Increate receive max coalesced bds to 80 to prevent system from being
[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 *
15e75dab 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.115 2008/07/13 01:12:41 dillon Exp $
8750964d
MD
35 */
36/*
37 * This header file contains structures used internally by the HAMMERFS
c60bb2c5 38 * implementation. See hammer_disk.h for on-disk structures.
8750964d
MD
39 */
40
427e5fc6
MD
41#include <sys/param.h>
42#include <sys/types.h>
43#include <sys/kernel.h>
42c7d26b 44#include <sys/conf.h>
427e5fc6 45#include <sys/systm.h>
8750964d
MD
46#include <sys/tree.h>
47#include <sys/malloc.h>
427e5fc6 48#include <sys/mount.h>
513ca7d7 49#include <sys/mountctl.h>
427e5fc6 50#include <sys/vnode.h>
42c7d26b 51#include <sys/proc.h>
e63644f0 52#include <sys/stat.h>
427e5fc6 53#include <sys/globaldata.h>
66325755
MD
54#include <sys/lockf.h>
55#include <sys/buf.h>
8cd0a023 56#include <sys/queue.h>
bcac4bbb 57#include <sys/ktr.h>
66325755
MD
58#include <sys/globaldata.h>
59
60#include <sys/buf2.h>
855942b6 61#include <sys/signal2.h>
427e5fc6 62#include "hammer_disk.h"
8750964d 63#include "hammer_mount.h"
7dc57964 64#include "hammer_ioctl.h"
8750964d
MD
65
66#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
67
68MALLOC_DECLARE(M_HAMMER);
69
bcac4bbb
MD
70/*
71 * Kernel trace
72 */
73#if !defined(KTR_HAMMER)
74#define KTR_HAMMER KTR_ALL
75#endif
76KTR_INFO_MASTER_EXTERN(hammer);
77
78/*
79 * Misc structures
80 */
66325755
MD
81struct hammer_mount;
82
8750964d
MD
83/*
84 * Key structure used for custom RB tree inode lookups. This prototypes
85 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
86 */
87typedef struct hammer_inode_info {
513ca7d7 88 int64_t obj_id; /* (key) object identifier */
8750964d 89 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
ddfdf542 90 u_int32_t obj_localization; /* (key) pseudo-fs */
43c665ae
MD
91 union {
92 struct hammer_btree_leaf_elm *leaf;
93 } u;
8750964d
MD
94} *hammer_inode_info_t;
95
b84de5af
MD
96typedef enum hammer_transaction_type {
97 HAMMER_TRANS_RO,
98 HAMMER_TRANS_STD,
99 HAMMER_TRANS_FLS
100} hammer_transaction_type_t;
101
66325755
MD
102/*
103 * HAMMER Transaction tracking
104 */
105struct hammer_transaction {
b84de5af 106 hammer_transaction_type_t type;
66325755
MD
107 struct hammer_mount *hmp;
108 hammer_tid_t tid;
ddfdf542 109 u_int64_t time;
dd94f1b1 110 u_int32_t time32;
2f85fa4d 111 int sync_lock_refs;
a89aec1b 112 struct hammer_volume *rootvol;
66325755
MD
113};
114
8cd0a023
MD
115typedef struct hammer_transaction *hammer_transaction_t;
116
66325755
MD
117/*
118 * HAMMER locks
119 */
427e5fc6 120struct hammer_lock {
0b075555 121 int refs; /* active references delay writes */
0b075555 122 int lockcount; /* lock count for exclusive/shared access */
427e5fc6 123 int wanted;
4a2796f3 124 int exwanted; /* number of threads waiting for ex lock */
427e5fc6 125 struct thread *locktd;
8750964d
MD
126};
127
427e5fc6
MD
128static __inline int
129hammer_islocked(struct hammer_lock *lock)
130{
8cd0a023 131 return(lock->lockcount != 0);
427e5fc6
MD
132}
133
0b075555
MD
134static __inline int
135hammer_isactive(struct hammer_lock *lock)
136{
137 return(lock->refs != 0);
138}
139
427e5fc6
MD
140static __inline int
141hammer_islastref(struct hammer_lock *lock)
142{
143 return(lock->refs == 1);
144}
c60bb2c5 145
6a37e7e4 146/*
7aa3b8a6 147 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
148 */
149static __inline int
7aa3b8a6 150hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 151{
7aa3b8a6
MD
152 if (lock->lockcount > 0 && lock->locktd == td)
153 return(1);
154 return(0);
6a37e7e4
MD
155}
156
ec4e8497 157/*
1f07f686 158 * Flush state, used by various structures
ec4e8497 159 */
1f07f686
MD
160typedef enum hammer_inode_state {
161 HAMMER_FST_IDLE,
162 HAMMER_FST_SETUP,
163 HAMMER_FST_FLUSH
164} hammer_inode_state_t;
ec4e8497 165
1f07f686 166TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 167
5fa5c92f
MD
168/*
169 * Pseudo-filesystem extended data tracking
170 */
171struct hammer_pfs_rb_tree;
172struct hammer_pseudofs_inmem;
173RB_HEAD(hammer_pfs_rb_tree, hammer_pseudofs_inmem);
174RB_PROTOTYPE2(hammer_pfs_rb_tree, hammer_pseudofs_inmem, rb_node,
175 hammer_pfs_rb_compare, u_int32_t);
176
177struct hammer_pseudofs_inmem {
178 RB_ENTRY(hammer_pseudofs_inmem) rb_node;
179 struct hammer_lock lock;
180 u_int32_t localization;
ea434b6f 181 hammer_tid_t create_tid;
842e7a70 182 int flags;
a56cb012 183 udev_t fsid_udev;
5fa5c92f
MD
184 struct hammer_pseudofs_data pfsd;
185};
186
187typedef struct hammer_pseudofs_inmem *hammer_pseudofs_inmem_t;
188
842e7a70
MD
189#define HAMMER_PFSM_DELETED 0x0001
190
0729c8c8
MD
191/*
192 * Cache object ids. A fixed number of objid cache structures are
193 * created to reserve object id's for newly created files in multiples
194 * of 100,000, localized to a particular directory, and recycled as
195 * needed. This allows parallel create operations in different
196 * directories to retain fairly localized object ids which in turn
197 * improves reblocking performance and layout.
198 */
bf3b416b 199#define OBJID_CACHE_SIZE 1024
0729c8c8
MD
200#define OBJID_CACHE_BULK 100000
201
202typedef struct hammer_objid_cache {
203 TAILQ_ENTRY(hammer_objid_cache) entry;
204 struct hammer_inode *dip;
205 hammer_tid_t next_tid;
206 int count;
207} *hammer_objid_cache_t;
208
bcac4bbb
MD
209/*
210 * Associate an inode with a B-Tree node to cache search start positions
211 */
212typedef struct hammer_node_cache {
7a61b85d 213 TAILQ_ENTRY(hammer_node_cache) entry;
bcac4bbb
MD
214 struct hammer_node *node;
215 struct hammer_inode *ip;
216} *hammer_node_cache_t;
217
218TAILQ_HEAD(hammer_node_cache_list, hammer_node_cache);
219
7a61b85d
MD
220/*
221 * Structure used to organize flush groups. Flush groups must be
222 * organized into chunks in order to avoid blowing out the UNDO FIFO.
223 * Without this a 'sync' could end up flushing 50,000 inodes in a single
224 * transaction.
225 */
226struct hammer_flush_group {
227 TAILQ_ENTRY(hammer_flush_group) flush_entry;
228 TAILQ_HEAD(, hammer_inode) flush_list;
229 int inode_count; /* inode load */
230 int total_count; /* record load */
231 int running; /* group is running */
232 int closed;
233 int refs;
234};
235
236typedef struct hammer_flush_group *hammer_flush_group_t;
237
238TAILQ_HEAD(hammer_flush_group_list, hammer_flush_group);
239
8750964d 240/*
8cd0a023
MD
241 * Structure used to represent an inode in-memory.
242 *
243 * The record and data associated with an inode may be out of sync with
244 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
245 * clear).
246 *
247 * An inode may also hold a cache of unsynchronized records, used for
248 * database and directories only. Unsynchronized regular file data is
249 * stored in the buffer cache.
250 *
251 * NOTE: A file which is created and destroyed within the initial
252 * synchronization period can wind up not doing any disk I/O at all.
253 *
254 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
255 */
256struct hammer_ino_rb_tree;
257struct hammer_inode;
258RB_HEAD(hammer_ino_rb_tree, hammer_inode);
259RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
260 hammer_ino_rb_compare, hammer_inode_info_t);
261
262struct hammer_rec_rb_tree;
263struct hammer_record;
264RB_HEAD(hammer_rec_rb_tree, hammer_record);
265RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
0832c9bb 266 hammer_rec_rb_compare, hammer_btree_leaf_elm_t);
8cd0a023
MD
267
268TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
269
270struct hammer_inode {
1f07f686
MD
271 RB_ENTRY(hammer_inode) rb_node;
272 hammer_inode_state_t flush_state;
7a61b85d 273 hammer_flush_group_t flush_group;
059819e3 274 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 275 struct hammer_record_list target_list; /* target of dependant recs */
adf01747 276 int64_t obj_id; /* (key) object identifier */
b84de5af 277 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
ddfdf542 278 u_int32_t obj_localization; /* (key) pseudo-fs */
b84de5af 279 struct hammer_mount *hmp;
0729c8c8 280 hammer_objid_cache_t objid_cache;
b84de5af
MD
281 int flags;
282 int error; /* flush error */
283 int cursor_ip_refs; /* sanity */
47637bff 284 int rsv_recs;
b84de5af 285 struct vnode *vp;
5fa5c92f 286 hammer_pseudofs_inmem_t pfsm;
b84de5af
MD
287 struct lockf advlock;
288 struct hammer_lock lock; /* sync copy interlock */
b84de5af 289 off_t trunc_off;
11ad5ade 290 struct hammer_btree_leaf_elm ino_leaf; /* in-memory cache */
b84de5af
MD
291 struct hammer_inode_data ino_data; /* in-memory cache */
292 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
bcac4bbb 293 struct hammer_node_cache cache[2]; /* search initiate cache */
b84de5af
MD
294
295 /*
296 * When a demark is created to synchronize an inode to
297 * disk, certain fields are copied so the front-end VOPs
298 * can continue to run in parallel with the synchronization
299 * occuring in the background.
300 */
301 int sync_flags; /* to-sync flags cache */
302 off_t sync_trunc_off; /* to-sync truncation */
a9d52b76 303 off_t save_trunc_off; /* write optimization */
11ad5ade 304 struct hammer_btree_leaf_elm sync_ino_leaf; /* to-sync cache */
b84de5af 305 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
306};
307
8cd0a023
MD
308typedef struct hammer_inode *hammer_inode_t;
309
66325755
MD
310#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
311
8cd0a023 312#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
ddfdf542 313 /* (not including atime/mtime) */
e63644f0 314#define HAMMER_INODE_RSV_INODES 0x0002 /* hmp->rsv_inodes bumped */
ea434b6f 315#define HAMMER_INODE_UNUSED0004 0x0004
1f07f686 316#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 317#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 318#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 319#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 320#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 321#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 322#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 323#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 324#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
b84de5af 325#define HAMMER_INODE_REFLUSH 0x2000 /* pipelined flush during flush */
9f5097dc 326#define HAMMER_INODE_RECLAIM 0x4000 /* trying to reclaim */
059819e3 327#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 328
b84de5af 329#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 330#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 331#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
ddfdf542 332#define HAMMER_INODE_ATIME 0x00100000 /* in-memory atime modified */
06ad81ff
MD
333#define HAMMER_INODE_MTIME 0x00200000 /* in-memory mtime modified */
334#define HAMMER_INODE_WOULDBLOCK 0x00400000 /* re-issue to new flush group */
b84de5af 335
11ad5ade
MD
336#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY| \
337 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
ddfdf542
MD
338 HAMMER_INODE_ATIME|HAMMER_INODE_MTIME| \
339 HAMMER_INODE_TRUNCATED|HAMMER_INODE_DELETING)
1f07f686
MD
340
341#define HAMMER_INODE_MODMASK_NOXDIRTY \
342 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 343
cb51be26 344#define HAMMER_FLUSH_GROUP_SIZE 64
66325755 345
f90dde4c 346#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 347#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 348
7bc5b8c2
MD
349/*
350 * Used by the inode reclaim code to pipeline reclaims and avoid
351 * blowing out kernel memory or letting the flusher get too far
352 * behind.
353 */
354struct hammer_reclaim {
355 TAILQ_ENTRY(hammer_reclaim) entry;
356 int okydoky;
357};
358
4a2796f3
MD
359#define HAMMER_RECLAIM_FLUSH 2000
360#define HAMMER_RECLAIM_WAIT 4000
9f5097dc 361
8750964d 362/*
1f07f686
MD
363 * Structure used to represent an unsynchronized record in-memory. These
364 * records typically represent directory entries. Only non-historical
365 * records are kept in-memory.
366 *
367 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
368 * on disk then neither are any records and the in-memory record tree
369 * represents the entire contents of the inode. If the inode is on disk
370 * then the on-disk B-Tree is scanned in parallel with the in-memory
371 * RB-Tree to synthesize the current state of the file.
372 *
1f07f686
MD
373 * Records are also used to enforce the ordering of directory create/delete
374 * operations. A new inode will not be flushed to disk unless its related
375 * directory entry is also being flushed at the same time. A directory entry
376 * will not be removed unless its related inode is also being removed at the
377 * same time.
8750964d 378 */
1f07f686 379typedef enum hammer_record_type {
e8599db1 380 HAMMER_MEM_RECORD_GENERAL, /* misc record */
930bf163 381 HAMMER_MEM_RECORD_INODE, /* inode record */
1f07f686 382 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
47637bff
MD
383 HAMMER_MEM_RECORD_DEL, /* negative delete-on-disk record */
384 HAMMER_MEM_RECORD_DATA /* bulk-data record w/on-disk ref */
1f07f686
MD
385} hammer_record_type_t;
386
8cd0a023
MD
387struct hammer_record {
388 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
389 TAILQ_ENTRY(hammer_record) target_entry;
390 hammer_inode_state_t flush_state;
7a61b85d 391 hammer_flush_group_t flush_group;
1f07f686 392 hammer_record_type_t type;
a89aec1b 393 struct hammer_lock lock;
0832c9bb 394 struct hammer_reserve *resv;
8cd0a023 395 struct hammer_inode *ip;
1f07f686 396 struct hammer_inode *target_ip;
11ad5ade 397 struct hammer_btree_leaf_elm leaf;
8cd0a023 398 union hammer_data_ondisk *data;
8cd0a023
MD
399 int flags;
400};
8750964d 401
8cd0a023
MD
402typedef struct hammer_record *hammer_record_t;
403
d36ec43b
MD
404/*
405 * Record flags. Note that FE can only be set by the frontend if the
406 * record has not been interlocked by the backend w/ BE.
407 */
8cd0a023
MD
408#define HAMMER_RECF_ALLOCDATA 0x0001
409#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 410#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b 411#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
11ad5ade 412#define HAMMER_RECF_UNUSED0010 0x0010
d36ec43b 413#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
0832c9bb 414#define HAMMER_RECF_WANTED 0x0040 /* wanted by the frontend */
1f07f686 415#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
8cd0a023 416
602c6cb8
MD
417/*
418 * hammer_delete_at_cursor() flags
419 */
420#define HAMMER_DELETE_ADJUST 0x0001
421#define HAMMER_DELETE_DESTROY 0x0002
422
8cd0a023 423/*
47197d71 424 * In-memory structures representing on-disk structures.
8cd0a023 425 */
8750964d 426struct hammer_volume;
427e5fc6 427struct hammer_buffer;
8cd0a023 428struct hammer_node;
e8599db1 429struct hammer_undo;
0832c9bb
MD
430struct hammer_reserve;
431
8750964d 432RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 433RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 434RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 435RB_HEAD(hammer_und_rb_tree, hammer_undo);
0832c9bb 436RB_HEAD(hammer_res_rb_tree, hammer_reserve);
8750964d
MD
437
438RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
439 hammer_vol_rb_compare, int32_t);
427e5fc6 440RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 441 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 442RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 443 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
444RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
445 hammer_und_rb_compare, hammer_off_t);
0832c9bb
MD
446RB_PROTOTYPE2(hammer_res_rb_tree, hammer_reserve, rb_node,
447 hammer_res_rb_compare, hammer_off_t);
8750964d 448
66325755
MD
449/*
450 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
451 *
452 * VOLUME - hammer_volume containing meta-data
453 * META_BUFFER - hammer_buffer containing meta-data
454 * DATA_BUFFER - hammer_buffer containing pure-data
455 *
456 * Dirty volume headers and dirty meta-data buffers are locked until the
457 * flusher can sequence them out. Dirty pure-data buffers can be written.
458 * Clean buffers can be passively released.
66325755 459 */
10a5d1ba
MD
460typedef enum hammer_io_type {
461 HAMMER_STRUCTURE_VOLUME,
462 HAMMER_STRUCTURE_META_BUFFER,
463 HAMMER_STRUCTURE_UNDO_BUFFER,
464 HAMMER_STRUCTURE_DATA_BUFFER
465} hammer_io_type_t;
66325755
MD
466
467union hammer_io_structure;
055f5ff8 468struct hammer_io;
66325755
MD
469
470struct worklist {
471 LIST_ENTRY(worklist) node;
472};
473
10a5d1ba
MD
474TAILQ_HEAD(hammer_io_list, hammer_io);
475typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 476
66325755 477struct hammer_io {
10a5d1ba
MD
478 struct worklist worklist;
479 struct hammer_lock lock;
480 enum hammer_io_type type;
481 struct hammer_mount *hmp;
482 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
483 hammer_io_list_t mod_list;
484 struct buf *bp;
4a2796f3
MD
485 int64_t offset; /* zone-2 offset */
486 int bytes; /* buffer cache buffer size */
10a5d1ba
MD
487 int loading; /* loading/unloading interlock */
488 int modify_refs;
489
66325755
MD
490 u_int modified : 1; /* bp's data was modified */
491 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
492 u_int running : 1; /* bp write IO in progress */
493 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 494 u_int validated : 1; /* ondisk has been validated */
b58c6388 495 u_int waitdep : 1; /* flush waits for dependancies */
51c35492 496 u_int recovered : 1; /* has recovery ref */
9f5097dc 497 u_int waitmod : 1; /* waiting for modify_refs */
cebe9493 498 u_int reclaim : 1; /* reclaim requested */
bcac4bbb 499 u_int gencrc : 1; /* crc needs to be generated */
66325755
MD
500};
501
8cd0a023
MD
502typedef struct hammer_io *hammer_io_t;
503
af209b0f
MD
504#define HAMMER_CLUSTER_SIZE (64 * 1024)
505#if HAMMER_CLUSTER_SIZE > MAXBSIZE
506#undef HAMMER_CLUSTER_SIZE
507#define HAMMER_CLUSTER_SIZE MAXBSIZE
508#endif
509#define HAMMER_CLUSTER_BUFS (HAMMER_CLUSTER_SIZE / HAMMER_BUFSIZE)
510
66325755 511/*
8cd0a023 512 * In-memory volume representing on-disk buffer
66325755 513 */
8750964d 514struct hammer_volume {
66325755 515 struct hammer_io io;
8750964d 516 RB_ENTRY(hammer_volume) rb_node;
8750964d
MD
517 struct hammer_volume_ondisk *ondisk;
518 int32_t vol_no;
fbc6e32a 519 int64_t nblocks; /* note: special calculation for statfs */
47197d71 520 int64_t buffer_base; /* base offset of buffer 0 */
2f85fa4d
MD
521 hammer_off_t maxbuf_off; /* Maximum buffer offset (zone-2) */
522 hammer_off_t maxraw_off; /* Maximum raw offset for device */
c60bb2c5 523 char *vol_name;
8750964d 524 struct vnode *devvp;
427e5fc6 525 int vol_flags;
8750964d
MD
526};
527
8cd0a023
MD
528typedef struct hammer_volume *hammer_volume_t;
529
66325755 530/*
8cd0a023
MD
531 * In-memory buffer (other then volume, super-cluster, or cluster),
532 * representing an on-disk buffer.
66325755 533 */
427e5fc6 534struct hammer_buffer {
66325755 535 struct hammer_io io;
427e5fc6 536 RB_ENTRY(hammer_buffer) rb_node;
47197d71 537 void *ondisk;
427e5fc6 538 struct hammer_volume *volume;
34d829f7 539 hammer_off_t zoneX_offset;
0832c9bb 540 hammer_off_t zone2_offset;
cebe9493 541 struct hammer_reserve *resv;
8cd0a023 542 struct hammer_node_list clist;
66325755
MD
543};
544
8cd0a023
MD
545typedef struct hammer_buffer *hammer_buffer_t;
546
547/*
548 * In-memory B-Tree node, representing an on-disk B-Tree node.
549 *
550 * This is a hang-on structure which is backed by a hammer_buffer,
551 * indexed by a hammer_cluster, and used for fine-grained locking of
552 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
553 * can contain multiple nodes representing wildly disassociated portions
554 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
555 *
556 * This structure uses a cluster-relative index to reduce the number
557 * of layers required to access it, and also because all on-disk B-Tree
558 * references are cluster-relative offsets.
559 */
560struct hammer_node {
561 struct hammer_lock lock; /* node-by-node lock */
562 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
563 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 564 hammer_off_t node_offset; /* full offset spec */
40043e7f 565 struct hammer_mount *hmp;
8cd0a023
MD
566 struct hammer_buffer *buffer; /* backing buffer */
567 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
b3bad96f 568 TAILQ_HEAD(, hammer_cursor) cursor_list; /* deadlock recovery */
bcac4bbb 569 struct hammer_node_cache_list cache_list; /* passive caches */
b3deaf57 570 int flags;
740d8317 571 int loading; /* load interlock */
8cd0a023
MD
572};
573
b3deaf57
MD
574#define HAMMER_NODE_DELETED 0x0001
575#define HAMMER_NODE_FLUSH 0x0002
bcac4bbb
MD
576#define HAMMER_NODE_CRCGOOD 0x0004
577#define HAMMER_NODE_NEEDSCRC 0x0008
c82af904 578#define HAMMER_NODE_NEEDSMIRROR 0x0010
b3deaf57 579
8cd0a023
MD
580typedef struct hammer_node *hammer_node_t;
581
b33e2cc0
MD
582/*
583 * List of locked nodes.
584 */
585struct hammer_node_locklist {
586 struct hammer_node_locklist *next;
587 hammer_node_t node;
588};
589
590typedef struct hammer_node_locklist *hammer_node_locklist_t;
591
592
8cd0a023
MD
593/*
594 * Common I/O management structure - embedded in in-memory structures
595 * which are backed by filesystem buffers.
596 */
66325755
MD
597union hammer_io_structure {
598 struct hammer_io io;
599 struct hammer_volume volume;
66325755 600 struct hammer_buffer buffer;
8750964d
MD
601};
602
055f5ff8
MD
603typedef union hammer_io_structure *hammer_io_structure_t;
604
cebe9493
MD
605/*
606 * The reserve structure prevents the blockmap from allocating
607 * out of a reserved bigblock. Such reservations are used by
608 * the direct-write mechanism.
609 *
610 * The structure is also used to hold off on reallocations of
611 * big blocks from the freemap until flush dependancies have
612 * been dealt with.
613 */
0832c9bb
MD
614struct hammer_reserve {
615 RB_ENTRY(hammer_reserve) rb_node;
cebe9493
MD
616 TAILQ_ENTRY(hammer_reserve) delay_entry;
617 int flush_group;
cb51be26 618 int flags;
0832c9bb 619 int refs;
cb51be26 620 int zone;
df301614 621 int append_off;
cebe9493 622 hammer_off_t zone_offset;
0832c9bb
MD
623};
624
625typedef struct hammer_reserve *hammer_reserve_t;
626
cb51be26
MD
627#define HAMMER_RESF_ONDELAY 0x0001
628
8cd0a023
MD
629#include "hammer_cursor.h"
630
e8599db1 631/*
cebe9493
MD
632 * The undo structure tracks recent undos to avoid laying down duplicate
633 * undos within a flush group, saving us a significant amount of overhead.
634 *
635 * This is strictly a heuristic.
e8599db1 636 */
7a61b85d
MD
637#define HAMMER_MAX_UNDOS 1024
638#define HAMMER_MAX_FLUSHERS 4
e8599db1
MD
639
640struct hammer_undo {
641 RB_ENTRY(hammer_undo) rb_node;
642 TAILQ_ENTRY(hammer_undo) lru_entry;
643 hammer_off_t offset;
644 int bytes;
645};
646
647typedef struct hammer_undo *hammer_undo_t;
648
af209b0f 649struct hammer_flusher_info;
7a61b85d 650TAILQ_HEAD(hammer_flusher_info_list, hammer_flusher_info);
da2da375
MD
651
652struct hammer_flusher {
653 int signal; /* flusher thread sequencer */
654 int act; /* currently active flush group */
655 int done; /* set to act when complete */
656 int next; /* next flush group */
657 int group_lock; /* lock sequencing of the next flush */
658 int exiting; /* request master exit */
da2da375
MD
659 thread_t td; /* master flusher thread */
660 hammer_tid_t tid; /* last flushed transaction id */
661 int finalize_want; /* serialize finalization */
662 struct hammer_lock finalize_lock; /* serialize finalization */
663 struct hammer_transaction trans; /* shared transaction */
7a61b85d
MD
664 struct hammer_flusher_info_list run_list;
665 struct hammer_flusher_info_list ready_list;
da2da375
MD
666};
667
8750964d
MD
668/*
669 * Internal hammer mount data structure
670 */
671struct hammer_mount {
672 struct mount *mp;
27ea2398 673 /*struct vnode *rootvp;*/
8750964d
MD
674 struct hammer_ino_rb_tree rb_inos_root;
675 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 676 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 677 struct hammer_und_rb_tree rb_undo_root;
0832c9bb
MD
678 struct hammer_res_rb_tree rb_resv_root;
679 struct hammer_buf_rb_tree rb_bufs_root;
5fa5c92f 680 struct hammer_pfs_rb_tree rb_pfsm_root;
8750964d 681 struct hammer_volume *rootvol;
47197d71
MD
682 struct hammer_base_elm root_btree_beg;
683 struct hammer_base_elm root_btree_end;
9f5097dc 684 int flags;
195c19a1
MD
685 int hflags;
686 int ronly;
687 int nvolumes;
d26d0ae9 688 int volume_iterator;
5de0c0e5 689 int masterid; /* -1 or 0-15 - clustering and mirroring */
e63644f0 690 int rsv_inodes; /* reserved space due to dirty inodes */
a7e9bef1 691 int64_t rsv_databytes; /* reserved space due to record data */
e63644f0 692 int rsv_recs; /* reserved space due to dirty records */
a7e9bef1 693 int rsv_fromdelay; /* bigblocks reserved due to flush delay */
7a61b85d 694 int undo_rec_limit; /* based on size of undo area */
4a2796f3
MD
695 int last_newrecords;
696 int count_newrecords;
da2da375 697
9f5097dc
MD
698 int inode_reclaims; /* inodes pending reclaim by flusher */
699 int count_inodes; /* total number of inodes */
af209b0f 700 int count_iqueued; /* inodes queued to flusher */
da2da375
MD
701
702 struct hammer_flusher flusher;
703
855942b6 704 u_int check_interrupt;
8750964d 705 uuid_t fsid;
10a5d1ba
MD
706 struct hammer_io_list volu_list; /* dirty undo buffers */
707 struct hammer_io_list undo_list; /* dirty undo buffers */
708 struct hammer_io_list data_list; /* dirty data buffers */
0832c9bb 709 struct hammer_io_list alt_data_list; /* dirty data buffers */
10a5d1ba
MD
710 struct hammer_io_list meta_list; /* dirty meta bufs */
711 struct hammer_io_list lose_list; /* loose buffers */
f5a07a7a
MD
712 int locked_dirty_space; /* meta/volu count */
713 int io_running_space;
0729c8c8 714 int objid_cache_count;
ddfdf542
MD
715 hammer_tid_t asof; /* snapshot mount */
716 hammer_off_t next_tid;
e63644f0
MD
717 int64_t copy_stat_freebigblocks; /* number of free bigblocks */
718
66325755 719 u_int32_t namekey_iterator;
513ca7d7 720 struct netexport export;
9480ff55 721 struct hammer_lock sync_lock;
c9b9e29d 722 struct hammer_lock free_lock;
d99d6bf5
MD
723 struct hammer_lock undo_lock;
724 struct hammer_lock blkmap_lock;
0729c8c8 725 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
726 struct hammer_undo undos[HAMMER_MAX_UNDOS];
727 int undo_alloc;
728 TAILQ_HEAD(, hammer_undo) undo_lru_list;
cebe9493 729 TAILQ_HEAD(, hammer_reserve) delay_list;
7a61b85d 730 struct hammer_flush_group_list flush_group_list;
0729c8c8 731 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
7bc5b8c2 732 TAILQ_HEAD(, hammer_reclaim) reclaim_list;
8750964d
MD
733};
734
8cd0a023 735typedef struct hammer_mount *hammer_mount_t;
8750964d 736
7bc5b8c2 737#define HAMMER_MOUNT_UNUSED0001 0x0001
9f5097dc 738
fbc6e32a
MD
739struct hammer_sync_info {
740 int error;
741 int waitfor;
742};
743
8750964d
MD
744#endif
745
a7e9bef1
MD
746/*
747 * checkspace slop (8MB chunks), higher numbers are more conservative.
748 */
93291532
MD
749#define HAMMER_CHKSPC_REBLOCK 25
750#define HAMMER_CHKSPC_MIRROR 20
751#define HAMMER_CHKSPC_WRITE 20
752#define HAMMER_CHKSPC_CREATE 20
753#define HAMMER_CHKSPC_REMOVE 10
a7e9bef1 754
8750964d
MD
755#if defined(_KERNEL)
756
757extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
758extern struct vop_ops hammer_spec_vops;
759extern struct vop_ops hammer_fifo_vops;
66325755 760extern struct bio_ops hammer_bioops;
427e5fc6 761
2f85fa4d 762extern int hammer_debug_io;
d5ef456e 763extern int hammer_debug_general;
77062c8a 764extern int hammer_debug_debug;
e8599db1 765extern int hammer_debug_inode;
7d683b0f 766extern int hammer_debug_locks;
b3deaf57 767extern int hammer_debug_btree;
d113fda1 768extern int hammer_debug_tid;
b33e2cc0 769extern int hammer_debug_recover;
46fe7ae1 770extern int hammer_debug_recover_faults;
a99b9ea2 771extern int hammer_debug_cluster_enable;
7a61b85d 772extern int hammer_count_fsyncs;
b3deaf57 773extern int hammer_count_inodes;
af209b0f 774extern int hammer_count_iqueued;
9f5097dc 775extern int hammer_count_reclaiming;
b3deaf57
MD
776extern int hammer_count_records;
777extern int hammer_count_record_datas;
778extern int hammer_count_volumes;
b3deaf57
MD
779extern int hammer_count_buffers;
780extern int hammer_count_nodes;
a7e9bef1 781extern int64_t hammer_count_extra_space_used;
cb51be26
MD
782extern int64_t hammer_stats_btree_lookups;
783extern int64_t hammer_stats_btree_searches;
784extern int64_t hammer_stats_btree_inserts;
785extern int64_t hammer_stats_btree_deletes;
786extern int64_t hammer_stats_btree_elements;
787extern int64_t hammer_stats_btree_splits;
788extern int64_t hammer_stats_btree_iterations;
789extern int64_t hammer_stats_record_iterations;
f5a07a7a 790extern int hammer_count_dirtybufspace;
a99b9ea2 791extern int hammer_count_refedbufs;
0832c9bb 792extern int hammer_count_reservations;
a99b9ea2
MD
793extern int hammer_count_io_running_read;
794extern int hammer_count_io_running_write;
795extern int hammer_count_io_locked;
f5a07a7a 796extern int hammer_limit_dirtybufspace;
af209b0f 797extern int hammer_limit_iqueued;
47637bff 798extern int hammer_limit_recs;
1f07f686 799extern int hammer_bio_count;
cb51be26
MD
800extern int hammer_verify_zone;
801extern int hammer_write_mode;
7d683b0f 802extern int64_t hammer_contention_count;
b3deaf57 803
8750964d
MD
804int hammer_vop_inactive(struct vop_inactive_args *);
805int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 806int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 807struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
adf01747 808 hammer_inode_t dip, int64_t obj_id,
ddfdf542
MD
809 hammer_tid_t asof, u_int32_t localization,
810 int flags, int *errorp);
43c665ae
MD
811void hammer_scan_inode_snapshots(hammer_mount_t hmp,
812 hammer_inode_info_t iinfo,
813 int (*callback)(hammer_inode_t ip, void *data),
814 void *data);
66325755
MD
815void hammer_put_inode(struct hammer_inode *ip);
816void hammer_put_inode_ref(struct hammer_inode *ip);
4a2796f3 817void hammer_inode_waitreclaims(hammer_mount_t hmp);
66325755 818
8cd0a023 819int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
51c35492
MD
820int hammer_adjust_volume_mode(hammer_volume_t volume, void *data __unused);
821
a89aec1b 822int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
8cd0a023
MD
823int hammer_install_volume(hammer_mount_t hmp, const char *volname);
824
45a014dc 825int hammer_ip_lookup(hammer_cursor_t cursor);
4e17f465 826int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 827int hammer_ip_next(hammer_cursor_t cursor);
8cd0a023 828int hammer_ip_resolve_data(hammer_cursor_t cursor);
e63644f0
MD
829int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_inode_t ip,
830 hammer_tid_t tid);
602c6cb8 831int hammer_delete_at_cursor(hammer_cursor_t cursor, int delete_flags,
842e7a70
MD
832 hammer_tid_t delete_tid, u_int32_t delete_ts,
833 int track, int64_t *stat_bytes);
b3deaf57 834int hammer_ip_check_directory_empty(hammer_transaction_t trans,
98f7132d 835 hammer_inode_t ip);
fbc6e32a 836int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
f36a9737
MD
837int hammer_queue_inodes_flusher(hammer_mount_t hmp, int waitfor);
838
8cd0a023 839hammer_record_t
11ad5ade 840 hammer_alloc_mem_record(hammer_inode_t ip, int data_len);
d36ec43b 841void hammer_flush_record_done(hammer_record_t record, int error);
af209b0f 842void hammer_wait_mem_record_ident(hammer_record_t record, const char *ident);
b3deaf57 843void hammer_rel_mem_record(hammer_record_t record);
b84de5af 844
6a37e7e4 845int hammer_cursor_up(hammer_cursor_t cursor);
f36a9737 846int hammer_cursor_up_locked(hammer_cursor_t cursor);
8cd0a023 847int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4 848int hammer_cursor_upgrade(hammer_cursor_t cursor);
7bc5b8c2 849int hammer_cursor_upgrade_node(hammer_cursor_t cursor);
6a37e7e4 850void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
851int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
852 int index);
af209b0f 853void hammer_lock_ex_ident(struct hammer_lock *lock, const char *ident);
8cd0a023
MD
854int hammer_lock_ex_try(struct hammer_lock *lock);
855void hammer_lock_sh(struct hammer_lock *lock);
47637bff 856int hammer_lock_sh_try(struct hammer_lock *lock);
6a37e7e4
MD
857int hammer_lock_upgrade(struct hammer_lock *lock);
858void hammer_lock_downgrade(struct hammer_lock *lock);
b3bad96f 859int hammer_lock_status(struct hammer_lock *lock);
427e5fc6 860void hammer_unlock(struct hammer_lock *lock);
66325755
MD
861void hammer_ref(struct hammer_lock *lock);
862void hammer_unref(struct hammer_lock *lock);
8cd0a023 863
2f85fa4d
MD
864void hammer_sync_lock_ex(hammer_transaction_t trans);
865void hammer_sync_lock_sh(hammer_transaction_t trans);
47637bff 866int hammer_sync_lock_sh_try(hammer_transaction_t trans);
2f85fa4d
MD
867void hammer_sync_unlock(hammer_transaction_t trans);
868
66325755 869u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023 870void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
ddfdf542
MD
871void hammer_time_to_timespec(u_int64_t xtime, struct timespec *ts);
872u_int64_t hammer_timespec_to_time(struct timespec *ts);
ea434b6f
MD
873hammer_tid_t hammer_str_to_tid(const char *str, int *ispfs,
874 u_int32_t *localizationp);
5de0c0e5 875hammer_tid_t hammer_alloc_objid(hammer_mount_t hmp, hammer_inode_t dip);
0729c8c8
MD
876void hammer_clear_objid(hammer_inode_t dip);
877void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 878
e8599db1 879int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
b3bad96f 880 int bytes);
e8599db1 881void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 882enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 883int hammer_get_dtype(u_int8_t obj_type);
66325755 884u_int8_t hammer_get_obj_type(enum vtype vtype);
5a930e66 885int64_t hammer_directory_namekey(const void *name, int len);
e63644f0 886int hammer_nohistory(hammer_inode_t ip);
427e5fc6 887
36f82b23 888int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
b3bad96f 889 hammer_node_cache_t cache, hammer_inode_t ip);
4e17f465 890void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 891void hammer_done_cursor(hammer_cursor_t cursor);
b3bad96f 892int hammer_recover_cursor(hammer_cursor_t cursor);
3f43fb33
MD
893void hammer_unlock_cursor(hammer_cursor_t cursor, int also_ip);
894int hammer_lock_cursor(hammer_cursor_t cursor, int also_ip);
895hammer_cursor_t hammer_push_cursor(hammer_cursor_t ocursor);
896void hammer_pop_cursor(hammer_cursor_t ocursor, hammer_cursor_t ncursor);
b3bad96f
MD
897
898void hammer_cursor_replaced_node(hammer_node_t onode, hammer_node_t nnode);
899void hammer_cursor_removed_node(hammer_node_t onode, hammer_node_t parent,
900 int index);
901void hammer_cursor_split_node(hammer_node_t onode, hammer_node_t nnode,
902 int index);
903void hammer_cursor_inserted_element(hammer_node_t node, int index);
904void hammer_cursor_deleted_element(hammer_node_t node, int index);
8cd0a023
MD
905
906int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 907int hammer_btree_first(hammer_cursor_t cursor);
32c90105 908int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
909int hammer_btree_extract(hammer_cursor_t cursor, int flags);
910int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 911int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
11ad5ade 912int hammer_btree_insert(hammer_cursor_t cursor,
602c6cb8 913 hammer_btree_leaf_elm_t elm, int *doprop);
8cd0a023 914int hammer_btree_delete(hammer_cursor_t cursor);
4c038e17
MD
915void hammer_btree_do_propagation(hammer_cursor_t cursor,
916 hammer_pseudofs_inmem_t pfsm,
602c6cb8 917 hammer_btree_leaf_elm_t leaf);
8cd0a023 918int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 919int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
920int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
921int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
922
2f85fa4d
MD
923int btree_set_parent(hammer_transaction_t trans, hammer_node_t node,
924 hammer_btree_elm_t elm);
b33e2cc0
MD
925int hammer_btree_lock_children(hammer_cursor_t cursor,
926 struct hammer_node_locklist **locklistp);
2f85fa4d 927void hammer_btree_unlock_children(struct hammer_node_locklist **locklistp);
bcac4bbb 928int hammer_btree_search_node(hammer_base_elm_t elm, hammer_node_ondisk_t node);
c82af904
MD
929hammer_node_t hammer_btree_get_parent(hammer_node_t node, int *parent_indexp,
930 int *errorp, int try_exclusive);
b33e2cc0 931
c0ade690
MD
932void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
933void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 934
47197d71
MD
935void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
936 int *errorp, struct hammer_buffer **bufferp);
937void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
938 int *errorp, struct hammer_buffer **bufferp);
4a2796f3
MD
939void *hammer_bread_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
940 int *errorp, struct hammer_buffer **bufferp);
941void *hammer_bnew_ext(struct hammer_mount *hmp, hammer_off_t off, int bytes,
942 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
943
944hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
427e5fc6 945
8cd0a023 946hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 947 int32_t vol_no, int *errorp);
4a2796f3
MD
948hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp, hammer_off_t buf_offset,
949 int bytes, int isnew, int *errorp);
0832c9bb
MD
950void hammer_del_buffers(hammer_mount_t hmp, hammer_off_t base_offset,
951 hammer_off_t zone2_offset, int bytes);
8cd0a023 952
fbc6e32a 953int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
954int hammer_ref_buffer(hammer_buffer_t buffer);
955void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
956
8cd0a023 957void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
958void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
959
513ca7d7
MD
960int hammer_vfs_export(struct mount *mp, int op,
961 const struct export_args *export);
19619882
MD
962hammer_node_t hammer_get_node(hammer_mount_t hmp, hammer_off_t node_offset,
963 int isnew, int *errorp);
740d8317 964void hammer_ref_node(hammer_node_t node);
055f5ff8 965hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
bcac4bbb 966 hammer_node_cache_t cache, int *errorp);
8cd0a023 967void hammer_rel_node(hammer_node_t node);
36f82b23
MD
968void hammer_delete_node(hammer_transaction_t trans,
969 hammer_node_t node);
bcac4bbb
MD
970void hammer_cache_node(hammer_node_cache_t cache,
971 hammer_node_t node);
972void hammer_uncache_node(hammer_node_cache_t cache);
8cd0a023
MD
973void hammer_flush_node(hammer_node_t node);
974
427e5fc6
MD
975void hammer_dup_buffer(struct hammer_buffer **bufferp,
976 struct hammer_buffer *buffer);
36f82b23 977hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
36f82b23 978void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf3b416b 979 u_int16_t rec_type, hammer_off_t *data_offsetp,
bf686dbe
MD
980 struct hammer_buffer **data_bufferp, int *errorp);
981
059819e3
MD
982int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
983 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 984
66325755 985void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 986void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 987
36f82b23
MD
988hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
989 hammer_off_t owner, int *errorp);
990void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 991 hammer_off_t owner, int *errorp);
a7e9bef1 992int hammer_checkspace(hammer_mount_t hmp, int slop);
36f82b23 993hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 994 int bytes, int *errorp);
0832c9bb
MD
995hammer_reserve_t hammer_blockmap_reserve(hammer_mount_t hmp, int zone,
996 int bytes, hammer_off_t *zone_offp, int *errorp);
997void hammer_blockmap_reserve_complete(hammer_mount_t hmp,
998 hammer_reserve_t resv);
cb51be26 999void hammer_reserve_clrdelay(hammer_mount_t hmp, hammer_reserve_t resv);
36f82b23
MD
1000void hammer_blockmap_free(hammer_transaction_t trans,
1001 hammer_off_t bmap_off, int bytes);
4a2796f3
MD
1002void hammer_blockmap_finalize(hammer_transaction_t trans,
1003 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
1004int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
1005 int *curp, int *errorp);
40043e7f
MD
1006hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1007 int *errorp);
bf686dbe
MD
1008hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
1009 int *errorp);
06ad81ff
MD
1010int64_t hammer_undo_used(hammer_transaction_t trans);
1011int64_t hammer_undo_space(hammer_transaction_t trans);
1f07f686
MD
1012int64_t hammer_undo_max(hammer_mount_t hmp);
1013
8cd0a023
MD
1014void hammer_start_transaction(struct hammer_transaction *trans,
1015 struct hammer_mount *hmp);
36f82b23
MD
1016void hammer_simple_transaction(struct hammer_transaction *trans,
1017 struct hammer_mount *hmp);
b84de5af
MD
1018void hammer_start_transaction_fls(struct hammer_transaction *trans,
1019 struct hammer_mount *hmp);
1020void hammer_done_transaction(struct hammer_transaction *trans);
66325755 1021
47637bff 1022void hammer_modify_inode(hammer_inode_t ip, int flags);
f90dde4c 1023void hammer_flush_inode(hammer_inode_t ip, int flags);
b84de5af
MD
1024void hammer_flush_inode_done(hammer_inode_t ip);
1025void hammer_wait_inode(hammer_inode_t ip);
1026
8cd0a023
MD
1027int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
1028 struct ucred *cred, struct hammer_inode *dip,
ea434b6f
MD
1029 hammer_pseudofs_inmem_t pfsm,
1030 struct hammer_inode **ipp);
a89aec1b 1031void hammer_rel_inode(hammer_inode_t ip, int flush);
51c35492 1032int hammer_reload_inode(hammer_inode_t ip, void *arg __unused);
af209b0f 1033int hammer_ino_rb_compare(hammer_inode_t ip1, hammer_inode_t ip2);
51c35492 1034
1f07f686
MD
1035int hammer_sync_inode(hammer_inode_t ip);
1036void hammer_test_inode(hammer_inode_t ip);
e8599db1 1037void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 1038
a89aec1b 1039int hammer_ip_add_directory(struct hammer_transaction *trans,
5a930e66 1040 hammer_inode_t dip, const char *name, int bytes,
8cd0a023 1041 hammer_inode_t nip);
a89aec1b 1042int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
1043 hammer_cursor_t cursor, hammer_inode_t dip,
1044 hammer_inode_t ip);
47637bff 1045hammer_record_t hammer_ip_add_bulk(hammer_inode_t ip, off_t file_offset,
0832c9bb 1046 void *data, int bytes, int *errorp);
47637bff 1047int hammer_ip_frontend_trunc(struct hammer_inode *ip, off_t file_size);
7a04d74f
MD
1048int hammer_ip_add_record(struct hammer_transaction *trans,
1049 hammer_record_t record);
4e17f465 1050int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
47637bff 1051 int64_t ran_beg, int64_t ran_end, int truncating);
a9d52b76 1052int hammer_ip_delete_clean(hammer_cursor_t cursor, hammer_inode_t ip,
4e17f465
MD
1053 int *countp);
1054int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
1055 int64_t offset, void *data, int bytes);
36f82b23 1056int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 1057int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
ea434b6f
MD
1058hammer_pseudofs_inmem_t hammer_load_pseudofs(hammer_transaction_t trans,
1059 u_int32_t localization, int *errorp);
1060int hammer_mkroot_pseudofs(hammer_transaction_t trans, struct ucred *cred,
1061 hammer_pseudofs_inmem_t pfsm);
1062int hammer_save_pseudofs(hammer_transaction_t trans,
1063 hammer_pseudofs_inmem_t pfsm);
842e7a70 1064int hammer_unload_pseudofs(hammer_transaction_t trans, u_int32_t localization);
5fa5c92f 1065void hammer_rel_pseudofs(hammer_mount_t hmp, hammer_pseudofs_inmem_t pfsm);
7dc57964
MD
1066int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
1067 struct ucred *cred);
1068
10a5d1ba
MD
1069void hammer_io_init(hammer_io_t io, hammer_mount_t hmp,
1070 enum hammer_io_type type);
2f85fa4d
MD
1071int hammer_io_read(struct vnode *devvp, struct hammer_io *io,
1072 hammer_off_t limit);
66325755 1073int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
47637bff 1074void hammer_io_inval(hammer_volume_t volume, hammer_off_t zone2_offset);
ecca949a 1075struct buf *hammer_io_release(struct hammer_io *io, int flush);
055f5ff8 1076void hammer_io_flush(struct hammer_io *io);
055f5ff8 1077void hammer_io_waitdep(struct hammer_io *io);
af209b0f 1078void hammer_io_wait_all(hammer_mount_t hmp, const char *ident);
4a2796f3 1079int hammer_io_direct_read(hammer_mount_t hmp, struct bio *bio);
47637bff
MD
1080int hammer_io_direct_write(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf,
1081 struct bio *bio);
43c665ae 1082void hammer_io_direct_uncache(hammer_mount_t hmp, hammer_btree_leaf_elm_t leaf);
9f5097dc
MD
1083void hammer_io_write_interlock(hammer_io_t io);
1084void hammer_io_done_interlock(hammer_io_t io);
4a2796f3 1085void hammer_io_clear_modify(struct hammer_io *io, int inval);
cebe9493 1086void hammer_io_clear_modlist(struct hammer_io *io);
36f82b23
MD
1087void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
1088 void *base, int len);
1089void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
1090 void *base, int len);
10a5d1ba
MD
1091void hammer_modify_volume_done(hammer_volume_t volume);
1092void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 1093
36f82b23
MD
1094int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
1095 struct hammer_ioc_reblock *reblock);
11ad5ade
MD
1096int hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
1097 struct hammer_ioc_prune *prune);
c82af904
MD
1098int hammer_ioc_mirror_read(hammer_transaction_t trans, hammer_inode_t ip,
1099 struct hammer_ioc_mirror_rw *mirror);
1100int hammer_ioc_mirror_write(hammer_transaction_t trans, hammer_inode_t ip,
1101 struct hammer_ioc_mirror_rw *mirror);
5fa5c92f 1102int hammer_ioc_set_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
ea434b6f 1103 struct ucred *cred, struct hammer_ioc_pseudofs_rw *pfs);
5fa5c92f
MD
1104int hammer_ioc_get_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1105 struct hammer_ioc_pseudofs_rw *pfs);
842e7a70
MD
1106int hammer_ioc_destroy_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1107 struct hammer_ioc_pseudofs_rw *pfs);
1108int hammer_ioc_downgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1109 struct hammer_ioc_pseudofs_rw *pfs);
1110int hammer_ioc_upgrade_pseudofs(hammer_transaction_t trans, hammer_inode_t ip,
1111 struct hammer_ioc_pseudofs_rw *pfs);
bf686dbe 1112
855942b6 1113int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 1114
059819e3
MD
1115void hammer_flusher_create(hammer_mount_t hmp);
1116void hammer_flusher_destroy(hammer_mount_t hmp);
1117void hammer_flusher_sync(hammer_mount_t hmp);
7a61b85d 1118int hammer_flusher_async(hammer_mount_t hmp, hammer_flush_group_t flg);
15e75dab 1119int hammer_flusher_async_one(hammer_mount_t hmp);
93291532 1120void hammer_flusher_wait(hammer_mount_t hmp, int seq);
06ad81ff 1121int hammer_flusher_meta_limit(hammer_mount_t hmp);
93291532 1122int hammer_flusher_meta_halflimit(hammer_mount_t hmp);
06ad81ff 1123int hammer_flusher_undo_exhausted(hammer_transaction_t trans, int quarter);
525aad3a 1124void hammer_flusher_clean_loose_ios(hammer_mount_t hmp);
6c1f89f4 1125void hammer_flusher_finalize(hammer_transaction_t trans, int final);
059819e3 1126
f90dde4c 1127int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
51c35492 1128void hammer_recover_flush_buffers(hammer_mount_t hmp,
06ad81ff 1129 hammer_volume_t root_volume, int final);
f90dde4c 1130
19619882
MD
1131void hammer_crc_set_blockmap(hammer_blockmap_t blockmap);
1132void hammer_crc_set_volume(hammer_volume_ondisk_t ondisk);
ddfdf542 1133void hammer_crc_set_leaf(void *data, hammer_btree_leaf_elm_t leaf);
19619882
MD
1134
1135int hammer_crc_test_blockmap(hammer_blockmap_t blockmap);
1136int hammer_crc_test_volume(hammer_volume_ondisk_t ondisk);
19619882 1137int hammer_crc_test_btree(hammer_node_ondisk_t ondisk);
ddfdf542 1138int hammer_crc_test_leaf(void *data, hammer_btree_leaf_elm_t leaf);
77062c8a 1139void hkprintf(const char *ctl, ...);
a56cb012
MD
1140udev_t hammer_fsid_to_udev(uuid_t *uuid);
1141
19619882 1142
4a2796f3
MD
1143int hammer_blocksize(int64_t file_offset);
1144int64_t hammer_blockdemarc(int64_t file_offset1, int64_t file_offset2);
1145
055f5ff8 1146#endif
427e5fc6 1147
af209b0f
MD
1148static __inline void
1149hammer_wait_mem_record(hammer_record_t record)
1150{
1151 hammer_wait_mem_record_ident(record, "hmmwai");
1152}
1153
1154static __inline void
1155hammer_lock_ex(struct hammer_lock *lock)
1156{
1157 hammer_lock_ex_ident(lock, "hmrlck");
1158}
1159
bcac4bbb
MD
1160/*
1161 * Indicate that a B-Tree node is being modified.
1162 */
8cd0a023 1163static __inline void
36f82b23 1164hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 1165{
36f82b23 1166 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
1167}
1168
1169static __inline void
36f82b23 1170hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 1171{
36f82b23
MD
1172 hammer_modify_buffer(trans, node->buffer,
1173 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
1174}
1175
bf686dbe 1176static __inline void
36f82b23
MD
1177hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
1178 void *base, int len)
bf686dbe 1179{
19619882
MD
1180 hammer_crc_t *crcptr;
1181
bf686dbe
MD
1182 KKASSERT((char *)base >= (char *)node->ondisk &&
1183 (char *)base + len <=
1184 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 1185 hammer_modify_buffer(trans, node->buffer, base, len);
19619882
MD
1186 crcptr = &node->ondisk->crc;
1187 hammer_modify_buffer(trans, node->buffer, crcptr, sizeof(hammer_crc_t));
1188 --node->buffer->io.modify_refs; /* only want one ref */
bf686dbe
MD
1189}
1190
bcac4bbb
MD
1191/*
1192 * Indicate that the specified modifications have been completed.
1193 *
1194 * Do not try to generate the crc here, it's very expensive to do and a
1195 * sequence of insertions or deletions can result in many calls to this
1196 * function on the same node.
1197 */
10a5d1ba
MD
1198static __inline void
1199hammer_modify_node_done(hammer_node_t node)
1200{
bcac4bbb
MD
1201 node->flags |= HAMMER_NODE_CRCGOOD;
1202 if ((node->flags & HAMMER_NODE_NEEDSCRC) == 0) {
1203 node->flags |= HAMMER_NODE_NEEDSCRC;
1204 node->buffer->io.gencrc = 1;
1205 hammer_ref_node(node);
1206 }
10a5d1ba
MD
1207 hammer_modify_buffer_done(node->buffer);
1208}
1209
e8599db1
MD
1210#define hammer_modify_volume_field(trans, vol, field) \
1211 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
1212 sizeof((vol)->ondisk->field))
1213
c9b9e29d
MD
1214#define hammer_modify_node_field(trans, node, field) \
1215 hammer_modify_node(trans, node, &(node)->ondisk->field, \
1216 sizeof((node)->ondisk->field))
1217