HAMMER 60C/many: Mirroring
[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 *
602c6cb8 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.102 2008/07/04 07:25:36 dillon Exp $
8750964d
MD
35 */
36/*
37 * This header file contains structures used internally by the HAMMERFS
c60bb2c5 38 * implementation. See hammer_disk.h for on-disk structures.
8750964d
MD
39 */
40
427e5fc6
MD
41#include <sys/param.h>
42#include <sys/types.h>
43#include <sys/kernel.h>
42c7d26b 44#include <sys/conf.h>
427e5fc6 45#include <sys/systm.h>
8750964d
MD
46#include <sys/tree.h>
47#include <sys/malloc.h>
427e5fc6 48#include <sys/mount.h>
513ca7d7 49#include <sys/mountctl.h>
427e5fc6 50#include <sys/vnode.h>
42c7d26b 51#include <sys/proc.h>
e63644f0 52#include <sys/stat.h>
427e5fc6 53#include <sys/globaldata.h>
66325755
MD
54#include <sys/lockf.h>
55#include <sys/buf.h>
8cd0a023 56#include <sys/queue.h>
bcac4bbb 57#include <sys/ktr.h>
66325755
MD
58#include <sys/globaldata.h>
59
60#include <sys/buf2.h>
855942b6 61#include <sys/signal2.h>
427e5fc6 62#include "hammer_disk.h"
8750964d 63#include "hammer_mount.h"
7dc57964 64#include "hammer_ioctl.h"
8750964d
MD
65
66#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
67
68MALLOC_DECLARE(M_HAMMER);
69
bcac4bbb
MD
70/*
71 * Kernel trace
72 */
73#if !defined(KTR_HAMMER)
74#define KTR_HAMMER KTR_ALL
75#endif
76KTR_INFO_MASTER_EXTERN(hammer);
77
78/*
79 * Misc structures
80 */
66325755
MD
81struct hammer_mount;
82
8750964d
MD
83/*
84 * Key structure used for custom RB tree inode lookups. This prototypes
85 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
86 */
87typedef struct hammer_inode_info {
513ca7d7 88 int64_t obj_id; /* (key) object identifier */
8750964d 89 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 90 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
91 union {
92 struct hammer_btree_leaf_elm *leaf;
93 } u;
8750964d
MD
94} *hammer_inode_info_t;
95
b84de5af
MD
96typedef enum hammer_transaction_type {
97 HAMMER_TRANS_RO,
98 HAMMER_TRANS_STD,
99 HAMMER_TRANS_FLS
100} hammer_transaction_type_t;
101
66325755
MD
102/*
103 * HAMMER Transaction tracking
104 */
105struct hammer_transaction {
b84de5af 106 hammer_transaction_type_t type;
66325755
MD
107 struct hammer_mount *hmp;
108 hammer_tid_t tid;
ddfdf542 109 u_int64_t time;
dd94f1b1 110 u_int32_t time32;
2f85fa4d 111 int sync_lock_refs;
a89aec1b 112 struct hammer_volume *rootvol;
66325755
MD
113};
114
8cd0a023
MD
115typedef struct hammer_transaction *hammer_transaction_t;
116
66325755
MD
117/*
118 * HAMMER locks
119 */
427e5fc6 120struct hammer_lock {
0b075555 121 int refs; /* active references delay writes */
0b075555 122 int lockcount; /* lock count for exclusive/shared access */
427e5fc6 123 int wanted;
4a2796f3 124 int exwanted; /* number of threads waiting for ex lock */
427e5fc6 125 struct thread *locktd;
8750964d
MD
126};
127
427e5fc6
MD
128static __inline int
129hammer_islocked(struct hammer_lock *lock)
130{
8cd0a023 131 return(lock->lockcount != 0);
427e5fc6
MD
132}
133
0b075555
MD
134static __inline int
135hammer_isactive(struct hammer_lock *lock)
136{
137 return(lock->refs != 0);
138}
139
427e5fc6
MD
140static __inline int
141hammer_islastref(struct hammer_lock *lock)
142{
143 return(lock->refs == 1);
144}
c60bb2c5 145
6a37e7e4 146/*
7aa3b8a6 147 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
148 */
149static __inline int
7aa3b8a6 150hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 151{
7aa3b8a6
MD
152 if (lock->lockcount > 0 && lock->locktd == td)
153 return(1);
154 return(0);
6a37e7e4
MD
155}
156
ec4e8497 157/*
1f07f686 158 * Flush state, used by various structures
ec4e8497 159 */
1f07f686
MD
160typedef enum hammer_inode_state {
161 HAMMER_FST_IDLE,
162 HAMMER_FST_SETUP,
163 HAMMER_FST_FLUSH
164} hammer_inode_state_t;
ec4e8497 165
1f07f686 166TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 167
5fa5c92f
MD
168/*
169 * Pseudo-filesystem extended data tracking
170 */
171struct hammer_pfs_rb_tree;
172struct hammer_pseudofs_inmem;
173RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
174RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
175 hammer_pfs_rb_compare, u_int32_t);
176
177struct hammer_pseudofs_inmem {
178 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
179 struct hammer_lock lock;
180 u_int32_t localization;
181 struct hammer_pseudofs_data pfsd;
182};
183
184typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
185
0729c8c8
MD
186/*
187 * Cache object ids. A fixed number of objid cache structures are
188 * created to reserve object id's for newly created files in multiples
189 * of 100,000, localized to a particular directory, and recycled as
190 * needed. This allows parallel create operations in different
191 * directories to retain fairly localized object ids which in turn
192 * improves reblocking performance and layout.
193 */
bf3b416b 194#define OBJID_CACHE_SIZE 1024
0729c8c8
MD
195#define OBJID_CACHE_BULK 100000
196
197typedef struct hammer_objid_cache {
198 TAILQ_ENTRY(hammer_objid_cache) entry;
199 struct hammer_inode *dip;
200 hammer_tid_t next_tid;
201 int count;
202} *hammer_objid_cache_t;
203
bcac4bbb
MD
204/*
205 * Associate an inode with a B-Tree node to cache search start positions
206 */
207typedef struct hammer_node_cache {
208 TAILQ_ENTRY(hammer_node_cache) entry;
209 struct hammer_node *node;
210 struct hammer_inode *ip;
211} *hammer_node_cache_t;
212
213TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
214
8750964d 215/*
8cd0a023
MD
216 * Structure used to represent an inode in-memory.
217 *
218 * The record and data associated with an inode may be out of sync with
219 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
220 * clear).
221 *
222 * An inode may also hold a cache of unsynchronized records, used for
223 * database and directories only. Unsynchronized regular file data is
224 * stored in the buffer cache.
225 *
226 * NOTE: A file which is created and destroyed within the initial
227 * synchronization period can wind up not doing any disk I/O at all.
228 *
229 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
230 */
231struct hammer_ino_rb_tree;
232struct hammer_inode;
233RB_HEAD(hammer_ino_rb_tree, hammer_inode);
234RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
235 hammer_ino_rb_compare, hammer_inode_info_t);
236
237struct hammer_rec_rb_tree;
238struct hammer_record;
239RB_HEAD(hammer_rec_rb_tree, hammer_record);
240RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 241 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
242
243TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
244
245struct hammer_inode {
1f07f686
MD
246 RB_ENTRY(hammer_inode) rb_node;
247 hammer_inode_state_t flush_state;
248 int flush_group;
059819e3 249 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 250 struct hammer_record_list target_list; /* target of dependant recs */
b84de5af
MD
251 u_int64_t obj_id; /* (key) object identifier */
252 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 253 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 254 struct hammer_mount *hmp;
0729c8c8 255 hammer_objid_cache_t objid_cache;
b84de5af
MD
256 int flags;
257 int error; /* flush error */
258 int cursor_ip_refs; /* sanity */
47637bff 259 int rsv_recs;
b84de5af 260 struct vnode *vp;
5fa5c92f 261 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
262 struct lockf advlock;
263 struct hammer_lock lock; /* sync copy interlock */
b84de5af 264 off_t trunc_off;
11ad5ade 265 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
266 struct hammer_inode_data ino_data; /* in-memory cache */
267 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
bcac4bbb 268 struct hammer_node_cache cache[2]; /* search initiate cache */
b84de5af
MD
269
270 /*
271 * When a demark is created to synchronize an inode to
272 * disk, certain fields are copied so the front-end VOPs
273 * can continue to run in parallel with the synchronization
274 * occuring in the background.
275 */
276 int sync_flags; /* to-sync flags cache */
277 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 278 off_t save_trunc_off; /* write optimization */
11ad5ade 279 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 280 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
281};
282
8cd0a023
MD
283typedef struct hammer_inode *hammer_inode_t;
284
66325755
MD
285#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
286
8cd0a023 287#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
ddfdf542 288 /* (not including atime/mtime) */
e63644f0 289#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
5fa5c92f 290#define HAMMER_INODE_PFSD 0x0004 /* obj_asof set based on pfsd */
1f07f686 291#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 292#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 293#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 294#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 295#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 296#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 297#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 298#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 299#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
b84de5af 300#define HAMMER_INODE_REFLUSH 0x2000 /* pipelined flush during flush */
9f5097dc 301#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 302#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 303
b84de5af 304#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 305#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 306#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 307#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
308#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
309#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
b84de5af 310
11ad5ade
MD
311#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \
312 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
313 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
314 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686
MD
315
316#define HAMMER_INODE_MODMASK_NOXDIRTY \
317 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 318
cb51be26 319#define HAMMER_FLUSH_GROUP_SIZE 64
66325755 320
f90dde4c 321#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 322#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 323
7bc5b8c2
MD
324/*
325 * Used by the inode reclaim code to pipeline reclaims and avoid
326 * blowing out kernel memory or letting the flusher get too far
327 * behind.
328 */
329struct hammer_reclaim {
330 TAILQ_ENTRY(hammer_reclaim) entry;
331 int okydoky;
332};
333
4a2796f3
MD
334#define HAMMER_RECLAIM_FLUSH 2000
335#define HAMMER_RECLAIM_WAIT 4000
9f5097dc 336
8750964d 337/*
1f07f686
MD
338 * Structure used to represent an unsynchronized record in-memory. These
339 * records typically represent directory entries. Only non-historical
340 * records are kept in-memory.
341 *
342 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
343 * on disk then neither are any records and the in-memory record tree
344 * represents the entire contents of the inode. If the inode is on disk
345 * then the on-disk B-Tree is scanned in parallel with the in-memory
346 * RB-Tree to synthesize the current state of the file.
347 *
1f07f686
MD
348 * Records are also used to enforce the ordering of directory create/delete
349 * operations. A new inode will not be flushed to disk unless its related
350 * directory entry is also being flushed at the same time. A directory entry
351 * will not be removed unless its related inode is also being removed at the
352 * same time.
8750964d 353 */
1f07f686 354typedef enum hammer_record_type {
e8599db1 355 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 356 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 357 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
358 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
359 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
360} hammer_record_type_t;
361
8cd0a023
MD
362struct hammer_record {
363 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
364 TAILQ_ENTRY(hammer_record) target_entry;
365 hammer_inode_state_t flush_state;
366 int flush_group;
367 hammer_record_type_t type;
a89aec1b 368 struct hammer_lock lock;
0832c9bb 369 struct hammer_reserve *resv;
8cd0a023 370 struct hammer_inode *ip;
1f07f686 371 struct hammer_inode *target_ip;
11ad5ade 372 struct hammer_btree_leaf_elm leaf;
8cd0a023 373 union hammer_data_ondisk *data;
8cd0a023
MD
374 int flags;
375};
8750964d 376
8cd0a023
MD
377typedef struct hammer_record *hammer_record_t;
378
d36ec43b
MD
379/*
380 * Record flags. Note that FE can only be set by the frontend if the
381 * record has not been interlocked by the backend w/ BE.
382 */
8cd0a023
MD
383#define HAMMER_RECF_ALLOCDATA 0x0001
384#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 385#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 386#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
11ad5ade 387#define HAMMER_RECF_UNUSED0010 0x0010
d36ec43b 388#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 389#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
1f07f686 390#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
8cd0a023 391
602c6cb8
MD
392/*
393 * hammer_delete_at_cursor() flags
394 */
395#define HAMMER_DELETE_ADJUST 0x0001
396#define HAMMER_DELETE_DESTROY 0x0002
397
8cd0a023 398/*
47197d71 399 * In-memory structures representing on-disk structures.
8cd0a023 400 */
8750964d 401struct hammer_volume;
427e5fc6 402struct hammer_buffer;
8cd0a023 403struct hammer_node;
e8599db1 404struct hammer_undo;
0832c9bb
MD
405struct hammer_reserve;
406
8750964d 407RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 408RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 409RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 410RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 411RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
412
413RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
414 hammer_vol_rb_compare, int32_t);
427e5fc6 415RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 416 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 417RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 418 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
419RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
420 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
421RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
422 hammer_res_rb_compare, hammer_off_t);
8750964d 423
66325755
MD
424/*
425 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
426 *
427 * VOLUME - hammer_volume containing meta-data
428 * META_BUFFER - hammer_buffer containing meta-data
429 * DATA_BUFFER - hammer_buffer containing pure-data
430 *
431 * Dirty volume headers and dirty meta-data buffers are locked until the
432 * flusher can sequence them out. Dirty pure-data buffers can be written.
433 * Clean buffers can be passively released.
66325755 434 */
10a5d1ba
MD
435typedef enum hammer_io_type {
436 HAMMER_STRUCTURE_VOLUME,
437 HAMMER_STRUCTURE_META_BUFFER,
438 HAMMER_STRUCTURE_UNDO_BUFFER,
439 HAMMER_STRUCTURE_DATA_BUFFER
440} hammer_io_type_t;
66325755
MD
441
442union hammer_io_structure;
055f5ff8 443struct hammer_io;
66325755
MD
444
445struct worklist {
446 LIST_ENTRY(worklist) node;
447};
448
10a5d1ba
MD
449TAILQ_HEAD(hammer_io_list, hammer_io);
450typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 451
66325755 452struct hammer_io {
10a5d1ba
MD
453 struct worklist worklist;
454 struct hammer_lock lock;
455 enum hammer_io_type type;
456 struct hammer_mount *hmp;
457 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
458 hammer_io_list_t mod_list;
459 struct buf *bp;
4a2796f3
MD
460 int64_t offset; /* zone-2 offset */
461 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
462 int loading; /* loading/unloading interlock */
463 int modify_refs;
464
66325755
MD
465 u_int modified : 1; /* bp's data was modified */
466 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
467 u_int running : 1; /* bp write IO in progress */
468 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 469 u_int validated : 1; /* ondisk has been validated */
b58c6388 470 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 471 u_int recovered : 1; /* has recovery ref */
9f5097dc 472 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 473 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 474 u_int gencrc : 1; /* crc needs to be generated */
66325755
MD
475};
476
8cd0a023
MD
477typedef struct hammer_io *hammer_io_t;
478
af209b0f
MD
479#define HAMMER_CLUSTER_SIZE (64 * 1024)
480#if HAMMER_CLUSTER_SIZE > MAXBSIZE
481#undef HAMMER_CLUSTER_SIZE
482#define HAMMER_CLUSTER_SIZE MAXBSIZE
483#endif
484#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
485
66325755 486/*
8cd0a023 487 * In-memory volume representing on-disk buffer
66325755 488 */
8750964d 489struct hammer_volume {
66325755 490 struct hammer_io io;
8750964d 491 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
492 struct hammer_volume_ondisk *ondisk;
493 int32_t vol_no;
fbc6e32a 494 int64_t nblocks; /* note: special calculation for statfs */
47197d71 495 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
496 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
497 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 498 char *vol_name;
8750964d 499 struct vnode *devvp;
427e5fc6 500 int vol_flags;
8750964d
MD
501};
502
8cd0a023
MD
503typedef struct hammer_volume *hammer_volume_t;
504
66325755 505/*
8cd0a023
MD
506 * In-memory buffer (other then volume, super-cluster, or cluster),
507 * representing an on-disk buffer.
66325755 508 */
427e5fc6 509struct hammer_buffer {
66325755 510 struct hammer_io io;
427e5fc6 511 RB_ENTRY(hammer_buffer) rb_node;
47197d71 512 void *ondisk;
427e5fc6 513 struct hammer_volume *volume;
34d829f7 514 hammer_off_t zoneX_offset;
0832c9bb 515 hammer_off_t zone2_offset;
cebe9493 516 struct hammer_reserve *resv;
8cd0a023 517 struct hammer_node_list clist;
66325755
MD
518};
519
8cd0a023
MD
520typedef struct hammer_buffer *hammer_buffer_t;
521
522/*
523 * In-memory B-Tree node, representing an on-disk B-Tree node.
524 *
525 * This is a hang-on structure which is backed by a hammer_buffer,
526 * indexed by a hammer_cluster, and used for fine-grained locking of
527 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
528 * can contain multiple nodes representing wildly disassociated portions
529 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
530 *
531 * This structure uses a cluster-relative index to reduce the number
532 * of layers required to access it, and also because all on-disk B-Tree
533 * references are cluster-relative offsets.
534 */
535struct hammer_node {
536 struct hammer_lock lock; /* node-by-node lock */
537 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
538 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 539 hammer_off_t node_offset; /* full offset spec */
40043e7f 540 struct hammer_mount *hmp;
8cd0a023
MD
541 struct hammer_buffer *buffer; /* backing buffer */
542 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
bcac4bbb 543 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 544 int flags;
740d8317 545 int loading; /* load interlock */
8cd0a023
MD
546};
547
b3deaf57
MD
548#define HAMMER_NODE_DELETED 0x0001
549#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
550#define HAMMER_NODE_CRCGOOD 0x0004
551#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 552#define HAMMER_NODE_NEEDSMIRROR 0x0010
b3deaf57 553
8cd0a023
MD
554typedef struct hammer_node *hammer_node_t;
555
b33e2cc0
MD
556/*
557 * List of locked nodes.
558 */
559struct hammer_node_locklist {
560 struct hammer_node_locklist *next;
561 hammer_node_t node;
562};
563
564typedef struct hammer_node_locklist *hammer_node_locklist_t;
565
566
8cd0a023
MD
567/*
568 * Common I/O management structure - embedded in in-memory structures
569 * which are backed by filesystem buffers.
570 */
66325755
MD
571union hammer_io_structure {
572 struct hammer_io io;
573 struct hammer_volume volume;
66325755 574 struct hammer_buffer buffer;
8750964d
MD
575};
576
055f5ff8
MD
577typedef union hammer_io_structure *hammer_io_structure_t;
578
cebe9493
MD
579/*
580 * The reserve structure prevents the blockmap from allocating
581 * out of a reserved bigblock. Such reservations are used by
582 * the direct-write mechanism.
583 *
584 * The structure is also used to hold off on reallocations of
585 * big blocks from the freemap until flush dependancies have
586 * been dealt with.
587 */
0832c9bb
MD
588struct hammer_reserve {
589 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
590 TAILQ_ENTRY(hammer_reserve) delay_entry;
591 int flush_group;
cb51be26 592 int flags;
0832c9bb 593 int refs;
cb51be26 594 int zone;
df301614 595 int append_off;
cebe9493 596 hammer_off_t zone_offset;
0832c9bb
MD
597};
598
599typedef struct hammer_reserve *hammer_reserve_t;
600
cb51be26
MD
601#define HAMMER_RESF_ONDELAY 0x0001
602
8cd0a023
MD
603#include "hammer_cursor.h"
604
e8599db1 605/*
cebe9493
MD
606 * The undo structure tracks recent undos to avoid laying down duplicate
607 * undos within a flush group, saving us a significant amount of overhead.
608 *
609 * This is strictly a heuristic.
e8599db1 610 */
da2da375
MD
611#define HAMMER_MAX_UNDOS 1024
612#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
613
614struct hammer_undo {
615 RB_ENTRY(hammer_undo) rb_node;
616 TAILQ_ENTRY(hammer_undo) lru_entry;
617 hammer_off_t offset;
618 int bytes;
619};
620
621typedef struct hammer_undo *hammer_undo_t;
622
af209b0f 623struct hammer_flusher_info;
da2da375
MD
624
625struct hammer_flusher {
626 int signal; /* flusher thread sequencer */
627 int act; /* currently active flush group */
628 int done; /* set to act when complete */
629 int next; /* next flush group */
630 int group_lock; /* lock sequencing of the next flush */
631 int exiting; /* request master exit */
632 int count; /* number of slave flushers */
633 int running; /* number of slave flushers running */
634 thread_t td; /* master flusher thread */
635 hammer_tid_t tid; /* last flushed transaction id */
636 int finalize_want; /* serialize finalization */
637 struct hammer_lock finalize_lock; /* serialize finalization */
638 struct hammer_transaction trans; /* shared transaction */
639 struct hammer_flusher_info *info[HAMMER_MAX_FLUSHERS];
640};
641
8750964d
MD
642/*
643 * Internal hammer mount data structure
644 */
645struct hammer_mount {
646 struct mount *mp;
27ea2398 647 /*struct vnode *rootvp;*/
8750964d
MD
648 struct hammer_ino_rb_tree rb_inos_root;
649 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 650 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 651 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
652 struct hammer_res_rb_tree rb_resv_root;
653 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 654 struct hammer_pfs_rb_tree rb_pfsm_root;
8750964d 655 struct hammer_volume *rootvol;
47197d71
MD
656 struct hammer_base_elm root_btree_beg;
657 struct hammer_base_elm root_btree_end;
9f5097dc 658 int flags;
195c19a1
MD
659 int hflags;
660 int ronly;
661 int nvolumes;
d26d0ae9 662 int volume_iterator;
5de0c0e5 663 int masterid; /* -1 or 0-15 - clustering and mirroring */
e63644f0 664 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 665 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 666 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 667 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
4a2796f3
MD
668 int last_newrecords;
669 int count_newrecords;
da2da375 670
9f5097dc
MD
671 int inode_reclaims; /* inodes pending reclaim by flusher */
672 int count_inodes; /* total number of inodes */
af209b0f 673 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
674
675 struct hammer_flusher flusher;
676
855942b6 677 u_int check_interrupt;
8750964d 678 uuid_t fsid;
66325755 679 udev_t fsid_udev;
10a5d1ba
MD
680 struct hammer_io_list volu_list; /* dirty undo buffers */
681 struct hammer_io_list undo_list; /* dirty undo buffers */
682 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 683 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
684 struct hammer_io_list meta_list; /* dirty meta bufs */
685 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a
MD
686 int locked_dirty_space; /* meta/volu count */
687 int io_running_space;
0729c8c8 688 int objid_cache_count;
ddfdf542
MD
689 hammer_tid_t asof; /* snapshot mount */
690 hammer_off_t next_tid;
e63644f0
MD
691 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
692
66325755 693 u_int32_t namekey_iterator;
513ca7d7 694 struct netexport export;
9480ff55 695 struct hammer_lock sync_lock;
c9b9e29d 696 struct hammer_lock free_lock;
d99d6bf5
MD
697 struct hammer_lock undo_lock;
698 struct hammer_lock blkmap_lock;
0729c8c8 699 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
700 struct hammer_undo undos[HAMMER_MAX_UNDOS];
701 int undo_alloc;
702 TAILQ_HEAD(, hammer_undo) undo_lru_list;
059819e3 703 TAILQ_HEAD(, hammer_inode) flush_list;
cebe9493 704 TAILQ_HEAD(, hammer_reserve) delay_list;
0729c8c8 705 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
7bc5b8c2 706 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
8750964d
MD
707};
708
8cd0a023 709typedef struct hammer_mount *hammer_mount_t;
8750964d 710
7bc5b8c2 711#define HAMMER_MOUNT_UNUSED0001 0x0001
9f5097dc 712
fbc6e32a
MD
713struct hammer_sync_info {
714 int error;
715 int waitfor;
716};
717
8750964d
MD
718#endif
719
a7e9bef1
MD
720/*
721 * checkspace slop (8MB chunks), higher numbers are more conservative.
722 */
723#define HAMMER_CHECKSPACE_SLOP_REBLOCK 25
724#define HAMMER_CHECKSPACE_SLOP_WRITE 20
725#define HAMMER_CHECKSPACE_SLOP_CREATE 15
726#define HAMMER_CHECKSPACE_SLOP_REMOVE 10
727
8750964d
MD
728#if defined(_KERNEL)
729
730extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
731extern struct vop_ops hammer_spec_vops;
732extern struct vop_ops hammer_fifo_vops;
66325755 733extern struct bio_ops hammer_bioops;
427e5fc6 734
2f85fa4d 735extern int hammer_debug_io;
d5ef456e 736extern int hammer_debug_general;
77062c8a 737extern int hammer_debug_debug;
e8599db1 738extern int hammer_debug_inode;
7d683b0f 739extern int hammer_debug_locks;
b3deaf57 740extern int hammer_debug_btree;
d113fda1 741extern int hammer_debug_tid;
b33e2cc0 742extern int hammer_debug_recover;
46fe7ae1 743extern int hammer_debug_recover_faults;
a99b9ea2 744extern int hammer_debug_cluster_enable;
b3deaf57 745extern int hammer_count_inodes;
af209b0f 746extern int hammer_count_iqueued;
9f5097dc 747extern int hammer_count_reclaiming;
b3deaf57
MD
748extern int hammer_count_records;
749extern int hammer_count_record_datas;
750extern int hammer_count_volumes;
b3deaf57
MD
751extern int hammer_count_buffers;
752extern int hammer_count_nodes;
a7e9bef1 753extern int64_t hammer_count_extra_space_used;
cb51be26
MD
754extern int64_t hammer_stats_btree_lookups;
755extern int64_t hammer_stats_btree_searches;
756extern int64_t hammer_stats_btree_inserts;
757extern int64_t hammer_stats_btree_deletes;
758extern int64_t hammer_stats_btree_elements;
759extern int64_t hammer_stats_btree_splits;
760extern int64_t hammer_stats_btree_iterations;
761extern int64_t hammer_stats_record_iterations;
f5a07a7a 762extern int hammer_count_dirtybufspace;
a99b9ea2 763extern int hammer_count_refedbufs;
0832c9bb 764extern int hammer_count_reservations;
a99b9ea2
MD
765extern int hammer_count_io_running_read;
766extern int hammer_count_io_running_write;
767extern int hammer_count_io_locked;
f5a07a7a 768extern int hammer_limit_dirtybufspace;
af209b0f 769extern int hammer_limit_iqueued;
47637bff 770extern int hammer_limit_recs;
1f07f686 771extern int hammer_bio_count;
cb51be26
MD
772extern int hammer_verify_zone;
773extern int hammer_write_mode;
7d683b0f 774extern int64_t hammer_contention_count;
b3deaf57 775
8750964d
MD
776int hammer_vop_inactive(struct vop_inactive_args *);
777int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 778int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 779struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
bcac4bbb 780 hammer_inode_t dip, u_int64_t obj_id,
ddfdf542
MD
781 hammer_tid_t asof, u_int32_t localization,
782 int flags, int *errorp);
43c665ae
MD
783void hammer_scan_inode_snapshots(hammer_mount_t hmp,
784 hammer_inode_info_t iinfo,
785 int (*callback)(hammer_inode_t ip, void *data),
786 void *data);
66325755
MD
787void hammer_put_inode(struct hammer_inode *ip);
788void hammer_put_inode_ref(struct hammer_inode *ip);
4a2796f3 789void hammer_inode_waitreclaims(hammer_mount_t hmp);
66325755 790
8cd0a023 791int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
792int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
793
a89aec1b 794int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
8cd0a023
MD
795int hammer_install_volume(hammer_mount_t hmp, const char *volname);
796
45a014dc 797int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 798int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 799int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 800int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
801int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
802 hammer_tid_t tid);
602c6cb8
MD
803int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
804 int64_t *stat_bytes);
b3deaf57 805int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 806 hammer_inode_t ip);
fbc6e32a 807int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
808int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
809
8cd0a023 810hammer_record_t
11ad5ade 811 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 812void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 813void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 814void hammer_rel_mem_record(hammer_record_t record);
b84de5af 815
6a37e7e4 816int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 817int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 818int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 819int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 820int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 821void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
822int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
823 int index);
af209b0f 824void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
825int hammer_lock_ex_try(struct hammer_lock *lock);
826void hammer_lock_sh(struct hammer_lock *lock);
4a2796f3 827void hammer_lock_sh_lowpri(struct hammer_lock *lock);
47637bff 828int hammer_lock_sh_try(struct hammer_lock *lock);
6a37e7e4
MD
829int hammer_lock_upgrade(struct hammer_lock *lock);
830void hammer_lock_downgrade(struct hammer_lock *lock);
427e5fc6 831void hammer_unlock(struct hammer_lock *lock);
66325755
MD
832void hammer_ref(struct hammer_lock *lock);
833void hammer_unref(struct hammer_lock *lock);
8cd0a023 834
2f85fa4d
MD
835void hammer_sync_lock_ex(hammer_transaction_t trans);
836void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 837int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
838void hammer_sync_unlock(hammer_transaction_t trans);
839
66325755 840u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 841void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
842void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
843u_int64_t hammer_timespec_to_time(struct timespec *ts);
d113fda1 844hammer_tid_t hammer_str_to_tid(const char *str);
5de0c0e5 845hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip);
0729c8c8
MD
846void hammer_clear_objid(hammer_inode_t dip);
847void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 848
e8599db1
MD
849int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
850 int bytes);
851void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 852enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 853int hammer_get_dtype(u_int8_t obj_type);
66325755 854u_int8_t hammer_get_obj_type(enum vtype vtype);
5a930e66 855int64_t hammer_directory_namekey(const void *name, int len);
e63644f0 856int hammer_nohistory(hammer_inode_t ip);
427e5fc6 857
36f82b23 858int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
bcac4bbb 859 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465
MD
860int hammer_reinit_cursor(hammer_cursor_t cursor);
861void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023
MD
862void hammer_done_cursor(hammer_cursor_t cursor);
863
864int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 865int hammer_btree_first(hammer_cursor_t cursor);
32c90105 866int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
867int hammer_btree_extract(hammer_cursor_t cursor, int flags);
868int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 869int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 870int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 871 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 872int hammer_btree_delete(hammer_cursor_t cursor);
602c6cb8
MD
873void hammer_btree_do_propagation(hammer_cursor_t cursor, hammer_inode_t ip,
874 hammer_btree_leaf_elm_t leaf);
8cd0a023 875int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 876int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
877int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
878int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
879
2f85fa4d
MD
880int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
881 hammer_btree_elm_t elm);
b33e2cc0
MD
882int hammer_btree_lock_children(hammer_cursor_t cursor,
883 struct hammer_node_locklist **locklistp);
2f85fa4d 884void hammer_btree_unlock_children(struct hammer_node_locklist **locklistp);
bcac4bbb 885int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
c82af904
MD
886hammer_node_t hammer_btree_get_parent(hammer_node_t node, int *parent_indexp,
887 int *errorp, int try_exclusive);
b33e2cc0 888
c0ade690
MD
889void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
890void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 891
47197d71
MD
892void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
893 int *errorp, struct hammer_buffer **bufferp);
894void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
895 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
896void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
897 int *errorp, struct hammer_buffer **bufferp);
898void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
899 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
900
901hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 902
8cd0a023 903hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 904 int32_t vol_no, int *errorp);
4a2796f3
MD
905hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
906 int bytes, int isnew, int *errorp);
0832c9bb
MD
907void hammer_del_buffers(hammer_mount_t hmp, hammer_off_t base_offset,
908 hammer_off_t zone2_offset, int bytes);
8cd0a023 909
fbc6e32a 910int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
911int hammer_ref_buffer(hammer_buffer_t buffer);
912void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
913
8cd0a023 914void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
915void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
916
513ca7d7
MD
917int hammer_vfs_export(struct mount *mp, int op,
918 const struct export_args *export);
19619882
MD
919hammer_node_t hammer_get_node(hammer_mount_t hmp, hammer_off_t node_offset,
920 int isnew, int *errorp);
740d8317 921void hammer_ref_node(hammer_node_t node);
055f5ff8 922hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
bcac4bbb 923 hammer_node_cache_t cache, int *errorp);
8cd0a023 924void hammer_rel_node(hammer_node_t node);
36f82b23
MD
925void hammer_delete_node(hammer_transaction_t trans,
926 hammer_node_t node);
bcac4bbb
MD
927void hammer_cache_node(hammer_node_cache_t cache,
928 hammer_node_t node);
929void hammer_uncache_node(hammer_node_cache_t cache);
8cd0a023
MD
930void hammer_flush_node(hammer_node_t node);
931
427e5fc6
MD
932void hammer_dup_buffer(struct hammer_buffer **bufferp,
933 struct hammer_buffer *buffer);
36f82b23 934hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
36f82b23 935void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 936 u_int16_t rec_type, hammer_off_t *data_offsetp,
bf686dbe
MD
937 struct hammer_buffer **data_bufferp, int *errorp);
938
059819e3
MD
939int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
940 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 941
66325755 942void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 943void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 944
36f82b23
MD
945hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
946 hammer_off_t owner, int *errorp);
947void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 948 hammer_off_t owner, int *errorp);
a7e9bef1 949int hammer_checkspace(hammer_mount_t hmp, int slop);
36f82b23 950hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 951 int bytes, int *errorp);
0832c9bb
MD
952hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
953 int bytes, hammer_off_t *zone_offp, int *errorp);
954void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
955 hammer_reserve_t resv);
cb51be26 956void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
957void hammer_blockmap_free(hammer_transaction_t trans,
958 hammer_off_t bmap_off, int bytes);
4a2796f3
MD
959void hammer_blockmap_finalize(hammer_transaction_t trans,
960 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
961int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
962 int *curp, int *errorp);
40043e7f
MD
963hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
964 int *errorp);
bf686dbe
MD
965hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
966 int *errorp);
06ad81ff
MD
967int64_t hammer_undo_used(hammer_transaction_t trans);
968int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686
MD
969int64_t hammer_undo_max(hammer_mount_t hmp);
970
8cd0a023
MD
971void hammer_start_transaction(struct hammer_transaction *trans,
972 struct hammer_mount *hmp);
36f82b23
MD
973void hammer_simple_transaction(struct hammer_transaction *trans,
974 struct hammer_mount *hmp);
b84de5af
MD
975void hammer_start_transaction_fls(struct hammer_transaction *trans,
976 struct hammer_mount *hmp);
977void hammer_done_transaction(struct hammer_transaction *trans);
66325755 978
47637bff 979void hammer_modify_inode(hammer_inode_t ip, int flags);
f90dde4c 980void hammer_flush_inode(hammer_inode_t ip, int flags);
b84de5af
MD
981void hammer_flush_inode_done(hammer_inode_t ip);
982void hammer_wait_inode(hammer_inode_t ip);
983
8cd0a023
MD
984int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
985 struct ucred *cred, struct hammer_inode *dip,
5a930e66 986 int pseudofs, struct hammer_inode **ipp);
a89aec1b 987void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 988int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 989int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
51c35492 990
1f07f686
MD
991int hammer_sync_inode(hammer_inode_t ip);
992void hammer_test_inode(hammer_inode_t ip);
e8599db1 993void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 994
a89aec1b 995int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 996 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 997 hammer_inode_t nip);
a89aec1b 998int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
999 hammer_cursor_t cursor, hammer_inode_t dip,
1000 hammer_inode_t ip);
47637bff 1001hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1002 void *data, int bytes, int *errorp);
47637bff 1003int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1004int hammer_ip_add_record(struct hammer_transaction *trans,
1005 hammer_record_t record);
4e17f465 1006int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1007 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1008int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1009 int *countp);
1010int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1011 int64_t offset, void *data, int bytes);
36f82b23 1012int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1013int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
5fa5c92f
MD
1014int hammer_load_pseudofs(hammer_transaction_t trans, hammer_inode_t ip);
1015int hammer_save_pseudofs(hammer_transaction_t trans, hammer_inode_t ip);
1016void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1017int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1018 struct ucred *cred);
1019
10a5d1ba
MD
1020void hammer_io_init(hammer_io_t io, hammer_mount_t hmp,
1021 enum hammer_io_type type);
2f85fa4d
MD
1022int hammer_io_read(struct vnode *devvp, struct hammer_io *io,
1023 hammer_off_t limit);
66325755 1024int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
47637bff 1025void hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1026struct buf *hammer_io_release(struct hammer_io *io, int flush);
055f5ff8 1027void hammer_io_flush(struct hammer_io *io);
055f5ff8 1028void hammer_io_waitdep(struct hammer_io *io);
af209b0f 1029void hammer_io_wait_all(hammer_mount_t hmp, const char *ident);
4a2796f3 1030int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio);
47637bff
MD
1031int hammer_io_direct_write(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf,
1032 struct bio *bio);
43c665ae 1033void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1034void hammer_io_write_interlock(hammer_io_t io);
1035void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1036void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1037void hammer_io_clear_modlist(struct hammer_io *io);
36f82b23
MD
1038void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1039 void *base, int len);
1040void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1041 void *base, int len);
10a5d1ba
MD
1042void hammer_modify_volume_done(hammer_volume_t volume);
1043void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1044
36f82b23
MD
1045int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1046 struct hammer_ioc_reblock *reblock);
11ad5ade
MD
1047int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1048 struct hammer_ioc_prune *prune);
c82af904
MD
1049int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1050 struct hammer_ioc_mirror_rw *mirror);
1051int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1052 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f
MD
1053int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1054 struct hammer_ioc_pseudofs_rw *pfs);
1055int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1056 struct hammer_ioc_pseudofs_rw *pfs);
bf686dbe 1057
855942b6 1058int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1059
059819e3
MD
1060void hammer_flusher_create(hammer_mount_t hmp);
1061void hammer_flusher_destroy(hammer_mount_t hmp);
1062void hammer_flusher_sync(hammer_mount_t hmp);
1063void hammer_flusher_async(hammer_mount_t hmp);
06ad81ff
MD
1064int hammer_flusher_meta_limit(hammer_mount_t hmp);
1065int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1066void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
059819e3 1067
f90dde4c 1068int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1069void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1070 hammer_volume_t root_volume, int final);
f90dde4c 1071
19619882
MD
1072void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1073void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1074void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1075
1076int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1077int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1078int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1079int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
77062c8a 1080void hkprintf(const char *ctl, ...);
19619882 1081
4a2796f3
MD
1082int hammer_blocksize(int64_t file_offset);
1083int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1084
055f5ff8 1085#endif
427e5fc6 1086
af209b0f
MD
1087static __inline void
1088hammer_wait_mem_record(hammer_record_t record)
1089{
1090 hammer_wait_mem_record_ident(record, "hmmwai");
1091}
1092
1093static __inline void
1094hammer_lock_ex(struct hammer_lock *lock)
1095{
1096 hammer_lock_ex_ident(lock, "hmrlck");
1097}
1098
bcac4bbb
MD
1099/*
1100 * Indicate that a B-Tree node is being modified.
1101 */
8cd0a023 1102static __inline void
36f82b23 1103hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1104{
36f82b23 1105 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1106}
1107
1108static __inline void
36f82b23 1109hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1110{
36f82b23
MD
1111 hammer_modify_buffer(trans, node->buffer,
1112 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1113}
1114
bf686dbe 1115static __inline void
36f82b23
MD
1116hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1117 void *base, int len)
bf686dbe 1118{
19619882
MD
1119 hammer_crc_t *crcptr;
1120
bf686dbe
MD
1121 KKASSERT((char *)base >= (char *)node->ondisk &&
1122 (char *)base + len <=
1123 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 1124 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1125 crcptr = &node->ondisk->crc;
1126 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1127 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1128}
1129
bcac4bbb
MD
1130/*
1131 * Indicate that the specified modifications have been completed.
1132 *
1133 * Do not try to generate the crc here, it's very expensive to do and a
1134 * sequence of insertions or deletions can result in many calls to this
1135 * function on the same node.
1136 */
10a5d1ba
MD
1137static __inline void
1138hammer_modify_node_done(hammer_node_t node)
1139{
bcac4bbb
MD
1140 node->flags |= HAMMER_NODE_CRCGOOD;
1141 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1142 node->flags |= HAMMER_NODE_NEEDSCRC;
1143 node->buffer->io.gencrc = 1;
1144 hammer_ref_node(node);
1145 }
10a5d1ba
MD
1146 hammer_modify_buffer_done(node->buffer);
1147}
1148
e8599db1
MD
1149#define hammer_modify_volume_field(trans, vol, field) \
1150 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1151 sizeof((vol)->ondisk->field))
1152
c9b9e29d
MD
1153#define hammer_modify_node_field(trans, node, field) \
1154 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1155 sizeof((node)->ondisk->field))
1156