The direct-write pipe code has a bug in it somewhere when the system is
[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 *
e8599db1 34 * $DragonFly: src/sys/vfs/hammer/hammer.h,v 1.59 2008/05/03 20:21:20 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>
427e5fc6 52#include <sys/globaldata.h>
66325755
MD
53#include <sys/lockf.h>
54#include <sys/buf.h>
8cd0a023 55#include <sys/queue.h>
66325755
MD
56#include <sys/globaldata.h>
57
58#include <sys/buf2.h>
855942b6 59#include <sys/signal2.h>
427e5fc6 60#include "hammer_disk.h"
8750964d 61#include "hammer_mount.h"
7dc57964 62#include "hammer_ioctl.h"
8750964d
MD
63
64#if defined(_KERNEL) || defined(_KERNEL_STRUCTURES)
65
66MALLOC_DECLARE(M_HAMMER);
67
66325755
MD
68struct hammer_mount;
69
8750964d
MD
70/*
71 * Key structure used for custom RB tree inode lookups. This prototypes
72 * the function hammer_ino_rb_tree_RB_LOOKUP_INFO(root, info).
73 */
74typedef struct hammer_inode_info {
513ca7d7 75 int64_t obj_id; /* (key) object identifier */
8750964d
MD
76 hammer_tid_t obj_asof; /* (key) snapshot transid or 0 */
77} *hammer_inode_info_t;
78
b84de5af
MD
79typedef enum hammer_transaction_type {
80 HAMMER_TRANS_RO,
81 HAMMER_TRANS_STD,
82 HAMMER_TRANS_FLS
83} hammer_transaction_type_t;
84
66325755
MD
85/*
86 * HAMMER Transaction tracking
87 */
88struct hammer_transaction {
b84de5af 89 hammer_transaction_type_t type;
66325755
MD
90 struct hammer_mount *hmp;
91 hammer_tid_t tid;
b84de5af 92 hammer_tid_t time;
a89aec1b 93 struct hammer_volume *rootvol;
66325755
MD
94};
95
8cd0a023
MD
96typedef struct hammer_transaction *hammer_transaction_t;
97
66325755
MD
98/*
99 * HAMMER locks
100 */
427e5fc6 101struct hammer_lock {
0b075555 102 int refs; /* active references delay writes */
0b075555 103 int lockcount; /* lock count for exclusive/shared access */
427e5fc6
MD
104 int wanted;
105 struct thread *locktd;
8750964d
MD
106};
107
427e5fc6
MD
108static __inline int
109hammer_islocked(struct hammer_lock *lock)
110{
8cd0a023 111 return(lock->lockcount != 0);
427e5fc6
MD
112}
113
0b075555
MD
114static __inline int
115hammer_isactive(struct hammer_lock *lock)
116{
117 return(lock->refs != 0);
118}
119
427e5fc6
MD
120static __inline int
121hammer_islastref(struct hammer_lock *lock)
122{
123 return(lock->refs == 1);
124}
c60bb2c5 125
6a37e7e4 126/*
7aa3b8a6 127 * Return if we specifically own the lock exclusively.
6a37e7e4
MD
128 */
129static __inline int
7aa3b8a6 130hammer_lock_excl_owned(struct hammer_lock *lock, thread_t td)
6a37e7e4 131{
7aa3b8a6
MD
132 if (lock->lockcount > 0 && lock->locktd == td)
133 return(1);
134 return(0);
6a37e7e4
MD
135}
136
ec4e8497 137/*
1f07f686 138 * Flush state, used by various structures
ec4e8497 139 */
1f07f686
MD
140typedef enum hammer_inode_state {
141 HAMMER_FST_IDLE,
142 HAMMER_FST_SETUP,
143 HAMMER_FST_FLUSH
144} hammer_inode_state_t;
ec4e8497 145
1f07f686 146TAILQ_HEAD(hammer_record_list, hammer_record);
ec4e8497 147
0729c8c8
MD
148/*
149 * Cache object ids. A fixed number of objid cache structures are
150 * created to reserve object id's for newly created files in multiples
151 * of 100,000, localized to a particular directory, and recycled as
152 * needed. This allows parallel create operations in different
153 * directories to retain fairly localized object ids which in turn
154 * improves reblocking performance and layout.
155 */
156#define OBJID_CACHE_SIZE 128
157#define OBJID_CACHE_BULK 100000
158
159typedef struct hammer_objid_cache {
160 TAILQ_ENTRY(hammer_objid_cache) entry;
161 struct hammer_inode *dip;
162 hammer_tid_t next_tid;
163 int count;
164} *hammer_objid_cache_t;
165
8750964d 166/*
8cd0a023
MD
167 * Structure used to represent an inode in-memory.
168 *
169 * The record and data associated with an inode may be out of sync with
170 * the disk (xDIRTY flags), or not even on the disk at all (ONDISK flag
171 * clear).
172 *
173 * An inode may also hold a cache of unsynchronized records, used for
174 * database and directories only. Unsynchronized regular file data is
175 * stored in the buffer cache.
176 *
177 * NOTE: A file which is created and destroyed within the initial
178 * synchronization period can wind up not doing any disk I/O at all.
179 *
180 * Finally, an inode may cache numerous disk-referencing B-Tree cursors.
8750964d
MD
181 */
182struct hammer_ino_rb_tree;
183struct hammer_inode;
184RB_HEAD(hammer_ino_rb_tree, hammer_inode);
185RB_PROTOTYPEX(hammer_ino_rb_tree, INFO, hammer_inode, rb_node,
8cd0a023
MD
186 hammer_ino_rb_compare, hammer_inode_info_t);
187
188struct hammer_rec_rb_tree;
189struct hammer_record;
190RB_HEAD(hammer_rec_rb_tree, hammer_record);
191RB_PROTOTYPEX(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
192 hammer_rec_rb_compare, hammer_base_elm_t);
193
194TAILQ_HEAD(hammer_node_list, hammer_node);
8750964d
MD
195
196struct hammer_inode {
1f07f686
MD
197 RB_ENTRY(hammer_inode) rb_node;
198 hammer_inode_state_t flush_state;
199 int flush_group;
059819e3 200 TAILQ_ENTRY(hammer_inode) flush_entry;
1f07f686 201 struct hammer_record_list target_list; /* target of dependant recs */
b84de5af
MD
202 u_int64_t obj_id; /* (key) object identifier */
203 hammer_tid_t obj_asof; /* (key) snapshot or 0 */
204 struct hammer_mount *hmp;
0729c8c8 205 hammer_objid_cache_t objid_cache;
b84de5af
MD
206 int flags;
207 int error; /* flush error */
208 int cursor_ip_refs; /* sanity */
209 struct vnode *vp;
210 struct lockf advlock;
211 struct hammer_lock lock; /* sync copy interlock */
212 TAILQ_HEAD(, bio) bio_list; /* BIOs to flush out */
213 TAILQ_HEAD(, bio) bio_alt_list; /* BIOs to flush out */
214 off_t trunc_off;
215 struct hammer_inode_record ino_rec; /* in-memory cache */
216 struct hammer_inode_data ino_data; /* in-memory cache */
217 struct hammer_rec_rb_tree rec_tree; /* in-memory cache */
61aeeb33 218 struct hammer_node *cache[2]; /* search initiate cache */
b84de5af
MD
219
220 /*
221 * When a demark is created to synchronize an inode to
222 * disk, certain fields are copied so the front-end VOPs
223 * can continue to run in parallel with the synchronization
224 * occuring in the background.
225 */
226 int sync_flags; /* to-sync flags cache */
227 off_t sync_trunc_off; /* to-sync truncation */
228 struct hammer_inode_record sync_ino_rec;/* to-sync cache */
229 struct hammer_inode_data sync_ino_data; /* to-sync cache */
8750964d
MD
230};
231
8cd0a023
MD
232typedef struct hammer_inode *hammer_inode_t;
233
66325755
MD
234#define VTOI(vp) ((struct hammer_inode *)(vp)->v_data)
235
8cd0a023
MD
236#define HAMMER_INODE_DDIRTY 0x0001 /* in-memory ino_data is dirty */
237#define HAMMER_INODE_RDIRTY 0x0002 /* in-memory ino_rec is dirty */
238#define HAMMER_INODE_ITIMES 0x0004 /* in-memory mtime/atime modified */
1f07f686 239#define HAMMER_INODE_XDIRTY 0x0008 /* in-memory records */
8cd0a023 240#define HAMMER_INODE_ONDISK 0x0010 /* inode is on-disk (else not yet) */
a89aec1b 241#define HAMMER_INODE_FLUSH 0x0020 /* flush on last ref */
869e8f55 242#define HAMMER_INODE_DELETED 0x0080 /* inode delete (backend) */
76376933 243#define HAMMER_INODE_DELONDISK 0x0100 /* delete synchronized to disk */
d113fda1 244#define HAMMER_INODE_RO 0x0200 /* read-only (because of as-of) */
e8599db1 245#define HAMMER_INODE_VHELD 0x0400 /* vnode held on sync */
0a72edae 246#define HAMMER_INODE_DONDISK 0x0800 /* data records may be on disk */
f3b0f382 247#define HAMMER_INODE_BUFS 0x1000 /* dirty high level bps present */
b84de5af 248#define HAMMER_INODE_REFLUSH 0x2000 /* pipelined flush during flush */
1f07f686 249#define HAMMER_INODE_WRITE_ALT 0x4000 /* strategy writes to alt bioq */
059819e3 250#define HAMMER_INODE_FLUSHW 0x8000 /* Someone waiting for flush */
8cd0a023 251
b84de5af 252#define HAMMER_INODE_TRUNCATED 0x00010000
869e8f55 253#define HAMMER_INODE_DELETING 0x00020000 /* inode delete request (frontend)*/
4e17f465 254#define HAMMER_INODE_RESIGNAL 0x00040000 /* re-signal on re-flush */
b84de5af 255
0b075555 256#define HAMMER_INODE_MODMASK (HAMMER_INODE_DDIRTY|HAMMER_INODE_RDIRTY| \
f3b0f382 257 HAMMER_INODE_XDIRTY|HAMMER_INODE_BUFS| \
1f07f686
MD
258 HAMMER_INODE_ITIMES|HAMMER_INODE_TRUNCATED|\
259 HAMMER_INODE_DELETING)
260
261#define HAMMER_INODE_MODMASK_NOXDIRTY \
262 (HAMMER_INODE_MODMASK & ~HAMMER_INODE_XDIRTY)
0b075555 263
8cd0a023 264#define HAMMER_MAX_INODE_CURSORS 4
66325755 265
f90dde4c 266#define HAMMER_FLUSH_SIGNAL 0x0001
4e17f465 267#define HAMMER_FLUSH_RECURSION 0x0002
f90dde4c 268
8750964d 269/*
1f07f686
MD
270 * Structure used to represent an unsynchronized record in-memory. These
271 * records typically represent directory entries. Only non-historical
272 * records are kept in-memory.
273 *
274 * Records are organized as a per-inode RB-Tree. If the inode is not
8cd0a023
MD
275 * on disk then neither are any records and the in-memory record tree
276 * represents the entire contents of the inode. If the inode is on disk
277 * then the on-disk B-Tree is scanned in parallel with the in-memory
278 * RB-Tree to synthesize the current state of the file.
279 *
1f07f686
MD
280 * Records are also used to enforce the ordering of directory create/delete
281 * operations. A new inode will not be flushed to disk unless its related
282 * directory entry is also being flushed at the same time. A directory entry
283 * will not be removed unless its related inode is also being removed at the
284 * same time.
8750964d 285 */
1f07f686 286typedef enum hammer_record_type {
e8599db1 287 HAMMER_MEM_RECORD_GENERAL, /* misc record */
1f07f686
MD
288 HAMMER_MEM_RECORD_ADD, /* positive memory cache record */
289 HAMMER_MEM_RECORD_DEL /* negative delete-on-disk record */
290} hammer_record_type_t;
291
8cd0a023
MD
292struct hammer_record {
293 RB_ENTRY(hammer_record) rb_node;
1f07f686
MD
294 TAILQ_ENTRY(hammer_record) target_entry;
295 hammer_inode_state_t flush_state;
296 int flush_group;
297 hammer_record_type_t type;
a89aec1b 298 struct hammer_lock lock;
8cd0a023 299 struct hammer_inode *ip;
1f07f686 300 struct hammer_inode *target_ip;
8cd0a023
MD
301 union hammer_record_ondisk rec;
302 union hammer_data_ondisk *data;
8cd0a023
MD
303 int flags;
304};
8750964d 305
8cd0a023
MD
306typedef struct hammer_record *hammer_record_t;
307
d36ec43b
MD
308/*
309 * Record flags. Note that FE can only be set by the frontend if the
310 * record has not been interlocked by the backend w/ BE.
311 */
8cd0a023
MD
312#define HAMMER_RECF_ALLOCDATA 0x0001
313#define HAMMER_RECF_ONRBTREE 0x0002
b84de5af 314#define HAMMER_RECF_DELETED_FE 0x0004 /* deleted (frontend) */
d36ec43b
MD
315#define HAMMER_RECF_DELETED_BE 0x0008 /* deleted (backend) */
316#define HAMMER_RECF_INBAND 0x0010
317#define HAMMER_RECF_INTERLOCK_BE 0x0020 /* backend interlock */
318#define HAMMER_RECF_WANTED 0x0040
1f07f686 319#define HAMMER_RECF_CONVERT_DELETE 0x0100 /* special case */
8cd0a023
MD
320
321/*
47197d71 322 * In-memory structures representing on-disk structures.
8cd0a023 323 */
8750964d 324struct hammer_volume;
427e5fc6 325struct hammer_buffer;
8cd0a023 326struct hammer_node;
e8599db1 327struct hammer_undo;
8750964d 328RB_HEAD(hammer_vol_rb_tree, hammer_volume);
427e5fc6 329RB_HEAD(hammer_buf_rb_tree, hammer_buffer);
8cd0a023 330RB_HEAD(hammer_nod_rb_tree, hammer_node);
e8599db1 331RB_HEAD(hammer_und_rb_tree, hammer_undo);
8750964d
MD
332
333RB_PROTOTYPE2(hammer_vol_rb_tree, hammer_volume, rb_node,
334 hammer_vol_rb_compare, int32_t);
427e5fc6 335RB_PROTOTYPE2(hammer_buf_rb_tree, hammer_buffer, rb_node,
47197d71 336 hammer_buf_rb_compare, hammer_off_t);
8cd0a023 337RB_PROTOTYPE2(hammer_nod_rb_tree, hammer_node, rb_node,
47197d71 338 hammer_nod_rb_compare, hammer_off_t);
e8599db1
MD
339RB_PROTOTYPE2(hammer_und_rb_tree, hammer_undo, rb_node,
340 hammer_und_rb_compare, hammer_off_t);
8750964d 341
66325755
MD
342/*
343 * IO management - embedded at the head of various in-memory structures
10a5d1ba
MD
344 *
345 * VOLUME - hammer_volume containing meta-data
346 * META_BUFFER - hammer_buffer containing meta-data
347 * DATA_BUFFER - hammer_buffer containing pure-data
348 *
349 * Dirty volume headers and dirty meta-data buffers are locked until the
350 * flusher can sequence them out. Dirty pure-data buffers can be written.
351 * Clean buffers can be passively released.
66325755 352 */
10a5d1ba
MD
353typedef enum hammer_io_type {
354 HAMMER_STRUCTURE_VOLUME,
355 HAMMER_STRUCTURE_META_BUFFER,
356 HAMMER_STRUCTURE_UNDO_BUFFER,
357 HAMMER_STRUCTURE_DATA_BUFFER
358} hammer_io_type_t;
66325755
MD
359
360union hammer_io_structure;
055f5ff8 361struct hammer_io;
66325755
MD
362
363struct worklist {
364 LIST_ENTRY(worklist) node;
365};
366
10a5d1ba
MD
367TAILQ_HEAD(hammer_io_list, hammer_io);
368typedef struct hammer_io_list *hammer_io_list_t;
055f5ff8 369
66325755 370struct hammer_io {
10a5d1ba
MD
371 struct worklist worklist;
372 struct hammer_lock lock;
373 enum hammer_io_type type;
374 struct hammer_mount *hmp;
375 TAILQ_ENTRY(hammer_io) mod_entry; /* list entry if modified */
376 hammer_io_list_t mod_list;
377 struct buf *bp;
378 int64_t offset;
379 int loading; /* loading/unloading interlock */
380 int modify_refs;
381
66325755
MD
382 u_int modified : 1; /* bp's data was modified */
383 u_int released : 1; /* bp released (w/ B_LOCKED set) */
055f5ff8
MD
384 u_int running : 1; /* bp write IO in progress */
385 u_int waiting : 1; /* someone is waiting on us */
b33e2cc0 386 u_int validated : 1; /* ondisk has been validated */
b58c6388
MD
387 u_int flush : 1; /* flush on last release */
388 u_int waitdep : 1; /* flush waits for dependancies */
66325755
MD
389};
390
8cd0a023
MD
391typedef struct hammer_io *hammer_io_t;
392
66325755 393/*
8cd0a023 394 * In-memory volume representing on-disk buffer
66325755 395 */
8750964d 396struct hammer_volume {
66325755 397 struct hammer_io io;
8750964d 398 RB_ENTRY(hammer_volume) rb_node;
47197d71 399 struct hammer_buf_rb_tree rb_bufs_root;
8750964d
MD
400 struct hammer_volume_ondisk *ondisk;
401 int32_t vol_no;
fbc6e32a 402 int64_t nblocks; /* note: special calculation for statfs */
47197d71
MD
403 int64_t buffer_base; /* base offset of buffer 0 */
404 hammer_off_t maxbuf_off; /* Maximum buffer offset */
c60bb2c5 405 char *vol_name;
8750964d 406 struct vnode *devvp;
427e5fc6 407 int vol_flags;
8750964d
MD
408};
409
8cd0a023
MD
410typedef struct hammer_volume *hammer_volume_t;
411
66325755 412/*
8cd0a023
MD
413 * In-memory buffer (other then volume, super-cluster, or cluster),
414 * representing an on-disk buffer.
66325755 415 */
427e5fc6 416struct hammer_buffer {
66325755 417 struct hammer_io io;
427e5fc6 418 RB_ENTRY(hammer_buffer) rb_node;
47197d71 419 void *ondisk;
427e5fc6 420 struct hammer_volume *volume;
34d829f7
MD
421 hammer_off_t zone2_offset;
422 hammer_off_t zoneX_offset;
8cd0a023 423 struct hammer_node_list clist;
66325755
MD
424};
425
8cd0a023
MD
426typedef struct hammer_buffer *hammer_buffer_t;
427
428/*
429 * In-memory B-Tree node, representing an on-disk B-Tree node.
430 *
431 * This is a hang-on structure which is backed by a hammer_buffer,
432 * indexed by a hammer_cluster, and used for fine-grained locking of
433 * B-Tree nodes in order to properly control lock ordering. A hammer_buffer
434 * can contain multiple nodes representing wildly disassociated portions
435 * of the B-Tree so locking cannot be done on a buffer-by-buffer basis.
436 *
437 * This structure uses a cluster-relative index to reduce the number
438 * of layers required to access it, and also because all on-disk B-Tree
439 * references are cluster-relative offsets.
440 */
441struct hammer_node {
442 struct hammer_lock lock; /* node-by-node lock */
443 TAILQ_ENTRY(hammer_node) entry; /* per-buffer linkage */
444 RB_ENTRY(hammer_node) rb_node; /* per-cluster linkage */
47197d71 445 hammer_off_t node_offset; /* full offset spec */
40043e7f 446 struct hammer_mount *hmp;
8cd0a023
MD
447 struct hammer_buffer *buffer; /* backing buffer */
448 hammer_node_ondisk_t ondisk; /* ptr to on-disk structure */
449 struct hammer_node **cache1; /* passive cache(s) */
450 struct hammer_node **cache2;
b3deaf57 451 int flags;
740d8317 452 int loading; /* load interlock */
8cd0a023
MD
453};
454
b3deaf57
MD
455#define HAMMER_NODE_DELETED 0x0001
456#define HAMMER_NODE_FLUSH 0x0002
b3deaf57 457
8cd0a023
MD
458typedef struct hammer_node *hammer_node_t;
459
b33e2cc0
MD
460/*
461 * List of locked nodes.
462 */
463struct hammer_node_locklist {
464 struct hammer_node_locklist *next;
465 hammer_node_t node;
466};
467
468typedef struct hammer_node_locklist *hammer_node_locklist_t;
469
470
8cd0a023
MD
471/*
472 * Common I/O management structure - embedded in in-memory structures
473 * which are backed by filesystem buffers.
474 */
66325755
MD
475union hammer_io_structure {
476 struct hammer_io io;
477 struct hammer_volume volume;
66325755 478 struct hammer_buffer buffer;
8750964d
MD
479};
480
055f5ff8
MD
481typedef union hammer_io_structure *hammer_io_structure_t;
482
bf686dbe
MD
483/*
484 * Allocation holes are recorded for a short period of time in an attempt
485 * to use up the space.
486 */
487
488#define HAMMER_MAX_HOLES 8
489
490struct hammer_hole;
491
492struct hammer_holes {
493 TAILQ_HEAD(, hammer_hole) list;
494 int count;
495};
496
497typedef struct hammer_holes *hammer_holes_t;
498
499struct hammer_hole {
500 TAILQ_ENTRY(hammer_hole) entry;
501 hammer_off_t offset;
502 int bytes;
503};
504
505typedef struct hammer_hole *hammer_hole_t;
427e5fc6 506
8cd0a023
MD
507#include "hammer_cursor.h"
508
e8599db1
MD
509/*
510 * Undo history tracking
511 */
512#define HAMMER_MAX_UNDOS 256
513
514struct hammer_undo {
515 RB_ENTRY(hammer_undo) rb_node;
516 TAILQ_ENTRY(hammer_undo) lru_entry;
517 hammer_off_t offset;
518 int bytes;
519};
520
521typedef struct hammer_undo *hammer_undo_t;
522
8750964d
MD
523/*
524 * Internal hammer mount data structure
525 */
526struct hammer_mount {
527 struct mount *mp;
27ea2398 528 /*struct vnode *rootvp;*/
8750964d
MD
529 struct hammer_ino_rb_tree rb_inos_root;
530 struct hammer_vol_rb_tree rb_vols_root;
40043e7f 531 struct hammer_nod_rb_tree rb_nods_root;
e8599db1 532 struct hammer_und_rb_tree rb_undo_root;
8750964d 533 struct hammer_volume *rootvol;
47197d71
MD
534 struct hammer_base_elm root_btree_beg;
535 struct hammer_base_elm root_btree_end;
d26d0ae9 536 char *zbuf; /* HAMMER_BUFSIZE bytes worth of all-zeros */
195c19a1
MD
537 int hflags;
538 int ronly;
539 int nvolumes;
d26d0ae9 540 int volume_iterator;
1f07f686
MD
541 int flusher_signal; /* flusher thread sequencer */
542 int flusher_act; /* currently active flush group */
543 int flusher_done; /* set to act when complete */
544 int flusher_next; /* next flush group */
4e17f465 545 int flusher_lock; /* lock sequencing of the next flush */
059819e3 546 int flusher_exiting;
f90dde4c 547 int reclaim_count;
059819e3 548 thread_t flusher_td;
855942b6 549 u_int check_interrupt;
8750964d 550 uuid_t fsid;
66325755 551 udev_t fsid_udev;
10a5d1ba
MD
552 struct hammer_io_list volu_list; /* dirty undo buffers */
553 struct hammer_io_list undo_list; /* dirty undo buffers */
554 struct hammer_io_list data_list; /* dirty data buffers */
555 struct hammer_io_list meta_list; /* dirty meta bufs */
556 struct hammer_io_list lose_list; /* loose buffers */
557 int locked_dirty_count; /* meta/volu count */
f90dde4c 558 int io_running_count;
0729c8c8 559 int objid_cache_count;
7f7c1f84 560 hammer_tid_t asof;
b84de5af 561 hammer_off_t next_tid;
66325755 562 u_int32_t namekey_iterator;
f03c9cf4 563 hammer_off_t zone_limits[HAMMER_MAX_ZONES];
513ca7d7 564 struct netexport export;
9480ff55 565 struct hammer_lock sync_lock;
40043e7f 566 struct lock blockmap_lock;
0729c8c8 567 struct hammer_blockmap blockmap[HAMMER_MAX_ZONES];
e8599db1
MD
568 struct hammer_holes holes[HAMMER_MAX_ZONES];
569 struct hammer_undo undos[HAMMER_MAX_UNDOS];
570 int undo_alloc;
571 TAILQ_HEAD(, hammer_undo) undo_lru_list;
059819e3 572 TAILQ_HEAD(, hammer_inode) flush_list;
0729c8c8 573 TAILQ_HEAD(, hammer_objid_cache) objid_cache_list;
8750964d
MD
574};
575
8cd0a023 576typedef struct hammer_mount *hammer_mount_t;
8750964d 577
fbc6e32a
MD
578struct hammer_sync_info {
579 int error;
580 int waitfor;
581};
582
8750964d
MD
583#endif
584
585#if defined(_KERNEL)
586
587extern struct vop_ops hammer_vnode_vops;
7a04d74f
MD
588extern struct vop_ops hammer_spec_vops;
589extern struct vop_ops hammer_fifo_vops;
66325755 590extern struct bio_ops hammer_bioops;
427e5fc6 591
d5ef456e 592extern int hammer_debug_general;
e8599db1 593extern int hammer_debug_inode;
7d683b0f 594extern int hammer_debug_locks;
b3deaf57 595extern int hammer_debug_btree;
d113fda1 596extern int hammer_debug_tid;
b33e2cc0 597extern int hammer_debug_recover;
46fe7ae1 598extern int hammer_debug_recover_faults;
b3deaf57
MD
599extern int hammer_count_inodes;
600extern int hammer_count_records;
601extern int hammer_count_record_datas;
602extern int hammer_count_volumes;
b3deaf57
MD
603extern int hammer_count_buffers;
604extern int hammer_count_nodes;
9480ff55
MD
605extern int hammer_count_dirtybufs;
606extern int hammer_limit_dirtybufs;
1f07f686 607extern int hammer_bio_count;
7d683b0f 608extern int64_t hammer_contention_count;
b3deaf57 609
8750964d
MD
610int hammer_vop_inactive(struct vop_inactive_args *);
611int hammer_vop_reclaim(struct vop_reclaim_args *);
e8599db1 612int hammer_get_vnode(struct hammer_inode *ip, struct vnode **vpp);
36f82b23 613struct hammer_inode *hammer_get_inode(hammer_transaction_t trans,
61aeeb33 614 struct hammer_node **cache,
d113fda1
MD
615 u_int64_t obj_id, hammer_tid_t asof, int flags,
616 int *errorp);
66325755
MD
617void hammer_put_inode(struct hammer_inode *ip);
618void hammer_put_inode_ref(struct hammer_inode *ip);
619
8cd0a023 620int hammer_unload_volume(hammer_volume_t volume, void *data __unused);
a89aec1b 621int hammer_unload_buffer(hammer_buffer_t buffer, void *data __unused);
8cd0a023
MD
622int hammer_install_volume(hammer_mount_t hmp, const char *volname);
623
a89aec1b 624int hammer_ip_lookup(hammer_cursor_t cursor, hammer_inode_t ip);
4e17f465 625int hammer_ip_first(hammer_cursor_t cursor);
a89aec1b 626int hammer_ip_next(hammer_cursor_t cursor);
40043e7f 627int hammer_ip_resolve_record_and_data(hammer_cursor_t cursor);
8cd0a023 628int hammer_ip_resolve_data(hammer_cursor_t cursor);
195c19a1 629int hammer_ip_delete_record(hammer_cursor_t cursor, hammer_tid_t tid);
7dc57964 630int hammer_delete_at_cursor(hammer_cursor_t cursor, int64_t *stat_bytes);
b3deaf57 631int hammer_ip_check_directory_empty(hammer_transaction_t trans,
1f07f686 632 hammer_cursor_t parent_cursor, hammer_inode_t ip);
fbc6e32a 633int hammer_sync_hmp(hammer_mount_t hmp, int waitfor);
fbc6e32a 634
8cd0a023 635hammer_record_t
40043e7f 636 hammer_alloc_mem_record(hammer_inode_t ip);
d36ec43b 637void hammer_flush_record_done(hammer_record_t record, int error);
b84de5af 638void hammer_wait_mem_record(hammer_record_t record);
b3deaf57 639void hammer_rel_mem_record(hammer_record_t record);
b84de5af 640
6a37e7e4 641int hammer_cursor_up(hammer_cursor_t cursor);
8cd0a023 642int hammer_cursor_down(hammer_cursor_t cursor);
6a37e7e4
MD
643int hammer_cursor_upgrade(hammer_cursor_t cursor);
644void hammer_cursor_downgrade(hammer_cursor_t cursor);
32c90105
MD
645int hammer_cursor_seek(hammer_cursor_t cursor, hammer_node_t node,
646 int index);
8cd0a023
MD
647void hammer_lock_ex(struct hammer_lock *lock);
648int hammer_lock_ex_try(struct hammer_lock *lock);
649void hammer_lock_sh(struct hammer_lock *lock);
6a37e7e4
MD
650int hammer_lock_upgrade(struct hammer_lock *lock);
651void hammer_lock_downgrade(struct hammer_lock *lock);
427e5fc6 652void hammer_unlock(struct hammer_lock *lock);
66325755
MD
653void hammer_ref(struct hammer_lock *lock);
654void hammer_unref(struct hammer_lock *lock);
8cd0a023 655
66325755 656u_int32_t hammer_to_unix_xid(uuid_t *uuid);
8cd0a023
MD
657void hammer_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
658void hammer_to_timespec(hammer_tid_t tid, struct timespec *ts);
659hammer_tid_t hammer_timespec_to_transid(struct timespec *ts);
7f7c1f84 660hammer_tid_t hammer_now_tid(void);
d113fda1 661hammer_tid_t hammer_str_to_tid(const char *str);
0729c8c8
MD
662hammer_tid_t hammer_alloc_objid(hammer_transaction_t trans, hammer_inode_t dip);
663void hammer_clear_objid(hammer_inode_t dip);
664void hammer_destroy_objid_cache(hammer_mount_t hmp);
8cd0a023 665
e8599db1
MD
666int hammer_enter_undo_history(hammer_mount_t hmp, hammer_off_t offset,
667 int bytes);
668void hammer_clear_undo_history(hammer_mount_t hmp);
66325755 669enum vtype hammer_get_vnode_type(u_int8_t obj_type);
6b4f890b 670int hammer_get_dtype(u_int8_t obj_type);
66325755
MD
671u_int8_t hammer_get_obj_type(enum vtype vtype);
672int64_t hammer_directory_namekey(void *name, int len);
427e5fc6 673
36f82b23 674int hammer_init_cursor(hammer_transaction_t trans, hammer_cursor_t cursor,
4e17f465
MD
675 struct hammer_node **cache, hammer_inode_t ip);
676int hammer_reinit_cursor(hammer_cursor_t cursor);
677void hammer_normalize_cursor(hammer_cursor_t cursor);
8cd0a023 678void hammer_done_cursor(hammer_cursor_t cursor);
6b4f890b 679void hammer_mem_done(hammer_cursor_t cursor);
8cd0a023
MD
680
681int hammer_btree_lookup(hammer_cursor_t cursor);
d26d0ae9 682int hammer_btree_first(hammer_cursor_t cursor);
32c90105 683int hammer_btree_last(hammer_cursor_t cursor);
8cd0a023
MD
684int hammer_btree_extract(hammer_cursor_t cursor, int flags);
685int hammer_btree_iterate(hammer_cursor_t cursor);
32c90105 686int hammer_btree_iterate_reverse(hammer_cursor_t cursor);
8cd0a023
MD
687int hammer_btree_insert(hammer_cursor_t cursor, hammer_btree_elm_t elm);
688int hammer_btree_delete(hammer_cursor_t cursor);
689int hammer_btree_cmp(hammer_base_elm_t key1, hammer_base_elm_t key2);
d26d0ae9 690int hammer_btree_chkts(hammer_tid_t ts, hammer_base_elm_t key);
32c90105
MD
691int hammer_btree_correct_rhb(hammer_cursor_t cursor, hammer_tid_t tid);
692int hammer_btree_correct_lhb(hammer_cursor_t cursor, hammer_tid_t tid);
693
fe7678ee 694
b33e2cc0
MD
695int hammer_btree_lock_children(hammer_cursor_t cursor,
696 struct hammer_node_locklist **locklistp);
b33e2cc0 697
c0ade690
MD
698void hammer_print_btree_node(hammer_node_ondisk_t ondisk);
699void hammer_print_btree_elm(hammer_btree_elm_t elm, u_int8_t type, int i);
8750964d 700
47197d71
MD
701void *hammer_bread(struct hammer_mount *hmp, hammer_off_t off,
702 int *errorp, struct hammer_buffer **bufferp);
703void *hammer_bnew(struct hammer_mount *hmp, hammer_off_t off,
704 int *errorp, struct hammer_buffer **bufferp);
8cd0a023
MD
705
706hammer_volume_t hammer_get_root_volume(hammer_mount_t hmp, int *errorp);
4e17f465 707int hammer_dowrite(hammer_cursor_t cursor, hammer_inode_t ip,
059819e3 708 struct bio *bio);
427e5fc6 709
8cd0a023 710hammer_volume_t hammer_get_volume(hammer_mount_t hmp,
427e5fc6 711 int32_t vol_no, int *errorp);
47197d71
MD
712hammer_buffer_t hammer_get_buffer(hammer_mount_t hmp,
713 hammer_off_t buf_offset, int isnew, int *errorp);
b58c6388 714void hammer_uncache_buffer(struct hammer_mount *hmp, hammer_off_t off);
8cd0a023 715
fbc6e32a 716int hammer_ref_volume(hammer_volume_t volume);
8cd0a023
MD
717int hammer_ref_buffer(hammer_buffer_t buffer);
718void hammer_flush_buffer_nodes(hammer_buffer_t buffer);
719
8cd0a023 720void hammer_rel_volume(hammer_volume_t volume, int flush);
8cd0a023
MD
721void hammer_rel_buffer(hammer_buffer_t buffer, int flush);
722
513ca7d7
MD
723int hammer_vfs_export(struct mount *mp, int op,
724 const struct export_args *export);
47197d71
MD
725hammer_node_t hammer_get_node(hammer_mount_t hmp,
726 hammer_off_t node_offset, int *errorp);
740d8317 727void hammer_ref_node(hammer_node_t node);
055f5ff8
MD
728hammer_node_t hammer_ref_node_safe(struct hammer_mount *hmp,
729 struct hammer_node **cache, int *errorp);
8cd0a023 730void hammer_rel_node(hammer_node_t node);
36f82b23
MD
731void hammer_delete_node(hammer_transaction_t trans,
732 hammer_node_t node);
8cd0a023
MD
733void hammer_cache_node(hammer_node_t node,
734 struct hammer_node **cache);
735void hammer_uncache_node(struct hammer_node **cache);
736void hammer_flush_node(hammer_node_t node);
737
427e5fc6
MD
738void hammer_dup_buffer(struct hammer_buffer **bufferp,
739 struct hammer_buffer *buffer);
36f82b23
MD
740hammer_node_t hammer_alloc_btree(hammer_transaction_t trans, int *errorp);
741void *hammer_alloc_record(hammer_transaction_t trans,
bf686dbe 742 hammer_off_t *rec_offp, u_int16_t rec_type,
40043e7f
MD
743 struct hammer_buffer **rec_bufferp,
744 int32_t data_len, void **datap,
745 struct hammer_buffer **data_bufferp, int *errorp);
36f82b23 746void *hammer_alloc_data(hammer_transaction_t trans, int32_t data_len,
bf686dbe
MD
747 hammer_off_t *data_offsetp,
748 struct hammer_buffer **data_bufferp, int *errorp);
749
059819e3
MD
750int hammer_generate_undo(hammer_transaction_t trans, hammer_io_t io,
751 hammer_off_t zone1_offset, void *base, int len);
427e5fc6 752
66325755 753void hammer_put_volume(struct hammer_volume *volume, int flush);
66325755 754void hammer_put_buffer(struct hammer_buffer *buffer, int flush);
427e5fc6 755
36f82b23
MD
756hammer_off_t hammer_freemap_alloc(hammer_transaction_t trans,
757 hammer_off_t owner, int *errorp);
758void hammer_freemap_free(hammer_transaction_t trans, hammer_off_t phys_offset,
c3be93f2 759 hammer_off_t owner, int *errorp);
36f82b23 760hammer_off_t hammer_blockmap_alloc(hammer_transaction_t trans, int zone,
40043e7f 761 int bytes, int *errorp);
36f82b23
MD
762void hammer_blockmap_free(hammer_transaction_t trans,
763 hammer_off_t bmap_off, int bytes);
bf686dbe
MD
764int hammer_blockmap_getfree(hammer_mount_t hmp, hammer_off_t bmap_off,
765 int *curp, int *errorp);
40043e7f
MD
766hammer_off_t hammer_blockmap_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
767 int *errorp);
bf686dbe
MD
768hammer_off_t hammer_undo_lookup(hammer_mount_t hmp, hammer_off_t bmap_off,
769 int *errorp);
1f07f686
MD
770int64_t hammer_undo_space(hammer_mount_t hmp);
771int64_t hammer_undo_max(hammer_mount_t hmp);
772
40043e7f 773
8cd0a023
MD
774void hammer_start_transaction(struct hammer_transaction *trans,
775 struct hammer_mount *hmp);
36f82b23
MD
776void hammer_simple_transaction(struct hammer_transaction *trans,
777 struct hammer_mount *hmp);
b84de5af
MD
778void hammer_start_transaction_fls(struct hammer_transaction *trans,
779 struct hammer_mount *hmp);
780void hammer_done_transaction(struct hammer_transaction *trans);
66325755
MD
781
782void hammer_modify_inode(struct hammer_transaction *trans,
8cd0a023 783 hammer_inode_t ip, int flags);
f90dde4c 784void hammer_flush_inode(hammer_inode_t ip, int flags);
b84de5af
MD
785void hammer_flush_inode_done(hammer_inode_t ip);
786void hammer_wait_inode(hammer_inode_t ip);
787
8cd0a023
MD
788int hammer_create_inode(struct hammer_transaction *trans, struct vattr *vap,
789 struct ucred *cred, struct hammer_inode *dip,
790 struct hammer_inode **ipp);
a89aec1b 791void hammer_rel_inode(hammer_inode_t ip, int flush);
1f07f686
MD
792int hammer_sync_inode(hammer_inode_t ip);
793void hammer_test_inode(hammer_inode_t ip);
e8599db1 794void hammer_inode_unloadable_check(hammer_inode_t ip, int getvp);
8cd0a023 795
a89aec1b 796int hammer_ip_add_directory(struct hammer_transaction *trans,
8cd0a023
MD
797 hammer_inode_t dip, struct namecache *ncp,
798 hammer_inode_t nip);
a89aec1b 799int hammer_ip_del_directory(struct hammer_transaction *trans,
8cd0a023
MD
800 hammer_cursor_t cursor, hammer_inode_t dip,
801 hammer_inode_t ip);
7a04d74f
MD
802int hammer_ip_add_record(struct hammer_transaction *trans,
803 hammer_record_t record);
4e17f465
MD
804int hammer_ip_delete_range(hammer_cursor_t cursor, hammer_inode_t ip,
805 int64_t ran_beg, int64_t ran_end);
806int hammer_ip_delete_range_all(hammer_cursor_t cursor, hammer_inode_t ip,
807 int *countp);
808int hammer_ip_sync_data(hammer_cursor_t cursor, hammer_inode_t ip,
809 int64_t offset, void *data, int bytes);
36f82b23 810int hammer_ip_sync_record(hammer_transaction_t trans, hammer_record_t rec);
4e17f465 811int hammer_ip_sync_record_cursor(hammer_cursor_t cursor, hammer_record_t rec);
66325755 812
7dc57964
MD
813int hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
814 struct ucred *cred);
815
10a5d1ba
MD
816void hammer_io_init(hammer_io_t io, hammer_mount_t hmp,
817 enum hammer_io_type type);
818void hammer_io_reinit(hammer_io_t io, enum hammer_io_type type);
66325755
MD
819int hammer_io_read(struct vnode *devvp, struct hammer_io *io);
820int hammer_io_new(struct vnode *devvp, struct hammer_io *io);
b58c6388 821void hammer_io_release(struct hammer_io *io);
055f5ff8 822void hammer_io_flush(struct hammer_io *io);
8cd0a023 823int hammer_io_checkflush(hammer_io_t io);
61aeeb33 824void hammer_io_clear_modify(struct hammer_io *io);
055f5ff8 825void hammer_io_waitdep(struct hammer_io *io);
66325755 826
36f82b23
MD
827void hammer_modify_volume(hammer_transaction_t trans, hammer_volume_t volume,
828 void *base, int len);
829void hammer_modify_buffer(hammer_transaction_t trans, hammer_buffer_t buffer,
830 void *base, int len);
10a5d1ba
MD
831void hammer_modify_volume_done(hammer_volume_t volume);
832void hammer_modify_buffer_done(hammer_buffer_t buffer);
0b075555 833
36f82b23
MD
834int hammer_ioc_reblock(hammer_transaction_t trans, hammer_inode_t ip,
835 struct hammer_ioc_reblock *reblock);
bf686dbe
MD
836
837void hammer_init_holes(hammer_mount_t hmp, hammer_holes_t holes);
838void hammer_free_holes(hammer_mount_t hmp, hammer_holes_t holes);
855942b6 839int hammer_signal_check(hammer_mount_t hmp);
bf686dbe 840
059819e3
MD
841void hammer_flusher_create(hammer_mount_t hmp);
842void hammer_flusher_destroy(hammer_mount_t hmp);
843void hammer_flusher_sync(hammer_mount_t hmp);
844void hammer_flusher_async(hammer_mount_t hmp);
845
f90dde4c
MD
846int hammer_recover(hammer_mount_t hmp, hammer_volume_t rootvol);
847
055f5ff8 848#endif
427e5fc6 849
8cd0a023 850static __inline void
36f82b23 851hammer_modify_node_noundo(hammer_transaction_t trans, hammer_node_t node)
bf686dbe 852{
36f82b23 853 hammer_modify_buffer(trans, node->buffer, NULL, 0);
bf686dbe
MD
854}
855
856static __inline void
36f82b23 857hammer_modify_node_all(hammer_transaction_t trans, struct hammer_node *node)
8cd0a023 858{
36f82b23
MD
859 hammer_modify_buffer(trans, node->buffer,
860 node->ondisk, sizeof(*node->ondisk));
427e5fc6
MD
861}
862
bf686dbe 863static __inline void
36f82b23
MD
864hammer_modify_node(hammer_transaction_t trans, hammer_node_t node,
865 void *base, int len)
bf686dbe
MD
866{
867 KKASSERT((char *)base >= (char *)node->ondisk &&
868 (char *)base + len <=
869 (char *)node->ondisk + sizeof(*node->ondisk));
36f82b23 870 hammer_modify_buffer(trans, node->buffer, base, len);
bf686dbe
MD
871}
872
10a5d1ba
MD
873static __inline void
874hammer_modify_node_done(hammer_node_t node)
875{
876 hammer_modify_buffer_done(node->buffer);
877}
878
bf686dbe 879static __inline void
36f82b23
MD
880hammer_modify_record(hammer_transaction_t trans, hammer_buffer_t buffer,
881 void *base, int len)
bf686dbe
MD
882{
883 KKASSERT((char *)base >= (char *)buffer->ondisk &&
884 (char *)base + len <= (char *)buffer->ondisk + HAMMER_BUFSIZE);
36f82b23 885 hammer_modify_buffer(trans, buffer, base, len);
bf686dbe
MD
886}
887
10a5d1ba
MD
888static __inline void
889hammer_modify_record_done(hammer_buffer_t buffer)
890{
891 hammer_modify_buffer_done(buffer);
892}
893
e8599db1
MD
894#define hammer_modify_volume_field(trans, vol, field) \
895 hammer_modify_volume(trans, vol, &(vol)->ondisk->field, \
896 sizeof((vol)->ondisk->field))
897
898