xref: /dragonfly/sys/vfs/hammer/hammer_ioctl.c (revision dca3c15d)
1 /*
2  * Copyright (c) 2008 The DragonFly Project.  All rights reserved.
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  *
34  * $DragonFly: src/sys/vfs/hammer/hammer_ioctl.c,v 1.32 2008/11/13 02:23:29 dillon Exp $
35  */
36 
37 #include "hammer.h"
38 
39 static int hammer_ioc_gethistory(hammer_transaction_t trans, hammer_inode_t ip,
40 				struct hammer_ioc_history *hist);
41 static int hammer_ioc_synctid(hammer_transaction_t trans, hammer_inode_t ip,
42 				struct hammer_ioc_synctid *std);
43 static int hammer_ioc_get_version(hammer_transaction_t trans,
44 				hammer_inode_t ip,
45 				struct hammer_ioc_version *ver);
46 static int hammer_ioc_set_version(hammer_transaction_t trans,
47 				hammer_inode_t ip,
48 				struct hammer_ioc_version *ver);
49 static int hammer_ioc_get_info(hammer_transaction_t trans,
50 				struct hammer_ioc_info *info);
51 static int hammer_ioc_add_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
52 				struct hammer_ioc_snapshot *snap);
53 static int hammer_ioc_del_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
54 				struct hammer_ioc_snapshot *snap);
55 static int hammer_ioc_get_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
56 				struct hammer_ioc_snapshot *snap);
57 static int hammer_ioc_get_config(hammer_transaction_t trans, hammer_inode_t ip,
58 				struct hammer_ioc_config *snap);
59 static int hammer_ioc_set_config(hammer_transaction_t trans, hammer_inode_t ip,
60 				struct hammer_ioc_config *snap);
61 
62 int
63 hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
64 	     struct ucred *cred)
65 {
66 	struct hammer_transaction trans;
67 	int error;
68 
69 	error = priv_check_cred(cred, PRIV_HAMMER_IOCTL, 0);
70 
71 	hammer_start_transaction(&trans, ip->hmp);
72 
73 	switch(com) {
74 	case HAMMERIOC_PRUNE:
75 		if (error == 0) {
76 			error = hammer_ioc_prune(&trans, ip,
77 					(struct hammer_ioc_prune *)data);
78 		}
79 		break;
80 	case HAMMERIOC_GETHISTORY:
81 		error = hammer_ioc_gethistory(&trans, ip,
82 					(struct hammer_ioc_history *)data);
83 		break;
84 	case HAMMERIOC_REBLOCK:
85 		if (error == 0) {
86 			error = hammer_ioc_reblock(&trans, ip,
87 					(struct hammer_ioc_reblock *)data);
88 		}
89 		break;
90 	case HAMMERIOC_REBALANCE:
91 		if (error == 0) {
92 			error = hammer_ioc_rebalance(&trans, ip,
93 					(struct hammer_ioc_rebalance *)data);
94 		}
95 		break;
96 	case HAMMERIOC_SYNCTID:
97 		error = hammer_ioc_synctid(&trans, ip,
98 					(struct hammer_ioc_synctid *)data);
99 		break;
100 	case HAMMERIOC_GET_PSEUDOFS:
101 		error = hammer_ioc_get_pseudofs(&trans, ip,
102 				    (struct hammer_ioc_pseudofs_rw *)data);
103 		break;
104 	case HAMMERIOC_SET_PSEUDOFS:
105 		if (error == 0) {
106 			error = hammer_ioc_set_pseudofs(&trans, ip, cred,
107 				    (struct hammer_ioc_pseudofs_rw *)data);
108 		}
109 		break;
110 	case HAMMERIOC_UPG_PSEUDOFS:
111 		if (error == 0) {
112 			error = hammer_ioc_upgrade_pseudofs(&trans, ip,
113 				    (struct hammer_ioc_pseudofs_rw *)data);
114 		}
115 		break;
116 	case HAMMERIOC_DGD_PSEUDOFS:
117 		if (error == 0) {
118 			error = hammer_ioc_downgrade_pseudofs(&trans, ip,
119 				    (struct hammer_ioc_pseudofs_rw *)data);
120 		}
121 		break;
122 	case HAMMERIOC_RMR_PSEUDOFS:
123 		if (error == 0) {
124 			error = hammer_ioc_destroy_pseudofs(&trans, ip,
125 				    (struct hammer_ioc_pseudofs_rw *)data);
126 		}
127 		break;
128 	case HAMMERIOC_WAI_PSEUDOFS:
129 		if (error == 0) {
130 			error = hammer_ioc_wait_pseudofs(&trans, ip,
131 				    (struct hammer_ioc_pseudofs_rw *)data);
132 		}
133 		break;
134 	case HAMMERIOC_MIRROR_READ:
135 		if (error == 0) {
136 			error = hammer_ioc_mirror_read(&trans, ip,
137 				    (struct hammer_ioc_mirror_rw *)data);
138 		}
139 		break;
140 	case HAMMERIOC_MIRROR_WRITE:
141 		if (error == 0) {
142 			error = hammer_ioc_mirror_write(&trans, ip,
143 				    (struct hammer_ioc_mirror_rw *)data);
144 		}
145 		break;
146 	case HAMMERIOC_GET_VERSION:
147 		error = hammer_ioc_get_version(&trans, ip,
148 				    (struct hammer_ioc_version *)data);
149 		break;
150 	case HAMMERIOC_GET_INFO:
151 		error = hammer_ioc_get_info(&trans,
152 				    (struct hammer_ioc_info *)data);
153 		break;
154 	case HAMMERIOC_SET_VERSION:
155 		if (error == 0) {
156 			error = hammer_ioc_set_version(&trans, ip,
157 					    (struct hammer_ioc_version *)data);
158 		}
159 		break;
160 	case HAMMERIOC_EXPAND:
161 		if (error == 0) {
162 			error = priv_check_cred(cred, PRIV_HAMMER_EXPAND, 0);
163 			if (error == 0)
164 				error = hammer_ioc_expand(&trans, ip,
165 					    (struct hammer_ioc_expand *)data);
166 		}
167 		break;
168 	case HAMMERIOC_ADD_SNAPSHOT:
169 		if (error == 0) {
170 			error = hammer_ioc_add_snapshot(
171 					&trans, ip, (struct hammer_ioc_snapshot *)data);
172 		}
173 		break;
174 	case HAMMERIOC_DEL_SNAPSHOT:
175 		if (error == 0) {
176 			error = hammer_ioc_del_snapshot(
177 					&trans, ip, (struct hammer_ioc_snapshot *)data);
178 		}
179 		break;
180 	case HAMMERIOC_GET_SNAPSHOT:
181 		error = hammer_ioc_get_snapshot(
182 					&trans, ip, (struct hammer_ioc_snapshot *)data);
183 		break;
184 	case HAMMERIOC_GET_CONFIG:
185 		error = hammer_ioc_get_config(
186 					&trans, ip, (struct hammer_ioc_config *)data);
187 		break;
188 	case HAMMERIOC_SET_CONFIG:
189 		if (error == 0) {
190 			error = hammer_ioc_set_config(
191 					&trans, ip, (struct hammer_ioc_config *)data);
192 		}
193 		break;
194 	default:
195 		error = EOPNOTSUPP;
196 		break;
197 	}
198 	hammer_done_transaction(&trans);
199 	return (error);
200 }
201 
202 /*
203  * Iterate through an object's inode or an object's records and record
204  * modification TIDs.
205  */
206 static void add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
207 			hammer_btree_elm_t elm);
208 
209 static
210 int
211 hammer_ioc_gethistory(hammer_transaction_t trans, hammer_inode_t ip,
212 		      struct hammer_ioc_history *hist)
213 {
214 	struct hammer_cursor cursor;
215 	hammer_btree_elm_t elm;
216 	int error;
217 
218 	/*
219 	 * Validate the structure and initialize for return.
220 	 */
221 	if (hist->beg_tid > hist->end_tid)
222 		return(EINVAL);
223 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
224 		if (hist->key > hist->nxt_key)
225 			return(EINVAL);
226 	}
227 
228 	hist->obj_id = ip->obj_id;
229 	hist->count = 0;
230 	hist->nxt_tid = hist->end_tid;
231 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_TID;
232 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_KEY;
233 	hist->head.flags &= ~HAMMER_IOC_HISTORY_EOF;
234 	hist->head.flags &= ~HAMMER_IOC_HISTORY_UNSYNCED;
235 	if ((ip->flags & HAMMER_INODE_MODMASK) &
236 	    ~(HAMMER_INODE_ATIME | HAMMER_INODE_MTIME)) {
237 		hist->head.flags |= HAMMER_IOC_HISTORY_UNSYNCED;
238 	}
239 
240 	/*
241 	 * Setup the cursor.  We can't handle undeletable records
242 	 * (create_tid of 0) at the moment.  A create_tid of 0 has
243 	 * a special meaning and cannot be specified in the cursor.
244 	 */
245 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
246 	if (error) {
247 		hammer_done_cursor(&cursor);
248 		return(error);
249 	}
250 
251 	cursor.key_beg.obj_id = hist->obj_id;
252 	cursor.key_beg.create_tid = hist->beg_tid;
253 	cursor.key_beg.delete_tid = 0;
254 	cursor.key_beg.obj_type = 0;
255 	if (cursor.key_beg.create_tid == HAMMER_MIN_TID)
256 		cursor.key_beg.create_tid = 1;
257 
258 	cursor.key_end.obj_id = hist->obj_id;
259 	cursor.key_end.create_tid = hist->end_tid;
260 	cursor.key_end.delete_tid = 0;
261 	cursor.key_end.obj_type = 0;
262 
263 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE;
264 
265 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
266 		/*
267 		 * key-range within the file.  For a regular file the
268 		 * on-disk key represents BASE+LEN, not BASE, so the
269 		 * first possible record containing the offset 'key'
270 		 * has an on-disk key of (key + 1).
271 		 */
272 		cursor.key_beg.key = hist->key;
273 		cursor.key_end.key = HAMMER_MAX_KEY;
274 		cursor.key_beg.localization = ip->obj_localization +
275 					      HAMMER_LOCALIZE_MISC;
276 		cursor.key_end.localization = ip->obj_localization +
277 					      HAMMER_LOCALIZE_MISC;
278 
279 		switch(ip->ino_data.obj_type) {
280 		case HAMMER_OBJTYPE_REGFILE:
281 			++cursor.key_beg.key;
282 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DATA;
283 			break;
284 		case HAMMER_OBJTYPE_DIRECTORY:
285 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DIRENTRY;
286 			cursor.key_beg.localization = ip->obj_localization +
287 						hammer_dir_localization(ip);
288 			cursor.key_end.localization = ip->obj_localization +
289 						hammer_dir_localization(ip);
290 			break;
291 		case HAMMER_OBJTYPE_DBFILE:
292 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DB;
293 			break;
294 		default:
295 			error = EINVAL;
296 			break;
297 		}
298 		cursor.key_end.rec_type = cursor.key_beg.rec_type;
299 	} else {
300 		/*
301 		 * The inode itself.
302 		 */
303 		cursor.key_beg.key = 0;
304 		cursor.key_end.key = 0;
305 		cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE;
306 		cursor.key_end.rec_type = HAMMER_RECTYPE_INODE;
307 		cursor.key_beg.localization = ip->obj_localization +
308 					      HAMMER_LOCALIZE_INODE;
309 		cursor.key_end.localization = ip->obj_localization +
310 					      HAMMER_LOCALIZE_INODE;
311 	}
312 
313 	error = hammer_btree_first(&cursor);
314 	while (error == 0) {
315 		elm = &cursor.node->ondisk->elms[cursor.index];
316 
317 		add_history(ip, hist, elm);
318 		if (hist->head.flags & (HAMMER_IOC_HISTORY_NEXT_TID |
319 				        HAMMER_IOC_HISTORY_NEXT_KEY |
320 				        HAMMER_IOC_HISTORY_EOF)) {
321 			break;
322 		}
323 		error = hammer_btree_iterate(&cursor);
324 	}
325 	if (error == ENOENT) {
326 		hist->head.flags |= HAMMER_IOC_HISTORY_EOF;
327 		error = 0;
328 	}
329 	hammer_done_cursor(&cursor);
330 	return(error);
331 }
332 
333 /*
334  * Add the scanned element to the ioctl return structure.  Some special
335  * casing is required for regular files to accomodate how data ranges are
336  * stored on-disk.
337  */
338 static void
339 add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
340 	    hammer_btree_elm_t elm)
341 {
342 	int i;
343 
344 	if (elm->base.btype != HAMMER_BTREE_TYPE_RECORD)
345 		return;
346 	if ((hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) &&
347 	    ip->ino_data.obj_type == HAMMER_OBJTYPE_REGFILE) {
348 		/*
349 		 * Adjust nxt_key
350 		 */
351 		if (hist->nxt_key > elm->leaf.base.key - elm->leaf.data_len &&
352 		    hist->key < elm->leaf.base.key - elm->leaf.data_len) {
353 			hist->nxt_key = elm->leaf.base.key - elm->leaf.data_len;
354 		}
355 		if (hist->nxt_key > elm->leaf.base.key)
356 			hist->nxt_key = elm->leaf.base.key;
357 
358 		/*
359 		 * Record is beyond MAXPHYS, there won't be any more records
360 		 * in the iteration covering the requested offset (key).
361 		 */
362 		if (elm->leaf.base.key >= MAXPHYS &&
363 		    elm->leaf.base.key - MAXPHYS > hist->key) {
364 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
365 		}
366 
367 		/*
368 		 * Data-range of record does not cover the key.
369 		 */
370 		if (elm->leaf.base.key - elm->leaf.data_len > hist->key)
371 			return;
372 
373 	} else if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
374 		/*
375 		 * Adjust nxt_key
376 		 */
377 		if (hist->nxt_key > elm->leaf.base.key &&
378 		    hist->key < elm->leaf.base.key) {
379 			hist->nxt_key = elm->leaf.base.key;
380 		}
381 
382 		/*
383 		 * Record is beyond the requested key.
384 		 */
385 		if (elm->leaf.base.key > hist->key)
386 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
387 	}
388 
389 	/*
390 	 * Add create_tid if it is in-bounds.
391 	 */
392 	i = hist->count;
393 	if ((i == 0 ||
394 	     elm->leaf.base.create_tid != hist->hist_ary[i - 1].tid) &&
395 	    elm->leaf.base.create_tid >= hist->beg_tid &&
396 	    elm->leaf.base.create_tid < hist->end_tid) {
397 		if (hist->count == HAMMER_MAX_HISTORY_ELMS) {
398 			hist->nxt_tid = elm->leaf.base.create_tid;
399 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
400 			return;
401 		}
402 		hist->hist_ary[i].tid = elm->leaf.base.create_tid;
403 		hist->hist_ary[i].time32 = elm->leaf.create_ts;
404 		++hist->count;
405 	}
406 
407 	/*
408 	 * Add delete_tid if it is in-bounds.  Note that different portions
409 	 * of the history may have overlapping data ranges with different
410 	 * delete_tid's.  If this case occurs the delete_tid may match the
411 	 * create_tid of a following record.  XXX
412 	 *
413 	 *	[        ]
414 	 *            [     ]
415 	 */
416 	i = hist->count;
417 	if (elm->leaf.base.delete_tid &&
418 	    elm->leaf.base.delete_tid >= hist->beg_tid &&
419 	    elm->leaf.base.delete_tid < hist->end_tid) {
420 		if (i == HAMMER_MAX_HISTORY_ELMS) {
421 			hist->nxt_tid = elm->leaf.base.delete_tid;
422 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
423 			return;
424 		}
425 		hist->hist_ary[i].tid = elm->leaf.base.delete_tid;
426 		hist->hist_ary[i].time32 = elm->leaf.delete_ts;
427 		++hist->count;
428 	}
429 }
430 
431 /*
432  * Acquire synchronization TID
433  */
434 static
435 int
436 hammer_ioc_synctid(hammer_transaction_t trans, hammer_inode_t ip,
437 		   struct hammer_ioc_synctid *std)
438 {
439 	hammer_mount_t hmp = ip->hmp;
440 	int error = 0;
441 
442 	switch(std->op) {
443 	case HAMMER_SYNCTID_NONE:
444 		std->tid = hmp->flusher.tid;	/* inaccurate */
445 		break;
446 	case HAMMER_SYNCTID_ASYNC:
447 		hammer_queue_inodes_flusher(hmp, MNT_NOWAIT);
448 		hammer_flusher_async(hmp, NULL);
449 		std->tid = hmp->flusher.tid;	/* inaccurate */
450 		break;
451 	case HAMMER_SYNCTID_SYNC1:
452 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
453 		hammer_flusher_sync(hmp);
454 		std->tid = hmp->flusher.tid;
455 		break;
456 	case HAMMER_SYNCTID_SYNC2:
457 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
458 		hammer_flusher_sync(hmp);
459 		std->tid = hmp->flusher.tid;
460 		hammer_flusher_sync(hmp);
461 		break;
462 	default:
463 		error = EOPNOTSUPP;
464 		break;
465 	}
466 	return(error);
467 }
468 
469 /*
470  * Retrieve version info.
471  *
472  * Load min_version, wip_version, and max_versino.  If cur_version is passed
473  * as 0 then load the current version into cur_version.  Load the description
474  * for cur_version into the description array.
475  *
476  * Returns 0 on success, EINVAL if cur_version is non-zero and set to an
477  * unsupported value.
478  */
479 static
480 int
481 hammer_ioc_get_version(hammer_transaction_t trans, hammer_inode_t ip,
482 		   struct hammer_ioc_version *ver)
483 {
484 	int error = 0;
485 
486 	ver->min_version = HAMMER_VOL_VERSION_MIN;
487 	ver->wip_version = HAMMER_VOL_VERSION_WIP;
488 	ver->max_version = HAMMER_VOL_VERSION_MAX;
489 	if (ver->cur_version == 0)
490 		ver->cur_version = trans->hmp->version;
491 	switch(ver->cur_version) {
492 	case 1:
493 		ksnprintf(ver->description, sizeof(ver->description),
494 			 "First HAMMER release (DragonFly 2.0+)");
495 		break;
496 	case 2:
497 		ksnprintf(ver->description, sizeof(ver->description),
498 			 "New directory entry layout (DragonFly 2.3+)");
499 		break;
500 	case 3:
501 		ksnprintf(ver->description, sizeof(ver->description),
502 			 "New snapshot management (DragonFly 2.5+)");
503 		break;
504 	default:
505 		ksnprintf(ver->description, sizeof(ver->description),
506 			 "Unknown");
507 		error = EINVAL;
508 		break;
509 	}
510 	return(error);
511 };
512 
513 /*
514  * Set version info
515  */
516 static
517 int
518 hammer_ioc_set_version(hammer_transaction_t trans, hammer_inode_t ip,
519 		   struct hammer_ioc_version *ver)
520 {
521 	struct hammer_cursor cursor;
522 	hammer_volume_t volume;
523 	int error;
524 
525 	if (ver->cur_version < trans->hmp->version)
526 		return(EINVAL);
527 	if (ver->cur_version == trans->hmp->version)
528 		return(0);
529 	if (ver->cur_version > HAMMER_VOL_VERSION_MAX)
530 		return(EINVAL);
531 	if (trans->hmp->ronly)
532 		return(EROFS);
533 
534 	/*
535 	 * Update the root volume header and the version cached in
536 	 * the hammer_mount structure.
537 	 */
538 	error = hammer_init_cursor(trans, &cursor, NULL, NULL);
539 	if (error)
540 		goto failed;
541 	hammer_sync_lock_sh(trans);
542 
543 	volume = hammer_get_root_volume(cursor.trans->hmp, &error);
544 	KKASSERT(error == 0);
545 	hammer_modify_volume_field(cursor.trans, volume, vol_version);
546 	volume->ondisk->vol_version = ver->cur_version;
547 	cursor.trans->hmp->version = ver->cur_version;
548 	hammer_modify_volume_done(volume);
549 	hammer_rel_volume(volume, 0);
550 
551 	hammer_sync_unlock(trans);
552 failed:
553 	ver->head.error = error;
554 	hammer_done_cursor(&cursor);
555 	return(0);
556 }
557 
558 /*
559  * Get information
560  */
561 static
562 int
563 hammer_ioc_get_info(hammer_transaction_t trans, struct hammer_ioc_info *info) {
564 
565 	struct hammer_volume_ondisk	*od = trans->hmp->rootvol->ondisk;
566 	struct hammer_mount 		*hm = trans->hmp;
567 
568 	/* Fill the structure with the necessary information */
569 	_hammer_checkspace(hm, HAMMER_CHKSPC_WRITE, &info->rsvbigblocks);
570 	info->rsvbigblocks = info->rsvbigblocks >> HAMMER_LARGEBLOCK_BITS;
571 	strlcpy(info->vol_name, od->vol_name, sizeof(od->vol_name));
572 
573 	info->vol_fsid = hm->fsid;
574 	info->vol_fstype = od->vol_fstype;
575 	info->version = hm->version;
576 
577 	info->inodes = od->vol0_stat_inodes;
578 	info->bigblocks = od->vol0_stat_bigblocks;
579 	info->freebigblocks = od->vol0_stat_freebigblocks;
580 	info->nvolumes = hm->nvolumes;
581 
582 	return 0;
583 }
584 
585 /*
586  * Add a snapshot transction id(s) to the list of snapshots.
587  *
588  * NOTE: Records are created with an allocated TID.  If a flush cycle
589  *	 is in progress the record may be synced in the current flush
590  *	 cycle and the volume header will reflect the allocation of the
591  *	 TID, but the synchronization point may not catch up to the
592  *	 TID until the next flush cycle.
593  */
594 static
595 int
596 hammer_ioc_add_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
597 			struct hammer_ioc_snapshot *snap)
598 {
599 	hammer_mount_t hmp = ip->hmp;
600 	struct hammer_btree_leaf_elm leaf;
601 	struct hammer_cursor cursor;
602 	int error;
603 
604 	/*
605 	 * Validate structure
606 	 */
607 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
608 		return (EINVAL);
609 	if (snap->index > snap->count)
610 		return (EINVAL);
611 
612 	hammer_lock_ex(&hmp->snapshot_lock);
613 again:
614 	/*
615 	 * Look for keys starting after the previous iteration, or at
616 	 * the beginning if snap->count is 0.
617 	 */
618 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
619 	if (error) {
620 		hammer_done_cursor(&cursor);
621 		return(error);
622 	}
623 
624 	cursor.asof = HAMMER_MAX_TID;
625 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
626 
627 	bzero(&leaf, sizeof(leaf));
628 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
629 	leaf.base.rec_type = HAMMER_RECTYPE_SNAPSHOT;
630 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
631 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
632 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
633 	leaf.data_len = sizeof(struct hammer_snapshot_data);
634 
635 	while (snap->index < snap->count) {
636 		leaf.base.key = (int64_t)snap->snaps[snap->index].tid;
637 		cursor.key_beg = leaf.base;
638 		error = hammer_btree_lookup(&cursor);
639 		if (error == 0) {
640 			error = EEXIST;
641 			break;
642 		}
643 
644 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
645 		error = hammer_create_at_cursor(&cursor, &leaf,
646 						&snap->snaps[snap->index],
647 						HAMMER_CREATE_MODE_SYS);
648 		if (error == EDEADLK) {
649 			hammer_done_cursor(&cursor);
650 			goto again;
651 		}
652 		cursor.flags |= HAMMER_CURSOR_ASOF;
653 		if (error)
654 			break;
655 		++snap->index;
656 	}
657 	snap->head.error = error;
658 	hammer_done_cursor(&cursor);
659 	hammer_unlock(&hmp->snapshot_lock);
660 	return(0);
661 }
662 
663 /*
664  * Delete snapshot transaction id(s) from the list of snapshots.
665  */
666 static
667 int
668 hammer_ioc_del_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
669 			struct hammer_ioc_snapshot *snap)
670 {
671 	hammer_mount_t hmp = ip->hmp;
672 	struct hammer_cursor cursor;
673 	int error;
674 
675 	/*
676 	 * Validate structure
677 	 */
678 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
679 		return (EINVAL);
680 	if (snap->index > snap->count)
681 		return (EINVAL);
682 
683 	hammer_lock_ex(&hmp->snapshot_lock);
684 again:
685 	/*
686 	 * Look for keys starting after the previous iteration, or at
687 	 * the beginning if snap->count is 0.
688 	 */
689 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
690 	if (error) {
691 		hammer_done_cursor(&cursor);
692 		return(error);
693 	}
694 
695 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
696 	cursor.key_beg.create_tid = 0;
697 	cursor.key_beg.delete_tid = 0;
698 	cursor.key_beg.obj_type = 0;
699 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
700 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
701 	cursor.asof = HAMMER_MAX_TID;
702 	cursor.flags |= HAMMER_CURSOR_ASOF;
703 
704 	while (snap->index < snap->count) {
705 		cursor.key_beg.key = (int64_t)snap->snaps[snap->index].tid;
706 		error = hammer_btree_lookup(&cursor);
707 		if (error)
708 			break;
709 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
710 		if (error)
711 			break;
712 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
713 						0, 0, 0, NULL);
714 		if (error == EDEADLK) {
715 			hammer_done_cursor(&cursor);
716 			goto again;
717 		}
718 		if (error)
719 			break;
720 		++snap->index;
721 	}
722 	snap->head.error = error;
723 	hammer_done_cursor(&cursor);
724 	hammer_unlock(&hmp->snapshot_lock);
725 	return(0);
726 }
727 
728 /*
729  * Retrieve as many snapshot ids as possible or until the array is
730  * full, starting after the last transction id passed in.  If count
731  * is 0 we retrieve starting at the beginning.
732  *
733  * NOTE: Because the b-tree key field is signed but transaction ids
734  *       are unsigned the returned list will be signed-sorted instead
735  *	 of unsigned sorted.  The Caller must still sort the aggregate
736  *	 results.
737  */
738 static
739 int
740 hammer_ioc_get_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
741 			struct hammer_ioc_snapshot *snap)
742 {
743 	struct hammer_cursor cursor;
744 	int error;
745 
746 	/*
747 	 * Validate structure
748 	 */
749 	if (snap->index != 0)
750 		return (EINVAL);
751 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
752 		return (EINVAL);
753 
754 	/*
755 	 * Look for keys starting after the previous iteration, or at
756 	 * the beginning if snap->count is 0.
757 	 */
758 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
759 	if (error) {
760 		hammer_done_cursor(&cursor);
761 		return(error);
762 	}
763 
764 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
765 	cursor.key_beg.create_tid = 0;
766 	cursor.key_beg.delete_tid = 0;
767 	cursor.key_beg.obj_type = 0;
768 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
769 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
770 	if (snap->count == 0)
771 		cursor.key_beg.key = HAMMER_MIN_KEY;
772 	else
773 		cursor.key_beg.key = (int64_t)snap->snaps[snap->count - 1].tid + 1;
774 
775 	cursor.key_end = cursor.key_beg;
776 	cursor.key_end.key = HAMMER_MAX_KEY;
777 	cursor.asof = HAMMER_MAX_TID;
778 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE | HAMMER_CURSOR_ASOF;
779 
780 	snap->count = 0;
781 
782 	error = hammer_btree_first(&cursor);
783 	while (error == 0 && snap->count < HAMMER_SNAPS_PER_IOCTL) {
784 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
785 		if (error)
786 			break;
787 		if (cursor.leaf->base.rec_type == HAMMER_RECTYPE_SNAPSHOT) {
788 			error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
789 							      HAMMER_CURSOR_GET_DATA);
790 			snap->snaps[snap->count] = cursor.data->snap;
791 			++snap->count;
792 		}
793 		error = hammer_btree_iterate(&cursor);
794 	}
795 
796 	if (error == ENOENT) {
797 		snap->head.flags |= HAMMER_IOC_SNAPSHOT_EOF;
798 		error = 0;
799 	}
800 	snap->head.error = error;
801 	hammer_done_cursor(&cursor);
802 	return(0);
803 }
804 
805 /*
806  * Retrieve the PFS hammer cleanup utility config record.  This is
807  * different (newer than) the PFS config.
808  */
809 static
810 int
811 hammer_ioc_get_config(hammer_transaction_t trans, hammer_inode_t ip,
812 			struct hammer_ioc_config *config)
813 {
814 	struct hammer_cursor cursor;
815 	int error;
816 
817 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
818 	if (error) {
819 		hammer_done_cursor(&cursor);
820 		return(error);
821 	}
822 
823 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
824 	cursor.key_beg.create_tid = 0;
825 	cursor.key_beg.delete_tid = 0;
826 	cursor.key_beg.obj_type = 0;
827 	cursor.key_beg.rec_type = HAMMER_RECTYPE_CONFIG;
828 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
829 	cursor.key_beg.key = 0;		/* config space page 0 */
830 
831 	cursor.asof = HAMMER_MAX_TID;
832 	cursor.flags |= HAMMER_CURSOR_ASOF;
833 
834 	error = hammer_btree_lookup(&cursor);
835 	if (error == 0) {
836 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
837 						      HAMMER_CURSOR_GET_DATA);
838 		if (error == 0)
839 			config->config = cursor.data->config;
840 	}
841 	/* error can be ENOENT */
842 	config->head.error = error;
843 	hammer_done_cursor(&cursor);
844 	return(0);
845 }
846 
847 /*
848  * Retrieve the PFS hammer cleanup utility config record.  This is
849  * different (newer than) the PFS config.
850  *
851  * This is kinda a hack.
852  */
853 static
854 int
855 hammer_ioc_set_config(hammer_transaction_t trans, hammer_inode_t ip,
856 			struct hammer_ioc_config *config)
857 {
858 	struct hammer_btree_leaf_elm leaf;
859 	struct hammer_cursor cursor;
860 	hammer_mount_t hmp = ip->hmp;
861 	int error;
862 
863 again:
864 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
865 	if (error) {
866 		hammer_done_cursor(&cursor);
867 		return(error);
868 	}
869 
870 	bzero(&leaf, sizeof(leaf));
871 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
872 	leaf.base.rec_type = HAMMER_RECTYPE_CONFIG;
873 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
874 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
875 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
876 	leaf.base.key = 0;	/* page 0 */
877 	leaf.data_len = sizeof(struct hammer_config_data);
878 
879 	cursor.key_beg = leaf.base;
880 
881 	cursor.asof = HAMMER_MAX_TID;
882 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
883 
884 	error = hammer_btree_lookup(&cursor);
885 	if (error == 0) {
886 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
887 						      HAMMER_CURSOR_GET_DATA);
888 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
889 						0, 0, 0, NULL);
890 		if (error == EDEADLK) {
891 			hammer_done_cursor(&cursor);
892 			goto again;
893 		}
894 	}
895 	if (error == ENOENT)
896 		error = 0;
897 	if (error == 0) {
898 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
899 		cursor.key_beg = leaf.base;
900 		error = hammer_create_at_cursor(&cursor, &leaf,
901 						&config->config,
902 						HAMMER_CREATE_MODE_SYS);
903 		if (error == EDEADLK) {
904 			hammer_done_cursor(&cursor);
905 			goto again;
906 		}
907 	}
908 	config->head.error = error;
909 	hammer_done_cursor(&cursor);
910 	return(0);
911 }
912