2 * Copyright (c) 2011-2012 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 * by Venkatesh Srinivas <vsrinivas@dragonflybsd.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * This header file contains structures used internally by the HAMMER2
38 * implementation. See hammer2_disk.h for on-disk structures.
41 #ifndef _VFS_HAMMER2_HAMMER2_H_
42 #define _VFS_HAMMER2_HAMMER2_H_
44 #include <sys/param.h>
45 #include <sys/types.h>
46 #include <sys/kernel.h>
48 #include <sys/systm.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/vnode.h>
54 #include <sys/mountctl.h>
57 #include <sys/thread.h>
58 #include <sys/globaldata.h>
59 #include <sys/lockf.h>
61 #include <sys/queue.h>
62 #include <sys/limits.h>
64 #include <sys/signal2.h>
67 #include "hammer2_disk.h"
68 #include "hammer2_mount.h"
69 #include "hammer2_ioctl.h"
70 #include "hammer2_ccms.h"
71 #include "hammer2_network.h"
76 struct hammer2_pfsmount;
81 * The chain structure tracks blockref recursions all the way to
82 * the root volume. These consist of indirect blocks, inodes,
83 * and eventually the volume header.
85 * The chain structure is embedded in the hammer2_mount, hammer2_inode,
86 * and other system memory structures. The chain structure typically
87 * implements the reference count and busy flag for the larger structure.
89 * It is always possible to track a chain element all the way back to the
90 * root by following the (parent) links. (index) is a type-dependent index
91 * in the parent indicating where in the parent the chain element resides.
93 * When a blockref is added or deleted the related chain element is marked
94 * modified and all of its parents are marked SUBMODIFIED (the parent
95 * recursion can stop once we hit a node that is already marked SUBMODIFIED).
96 * A deleted chain element must remain intact until synchronized against
99 * The blockref at (parent, index) is not adjusted until the modified chain
100 * element is flushed and unmarked. Until then the child's blockref may
101 * not match the blockref at (parent, index).
103 RB_HEAD(hammer2_chain_tree, hammer2_chain);
104 RB_HEAD(hammer2_state_tree, hammer2_state);
105 TAILQ_HEAD(flush_deferral_list, hammer2_chain);
107 struct hammer2_chain {
108 ccms_cst_t cst; /* attr or data cst */
109 struct hammer2_blockref bref;
110 struct hammer2_blockref bref_flush; /* synchronized w/MOVED bit */
111 struct hammer2_chain *parent; /* return chain to root */
112 struct hammer2_chain_tree rbhead;
113 struct hammer2_state *state; /* if active cache msg */
114 RB_ENTRY(hammer2_chain) rbnode;
115 TAILQ_ENTRY(hammer2_chain) flush_node; /* flush deferral list */
117 struct hammer2_inode *ip;
118 struct hammer2_indblock *np;
119 struct hammer2_data *dp;
123 struct buf *bp; /* buffer cache (ro) */
124 hammer2_media_data_t *data; /* modified copy of data (rw) */
125 u_int bytes; /* physical size of data */
126 int index; /* index in parent */
131 typedef struct hammer2_chain hammer2_chain_t;
133 int hammer2_chain_cmp(hammer2_chain_t *chain1, hammer2_chain_t *chain2);
134 RB_PROTOTYPE(hammer2_chain_tree, hammer2_chain, rbnode, hammer2_chain_cmp);
137 * MOVED - This bit is set during the flush when the MODIFIED bit is cleared,
138 * indicating that the parent's blocktable must inherit a change to
139 * the bref (typically a block reallocation)
141 * It must also be set in situations where a chain is not MODIFIED
142 * but whos bref has changed (typically due to fields other than
143 * a block reallocation).
145 #define HAMMER2_CHAIN_MODIFIED 0x00000001 /* active mods */
146 #define HAMMER2_CHAIN_DIRTYEMBED 0x00000002 /* inode embedded */
147 #define HAMMER2_CHAIN_DIRTYBP 0x00000004 /* dirty on unlock */
148 #define HAMMER2_CHAIN_SUBMODIFIED 0x00000008 /* 1+ subs modified */
149 #define HAMMER2_CHAIN_DELETED 0x00000010
150 #define HAMMER2_CHAIN_INITIAL 0x00000020 /* initial create */
151 #define HAMMER2_CHAIN_FLUSHED 0x00000040 /* flush on unlock */
152 #define HAMMER2_CHAIN_MOVED 0x00000080 /* bref changed */
153 #define HAMMER2_CHAIN_IOFLUSH 0x00000100 /* bawrite on put */
154 #define HAMMER2_CHAIN_DEFERRED 0x00000200 /* on a deferral list*/
155 #define HAMMER2_CHAIN_DESTROYED 0x00000400 /* destroying */
156 #define HAMMER2_CHAIN_MODIFIED_AUX 0x00000800 /* hmp->vchain only */
157 #define HAMMER2_CHAIN_MODIFY_TID 0x00001000 /* mod updates field */
158 #define HAMMER2_CHAIN_MOUNTED 0x00002000 /* PFS is mounted */
161 * Flags passed to hammer2_chain_lookup() and hammer2_chain_next()
163 #define HAMMER2_LOOKUP_NOLOCK 0x00000001 /* ref only */
164 #define HAMMER2_LOOKUP_NODATA 0x00000002 /* data left NULL */
165 #define HAMMER2_LOOKUP_SHARED 0x00000100
168 * Flags passed to hammer2_chain_modify() and hammer2_chain_resize()
170 * NOTE: OPTDATA allows us to avoid instantiating buffers for INDIRECT
171 * blocks in the INITIAL-create state.
173 * NOTE: NO_MODIFY_TID tells the function to not set HAMMER2_CHAIN_MODIFY_TID
174 * when marking the chain modified (used when a sub-chain modification
175 * propagates upward).
177 #define HAMMER2_MODIFY_NOSUB 0x00000001 /* do not set SUBMOD */
178 #define HAMMER2_MODIFY_OPTDATA 0x00000002 /* data can be NULL */
179 #define HAMMER2_MODIFY_NO_MODIFY_TID 0x00000004
182 * Flags passed to hammer2_chain_lock()
184 #define HAMMER2_RESOLVE_NEVER 1
185 #define HAMMER2_RESOLVE_MAYBE 2
186 #define HAMMER2_RESOLVE_ALWAYS 3
187 #define HAMMER2_RESOLVE_MASK 0x0F
189 #define HAMMER2_RESOLVE_SHARED 0x10
192 * Cluster different types of storage together for allocations
194 #define HAMMER2_FREECACHE_INODE 0
195 #define HAMMER2_FREECACHE_INDIR 1
196 #define HAMMER2_FREECACHE_DATA 2
197 #define HAMMER2_FREECACHE_UNUSED3 3
198 #define HAMMER2_FREECACHE_TYPES 4
201 * BMAP read-ahead maximum parameters
203 #define HAMMER2_BMAP_COUNT 16 /* max bmap read-ahead */
204 #define HAMMER2_BMAP_BYTES (HAMMER2_PBUFSIZE * HAMMER2_BMAP_COUNT)
209 #define HAMMER2_FLUSH_DEPTH_LIMIT 40 /* stack recursion limit */
212 * HAMMER2 IN-MEMORY CACHE OF MEDIA STRUCTURES
214 * There is an in-memory representation of all on-media data structure.
216 * When accessed read-only the data will be mapped to the related buffer
219 * When accessed read-write (marked modified) a kmalloc()'d copy of the
220 * is created which can then be modified. The copy is destroyed when a
221 * filesystem block is allocated to replace it.
223 * Active inodes (those with vnodes attached) will maintain the kmalloc()'d
224 * copy for both the read-only and the read-write case. The combination of
225 * (bp) and (data) determines whether (data) was allocated or not.
227 * The in-memory representation may remain cached (for example in order to
228 * placemark clustering locks) even after the related data has been
235 * NOTE: The inode's attribute CST which is also used to lock the inode
236 * is embedded in the chain (chain.cst) and aliased w/ attr_cst.
238 struct hammer2_inode {
239 struct hammer2_mount *hmp; /* Global mount */
240 struct hammer2_pfsmount *pmp; /* PFS mount */
241 struct hammer2_inode *pip; /* parent inode */
243 ccms_cst_t topo_cst; /* directory topology cst */
244 hammer2_chain_t chain;
245 struct hammer2_inode_data ip_data;
246 struct lockf advlock;
247 u_int depth; /* directory depth */
248 hammer2_off_t delta_dcount; /* adjust data_count */
249 hammer2_off_t delta_icount; /* adjust inode_count */
252 typedef struct hammer2_inode hammer2_inode_t;
255 * A hammer2 indirect block
257 struct hammer2_indblock {
258 hammer2_chain_t chain;
261 typedef struct hammer2_indblock hammer2_indblock_t;
264 * A hammer2 data block
266 struct hammer2_data {
267 hammer2_chain_t chain;
270 typedef struct hammer2_data hammer2_data_t;
272 struct hammer2_freecache {
274 hammer2_off_t single;
277 typedef struct hammer2_freecache hammer2_freecache_t;
280 * Global (per device) mount structure for device (aka vp->v_mount->hmp)
282 struct hammer2_mount {
283 struct vnode *devvp; /* device vnode */
284 int ronly; /* read-only mount */
285 int pmp_count; /* PFS mounts backed by us */
286 TAILQ_ENTRY(hammer2_mount) mntentry; /* hammer2_mntlist */
288 struct malloc_type *minode;
292 struct malloc_type *mchain;
295 hammer2_chain_t vchain; /* anchor chain */
296 hammer2_chain_t *schain; /* super-root */
297 struct lock alloclk; /* lockmgr lock */
298 struct lock voldatalk; /* lockmgr lock */
300 hammer2_volume_data_t voldata;
301 hammer2_freecache_t freecache[HAMMER2_FREECACHE_TYPES]
302 [HAMMER2_MAX_RADIX+1];
305 typedef struct hammer2_mount hammer2_mount_t;
308 * Per-PFS mount structure for device (aka vp->v_mount)
310 struct hammer2_pfsmount {
311 struct mount *mp; /* kernel mount */
312 struct hammer2_mount *hmp; /* device global mount */
313 hammer2_chain_t *rchain; /* PFS root chain */
314 hammer2_inode_t *iroot; /* PFS root inode */
315 struct malloc_type *mmsg;
316 ccms_domain_t ccms_dom;
317 struct netexport export; /* nfs export */
318 int ronly; /* read-only mount */
319 struct file *msg_fp; /* cluster pipe->userland */
320 thread_t msgrd_td; /* cluster thread */
321 thread_t msgwr_td; /* cluster thread */
322 int msg_ctl; /* wakeup flags */
323 int msg_seq; /* cluster msg sequence id */
324 uint32_t msgid_iterator;
325 struct lock msglk; /* lockmgr lock */
326 TAILQ_HEAD(, hammer2_msg) msgq; /* transmit queue */
327 struct hammer2_state *freerd_state; /* allocation cache */
328 struct hammer2_state *freewr_state; /* allocation cache */
329 struct hammer2_state_tree staterd_tree; /* active messages */
330 struct hammer2_state_tree statewr_tree; /* active messages */
333 typedef struct hammer2_pfsmount hammer2_pfsmount_t;
335 #define HAMMER2_CLUSTERCTL_KILL 0x0001
338 * In-memory message structure for hammer2.
340 * Persistent cache state messages will be associated with a hammer2_chain.
342 * NOTE! If REPLY is set then the source and target fields in the message
343 * are swapped. That is, source and target remain unchanged whether
344 * the message is a command from side A or a reply from side B.
345 * The message is routed based on target if REPLY is not set, and on
346 * source if REPLY is set.
348 struct hammer2_state {
349 RB_ENTRY(hammer2_state) rbnode; /* indexed by msgid */
350 struct hammer2_pfsmount *pmp;
351 uint32_t txcmd; /* mostly for CMDF flags */
352 uint32_t rxcmd; /* mostly for CMDF flags */
353 uint64_t spanid; /* spanning tree routing */
354 uint64_t msgid; /* {spanid,msgid} uniq */
357 struct hammer2_chain *chain; /* msg associated w/chain */
358 struct hammer2_msg *msg;
359 int (*func)(struct hammer2_state *, struct hammer2_msg *);
362 hammer2_pfsmount_t *pmp;
366 #define HAMMER2_STATE_INSERTED 0x0001
367 #define HAMMER2_STATE_DYNAMIC 0x0002
370 TAILQ_ENTRY(hammer2_msg) qentry; /* serialized queue */
371 struct hammer2_state *state;
375 hammer2_msg_any_t any;
378 typedef struct hammer2_state hammer2_state_t;
379 typedef struct hammer2_msg hammer2_msg_t;
381 int hammer2_state_cmp(hammer2_state_t *state1, hammer2_state_t *state2);
382 RB_PROTOTYPE(hammer2_state_tree, hammer2_state, rbnode, hammer2_state_cmp);
387 MALLOC_DECLARE(M_HAMMER2);
389 #define VTOI(vp) ((hammer2_inode_t *)(vp)->v_data)
390 #define ITOV(ip) ((ip)->vp)
394 MPTOPMP(struct mount *mp)
396 return ((hammer2_pfsmount_t *)mp->mnt_data);
401 MPTOHMP(struct mount *mp)
403 return (((hammer2_pfsmount_t *)mp->mnt_data)->hmp);
406 extern struct vop_ops hammer2_vnode_vops;
407 extern struct vop_ops hammer2_spec_vops;
408 extern struct vop_ops hammer2_fifo_vops;
410 extern int hammer2_debug;
411 extern int hammer2_cluster_enable;
412 extern int hammer2_hardlink_enable;
413 extern long hammer2_iod_file_read;
414 extern long hammer2_iod_meta_read;
415 extern long hammer2_iod_indr_read;
416 extern long hammer2_iod_file_write;
417 extern long hammer2_iod_meta_write;
418 extern long hammer2_iod_indr_write;
419 extern long hammer2_iod_volu_write;
420 extern long hammer2_ioa_file_read;
421 extern long hammer2_ioa_meta_read;
422 extern long hammer2_ioa_indr_read;
423 extern long hammer2_ioa_file_write;
424 extern long hammer2_ioa_meta_write;
425 extern long hammer2_ioa_indr_write;
426 extern long hammer2_ioa_volu_write;
431 void hammer2_inode_lock_ex(hammer2_inode_t *ip);
432 void hammer2_inode_unlock_ex(hammer2_inode_t *ip);
433 void hammer2_inode_lock_sh(hammer2_inode_t *ip);
434 void hammer2_inode_unlock_sh(hammer2_inode_t *ip);
435 void hammer2_inode_busy(hammer2_inode_t *ip);
436 void hammer2_inode_unbusy(hammer2_inode_t *ip);
437 void hammer2_voldata_lock(hammer2_mount_t *hmp);
438 void hammer2_voldata_unlock(hammer2_mount_t *hmp);
440 void hammer2_mount_exlock(hammer2_mount_t *hmp);
441 void hammer2_mount_shlock(hammer2_mount_t *hmp);
442 void hammer2_mount_unlock(hammer2_mount_t *hmp);
444 int hammer2_get_dtype(hammer2_inode_t *ip);
445 int hammer2_get_vtype(hammer2_inode_t *ip);
446 u_int8_t hammer2_get_obj_type(enum vtype vtype);
447 void hammer2_time_to_timespec(u_int64_t xtime, struct timespec *ts);
448 u_int64_t hammer2_timespec_to_time(struct timespec *ts);
449 u_int32_t hammer2_to_unix_xid(uuid_t *uuid);
450 void hammer2_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
452 hammer2_key_t hammer2_dirhash(const unsigned char *name, size_t len);
453 int hammer2_bytes_to_radix(size_t bytes);
455 int hammer2_calc_logical(hammer2_inode_t *ip, hammer2_off_t uoff,
456 hammer2_key_t *lbasep, hammer2_key_t *leofp);
457 void hammer2_update_time(uint64_t *timep);
462 struct vnode *hammer2_igetv(hammer2_inode_t *ip, int *errorp);
464 void hammer2_inode_lock_nlinks(hammer2_inode_t *ip);
465 void hammer2_inode_unlock_nlinks(hammer2_inode_t *ip);
466 hammer2_inode_t *hammer2_inode_alloc(hammer2_pfsmount_t *pmp, void *data);
467 void hammer2_inode_free(hammer2_inode_t *ip);
468 void hammer2_inode_ref(hammer2_inode_t *ip);
469 void hammer2_inode_drop(hammer2_inode_t *ip);
470 int hammer2_inode_calc_alloc(hammer2_key_t filesize);
472 int hammer2_inode_create(hammer2_inode_t *dip,
473 struct vattr *vap, struct ucred *cred,
474 const uint8_t *name, size_t name_len,
475 hammer2_inode_t **nipp);
477 int hammer2_inode_duplicate(hammer2_inode_t *dip,
478 hammer2_inode_t *oip, hammer2_inode_t **nipp,
479 const uint8_t *name, size_t name_len);
480 int hammer2_inode_connect(hammer2_inode_t *dip, hammer2_inode_t *oip,
481 const uint8_t *name, size_t name_len);
483 int hammer2_unlink_file(hammer2_inode_t *dip,
484 const uint8_t *name, size_t name_len,
485 int isdir, hammer2_inode_t *retain_ip);
486 int hammer2_hardlink_consolidate(hammer2_inode_t **ipp, hammer2_inode_t *tdip);
487 int hammer2_hardlink_deconsolidate(hammer2_inode_t *dip,
488 hammer2_chain_t **chainp, hammer2_inode_t **ipp);
489 int hammer2_hardlink_find(hammer2_inode_t *dip, hammer2_chain_t **chainp,
490 hammer2_inode_t **ipp);
495 void hammer2_modify_volume(hammer2_mount_t *hmp);
496 hammer2_chain_t *hammer2_chain_alloc(hammer2_mount_t *hmp,
497 hammer2_blockref_t *bref);
498 void hammer2_chain_free(hammer2_mount_t *hmp, hammer2_chain_t *chain);
499 void hammer2_chain_ref(hammer2_mount_t *hmp, hammer2_chain_t *chain);
500 void hammer2_chain_drop(hammer2_mount_t *hmp, hammer2_chain_t *chain);
501 int hammer2_chain_lock(hammer2_mount_t *hmp, hammer2_chain_t *chain, int how);
502 void hammer2_chain_moved(hammer2_mount_t *hmp, hammer2_chain_t *chain);
503 void hammer2_chain_modify(hammer2_mount_t *hmp, hammer2_chain_t *chain,
505 void hammer2_chain_resize(hammer2_inode_t *ip, hammer2_chain_t *chain,
506 int nradix, int flags);
507 void hammer2_chain_unlock(hammer2_mount_t *hmp, hammer2_chain_t *chain);
508 hammer2_chain_t *hammer2_chain_find(hammer2_mount_t *hmp,
509 hammer2_chain_t *parent, int index);
510 hammer2_chain_t *hammer2_chain_get(hammer2_mount_t *hmp,
511 hammer2_chain_t *parent,
512 int index, int flags);
513 hammer2_chain_t *hammer2_chain_lookup(hammer2_mount_t *hmp,
514 hammer2_chain_t **parentp,
515 hammer2_key_t key_beg, hammer2_key_t key_end,
517 hammer2_chain_t *hammer2_chain_next(hammer2_mount_t *hmp,
518 hammer2_chain_t **parentp,
519 hammer2_chain_t *chain,
520 hammer2_key_t key_beg, hammer2_key_t key_end,
522 hammer2_chain_t *hammer2_chain_create(hammer2_mount_t *hmp,
523 hammer2_chain_t *parent,
524 hammer2_chain_t *chain,
525 hammer2_key_t key, int keybits,
526 int type, size_t bytes);
527 void hammer2_chain_delete(hammer2_mount_t *hmp, hammer2_chain_t *parent,
528 hammer2_chain_t *chain, int retain);
529 void hammer2_chain_flush(hammer2_mount_t *hmp, hammer2_chain_t *chain,
530 hammer2_tid_t modify_tid);
531 void hammer2_chain_commit(hammer2_mount_t *hmp, hammer2_chain_t *chain);
536 int hammer2_ioctl(hammer2_inode_t *ip, u_long com, void *data,
537 int fflag, struct ucred *cred);
542 int hammer2_state_msgrx(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
543 int hammer2_state_msgtx(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
544 void hammer2_state_cleanuprx(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
545 void hammer2_state_cleanuptx(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
546 int hammer2_msg_execute(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
547 void hammer2_state_free(hammer2_state_t *state);
548 void hammer2_msg_free(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
549 hammer2_msg_t *hammer2_msg_alloc(hammer2_pfsmount_t *pmp, uint64_t spanid,
551 void hammer2_msg_write(hammer2_pfsmount_t *pmp,
553 int (*func)(hammer2_state_t *, hammer2_msg_t *),
555 void hammer2_msg_reply(hammer2_pfsmount_t *pmp,
556 hammer2_msg_t *msg, uint32_t error);
557 void hammer2_msg_result(hammer2_pfsmount_t *pmp,
558 hammer2_msg_t *msg, uint32_t error);
563 int hammer2_msg_dbg_rcvmsg(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
564 int hammer2_msg_adhoc_input(hammer2_pfsmount_t *pmp, hammer2_msg_t *msg);
569 hammer2_off_t hammer2_freemap_alloc(hammer2_mount_t *hmp,
570 int type, size_t bytes);
571 void hammer2_freemap_free(hammer2_mount_t *hmp, hammer2_off_t data_off,
574 #endif /* !_KERNEL */
575 #endif /* !_VFS_HAMMER2_HAMMER2_H_ */