Make two more changes to the ata request queue sorting code.
[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 *
748efb59 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.127 2008/08/29 20:19:08 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);
9b45155d 69MALLOC_DECLARE(M_HAMMER_INO);
8750964d 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;
a89aec1b 113 struct hammer_volume *rootvol;
66325755
MD
114};
115
8cd0a023
MD
116typedef struct hammer_transaction *hammer_transaction_t;
117
66325755
MD
118/*
119 * HAMMER locks
120 */
427e5fc6 121struct hammer_lock {
0b075555 122 int refs; /* active references delay writes */
0b075555 123 int lockcount; /* lock count for exclusive/shared access */
427e5fc6 124 int wanted;
4a2796f3 125 int exwanted; /* number of threads waiting for ex lock */
427e5fc6 126 struct thread *locktd;
8750964d
MD
127};
128
427e5fc6
MD
129static __inline int
130hammer_islocked(struct hammer_lock *lock)
131{
8cd0a023 132 return(lock->lockcount != 0);
427e5fc6
MD
133}
134
0b075555
MD
135static __inline int
136hammer_isactive(struct hammer_lock *lock)
137{
138 return(lock->refs != 0);
139}
140
427e5fc6
MD
141static __inline int
142hammer_islastref(struct hammer_lock *lock)
143{
144 return(lock->refs == 1);
145}
c60bb2c5 146
6a37e7e4 147/*
7aa3b8a6 148 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
149 */
150static __inline int
7aa3b8a6 151hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 152{
7aa3b8a6
MD
153 if (lock->lockcount > 0 && lock->locktd == td)
154 return(1);
155 return(0);
6a37e7e4
MD
156}
157
ec4e8497 158/*
1f07f686 159 * Flush state, used by various structures
ec4e8497 160 */
1f07f686
MD
161typedef enum hammer_inode_state {
162 HAMMER_FST_IDLE,
163 HAMMER_FST_SETUP,
164 HAMMER_FST_FLUSH
165} hammer_inode_state_t;
ec4e8497 166
1f07f686 167TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 168
5fa5c92f
MD
169/*
170 * Pseudo-filesystem extended data tracking
171 */
172struct hammer_pfs_rb_tree;
173struct hammer_pseudofs_inmem;
174RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
175RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
176 hammer_pfs_rb_compare, u_int32_t);
177
178struct hammer_pseudofs_inmem {
179 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
180 struct hammer_lock lock;
181 u_int32_t localization;
ea434b6f 182 hammer_tid_t create_tid;
842e7a70 183 int flags;
a56cb012 184 udev_t fsid_udev;
5fa5c92f
MD
185 struct hammer_pseudofs_data pfsd;
186};
187
188typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
189
842e7a70
MD
190#define HAMMER_PFSM_DELETED 0x0001
191
0729c8c8
MD
192/*
193 * Cache object ids. A fixed number of objid cache structures are
194 * created to reserve object id's for newly created files in multiples
195 * of 100,000, localized to a particular directory, and recycled as
196 * needed. This allows parallel create operations in different
197 * directories to retain fairly localized object ids which in turn
198 * improves reblocking performance and layout.
199 */
bf3b416b 200#define OBJID_CACHE_SIZE 1024
0729c8c8
MD
201#define OBJID_CACHE_BULK 100000
202
203typedef struct hammer_objid_cache {
204 TAILQ_ENTRY(hammer_objid_cache) entry;
205 struct hammer_inode *dip;
206 hammer_tid_t next_tid;
207 int count;
208} *hammer_objid_cache_t;
209
bcac4bbb
MD
210/*
211 * Associate an inode with a B-Tree node to cache search start positions
212 */
213typedef struct hammer_node_cache {
7a61b85d 214 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
215 struct hammer_node *node;
216 struct hammer_inode *ip;
217} *hammer_node_cache_t;
218
219TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
220
7a61b85d
MD
221/*
222 * Structure used to organize flush groups. Flush groups must be
223 * organized into chunks in order to avoid blowing out the UNDO FIFO.
224 * Without this a 'sync' could end up flushing 50,000 inodes in a single
225 * transaction.
226 */
227struct hammer_flush_group {
228 TAILQ_ENTRY(hammer_flush_group) flush_entry;
229 TAILQ_HEAD(, hammer_inode) flush_list;
230 int inode_count; /* inode load */
231 int total_count; /* record load */
232 int running; /* group is running */
233 int closed;
234 int refs;
235};
236
237typedef struct hammer_flush_group *hammer_flush_group_t;
238
239TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
240
8750964d 241/*
8cd0a023
MD
242 * Structure used to represent an inode in-memory.
243 *
244 * The record and data associated with an inode may be out of sync with
245 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
246 * clear).
247 *
248 * An inode may also hold a cache of unsynchronized records, used for
249 * database and directories only. Unsynchronized regular file data is
250 * stored in the buffer cache.
251 *
252 * NOTE: A file which is created and destroyed within the initial
253 * synchronization period can wind up not doing any disk I/O at all.
254 *
255 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
256 */
257struct hammer_ino_rb_tree;
258struct hammer_inode;
259RB_HEAD(hammer_ino_rb_tree, hammer_inode);
260RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
261 hammer_ino_rb_compare, hammer_inode_info_t);
262
263struct hammer_rec_rb_tree;
264struct hammer_record;
265RB_HEAD(hammer_rec_rb_tree, hammer_record);
266RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 267 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
268
269TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
270
271struct hammer_inode {
1f07f686
MD
272 RB_ENTRY(hammer_inode) rb_node;
273 hammer_inode_state_t flush_state;
7a61b85d 274 hammer_flush_group_t flush_group;
059819e3 275 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 276 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 277 int64_t obj_id; /* (key) object identifier */
b84de5af 278 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 279 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 280 struct hammer_mount *hmp;
0729c8c8 281 hammer_objid_cache_t objid_cache;
b84de5af
MD
282 int flags;
283 int error; /* flush error */
284 int cursor_ip_refs; /* sanity */
47637bff 285 int rsv_recs;
b84de5af 286 struct vnode *vp;
5fa5c92f 287 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
288 struct lockf advlock;
289 struct hammer_lock lock; /* sync copy interlock */
b84de5af 290 off_t trunc_off;
11ad5ade 291 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
292 struct hammer_inode_data ino_data; /* in-memory cache */
293 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
bcac4bbb 294 struct hammer_node_cache cache[2]; /* search initiate cache */
b84de5af
MD
295
296 /*
297 * When a demark is created to synchronize an inode to
298 * disk, certain fields are copied so the front-end VOPs
299 * can continue to run in parallel with the synchronization
300 * occuring in the background.
301 */
302 int sync_flags; /* to-sync flags cache */
303 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 304 off_t save_trunc_off; /* write optimization */
11ad5ade 305 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 306 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
307};
308
8cd0a023
MD
309typedef struct hammer_inode *hammer_inode_t;
310
66325755
MD
311#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
312
8cd0a023 313#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
ddfdf542 314 /* (not including atime/mtime) */
e63644f0 315#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
7b6ccb11 316#define HAMMER_INODE_CONN_DOWN 0x0004 /* include in downward recursion */
1f07f686 317#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 318#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 319#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 320#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 321#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 322#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 323#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 324#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 325#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
f153644d 326#define HAMMER_INODE_REFLUSH 0x2000 /* flush on dependancy / reflush */
9f5097dc 327#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 328#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 329
b84de5af 330#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 331#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 332#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 333#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
334#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
335#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
b84de5af 336
11ad5ade
MD
337#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \
338 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
339 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
340 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686
MD
341
342#define HAMMER_INODE_MODMASK_NOXDIRTY \
343 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 344
cb51be26 345#define HAMMER_FLUSH_GROUP_SIZE 64
66325755 346
f90dde4c 347#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 348#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 349
7bc5b8c2
MD
350/*
351 * Used by the inode reclaim code to pipeline reclaims and avoid
352 * blowing out kernel memory or letting the flusher get too far
353 * behind.
354 */
355struct hammer_reclaim {
356 TAILQ_ENTRY(hammer_reclaim) entry;
357 int okydoky;
358};
359
4a2796f3
MD
360#define HAMMER_RECLAIM_FLUSH 2000
361#define HAMMER_RECLAIM_WAIT 4000
9f5097dc 362
8750964d 363/*
1f07f686
MD
364 * Structure used to represent an unsynchronized record in-memory. These
365 * records typically represent directory entries. Only non-historical
366 * records are kept in-memory.
367 *
368 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
369 * on disk then neither are any records and the in-memory record tree
370 * represents the entire contents of the inode. If the inode is on disk
371 * then the on-disk B-Tree is scanned in parallel with the in-memory
372 * RB-Tree to synthesize the current state of the file.
373 *
1f07f686
MD
374 * Records are also used to enforce the ordering of directory create/delete
375 * operations. A new inode will not be flushed to disk unless its related
376 * directory entry is also being flushed at the same time. A directory entry
377 * will not be removed unless its related inode is also being removed at the
378 * same time.
8750964d 379 */
1f07f686 380typedef enum hammer_record_type {
e8599db1 381 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 382 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 383 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
384 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
385 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
386} hammer_record_type_t;
387
8cd0a023
MD
388struct hammer_record {
389 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
390 TAILQ_ENTRY(hammer_record) target_entry;
391 hammer_inode_state_t flush_state;
7a61b85d 392 hammer_flush_group_t flush_group;
1f07f686 393 hammer_record_type_t type;
a89aec1b 394 struct hammer_lock lock;
0832c9bb 395 struct hammer_reserve *resv;
8cd0a023 396 struct hammer_inode *ip;
1f07f686 397 struct hammer_inode *target_ip;
11ad5ade 398 struct hammer_btree_leaf_elm leaf;
8cd0a023 399 union hammer_data_ondisk *data;
8cd0a023 400 int flags;
e469566b 401 hammer_off_t zone2_offset; /* direct-write only */
8cd0a023 402};
8750964d 403
8cd0a023
MD
404typedef struct hammer_record *hammer_record_t;
405
d36ec43b
MD
406/*
407 * Record flags. Note that FE can only be set by the frontend if the
408 * record has not been interlocked by the backend w/ BE.
409 */
8cd0a023
MD
410#define HAMMER_RECF_ALLOCDATA 0x0001
411#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 412#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 413#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
1b0ab2c3 414#define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */
d36ec43b 415#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 416#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
1b0ab2c3
MD
417#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
418#define HAMMER_RECF_DIRECT_IO 0x0200 /* related direct I/O running*/
419#define HAMMER_RECF_DIRECT_WAIT 0x0400 /* related direct I/O running*/
e469566b 420#define HAMMER_RECF_DIRECT_INVAL 0x0800 /* buffer alias invalidation */
8cd0a023 421
602c6cb8
MD
422/*
423 * hammer_delete_at_cursor() flags
424 */
425#define HAMMER_DELETE_ADJUST 0x0001
426#define HAMMER_DELETE_DESTROY 0x0002
427
8cd0a023 428/*
47197d71 429 * In-memory structures representing on-disk structures.
8cd0a023 430 */
8750964d 431struct hammer_volume;
427e5fc6 432struct hammer_buffer;
8cd0a023 433struct hammer_node;
e8599db1 434struct hammer_undo;
0832c9bb
MD
435struct hammer_reserve;
436
8750964d 437RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 438RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 439RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 440RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 441RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
442
443RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
444 hammer_vol_rb_compare, int32_t);
427e5fc6 445RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 446 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 447RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 448 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
449RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
450 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
451RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
452 hammer_res_rb_compare, hammer_off_t);
8750964d 453
66325755
MD
454/*
455 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
456 *
457 * VOLUME - hammer_volume containing meta-data
458 * META_BUFFER - hammer_buffer containing meta-data
459 * DATA_BUFFER - hammer_buffer containing pure-data
460 *
461 * Dirty volume headers and dirty meta-data buffers are locked until the
462 * flusher can sequence them out. Dirty pure-data buffers can be written.
463 * Clean buffers can be passively released.
66325755 464 */
10a5d1ba
MD
465typedef enum hammer_io_type {
466 HAMMER_STRUCTURE_VOLUME,
467 HAMMER_STRUCTURE_META_BUFFER,
468 HAMMER_STRUCTURE_UNDO_BUFFER,
469 HAMMER_STRUCTURE_DATA_BUFFER
470} hammer_io_type_t;
66325755
MD
471
472union hammer_io_structure;
055f5ff8 473struct hammer_io;
66325755
MD
474
475struct worklist {
476 LIST_ENTRY(worklist) node;
477};
478
10a5d1ba
MD
479TAILQ_HEAD(hammer_io_list, hammer_io);
480typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 481
66325755 482struct hammer_io {
10a5d1ba
MD
483 struct worklist worklist;
484 struct hammer_lock lock;
485 enum hammer_io_type type;
486 struct hammer_mount *hmp;
748efb59 487 struct hammer_volume *volume;
10a5d1ba
MD
488 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
489 hammer_io_list_t mod_list;
490 struct buf *bp;
4a2796f3
MD
491 int64_t offset; /* zone-2 offset */
492 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
493 int loading; /* loading/unloading interlock */
494 int modify_refs;
495
66325755
MD
496 u_int modified : 1; /* bp's data was modified */
497 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
498 u_int running : 1; /* bp write IO in progress */
499 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 500 u_int validated : 1; /* ondisk has been validated */
b58c6388 501 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 502 u_int recovered : 1; /* has recovery ref */
9f5097dc 503 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 504 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 505 u_int gencrc : 1; /* crc needs to be generated */
cdb6e4e6 506 u_int ioerror : 1; /* abort on io-error */
66325755
MD
507};
508
8cd0a023
MD
509typedef struct hammer_io *hammer_io_t;
510
af209b0f
MD
511#define HAMMER_CLUSTER_SIZE (64 * 1024)
512#if HAMMER_CLUSTER_SIZE > MAXBSIZE
513#undef HAMMER_CLUSTER_SIZE
514#define HAMMER_CLUSTER_SIZE MAXBSIZE
515#endif
516#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
517
66325755 518/*
8cd0a023 519 * In-memory volume representing on-disk buffer
66325755 520 */
8750964d 521struct hammer_volume {
66325755 522 struct hammer_io io;
8750964d 523 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
524 struct hammer_volume_ondisk *ondisk;
525 int32_t vol_no;
fbc6e32a 526 int64_t nblocks; /* note: special calculation for statfs */
47197d71 527 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
528 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
529 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 530 char *vol_name;
8750964d 531 struct vnode *devvp;
427e5fc6 532 int vol_flags;
8750964d
MD
533};
534
8cd0a023
MD
535typedef struct hammer_volume *hammer_volume_t;
536
66325755 537/*
8cd0a023
MD
538 * In-memory buffer (other then volume, super-cluster, or cluster),
539 * representing an on-disk buffer.
66325755 540 */
427e5fc6 541struct hammer_buffer {
66325755 542 struct hammer_io io;
427e5fc6 543 RB_ENTRY(hammer_buffer) rb_node;
47197d71 544 void *ondisk;
34d829f7 545 hammer_off_t zoneX_offset;
0832c9bb 546 hammer_off_t zone2_offset;
cebe9493 547 struct hammer_reserve *resv;
8cd0a023 548 struct hammer_node_list clist;
66325755
MD
549};
550
8cd0a023
MD
551typedef struct hammer_buffer *hammer_buffer_t;
552
553/*
554 * In-memory B-Tree node, representing an on-disk B-Tree node.
555 *
556 * This is a hang-on structure which is backed by a hammer_buffer,
557 * indexed by a hammer_cluster, and used for fine-grained locking of
558 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
559 * can contain multiple nodes representing wildly disassociated portions
560 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
561 *
562 * This structure uses a cluster-relative index to reduce the number
563 * of layers required to access it, and also because all on-disk B-Tree
564 * references are cluster-relative offsets.
565 */
566struct hammer_node {
567 struct hammer_lock lock; /* node-by-node lock */
568 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
569 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 570 hammer_off_t node_offset; /* full offset spec */
40043e7f 571 struct hammer_mount *hmp;
8cd0a023
MD
572 struct hammer_buffer *buffer; /* backing buffer */
573 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 574 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 575 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 576 int flags;
740d8317 577 int loading; /* load interlock */
8cd0a023
MD
578};
579
b3deaf57
MD
580#define HAMMER_NODE_DELETED 0x0001
581#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
582#define HAMMER_NODE_CRCGOOD 0x0004
583#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 584#define HAMMER_NODE_NEEDSMIRROR 0x0010
b3deaf57 585
8cd0a023
MD
586typedef struct hammer_node *hammer_node_t;
587
b33e2cc0
MD
588/*
589 * List of locked nodes.
590 */
591struct hammer_node_locklist {
592 struct hammer_node_locklist *next;
593 hammer_node_t node;
594};
595
596typedef struct hammer_node_locklist *hammer_node_locklist_t;
597
598
8cd0a023
MD
599/*
600 * Common I/O management structure - embedded in in-memory structures
601 * which are backed by filesystem buffers.
602 */
66325755
MD
603union hammer_io_structure {
604 struct hammer_io io;
605 struct hammer_volume volume;
66325755 606 struct hammer_buffer buffer;
8750964d
MD
607};
608
055f5ff8
MD
609typedef union hammer_io_structure *hammer_io_structure_t;
610
cebe9493
MD
611/*
612 * The reserve structure prevents the blockmap from allocating
613 * out of a reserved bigblock. Such reservations are used by
614 * the direct-write mechanism.
615 *
616 * The structure is also used to hold off on reallocations of
617 * big blocks from the freemap until flush dependancies have
618 * been dealt with.
619 */
0832c9bb
MD
620struct hammer_reserve {
621 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
622 TAILQ_ENTRY(hammer_reserve) delay_entry;
623 int flush_group;
cb51be26 624 int flags;
0832c9bb 625 int refs;
cb51be26 626 int zone;
1b0ab2c3 627 int bytes_freed;
df301614 628 int append_off;
cebe9493 629 hammer_off_t zone_offset;
0832c9bb
MD
630};
631
632typedef struct hammer_reserve *hammer_reserve_t;
633
cb51be26
MD
634#define HAMMER_RESF_ONDELAY 0x0001
635
8cd0a023
MD
636#include "hammer_cursor.h"
637
e8599db1 638/*
cebe9493
MD
639 * The undo structure tracks recent undos to avoid laying down duplicate
640 * undos within a flush group, saving us a significant amount of overhead.
641 *
642 * This is strictly a heuristic.
e8599db1 643 */
7a61b85d
MD
644#define HAMMER_MAX_UNDOS 1024
645#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
646
647struct hammer_undo {
648 RB_ENTRY(hammer_undo) rb_node;
649 TAILQ_ENTRY(hammer_undo) lru_entry;
650 hammer_off_t offset;
651 int bytes;
652};
653
654typedef struct hammer_undo *hammer_undo_t;
655
af209b0f 656struct hammer_flusher_info;
7a61b85d 657TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
658
659struct hammer_flusher {
660 int signal; /* flusher thread sequencer */
661 int act; /* currently active flush group */
662 int done; /* set to act when complete */
663 int next; /* next flush group */
664 int group_lock; /* lock sequencing of the next flush */
665 int exiting; /* request master exit */
da2da375
MD
666 thread_t td; /* master flusher thread */
667 hammer_tid_t tid; /* last flushed transaction id */
668 int finalize_want; /* serialize finalization */
669 struct hammer_lock finalize_lock; /* serialize finalization */
670 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
671 struct hammer_flusher_info_list run_list;
672 struct hammer_flusher_info_list ready_list;
da2da375
MD
673};
674
8750964d
MD
675/*
676 * Internal hammer mount data structure
677 */
678struct hammer_mount {
679 struct mount *mp;
27ea2398 680 /*struct vnode *rootvp;*/
8750964d
MD
681 struct hammer_ino_rb_tree rb_inos_root;
682 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 683 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 684 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
685 struct hammer_res_rb_tree rb_resv_root;
686 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 687 struct hammer_pfs_rb_tree rb_pfsm_root;
8750964d 688 struct hammer_volume *rootvol;
47197d71
MD
689 struct hammer_base_elm root_btree_beg;
690 struct hammer_base_elm root_btree_end;
cdb6e4e6 691 int flags; /* HAMMER_MOUNT_xxx flags */
195c19a1
MD
692 int hflags;
693 int ronly;
694 int nvolumes;
d26d0ae9 695 int volume_iterator;
732a1697 696 int master_id; /* -1 or 0-15 - clustering and mirroring */
e63644f0 697 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 698 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 699 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 700 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 701 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
702 int last_newrecords;
703 int count_newrecords;
da2da375 704
9f5097dc
MD
705 int inode_reclaims; /* inodes pending reclaim by flusher */
706 int count_inodes; /* total number of inodes */
af209b0f 707 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
708
709 struct hammer_flusher flusher;
710
855942b6 711 u_int check_interrupt;
8750964d 712 uuid_t fsid;
10a5d1ba
MD
713 struct hammer_io_list volu_list; /* dirty undo buffers */
714 struct hammer_io_list undo_list; /* dirty undo buffers */
715 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 716 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
717 struct hammer_io_list meta_list; /* dirty meta bufs */
718 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a
MD
719 int locked_dirty_space; /* meta/volu count */
720 int io_running_space;
0729c8c8 721 int objid_cache_count;
cdb6e4e6
MD
722 int error; /* critical I/O error */
723 struct krate krate; /* rate limited kprintf */
ddfdf542 724 hammer_tid_t asof; /* snapshot mount */
4889cbd4
MD
725 hammer_tid_t next_tid;
726 hammer_tid_t flush_tid1; /* flusher tid sequencing */
727 hammer_tid_t flush_tid2; /* flusher tid sequencing */
e63644f0
MD
728 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
729
66325755 730 u_int32_t namekey_iterator;
513ca7d7 731 struct netexport export;
9480ff55 732 struct hammer_lock sync_lock;
c9b9e29d 733 struct hammer_lock free_lock;
d99d6bf5
MD
734 struct hammer_lock undo_lock;
735 struct hammer_lock blkmap_lock;
0729c8c8 736 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
737 struct hammer_undo undos[HAMMER_MAX_UNDOS];
738 int undo_alloc;
739 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 740 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 741 struct hammer_flush_group_list flush_group_list;
7b6ccb11 742 hammer_flush_group_t next_flush_group;
0729c8c8 743 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
7bc5b8c2 744 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
8750964d
MD
745};
746
8cd0a023 747typedef struct hammer_mount *hammer_mount_t;
8750964d 748
cdb6e4e6 749#define HAMMER_MOUNT_CRITICAL_ERROR 0x0001
9f5097dc 750
fbc6e32a
MD
751struct hammer_sync_info {
752 int error;
753 int waitfor;
754};
755
8750964d
MD
756#endif
757
a7e9bef1
MD
758/*
759 * checkspace slop (8MB chunks), higher numbers are more conservative.
760 */
93291532
MD
761#define HAMMER_CHKSPC_REBLOCK 25
762#define HAMMER_CHKSPC_MIRROR 20
763#define HAMMER_CHKSPC_WRITE 20
764#define HAMMER_CHKSPC_CREATE 20
765#define HAMMER_CHKSPC_REMOVE 10
7b6ccb11 766#define HAMMER_CHKSPC_EMERGENCY 0
a7e9bef1 767
8750964d
MD
768#if defined(_KERNEL)
769
770extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
771extern struct vop_ops hammer_spec_vops;
772extern struct vop_ops hammer_fifo_vops;
66325755 773extern struct bio_ops hammer_bioops;
427e5fc6 774
2f85fa4d 775extern int hammer_debug_io;
d5ef456e 776extern int hammer_debug_general;
77062c8a 777extern int hammer_debug_debug;
e8599db1 778extern int hammer_debug_inode;
7d683b0f 779extern int hammer_debug_locks;
b3deaf57 780extern int hammer_debug_btree;
d113fda1 781extern int hammer_debug_tid;
b33e2cc0 782extern int hammer_debug_recover;
46fe7ae1 783extern int hammer_debug_recover_faults;
1b0ab2c3 784extern int hammer_cluster_enable;
7a61b85d 785extern int hammer_count_fsyncs;
b3deaf57 786extern int hammer_count_inodes;
af209b0f 787extern int hammer_count_iqueued;
9f5097dc 788extern int hammer_count_reclaiming;
b3deaf57
MD
789extern int hammer_count_records;
790extern int hammer_count_record_datas;
791extern int hammer_count_volumes;
b3deaf57
MD
792extern int hammer_count_buffers;
793extern int hammer_count_nodes;
a7e9bef1 794extern int64_t hammer_count_extra_space_used;
cb51be26
MD
795extern int64_t hammer_stats_btree_lookups;
796extern int64_t hammer_stats_btree_searches;
797extern int64_t hammer_stats_btree_inserts;
798extern int64_t hammer_stats_btree_deletes;
799extern int64_t hammer_stats_btree_elements;
800extern int64_t hammer_stats_btree_splits;
801extern int64_t hammer_stats_btree_iterations;
802extern int64_t hammer_stats_record_iterations;
ce0138a6
MD
803extern int64_t hammer_stats_file_read;
804extern int64_t hammer_stats_file_write;
805extern int64_t hammer_stats_file_iopsr;
806extern int64_t hammer_stats_file_iopsw;
807extern int64_t hammer_stats_disk_read;
808extern int64_t hammer_stats_disk_write;
809extern int64_t hammer_stats_inode_flushes;
810extern int64_t hammer_stats_commits;
f5a07a7a 811extern int hammer_count_dirtybufspace;
a99b9ea2 812extern int hammer_count_refedbufs;
0832c9bb 813extern int hammer_count_reservations;
a99b9ea2
MD
814extern int hammer_count_io_running_read;
815extern int hammer_count_io_running_write;
816extern int hammer_count_io_locked;
f5a07a7a 817extern int hammer_limit_dirtybufspace;
af209b0f 818extern int hammer_limit_iqueued;
47637bff 819extern int hammer_limit_recs;
1f07f686 820extern int hammer_bio_count;
cb51be26 821extern int hammer_verify_zone;
1b0ab2c3 822extern int hammer_verify_data;
cb51be26 823extern int hammer_write_mode;
7d683b0f 824extern int64_t hammer_contention_count;
b3deaf57 825
cdb6e4e6
MD
826void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
827 int error, const char *msg);
8750964d
MD
828int hammer_vop_inactive(struct vop_inactive_args *);
829int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 830int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 831struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 832 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
833 hammer_tid_t asof, u_int32_t localization,
834 int flags, int *errorp);
43c665ae
MD
835void hammer_scan_inode_snapshots(hammer_mount_t hmp,
836 hammer_inode_info_t iinfo,
837 int (*callback)(hammer_inode_t ip, void *data),
838 void *data);
66325755
MD
839void hammer_put_inode(struct hammer_inode *ip);
840void hammer_put_inode_ref(struct hammer_inode *ip);
4a2796f3 841void hammer_inode_waitreclaims(hammer_mount_t hmp);
66325755 842
8cd0a023 843int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
844int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
845
a89aec1b 846int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
7c19b529
MN
847int hammer_install_volume(hammer_mount_t hmp, const char *volname,
848 struct vnode *devvp);
1b0ab2c3 849int hammer_mountcheck_volumes(hammer_mount_t hmp);
8cd0a023 850
e469566b 851int hammer_mem_add(hammer_record_t record);
45a014dc 852int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 853int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 854int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 855int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
856int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
857 hammer_tid_t tid);
602c6cb8 858int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
859 hammer_tid_t delete_tid, u_int32_t delete_ts,
860 int track, int64_t *stat_bytes);
b3deaf57 861int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 862 hammer_inode_t ip);
fbc6e32a 863int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
864int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
865
8cd0a023 866hammer_record_t
11ad5ade 867 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 868void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 869void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 870void hammer_rel_mem_record(hammer_record_t record);
b84de5af 871
6a37e7e4 872int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 873int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 874int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 875int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 876int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 877void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
878int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
879 int index);
af209b0f 880void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
881int hammer_lock_ex_try(struct hammer_lock *lock);
882void hammer_lock_sh(struct hammer_lock *lock);
47637bff 883int hammer_lock_sh_try(struct hammer_lock *lock);
6a37e7e4
MD
884int hammer_lock_upgrade(struct hammer_lock *lock);
885void hammer_lock_downgrade(struct hammer_lock *lock);
b3bad96f 886int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 887void hammer_unlock(struct hammer_lock *lock);
66325755
MD
888void hammer_ref(struct hammer_lock *lock);
889void hammer_unref(struct hammer_lock *lock);
8cd0a023 890
2f85fa4d
MD
891void hammer_sync_lock_ex(hammer_transaction_t trans);
892void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 893int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
894void hammer_sync_unlock(hammer_transaction_t trans);
895
66325755 896u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 897void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
898void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
899u_int64_t hammer_timespec_to_time(struct timespec *ts);
ea434b6f
MD
900hammer_tid_t hammer_str_to_tid(const char *str, int *ispfs,
901 u_int32_t *localizationp);
5de0c0e5 902hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip);
0729c8c8
MD
903void hammer_clear_objid(hammer_inode_t dip);
904void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 905
e8599db1 906int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 907 int bytes);
e8599db1 908void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 909enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 910int hammer_get_dtype(u_int8_t obj_type);
66325755 911u_int8_t hammer_get_obj_type(enum vtype vtype);
5a930e66 912int64_t hammer_directory_namekey(const void *name, int len);
e63644f0 913int hammer_nohistory(hammer_inode_t ip);
427e5fc6 914
36f82b23 915int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 916 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 917void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 918void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 919int hammer_recover_cursor(hammer_cursor_t cursor);
3f43fb33
MD
920void hammer_unlock_cursor(hammer_cursor_t cursor, int also_ip);
921int hammer_lock_cursor(hammer_cursor_t cursor, int also_ip);
922hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
923void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
924
925void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
926void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
927 int index);
928void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
929 int index);
930void hammer_cursor_inserted_element(hammer_node_t node, int index);
931void hammer_cursor_deleted_element(hammer_node_t node, int index);
8cd0a023
MD
932
933int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 934int hammer_btree_first(hammer_cursor_t cursor);
32c90105 935int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
936int hammer_btree_extract(hammer_cursor_t cursor, int flags);
937int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 938int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 939int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 940 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 941int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
942void hammer_btree_do_propagation(hammer_cursor_t cursor,
943 hammer_pseudofs_inmem_t pfsm,
602c6cb8 944 hammer_btree_leaf_elm_t leaf);
8cd0a023 945int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 946int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
947int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
948int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
949
2f85fa4d
MD
950int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
951 hammer_btree_elm_t elm);
b33e2cc0
MD
952int hammer_btree_lock_children(hammer_cursor_t cursor,
953 struct hammer_node_locklist **locklistp);
2f85fa4d 954void hammer_btree_unlock_children(struct hammer_node_locklist **locklistp);
bcac4bbb 955int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
c82af904
MD
956hammer_node_t hammer_btree_get_parent(hammer_node_t node, int *parent_indexp,
957 int *errorp, int try_exclusive);
b33e2cc0 958
c0ade690
MD
959void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
960void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 961
47197d71
MD
962void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
963 int *errorp, struct hammer_buffer **bufferp);
964void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
965 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
966void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
967 int *errorp, struct hammer_buffer **bufferp);
968void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
969 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
970
971hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 972
8cd0a023 973hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 974 int32_t vol_no, int *errorp);
4a2796f3
MD
975hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
976 int bytes, int isnew, int *errorp);
1b0ab2c3
MD
977void hammer_sync_buffers(hammer_mount_t hmp,
978 hammer_off_t base_offset, int bytes);
979void hammer_del_buffers(hammer_mount_t hmp,
980 hammer_off_t base_offset,
0832c9bb 981 hammer_off_t zone2_offset, int bytes);
8cd0a023 982
fbc6e32a 983int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
984int hammer_ref_buffer(hammer_buffer_t buffer);
985void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
986
8cd0a023 987void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
988void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
989
513ca7d7
MD
990int hammer_vfs_export(struct mount *mp, int op,
991 const struct export_args *export);
19619882
MD
992hammer_node_t hammer_get_node(hammer_mount_t hmp, hammer_off_t node_offset,
993 int isnew, int *errorp);
740d8317 994void hammer_ref_node(hammer_node_t node);
055f5ff8 995hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
bcac4bbb 996 hammer_node_cache_t cache, int *errorp);
8cd0a023 997void hammer_rel_node(hammer_node_t node);
36f82b23
MD
998void hammer_delete_node(hammer_transaction_t trans,
999 hammer_node_t node);
bcac4bbb
MD
1000void hammer_cache_node(hammer_node_cache_t cache,
1001 hammer_node_t node);
1002void hammer_uncache_node(hammer_node_cache_t cache);
8cd0a023
MD
1003void hammer_flush_node(hammer_node_t node);
1004
427e5fc6
MD
1005void hammer_dup_buffer(struct hammer_buffer **bufferp,
1006 struct hammer_buffer *buffer);
36f82b23 1007hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
36f82b23 1008void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 1009 u_int16_t rec_type, hammer_off_t *data_offsetp,
bf686dbe
MD
1010 struct hammer_buffer **data_bufferp, int *errorp);
1011
059819e3
MD
1012int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
1013 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 1014
66325755 1015void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 1016void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 1017
36f82b23
MD
1018hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
1019 hammer_off_t owner, int *errorp);
1020void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 1021 hammer_off_t owner, int *errorp);
a7e9bef1 1022int hammer_checkspace(hammer_mount_t hmp, int slop);
36f82b23 1023hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 1024 int bytes, int *errorp);
0832c9bb
MD
1025hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
1026 int bytes, hammer_off_t *zone_offp, int *errorp);
1b0ab2c3
MD
1027void hammer_blockmap_reserve_undo(hammer_reserve_t resv,
1028 hammer_off_t zone_offset, int bytes);
0832c9bb
MD
1029void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
1030 hammer_reserve_t resv);
cb51be26 1031void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1032void hammer_blockmap_free(hammer_transaction_t trans,
1033 hammer_off_t bmap_off, int bytes);
cdb6e4e6 1034int hammer_blockmap_finalize(hammer_transaction_t trans,
4a2796f3 1035 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1036int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1037 int *curp, int *errorp);
40043e7f
MD
1038hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1039 int *errorp);
bf686dbe
MD
1040hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1041 int *errorp);
06ad81ff
MD
1042int64_t hammer_undo_used(hammer_transaction_t trans);
1043int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686
MD
1044int64_t hammer_undo_max(hammer_mount_t hmp);
1045
8cd0a023
MD
1046void hammer_start_transaction(struct hammer_transaction *trans,
1047 struct hammer_mount *hmp);
36f82b23
MD
1048void hammer_simple_transaction(struct hammer_transaction *trans,
1049 struct hammer_mount *hmp);
b84de5af
MD
1050void hammer_start_transaction_fls(struct hammer_transaction *trans,
1051 struct hammer_mount *hmp);
1052void hammer_done_transaction(struct hammer_transaction *trans);
66325755 1053
47637bff 1054void hammer_modify_inode(hammer_inode_t ip, int flags);
f90dde4c 1055void hammer_flush_inode(hammer_inode_t ip, int flags);
cdb6e4e6 1056void hammer_flush_inode_done(hammer_inode_t ip, int error);
b84de5af
MD
1057void hammer_wait_inode(hammer_inode_t ip);
1058
8cd0a023
MD
1059int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1060 struct ucred *cred, struct hammer_inode *dip,
ea434b6f
MD
1061 hammer_pseudofs_inmem_t pfsm,
1062 struct hammer_inode **ipp);
a89aec1b 1063void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1064int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1065int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
cdb6e4e6 1066int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused);
51c35492 1067
02325004 1068int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip);
7b6ccb11 1069void hammer_test_inode(hammer_inode_t dip);
e8599db1 1070void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 1071
a89aec1b 1072int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1073 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1074 hammer_inode_t nip);
a89aec1b 1075int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1076 hammer_cursor_t cursor, hammer_inode_t dip,
1077 hammer_inode_t ip);
47637bff 1078hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1079 void *data, int bytes, int *errorp);
47637bff 1080int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1081int hammer_ip_add_record(struct hammer_transaction *trans,
1082 hammer_record_t record);
4e17f465 1083int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1084 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1085int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1086 int *countp);
1087int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1088 int64_t offset, void *data, int bytes);
36f82b23 1089int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1090int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1091hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1092 u_int32_t localization, int *errorp);
1093int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1094 hammer_pseudofs_inmem_t pfsm);
1095int hammer_save_pseudofs(hammer_transaction_t trans,
1096 hammer_pseudofs_inmem_t pfsm);
842e7a70 1097int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1098void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1099int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1100 struct ucred *cred);
1101
748efb59 1102void hammer_io_init(hammer_io_t io, hammer_volume_t volume,
10a5d1ba 1103 enum hammer_io_type type);
2f85fa4d
MD
1104int hammer_io_read(struct vnode *devvp, struct hammer_io *io,
1105 hammer_off_t limit);
66325755 1106int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
47637bff 1107void hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1108struct buf *hammer_io_release(struct hammer_io *io, int flush);
055f5ff8 1109void hammer_io_flush(struct hammer_io *io);
1b0ab2c3 1110void hammer_io_wait(struct hammer_io *io);
055f5ff8 1111void hammer_io_waitdep(struct hammer_io *io);
af209b0f 1112void hammer_io_wait_all(hammer_mount_t hmp, const char *ident);
1b0ab2c3
MD
1113int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio,
1114 hammer_btree_leaf_elm_t leaf);
1115int hammer_io_direct_write(hammer_mount_t hmp, hammer_record_t record,
1116 struct bio *bio);
1117void hammer_io_direct_wait(hammer_record_t record);
43c665ae 1118void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1119void hammer_io_write_interlock(hammer_io_t io);
1120void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1121void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1122void hammer_io_clear_modlist(struct hammer_io *io);
748efb59
MD
1123void hammer_io_flush_sync(hammer_mount_t hmp);
1124
36f82b23
MD
1125void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1126 void *base, int len);
1127void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1128 void *base, int len);
10a5d1ba
MD
1129void hammer_modify_volume_done(hammer_volume_t volume);
1130void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1131
36f82b23
MD
1132int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1133 struct hammer_ioc_reblock *reblock);
11ad5ade
MD
1134int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1135 struct hammer_ioc_prune *prune);
c82af904
MD
1136int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1137 struct hammer_ioc_mirror_rw *mirror);
1138int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1139 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1140int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1141 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1142int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1143 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1144int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1145 struct hammer_ioc_pseudofs_rw *pfs);
1146int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1147 struct hammer_ioc_pseudofs_rw *pfs);
1148int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1149 struct hammer_ioc_pseudofs_rw *pfs);
4889cbd4
MD
1150int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1151 struct hammer_ioc_pseudofs_rw *pfs);
bf686dbe 1152
855942b6 1153int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1154
059819e3
MD
1155void hammer_flusher_create(hammer_mount_t hmp);
1156void hammer_flusher_destroy(hammer_mount_t hmp);
1157void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1158int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1159int hammer_flusher_async_one(hammer_mount_t hmp);
93291532 1160void hammer_flusher_wait(hammer_mount_t hmp, int seq);
06ad81ff 1161int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1162int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1163int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1164void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1165void hammer_flusher_finalize(hammer_transaction_t trans, int final);
1b0ab2c3
MD
1166int hammer_flusher_haswork(hammer_mount_t hmp);
1167
059819e3 1168
f90dde4c 1169int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1170void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1171 hammer_volume_t root_volume, int final);
f90dde4c 1172
19619882
MD
1173void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1174void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1175void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1176
1177int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1178int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1179int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1180int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
77062c8a 1181void hkprintf(const char *ctl, ...);
a56cb012
MD
1182udev_t hammer_fsid_to_udev(uuid_t *uuid);
1183
19619882 1184
4a2796f3
MD
1185int hammer_blocksize(int64_t file_offset);
1186int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1187
055f5ff8 1188#endif
427e5fc6 1189
af209b0f
MD
1190static __inline void
1191hammer_wait_mem_record(hammer_record_t record)
1192{
1193 hammer_wait_mem_record_ident(record, "hmmwai");
1194}
1195
1196static __inline void
1197hammer_lock_ex(struct hammer_lock *lock)
1198{
1199 hammer_lock_ex_ident(lock, "hmrlck");
1200}
1201
bcac4bbb
MD
1202/*
1203 * Indicate that a B-Tree node is being modified.
1204 */
8cd0a023 1205static __inline void
36f82b23 1206hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1207{
36f82b23 1208 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1209}
1210
1211static __inline void
36f82b23 1212hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1213{
36f82b23
MD
1214 hammer_modify_buffer(trans, node->buffer,
1215 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1216}
1217
bf686dbe 1218static __inline void
36f82b23
MD
1219hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1220 void *base, int len)
bf686dbe 1221{
19619882
MD
1222 hammer_crc_t *crcptr;
1223
bf686dbe
MD
1224 KKASSERT((char *)base >= (char *)node->ondisk &&
1225 (char *)base + len <=
1226 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 1227 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1228 crcptr = &node->ondisk->crc;
1229 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1230 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1231}
1232
bcac4bbb
MD
1233/*
1234 * Indicate that the specified modifications have been completed.
1235 *
1236 * Do not try to generate the crc here, it's very expensive to do and a
1237 * sequence of insertions or deletions can result in many calls to this
1238 * function on the same node.
1239 */
10a5d1ba
MD
1240static __inline void
1241hammer_modify_node_done(hammer_node_t node)
1242{
bcac4bbb
MD
1243 node->flags |= HAMMER_NODE_CRCGOOD;
1244 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1245 node->flags |= HAMMER_NODE_NEEDSCRC;
1246 node->buffer->io.gencrc = 1;
1247 hammer_ref_node(node);
1248 }
10a5d1ba
MD
1249 hammer_modify_buffer_done(node->buffer);
1250}
1251
e8599db1
MD
1252#define hammer_modify_volume_field(trans, vol, field) \
1253 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1254 sizeof((vol)->ondisk->field))
1255
c9b9e29d
MD
1256#define hammer_modify_node_field(trans, node, field) \
1257 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1258 sizeof((node)->ondisk->field))
1259