hammer - Add tunable vfs.hammer.skip_redo
[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.
8750964d
MD
33 */
34/*
35 * This header file contains structures used internally by the HAMMERFS
c60bb2c5 36 * implementation. See hammer_disk.h for on-disk structures.
8750964d
MD
37 */
38
427e5fc6
MD
39#include <sys/param.h>
40#include <sys/types.h>
41#include <sys/kernel.h>
42c7d26b 42#include <sys/conf.h>
427e5fc6 43#include <sys/systm.h>
8750964d
MD
44#include <sys/tree.h>
45#include <sys/malloc.h>
427e5fc6 46#include <sys/mount.h>
513ca7d7 47#include <sys/mountctl.h>
427e5fc6 48#include <sys/vnode.h>
42c7d26b 49#include <sys/proc.h>
895c1f85 50#include <sys/priv.h>
e63644f0 51#include <sys/stat.h>
427e5fc6 52#include <sys/globaldata.h>
66325755
MD
53#include <sys/lockf.h>
54#include <sys/buf.h>
8cd0a023 55#include <sys/queue.h>
bcac4bbb 56#include <sys/ktr.h>
66325755 57#include <sys/globaldata.h>
9192654c 58#include <sys/limits.h>
c652be54 59#include <vm/vm_extern.h>
66325755
MD
60
61#include <sys/buf2.h>
855942b6 62#include <sys/signal2.h>
81e764f0
MD
63#include <vm/vm_page2.h>
64
427e5fc6 65#include "hammer_disk.h"
8750964d 66#include "hammer_mount.h"
7dc57964 67#include "hammer_ioctl.h"
8750964d
MD
68
69#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
70
71MALLOC_DECLARE(M_HAMMER);
72
bcac4bbb
MD
73/*
74 * Kernel trace
75 */
76#if !defined(KTR_HAMMER)
77#define KTR_HAMMER KTR_ALL
78#endif
879a1b60 79/* KTR_INFO_MASTER_EXTERN(hammer); */
bcac4bbb
MD
80
81/*
82 * Misc structures
83 */
66325755
MD
84struct hammer_mount;
85
8750964d
MD
86/*
87 * Key structure used for custom RB tree inode lookups. This prototypes
88 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
89 */
90typedef struct hammer_inode_info {
513ca7d7 91 int64_t obj_id; /* (key) object identifier */
8750964d 92 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 93 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
94 union {
95 struct hammer_btree_leaf_elm *leaf;
96 } u;
8750964d
MD
97} *hammer_inode_info_t;
98
b84de5af
MD
99typedef enum hammer_transaction_type {
100 HAMMER_TRANS_RO,
101 HAMMER_TRANS_STD,
102 HAMMER_TRANS_FLS
103} hammer_transaction_type_t;
104
66325755
MD
105/*
106 * HAMMER Transaction tracking
107 */
108struct hammer_transaction {
b84de5af 109 hammer_transaction_type_t type;
66325755
MD
110 struct hammer_mount *hmp;
111 hammer_tid_t tid;
ddfdf542 112 u_int64_t time;
dd94f1b1 113 u_int32_t time32;
2f85fa4d 114 int sync_lock_refs;
21fde338 115 int flags;
a89aec1b 116 struct hammer_volume *rootvol;
66325755
MD
117};
118
8cd0a023
MD
119typedef struct hammer_transaction *hammer_transaction_t;
120
21fde338 121#define HAMMER_TRANSF_NEWINODE 0x0001
82010f9f 122#define HAMMER_TRANSF_DIDIO 0x0002
4c286c36 123#define HAMMER_TRANSF_CRCDOM 0x0004 /* EDOM on CRC error, less critical */
21fde338 124
66325755
MD
125/*
126 * HAMMER locks
127 */
427e5fc6 128struct hammer_lock {
250aec18 129 volatile u_int refs; /* active references */
899eb297 130 volatile u_int lockval; /* lock count and control bits */
250aec18
MD
131 struct thread *lowner; /* owner if exclusively held */
132 struct thread *rowner; /* owner if exclusively held */
8750964d
MD
133};
134
250aec18
MD
135#define HAMMER_REFS_LOCKED 0x40000000 /* transition check */
136#define HAMMER_REFS_WANTED 0x20000000 /* transition check */
137#define HAMMER_REFS_CHECK 0x10000000 /* transition check */
138
139#define HAMMER_REFS_FLAGS (HAMMER_REFS_LOCKED | \
140 HAMMER_REFS_WANTED | \
141 HAMMER_REFS_CHECK)
142
899eb297 143#define HAMMER_LOCKF_EXCLUSIVE 0x40000000
250aec18 144#define HAMMER_LOCKF_WANTED 0x20000000
899eb297
MD
145
146static __inline int
147hammer_notlocked(struct hammer_lock *lock)
148{
149 return(lock->lockval == 0);
150}
151
427e5fc6
MD
152static __inline int
153hammer_islocked(struct hammer_lock *lock)
154{
899eb297 155 return(lock->lockval != 0);
427e5fc6
MD
156}
157
250aec18
MD
158/*
159 * Returns the number of refs on the object.
160 */
427e5fc6 161static __inline int
0b075555
MD
162hammer_isactive(struct hammer_lock *lock)
163{
250aec18
MD
164 return(lock->refs & ~HAMMER_REFS_FLAGS);
165}
166
167static __inline int
168hammer_oneref(struct hammer_lock *lock)
169{
170 return((lock->refs & ~HAMMER_REFS_FLAGS) == 1);
0b075555
MD
171}
172
173static __inline int
250aec18 174hammer_norefs(struct hammer_lock *lock)
427e5fc6 175{
250aec18
MD
176 return((lock->refs & ~HAMMER_REFS_FLAGS) == 0);
177}
178
179static __inline int
180hammer_norefsorlock(struct hammer_lock *lock)
181{
182 return(lock->refs == 0);
183}
184
185static __inline int
186hammer_refsorlock(struct hammer_lock *lock)
187{
188 return(lock->refs != 0);
427e5fc6 189}
c60bb2c5 190
8750964d 191/*
7aa3b8a6 192 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
193 */
194static __inline int
7aa3b8a6 195hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 196{
899eb297 197 if ((lock->lockval & HAMMER_LOCKF_EXCLUSIVE) &&
250aec18 198 lock->lowner == td) {
7aa3b8a6 199 return(1);
899eb297 200 }
7aa3b8a6 201 return(0);
6a37e7e4
MD
202}
203
204/*
1f07f686 205 * Flush state, used by various structures
ec4e8497 206 */
1f07f686
MD
207typedef enum hammer_inode_state {
208 HAMMER_FST_IDLE,
209 HAMMER_FST_SETUP,
210 HAMMER_FST_FLUSH
211} hammer_inode_state_t;
ec4e8497 212
1f07f686 213TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497
MD
214
215/*
5fa5c92f
MD
216 * Pseudo-filesystem extended data tracking
217 */
218struct hammer_pfs_rb_tree;
219struct hammer_pseudofs_inmem;
220RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
221RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
222 hammer_pfs_rb_compare, u_int32_t);
223
224struct hammer_pseudofs_inmem {
225 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
226 struct hammer_lock lock;
227 u_int32_t localization;
ea434b6f 228 hammer_tid_t create_tid;
842e7a70 229 int flags;
a56cb012 230 udev_t fsid_udev;
5fa5c92f
MD
231 struct hammer_pseudofs_data pfsd;
232};
233
234typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
235
842e7a70
MD
236#define HAMMER_PFSM_DELETED 0x0001
237
5fa5c92f 238/*
0729c8c8
MD
239 * Cache object ids. A fixed number of objid cache structures are
240 * created to reserve object id's for newly created files in multiples
241 * of 100,000, localized to a particular directory, and recycled as
242 * needed. This allows parallel create operations in different
243 * directories to retain fairly localized object ids which in turn
244 * improves reblocking performance and layout.
245 */
7d29aec0 246#define OBJID_CACHE_SIZE 2048
5a64efa1
MD
247#define OBJID_CACHE_BULK_BITS 10 /* 10 bits (1024) */
248#define OBJID_CACHE_BULK (32 * 32) /* two level (1024) */
249#define OBJID_CACHE_BULK_MASK (OBJID_CACHE_BULK - 1)
250#define OBJID_CACHE_BULK_MASK64 ((u_int64_t)(OBJID_CACHE_BULK - 1))
0729c8c8
MD
251
252typedef struct hammer_objid_cache {
253 TAILQ_ENTRY(hammer_objid_cache) entry;
254 struct hammer_inode *dip;
5a64efa1 255 hammer_tid_t base_tid;
0729c8c8 256 int count;
5a64efa1
MD
257 u_int32_t bm0;
258 u_int32_t bm1[32];
0729c8c8
MD
259} *hammer_objid_cache_t;
260
261/*
bcac4bbb
MD
262 * Associate an inode with a B-Tree node to cache search start positions
263 */
264typedef struct hammer_node_cache {
7a61b85d 265 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
266 struct hammer_node *node;
267 struct hammer_inode *ip;
268} *hammer_node_cache_t;
269
270TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
271
272/*
507df98a
ID
273 * Live dedup cache
274 */
275struct hammer_dedup_crc_rb_tree;
276RB_HEAD(hammer_dedup_crc_rb_tree, hammer_dedup_cache);
277RB_PROTOTYPE2(hammer_dedup_crc_rb_tree, hammer_dedup_cache, crc_entry,
278 hammer_dedup_crc_rb_compare, hammer_crc_t);
279
280struct hammer_dedup_off_rb_tree;
281RB_HEAD(hammer_dedup_off_rb_tree, hammer_dedup_cache);
282RB_PROTOTYPE2(hammer_dedup_off_rb_tree, hammer_dedup_cache, off_entry,
283 hammer_dedup_off_rb_compare, hammer_off_t);
284
285#define DEDUP_CACHE_SIZE 4096 /* XXX make it a dynamic tunable */
286
287typedef struct hammer_dedup_cache {
288 RB_ENTRY(hammer_dedup_cache) crc_entry;
289 RB_ENTRY(hammer_dedup_cache) off_entry;
290 TAILQ_ENTRY(hammer_dedup_cache) lru_entry;
291 struct hammer_mount *hmp;
292 int64_t obj_id;
293 u_int32_t localization;
294 off_t file_offset;
295 int bytes;
296 hammer_off_t data_offset;
297 hammer_crc_t crc;
298} *hammer_dedup_cache_t;
299
300/*
7a61b85d
MD
301 * Structure used to organize flush groups. Flush groups must be
302 * organized into chunks in order to avoid blowing out the UNDO FIFO.
303 * Without this a 'sync' could end up flushing 50,000 inodes in a single
304 * transaction.
305 */
ff003b11
MD
306struct hammer_fls_rb_tree;
307RB_HEAD(hammer_fls_rb_tree, hammer_inode);
308RB_PROTOTYPE(hammer_fls_rb_tree, hammer_inode, rb_flsnode,
309 hammer_ino_rb_compare);
310
7a61b85d
MD
311struct hammer_flush_group {
312 TAILQ_ENTRY(hammer_flush_group) flush_entry;
ff003b11 313 struct hammer_fls_rb_tree flush_tree;
37646115 314 int seq; /* our seq no */
7a61b85d
MD
315 int total_count; /* record load */
316 int running; /* group is running */
317 int closed;
318 int refs;
319};
320
321typedef struct hammer_flush_group *hammer_flush_group_t;
322
323TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
324
325/*
8cd0a023
MD
326 * Structure used to represent an inode in-memory.
327 *
328 * The record and data associated with an inode may be out of sync with
329 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
330 * clear).
331 *
332 * An inode may also hold a cache of unsynchronized records, used for
333 * database and directories only. Unsynchronized regular file data is
334 * stored in the buffer cache.
335 *
336 * NOTE: A file which is created and destroyed within the initial
337 * synchronization period can wind up not doing any disk I/O at all.
338 *
339 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
340 */
341struct hammer_ino_rb_tree;
342struct hammer_inode;
343RB_HEAD(hammer_ino_rb_tree, hammer_inode);
344RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
345 hammer_ino_rb_compare, hammer_inode_info_t);
346
73896937
MD
347struct hammer_redo_rb_tree;
348RB_HEAD(hammer_redo_rb_tree, hammer_inode);
349RB_PROTOTYPE2(hammer_redo_rb_tree, hammer_inode, rb_redonode,
350 hammer_redo_rb_compare, hammer_off_t);
351
8cd0a023
MD
352struct hammer_rec_rb_tree;
353struct hammer_record;
354RB_HEAD(hammer_rec_rb_tree, hammer_record);
355RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 356 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
357
358TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
359
360struct hammer_inode {
1f07f686
MD
361 RB_ENTRY(hammer_inode) rb_node;
362 hammer_inode_state_t flush_state;
7a61b85d 363 hammer_flush_group_t flush_group;
ff003b11 364 RB_ENTRY(hammer_inode) rb_flsnode; /* when on flush list */
73896937 365 RB_ENTRY(hammer_inode) rb_redonode; /* when INODE_RDIRTY is set */
1f07f686 366 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 367 int64_t obj_id; /* (key) object identifier */
b84de5af 368 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 369 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 370 struct hammer_mount *hmp;
0729c8c8 371 hammer_objid_cache_t objid_cache;
b84de5af
MD
372 int flags;
373 int error; /* flush error */
374 int cursor_ip_refs; /* sanity */
e2a02b72 375 int cursor_exclreq_count;
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) */
a117fbeb 441#define HAMMER_INODE_RECSW 0x0400 /* waiting on data record flush */
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 */
e2a02b72 458#define HAMMER_INODE_SLAVEFLUSH 0x08000000 /* being flushed by slave */
b84de5af 459
9192654c 460#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY|HAMMER_INODE_SDIRTY| \
11ad5ade 461 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
462 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
463 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686 464
9192654c 465#define HAMMER_INODE_MODMASK_NOXDIRTY \
1f07f686 466 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 467
9192654c
MD
468#define HAMMER_INODE_MODMASK_NOREDO \
469 (HAMMER_INODE_DDIRTY| \
470 HAMMER_INODE_XDIRTY| \
471 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
472
f90dde4c 473#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 474#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 475
7bc5b8c2
MD
476/*
477 * Used by the inode reclaim code to pipeline reclaims and avoid
478 * blowing out kernel memory or letting the flusher get too far
82010f9f
MD
479 * behind. The reclaim wakes up when count reaches 0 or the
480 * timer expires.
7bc5b8c2
MD
481 */
482struct hammer_reclaim {
483 TAILQ_ENTRY(hammer_reclaim) entry;
82010f9f 484 int count;
7bc5b8c2
MD
485};
486
8750964d 487/*
e98f1b96
MD
488 * Track who is creating the greatest burden on the
489 * inode cache.
490 */
491struct hammer_inostats {
492 pid_t pid; /* track user process */
493 int ltick; /* last tick */
494 int count; /* count (degenerates) */
495};
496
497#define HAMMER_INOSTATS_HSIZE 32
498#define HAMMER_INOSTATS_HMASK (HAMMER_INOSTATS_HSIZE - 1)
499
500/*
1f07f686
MD
501 * Structure used to represent an unsynchronized record in-memory. These
502 * records typically represent directory entries. Only non-historical
503 * records are kept in-memory.
504 *
505 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
506 * on disk then neither are any records and the in-memory record tree
507 * represents the entire contents of the inode. If the inode is on disk
508 * then the on-disk B-Tree is scanned in parallel with the in-memory
509 * RB-Tree to synthesize the current state of the file.
510 *
1f07f686
MD
511 * Records are also used to enforce the ordering of directory create/delete
512 * operations. A new inode will not be flushed to disk unless its related
513 * directory entry is also being flushed at the same time. A directory entry
514 * will not be removed unless its related inode is also being removed at the
515 * same time.
8750964d 516 */
1f07f686 517typedef enum hammer_record_type {
e8599db1 518 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 519 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 520 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
521 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
522 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
523} hammer_record_type_t;
524
8cd0a023
MD
525struct hammer_record {
526 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
527 TAILQ_ENTRY(hammer_record) target_entry;
528 hammer_inode_state_t flush_state;
7a61b85d 529 hammer_flush_group_t flush_group;
1f07f686 530 hammer_record_type_t type;
a89aec1b 531 struct hammer_lock lock;
0832c9bb 532 struct hammer_reserve *resv;
8cd0a023 533 struct hammer_inode *ip;
1f07f686 534 struct hammer_inode *target_ip;
11ad5ade 535 struct hammer_btree_leaf_elm leaf;
8cd0a023 536 union hammer_data_ondisk *data;
8cd0a023 537 int flags;
77912481 538 int gflags;
e469566b 539 hammer_off_t zone2_offset; /* direct-write only */
8cd0a023 540};
8750964d 541
8cd0a023
MD
542typedef struct hammer_record *hammer_record_t;
543
d36ec43b
MD
544/*
545 * Record flags. Note that FE can only be set by the frontend if the
546 * record has not been interlocked by the backend w/ BE.
547 */
8cd0a023
MD
548#define HAMMER_RECF_ALLOCDATA 0x0001
549#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 550#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 551#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
1b0ab2c3 552#define HAMMER_RECF_COMMITTED 0x0010 /* committed to the B-Tree */
d36ec43b 553#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 554#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
507df98a 555#define HAMMER_RECF_DEDUPED 0x0080 /* will be live-dedup'ed */
1b0ab2c3 556#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
47f363f1 557#define HAMMER_RECF_REDO 0x1000 /* REDO was laid down */
8cd0a023
MD
558
559/*
77912481
MD
560 * These flags must be separate to deal with SMP races
561 */
562#define HAMMER_RECG_DIRECT_IO 0x0001 /* related direct I/O running*/
563#define HAMMER_RECG_DIRECT_WAIT 0x0002 /* related direct I/O running*/
564#define HAMMER_RECG_DIRECT_INVAL 0x0004 /* buffer alias invalidation */
565/*
83f2a3aa 566 * hammer_create_at_cursor() and hammer_delete_at_cursor() flags.
602c6cb8 567 */
83f2a3aa
MD
568#define HAMMER_CREATE_MODE_UMIRROR 0x0001
569#define HAMMER_CREATE_MODE_SYS 0x0002
570
602c6cb8
MD
571#define HAMMER_DELETE_ADJUST 0x0001
572#define HAMMER_DELETE_DESTROY 0x0002
573
574/*
47197d71 575 * In-memory structures representing on-disk structures.
8cd0a023 576 */
8750964d 577struct hammer_volume;
427e5fc6 578struct hammer_buffer;
8cd0a023 579struct hammer_node;
e8599db1 580struct hammer_undo;
0832c9bb
MD
581struct hammer_reserve;
582
8750964d 583RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 584RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 585RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 586RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 587RB_HEAD(hammer_res_rb_tree, hammer_reserve);
1afb73cf 588RB_HEAD(hammer_mod_rb_tree, hammer_io);
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);
1afb73cf
MD
600RB_PROTOTYPE2(hammer_mod_rb_tree, hammer_io, rb_node,
601 hammer_mod_rb_compare, hammer_off_t);
8750964d 602
66325755
MD
603/*
604 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
605 *
606 * VOLUME - hammer_volume containing meta-data
607 * META_BUFFER - hammer_buffer containing meta-data
608 * DATA_BUFFER - hammer_buffer containing pure-data
609 *
610 * Dirty volume headers and dirty meta-data buffers are locked until the
611 * flusher can sequence them out. Dirty pure-data buffers can be written.
612 * Clean buffers can be passively released.
66325755 613 */
10a5d1ba
MD
614typedef enum hammer_io_type {
615 HAMMER_STRUCTURE_VOLUME,
616 HAMMER_STRUCTURE_META_BUFFER,
617 HAMMER_STRUCTURE_UNDO_BUFFER,
eddadaee
MD
618 HAMMER_STRUCTURE_DATA_BUFFER,
619 HAMMER_STRUCTURE_DUMMY
10a5d1ba 620} hammer_io_type_t;
66325755
MD
621
622union hammer_io_structure;
055f5ff8 623struct hammer_io;
66325755
MD
624
625struct worklist {
626 LIST_ENTRY(worklist) node;
627};
628
10a5d1ba
MD
629TAILQ_HEAD(hammer_io_list, hammer_io);
630typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 631
66325755 632struct hammer_io {
10a5d1ba
MD
633 struct worklist worklist;
634 struct hammer_lock lock;
635 enum hammer_io_type type;
636 struct hammer_mount *hmp;
748efb59 637 struct hammer_volume *volume;
1afb73cf 638 RB_ENTRY(hammer_io) rb_node; /* if modified */
eddadaee 639 TAILQ_ENTRY(hammer_io) iorun_entry; /* iorun_list */
1afb73cf 640 struct hammer_mod_rb_tree *mod_root;
10a5d1ba 641 struct buf *bp;
4a2796f3
MD
642 int64_t offset; /* zone-2 offset */
643 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
644 int modify_refs;
645
77912481
MD
646 /*
647 * These can be modified at any time by the backend while holding
648 * io_token, due to bio_done and hammer_io_complete() callbacks.
649 */
055f5ff8
MD
650 u_int running : 1; /* bp write IO in progress */
651 u_int waiting : 1; /* someone is waiting on us */
77912481
MD
652 u_int ioerror : 1; /* abort on io-error */
653 u_int unusedA : 29;
654
655 /*
656 * These can only be modified by the frontend while holding
657 * fs_token, or by the backend while holding the io interlocked
658 * with no references (which will block the frontend when it
659 * tries to reference it).
660 *
661 * WARNING! SMP RACES will create havoc if the callbacks ever tried
662 * to modify any of these outside the above restrictions.
663 */
664 u_int modified : 1; /* bp's data was modified */
665 u_int released : 1; /* bp released (w/ B_LOCKED set) */
b33e2cc0 666 u_int validated : 1; /* ondisk has been validated */
b58c6388 667 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 668 u_int recovered : 1; /* has recovery ref */
9f5097dc 669 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 670 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 671 u_int gencrc : 1; /* crc needs to be generated */
77912481 672 u_int unusedB : 24;
66325755
MD
673};
674
8cd0a023
MD
675typedef struct hammer_io *hammer_io_t;
676
af209b0f
MD
677#define HAMMER_CLUSTER_SIZE (64 * 1024)
678#if HAMMER_CLUSTER_SIZE > MAXBSIZE
679#undef HAMMER_CLUSTER_SIZE
680#define HAMMER_CLUSTER_SIZE MAXBSIZE
681#endif
682#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
683
66325755 684/*
8cd0a023 685 * In-memory volume representing on-disk buffer
66325755 686 */
8750964d 687struct hammer_volume {
66325755 688 struct hammer_io io;
8750964d 689 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
690 struct hammer_volume_ondisk *ondisk;
691 int32_t vol_no;
fbc6e32a 692 int64_t nblocks; /* note: special calculation for statfs */
47197d71 693 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
694 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
695 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 696 char *vol_name;
8750964d 697 struct vnode *devvp;
427e5fc6 698 int vol_flags;
8750964d
MD
699};
700
8cd0a023
MD
701typedef struct hammer_volume *hammer_volume_t;
702
66325755 703/*
8cd0a023
MD
704 * In-memory buffer (other then volume, super-cluster, or cluster),
705 * representing an on-disk buffer.
66325755 706 */
427e5fc6 707struct hammer_buffer {
66325755 708 struct hammer_io io;
427e5fc6 709 RB_ENTRY(hammer_buffer) rb_node;
47197d71 710 void *ondisk;
34d829f7 711 hammer_off_t zoneX_offset;
0832c9bb 712 hammer_off_t zone2_offset;
cebe9493 713 struct hammer_reserve *resv;
8cd0a023 714 struct hammer_node_list clist;
66325755
MD
715};
716
8cd0a023
MD
717typedef struct hammer_buffer *hammer_buffer_t;
718
719/*
720 * In-memory B-Tree node, representing an on-disk B-Tree node.
721 *
722 * This is a hang-on structure which is backed by a hammer_buffer,
723 * indexed by a hammer_cluster, and used for fine-grained locking of
724 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
725 * can contain multiple nodes representing wildly disassociated portions
726 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
727 *
728 * This structure uses a cluster-relative index to reduce the number
729 * of layers required to access it, and also because all on-disk B-Tree
730 * references are cluster-relative offsets.
731 */
732struct hammer_node {
733 struct hammer_lock lock; /* node-by-node lock */
734 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
735 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 736 hammer_off_t node_offset; /* full offset spec */
40043e7f 737 struct hammer_mount *hmp;
8cd0a023
MD
738 struct hammer_buffer *buffer; /* backing buffer */
739 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 740 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 741 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 742 int flags;
e2a02b72 743 int cursor_exclreq_count;
8cd0a023
MD
744};
745
b3deaf57
MD
746#define HAMMER_NODE_DELETED 0x0001
747#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
748#define HAMMER_NODE_CRCGOOD 0x0004
749#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 750#define HAMMER_NODE_NEEDSMIRROR 0x0010
4c286c36 751#define HAMMER_NODE_CRCBAD 0x0020
fa2b9a03 752#define HAMMER_NODE_NONLINEAR 0x0040 /* linear heuristic */
4c286c36
MD
753
754#define HAMMER_NODE_CRCANY (HAMMER_NODE_CRCGOOD | HAMMER_NODE_CRCBAD)
b3deaf57 755
8cd0a023
MD
756typedef struct hammer_node *hammer_node_t;
757
758/*
1775b6a0
MD
759 * List of locked nodes. This structure is used to lock potentially large
760 * numbers of nodes as an aid for complex B-Tree operations.
b33e2cc0 761 */
1775b6a0
MD
762struct hammer_node_lock;
763TAILQ_HEAD(hammer_node_lock_list, hammer_node_lock);
764
765struct hammer_node_lock {
766 TAILQ_ENTRY(hammer_node_lock) entry;
767 struct hammer_node_lock_list list;
768 struct hammer_node_lock *parent;
b33e2cc0 769 hammer_node_t node;
1775b6a0
MD
770 hammer_node_ondisk_t copy; /* copy of on-disk data */
771 int index; /* index of this node in parent */
772 int count; /* count children */
773 int flags;
b33e2cc0
MD
774};
775
1775b6a0 776typedef struct hammer_node_lock *hammer_node_lock_t;
b33e2cc0 777
1775b6a0 778#define HAMMER_NODE_LOCK_UPDATED 0x0001
24cf83d2 779#define HAMMER_NODE_LOCK_LCACHE 0x0002
b33e2cc0
MD
780
781/*
8cd0a023
MD
782 * Common I/O management structure - embedded in in-memory structures
783 * which are backed by filesystem buffers.
784 */
66325755
MD
785union hammer_io_structure {
786 struct hammer_io io;
787 struct hammer_volume volume;
66325755 788 struct hammer_buffer buffer;
8750964d
MD
789};
790
055f5ff8
MD
791typedef union hammer_io_structure *hammer_io_structure_t;
792
bf686dbe 793/*
cebe9493
MD
794 * The reserve structure prevents the blockmap from allocating
795 * out of a reserved bigblock. Such reservations are used by
796 * the direct-write mechanism.
797 *
798 * The structure is also used to hold off on reallocations of
799 * big blocks from the freemap until flush dependancies have
800 * been dealt with.
801 */
0832c9bb
MD
802struct hammer_reserve {
803 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
804 TAILQ_ENTRY(hammer_reserve) delay_entry;
805 int flush_group;
cb51be26 806 int flags;
0832c9bb 807 int refs;
cb51be26 808 int zone;
df301614 809 int append_off;
507df98a 810 int32_t bytes_free;
cebe9493 811 hammer_off_t zone_offset;
0832c9bb
MD
812};
813
814typedef struct hammer_reserve *hammer_reserve_t;
815
cb51be26 816#define HAMMER_RESF_ONDELAY 0x0001
5e435c92 817#define HAMMER_RESF_LAYER2FREE 0x0002
cb51be26 818
8cd0a023
MD
819#include "hammer_cursor.h"
820
8750964d 821/*
cebe9493
MD
822 * The undo structure tracks recent undos to avoid laying down duplicate
823 * undos within a flush group, saving us a significant amount of overhead.
824 *
825 * This is strictly a heuristic.
e8599db1 826 */
7a61b85d
MD
827#define HAMMER_MAX_UNDOS 1024
828#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
829
830struct hammer_undo {
831 RB_ENTRY(hammer_undo) rb_node;
832 TAILQ_ENTRY(hammer_undo) lru_entry;
833 hammer_off_t offset;
834 int bytes;
835};
836
837typedef struct hammer_undo *hammer_undo_t;
838
af209b0f 839struct hammer_flusher_info;
7a61b85d 840TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
841
842struct hammer_flusher {
843 int signal; /* flusher thread sequencer */
e86903d8 844 int done; /* last completed flush group */
37646115 845 int next; /* next unallocated flg seqno */
da2da375
MD
846 int group_lock; /* lock sequencing of the next flush */
847 int exiting; /* request master exit */
da2da375
MD
848 thread_t td; /* master flusher thread */
849 hammer_tid_t tid; /* last flushed transaction id */
850 int finalize_want; /* serialize finalization */
851 struct hammer_lock finalize_lock; /* serialize finalization */
852 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
853 struct hammer_flusher_info_list run_list;
854 struct hammer_flusher_info_list ready_list;
da2da375
MD
855};
856
9192654c
MD
857#define HAMMER_FLUSH_UNDOS_RELAXED 0
858#define HAMMER_FLUSH_UNDOS_FORCED 1
859#define HAMMER_FLUSH_UNDOS_AUTO 2
da2da375 860/*
8750964d
MD
861 * Internal hammer mount data structure
862 */
863struct hammer_mount {
864 struct mount *mp;
27ea2398 865 /*struct vnode *rootvp;*/
8750964d 866 struct hammer_ino_rb_tree rb_inos_root;
73896937 867 struct hammer_redo_rb_tree rb_redo_root;
8750964d 868 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 869 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 870 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
871 struct hammer_res_rb_tree rb_resv_root;
872 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 873 struct hammer_pfs_rb_tree rb_pfsm_root;
507df98a
ID
874
875 struct hammer_dedup_crc_rb_tree rb_dedup_crc_root;
876 struct hammer_dedup_off_rb_tree rb_dedup_off_root;
877
8750964d 878 struct hammer_volume *rootvol;
47197d71
MD
879 struct hammer_base_elm root_btree_beg;
880 struct hammer_base_elm root_btree_end;
bac808fe
MD
881
882 struct malloc_type *m_misc;
883 struct malloc_type *m_inodes;
884
cdb6e4e6 885 int flags; /* HAMMER_MOUNT_xxx flags */
195c19a1
MD
886 int hflags;
887 int ronly;
888 int nvolumes;
d26d0ae9 889 int volume_iterator;
732a1697 890 int master_id; /* -1 or 0-15 - clustering and mirroring */
44a83111 891 int version; /* hammer filesystem version to use */
e63644f0 892 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 893 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 894 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 895 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 896 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
897 int last_newrecords;
898 int count_newrecords;
da2da375 899
865c9609
MN
900 int volume_to_remove; /* volume that is currently being removed */
901
9f5097dc 902 int count_inodes; /* total number of inodes */
af209b0f 903 int count_iqueued; /* inodes queued to flusher */
e2a02b72 904 int count_reclaims; /* inodes pending reclaim by flusher */
da2da375
MD
905
906 struct hammer_flusher flusher;
907
855942b6 908 u_int check_interrupt;
3e583440 909 u_int check_yield;
8750964d 910 uuid_t fsid;
1afb73cf
MD
911 struct hammer_mod_rb_tree volu_root; /* dirty undo buffers */
912 struct hammer_mod_rb_tree undo_root; /* dirty undo buffers */
913 struct hammer_mod_rb_tree data_root; /* dirty data buffers */
914 struct hammer_mod_rb_tree meta_root; /* dirty meta bufs */
915 struct hammer_mod_rb_tree lose_root; /* loose buffers */
3583bbb4
MD
916 long locked_dirty_space; /* meta/volu count */
917 long io_running_space; /* io_token */
3038a8ca 918 int unused01;
0729c8c8 919 int objid_cache_count;
507df98a 920 int dedup_cache_count;
cdb6e4e6
MD
921 int error; /* critical I/O error */
922 struct krate krate; /* rate limited kprintf */
ddfdf542 923 hammer_tid_t asof; /* snapshot mount */
4889cbd4
MD
924 hammer_tid_t next_tid;
925 hammer_tid_t flush_tid1; /* flusher tid sequencing */
926 hammer_tid_t flush_tid2; /* flusher tid sequencing */
e63644f0 927 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
02428fb6 928 u_int32_t undo_seqno; /* UNDO/REDO FIFO seqno */
c58123da
MD
929 u_int32_t recover_stage2_seqno; /* REDO recovery seqno */
930 hammer_off_t recover_stage2_offset; /* REDO recovery offset */
e63644f0 931
513ca7d7 932 struct netexport export;
9480ff55 933 struct hammer_lock sync_lock;
c9b9e29d 934 struct hammer_lock free_lock;
d99d6bf5
MD
935 struct hammer_lock undo_lock;
936 struct hammer_lock blkmap_lock;
83f2a3aa 937 struct hammer_lock snapshot_lock;
52e547e3 938 struct hammer_lock volume_lock;
0729c8c8 939 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
940 struct hammer_undo undos[HAMMER_MAX_UNDOS];
941 int undo_alloc;
942 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 943 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 944 struct hammer_flush_group_list flush_group_list;
37646115 945 hammer_flush_group_t fill_flush_group;
7b6ccb11 946 hammer_flush_group_t next_flush_group;
0729c8c8 947 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
507df98a
ID
948 TAILQ_HEAD(, hammer_dedup_cache) dedup_lru_list;
949 hammer_dedup_cache_t dedup_free_cache;
7bc5b8c2 950 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
eddadaee 951 TAILQ_HEAD(, hammer_io) iorun_list;
e98f1b96 952
b0aab9b9
MD
953 struct lwkt_token fs_token; /* high level */
954 struct lwkt_token io_token; /* low level (IO callback) */
955
e98f1b96 956 struct hammer_inostats inostats[HAMMER_INOSTATS_HSIZE];
8750964d
MD
957};
958
8cd0a023 959typedef struct hammer_mount *hammer_mount_t;
8750964d 960
cdb6e4e6 961#define HAMMER_MOUNT_CRITICAL_ERROR 0x0001
82010f9f 962#define HAMMER_MOUNT_FLUSH_RECOVERY 0x0002
47f363f1 963#define HAMMER_MOUNT_REDO_SYNC 0x0004
c58123da
MD
964#define HAMMER_MOUNT_REDO_RECOVERY_REQ 0x0008
965#define HAMMER_MOUNT_REDO_RECOVERY_RUN 0x0010
9f5097dc 966
fbc6e32a
MD
967struct hammer_sync_info {
968 int error;
969 int waitfor;
970};
971
f6468e9e
MD
972/*
973 * Minium buffer cache bufs required to rebalance the B-Tree.
974 * This is because we must hold the children and the children's children
975 * locked. Even this might not be enough if things are horribly out
976 * of balance.
977 */
978#define HAMMER_REBALANCE_MIN_BUFS \
979 (HAMMER_BTREE_LEAF_ELMS * HAMMER_BTREE_LEAF_ELMS)
980
981
8750964d
MD
982#endif
983
a7e9bef1
MD
984/*
985 * checkspace slop (8MB chunks), higher numbers are more conservative.
986 */
93291532
MD
987#define HAMMER_CHKSPC_REBLOCK 25
988#define HAMMER_CHKSPC_MIRROR 20
989#define HAMMER_CHKSPC_WRITE 20
990#define HAMMER_CHKSPC_CREATE 20
991#define HAMMER_CHKSPC_REMOVE 10
7b6ccb11 992#define HAMMER_CHKSPC_EMERGENCY 0
a7e9bef1 993
8750964d
MD
994#if defined(_KERNEL)
995
996extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
997extern struct vop_ops hammer_spec_vops;
998extern struct vop_ops hammer_fifo_vops;
66325755 999extern struct bio_ops hammer_bioops;
427e5fc6 1000
dbd4f600 1001extern int hammer_skip_redo;
2f85fa4d 1002extern int hammer_debug_io;
d5ef456e 1003extern int hammer_debug_general;
77062c8a 1004extern int hammer_debug_debug;
e8599db1 1005extern int hammer_debug_inode;
7d683b0f 1006extern int hammer_debug_locks;
b3deaf57 1007extern int hammer_debug_btree;
d113fda1 1008extern int hammer_debug_tid;
b33e2cc0 1009extern int hammer_debug_recover;
46fe7ae1 1010extern int hammer_debug_recover_faults;
fc73edd8 1011extern int hammer_debug_critical;
1b0ab2c3 1012extern int hammer_cluster_enable;
507df98a 1013extern int hammer_live_dedup;
e2a02b72 1014extern int hammer_tdmux_ticks;
7a61b85d 1015extern int hammer_count_fsyncs;
b3deaf57 1016extern int hammer_count_inodes;
af209b0f 1017extern int hammer_count_iqueued;
e2a02b72 1018extern int hammer_count_reclaims;
b3deaf57
MD
1019extern int hammer_count_records;
1020extern int hammer_count_record_datas;
1021extern int hammer_count_volumes;
b3deaf57
MD
1022extern int hammer_count_buffers;
1023extern int hammer_count_nodes;
a7e9bef1 1024extern int64_t hammer_count_extra_space_used;
cb51be26
MD
1025extern int64_t hammer_stats_btree_lookups;
1026extern int64_t hammer_stats_btree_searches;
1027extern int64_t hammer_stats_btree_inserts;
1028extern int64_t hammer_stats_btree_deletes;
1029extern int64_t hammer_stats_btree_elements;
1030extern int64_t hammer_stats_btree_splits;
1031extern int64_t hammer_stats_btree_iterations;
39d8fd63 1032extern int64_t hammer_stats_btree_root_iterations;
cb51be26 1033extern int64_t hammer_stats_record_iterations;
ce0138a6
MD
1034extern int64_t hammer_stats_file_read;
1035extern int64_t hammer_stats_file_write;
1036extern int64_t hammer_stats_file_iopsr;
1037extern int64_t hammer_stats_file_iopsw;
1038extern int64_t hammer_stats_disk_read;
1039extern int64_t hammer_stats_disk_write;
1040extern int64_t hammer_stats_inode_flushes;
1041extern int64_t hammer_stats_commits;
89e744ce 1042extern int64_t hammer_stats_undo;
6048b411 1043extern int64_t hammer_stats_redo;
3583bbb4 1044extern long hammer_count_dirtybufspace;
a99b9ea2 1045extern int hammer_count_refedbufs;
0832c9bb 1046extern int hammer_count_reservations;
3583bbb4
MD
1047extern long hammer_count_io_running_read;
1048extern long hammer_count_io_running_write;
a99b9ea2 1049extern int hammer_count_io_locked;
3583bbb4 1050extern long hammer_limit_dirtybufspace;
47637bff 1051extern int hammer_limit_recs;
de996e86 1052extern int hammer_limit_inode_recs;
e2a02b72 1053extern int hammer_limit_reclaims;
e2ef7a95 1054extern int hammer_live_dedup_cache_size;
9192654c 1055extern int hammer_limit_redo;
1f07f686 1056extern int hammer_bio_count;
cb51be26 1057extern int hammer_verify_zone;
1b0ab2c3 1058extern int hammer_verify_data;
cb51be26 1059extern int hammer_write_mode;
b4f86ea3 1060extern int hammer_double_buffer;
3e583440 1061extern int hammer_yield_check;
6f3d87c0 1062extern int hammer_fsync_mode;
21fde338 1063extern int hammer_autoflush;
7d683b0f 1064extern int64_t hammer_contention_count;
b3deaf57 1065
507df98a
ID
1066extern int64_t hammer_live_dedup_vnode_bcmps;
1067extern int64_t hammer_live_dedup_device_bcmps;
1068extern int64_t hammer_live_dedup_findblk_failures;
1069extern int64_t hammer_live_dedup_bmap_saves;
1070
cdb6e4e6
MD
1071void hammer_critical_error(hammer_mount_t hmp, hammer_inode_t ip,
1072 int error, const char *msg);
8750964d
MD
1073int hammer_vop_inactive(struct vop_inactive_args *);
1074int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 1075int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 1076struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 1077 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
1078 hammer_tid_t asof, u_int32_t localization,
1079 int flags, int *errorp);
4c286c36
MD
1080struct hammer_inode *hammer_get_dummy_inode(hammer_transaction_t trans,
1081 hammer_inode_t dip, int64_t obj_id,
1082 hammer_tid_t asof, u_int32_t localization,
1083 int flags, int *errorp);
39d8fd63
MD
1084struct hammer_inode *hammer_find_inode(hammer_transaction_t trans,
1085 int64_t obj_id, hammer_tid_t asof,
1086 u_int32_t localization);
43c665ae
MD
1087void hammer_scan_inode_snapshots(hammer_mount_t hmp,
1088 hammer_inode_info_t iinfo,
1089 int (*callback)(hammer_inode_t ip, void *data),
1090 void *data);
66325755
MD
1091void hammer_put_inode(struct hammer_inode *ip);
1092void hammer_put_inode_ref(struct hammer_inode *ip);
e98f1b96 1093void hammer_inode_waitreclaims(hammer_transaction_t trans);
66325755 1094
8cd0a023 1095int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
1096int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
1097
07be83b8 1098int hammer_unload_buffer(hammer_buffer_t buffer, void *data);
7c19b529
MN
1099int hammer_install_volume(hammer_mount_t hmp, const char *volname,
1100 struct vnode *devvp);
1b0ab2c3 1101int hammer_mountcheck_volumes(hammer_mount_t hmp);
8cd0a023 1102
e469566b 1103int hammer_mem_add(hammer_record_t record);
45a014dc 1104int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 1105int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 1106int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 1107int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
1108int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
1109 hammer_tid_t tid);
83f2a3aa
MD
1110int hammer_create_at_cursor(hammer_cursor_t cursor,
1111 hammer_btree_leaf_elm_t leaf, void *udata, int mode);
602c6cb8 1112int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
1113 hammer_tid_t delete_tid, u_int32_t delete_ts,
1114 int track, int64_t *stat_bytes);
b3deaf57 1115int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 1116 hammer_inode_t ip);
fbc6e32a 1117int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
1118int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
1119
8cd0a023 1120hammer_record_t
11ad5ade 1121 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 1122void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 1123void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 1124void hammer_rel_mem_record(hammer_record_t record);
b84de5af 1125
6a37e7e4 1126int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 1127int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 1128int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 1129int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 1130int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 1131void hammer_cursor_downgrade(hammer_cursor_t cursor);
bb29b5d8
MD
1132int hammer_cursor_upgrade2(hammer_cursor_t c1, hammer_cursor_t c2);
1133void hammer_cursor_downgrade2(hammer_cursor_t c1, hammer_cursor_t c2);
32c90105
MD
1134int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
1135 int index);
af209b0f 1136void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
1137int hammer_lock_ex_try(struct hammer_lock *lock);
1138void hammer_lock_sh(struct hammer_lock *lock);
47637bff 1139int hammer_lock_sh_try(struct hammer_lock *lock);
bb29b5d8
MD
1140int hammer_lock_upgrade(struct hammer_lock *lock, int shcount);
1141void hammer_lock_downgrade(struct hammer_lock *lock, int shcount);
b3bad96f 1142int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 1143void hammer_unlock(struct hammer_lock *lock);
66325755 1144void hammer_ref(struct hammer_lock *lock);
250aec18
MD
1145int hammer_ref_interlock(struct hammer_lock *lock);
1146int hammer_ref_interlock_true(struct hammer_lock *lock);
1147void hammer_ref_interlock_done(struct hammer_lock *lock);
1148void hammer_rel(struct hammer_lock *lock);
1149int hammer_rel_interlock(struct hammer_lock *lock, int locked);
1150void hammer_rel_interlock_done(struct hammer_lock *lock, int orig_locked);
1151int hammer_get_interlock(struct hammer_lock *lock);
1152int hammer_try_interlock_norefs(struct hammer_lock *lock);
1153void hammer_put_interlock(struct hammer_lock *lock, int error);
8cd0a023 1154
2f85fa4d
MD
1155void hammer_sync_lock_ex(hammer_transaction_t trans);
1156void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 1157int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
1158void hammer_sync_unlock(hammer_transaction_t trans);
1159
66325755 1160u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 1161void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
1162void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
1163u_int64_t hammer_timespec_to_time(struct timespec *ts);
bc6c1f13
MD
1164int hammer_str_to_tid(const char *str, int *ispfsp,
1165 hammer_tid_t *tidp, u_int32_t *localizationp);
1166int hammer_is_atatext(const char *name, int len);
5a64efa1
MD
1167hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip,
1168 int64_t namekey);
0729c8c8
MD
1169void hammer_clear_objid(hammer_inode_t dip);
1170void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 1171
507df98a
ID
1172int hammer_dedup_crc_rb_compare(hammer_dedup_cache_t dc1,
1173 hammer_dedup_cache_t dc2);
1174int hammer_dedup_off_rb_compare(hammer_dedup_cache_t dc1,
1175 hammer_dedup_cache_t dc2);
1176hammer_dedup_cache_t hammer_dedup_cache_add(hammer_inode_t ip,
1177 hammer_btree_leaf_elm_t leaf);
1178hammer_dedup_cache_t hammer_dedup_cache_lookup(hammer_mount_t hmp,
1179 hammer_crc_t crc);
1180void hammer_dedup_cache_inval(hammer_mount_t hmp, hammer_off_t base_offset);
1181void hammer_destroy_dedup_cache(hammer_mount_t hmp);
1182void hammer_dump_dedup_cache(hammer_mount_t hmp);
1183int hammer_dedup_validate(hammer_dedup_cache_t dcp, int zone, int bytes,
1184 void *data);
1185
e8599db1 1186int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 1187 int bytes);
e8599db1 1188void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 1189enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 1190int hammer_get_dtype(u_int8_t obj_type);
66325755 1191u_int8_t hammer_get_obj_type(enum vtype vtype);
5e435c92
MD
1192int64_t hammer_directory_namekey(hammer_inode_t dip, const void *name, int len,
1193 u_int32_t *max_iterationsp);
e63644f0 1194int hammer_nohistory(hammer_inode_t ip);
427e5fc6 1195
36f82b23 1196int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 1197 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 1198void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 1199void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 1200int hammer_recover_cursor(hammer_cursor_t cursor);
982be4bf
MD
1201void hammer_unlock_cursor(hammer_cursor_t cursor);
1202int hammer_lock_cursor(hammer_cursor_t cursor);
3f43fb33
MD
1203hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
1204void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
1205
1206void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
1207void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
1208 int index);
1209void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
1210 int index);
bbb01e14
MD
1211void hammer_cursor_moved_element(hammer_node_t oparent, int pindex,
1212 hammer_node_t onode, int oindex,
1213 hammer_node_t nnode, int nindex);
1775b6a0
MD
1214void hammer_cursor_parent_changed(hammer_node_t node, hammer_node_t oparent,
1215 hammer_node_t nparent, int nindex);
b3bad96f
MD
1216void hammer_cursor_inserted_element(hammer_node_t node, int index);
1217void hammer_cursor_deleted_element(hammer_node_t node, int index);
b9107f58 1218void hammer_cursor_invalidate_cache(hammer_cursor_t cursor);
8cd0a023
MD
1219
1220int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 1221int hammer_btree_first(hammer_cursor_t cursor);
32c90105 1222int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
1223int hammer_btree_extract(hammer_cursor_t cursor, int flags);
1224int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 1225int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 1226int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 1227 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 1228int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
1229void hammer_btree_do_propagation(hammer_cursor_t cursor,
1230 hammer_pseudofs_inmem_t pfsm,
602c6cb8 1231 hammer_btree_leaf_elm_t leaf);
8cd0a023 1232int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 1233int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
1234int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
1235int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
1236
2f85fa4d
MD
1237int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
1238 hammer_btree_elm_t elm);
1775b6a0 1239void hammer_node_lock_init(hammer_node_lock_t parent, hammer_node_t node);
24cf83d2
MD
1240void hammer_btree_lcache_init(hammer_mount_t hmp, hammer_node_lock_t lcache,
1241 int depth);
1242void hammer_btree_lcache_free(hammer_mount_t hmp, hammer_node_lock_t lcache);
1775b6a0 1243int hammer_btree_lock_children(hammer_cursor_t cursor, int depth,
24cf83d2
MD
1244 hammer_node_lock_t parent,
1245 hammer_node_lock_t lcache);
1775b6a0
MD
1246void hammer_btree_lock_copy(hammer_cursor_t cursor,
1247 hammer_node_lock_t parent);
7ddc70d1 1248int hammer_btree_sync_copy(hammer_cursor_t cursor,
1775b6a0 1249 hammer_node_lock_t parent);
24cf83d2
MD
1250void hammer_btree_unlock_children(hammer_mount_t hmp,
1251 hammer_node_lock_t parent,
1252 hammer_node_lock_t lcache);
bcac4bbb 1253int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
82010f9f
MD
1254hammer_node_t hammer_btree_get_parent(hammer_transaction_t trans,
1255 hammer_node_t node, int *parent_indexp,
c82af904 1256 int *errorp, int try_exclusive);
b33e2cc0 1257
c0ade690
MD
1258void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
1259void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 1260
47197d71
MD
1261void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
1262 int *errorp, struct hammer_buffer **bufferp);
1263void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
1264 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
1265void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
1266 int *errorp, struct hammer_buffer **bufferp);
1267void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
1268 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
1269
1270hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 1271
8cd0a023 1272hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 1273 int32_t vol_no, int *errorp);
4a2796f3
MD
1274hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
1275 int bytes, int isnew, int *errorp);
1b0ab2c3
MD
1276void hammer_sync_buffers(hammer_mount_t hmp,
1277 hammer_off_t base_offset, int bytes);
362ec2dc 1278int hammer_del_buffers(hammer_mount_t hmp,
1b0ab2c3 1279 hammer_off_t base_offset,
362ec2dc
MD
1280 hammer_off_t zone2_offset, int bytes,
1281 int report_conflicts);
8cd0a023 1282
fbc6e32a 1283int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
1284int hammer_ref_buffer(hammer_buffer_t buffer);
1285void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
1286
250aec18
MD
1287void hammer_rel_volume(hammer_volume_t volume, int locked);
1288void hammer_rel_buffer(hammer_buffer_t buffer, int locked);
8cd0a023 1289
513ca7d7
MD
1290int hammer_vfs_export(struct mount *mp, int op,
1291 const struct export_args *export);
82010f9f
MD
1292hammer_node_t hammer_get_node(hammer_transaction_t trans,
1293 hammer_off_t node_offset, int isnew, int *errorp);
740d8317 1294void hammer_ref_node(hammer_node_t node);
4c286c36 1295hammer_node_t hammer_ref_node_safe(hammer_transaction_t trans,
bcac4bbb 1296 hammer_node_cache_t cache, int *errorp);
8cd0a023 1297void hammer_rel_node(hammer_node_t node);
36f82b23
MD
1298void hammer_delete_node(hammer_transaction_t trans,
1299 hammer_node_t node);
bcac4bbb
MD
1300void hammer_cache_node(hammer_node_cache_t cache,
1301 hammer_node_t node);
1302void hammer_uncache_node(hammer_node_cache_t cache);
250aec18 1303void hammer_flush_node(hammer_node_t node, int locked);
8cd0a023 1304
427e5fc6
MD
1305void hammer_dup_buffer(struct hammer_buffer **bufferp,
1306 struct hammer_buffer *buffer);
df2ccbac
MD
1307hammer_node_t hammer_alloc_btree(hammer_transaction_t trans,
1308 hammer_off_t hint, int *errorp);
36f82b23 1309void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 1310 u_int16_t rec_type, hammer_off_t *data_offsetp,
df2ccbac
MD
1311 struct hammer_buffer **data_bufferp,
1312 hammer_off_t hint, int *errorp);
bf686dbe 1313
02428fb6 1314int hammer_generate_undo(hammer_transaction_t trans,
059819e3 1315 hammer_off_t zone1_offset, void *base, int len);
6048b411 1316int hammer_generate_redo(hammer_transaction_t trans, hammer_inode_t ip,
47f363f1
MD
1317 hammer_off_t file_offset, u_int32_t flags,
1318 void *base, int len);
1319void hammer_generate_redo_sync(hammer_transaction_t trans);
73896937
MD
1320void hammer_redo_fifo_start_flush(hammer_inode_t ip);
1321void hammer_redo_fifo_end_flush(hammer_inode_t ip);
1322
47f363f1 1323void hammer_format_undo(void *base, u_int32_t seqno);
02428fb6 1324int hammer_upgrade_undo_4(hammer_transaction_t trans);
427e5fc6 1325
66325755 1326void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 1327void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 1328
36f82b23
MD
1329hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
1330 hammer_off_t owner, int *errorp);
1331void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 1332 hammer_off_t owner, int *errorp);
0f65be10 1333int _hammer_checkspace(hammer_mount_t hmp, int slop, int64_t *resp);
36f82b23 1334hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
df2ccbac 1335 int bytes, hammer_off_t hint, int *errorp);
0832c9bb
MD
1336hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
1337 int bytes, hammer_off_t *zone_offp, int *errorp);
507df98a
ID
1338hammer_reserve_t hammer_blockmap_reserve_dedup(hammer_mount_t hmp, int zone,
1339 int bytes, hammer_off_t zone_offset, int *errorp);
0832c9bb
MD
1340void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
1341 hammer_reserve_t resv);
cb51be26 1342void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1343void hammer_blockmap_free(hammer_transaction_t trans,
1344 hammer_off_t bmap_off, int bytes);
bb29b5d8
MD
1345int hammer_blockmap_dedup(hammer_transaction_t trans,
1346 hammer_off_t bmap_off, int bytes);
cdb6e4e6 1347int hammer_blockmap_finalize(hammer_transaction_t trans,
5e435c92 1348 hammer_reserve_t resv,
4a2796f3 1349 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1350int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1351 int *curp, int *errorp);
40043e7f
MD
1352hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1353 int *errorp);
bf686dbe
MD
1354hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1355 int *errorp);
06ad81ff
MD
1356int64_t hammer_undo_used(hammer_transaction_t trans);
1357int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686 1358int64_t hammer_undo_max(hammer_mount_t hmp);
710733a6 1359int hammer_undo_reclaim(hammer_io_t io);
1f07f686 1360
8cd0a023
MD
1361void hammer_start_transaction(struct hammer_transaction *trans,
1362 struct hammer_mount *hmp);
36f82b23
MD
1363void hammer_simple_transaction(struct hammer_transaction *trans,
1364 struct hammer_mount *hmp);
b84de5af
MD
1365void hammer_start_transaction_fls(struct hammer_transaction *trans,
1366 struct hammer_mount *hmp);
1367void hammer_done_transaction(struct hammer_transaction *trans);
83f2a3aa 1368hammer_tid_t hammer_alloc_tid(hammer_mount_t hmp, int count);
66325755 1369
e98f1b96 1370void hammer_modify_inode(hammer_transaction_t trans, hammer_inode_t ip, int flags);
f90dde4c 1371void hammer_flush_inode(hammer_inode_t ip, int flags);
cdb6e4e6 1372void hammer_flush_inode_done(hammer_inode_t ip, int error);
b84de5af
MD
1373void hammer_wait_inode(hammer_inode_t ip);
1374
8cd0a023
MD
1375int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1376 struct ucred *cred, struct hammer_inode *dip,
5a64efa1 1377 const char *name, int namelen,
ea434b6f
MD
1378 hammer_pseudofs_inmem_t pfsm,
1379 struct hammer_inode **ipp);
a89aec1b 1380void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1381int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1382int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
73896937 1383int hammer_redo_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
cdb6e4e6 1384int hammer_destroy_inode_callback(hammer_inode_t ip, void *data __unused);
51c35492 1385
02325004 1386int hammer_sync_inode(hammer_transaction_t trans, hammer_inode_t ip);
7b6ccb11 1387void hammer_test_inode(hammer_inode_t dip);
e8599db1 1388void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
70125e78 1389int hammer_update_atime_quick(hammer_inode_t ip);
8cd0a023 1390
a89aec1b 1391int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1392 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1393 hammer_inode_t nip);
a89aec1b 1394int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1395 hammer_cursor_t cursor, hammer_inode_t dip,
1396 hammer_inode_t ip);
6362a262 1397void hammer_ip_replace_bulk(hammer_mount_t hmp, hammer_record_t record);
47637bff 1398hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1399 void *data, int bytes, int *errorp);
47637bff 1400int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1401int hammer_ip_add_record(struct hammer_transaction *trans,
1402 hammer_record_t record);
4e17f465 1403int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1404 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1405int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1406 int *countp);
1407int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1408 int64_t offset, void *data, int bytes);
36f82b23 1409int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1410int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1411hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1412 u_int32_t localization, int *errorp);
1413int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1414 hammer_pseudofs_inmem_t pfsm);
1415int hammer_save_pseudofs(hammer_transaction_t trans,
1416 hammer_pseudofs_inmem_t pfsm);
842e7a70 1417int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1418void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1419int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1420 struct ucred *cred);
1421
748efb59 1422void hammer_io_init(hammer_io_t io, hammer_volume_t volume,
10a5d1ba 1423 enum hammer_io_type type);
b7de8aa5 1424int hammer_io_read(struct vnode *devvp, struct hammer_io *io, int limit);
0e8bd897 1425void hammer_io_advance(struct hammer_io *io);
66325755 1426int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
362ec2dc 1427int hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1428struct buf *hammer_io_release(struct hammer_io *io, int flush);
710733a6 1429void hammer_io_flush(struct hammer_io *io, int reclaim);
1b0ab2c3 1430void hammer_io_wait(struct hammer_io *io);
055f5ff8 1431void hammer_io_waitdep(struct hammer_io *io);
eddadaee 1432void hammer_io_wait_all(hammer_mount_t hmp, const char *ident, int doflush);
1b0ab2c3
MD
1433int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio,
1434 hammer_btree_leaf_elm_t leaf);
9a98f3cc
MD
1435int hammer_io_indirect_read(hammer_mount_t hmp, struct bio *bio,
1436 hammer_btree_leaf_elm_t leaf);
6362a262
MD
1437int hammer_io_direct_write(hammer_mount_t hmp, struct bio *bio,
1438 hammer_record_t record);
1b0ab2c3 1439void hammer_io_direct_wait(hammer_record_t record);
43c665ae 1440void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1441void hammer_io_write_interlock(hammer_io_t io);
1442void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1443void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1444void hammer_io_clear_modlist(struct hammer_io *io);
748efb59 1445void hammer_io_flush_sync(hammer_mount_t hmp);
2faf0737 1446void hammer_io_clear_error(struct hammer_io *io);
77912481 1447void hammer_io_clear_error_noassert(struct hammer_io *io);
b8a41159 1448void hammer_io_notmeta(hammer_buffer_t buffer);
ba298df1 1449void hammer_io_limit_backlog(hammer_mount_t hmp);
748efb59 1450
36f82b23
MD
1451void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1452 void *base, int len);
1453void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1454 void *base, int len);
10a5d1ba
MD
1455void hammer_modify_volume_done(hammer_volume_t volume);
1456void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1457
36f82b23
MD
1458int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1459 struct hammer_ioc_reblock *reblock);
1775b6a0
MD
1460int hammer_ioc_rebalance(hammer_transaction_t trans, hammer_inode_t ip,
1461 struct hammer_ioc_rebalance *rebal);
11ad5ade
MD
1462int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1463 struct hammer_ioc_prune *prune);
c82af904
MD
1464int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1465 struct hammer_ioc_mirror_rw *mirror);
1466int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1467 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1468int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1469 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1470int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1471 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1472int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1473 struct hammer_ioc_pseudofs_rw *pfs);
1474int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1475 struct hammer_ioc_pseudofs_rw *pfs);
1476int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1477 struct hammer_ioc_pseudofs_rw *pfs);
4889cbd4
MD
1478int hammer_ioc_wait_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1479 struct hammer_ioc_pseudofs_rw *pfs);
d121f61c 1480int hammer_ioc_volume_add(hammer_transaction_t trans, hammer_inode_t ip,
865c9609
MN
1481 struct hammer_ioc_volume *ioc);
1482int hammer_ioc_volume_del(hammer_transaction_t trans, hammer_inode_t ip,
1483 struct hammer_ioc_volume *ioc);
e914c91d
SK
1484int hammer_ioc_volume_list(hammer_transaction_t trans, hammer_inode_t ip,
1485 struct hammer_ioc_volume_list *ioc);
bb29b5d8
MD
1486int hammer_ioc_dedup(hammer_transaction_t trans, hammer_inode_t ip,
1487 struct hammer_ioc_dedup *dedup);
bf686dbe 1488
855942b6 1489int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1490
059819e3
MD
1491void hammer_flusher_create(hammer_mount_t hmp);
1492void hammer_flusher_destroy(hammer_mount_t hmp);
1493void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1494int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1495int hammer_flusher_async_one(hammer_mount_t hmp);
e2a02b72 1496int hammer_flusher_running(hammer_mount_t hmp);
93291532 1497void hammer_flusher_wait(hammer_mount_t hmp, int seq);
82010f9f 1498void hammer_flusher_wait_next(hammer_mount_t hmp);
06ad81ff 1499int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1500int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1501int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1502void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1503void hammer_flusher_finalize(hammer_transaction_t trans, int final);
1b0ab2c3 1504int hammer_flusher_haswork(hammer_mount_t hmp);
6048b411 1505void hammer_flusher_flush_undos(hammer_mount_t hmp, int already_flushed);
059819e3 1506
02428fb6
MD
1507int hammer_recover_stage1(hammer_mount_t hmp, hammer_volume_t rootvol);
1508int hammer_recover_stage2(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1509void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1510 hammer_volume_t root_volume, int final);
f90dde4c 1511
19619882
MD
1512void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1513void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1514void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1515
1516int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1517int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1518int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1519int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
c6b20585 1520void hkprintf(const char *ctl, ...) __printflike(1, 2);
a56cb012
MD
1521udev_t hammer_fsid_to_udev(uuid_t *uuid);
1522
19619882 1523
4a2796f3 1524int hammer_blocksize(int64_t file_offset);
6362a262 1525int hammer_blockoff(int64_t file_offset);
4a2796f3
MD
1526int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1527
0f65be10
MD
1528/*
1529 * Shortcut for _hammer_checkspace(), used all over the code.
1530 */
1531static __inline int
1532hammer_checkspace(hammer_mount_t hmp, int slop)
1533{
1534 return(_hammer_checkspace(hmp, slop, NULL));
1535}
1536
055f5ff8 1537#endif
427e5fc6 1538
8cd0a023 1539static __inline void
af209b0f
MD
1540hammer_wait_mem_record(hammer_record_t record)
1541{
1542 hammer_wait_mem_record_ident(record, "hmmwai");
1543}
1544
1545static __inline void
1546hammer_lock_ex(struct hammer_lock *lock)
1547{
1548 hammer_lock_ex_ident(lock, "hmrlck");
1549}
1550
bcac4bbb
MD
1551/*
1552 * Indicate that a B-Tree node is being modified.
1553 */
af209b0f 1554static __inline void
36f82b23 1555hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1556{
4c286c36 1557 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23 1558 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1559}
1560
1561static __inline void
36f82b23 1562hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1563{
4c286c36 1564 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23
MD
1565 hammer_modify_buffer(trans, node->buffer,
1566 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1567}
1568
bf686dbe 1569static __inline void
36f82b23
MD
1570hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1571 void *base, int len)
bf686dbe 1572{
19619882
MD
1573 hammer_crc_t *crcptr;
1574
bf686dbe
MD
1575 KKASSERT((char *)base >= (char *)node->ondisk &&
1576 (char *)base + len <=
1577 (char *)node->ondisk + sizeof(*node->ondisk));
4c286c36 1578 KKASSERT((node->flags & HAMMER_NODE_CRCBAD) == 0);
36f82b23 1579 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1580 crcptr = &node->ondisk->crc;
1581 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1582 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1583}
1584
bcac4bbb
MD
1585/*
1586 * Indicate that the specified modifications have been completed.
1587 *
1588 * Do not try to generate the crc here, it's very expensive to do and a
1589 * sequence of insertions or deletions can result in many calls to this
1590 * function on the same node.
1591 */
bf686dbe 1592static __inline void
10a5d1ba
MD
1593hammer_modify_node_done(hammer_node_t node)
1594{
bcac4bbb
MD
1595 node->flags |= HAMMER_NODE_CRCGOOD;
1596 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1597 node->flags |= HAMMER_NODE_NEEDSCRC;
1598 node->buffer->io.gencrc = 1;
1599 hammer_ref_node(node);
1600 }
10a5d1ba
MD
1601 hammer_modify_buffer_done(node->buffer);
1602}
1603
e8599db1
MD
1604#define hammer_modify_volume_field(trans, vol, field) \
1605 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1606 sizeof((vol)->ondisk->field))
1607
c9b9e29d
MD
1608#define hammer_modify_node_field(trans, node, field) \
1609 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1610 sizeof((node)->ondisk->field))
1611
beec5dc4
MD
1612/*
1613 * The HAMMER_INODE_CAP_DIR_LOCAL_INO capability is set on newly
1614 * created directories for HAMMER version 2 or greater and causes
1615 * directory entries to be placed the inode localization zone in
1616 * the B-Tree instead of the misc zone.
1617 *
1618 * This greatly improves localization between directory entries and
1619 * inodes
1620 */
1621static __inline u_int32_t
1622hammer_dir_localization(hammer_inode_t dip)
1623{
1624 if (dip->ino_data.cap_flags & HAMMER_INODE_CAP_DIR_LOCAL_INO)
1625 return(HAMMER_LOCALIZE_INODE);
1626 else
1627 return(HAMMER_LOCALIZE_MISC);
1628}