HAMMER - Add live dedup sysctl and support
[dragonfly.git] / sys / vfs / hammer / hammer.h
CommitLineData
8750964d 1/*
1f07f686 2 * Copyright (c) 2007-2008 The DragonFly Project. All rights reserved.
8750964d
MD
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
44a83111 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.130 2008/11/13 02:18:43 dillon Exp $
8750964d
MD
35 */
36/*
37 * This header file contains structures used internally by the HAMMERFS
c60bb2c5 38 * implementation. See hammer_disk.h for on-disk structures.
8750964d
MD
39 */
40
427e5fc6
MD
41#include <sys/param.h>
42#include <sys/types.h>
43#include <sys/kernel.h>
42c7d26b 44#include <sys/conf.h>
427e5fc6 45#include <sys/systm.h>
8750964d
MD
46#include <sys/tree.h>
47#include <sys/malloc.h>
427e5fc6 48#include <sys/mount.h>
513ca7d7 49#include <sys/mountctl.h>
427e5fc6 50#include <sys/vnode.h>
42c7d26b 51#include <sys/proc.h>
895c1f85 52#include <sys/priv.h>
e63644f0 53#include <sys/stat.h>
427e5fc6 54#include <sys/globaldata.h>
66325755
MD
55#include <sys/lockf.h>
56#include <sys/buf.h>
8cd0a023 57#include <sys/queue.h>
bcac4bbb 58#include <sys/ktr.h>
66325755 59#include <sys/globaldata.h>
9192654c 60#include <sys/limits.h>
c652be54 61#include <vm/vm_extern.h>
66325755
MD
62
63#include <sys/buf2.h>
855942b6 64#include <sys/signal2.h>
2247fe02 65#include <sys/mplock2.h>
427e5fc6 66#include "hammer_disk.h"
8750964d 67#include "hammer_mount.h"
7dc57964 68#include "hammer_ioctl.h"
8750964d
MD
69
70#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
71
72MALLOC_DECLARE(M_HAMMER);
73
bcac4bbb
MD
74/*
75 * Kernel trace
76 */
77#if !defined(KTR_HAMMER)
78#define KTR_HAMMER KTR_ALL
79#endif
80KTR_INFO_MASTER_EXTERN(hammer);
81
82/*
83 * Misc structures
84 */
66325755
MD
85struct hammer_mount;
86
8750964d
MD
87/*
88 * Key structure used for custom RB tree inode lookups. This prototypes
89 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
90 */
91typedef struct hammer_inode_info {
513ca7d7 92 int64_t obj_id; /* (key) object identifier */
8750964d 93 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 94 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
95 union {
96 struct hammer_btree_leaf_elm *leaf;
97 } u;
8750964d
MD
98} *hammer_inode_info_t;
99
b84de5af
MD
100typedef enum hammer_transaction_type {
101 HAMMER_TRANS_RO,
102 HAMMER_TRANS_STD,
103 HAMMER_TRANS_FLS
104} hammer_transaction_type_t;
105
66325755
MD
106/*
107 * HAMMER Transaction tracking
108 */
109struct hammer_transaction {
b84de5af 110 hammer_transaction_type_t type;
66325755
MD
111 struct hammer_mount *hmp;
112 hammer_tid_t tid;
ddfdf542 113 u_int64_t time;
dd94f1b1 114 u_int32_t time32;
2f85fa4d 115 int sync_lock_refs;
21fde338 116 int flags;
a89aec1b 117 struct hammer_volume *rootvol;
66325755
MD
118};
119
8cd0a023
MD
120typedef struct hammer_transaction *hammer_transaction_t;
121
21fde338 122#define HAMMER_TRANSF_NEWINODE 0x0001
82010f9f 123#define HAMMER_TRANSF_DIDIO 0x0002
4c286c36 124#define HAMMER_TRANSF_CRCDOM 0x0004 /* EDOM on CRC error, less critical */
21fde338 125
66325755
MD
126/*
127 * HAMMER locks
128 */
427e5fc6 129struct hammer_lock {
250aec18 130 volatile u_int refs; /* active references */
899eb297 131 volatile u_int lockval; /* lock count and control bits */
250aec18
MD
132 struct thread *lowner; /* owner if exclusively held */
133 struct thread *rowner; /* owner if exclusively held */
8750964d
MD
134};
135
250aec18
MD
136#define HAMMER_REFS_LOCKED 0x40000000 /* transition check */
137#define HAMMER_REFS_WANTED 0x20000000 /* transition check */
138#define HAMMER_REFS_CHECK 0x10000000 /* transition check */
139
140#define HAMMER_REFS_FLAGS (HAMMER_REFS_LOCKED | \
141 HAMMER_REFS_WANTED | \
142 HAMMER_REFS_CHECK)
143
899eb297 144#define HAMMER_LOCKF_EXCLUSIVE 0x40000000
250aec18 145#define HAMMER_LOCKF_WANTED 0x20000000
899eb297
MD
146
147static __inline int
148hammer_notlocked(struct hammer_lock *lock)
149{
150 return(lock->lockval == 0);
151}
152
427e5fc6
MD
153static __inline int
154hammer_islocked(struct hammer_lock *lock)
155{
899eb297 156 return(lock->lockval != 0);
427e5fc6
MD
157}
158
250aec18
MD
159/*
160 * Returns the number of refs on the object.
161 */
427e5fc6 162static __inline int
0b075555
MD
163hammer_isactive(struct hammer_lock *lock)
164{
250aec18
MD
165 return(lock->refs & ~HAMMER_REFS_FLAGS);
166}
167
168static __inline int
169hammer_oneref(struct hammer_lock *lock)
170{
171 return((lock->refs & ~HAMMER_REFS_FLAGS) == 1);
0b075555
MD
172}
173
174static __inline int
250aec18 175hammer_norefs(struct hammer_lock *lock)
427e5fc6 176{
250aec18
MD
177 return((lock->refs & ~HAMMER_REFS_FLAGS) == 0);
178}
179
180static __inline int
181hammer_norefsorlock(struct hammer_lock *lock)
182{
183 return(lock->refs == 0);
184}
185
186static __inline int
187hammer_refsorlock(struct hammer_lock *lock)
188{
189 return(lock->refs != 0);
427e5fc6 190}
c60bb2c5 191
8750964d 192/*
7aa3b8a6 193 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
194 */
195static __inline int
7aa3b8a6 196hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 197{
899eb297 198 if ((lock->lockval & HAMMER_LOCKF_EXCLUSIVE) &&
250aec18 199 lock->lowner == td) {
7aa3b8a6 200 return(1);
899eb297 201 }
7aa3b8a6 202 return(0);
6a37e7e4
MD
203}
204
205/*
1f07f686 206 * Flush state, used by various structures
ec4e8497 207 */
1f07f686
MD
208typedef enum hammer_inode_state {
209 HAMMER_FST_IDLE,
210 HAMMER_FST_SETUP,
211 HAMMER_FST_FLUSH
212} hammer_inode_state_t;
ec4e8497 213
1f07f686 214TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497
MD
215
216/*
5fa5c92f
MD
217 * Pseudo-filesystem extended data tracking
218 */
219struct hammer_pfs_rb_tree;
220struct hammer_pseudofs_inmem;
221RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
222RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
223 hammer_pfs_rb_compare, u_int32_t);
224
225struct hammer_pseudofs_inmem {
226 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
227 struct hammer_lock lock;
228 u_int32_t localization;
ea434b6f 229 hammer_tid_t create_tid;
842e7a70 230 int flags;
a56cb012 231 udev_t fsid_udev;
5fa5c92f
MD
232 struct hammer_pseudofs_data pfsd;
233};
234
235typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
236
842e7a70
MD
237#define HAMMER_PFSM_DELETED 0x0001
238
5fa5c92f 239/*
0729c8c8
MD
240 * Cache object ids. A fixed number of objid cache structures are
241 * created to reserve object id's for newly created files in multiples
242 * of 100,000, localized to a particular directory, and recycled as
243 * needed. This allows parallel create operations in different
244 * directories to retain fairly localized object ids which in turn
245 * improves reblocking performance and layout.
246 */
bf3b416b 247#define OBJID_CACHE_SIZE 1024
5a64efa1
MD
248#define OBJID_CACHE_BULK_BITS 10 /* 10 bits (1024) */
249#define OBJID_CACHE_BULK (32 * 32) /* two level (1024) */
250#define OBJID_CACHE_BULK_MASK (OBJID_CACHE_BULK - 1)
251#define OBJID_CACHE_BULK_MASK64 ((u_int64_t)(OBJID_CACHE_BULK - 1))
0729c8c8
MD
252
253typedef struct hammer_objid_cache {
254 TAILQ_ENTRY(hammer_objid_cache) entry;
255 struct hammer_inode *dip;
5a64efa1 256 hammer_tid_t base_tid;
0729c8c8 257 int count;
5a64efa1
MD
258 u_int32_t bm0;
259 u_int32_t bm1[32];
0729c8c8
MD
260} *hammer_objid_cache_t;
261
262/*
bcac4bbb
MD
263 * Associate an inode with a B-Tree node to cache search start positions
264 */
265typedef struct hammer_node_cache {
7a61b85d 266 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
267 struct hammer_node *node;
268 struct hammer_inode *ip;
269} *hammer_node_cache_t;
270
271TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
272
273/*
507df98a
ID
274 * Live dedup cache
275 */
276struct hammer_dedup_crc_rb_tree;
277RB_HEAD(hammer_dedup_crc_rb_tree, hammer_dedup_cache);
278RB_PROTOTYPE2(hammer_dedup_crc_rb_tree, hammer_dedup_cache, crc_entry,
279 hammer_dedup_crc_rb_compare, hammer_crc_t);
280
281struct hammer_dedup_off_rb_tree;
282RB_HEAD(hammer_dedup_off_rb_tree, hammer_dedup_cache);
283RB_PROTOTYPE2(hammer_dedup_off_rb_tree, hammer_dedup_cache, off_entry,
284 hammer_dedup_off_rb_compare, hammer_off_t);
285
286#define DEDUP_CACHE_SIZE 4096 /* XXX make it a dynamic tunable */
287
288typedef struct hammer_dedup_cache {
289 RB_ENTRY(hammer_dedup_cache) crc_entry;
290 RB_ENTRY(hammer_dedup_cache) off_entry;
291 TAILQ_ENTRY(hammer_dedup_cache) lru_entry;
292 struct hammer_mount *hmp;
293 int64_t obj_id;
294 u_int32_t localization;
295 off_t file_offset;
296 int bytes;
297 hammer_off_t data_offset;
298 hammer_crc_t crc;
299} *hammer_dedup_cache_t;
300
301/*
7a61b85d
MD
302 * Structure used to organize flush groups. Flush groups must be
303 * organized into chunks in order to avoid blowing out the UNDO FIFO.
304 * Without this a 'sync' could end up flushing 50,000 inodes in a single
305 * transaction.
306 */
ff003b11
MD
307struct hammer_fls_rb_tree;
308RB_HEAD(hammer_fls_rb_tree, hammer_inode);
309RB_PROTOTYPE(hammer_fls_rb_tree, hammer_inode, rb_flsnode,
310 hammer_ino_rb_compare);
311
7a61b85d
MD
312struct hammer_flush_group {
313 TAILQ_ENTRY(hammer_flush_group) flush_entry;
ff003b11
MD
314 struct hammer_fls_rb_tree flush_tree;
315 int unused01;
7a61b85d
MD
316 int total_count; /* record load */
317 int running; /* group is running */
318 int closed;
319 int refs;
320};
321
322typedef struct hammer_flush_group *hammer_flush_group_t;
323
324TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
325
326/*
8cd0a023
MD
327 * Structure used to represent an inode in-memory.
328 *
329 * The record and data associated with an inode may be out of sync with
330 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
331 * clear).
332 *
333 * An inode may also hold a cache of unsynchronized records, used for
334 * database and directories only. Unsynchronized regular file data is
335 * stored in the buffer cache.
336 *
337 * NOTE: A file which is created and destroyed within the initial
338 * synchronization period can wind up not doing any disk I/O at all.
339 *
340 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
341 */
342struct hammer_ino_rb_tree;
343struct hammer_inode;
344RB_HEAD(hammer_ino_rb_tree, hammer_inode);
345RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
346 hammer_ino_rb_compare, hammer_inode_info_t);
347
73896937
MD
348struct hammer_redo_rb_tree;
349RB_HEAD(hammer_redo_rb_tree, hammer_inode);
350RB_PROTOTYPE2(hammer_redo_rb_tree, hammer_inode, rb_redonode,
351 hammer_redo_rb_compare, hammer_off_t);
352
8cd0a023
MD
353struct hammer_rec_rb_tree;
354struct hammer_record;
355RB_HEAD(hammer_rec_rb_tree, hammer_record);
356RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 357 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
358
359TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
360
361struct hammer_inode {
1f07f686
MD
362 RB_ENTRY(hammer_inode) rb_node;
363 hammer_inode_state_t flush_state;
7a61b85d 364 hammer_flush_group_t flush_group;
ff003b11 365 RB_ENTRY(hammer_inode) rb_flsnode; /* when on flush list */
73896937 366 RB_ENTRY(hammer_inode) rb_redonode; /* when INODE_RDIRTY is set */
1f07f686 367 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 368 int64_t obj_id; /* (key) object identifier */
b84de5af 369 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 370 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 371 struct hammer_mount *hmp;
0729c8c8 372 hammer_objid_cache_t objid_cache;
b84de5af
MD
373 int flags;
374 int error; /* flush error */
375 int cursor_ip_refs; /* sanity */
47637bff 376 int rsv_recs;
b84de5af 377 struct vnode *vp;
5fa5c92f 378 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
379 struct lockf advlock;
380 struct hammer_lock lock; /* sync copy interlock */
b84de5af 381 off_t trunc_off;
11ad5ade 382 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
383 struct hammer_inode_data ino_data; /* in-memory cache */
384 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
3214ade6 385 int rec_generation;
39d8fd63 386 struct hammer_node_cache cache[4]; /* search initiate cache */
b84de5af
MD
387
388 /*
389 * When a demark is created to synchronize an inode to
390 * disk, certain fields are copied so the front-end VOPs
391 * can continue to run in parallel with the synchronization
392 * occuring in the background.
393 */
394 int sync_flags; /* to-sync flags cache */
395 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 396 off_t save_trunc_off; /* write optimization */
11ad5ade 397 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 398 struct hammer_inode_data sync_ino_data; /* to-sync cache */
9192654c 399 size_t redo_count;
73896937
MD
400
401 /*
402 * Track the earliest offset in the UNDO/REDO FIFO containing
403 * REDO records. This is staged to the backend during flush
404 * sequences. While the inode is staged redo_fifo_next is used
405 * to track the earliest offset for rotation into redo_fifo_start
406 * on completion of the flush.
407 */
408 hammer_off_t redo_fifo_start;
409 hammer_off_t redo_fifo_next;
8750964d
MD
410};
411
8cd0a023
MD
412typedef struct hammer_inode *hammer_inode_t;
413
66325755
MD
414#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
415
9192654c
MD
416/*
417 * NOTE: DDIRTY does not include atime or mtime and does not include
418 * write-append size changes. SDIRTY handles write-append size
419 * changes.
47f363f1
MD
420 *
421 * REDO indicates that REDO logging is active, creating a definitive
422 * stream of REDO records in the UNDO/REDO log for writes and
423 * truncations, including boundary records when/if REDO is turned off.
424 * REDO is typically enabled by fsync() and turned off if excessive
425 * writes without an fsync() occurs.
426 *
427 * RDIRTY indicates that REDO records were laid down in the UNDO/REDO
428 * FIFO (even if REDO is turned off some might still be active) and
429 * still being tracked for this inode. See hammer_redo.c
9192654c 430 */
ddfdf542 431 /* (not including atime/mtime) */
9192654c 432#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
e63644f0 433#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
7b6ccb11 434#define HAMMER_INODE_CONN_DOWN 0x0004 /* include in downward recursion */
1f07f686 435#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 436#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 437#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 438#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 439#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 440#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e0092341 441#define HAMMER_INODE_UNUSED0400 0x0400
0a72edae 442#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 443#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
f153644d 444#define HAMMER_INODE_REFLUSH 0x2000 /* flush on dependancy / reflush */
9f5097dc 445#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 446#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 447
b84de5af 448#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 449#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 450#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 451#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
452#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
453#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
4c286c36 454#define HAMMER_INODE_DUMMY 0x00800000 /* dummy inode covering bad file */
9192654c 455#define HAMMER_INODE_SDIRTY 0x01000000 /* in-memory ino_data.size is dirty*/
47f363f1
MD
456#define HAMMER_INODE_REDO 0x02000000 /* REDO logging active */
457#define HAMMER_INODE_RDIRTY 0x04000000 /* REDO records active in fifo */
b84de5af 458
9192654c 459#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY|HAMMER_INODE_SDIRTY| \
11ad5ade 460 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
461 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
462 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686 463
9192654c 464#define HAMMER_INODE_MODMASK_NOXDIRTY \
1f07f686 465 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 466
9192654c
MD
467#define HAMMER_INODE_MODMASK_NOREDO \
468 (HAMMER_INODE_DDIRTY| \
469 HAMMER_INODE_XDIRTY| \
470 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
471
f90dde4c 472#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 473#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 474
7bc5b8c2
MD
475/*
476 * Used by the inode reclaim code to pipeline reclaims and avoid
477 * blowing out kernel memory or letting the flusher get too far
82010f9f
MD
478 * behind. The reclaim wakes up when count reaches 0 or the
479 * timer expires.
7bc5b8c2
MD
480 */
481struct hammer_reclaim {
482 TAILQ_ENTRY(hammer_reclaim) entry;
82010f9f 483 int count;
7bc5b8c2
MD
484};
485
ff003b11 486#define HAMMER_RECLAIM_WAIT 4000 /* default vfs.hammer.limit_reclaim */
9f5097dc 487
8750964d 488/*
e98f1b96
MD
489 * Track who is creating the greatest burden on the
490 * inode cache.
491 */
492struct hammer_inostats {
493 pid_t pid; /* track user process */
494 int ltick; /* last tick */
495 int count; /* count (degenerates) */
496};
497
498#define HAMMER_INOSTATS_HSIZE 32
499#define HAMMER_INOSTATS_HMASK (HAMMER_INOSTATS_HSIZE - 1)
500
501/*
1f07f686
MD
502 * Structure used to represent an unsynchronized record in-memory. These
503 * records typically represent directory entries. Only non-historical
504 * records are kept in-memory.
505 *
506 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
507 * on disk then neither are any records and the in-memory record tree
508 * represents the entire contents of the inode. If the inode is on disk
509 * then the on-disk B-Tree is scanned in parallel with the in-memory
510 * RB-Tree to synthesize the current state of the file.
511 *
1f07f686
MD
512 * Records are also used to enforce the ordering of directory create/delete
513 * operations. A new inode will not be flushed to disk unless its related
514 * directory entry is also being flushed at the same time. A directory entry
515 * will not be removed unless its related inode is also being removed at the
516 * same time.
8750964d 517 */
1f07f686 518typedef enum hammer_record_type {
e8599db1 519 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 520 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 521 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
522 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
523 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
524} hammer_record_type_t;
525
8cd0a023
MD
526struct hammer_record {
527 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
528 TAILQ_ENTRY(hammer_record) target_entry;
529 hammer_inode_state_t flush_state;
7a61b85d 530 hammer_flush_group_t flush_group;
1f07f686 531 hammer_record_type_t type;
a89aec1b 532 struct hammer_lock lock;
0832c9bb 533 struct hammer_reserve *resv;
8cd0a023 534 struct hammer_inode *ip;
1f07f686 535 struct hammer_inode *target_ip;
11ad5ade 536 struct hammer_btree_leaf_elm leaf;
8cd0a023 537 union hammer_data_ondisk *data;
8cd0a023 538 int flags;
77912481 539 int gflags;
e469566b 540 hammer_off_t zone2_offset; /* direct-write only */
8cd0a023 541};
8750964d 542
8cd0a023
MD
543typedef struct hammer_record *hammer_record_t;
544
d36ec43b
MD
545/*
546 * Record flags. Note that FE can only be set by the frontend if the
547 * record has not been interlocked by the backend w/ BE.
548 */
8cd0a023
MD
549#define HAMMER_RECF_ALLOCDATA 0x0001
550#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 551#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 552#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
1b0ab2c3 553#define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */
d36ec43b 554#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 555#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
507df98a 556#define HAMMER_RECF_DEDUPED 0x0080 /* will be live-dedup'ed */
1b0ab2c3 557#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
47f363f1 558#define HAMMER_RECF_REDO 0x1000 /* REDO was laid down */
8cd0a023
MD
559
560/*
77912481
MD
561 * These flags must be separate to deal with SMP races
562 */
563#define HAMMER_RECG_DIRECT_IO 0x0001 /* related direct I/O running*/
564#define HAMMER_RECG_DIRECT_WAIT 0x0002 /* related direct I/O running*/
565#define HAMMER_RECG_DIRECT_INVAL 0x0004 /* buffer alias invalidation */
566/*
83f2a3aa 567 * hammer_create_at_cursor() and hammer_delete_at_cursor() flags.
602c6cb8 568 */
83f2a3aa
MD
569#define HAMMER_CREATE_MODE_UMIRROR 0x0001
570#define HAMMER_CREATE_MODE_SYS 0x0002
571
602c6cb8
MD
572#define HAMMER_DELETE_ADJUST 0x0001
573#define HAMMER_DELETE_DESTROY 0x0002
574
575/*
47197d71 576 * In-memory structures representing on-disk structures.
8cd0a023 577 */
8750964d 578struct hammer_volume;
427e5fc6 579struct hammer_buffer;
8cd0a023 580struct hammer_node;
e8599db1 581struct hammer_undo;
0832c9bb
MD
582struct hammer_reserve;
583
8750964d 584RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 585RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 586RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 587RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 588RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
589
590RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
591 hammer_vol_rb_compare, int32_t);
427e5fc6 592RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 593 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 594RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 595 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
596RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
597 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
598RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
599 hammer_res_rb_compare, hammer_off_t);
8750964d 600
66325755
MD
601/*
602 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
603 *
604 * VOLUME - hammer_volume containing meta-data
605 * META_BUFFER - hammer_buffer containing meta-data
606 * DATA_BUFFER - hammer_buffer containing pure-data
607 *
608 * Dirty volume headers and dirty meta-data buffers are locked until the
609 * flusher can sequence them out. Dirty pure-data buffers can be written.
610 * Clean buffers can be passively released.
66325755 611 */
10a5d1ba
MD
612typedef enum hammer_io_type {
613 HAMMER_STRUCTURE_VOLUME,
614 HAMMER_STRUCTURE_META_BUFFER,
615 HAMMER_STRUCTURE_UNDO_BUFFER,
eddadaee
MD
616 HAMMER_STRUCTURE_DATA_BUFFER,
617 HAMMER_STRUCTURE_DUMMY
10a5d1ba 618} hammer_io_type_t;
66325755
MD
619
620union hammer_io_structure;
055f5ff8 621struct hammer_io;
66325755
MD
622
623struct worklist {
624 LIST_ENTRY(worklist) node;
625};
626
10a5d1ba
MD
627TAILQ_HEAD(hammer_io_list, hammer_io);
628typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 629
66325755 630struct hammer_io {
10a5d1ba
MD
631 struct worklist worklist;
632 struct hammer_lock lock;
633 enum hammer_io_type type;
634 struct hammer_mount *hmp;
748efb59 635 struct hammer_volume *volume;
10a5d1ba 636 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
eddadaee 637 TAILQ_ENTRY(hammer_io) iorun_entry; /* iorun_list */
10a5d1ba
MD
638 hammer_io_list_t mod_list;
639 struct buf *bp;
4a2796f3
MD
640 int64_t offset; /* zone-2 offset */
641 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
642 int modify_refs;
643
77912481
MD
644 /*
645 * These can be modified at any time by the backend while holding
646 * io_token, due to bio_done and hammer_io_complete() callbacks.
647 */
055f5ff8
MD
648 u_int running : 1; /* bp write IO in progress */
649 u_int waiting : 1; /* someone is waiting on us */
77912481
MD
650 u_int ioerror : 1; /* abort on io-error */
651 u_int unusedA : 29;
652
653 /*
654 * These can only be modified by the frontend while holding
655 * fs_token, or by the backend while holding the io interlocked
656 * with no references (which will block the frontend when it
657 * tries to reference it).
658 *
659 * WARNING! SMP RACES will create havoc if the callbacks ever tried
660 * to modify any of these outside the above restrictions.
661 */
662 u_int modified : 1; /* bp's data was modified */
663 u_int released : 1; /* bp released (w/ B_LOCKED set) */
b33e2cc0 664 u_int validated : 1; /* ondisk has been validated */
b58c6388 665 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 666 u_int recovered : 1; /* has recovery ref */
9f5097dc 667 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 668 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 669 u_int gencrc : 1; /* crc needs to be generated */
77912481 670 u_int unusedB : 24;
66325755
MD
671};
672
8cd0a023
MD
673typedef struct hammer_io *hammer_io_t;
674
af209b0f
MD
675#define HAMMER_CLUSTER_SIZE (64 * 1024)
676#if HAMMER_CLUSTER_SIZE > MAXBSIZE
677#undef HAMMER_CLUSTER_SIZE
678#define HAMMER_CLUSTER_SIZE MAXBSIZE
679#endif
680#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
681
66325755 682/*
8cd0a023 683 * In-memory volume representing on-disk buffer
66325755 684 */
8750964d 685struct hammer_volume {
66325755 686 struct hammer_io io;
8750964d 687 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
688 struct hammer_volume_ondisk *ondisk;
689 int32_t vol_no;
fbc6e32a 690 int64_t nblocks; /* note: special calculation for statfs */
47197d71 691 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
692 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
693 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 694 char *vol_name;
8750964d 695 struct vnode *devvp;
427e5fc6 696 int vol_flags;
8750964d
MD
697};
698
8cd0a023
MD
699typedef struct hammer_volume *hammer_volume_t;
700
66325755 701/*
8cd0a023
MD
702 * In-memory buffer (other then volume, super-cluster, or cluster),
703 * representing an on-disk buffer.
66325755 704 */
427e5fc6 705struct hammer_buffer {
66325755 706 struct hammer_io io;
427e5fc6 707 RB_ENTRY(hammer_buffer) rb_node;
47197d71 708 void *ondisk;
34d829f7 709 hammer_off_t zoneX_offset;
0832c9bb 710 hammer_off_t zone2_offset;
cebe9493 711 struct hammer_reserve *resv;
8cd0a023 712 struct hammer_node_list clist;
66325755
MD
713};
714
8cd0a023
MD
715typedef struct hammer_buffer *hammer_buffer_t;
716
717/*
718 * In-memory B-Tree node, representing an on-disk B-Tree node.
719 *
720 * This is a hang-on structure which is backed by a hammer_buffer,
721 * indexed by a hammer_cluster, and used for fine-grained locking of
722 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
723 * can contain multiple nodes representing wildly disassociated portions
724 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
725 *
726 * This structure uses a cluster-relative index to reduce the number
727 * of layers required to access it, and also because all on-disk B-Tree
728 * references are cluster-relative offsets.
729 */
730struct hammer_node {
731 struct hammer_lock lock; /* node-by-node lock */
732 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
733 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 734 hammer_off_t node_offset; /* full offset spec */
40043e7f 735 struct hammer_mount *hmp;
8cd0a023
MD
736 struct hammer_buffer *buffer; /* backing buffer */
737 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 738 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 739 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 740 int flags;
8cd0a023
MD
741};
742
b3deaf57
MD
743#define HAMMER_NODE_DELETED 0x0001
744#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
745#define HAMMER_NODE_CRCGOOD 0x0004
746#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 747#define HAMMER_NODE_NEEDSMIRROR 0x0010
4c286c36 748#define HAMMER_NODE_CRCBAD 0x0020
fa2b9a03 749#define HAMMER_NODE_NONLINEAR 0x0040 /* linear heuristic */
4c286c36
MD
750
751#define HAMMER_NODE_CRCANY (HAMMER_NODE_CRCGOOD | HAMMER_NODE_CRCBAD)
b3deaf57 752
8cd0a023
MD
753typedef struct hammer_node *hammer_node_t;
754
755/*
1775b6a0
MD
756 * List of locked nodes. This structure is used to lock potentially large
757 * numbers of nodes as an aid for complex B-Tree operations.
b33e2cc0 758 */
1775b6a0
MD
759struct hammer_node_lock;
760TAILQ_HEAD(hammer_node_lock_list, hammer_node_lock);
761
762struct hammer_node_lock {
763 TAILQ_ENTRY(hammer_node_lock) entry;
764 struct hammer_node_lock_list list;
765 struct hammer_node_lock *parent;
b33e2cc0 766 hammer_node_t node;
1775b6a0
MD
767 hammer_node_ondisk_t copy; /* copy of on-disk data */
768 int index; /* index of this node in parent */
769 int count; /* count children */
770 int flags;
b33e2cc0
MD
771};
772
1775b6a0 773typedef struct hammer_node_lock *hammer_node_lock_t;
b33e2cc0 774
1775b6a0 775#define HAMMER_NODE_LOCK_UPDATED 0x0001
24cf83d2 776#define HAMMER_NODE_LOCK_LCACHE 0x0002
b33e2cc0
MD
777
778/*
8cd0a023
MD
779 * Common I/O management structure - embedded in in-memory structures
780 * which are backed by filesystem buffers.
781 */
66325755
MD
782union hammer_io_structure {
783 struct hammer_io io;
784 struct hammer_volume volume;
66325755 785 struct hammer_buffer buffer;
8750964d
MD
786};
787
055f5ff8
MD
788typedef union hammer_io_structure *hammer_io_structure_t;
789
bf686dbe 790/*
cebe9493
MD
791 * The reserve structure prevents the blockmap from allocating
792 * out of a reserved bigblock. Such reservations are used by
793 * the direct-write mechanism.
794 *
795 * The structure is also used to hold off on reallocations of
796 * big blocks from the freemap until flush dependancies have
797 * been dealt with.
798 */
0832c9bb
MD
799struct hammer_reserve {
800 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
801 TAILQ_ENTRY(hammer_reserve) delay_entry;
802 int flush_group;
cb51be26 803 int flags;
0832c9bb 804 int refs;
cb51be26 805 int zone;
df301614 806 int append_off;
507df98a 807 int32_t bytes_free;
cebe9493 808 hammer_off_t zone_offset;
0832c9bb
MD
809};
810
811typedef struct hammer_reserve *hammer_reserve_t;
812
cb51be26 813#define HAMMER_RESF_ONDELAY 0x0001
5e435c92 814#define HAMMER_RESF_LAYER2FREE 0x0002
cb51be26 815
8cd0a023
MD
816#include "hammer_cursor.h"
817
8750964d 818/*
cebe9493
MD
819 * The undo structure tracks recent undos to avoid laying down duplicate
820 * undos within a flush group, saving us a significant amount of overhead.
821 *
822 * This is strictly a heuristic.
e8599db1 823 */
7a61b85d
MD
824#define HAMMER_MAX_UNDOS 1024
825#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
826
827struct hammer_undo {
828 RB_ENTRY(hammer_undo) rb_node;
829 TAILQ_ENTRY(hammer_undo) lru_entry;
830 hammer_off_t offset;
831 int bytes;
832};
833
834typedef struct hammer_undo *hammer_undo_t;
835
af209b0f 836struct hammer_flusher_info;
7a61b85d 837TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
838
839struct hammer_flusher {
840 int signal; /* flusher thread sequencer */
841 int act; /* currently active flush group */
842 int done; /* set to act when complete */
843 int next; /* next flush group */
844 int group_lock; /* lock sequencing of the next flush */
845 int exiting; /* request master exit */
da2da375
MD
846 thread_t td; /* master flusher thread */
847 hammer_tid_t tid; /* last flushed transaction id */
848 int finalize_want; /* serialize finalization */
849 struct hammer_lock finalize_lock; /* serialize finalization */
850 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
851 struct hammer_flusher_info_list run_list;
852 struct hammer_flusher_info_list ready_list;
da2da375
MD
853};
854
9192654c
MD
855#define HAMMER_FLUSH_UNDOS_RELAXED 0
856#define HAMMER_FLUSH_UNDOS_FORCED 1
857#define HAMMER_FLUSH_UNDOS_AUTO 2
da2da375 858/*
8750964d
MD
859 * Internal hammer mount data structure
860 */
861struct hammer_mount {
862 struct mount *mp;
27ea2398 863 /*struct vnode *rootvp;*/
8750964d 864 struct hammer_ino_rb_tree rb_inos_root;
73896937 865 struct hammer_redo_rb_tree rb_redo_root;
8750964d 866 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 867 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 868 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
869 struct hammer_res_rb_tree rb_resv_root;
870 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 871 struct hammer_pfs_rb_tree rb_pfsm_root;
507df98a
ID
872
873 struct hammer_dedup_crc_rb_tree rb_dedup_crc_root;
874 struct hammer_dedup_off_rb_tree rb_dedup_off_root;
875
8750964d 876 struct hammer_volume *rootvol;
47197d71
MD
877 struct hammer_base_elm root_btree_beg;
878 struct hammer_base_elm root_btree_end;
bac808fe
MD
879
880 struct malloc_type *m_misc;
881 struct malloc_type *m_inodes;
882
cdb6e4e6 883 int flags; /* HAMMER_MOUNT_xxx flags */
195c19a1
MD
884 int hflags;
885 int ronly;
886 int nvolumes;
d26d0ae9 887 int volume_iterator;
732a1697 888 int master_id; /* -1 or 0-15 - clustering and mirroring */
44a83111 889 int version; /* hammer filesystem version to use */
e63644f0 890 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 891 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 892 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 893 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 894 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
895 int last_newrecords;
896 int count_newrecords;
da2da375 897
865c9609
MN
898 int volume_to_remove; /* volume that is currently being removed */
899
9f5097dc
MD
900 int inode_reclaims; /* inodes pending reclaim by flusher */
901 int count_inodes; /* total number of inodes */
af209b0f 902 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
903
904 struct hammer_flusher flusher;
905
855942b6 906 u_int check_interrupt;
3e583440 907 u_int check_yield;
8750964d 908 uuid_t fsid;
10a5d1ba
MD
909 struct hammer_io_list volu_list; /* dirty undo buffers */
910 struct hammer_io_list undo_list; /* dirty undo buffers */
911 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 912 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
913 struct hammer_io_list meta_list; /* dirty meta bufs */
914 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a 915 int locked_dirty_space; /* meta/volu count */
b0aab9b9
MD
916 int io_running_space; /* io_token */
917 int io_running_wakeup; /* io_token */
0729c8c8 918 int objid_cache_count;
507df98a 919 int dedup_cache_count;
cdb6e4e6
MD
920 int error; /* critical I/O error */
921 struct krate krate; /* rate limited kprintf */
ddfdf542 922 hammer_tid_t asof; /* snapshot mount */
4889cbd4
MD
923 hammer_tid_t next_tid;
924 hammer_tid_t flush_tid1; /* flusher tid sequencing */
925 hammer_tid_t flush_tid2; /* flusher tid sequencing */
e63644f0 926 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
02428fb6 927 u_int32_t undo_seqno; /* UNDO/REDO FIFO seqno */
c58123da
MD
928 u_int32_t recover_stage2_seqno; /* REDO recovery seqno */
929 hammer_off_t recover_stage2_offset; /* REDO recovery offset */
e63644f0 930
513ca7d7 931 struct netexport export;
9480ff55 932 struct hammer_lock sync_lock;
c9b9e29d 933 struct hammer_lock free_lock;
d99d6bf5
MD
934 struct hammer_lock undo_lock;
935 struct hammer_lock blkmap_lock;
83f2a3aa 936 struct hammer_lock snapshot_lock;
52e547e3 937 struct hammer_lock volume_lock;
0729c8c8 938 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
939 struct hammer_undo undos[HAMMER_MAX_UNDOS];
940 int undo_alloc;
941 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 942 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 943 struct hammer_flush_group_list flush_group_list;
7b6ccb11 944 hammer_flush_group_t next_flush_group;
0729c8c8 945 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
507df98a
ID
946 TAILQ_HEAD(, hammer_dedup_cache) dedup_lru_list;
947 hammer_dedup_cache_t dedup_free_cache;
7bc5b8c2 948 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
eddadaee 949 TAILQ_HEAD(, hammer_io) iorun_list;
e98f1b96 950
b0aab9b9
MD
951 struct lwkt_token fs_token; /* high level */
952 struct lwkt_token io_token; /* low level (IO callback) */
953
e98f1b96 954 struct hammer_inostats inostats[HAMMER_INOSTATS_HSIZE];
8750964d
MD
955};
956
8cd0a023 957typedef struct hammer_mount *hammer_mount_t;
8750964d 958
cdb6e4e6 959#define HAMMER_MOUNT_CRITICAL_ERROR 0x0001
82010f9f 960#define HAMMER_MOUNT_FLUSH_RECOVERY 0x0002
47f363f1 961#define HAMMER_MOUNT_REDO_SYNC 0x0004
c58123da
MD
962#define HAMMER_MOUNT_REDO_RECOVERY_REQ 0x0008
963#define HAMMER_MOUNT_REDO_RECOVERY_RUN 0x0010
9f5097dc 964
fbc6e32a
MD
965struct hammer_sync_info {
966 int error;
967 int waitfor;
968};
969
f6468e9e
MD
970/*
971 * Minium buffer cache bufs required to rebalance the B-Tree.
972 * This is because we must hold the children and the children's children
973 * locked. Even this might not be enough if things are horribly out
974 * of balance.
975 */
976#define HAMMER_REBALANCE_MIN_BUFS \
977 (HAMMER_BTREE_LEAF_ELMS * HAMMER_BTREE_LEAF_ELMS)
978
979
8750964d
MD
980#endif
981
a7e9bef1
MD
982/*
983 * checkspace slop (8MB chunks), higher numbers are more conservative.
984 */
93291532
MD
985#define HAMMER_CHKSPC_REBLOCK 25
986#define HAMMER_CHKSPC_MIRROR 20
987#define HAMMER_CHKSPC_WRITE 20
988#define HAMMER_CHKSPC_CREATE 20
989#define HAMMER_CHKSPC_REMOVE 10
7b6ccb11 990#define HAMMER_CHKSPC_EMERGENCY 0
a7e9bef1 991
8750964d
MD
992#if defined(_KERNEL)
993
994extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
995extern struct vop_ops hammer_spec_vops;
996extern struct vop_ops hammer_fifo_vops;
66325755 997extern struct bio_ops hammer_bioops;
427e5fc6 998
2f85fa4d 999extern int hammer_debug_io;
d5ef456e 1000extern int hammer_debug_general;
77062c8a 1001extern int hammer_debug_debug;
e8599db1 1002extern int hammer_debug_inode;
7d683b0f 1003extern int hammer_debug_locks;
b3deaf57 1004extern int hammer_debug_btree;
d113fda1 1005extern int hammer_debug_tid;
b33e2cc0 1006extern int hammer_debug_recover;
46fe7ae1 1007extern int hammer_debug_recover_faults;
fc73edd8 1008extern int hammer_debug_critical;
1b0ab2c3 1009extern int hammer_cluster_enable;
507df98a 1010extern int hammer_live_dedup;
7a61b85d 1011extern int hammer_count_fsyncs;
b3deaf57 1012extern int hammer_count_inodes;
af209b0f 1013extern int hammer_count_iqueued;
9f5097dc 1014extern int hammer_count_reclaiming;
b3deaf57
MD
1015extern int hammer_count_records;
1016extern int hammer_count_record_datas;
1017extern int hammer_count_volumes;
b3deaf57
MD
1018extern int hammer_count_buffers;
1019extern int hammer_count_nodes;
a7e9bef1 1020extern int64_t hammer_count_extra_space_used;
cb51be26
MD
1021extern int64_t hammer_stats_btree_lookups;
1022extern int64_t hammer_stats_btree_searches;
1023extern int64_t hammer_stats_btree_inserts;
1024extern int64_t hammer_stats_btree_deletes;
1025extern int64_t hammer_stats_btree_elements;
1026extern int64_t hammer_stats_btree_splits;
1027extern int64_t hammer_stats_btree_iterations;
39d8fd63 1028extern int64_t hammer_stats_btree_root_iterations;
cb51be26 1029extern int64_t hammer_stats_record_iterations;
ce0138a6
MD
1030extern int64_t hammer_stats_file_read;
1031extern int64_t hammer_stats_file_write;
1032extern int64_t hammer_stats_file_iopsr;
1033extern int64_t hammer_stats_file_iopsw;
1034extern int64_t hammer_stats_disk_read;
1035extern int64_t hammer_stats_disk_write;
1036extern int64_t hammer_stats_inode_flushes;
1037extern int64_t hammer_stats_commits;
89e744ce 1038extern int64_t hammer_stats_undo;
6048b411 1039extern int64_t hammer_stats_redo;
f5a07a7a 1040extern int hammer_count_dirtybufspace;
a99b9ea2 1041extern int hammer_count_refedbufs;
0832c9bb 1042extern int hammer_count_reservations;
a99b9ea2
MD
1043extern int hammer_count_io_running_read;
1044extern int hammer_count_io_running_write;
1045extern int hammer_count_io_locked;
f5a07a7a 1046extern int hammer_limit_dirtybufspace;
ba298df1 1047extern int hammer_limit_running_io;
47637bff 1048extern int hammer_limit_recs;
de996e86 1049extern int hammer_limit_inode_recs;
ff003b11 1050extern int hammer_limit_reclaim;
9192654c 1051extern int hammer_limit_redo;
1f07f686 1052extern int hammer_bio_count;
cb51be26 1053extern int hammer_verify_zone;
1b0ab2c3 1054extern int hammer_verify_data;
cb51be26 1055extern int hammer_write_mode;
3e583440 1056extern int hammer_yield_check;
6f3d87c0 1057extern int hammer_fsync_mode;
21fde338 1058extern int hammer_autoflush;
7d683b0f 1059extern int64_t hammer_contention_count;
b3deaf57 1060
507df98a
ID
1061extern int64_t hammer_live_dedup_vnode_bcmps;
1062extern int64_t hammer_live_dedup_device_bcmps;
1063extern int64_t hammer_live_dedup_findblk_failures;
1064extern int64_t hammer_live_dedup_bmap_saves;
1065
cdb6e4e6
MD
1066void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
1067 int error, const char *msg);
8750964d
MD
1068int hammer_vop_inactive(struct vop_inactive_args *);
1069int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 1070int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 1071struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 1072 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
1073 hammer_tid_t asof, u_int32_t localization,
1074 int flags, int *errorp);
4c286c36
MD
1075struct hammer_inode *hammer_get_dummy_inode(hammer_transaction_t trans,
1076 hammer_inode_t dip, int64_t obj_id,
1077 hammer_tid_t asof, u_int32_t localization,
1078 int flags, int *errorp);
39d8fd63
MD
1079struct hammer_inode *hammer_find_inode(hammer_transaction_t trans,
1080 int64_t obj_id, hammer_tid_t asof,
1081 u_int32_t localization);
43c665ae
MD
1082void hammer_scan_inode_snapshots(hammer_mount_t hmp,
1083 hammer_inode_info_t iinfo,
1084 int (*callback)(hammer_inode_t ip, void *data),
1085 void *data);
66325755
MD
1086void hammer_put_inode(struct hammer_inode *ip);
1087void hammer_put_inode_ref(struct hammer_inode *ip);
e98f1b96 1088void hammer_inode_waitreclaims(hammer_transaction_t trans);
66325755 1089
8cd0a023 1090int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
1091int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
1092
07be83b8 1093int hammer_unload_buffer(hammer_buffer_t buffer, void *data);
7c19b529
MN
1094int hammer_install_volume(hammer_mount_t hmp, const char *volname,
1095 struct vnode *devvp);
1b0ab2c3 1096int hammer_mountcheck_volumes(hammer_mount_t hmp);
8cd0a023 1097
e469566b 1098int hammer_mem_add(hammer_record_t record);
45a014dc 1099int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 1100int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 1101int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 1102int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
1103int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
1104 hammer_tid_t tid);
83f2a3aa
MD
1105int hammer_create_at_cursor(hammer_cursor_t cursor,
1106 hammer_btree_leaf_elm_t leaf, void *udata, int mode);
602c6cb8 1107int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
1108 hammer_tid_t delete_tid, u_int32_t delete_ts,
1109 int track, int64_t *stat_bytes);
b3deaf57 1110int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 1111 hammer_inode_t ip);
fbc6e32a 1112int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
1113int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
1114
8cd0a023 1115hammer_record_t
11ad5ade 1116 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 1117void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 1118void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 1119void hammer_rel_mem_record(hammer_record_t record);
b84de5af 1120
6a37e7e4 1121int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 1122int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 1123int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 1124int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 1125int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 1126void hammer_cursor_downgrade(hammer_cursor_t cursor);
bb29b5d8
MD
1127int hammer_cursor_upgrade2(hammer_cursor_t c1, hammer_cursor_t c2);
1128void hammer_cursor_downgrade2(hammer_cursor_t c1, hammer_cursor_t c2);
32c90105
MD
1129int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
1130 int index);
af209b0f 1131void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
1132int hammer_lock_ex_try(struct hammer_lock *lock);
1133void hammer_lock_sh(struct hammer_lock *lock);
47637bff 1134int hammer_lock_sh_try(struct hammer_lock *lock);
bb29b5d8
MD
1135int hammer_lock_upgrade(struct hammer_lock *lock, int shcount);
1136void hammer_lock_downgrade(struct hammer_lock *lock, int shcount);
b3bad96f 1137int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 1138void hammer_unlock(struct hammer_lock *lock);
66325755 1139void hammer_ref(struct hammer_lock *lock);
250aec18
MD
1140int hammer_ref_interlock(struct hammer_lock *lock);
1141int hammer_ref_interlock_true(struct hammer_lock *lock);
1142void hammer_ref_interlock_done(struct hammer_lock *lock);
1143void hammer_rel(struct hammer_lock *lock);
1144int hammer_rel_interlock(struct hammer_lock *lock, int locked);
1145void hammer_rel_interlock_done(struct hammer_lock *lock, int orig_locked);
1146int hammer_get_interlock(struct hammer_lock *lock);
1147int hammer_try_interlock_norefs(struct hammer_lock *lock);
1148void hammer_put_interlock(struct hammer_lock *lock, int error);
8cd0a023 1149
2f85fa4d
MD
1150void hammer_sync_lock_ex(hammer_transaction_t trans);
1151void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 1152int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
1153void hammer_sync_unlock(hammer_transaction_t trans);
1154
66325755 1155u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 1156void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
1157void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
1158u_int64_t hammer_timespec_to_time(struct timespec *ts);
bc6c1f13
MD
1159int hammer_str_to_tid(const char *str, int *ispfsp,
1160 hammer_tid_t *tidp, u_int32_t *localizationp);
1161int hammer_is_atatext(const char *name, int len);
5a64efa1
MD
1162hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip,
1163 int64_t namekey);
0729c8c8
MD
1164void hammer_clear_objid(hammer_inode_t dip);
1165void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 1166
507df98a
ID
1167int hammer_dedup_crc_rb_compare(hammer_dedup_cache_t dc1,
1168 hammer_dedup_cache_t dc2);
1169int hammer_dedup_off_rb_compare(hammer_dedup_cache_t dc1,
1170 hammer_dedup_cache_t dc2);
1171hammer_dedup_cache_t hammer_dedup_cache_add(hammer_inode_t ip,
1172 hammer_btree_leaf_elm_t leaf);
1173hammer_dedup_cache_t hammer_dedup_cache_lookup(hammer_mount_t hmp,
1174 hammer_crc_t crc);
1175void hammer_dedup_cache_inval(hammer_mount_t hmp, hammer_off_t base_offset);
1176void hammer_destroy_dedup_cache(hammer_mount_t hmp);
1177void hammer_dump_dedup_cache(hammer_mount_t hmp);
1178int hammer_dedup_validate(hammer_dedup_cache_t dcp, int zone, int bytes,
1179 void *data);
1180
e8599db1 1181int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 1182 int bytes);
e8599db1 1183void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 1184enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 1185int hammer_get_dtype(u_int8_t obj_type);
66325755 1186u_int8_t hammer_get_obj_type(enum vtype vtype);
5e435c92
MD
1187int64_t hammer_directory_namekey(hammer_inode_t dip, const void *name, int len,
1188 u_int32_t *max_iterationsp);
e63644f0 1189int hammer_nohistory(hammer_inode_t ip);
427e5fc6 1190
36f82b23 1191int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 1192 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 1193void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 1194void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 1195int hammer_recover_cursor(hammer_cursor_t cursor);
982be4bf
MD
1196void hammer_unlock_cursor(hammer_cursor_t cursor);
1197int hammer_lock_cursor(hammer_cursor_t cursor);
3f43fb33
MD
1198hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
1199void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
1200
1201void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
1202void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
1203 int index);
1204void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
1205 int index);
bbb01e14
MD
1206void hammer_cursor_moved_element(hammer_node_t oparent, int pindex,
1207 hammer_node_t onode, int oindex,
1208 hammer_node_t nnode, int nindex);
1775b6a0
MD
1209void hammer_cursor_parent_changed(hammer_node_t node, hammer_node_t oparent,
1210 hammer_node_t nparent, int nindex);
b3bad96f
MD
1211void hammer_cursor_inserted_element(hammer_node_t node, int index);
1212void hammer_cursor_deleted_element(hammer_node_t node, int index);
b9107f58 1213void hammer_cursor_invalidate_cache(hammer_cursor_t cursor);
8cd0a023
MD
1214
1215int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 1216int hammer_btree_first(hammer_cursor_t cursor);
32c90105 1217int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
1218int hammer_btree_extract(hammer_cursor_t cursor, int flags);
1219int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 1220int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 1221int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 1222 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 1223int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
1224void hammer_btree_do_propagation(hammer_cursor_t cursor,
1225 hammer_pseudofs_inmem_t pfsm,
602c6cb8 1226 hammer_btree_leaf_elm_t leaf);
8cd0a023 1227int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 1228int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
1229int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
1230int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
1231
2f85fa4d
MD
1232int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
1233 hammer_btree_elm_t elm);
1775b6a0 1234void hammer_node_lock_init(hammer_node_lock_t parent, hammer_node_t node);
24cf83d2
MD
1235void hammer_btree_lcache_init(hammer_mount_t hmp, hammer_node_lock_t lcache,
1236 int depth);
1237void hammer_btree_lcache_free(hammer_mount_t hmp, hammer_node_lock_t lcache);
1775b6a0 1238int hammer_btree_lock_children(hammer_cursor_t cursor, int depth,
24cf83d2
MD
1239 hammer_node_lock_t parent,
1240 hammer_node_lock_t lcache);
1775b6a0
MD
1241void hammer_btree_lock_copy(hammer_cursor_t cursor,
1242 hammer_node_lock_t parent);
7ddc70d1 1243int hammer_btree_sync_copy(hammer_cursor_t cursor,
1775b6a0 1244 hammer_node_lock_t parent);
24cf83d2
MD
1245void hammer_btree_unlock_children(hammer_mount_t hmp,
1246 hammer_node_lock_t parent,
1247 hammer_node_lock_t lcache);
bcac4bbb 1248int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
82010f9f
MD
1249hammer_node_t hammer_btree_get_parent(hammer_transaction_t trans,
1250 hammer_node_t node, int *parent_indexp,
c82af904 1251 int *errorp, int try_exclusive);
b33e2cc0 1252
c0ade690
MD
1253void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
1254void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 1255
47197d71
MD
1256void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
1257 int *errorp, struct hammer_buffer **bufferp);
1258void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
1259 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
1260void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
1261 int *errorp, struct hammer_buffer **bufferp);
1262void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
1263 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
1264
1265hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 1266
8cd0a023 1267hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 1268 int32_t vol_no, int *errorp);
4a2796f3
MD
1269hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
1270 int bytes, int isnew, int *errorp);
1b0ab2c3
MD
1271void hammer_sync_buffers(hammer_mount_t hmp,
1272 hammer_off_t base_offset, int bytes);
362ec2dc 1273int hammer_del_buffers(hammer_mount_t hmp,
1b0ab2c3 1274 hammer_off_t base_offset,
362ec2dc
MD
1275 hammer_off_t zone2_offset, int bytes,
1276 int report_conflicts);
8cd0a023 1277
fbc6e32a 1278int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
1279int hammer_ref_buffer(hammer_buffer_t buffer);
1280void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
1281
250aec18
MD
1282void hammer_rel_volume(hammer_volume_t volume, int locked);
1283void hammer_rel_buffer(hammer_buffer_t buffer, int locked);
8cd0a023 1284
513ca7d7
MD
1285int hammer_vfs_export(struct mount *mp, int op,
1286 const struct export_args *export);
82010f9f
MD
1287hammer_node_t hammer_get_node(hammer_transaction_t trans,
1288 hammer_off_t node_offset, int isnew, int *errorp);
740d8317 1289void hammer_ref_node(hammer_node_t node);
4c286c36 1290hammer_node_t hammer_ref_node_safe(hammer_transaction_t trans,
bcac4bbb 1291 hammer_node_cache_t cache, int *errorp);
8cd0a023 1292void hammer_rel_node(hammer_node_t node);
36f82b23
MD
1293void hammer_delete_node(hammer_transaction_t trans,
1294 hammer_node_t node);
bcac4bbb
MD
1295void hammer_cache_node(hammer_node_cache_t cache,
1296 hammer_node_t node);
1297void hammer_uncache_node(hammer_node_cache_t cache);
250aec18 1298void hammer_flush_node(hammer_node_t node, int locked);
8cd0a023 1299
427e5fc6
MD
1300void hammer_dup_buffer(struct hammer_buffer **bufferp,
1301 struct hammer_buffer *buffer);
df2ccbac
MD
1302hammer_node_t hammer_alloc_btree(hammer_transaction_t trans,
1303 hammer_off_t hint, int *errorp);
36f82b23 1304void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 1305 u_int16_t rec_type, hammer_off_t *data_offsetp,
df2ccbac
MD
1306 struct hammer_buffer **data_bufferp,
1307 hammer_off_t hint, int *errorp);
bf686dbe 1308
02428fb6 1309int hammer_generate_undo(hammer_transaction_t trans,
059819e3 1310 hammer_off_t zone1_offset, void *base, int len);
6048b411 1311int hammer_generate_redo(hammer_transaction_t trans, hammer_inode_t ip,
47f363f1
MD
1312 hammer_off_t file_offset, u_int32_t flags,
1313 void *base, int len);
1314void hammer_generate_redo_sync(hammer_transaction_t trans);
73896937
MD
1315void hammer_redo_fifo_start_flush(hammer_inode_t ip);
1316void hammer_redo_fifo_end_flush(hammer_inode_t ip);
1317
47f363f1 1318void hammer_format_undo(void *base, u_int32_t seqno);
02428fb6 1319int hammer_upgrade_undo_4(hammer_transaction_t trans);
427e5fc6 1320
66325755 1321void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 1322void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 1323
36f82b23
MD
1324hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
1325 hammer_off_t owner, int *errorp);
1326void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 1327 hammer_off_t owner, int *errorp);
0f65be10 1328int _hammer_checkspace(hammer_mount_t hmp, int slop, int64_t *resp);
36f82b23 1329hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
df2ccbac 1330 int bytes, hammer_off_t hint, int *errorp);
0832c9bb
MD
1331hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
1332 int bytes, hammer_off_t *zone_offp, int *errorp);
507df98a
ID
1333hammer_reserve_t hammer_blockmap_reserve_dedup(hammer_mount_t hmp, int zone,
1334 int bytes, hammer_off_t zone_offset, int *errorp);
0832c9bb
MD
1335void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
1336 hammer_reserve_t resv);
cb51be26 1337void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1338void hammer_blockmap_free(hammer_transaction_t trans,
1339 hammer_off_t bmap_off, int bytes);
bb29b5d8
MD
1340int hammer_blockmap_dedup(hammer_transaction_t trans,
1341 hammer_off_t bmap_off, int bytes);
cdb6e4e6 1342int hammer_blockmap_finalize(hammer_transaction_t trans,
5e435c92 1343 hammer_reserve_t resv,
4a2796f3 1344 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1345int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1346 int *curp, int *errorp);
40043e7f
MD
1347hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1348 int *errorp);
bf686dbe
MD
1349hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1350 int *errorp);
06ad81ff
MD
1351int64_t hammer_undo_used(hammer_transaction_t trans);
1352int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686 1353int64_t hammer_undo_max(hammer_mount_t hmp);
710733a6 1354int hammer_undo_reclaim(hammer_io_t io);
1f07f686 1355
8cd0a023
MD
1356void hammer_start_transaction(struct hammer_transaction *trans,
1357 struct hammer_mount *hmp);
36f82b23
MD
1358void hammer_simple_transaction(struct hammer_transaction *trans,
1359 struct hammer_mount *hmp);
b84de5af
MD
1360void hammer_start_transaction_fls(struct hammer_transaction *trans,
1361 struct hammer_mount *hmp);
1362void hammer_done_transaction(struct hammer_transaction *trans);
83f2a3aa 1363hammer_tid_t hammer_alloc_tid(hammer_mount_t hmp, int count);
66325755 1364
e98f1b96 1365void hammer_modify_inode(hammer_transaction_t trans, hammer_inode_t ip, int flags);
f90dde4c 1366void hammer_flush_inode(hammer_inode_t ip, int flags);
cdb6e4e6 1367void hammer_flush_inode_done(hammer_inode_t ip, int error);
b84de5af
MD
1368void hammer_wait_inode(hammer_inode_t ip);
1369
8cd0a023
MD
1370int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1371 struct ucred *cred, struct hammer_inode *dip,
5a64efa1 1372 const char *name, int namelen,
ea434b6f
MD
1373 hammer_pseudofs_inmem_t pfsm,
1374 struct hammer_inode **ipp);
a89aec1b 1375void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1376int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1377int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
73896937 1378int hammer_redo_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
cdb6e4e6 1379int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused);
51c35492 1380
02325004 1381int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip);
7b6ccb11 1382void hammer_test_inode(hammer_inode_t dip);
e8599db1 1383void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 1384
a89aec1b 1385int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1386 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1387 hammer_inode_t nip);
a89aec1b 1388int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1389 hammer_cursor_t cursor, hammer_inode_t dip,
1390 hammer_inode_t ip);
6362a262 1391void hammer_ip_replace_bulk(hammer_mount_t hmp, hammer_record_t record);
47637bff 1392hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1393 void *data, int bytes, int *errorp);
47637bff 1394int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1395int hammer_ip_add_record(struct hammer_transaction *trans,
1396 hammer_record_t record);
4e17f465 1397int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1398 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1399int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1400 int *countp);
1401int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1402 int64_t offset, void *data, int bytes);
36f82b23 1403int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1404int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1405hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1406 u_int32_t localization, int *errorp);
1407int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1408 hammer_pseudofs_inmem_t pfsm);
1409int hammer_save_pseudofs(hammer_transaction_t trans,
1410 hammer_pseudofs_inmem_t pfsm);
842e7a70 1411int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1412void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1413int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1414 struct ucred *cred);
1415
748efb59 1416void hammer_io_init(hammer_io_t io, hammer_volume_t volume,
10a5d1ba 1417 enum hammer_io_type type);
b7de8aa5 1418int hammer_io_read(struct vnode *devvp, struct hammer_io *io, int limit);
0e8bd897 1419void hammer_io_advance(struct hammer_io *io);
66325755 1420int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
362ec2dc 1421int hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1422struct buf *hammer_io_release(struct hammer_io *io, int flush);
710733a6 1423void hammer_io_flush(struct hammer_io *io, int reclaim);
1b0ab2c3 1424void hammer_io_wait(struct hammer_io *io);
055f5ff8 1425void hammer_io_waitdep(struct hammer_io *io);
eddadaee 1426void hammer_io_wait_all(hammer_mount_t hmp, const char *ident, int doflush);
1b0ab2c3
MD
1427int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio,
1428 hammer_btree_leaf_elm_t leaf);
6362a262
MD
1429int hammer_io_direct_write(hammer_mount_t hmp, struct bio *bio,
1430 hammer_record_t record);
1b0ab2c3 1431void hammer_io_direct_wait(hammer_record_t record);
43c665ae 1432void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1433void hammer_io_write_interlock(hammer_io_t io);
1434void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1435void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1436void hammer_io_clear_modlist(struct hammer_io *io);
748efb59 1437void hammer_io_flush_sync(hammer_mount_t hmp);
2faf0737 1438void hammer_io_clear_error(struct hammer_io *io);
77912481 1439void hammer_io_clear_error_noassert(struct hammer_io *io);
b8a41159 1440void hammer_io_notmeta(hammer_buffer_t buffer);
ba298df1 1441void hammer_io_limit_backlog(hammer_mount_t hmp);
748efb59 1442
36f82b23
MD
1443void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1444 void *base, int len);
1445void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1446 void *base, int len);
10a5d1ba
MD
1447void hammer_modify_volume_done(hammer_volume_t volume);
1448void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1449
36f82b23
MD
1450int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1451 struct hammer_ioc_reblock *reblock);
1775b6a0
MD
1452int hammer_ioc_rebalance(hammer_transaction_t trans, hammer_inode_t ip,
1453 struct hammer_ioc_rebalance *rebal);
11ad5ade
MD
1454int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1455 struct hammer_ioc_prune *prune);
c82af904
MD
1456int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1457 struct hammer_ioc_mirror_rw *mirror);
1458int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1459 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1460int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1461 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1462int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1463 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1464int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1465 struct hammer_ioc_pseudofs_rw *pfs);
1466int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1467 struct hammer_ioc_pseudofs_rw *pfs);
1468int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1469 struct hammer_ioc_pseudofs_rw *pfs);
4889cbd4
MD
1470int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1471 struct hammer_ioc_pseudofs_rw *pfs);
d121f61c 1472int hammer_ioc_volume_add(hammer_transaction_t trans, hammer_inode_t ip,
865c9609
MN
1473 struct hammer_ioc_volume *ioc);
1474int hammer_ioc_volume_del(hammer_transaction_t trans, hammer_inode_t ip,
1475 struct hammer_ioc_volume *ioc);
e914c91d
SK
1476int hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
1477 struct hammer_ioc_volume_list *ioc);
bb29b5d8
MD
1478int hammer_ioc_dedup(hammer_transaction_t trans, hammer_inode_t ip,
1479 struct hammer_ioc_dedup *dedup);
bf686dbe 1480
855942b6 1481int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1482
059819e3
MD
1483void hammer_flusher_create(hammer_mount_t hmp);
1484void hammer_flusher_destroy(hammer_mount_t hmp);
1485void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1486int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1487int hammer_flusher_async_one(hammer_mount_t hmp);
93291532 1488void hammer_flusher_wait(hammer_mount_t hmp, int seq);
82010f9f 1489void hammer_flusher_wait_next(hammer_mount_t hmp);
06ad81ff 1490int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1491int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1492int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1493void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1494void hammer_flusher_finalize(hammer_transaction_t trans, int final);
1b0ab2c3 1495int hammer_flusher_haswork(hammer_mount_t hmp);
6048b411 1496void hammer_flusher_flush_undos(hammer_mount_t hmp, int already_flushed);
059819e3 1497
02428fb6
MD
1498int hammer_recover_stage1(hammer_mount_t hmp, hammer_volume_t rootvol);
1499int hammer_recover_stage2(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1500void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1501 hammer_volume_t root_volume, int final);
f90dde4c 1502
19619882
MD
1503void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1504void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1505void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1506
1507int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1508int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1509int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1510int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
c6b20585 1511void hkprintf(const char *ctl, ...) __printflike(1, 2);
a56cb012
MD
1512udev_t hammer_fsid_to_udev(uuid_t *uuid);
1513
19619882 1514
4a2796f3 1515int hammer_blocksize(int64_t file_offset);
6362a262 1516int hammer_blockoff(int64_t file_offset);
4a2796f3
MD
1517int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1518
0f65be10
MD
1519/*
1520 * Shortcut for _hammer_checkspace(), used all over the code.
1521 */
1522static __inline int
1523hammer_checkspace(hammer_mount_t hmp, int slop)
1524{
1525 return(_hammer_checkspace(hmp, slop, NULL));
1526}
1527
055f5ff8 1528#endif
427e5fc6 1529
8cd0a023 1530static __inline void
af209b0f
MD
1531hammer_wait_mem_record(hammer_record_t record)
1532{
1533 hammer_wait_mem_record_ident(record, "hmmwai");
1534}
1535
1536static __inline void
1537hammer_lock_ex(struct hammer_lock *lock)
1538{
1539 hammer_lock_ex_ident(lock, "hmrlck");
1540}
1541
bcac4bbb
MD
1542/*
1543 * Indicate that a B-Tree node is being modified.
1544 */
af209b0f 1545static __inline void
36f82b23 1546hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1547{
4c286c36 1548 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23 1549 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1550}
1551
1552static __inline void
36f82b23 1553hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1554{
4c286c36 1555 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23
MD
1556 hammer_modify_buffer(trans, node->buffer,
1557 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1558}
1559
bf686dbe 1560static __inline void
36f82b23
MD
1561hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1562 void *base, int len)
bf686dbe 1563{
19619882
MD
1564 hammer_crc_t *crcptr;
1565
bf686dbe
MD
1566 KKASSERT((char *)base >= (char *)node->ondisk &&
1567 (char *)base + len <=
1568 (char *)node->ondisk + sizeof(*node->ondisk));
4c286c36 1569 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23 1570 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1571 crcptr = &node->ondisk->crc;
1572 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1573 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1574}
1575
bcac4bbb
MD
1576/*
1577 * Indicate that the specified modifications have been completed.
1578 *
1579 * Do not try to generate the crc here, it's very expensive to do and a
1580 * sequence of insertions or deletions can result in many calls to this
1581 * function on the same node.
1582 */
bf686dbe 1583static __inline void
10a5d1ba
MD
1584hammer_modify_node_done(hammer_node_t node)
1585{
bcac4bbb
MD
1586 node->flags |= HAMMER_NODE_CRCGOOD;
1587 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1588 node->flags |= HAMMER_NODE_NEEDSCRC;
1589 node->buffer->io.gencrc = 1;
1590 hammer_ref_node(node);
1591 }
10a5d1ba
MD
1592 hammer_modify_buffer_done(node->buffer);
1593}
1594
e8599db1
MD
1595#define hammer_modify_volume_field(trans, vol, field) \
1596 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1597 sizeof((vol)->ondisk->field))
1598
c9b9e29d
MD
1599#define hammer_modify_node_field(trans, node, field) \
1600 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1601 sizeof((node)->ondisk->field))
1602
beec5dc4
MD
1603/*
1604 * The HAMMER_INODE_CAP_DIR_LOCAL_INO capability is set on newly
1605 * created directories for HAMMER version 2 or greater and causes
1606 * directory entries to be placed the inode localization zone in
1607 * the B-Tree instead of the misc zone.
1608 *
1609 * This greatly improves localization between directory entries and
1610 * inodes
1611 */
1612static __inline u_int32_t
1613hammer_dir_localization(hammer_inode_t dip)
1614{
1615 if (dip->ino_data.cap_flags & HAMMER_INODE_CAP_DIR_LOCAL_INO)
1616 return(HAMMER_LOCALIZE_INODE);
1617 else
1618 return(HAMMER_LOCALIZE_MISC);
1619}