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;
82 * The chain structure tracks blockref recursions all the way to
83 * the root volume. These consist of indirect blocks, inodes,
84 * and eventually the volume header.
86 * The chain structure is embedded in the hammer2_mount, hammer2_inode,
87 * and other system memory structures. The chain structure typically
88 * implements the reference count and busy flag for the larger structure.
90 * It is always possible to track a chain element all the way back to the
91 * root by following the (parent) links. (index) is a type-dependent index
92 * in the parent indicating where in the parent the chain element resides.
94 * When a blockref is added or deleted the related chain element is marked
95 * modified and all of its parents are marked SUBMODIFIED (the parent
96 * recursion can stop once we hit a node that is already marked SUBMODIFIED).
97 * A deleted chain element must remain intact until synchronized against
100 * The blockref at (parent, index) is not adjusted until the modified chain
101 * element is flushed and unmarked. Until then the child's blockref may
102 * not match the blockref at (parent, index).
104 RB_HEAD(hammer2_chain_tree, hammer2_chain);
105 RB_HEAD(hammer2_state_tree, hammer2_state);
106 TAILQ_HEAD(flush_deferral_list, hammer2_chain);
108 struct hammer2_chain {
109 ccms_cst_t cst; /* attr or data cst */
110 struct hammer2_blockref bref;
111 struct hammer2_blockref bref_flush; /* synchronized w/MOVED bit */
112 struct hammer2_chain *parent; /* return chain to root */
113 struct hammer2_chain_tree rbhead;
114 struct hammer2_state *state; /* if active cache msg */
115 RB_ENTRY(hammer2_chain) rbnode;
116 TAILQ_ENTRY(hammer2_chain) flush_node; /* flush deferral list */
118 struct hammer2_inode *ip;
119 struct hammer2_indblock *np;
120 struct hammer2_data *dp;
124 struct buf *bp; /* buffer cache (ro) */
125 hammer2_media_data_t *data; /* modified copy of data (rw) */
126 u_int bytes; /* physical size of data */
127 int index; /* index in parent */
132 typedef struct hammer2_chain hammer2_chain_t;
134 int hammer2_chain_cmp(hammer2_chain_t *chain1, hammer2_chain_t *chain2);
135 RB_PROTOTYPE(hammer2_chain_tree, hammer2_chain, rbnode, hammer2_chain_cmp);
138 * MOVED - This bit is set during the flush when the MODIFIED bit is cleared,
139 * indicating that the parent's blocktable must inherit a change to
140 * the bref (typically a block reallocation)
142 * It must also be set in situations where a chain is not MODIFIED
143 * but whos bref has changed (typically due to fields other than
144 * a block reallocation).
146 #define HAMMER2_CHAIN_MODIFIED 0x00000001 /* active mods */
147 #define HAMMER2_CHAIN_DIRTYEMBED 0x00000002 /* inode embedded */
148 #define HAMMER2_CHAIN_DIRTYBP 0x00000004 /* dirty on unlock */
149 #define HAMMER2_CHAIN_SUBMODIFIED 0x00000008 /* 1+ subs modified */
150 #define HAMMER2_CHAIN_DELETED 0x00000010
151 #define HAMMER2_CHAIN_INITIAL 0x00000020 /* initial create */
152 #define HAMMER2_CHAIN_FLUSHED 0x00000040 /* flush on unlock */
153 #define HAMMER2_CHAIN_MOVED 0x00000080 /* bref changed */
154 #define HAMMER2_CHAIN_IOFLUSH 0x00000100 /* bawrite on put */
155 #define HAMMER2_CHAIN_DEFERRED 0x00000200 /* on a deferral list*/
156 #define HAMMER2_CHAIN_DESTROYED 0x00000400 /* destroying */
157 #define HAMMER2_CHAIN_MODIFIED_AUX 0x00000800 /* hmp->vchain only */
158 #define HAMMER2_CHAIN_MODIFY_TID 0x00001000 /* mod updates field */
159 #define HAMMER2_CHAIN_MOUNTED 0x00002000 /* PFS is mounted */
162 * Flags passed to hammer2_chain_lookup() and hammer2_chain_next()
164 #define HAMMER2_LOOKUP_NOLOCK 0x00000001 /* ref only */
165 #define HAMMER2_LOOKUP_NODATA 0x00000002 /* data left NULL */
166 #define HAMMER2_LOOKUP_SHARED 0x00000100
169 * Flags passed to hammer2_chain_modify() and hammer2_chain_resize()
171 * NOTE: OPTDATA allows us to avoid instantiating buffers for INDIRECT
172 * blocks in the INITIAL-create state.
174 * NOTE: NO_MODIFY_TID tells the function to not set HAMMER2_CHAIN_MODIFY_TID
175 * when marking the chain modified (used when a sub-chain modification
176 * propagates upward).
178 #define HAMMER2_MODIFY_NOSUB 0x00000001 /* do not set SUBMOD */
179 #define HAMMER2_MODIFY_OPTDATA 0x00000002 /* data can be NULL */
180 #define HAMMER2_MODIFY_NO_MODIFY_TID 0x00000004
183 * Flags passed to hammer2_chain_lock()
185 #define HAMMER2_RESOLVE_NEVER 1
186 #define HAMMER2_RESOLVE_MAYBE 2
187 #define HAMMER2_RESOLVE_ALWAYS 3
188 #define HAMMER2_RESOLVE_MASK 0x0F
190 #define HAMMER2_RESOLVE_SHARED 0x10
193 * Cluster different types of storage together for allocations
195 #define HAMMER2_FREECACHE_INODE 0
196 #define HAMMER2_FREECACHE_INDIR 1
197 #define HAMMER2_FREECACHE_DATA 2
198 #define HAMMER2_FREECACHE_UNUSED3 3
199 #define HAMMER2_FREECACHE_TYPES 4
202 * BMAP read-ahead maximum parameters
204 #define HAMMER2_BMAP_COUNT 16 /* max bmap read-ahead */
205 #define HAMMER2_BMAP_BYTES (HAMMER2_PBUFSIZE * HAMMER2_BMAP_COUNT)
210 #define HAMMER2_FLUSH_DEPTH_LIMIT 40 /* stack recursion limit */
213 * HAMMER2 IN-MEMORY CACHE OF MEDIA STRUCTURES
215 * There is an in-memory representation of all on-media data structure.
217 * When accessed read-only the data will be mapped to the related buffer
220 * When accessed read-write (marked modified) a kmalloc()'d copy of the
221 * is created which can then be modified. The copy is destroyed when a
222 * filesystem block is allocated to replace it.
224 * Active inodes (those with vnodes attached) will maintain the kmalloc()'d
225 * copy for both the read-only and the read-write case. The combination of
226 * (bp) and (data) determines whether (data) was allocated or not.
228 * The in-memory representation may remain cached (for example in order to
229 * placemark clustering locks) even after the related data has been
236 * NOTE: The inode's attribute CST which is also used to lock the inode
237 * is embedded in the chain (chain.cst) and aliased w/ attr_cst.
239 struct hammer2_inode {
240 struct hammer2_mount *hmp; /* Global mount */
241 struct hammer2_pfsmount *pmp; /* PFS mount */
242 struct hammer2_inode *pip; /* parent inode */
244 ccms_cst_t topo_cst; /* directory topology cst */
245 hammer2_chain_t chain;
246 struct hammer2_inode_data ip_data;
247 struct lockf advlock;
248 u_int depth; /* directory depth */
249 hammer2_off_t delta_dcount; /* adjust data_count */
250 hammer2_off_t delta_icount; /* adjust inode_count */
253 typedef struct hammer2_inode hammer2_inode_t;
256 * A hammer2 indirect block
258 struct hammer2_indblock {
259 hammer2_chain_t chain;
262 typedef struct hammer2_indblock hammer2_indblock_t;
265 * A hammer2 data block
267 struct hammer2_data {
268 hammer2_chain_t chain;
271 typedef struct hammer2_data hammer2_data_t;
273 struct hammer2_freecache {
275 hammer2_off_t single;
278 typedef struct hammer2_freecache hammer2_freecache_t;
281 * Structure used to represent a virtual circuit for a messaging
282 * route. Typically associated from hammer2_state but the hammer2_pfsmount
283 * structure also has one to represent the point-to-point link.
285 struct hammer2_router {
286 struct hammer2_pfsmount *pmp;
287 struct hammer2_state *state; /* received LNK_SPAN state */
288 uint64_t target; /* target */
291 typedef struct hammer2_router hammer2_router_t;
294 * Global (per device) mount structure for device (aka vp->v_mount->hmp)
296 struct hammer2_mount {
297 struct vnode *devvp; /* device vnode */
298 int ronly; /* read-only mount */
299 int pmp_count; /* PFS mounts backed by us */
300 TAILQ_ENTRY(hammer2_mount) mntentry; /* hammer2_mntlist */
302 struct malloc_type *minode;
306 struct malloc_type *mchain;
309 hammer2_chain_t vchain; /* anchor chain */
310 hammer2_chain_t *schain; /* super-root */
311 struct lock alloclk; /* lockmgr lock */
312 struct lock voldatalk; /* lockmgr lock */
314 hammer2_volume_data_t voldata;
315 hammer2_freecache_t freecache[HAMMER2_FREECACHE_TYPES]
316 [HAMMER2_MAX_RADIX+1];
319 typedef struct hammer2_mount hammer2_mount_t;
322 * Per-PFS mount structure for device (aka vp->v_mount)
324 struct hammer2_pfsmount {
325 struct mount *mp; /* kernel mount */
326 struct hammer2_mount *hmp; /* device global mount */
327 hammer2_chain_t *rchain; /* PFS root chain */
328 hammer2_inode_t *iroot; /* PFS root inode */
329 struct malloc_type *mmsg;
330 ccms_domain_t ccms_dom;
331 struct netexport export; /* nfs export */
332 int ronly; /* read-only mount */
333 struct file *msg_fp; /* cluster pipe->userland */
334 thread_t msgrd_td; /* cluster thread */
335 thread_t msgwr_td; /* cluster thread */
336 int msg_ctl; /* wakeup flags */
337 int msg_seq; /* cluster msg sequence id */
338 uint32_t msgid_iterator;
339 struct lock msglk; /* lockmgr lock */
340 TAILQ_HEAD(, hammer2_msg) msgq; /* transmit queue */
341 struct hammer2_state *freerd_state; /* allocation cache */
342 struct hammer2_state *freewr_state; /* allocation cache */
343 struct hammer2_state_tree staterd_tree; /* active messages */
344 struct hammer2_state_tree statewr_tree; /* active messages */
345 struct hammer2_router router;
348 typedef struct hammer2_pfsmount hammer2_pfsmount_t;
350 #define HAMMER2_CLUSTERCTL_KILL 0x0001
353 * Transactional state structure, representing an open transaction. The
354 * transaction might represent a cache state (and thus have a chain
355 * association), or a VOP op, LNK_SPAN, or other things.
357 struct hammer2_state {
358 RB_ENTRY(hammer2_state) rbnode; /* indexed by msgid */
359 struct hammer2_pfsmount *pmp;
360 struct hammer2_router *router; /* related LNK_SPAN route */
361 uint32_t txcmd; /* mostly for CMDF flags */
362 uint32_t rxcmd; /* mostly for CMDF flags */
363 uint64_t msgid; /* {spanid,msgid} uniq */
366 struct hammer2_chain *chain; /* msg associated w/chain */
367 struct hammer2_msg *msg;
368 int (*func)(struct hammer2_state *, struct hammer2_msg *);
371 hammer2_pfsmount_t *pmp;
375 #define HAMMER2_STATE_INSERTED 0x0001
376 #define HAMMER2_STATE_DYNAMIC 0x0002
379 TAILQ_ENTRY(hammer2_msg) qentry; /* serialized queue */
380 struct hammer2_router *router;
381 struct hammer2_state *state;
385 hammer2_msg_any_t any;
388 typedef struct hammer2_link hammer2_link_t;
389 typedef struct hammer2_state hammer2_state_t;
390 typedef struct hammer2_msg hammer2_msg_t;
392 int hammer2_state_cmp(hammer2_state_t *state1, hammer2_state_t *state2);
393 RB_PROTOTYPE(hammer2_state_tree, hammer2_state, rbnode, hammer2_state_cmp);
398 MALLOC_DECLARE(M_HAMMER2);
400 #define VTOI(vp) ((hammer2_inode_t *)(vp)->v_data)
401 #define ITOV(ip) ((ip)->vp)
405 MPTOPMP(struct mount *mp)
407 return ((hammer2_pfsmount_t *)mp->mnt_data);
412 MPTOHMP(struct mount *mp)
414 return (((hammer2_pfsmount_t *)mp->mnt_data)->hmp);
417 extern struct vop_ops hammer2_vnode_vops;
418 extern struct vop_ops hammer2_spec_vops;
419 extern struct vop_ops hammer2_fifo_vops;
421 extern int hammer2_debug;
422 extern int hammer2_cluster_enable;
423 extern int hammer2_hardlink_enable;
424 extern long hammer2_iod_file_read;
425 extern long hammer2_iod_meta_read;
426 extern long hammer2_iod_indr_read;
427 extern long hammer2_iod_file_write;
428 extern long hammer2_iod_meta_write;
429 extern long hammer2_iod_indr_write;
430 extern long hammer2_iod_volu_write;
431 extern long hammer2_ioa_file_read;
432 extern long hammer2_ioa_meta_read;
433 extern long hammer2_ioa_indr_read;
434 extern long hammer2_ioa_file_write;
435 extern long hammer2_ioa_meta_write;
436 extern long hammer2_ioa_indr_write;
437 extern long hammer2_ioa_volu_write;
442 void hammer2_inode_lock_ex(hammer2_inode_t *ip);
443 void hammer2_inode_unlock_ex(hammer2_inode_t *ip);
444 void hammer2_inode_lock_sh(hammer2_inode_t *ip);
445 void hammer2_inode_unlock_sh(hammer2_inode_t *ip);
446 void hammer2_inode_busy(hammer2_inode_t *ip);
447 void hammer2_inode_unbusy(hammer2_inode_t *ip);
448 void hammer2_voldata_lock(hammer2_mount_t *hmp);
449 void hammer2_voldata_unlock(hammer2_mount_t *hmp);
451 void hammer2_mount_exlock(hammer2_mount_t *hmp);
452 void hammer2_mount_shlock(hammer2_mount_t *hmp);
453 void hammer2_mount_unlock(hammer2_mount_t *hmp);
455 int hammer2_get_dtype(hammer2_inode_t *ip);
456 int hammer2_get_vtype(hammer2_inode_t *ip);
457 u_int8_t hammer2_get_obj_type(enum vtype vtype);
458 void hammer2_time_to_timespec(u_int64_t xtime, struct timespec *ts);
459 u_int64_t hammer2_timespec_to_time(struct timespec *ts);
460 u_int32_t hammer2_to_unix_xid(uuid_t *uuid);
461 void hammer2_guid_to_uuid(uuid_t *uuid, u_int32_t guid);
463 hammer2_key_t hammer2_dirhash(const unsigned char *name, size_t len);
464 int hammer2_bytes_to_radix(size_t bytes);
466 int hammer2_calc_logical(hammer2_inode_t *ip, hammer2_off_t uoff,
467 hammer2_key_t *lbasep, hammer2_key_t *leofp);
468 void hammer2_update_time(uint64_t *timep);
473 struct vnode *hammer2_igetv(hammer2_inode_t *ip, int *errorp);
475 void hammer2_inode_lock_nlinks(hammer2_inode_t *ip);
476 void hammer2_inode_unlock_nlinks(hammer2_inode_t *ip);
477 hammer2_inode_t *hammer2_inode_alloc(hammer2_pfsmount_t *pmp, void *data);
478 void hammer2_inode_free(hammer2_inode_t *ip);
479 void hammer2_inode_ref(hammer2_inode_t *ip);
480 void hammer2_inode_drop(hammer2_inode_t *ip);
481 int hammer2_inode_calc_alloc(hammer2_key_t filesize);
483 int hammer2_inode_create(hammer2_inode_t *dip,
484 struct vattr *vap, struct ucred *cred,
485 const uint8_t *name, size_t name_len,
486 hammer2_inode_t **nipp);
488 int hammer2_inode_duplicate(hammer2_inode_t *dip,
489 hammer2_inode_t *oip, hammer2_inode_t **nipp,
490 const uint8_t *name, size_t name_len);
491 int hammer2_inode_connect(hammer2_inode_t *dip, hammer2_inode_t *oip,
492 const uint8_t *name, size_t name_len);
494 int hammer2_unlink_file(hammer2_inode_t *dip,
495 const uint8_t *name, size_t name_len,
496 int isdir, hammer2_inode_t *retain_ip);
497 int hammer2_hardlink_consolidate(hammer2_inode_t **ipp, hammer2_inode_t *tdip);
498 int hammer2_hardlink_deconsolidate(hammer2_inode_t *dip,
499 hammer2_chain_t **chainp, hammer2_inode_t **ipp);
500 int hammer2_hardlink_find(hammer2_inode_t *dip, hammer2_chain_t **chainp,
501 hammer2_inode_t **ipp);
506 void hammer2_modify_volume(hammer2_mount_t *hmp);
507 hammer2_chain_t *hammer2_chain_alloc(hammer2_mount_t *hmp,
508 hammer2_blockref_t *bref);
509 void hammer2_chain_free(hammer2_mount_t *hmp, hammer2_chain_t *chain);
510 void hammer2_chain_ref(hammer2_mount_t *hmp, hammer2_chain_t *chain);
511 void hammer2_chain_drop(hammer2_mount_t *hmp, hammer2_chain_t *chain);
512 int hammer2_chain_lock(hammer2_mount_t *hmp, hammer2_chain_t *chain, int how);
513 void hammer2_chain_moved(hammer2_mount_t *hmp, hammer2_chain_t *chain);
514 void hammer2_chain_modify(hammer2_mount_t *hmp, hammer2_chain_t *chain,
516 void hammer2_chain_resize(hammer2_inode_t *ip, hammer2_chain_t *chain,
517 int nradix, int flags);
518 void hammer2_chain_unlock(hammer2_mount_t *hmp, hammer2_chain_t *chain);
519 hammer2_chain_t *hammer2_chain_find(hammer2_mount_t *hmp,
520 hammer2_chain_t *parent, int index);
521 hammer2_chain_t *hammer2_chain_get(hammer2_mount_t *hmp,
522 hammer2_chain_t *parent,
523 int index, int flags);
524 hammer2_chain_t *hammer2_chain_lookup(hammer2_mount_t *hmp,
525 hammer2_chain_t **parentp,
526 hammer2_key_t key_beg, hammer2_key_t key_end,
528 hammer2_chain_t *hammer2_chain_next(hammer2_mount_t *hmp,
529 hammer2_chain_t **parentp,
530 hammer2_chain_t *chain,
531 hammer2_key_t key_beg, hammer2_key_t key_end,
533 hammer2_chain_t *hammer2_chain_create(hammer2_mount_t *hmp,
534 hammer2_chain_t *parent,
535 hammer2_chain_t *chain,
536 hammer2_key_t key, int keybits,
537 int type, size_t bytes);
538 void hammer2_chain_delete(hammer2_mount_t *hmp, hammer2_chain_t *parent,
539 hammer2_chain_t *chain, int retain);
540 void hammer2_chain_flush(hammer2_mount_t *hmp, hammer2_chain_t *chain,
541 hammer2_tid_t modify_tid);
542 void hammer2_chain_commit(hammer2_mount_t *hmp, hammer2_chain_t *chain);
547 int hammer2_ioctl(hammer2_inode_t *ip, u_long com, void *data,
548 int fflag, struct ucred *cred);
553 int hammer2_state_msgrx(hammer2_msg_t *msg);
554 int hammer2_state_msgtx(hammer2_msg_t *msg);
555 void hammer2_state_cleanuprx(hammer2_msg_t *msg);
556 void hammer2_state_cleanuptx(hammer2_msg_t *msg);
557 int hammer2_msg_execute(hammer2_msg_t *msg);
558 void hammer2_state_free(hammer2_state_t *state);
559 void hammer2_msg_free(hammer2_msg_t *msg);
560 hammer2_msg_t *hammer2_msg_alloc(hammer2_router_t *router, uint32_t cmd);
561 void hammer2_msg_write(hammer2_msg_t *msg,
562 int (*func)(hammer2_state_t *, hammer2_msg_t *),
564 void hammer2_msg_reply(hammer2_msg_t *msg, uint32_t error);
565 void hammer2_msg_result(hammer2_msg_t *msg, uint32_t error);
570 int hammer2_msg_dbg_rcvmsg(hammer2_msg_t *msg);
571 int hammer2_msg_adhoc_input(hammer2_msg_t *msg);
576 hammer2_off_t hammer2_freemap_alloc(hammer2_mount_t *hmp,
577 int type, size_t bytes);
578 void hammer2_freemap_free(hammer2_mount_t *hmp, hammer2_off_t data_off,
581 #endif /* !_KERNEL */
582 #endif /* !_VFS_HAMMER2_HAMMER2_H_ */