /* master flusher thread */
672 hammer_tid_t tid; /* last flushed transaction id */
673 int finalize_want; /* serialize finalization */
674 struct hammer_lock finalize_lock; /* serialize finalization */
675 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
676 struct hammer_flusher_info_list run_list;
677 struct hammer_flusher_info_list ready_list;
da2da375
MD
678};
679
8750964d
MD
680/*
681 * Internal hammer mount data structure
682 */
683struct hammer_mount {
684 struct mount *mp;
27ea2398 685 /*struct vnode *rootvp;*/
8750964d
MD
686 struct hammer_ino_rb_tree rb_inos_root;
687 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 688 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 689 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
690 struct hammer_res_rb_tree rb_resv_root;
691 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 692 struct hammer_pfs_rb_tree rb_pfsm_root;
8750964d 693 struct hammer_volume *rootvol;
47197d71
MD
694 struct hammer_base_elm root_btree_beg;
695 struct hammer_base_elm root_btree_end;
bac808fe
MD
696
697 struct malloc_type *m_misc;
698 struct malloc_type *m_inodes;
699
cdb6e4e6 700 int flags; /* HAMMER_MOUNT_xxx flags */
195c19a1
MD
701 int hflags;
702 int ronly;
703 int nvolumes;
d26d0ae9 704 int volume_iterator;
732a1697 705 int master_id; /* -1 or 0-15 - clustering and mirroring */
44a83111 706 int version; /* hammer filesystem version to use */
e63644f0 707 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 708 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 709 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 710 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 711 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
712 int last_newrecords;
713 int count_newrecords;
da2da375 714
9f5097dc
MD
715 int inode_reclaims; /* inodes pending reclaim by flusher */
716 int count_inodes; /* total number of inodes */
af209b0f 717 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
718
719 struct hammer_flusher flusher;
720
855942b6 721 u_int check_interrupt;
8750964d 722 uuid_t fsid;
10a5d1ba
MD
723 struct hammer_io_list volu_list; /* dirty undo buffers */
724 struct hammer_io_list undo_list; /* dirty undo buffers */
725 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 726 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
727 struct hammer_io_list meta_list; /* dirty meta bufs */
728 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a
MD
729 int locked_dirty_space; /* meta/volu count */
730 int io_running_space;
0729c8c8 731 int objid_cache_count;
cdb6e4e6
MD
732 int error; /* critical I/O error */
733 struct krate krate; /* rate limited kprintf */
ddfdf542 734 hammer_tid_t asof; /* snapshot mount */
4889cbd4
MD
735 hammer_tid_t next_tid;
736 hammer_tid_t flush_tid1; /* flusher tid sequencing */
737 hammer_tid_t flush_tid2; /* flusher tid sequencing */
e63644f0
MD
738 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
739
513ca7d7 740 struct netexport export;
9480ff55 741 struct hammer_lock sync_lock;
c9b9e29d 742 struct hammer_lock free_lock;
d99d6bf5
MD
743 struct hammer_lock undo_lock;
744 struct hammer_lock blkmap_lock;
0729c8c8 745 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
746 struct hammer_undo undos[HAMMER_MAX_UNDOS];
747 int undo_alloc;
748 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 749 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 750 struct hammer_flush_group_list flush_group_list;
7b6ccb11 751 hammer_flush_group_t next_flush_group;
0729c8c8 752 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
7bc5b8c2 753 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
8750964d
MD
754};
755
8cd0a023 756typedef struct hammer_mount *hammer_mount_t;
8750964d 757
cdb6e4e6 758#define HAMMER_MOUNT_CRITICAL_ERROR 0x0001
82010f9f 759#define HAMMER_MOUNT_FLUSH_RECOVERY 0x0002
9f5097dc 760
fbc6e32a
MD
761struct hammer_sync_info {
762 int error;
763 int waitfor;
764};
765
8750964d
MD
766#endif
767
a7e9bef1
MD
768/*
769 * checkspace slop (8MB chunks), higher numbers are more conservative.
770 */
93291532
MD
771#define HAMMER_CHKSPC_REBLOCK 25
772#define HAMMER_CHKSPC_MIRROR 20
773#define HAMMER_CHKSPC_WRITE 20
774#define HAMMER_CHKSPC_CREATE 20
775#define HAMMER_CHKSPC_REMOVE 10
7b6ccb11 776#define HAMMER_CHKSPC_EMERGENCY 0
a7e9bef1 777
8750964d
MD
778#if defined(_KERNEL)
779
780extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
781extern struct vop_ops hammer_spec_vops;
782extern struct vop_ops hammer_fifo_vops;
66325755 783extern struct bio_ops hammer_bioops;
427e5fc6 784
2f85fa4d 785extern int hammer_debug_io;
d5ef456e 786extern int hammer_debug_general;
77062c8a 787extern int hammer_debug_debug;
e8599db1 788extern int hammer_debug_inode;
7d683b0f 789extern int hammer_debug_locks;
b3deaf57 790extern int hammer_debug_btree;
d113fda1 791extern int hammer_debug_tid;
b33e2cc0 792extern int hammer_debug_recover;
46fe7ae1 793extern int hammer_debug_recover_faults;
1b0ab2c3 794extern int hammer_cluster_enable;
7a61b85d 795extern int hammer_count_fsyncs;
b3deaf57 796extern int hammer_count_inodes;
af209b0f 797extern int hammer_count_iqueued;
9f5097dc 798extern int hammer_count_reclaiming;
b3deaf57
MD
799extern int hammer_count_records;
800extern int hammer_count_record_datas;
801extern int hammer_count_volumes;
b3deaf57
MD
802extern int hammer_count_buffers;
803extern int hammer_count_nodes;
a7e9bef1 804extern int64_t hammer_count_extra_space_used;
cb51be26
MD
805extern int64_t hammer_stats_btree_lookups;
806extern int64_t hammer_stats_btree_searches;
807extern int64_t hammer_stats_btree_inserts;
808extern int64_t hammer_stats_btree_deletes;
809extern int64_t hammer_stats_btree_elements;
810extern int64_t hammer_stats_btree_splits;
811extern int64_t hammer_stats_btree_iterations;
812extern int64_t hammer_stats_record_iterations;
ce0138a6
MD
813extern int64_t hammer_stats_file_read;
814extern int64_t hammer_stats_file_write;
815extern int64_t hammer_stats_file_iopsr;
816extern int64_t hammer_stats_file_iopsw;
817extern int64_t hammer_stats_disk_read;
818extern int64_t hammer_stats_disk_write;
819extern int64_t hammer_stats_inode_flushes;
820extern int64_t hammer_stats_commits;
f5a07a7a 821extern int hammer_count_dirtybufspace;
a99b9ea2 822extern int hammer_count_refedbufs;
0832c9bb 823extern int hammer_count_reservations;
a99b9ea2
MD
824extern int hammer_count_io_running_read;
825extern int hammer_count_io_running_write;
826extern int hammer_count_io_locked;
f5a07a7a 827extern int hammer_limit_dirtybufspace;
47637bff 828extern int hammer_limit_recs;
1f07f686 829extern int hammer_bio_count;
cb51be26 830extern int hammer_verify_zone;
1b0ab2c3 831extern int hammer_verify_data;
cb51be26 832extern int hammer_write_mode;
21fde338 833extern int hammer_autoflush;
7d683b0f 834extern int64_t hammer_contention_count;
b3deaf57 835
cdb6e4e6
MD
836void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
837 int error, const char *msg);
8750964d
MD
838int hammer_vop_inactive(struct vop_inactive_args *);
839int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 840int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 841struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 842 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
843 hammer_tid_t asof, u_int32_t localization,
844 int flags, int *errorp);
43c665ae
MD
845void hammer_scan_inode_snapshots(hammer_mount_t hmp,
846 hammer_inode_info_t iinfo,
847 int (*callback)(hammer_inode_t ip, void *data),
848 void *data);
66325755
MD
849void hammer_put_inode(struct hammer_inode *ip);
850void hammer_put_inode_ref(struct hammer_inode *ip);
4a2796f3 851void hammer_inode_waitreclaims(hammer_mount_t hmp);
82010f9f 852void hammer_inode_waithard(hammer_mount_t hmp);
66325755 853
8cd0a023 854int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
855int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
856
a89aec1b 857int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
7c19b529
MN
858int hammer_install_volume(hammer_mount_t hmp, const char *volname,
859 struct vnode *devvp);
1b0ab2c3 860int hammer_mountcheck_volumes(hammer_mount_t hmp);
8cd0a023 861
e469566b 862int hammer_mem_add(hammer_record_t record);
45a014dc 863int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 864int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 865int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 866int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
867int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
868 hammer_tid_t tid);
602c6cb8 869int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
870 hammer_tid_t delete_tid, u_int32_t delete_ts,
871 int track, int64_t *stat_bytes);
b3deaf57 872int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 873 hammer_inode_t ip);
fbc6e32a 874int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
875int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
876
8cd0a023 877hammer_record_t
11ad5ade 878 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 879void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 880void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 881void hammer_rel_mem_record(hammer_record_t record);
b84de5af 882
6a37e7e4 883int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 884int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 885int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 886int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 887int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 888void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
889int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
890 int index);
af209b0f 891void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
892int hammer_lock_ex_try(struct hammer_lock *lock);
893void hammer_lock_sh(struct hammer_lock *lock);
47637bff 894int hammer_lock_sh_try(struct hammer_lock *lock);
6a37e7e4
MD
895int hammer_lock_upgrade(struct hammer_lock *lock);
896void hammer_lock_downgrade(struct hammer_lock *lock);
b3bad96f 897int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 898void hammer_unlock(struct hammer_lock *lock);
66325755
MD
899void hammer_ref(struct hammer_lock *lock);
900void hammer_unref(struct hammer_lock *lock);
8cd0a023 901
2f85fa4d
MD
902void hammer_sync_lock_ex(hammer_transaction_t trans);
903void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 904int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
905void hammer_sync_unlock(hammer_transaction_t trans);
906
66325755 907u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 908void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
909void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
910u_int64_t hammer_timespec_to_time(struct timespec *ts);
bc6c1f13
MD
911int hammer_str_to_tid(const char *str, int *ispfsp,
912 hammer_tid_t *tidp, u_int32_t *localizationp);
913int hammer_is_atatext(const char *name, int len);
5de0c0e5 914hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip);
0729c8c8
MD
915void hammer_clear_objid(hammer_inode_t dip);
916void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 917
e8599db1 918int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 919 int bytes);
e8599db1 920void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 921enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 922int hammer_get_dtype(u_int8_t obj_type);
66325755 923u_int8_t hammer_get_obj_type(enum vtype vtype);
5e435c92
MD
924int64_t hammer_directory_namekey(hammer_inode_t dip, const void *name, int len,
925 u_int32_t *max_iterationsp);
e63644f0 926int hammer_nohistory(hammer_inode_t ip);
427e5fc6 927
36f82b23 928int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 929 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 930void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 931void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 932int hammer_recover_cursor(hammer_cursor_t cursor);
982be4bf
MD
933void hammer_unlock_cursor(hammer_cursor_t cursor);
934int hammer_lock_cursor(hammer_cursor_t cursor);
3f43fb33
MD
935hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
936void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
937
938void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
939void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
940 int index);
941void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
942 int index);
943void hammer_cursor_inserted_element(hammer_node_t node, int index);
944void hammer_cursor_deleted_element(hammer_node_t node, int index);
8cd0a023
MD
945
946int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 947int hammer_btree_first(hammer_cursor_t cursor);
32c90105 948int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
949int hammer_btree_extract(hammer_cursor_t cursor, int flags);
950int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 951int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 952int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 953 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 954int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
955void hammer_btree_do_propagation(hammer_cursor_t cursor,
956 hammer_pseudofs_inmem_t pfsm,
602c6cb8 957 hammer_btree_leaf_elm_t leaf);
8cd0a023 958int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 959int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
960int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
961int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
962
2f85fa4d
MD
963int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
964 hammer_btree_elm_t elm);
b33e2cc0
MD
965int hammer_btree_lock_children(hammer_cursor_t cursor,
966 struct hammer_node_locklist **locklistp);
bac808fe
MD
967void hammer_btree_unlock_children(hammer_cursor_t cursor,
968 struct hammer_node_locklist **locklistp);
bcac4bbb 969int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
82010f9f
MD
970hammer_node_t hammer_btree_get_parent(hammer_transaction_t trans,
971 hammer_node_t node, int *parent_indexp,
c82af904 972 int *errorp, int try_exclusive);
b33e2cc0 973
c0ade690
MD
974void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
975void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 976
47197d71
MD
977void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
978 int *errorp, struct hammer_buffer **bufferp);
979void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
980 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
981void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
982 int *errorp, struct hammer_buffer **bufferp);
983void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
984 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
985
986hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 987
8cd0a023 988hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 989 int32_t vol_no, int *errorp);
4a2796f3
MD
990hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
991 int bytes, int isnew, int *errorp);
1b0ab2c3
MD
992void hammer_sync_buffers(hammer_mount_t hmp,
993 hammer_off_t base_offset, int bytes);
994void hammer_del_buffers(hammer_mount_t hmp,
995 hammer_off_t base_offset,
0832c9bb 996 hammer_off_t zone2_offset, int bytes);
8cd0a023 997
fbc6e32a 998int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
999int hammer_ref_buffer(hammer_buffer_t buffer);
1000void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
1001
8cd0a023 1002void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
1003void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
1004
513ca7d7
MD
1005int hammer_vfs_export(struct mount *mp, int op,
1006 const struct export_args *export);
82010f9f
MD
1007hammer_node_t hammer_get_node(hammer_transaction_t trans,
1008 hammer_off_t node_offset, int isnew, int *errorp);
740d8317 1009void hammer_ref_node(hammer_node_t node);
055f5ff8 1010hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
bcac4bbb 1011 hammer_node_cache_t cache, int *errorp);
8cd0a023 1012void hammer_rel_node(hammer_node_t node);
36f82b23
MD
1013void hammer_delete_node(hammer_transaction_t trans,
1014 hammer_node_t node);
bcac4bbb
MD
1015void hammer_cache_node(hammer_node_cache_t cache,
1016 hammer_node_t node);
1017void hammer_uncache_node(hammer_node_cache_t cache);
8cd0a023
MD
1018void hammer_flush_node(hammer_node_t node);
1019
427e5fc6
MD
1020void hammer_dup_buffer(struct hammer_buffer **bufferp,
1021 struct hammer_buffer *buffer);
36f82b23 1022hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
36f82b23 1023void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 1024 u_int16_t rec_type, hammer_off_t *data_offsetp,
bf686dbe
MD
1025 struct hammer_buffer **data_bufferp, int *errorp);
1026
059819e3
MD
1027int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
1028 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 1029
66325755 1030void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 1031void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 1032
36f82b23
MD
1033hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
1034 hammer_off_t owner, int *errorp);
1035void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 1036 hammer_off_t owner, int *errorp);
a7e9bef1 1037int hammer_checkspace(hammer_mount_t hmp, int slop);
36f82b23 1038hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 1039 int bytes, int *errorp);
0832c9bb
MD
1040hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
1041 int bytes, hammer_off_t *zone_offp, int *errorp);
5e435c92 1042void hammer_blockmap_reserve_undo(hammer_mount_t hmp, hammer_reserve_t resv,
1b0ab2c3 1043 hammer_off_t zone_offset, int bytes);
0832c9bb
MD
1044void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
1045 hammer_reserve_t resv);
cb51be26 1046void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1047void hammer_blockmap_free(hammer_transaction_t trans,
1048 hammer_off_t bmap_off, int bytes);
cdb6e4e6 1049int hammer_blockmap_finalize(hammer_transaction_t trans,
5e435c92 1050 hammer_reserve_t resv,
4a2796f3 1051 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1052int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1053 int *curp, int *errorp);
40043e7f
MD
1054hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1055 int *errorp);
bf686dbe
MD
1056hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1057 int *errorp);
06ad81ff
MD
1058int64_t hammer_undo_used(hammer_transaction_t trans);
1059int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686
MD
1060int64_t hammer_undo_max(hammer_mount_t hmp);
1061
8cd0a023
MD
1062void hammer_start_transaction(struct hammer_transaction *trans,
1063 struct hammer_mount *hmp);
36f82b23
MD
1064void hammer_simple_transaction(struct hammer_transaction *trans,
1065 struct hammer_mount *hmp);
b84de5af
MD
1066void hammer_start_transaction_fls(struct hammer_transaction *trans,
1067 struct hammer_mount *hmp);
1068void hammer_done_transaction(struct hammer_transaction *trans);
66325755 1069
47637bff 1070void hammer_modify_inode(hammer_inode_t ip, int flags);
f90dde4c 1071void hammer_flush_inode(hammer_inode_t ip, int flags);
cdb6e4e6 1072void hammer_flush_inode_done(hammer_inode_t ip, int error);
b84de5af
MD
1073void hammer_wait_inode(hammer_inode_t ip);
1074
8cd0a023
MD
1075int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1076 struct ucred *cred, struct hammer_inode *dip,
ea434b6f
MD
1077 hammer_pseudofs_inmem_t pfsm,
1078 struct hammer_inode **ipp);
a89aec1b 1079void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1080int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1081int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
cdb6e4e6 1082int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused);
51c35492 1083
02325004 1084int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip);
7b6ccb11 1085void hammer_test_inode(hammer_inode_t dip);
e8599db1 1086void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 1087
a89aec1b 1088int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1089 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1090 hammer_inode_t nip);
a89aec1b 1091int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1092 hammer_cursor_t cursor, hammer_inode_t dip,
1093 hammer_inode_t ip);
47637bff 1094hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1095 void *data, int bytes, int *errorp);
47637bff 1096int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1097int hammer_ip_add_record(struct hammer_transaction *trans,
1098 hammer_record_t record);
4e17f465 1099int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1100 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1101int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1102 int *countp);
1103int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1104 int64_t offset, void *data, int bytes);
36f82b23 1105int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1106int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1107hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1108 u_int32_t localization, int *errorp);
1109int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1110 hammer_pseudofs_inmem_t pfsm);
1111int hammer_save_pseudofs(hammer_transaction_t trans,
1112 hammer_pseudofs_inmem_t pfsm);
842e7a70 1113int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1114void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1115int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1116 struct ucred *cred);
1117
748efb59 1118void hammer_io_init(hammer_io_t io, hammer_volume_t volume,
10a5d1ba 1119 enum hammer_io_type type);
2f85fa4d
MD
1120int hammer_io_read(struct vnode *devvp, struct hammer_io *io,
1121 hammer_off_t limit);
66325755 1122int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
47637bff 1123void hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1124struct buf *hammer_io_release(struct hammer_io *io, int flush);
055f5ff8 1125void hammer_io_flush(struct hammer_io *io);
1b0ab2c3 1126void hammer_io_wait(struct hammer_io *io);
055f5ff8 1127void hammer_io_waitdep(struct hammer_io *io);
af209b0f 1128void hammer_io_wait_all(hammer_mount_t hmp, const char *ident);
1b0ab2c3
MD
1129int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio,
1130 hammer_btree_leaf_elm_t leaf);
1131int hammer_io_direct_write(hammer_mount_t hmp, hammer_record_t record,
1132 struct bio *bio);
1133void hammer_io_direct_wait(hammer_record_t record);
43c665ae 1134void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1135void hammer_io_write_interlock(hammer_io_t io);
1136void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1137void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1138void hammer_io_clear_modlist(struct hammer_io *io);
748efb59
MD
1139void hammer_io_flush_sync(hammer_mount_t hmp);
1140
36f82b23
MD
1141void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1142 void *base, int len);
1143void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1144 void *base, int len);
10a5d1ba
MD
1145void hammer_modify_volume_done(hammer_volume_t volume);
1146void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1147
36f82b23
MD
1148int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1149 struct hammer_ioc_reblock *reblock);
11ad5ade
MD
1150int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1151 struct hammer_ioc_prune *prune);
c82af904
MD
1152int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1153 struct hammer_ioc_mirror_rw *mirror);
1154int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1155 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1156int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1157 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1158int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1159 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1160int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1161 struct hammer_ioc_pseudofs_rw *pfs);
1162int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1163 struct hammer_ioc_pseudofs_rw *pfs);
1164int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1165 struct hammer_ioc_pseudofs_rw *pfs);
4889cbd4
MD
1166int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1167 struct hammer_ioc_pseudofs_rw *pfs);
bf686dbe 1168
855942b6 1169int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1170
059819e3
MD
1171void hammer_flusher_create(hammer_mount_t hmp);
1172void hammer_flusher_destroy(hammer_mount_t hmp);
1173void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1174int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1175int hammer_flusher_async_one(hammer_mount_t hmp);
93291532 1176void hammer_flusher_wait(hammer_mount_t hmp, int seq);
82010f9f 1177void hammer_flusher_wait_next(hammer_mount_t hmp);
06ad81ff 1178int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1179int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1180int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1181void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1182void hammer_flusher_finalize(hammer_transaction_t trans, int final);
1b0ab2c3
MD
1183int hammer_flusher_haswork(hammer_mount_t hmp);
1184
059819e3 1185
f90dde4c 1186int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1187void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1188 hammer_volume_t root_volume, int final);
f90dde4c 1189
19619882
MD
1190void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1191void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1192void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1193
1194int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1195int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1196int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1197int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
77062c8a 1198void hkprintf(const char *ctl, ...);
a56cb012
MD
1199udev_t hammer_fsid_to_udev(uuid_t *uuid);
1200
19619882 1201
4a2796f3
MD
1202int hammer_blocksize(int64_t file_offset);
1203int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1204
055f5ff8 1205#endif
427e5fc6 1206
af209b0f
MD
1207static __inline void
1208hammer_wait_mem_record(hammer_record_t record)
1209{
1210 hammer_wait_mem_record_ident(record, "hmmwai");
1211}
1212
1213static __inline void
1214hammer_lock_ex(struct hammer_lock *lock)
1215{
1216 hammer_lock_ex_ident(lock, "hmrlck");
1217}
1218
bcac4bbb
MD
1219/*
1220 * Indicate that a B-Tree node is being modified.
1221 */
8cd0a023 1222static __inline void
36f82b23 1223hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1224{
36f82b23 1225 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1226}
1227
1228static __inline void
36f82b23 1229hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1230{
36f82b23
MD
1231 hammer_modify_buffer(trans, node->buffer,
1232 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1233}
1234
bf686dbe 1235static __inline void
36f82b23
MD
1236hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1237 void *base, int len)
bf686dbe 1238{
19619882
MD
1239 hammer_crc_t *crcptr;
1240
bf686dbe
MD
1241 KKASSERT((char *)base >= (char *)node->ondisk &&
1242 (char *)base + len <=
1243 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 1244 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1245 crcptr = &node->ondisk->crc;
1246 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1247 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1248}
1249
bcac4bbb
MD
1250/*
1251 * Indicate that the specified modifications have been completed.
1252 *
1253 * Do not try to generate the crc here, it's very expensive to do and a
1254 * sequence of insertions or deletions can result in many calls to this
1255 * function on the same node.
1256 */
10a5d1ba
MD
1257static __inline void
1258hammer_modify_node_done(hammer_node_t node)
1259{
bcac4bbb
MD
1260 node->flags |= HAMMER_NODE_CRCGOOD;
1261 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1262 node->flags |= HAMMER_NODE_NEEDSCRC;
1263 node->buffer->io.gencrc = 1;
1264 hammer_ref_node(node);
1265 }
10a5d1ba
MD
1266 hammer_modify_buffer_done(node->buffer);
1267}
1268
e8599db1
MD
1269#define hammer_modify_volume_field(trans, vol, field) \
1270 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1271 sizeof((vol)->ondisk->field))
1272
c9b9e29d
MD
1273#define hammer_modify_node_field(trans, node, field) \
1274 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1275 sizeof((node)->ondisk->field))
1276