xref: /dragonfly/sys/vfs/hammer/hammer_ioctl.c (revision 0dace59e)
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_pfs_iterate(hammer_transaction_t trans,
52 				struct hammer_ioc_pfs_iterate *pi);
53 static int hammer_ioc_add_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
54 				struct hammer_ioc_snapshot *snap);
55 static int hammer_ioc_del_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
56 				struct hammer_ioc_snapshot *snap);
57 static int hammer_ioc_get_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
58 				struct hammer_ioc_snapshot *snap);
59 static int hammer_ioc_get_config(hammer_transaction_t trans, hammer_inode_t ip,
60 				struct hammer_ioc_config *snap);
61 static int hammer_ioc_set_config(hammer_transaction_t trans, hammer_inode_t ip,
62 				struct hammer_ioc_config *snap);
63 static int hammer_ioc_get_data(hammer_transaction_t trans, hammer_inode_t ip,
64 				struct hammer_ioc_data *data);
65 
66 int
67 hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
68 	     struct ucred *cred)
69 {
70 	struct hammer_transaction trans;
71 	int error;
72 
73 	error = priv_check_cred(cred, PRIV_HAMMER_IOCTL, 0);
74 
75 	hammer_start_transaction(&trans, ip->hmp);
76 
77 	switch(com) {
78 	case HAMMERIOC_PRUNE:
79 		if (error == 0) {
80 			error = hammer_ioc_prune(&trans, ip,
81 					(struct hammer_ioc_prune *)data);
82 		}
83 		break;
84 	case HAMMERIOC_GETHISTORY:
85 		error = hammer_ioc_gethistory(&trans, ip,
86 					(struct hammer_ioc_history *)data);
87 		break;
88 	case HAMMERIOC_REBLOCK:
89 		if (error == 0) {
90 			error = hammer_ioc_reblock(&trans, ip,
91 					(struct hammer_ioc_reblock *)data);
92 		}
93 		break;
94 	case HAMMERIOC_REBALANCE:
95 		/*
96 		 * Rebalancing needs to lock a lot of B-Tree nodes.  The
97 		 * children and children's children.  Systems with very
98 		 * little memory will not be able to do it.
99 		 */
100 		if (error == 0 && nbuf < HAMMER_REBALANCE_MIN_BUFS) {
101 			kprintf("hammer: System has insufficient buffers "
102 				"to rebalance the tree.  nbuf < %d\n",
103 				HAMMER_REBALANCE_MIN_BUFS);
104 			error = ENOSPC;
105 		}
106 		if (error == 0) {
107 			error = hammer_ioc_rebalance(&trans, ip,
108 					(struct hammer_ioc_rebalance *)data);
109 		}
110 		break;
111 	case HAMMERIOC_SYNCTID:
112 		error = hammer_ioc_synctid(&trans, ip,
113 					(struct hammer_ioc_synctid *)data);
114 		break;
115 	case HAMMERIOC_GET_PSEUDOFS:
116 		error = hammer_ioc_get_pseudofs(&trans, ip,
117 				    (struct hammer_ioc_pseudofs_rw *)data);
118 		break;
119 	case HAMMERIOC_SET_PSEUDOFS:
120 		if (error == 0) {
121 			error = hammer_ioc_set_pseudofs(&trans, ip, cred,
122 				    (struct hammer_ioc_pseudofs_rw *)data);
123 		}
124 		break;
125 	case HAMMERIOC_UPG_PSEUDOFS:
126 		if (error == 0) {
127 			error = hammer_ioc_upgrade_pseudofs(&trans, ip,
128 				    (struct hammer_ioc_pseudofs_rw *)data);
129 		}
130 		break;
131 	case HAMMERIOC_DGD_PSEUDOFS:
132 		if (error == 0) {
133 			error = hammer_ioc_downgrade_pseudofs(&trans, ip,
134 				    (struct hammer_ioc_pseudofs_rw *)data);
135 		}
136 		break;
137 	case HAMMERIOC_RMR_PSEUDOFS:
138 		if (error == 0) {
139 			error = hammer_ioc_destroy_pseudofs(&trans, ip,
140 				    (struct hammer_ioc_pseudofs_rw *)data);
141 		}
142 		break;
143 	case HAMMERIOC_WAI_PSEUDOFS:
144 		if (error == 0) {
145 			error = hammer_ioc_wait_pseudofs(&trans, ip,
146 				    (struct hammer_ioc_pseudofs_rw *)data);
147 		}
148 		break;
149 	case HAMMERIOC_MIRROR_READ:
150 		if (error == 0) {
151 			error = hammer_ioc_mirror_read(&trans, ip,
152 				    (struct hammer_ioc_mirror_rw *)data);
153 		}
154 		break;
155 	case HAMMERIOC_MIRROR_WRITE:
156 		if (error == 0) {
157 			error = hammer_ioc_mirror_write(&trans, ip,
158 				    (struct hammer_ioc_mirror_rw *)data);
159 		}
160 		break;
161 	case HAMMERIOC_GET_VERSION:
162 		error = hammer_ioc_get_version(&trans, ip,
163 				    (struct hammer_ioc_version *)data);
164 		break;
165 	case HAMMERIOC_GET_INFO:
166 		error = hammer_ioc_get_info(&trans,
167 				    (struct hammer_ioc_info *)data);
168 		break;
169 	case HAMMERIOC_SET_VERSION:
170 		if (error == 0) {
171 			error = hammer_ioc_set_version(&trans, ip,
172 					    (struct hammer_ioc_version *)data);
173 		}
174 		break;
175 	case HAMMERIOC_ADD_VOLUME:
176 		if (error == 0) {
177 			error = priv_check_cred(cred, PRIV_HAMMER_VOLUME, 0);
178 			if (error == 0)
179 				error = hammer_ioc_volume_add(&trans, ip,
180 					    (struct hammer_ioc_volume *)data);
181 		}
182 		break;
183 	case HAMMERIOC_DEL_VOLUME:
184 		if (error == 0) {
185 			error = priv_check_cred(cred, PRIV_HAMMER_VOLUME, 0);
186 			if (error == 0)
187 				error = hammer_ioc_volume_del(&trans, ip,
188 					    (struct hammer_ioc_volume *)data);
189 		}
190 		break;
191 	case HAMMERIOC_LIST_VOLUMES:
192 		error = hammer_ioc_volume_list(&trans, ip,
193 		    (struct hammer_ioc_volume_list *)data);
194 		break;
195 	case HAMMERIOC_ADD_SNAPSHOT:
196 		if (error == 0) {
197 			error = hammer_ioc_add_snapshot(
198 					&trans, ip, (struct hammer_ioc_snapshot *)data);
199 		}
200 		break;
201 	case HAMMERIOC_DEL_SNAPSHOT:
202 		if (error == 0) {
203 			error = hammer_ioc_del_snapshot(
204 					&trans, ip, (struct hammer_ioc_snapshot *)data);
205 		}
206 		break;
207 	case HAMMERIOC_GET_SNAPSHOT:
208 		error = hammer_ioc_get_snapshot(
209 					&trans, ip, (struct hammer_ioc_snapshot *)data);
210 		break;
211 	case HAMMERIOC_GET_CONFIG:
212 		error = hammer_ioc_get_config(
213 					&trans, ip, (struct hammer_ioc_config *)data);
214 		break;
215 	case HAMMERIOC_SET_CONFIG:
216 		if (error == 0) {
217 			error = hammer_ioc_set_config(
218 					&trans, ip, (struct hammer_ioc_config *)data);
219 		}
220 		break;
221 	case HAMMERIOC_DEDUP:
222 		if (error == 0) {
223 			error = hammer_ioc_dedup(
224 					&trans, ip, (struct hammer_ioc_dedup *)data);
225 		}
226 		break;
227 	case HAMMERIOC_GET_DATA:
228 		if (error == 0) {
229 			error = hammer_ioc_get_data(
230 					&trans, ip, (struct hammer_ioc_data *)data);
231 		}
232 		break;
233 	case HAMMERIOC_PFS_ITERATE:
234 		if (error == 0) {
235 			error = hammer_ioc_pfs_iterate(
236 				&trans, (struct hammer_ioc_pfs_iterate *)data);
237 		}
238 		break;
239 	default:
240 		error = EOPNOTSUPP;
241 		break;
242 	}
243 	hammer_done_transaction(&trans);
244 	return (error);
245 }
246 
247 /*
248  * Iterate through an object's inode or an object's records and record
249  * modification TIDs.
250  */
251 static void add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
252 			hammer_btree_elm_t elm);
253 
254 static
255 int
256 hammer_ioc_gethistory(hammer_transaction_t trans, hammer_inode_t ip,
257 		      struct hammer_ioc_history *hist)
258 {
259 	struct hammer_cursor cursor;
260 	hammer_btree_elm_t elm;
261 	int error;
262 
263 	/*
264 	 * Validate the structure and initialize for return.
265 	 */
266 	if (hist->beg_tid > hist->end_tid)
267 		return(EINVAL);
268 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
269 		if (hist->key > hist->nxt_key)
270 			return(EINVAL);
271 	}
272 
273 	hist->obj_id = ip->obj_id;
274 	hist->count = 0;
275 	hist->nxt_tid = hist->end_tid;
276 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_TID;
277 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_KEY;
278 	hist->head.flags &= ~HAMMER_IOC_HISTORY_EOF;
279 	hist->head.flags &= ~HAMMER_IOC_HISTORY_UNSYNCED;
280 	if ((ip->flags & HAMMER_INODE_MODMASK) &
281 	    ~(HAMMER_INODE_ATIME | HAMMER_INODE_MTIME)) {
282 		hist->head.flags |= HAMMER_IOC_HISTORY_UNSYNCED;
283 	}
284 
285 	/*
286 	 * Setup the cursor.  We can't handle undeletable records
287 	 * (create_tid of 0) at the moment.  A create_tid of 0 has
288 	 * a special meaning and cannot be specified in the cursor.
289 	 */
290 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
291 	if (error) {
292 		hammer_done_cursor(&cursor);
293 		return(error);
294 	}
295 
296 	cursor.key_beg.obj_id = hist->obj_id;
297 	cursor.key_beg.create_tid = hist->beg_tid;
298 	cursor.key_beg.delete_tid = 0;
299 	cursor.key_beg.obj_type = 0;
300 	if (cursor.key_beg.create_tid == HAMMER_MIN_TID)
301 		cursor.key_beg.create_tid = 1;
302 
303 	cursor.key_end.obj_id = hist->obj_id;
304 	cursor.key_end.create_tid = hist->end_tid;
305 	cursor.key_end.delete_tid = 0;
306 	cursor.key_end.obj_type = 0;
307 
308 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE;
309 
310 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
311 		/*
312 		 * key-range within the file.  For a regular file the
313 		 * on-disk key represents BASE+LEN, not BASE, so the
314 		 * first possible record containing the offset 'key'
315 		 * has an on-disk key of (key + 1).
316 		 */
317 		cursor.key_beg.key = hist->key;
318 		cursor.key_end.key = HAMMER_MAX_KEY;
319 		cursor.key_beg.localization = ip->obj_localization +
320 					      HAMMER_LOCALIZE_MISC;
321 		cursor.key_end.localization = ip->obj_localization +
322 					      HAMMER_LOCALIZE_MISC;
323 
324 		switch(ip->ino_data.obj_type) {
325 		case HAMMER_OBJTYPE_REGFILE:
326 			++cursor.key_beg.key;
327 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DATA;
328 			break;
329 		case HAMMER_OBJTYPE_DIRECTORY:
330 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DIRENTRY;
331 			cursor.key_beg.localization = ip->obj_localization +
332 						hammer_dir_localization(ip);
333 			cursor.key_end.localization = ip->obj_localization +
334 						hammer_dir_localization(ip);
335 			break;
336 		case HAMMER_OBJTYPE_DBFILE:
337 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DB;
338 			break;
339 		default:
340 			error = EINVAL;
341 			break;
342 		}
343 		cursor.key_end.rec_type = cursor.key_beg.rec_type;
344 	} else {
345 		/*
346 		 * The inode itself.
347 		 */
348 		cursor.key_beg.key = 0;
349 		cursor.key_end.key = 0;
350 		cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE;
351 		cursor.key_end.rec_type = HAMMER_RECTYPE_INODE;
352 		cursor.key_beg.localization = ip->obj_localization +
353 					      HAMMER_LOCALIZE_INODE;
354 		cursor.key_end.localization = ip->obj_localization +
355 					      HAMMER_LOCALIZE_INODE;
356 	}
357 
358 	error = hammer_btree_first(&cursor);
359 	while (error == 0) {
360 		elm = &cursor.node->ondisk->elms[cursor.index];
361 
362 		add_history(ip, hist, elm);
363 		if (hist->head.flags & (HAMMER_IOC_HISTORY_NEXT_TID |
364 				        HAMMER_IOC_HISTORY_NEXT_KEY |
365 				        HAMMER_IOC_HISTORY_EOF)) {
366 			break;
367 		}
368 		error = hammer_btree_iterate(&cursor);
369 	}
370 	if (error == ENOENT) {
371 		hist->head.flags |= HAMMER_IOC_HISTORY_EOF;
372 		error = 0;
373 	}
374 	hammer_done_cursor(&cursor);
375 	return(error);
376 }
377 
378 /*
379  * Add the scanned element to the ioctl return structure.  Some special
380  * casing is required for regular files to accomodate how data ranges are
381  * stored on-disk.
382  */
383 static void
384 add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
385 	    hammer_btree_elm_t elm)
386 {
387 	int i;
388 
389 	if (elm->base.btype != HAMMER_BTREE_TYPE_RECORD)
390 		return;
391 	if ((hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) &&
392 	    ip->ino_data.obj_type == HAMMER_OBJTYPE_REGFILE) {
393 		/*
394 		 * Adjust nxt_key
395 		 */
396 		if (hist->nxt_key > elm->leaf.base.key - elm->leaf.data_len &&
397 		    hist->key < elm->leaf.base.key - elm->leaf.data_len) {
398 			hist->nxt_key = elm->leaf.base.key - elm->leaf.data_len;
399 		}
400 		if (hist->nxt_key > elm->leaf.base.key)
401 			hist->nxt_key = elm->leaf.base.key;
402 
403 		/*
404 		 * Record is beyond MAXPHYS, there won't be any more records
405 		 * in the iteration covering the requested offset (key).
406 		 */
407 		if (elm->leaf.base.key >= MAXPHYS &&
408 		    elm->leaf.base.key - MAXPHYS > hist->key) {
409 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
410 		}
411 
412 		/*
413 		 * Data-range of record does not cover the key.
414 		 */
415 		if (elm->leaf.base.key - elm->leaf.data_len > hist->key)
416 			return;
417 
418 	} else if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
419 		/*
420 		 * Adjust nxt_key
421 		 */
422 		if (hist->nxt_key > elm->leaf.base.key &&
423 		    hist->key < elm->leaf.base.key) {
424 			hist->nxt_key = elm->leaf.base.key;
425 		}
426 
427 		/*
428 		 * Record is beyond the requested key.
429 		 */
430 		if (elm->leaf.base.key > hist->key)
431 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
432 	}
433 
434 	/*
435 	 * Add create_tid if it is in-bounds.
436 	 */
437 	i = hist->count;
438 	if ((i == 0 ||
439 	     elm->leaf.base.create_tid != hist->hist_ary[i - 1].tid) &&
440 	    elm->leaf.base.create_tid >= hist->beg_tid &&
441 	    elm->leaf.base.create_tid < hist->end_tid) {
442 		if (hist->count == HAMMER_MAX_HISTORY_ELMS) {
443 			hist->nxt_tid = elm->leaf.base.create_tid;
444 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
445 			return;
446 		}
447 		hist->hist_ary[i].tid = elm->leaf.base.create_tid;
448 		hist->hist_ary[i].time32 = elm->leaf.create_ts;
449 		++hist->count;
450 	}
451 
452 	/*
453 	 * Add delete_tid if it is in-bounds.  Note that different portions
454 	 * of the history may have overlapping data ranges with different
455 	 * delete_tid's.  If this case occurs the delete_tid may match the
456 	 * create_tid of a following record.  XXX
457 	 *
458 	 *	[        ]
459 	 *            [     ]
460 	 */
461 	i = hist->count;
462 	if (elm->leaf.base.delete_tid &&
463 	    elm->leaf.base.delete_tid >= hist->beg_tid &&
464 	    elm->leaf.base.delete_tid < hist->end_tid) {
465 		if (i == HAMMER_MAX_HISTORY_ELMS) {
466 			hist->nxt_tid = elm->leaf.base.delete_tid;
467 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
468 			return;
469 		}
470 		hist->hist_ary[i].tid = elm->leaf.base.delete_tid;
471 		hist->hist_ary[i].time32 = elm->leaf.delete_ts;
472 		++hist->count;
473 	}
474 }
475 
476 /*
477  * Acquire synchronization TID
478  */
479 static
480 int
481 hammer_ioc_synctid(hammer_transaction_t trans, hammer_inode_t ip,
482 		   struct hammer_ioc_synctid *std)
483 {
484 	hammer_mount_t hmp = ip->hmp;
485 	int error = 0;
486 
487 	switch(std->op) {
488 	case HAMMER_SYNCTID_NONE:
489 		std->tid = hmp->flusher.tid;	/* inaccurate */
490 		break;
491 	case HAMMER_SYNCTID_ASYNC:
492 		hammer_queue_inodes_flusher(hmp, MNT_NOWAIT);
493 		hammer_flusher_async(hmp, NULL);
494 		std->tid = hmp->flusher.tid;	/* inaccurate */
495 		break;
496 	case HAMMER_SYNCTID_SYNC1:
497 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
498 		hammer_flusher_sync(hmp);
499 		std->tid = hmp->flusher.tid;
500 		break;
501 	case HAMMER_SYNCTID_SYNC2:
502 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
503 		hammer_flusher_sync(hmp);
504 		std->tid = hmp->flusher.tid;
505 		hammer_flusher_sync(hmp);
506 		break;
507 	default:
508 		error = EOPNOTSUPP;
509 		break;
510 	}
511 	return(error);
512 }
513 
514 /*
515  * Retrieve version info.
516  *
517  * Load min_version, wip_version, and max_versino.  If cur_version is passed
518  * as 0 then load the current version into cur_version.  Load the description
519  * for cur_version into the description array.
520  *
521  * Returns 0 on success, EINVAL if cur_version is non-zero and set to an
522  * unsupported value.
523  */
524 static
525 int
526 hammer_ioc_get_version(hammer_transaction_t trans, hammer_inode_t ip,
527 		   struct hammer_ioc_version *ver)
528 {
529 	int error = 0;
530 
531 	ver->min_version = HAMMER_VOL_VERSION_MIN;
532 	ver->wip_version = HAMMER_VOL_VERSION_WIP;
533 	ver->max_version = HAMMER_VOL_VERSION_MAX;
534 	if (ver->cur_version == 0)
535 		ver->cur_version = trans->hmp->version;
536 	switch(ver->cur_version) {
537 	case 1:
538 		ksnprintf(ver->description, sizeof(ver->description),
539 			 "First HAMMER release (DragonFly 2.0+)");
540 		break;
541 	case 2:
542 		ksnprintf(ver->description, sizeof(ver->description),
543 			 "New directory entry layout (DragonFly 2.3+)");
544 		break;
545 	case 3:
546 		ksnprintf(ver->description, sizeof(ver->description),
547 			 "New snapshot management (DragonFly 2.5+)");
548 		break;
549 	case 4:
550 		ksnprintf(ver->description, sizeof(ver->description),
551 			 "New undo/flush, faster flush/sync (DragonFly 2.5+)");
552 		break;
553 	case 5:
554 		ksnprintf(ver->description, sizeof(ver->description),
555 			 "Adjustments for dedup support (DragonFly 2.9+)");
556 		break;
557 	case 6:
558 		ksnprintf(ver->description, sizeof(ver->description),
559 			  "Directory Hash ALG1 (tmp/rename resistance)");
560 		break;
561 	default:
562 		ksnprintf(ver->description, sizeof(ver->description),
563 			 "Unknown");
564 		error = EINVAL;
565 		break;
566 	}
567 	return(error);
568 };
569 
570 /*
571  * Set version info
572  */
573 static
574 int
575 hammer_ioc_set_version(hammer_transaction_t trans, hammer_inode_t ip,
576 		   struct hammer_ioc_version *ver)
577 {
578 	hammer_mount_t hmp = trans->hmp;
579 	struct hammer_cursor cursor;
580 	hammer_volume_t volume;
581 	int error;
582 	int over = hmp->version;
583 
584 	/*
585 	 * Generally do not allow downgrades.  However, version 4 can
586 	 * be downgraded to version 3.
587 	 */
588 	if (ver->cur_version < hmp->version) {
589 		if (!(ver->cur_version == 3 && hmp->version == 4))
590 			return(EINVAL);
591 	}
592 	if (ver->cur_version == hmp->version)
593 		return(0);
594 	if (ver->cur_version > HAMMER_VOL_VERSION_MAX)
595 		return(EINVAL);
596 	if (hmp->ronly)
597 		return(EROFS);
598 
599 	/*
600 	 * Update the root volume header and the version cached in
601 	 * the hammer_mount structure.
602 	 */
603 	error = hammer_init_cursor(trans, &cursor, NULL, NULL);
604 	if (error)
605 		goto failed;
606 	hammer_lock_ex(&hmp->flusher.finalize_lock);
607 	hammer_sync_lock_ex(trans);
608 	hmp->version = ver->cur_version;
609 
610 	/*
611 	 * If upgrading from version < 4 to version >= 4 the UNDO FIFO
612 	 * must be reinitialized.
613 	 */
614 	if (over < HAMMER_VOL_VERSION_FOUR &&
615 	    ver->cur_version >= HAMMER_VOL_VERSION_FOUR) {
616 		kprintf("upgrade undo to version 4\n");
617 		error = hammer_upgrade_undo_4(trans);
618 		if (error)
619 			goto failed;
620 	}
621 
622 	/*
623 	 * Adjust the version in the volume header
624 	 */
625 	volume = hammer_get_root_volume(hmp, &error);
626 	KKASSERT(error == 0);
627 	hammer_modify_volume_field(cursor.trans, volume, vol_version);
628 	volume->ondisk->vol_version = ver->cur_version;
629 	hammer_modify_volume_done(volume);
630 	hammer_rel_volume(volume, 0);
631 
632 	hammer_sync_unlock(trans);
633 	hammer_unlock(&hmp->flusher.finalize_lock);
634 failed:
635 	ver->head.error = error;
636 	hammer_done_cursor(&cursor);
637 	return(0);
638 }
639 
640 /*
641  * Get information
642  */
643 static
644 int
645 hammer_ioc_get_info(hammer_transaction_t trans, struct hammer_ioc_info *info) {
646 
647 	struct hammer_volume_ondisk	*od = trans->hmp->rootvol->ondisk;
648 	struct hammer_mount 		*hm = trans->hmp;
649 
650 	/* Fill the structure with the necessary information */
651 	_hammer_checkspace(hm, HAMMER_CHKSPC_WRITE, &info->rsvbigblocks);
652 	info->rsvbigblocks = info->rsvbigblocks >> HAMMER_LARGEBLOCK_BITS;
653 	strlcpy(info->vol_name, od->vol_name, sizeof(od->vol_name));
654 
655 	info->vol_fsid = hm->fsid;
656 	info->vol_fstype = od->vol_fstype;
657 	info->version = hm->version;
658 
659 	info->inodes = od->vol0_stat_inodes;
660 	info->bigblocks = od->vol0_stat_bigblocks;
661 	info->freebigblocks = od->vol0_stat_freebigblocks;
662 	info->nvolumes = hm->nvolumes;
663 
664 	return 0;
665 }
666 
667 /*
668  * Add a snapshot transction id(s) to the list of snapshots.
669  *
670  * NOTE: Records are created with an allocated TID.  If a flush cycle
671  *	 is in progress the record may be synced in the current flush
672  *	 cycle and the volume header will reflect the allocation of the
673  *	 TID, but the synchronization point may not catch up to the
674  *	 TID until the next flush cycle.
675  */
676 static
677 int
678 hammer_ioc_add_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
679 			struct hammer_ioc_snapshot *snap)
680 {
681 	hammer_mount_t hmp = ip->hmp;
682 	struct hammer_btree_leaf_elm leaf;
683 	struct hammer_cursor cursor;
684 	int error;
685 
686 	/*
687 	 * Validate structure
688 	 */
689 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
690 		return (EINVAL);
691 	if (snap->index > snap->count)
692 		return (EINVAL);
693 
694 	hammer_lock_ex(&hmp->snapshot_lock);
695 again:
696 	/*
697 	 * Look for keys starting after the previous iteration, or at
698 	 * the beginning if snap->count is 0.
699 	 */
700 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
701 	if (error) {
702 		hammer_done_cursor(&cursor);
703 		return(error);
704 	}
705 
706 	cursor.asof = HAMMER_MAX_TID;
707 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
708 
709 	bzero(&leaf, sizeof(leaf));
710 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
711 	leaf.base.rec_type = HAMMER_RECTYPE_SNAPSHOT;
712 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
713 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
714 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
715 	leaf.data_len = sizeof(struct hammer_snapshot_data);
716 
717 	while (snap->index < snap->count) {
718 		leaf.base.key = (int64_t)snap->snaps[snap->index].tid;
719 		cursor.key_beg = leaf.base;
720 		error = hammer_btree_lookup(&cursor);
721 		if (error == 0) {
722 			error = EEXIST;
723 			break;
724 		}
725 
726 		/*
727 		 * NOTE: Must reload key_beg after an ASOF search because
728 		 *	 the create_tid may have been modified during the
729 		 *	 search.
730 		 */
731 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
732 		cursor.key_beg = leaf.base;
733 		error = hammer_create_at_cursor(&cursor, &leaf,
734 						&snap->snaps[snap->index],
735 						HAMMER_CREATE_MODE_SYS);
736 		if (error == EDEADLK) {
737 			hammer_done_cursor(&cursor);
738 			goto again;
739 		}
740 		cursor.flags |= HAMMER_CURSOR_ASOF;
741 		if (error)
742 			break;
743 		++snap->index;
744 	}
745 	snap->head.error = error;
746 	hammer_done_cursor(&cursor);
747 	hammer_unlock(&hmp->snapshot_lock);
748 	return(0);
749 }
750 
751 /*
752  * Delete snapshot transaction id(s) from the list of snapshots.
753  */
754 static
755 int
756 hammer_ioc_del_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
757 			struct hammer_ioc_snapshot *snap)
758 {
759 	hammer_mount_t hmp = ip->hmp;
760 	struct hammer_cursor cursor;
761 	int error;
762 
763 	/*
764 	 * Validate structure
765 	 */
766 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
767 		return (EINVAL);
768 	if (snap->index > snap->count)
769 		return (EINVAL);
770 
771 	hammer_lock_ex(&hmp->snapshot_lock);
772 again:
773 	/*
774 	 * Look for keys starting after the previous iteration, or at
775 	 * the beginning if snap->count is 0.
776 	 */
777 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
778 	if (error) {
779 		hammer_done_cursor(&cursor);
780 		return(error);
781 	}
782 
783 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
784 	cursor.key_beg.create_tid = 0;
785 	cursor.key_beg.delete_tid = 0;
786 	cursor.key_beg.obj_type = 0;
787 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
788 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
789 	cursor.asof = HAMMER_MAX_TID;
790 	cursor.flags |= HAMMER_CURSOR_ASOF;
791 
792 	while (snap->index < snap->count) {
793 		cursor.key_beg.key = (int64_t)snap->snaps[snap->index].tid;
794 		error = hammer_btree_lookup(&cursor);
795 		if (error)
796 			break;
797 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
798 		if (error)
799 			break;
800 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
801 						0, 0, 0, NULL);
802 		if (error == EDEADLK) {
803 			hammer_done_cursor(&cursor);
804 			goto again;
805 		}
806 		if (error)
807 			break;
808 		++snap->index;
809 	}
810 	snap->head.error = error;
811 	hammer_done_cursor(&cursor);
812 	hammer_unlock(&hmp->snapshot_lock);
813 	return(0);
814 }
815 
816 /*
817  * Retrieve as many snapshot ids as possible or until the array is
818  * full, starting after the last transction id passed in.  If count
819  * is 0 we retrieve starting at the beginning.
820  *
821  * NOTE: Because the b-tree key field is signed but transaction ids
822  *       are unsigned the returned list will be signed-sorted instead
823  *	 of unsigned sorted.  The Caller must still sort the aggregate
824  *	 results.
825  */
826 static
827 int
828 hammer_ioc_get_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
829 			struct hammer_ioc_snapshot *snap)
830 {
831 	struct hammer_cursor cursor;
832 	int error;
833 
834 	/*
835 	 * Validate structure
836 	 */
837 	if (snap->index != 0)
838 		return (EINVAL);
839 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
840 		return (EINVAL);
841 
842 	/*
843 	 * Look for keys starting after the previous iteration, or at
844 	 * the beginning if snap->count is 0.
845 	 */
846 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
847 	if (error) {
848 		hammer_done_cursor(&cursor);
849 		return(error);
850 	}
851 
852 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
853 	cursor.key_beg.create_tid = 0;
854 	cursor.key_beg.delete_tid = 0;
855 	cursor.key_beg.obj_type = 0;
856 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
857 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
858 	if (snap->count == 0)
859 		cursor.key_beg.key = HAMMER_MIN_KEY;
860 	else
861 		cursor.key_beg.key = (int64_t)snap->snaps[snap->count - 1].tid + 1;
862 
863 	cursor.key_end = cursor.key_beg;
864 	cursor.key_end.key = HAMMER_MAX_KEY;
865 	cursor.asof = HAMMER_MAX_TID;
866 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE | HAMMER_CURSOR_ASOF;
867 
868 	snap->count = 0;
869 
870 	error = hammer_btree_first(&cursor);
871 	while (error == 0 && snap->count < HAMMER_SNAPS_PER_IOCTL) {
872 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
873 		if (error)
874 			break;
875 		if (cursor.leaf->base.rec_type == HAMMER_RECTYPE_SNAPSHOT) {
876 			error = hammer_btree_extract(
877 					     &cursor, HAMMER_CURSOR_GET_LEAF |
878 						      HAMMER_CURSOR_GET_DATA);
879 			snap->snaps[snap->count] = cursor.data->snap;
880 
881 			/*
882 			 * The snap data tid should match the key but might
883 			 * not due to a bug in the HAMMER v3 conversion code.
884 			 *
885 			 * This error will work itself out over time but we
886 			 * have to force a match or the snapshot will not
887 			 * be deletable.
888 			 */
889 			if (cursor.data->snap.tid !=
890 			    (hammer_tid_t)cursor.leaf->base.key) {
891 				kprintf("HAMMER: lo=%08x snapshot key "
892 					"0x%016jx data mismatch 0x%016jx\n",
893 					cursor.key_beg.localization,
894 					(uintmax_t)cursor.data->snap.tid,
895 					cursor.leaf->base.key);
896 				kprintf("HAMMER: Probably left over from the "
897 					"original v3 conversion, hammer "
898 					"cleanup should get it eventually\n");
899 				snap->snaps[snap->count].tid =
900 					cursor.leaf->base.key;
901 			}
902 			++snap->count;
903 		}
904 		error = hammer_btree_iterate(&cursor);
905 	}
906 
907 	if (error == ENOENT) {
908 		snap->head.flags |= HAMMER_IOC_SNAPSHOT_EOF;
909 		error = 0;
910 	}
911 	snap->head.error = error;
912 	hammer_done_cursor(&cursor);
913 	return(0);
914 }
915 
916 /*
917  * Retrieve the PFS hammer cleanup utility config record.  This is
918  * different (newer than) the PFS config.
919  */
920 static
921 int
922 hammer_ioc_get_config(hammer_transaction_t trans, hammer_inode_t ip,
923 			struct hammer_ioc_config *config)
924 {
925 	struct hammer_cursor cursor;
926 	int error;
927 
928 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
929 	if (error) {
930 		hammer_done_cursor(&cursor);
931 		return(error);
932 	}
933 
934 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
935 	cursor.key_beg.create_tid = 0;
936 	cursor.key_beg.delete_tid = 0;
937 	cursor.key_beg.obj_type = 0;
938 	cursor.key_beg.rec_type = HAMMER_RECTYPE_CONFIG;
939 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
940 	cursor.key_beg.key = 0;		/* config space page 0 */
941 
942 	cursor.asof = HAMMER_MAX_TID;
943 	cursor.flags |= HAMMER_CURSOR_ASOF;
944 
945 	error = hammer_btree_lookup(&cursor);
946 	if (error == 0) {
947 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
948 						      HAMMER_CURSOR_GET_DATA);
949 		if (error == 0)
950 			config->config = cursor.data->config;
951 	}
952 	/* error can be ENOENT */
953 	config->head.error = error;
954 	hammer_done_cursor(&cursor);
955 	return(0);
956 }
957 
958 /*
959  * Retrieve the PFS hammer cleanup utility config record.  This is
960  * different (newer than) the PFS config.
961  *
962  * This is kinda a hack.
963  */
964 static
965 int
966 hammer_ioc_set_config(hammer_transaction_t trans, hammer_inode_t ip,
967 			struct hammer_ioc_config *config)
968 {
969 	struct hammer_btree_leaf_elm leaf;
970 	struct hammer_cursor cursor;
971 	hammer_mount_t hmp = ip->hmp;
972 	int error;
973 
974 again:
975 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
976 	if (error) {
977 		hammer_done_cursor(&cursor);
978 		return(error);
979 	}
980 
981 	bzero(&leaf, sizeof(leaf));
982 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
983 	leaf.base.rec_type = HAMMER_RECTYPE_CONFIG;
984 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
985 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
986 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
987 	leaf.base.key = 0;	/* page 0 */
988 	leaf.data_len = sizeof(struct hammer_config_data);
989 
990 	cursor.key_beg = leaf.base;
991 
992 	cursor.asof = HAMMER_MAX_TID;
993 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
994 
995 	error = hammer_btree_lookup(&cursor);
996 	if (error == 0) {
997 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
998 						      HAMMER_CURSOR_GET_DATA);
999 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
1000 						0, 0, 0, NULL);
1001 		if (error == EDEADLK) {
1002 			hammer_done_cursor(&cursor);
1003 			goto again;
1004 		}
1005 	}
1006 	if (error == ENOENT)
1007 		error = 0;
1008 	if (error == 0) {
1009 		/*
1010 		 * NOTE: Must reload key_beg after an ASOF search because
1011 		 *	 the create_tid may have been modified during the
1012 		 *	 search.
1013 		 */
1014 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
1015 		cursor.key_beg = leaf.base;
1016 		error = hammer_create_at_cursor(&cursor, &leaf,
1017 						&config->config,
1018 						HAMMER_CREATE_MODE_SYS);
1019 		if (error == EDEADLK) {
1020 			hammer_done_cursor(&cursor);
1021 			goto again;
1022 		}
1023 	}
1024 	config->head.error = error;
1025 	hammer_done_cursor(&cursor);
1026 	return(0);
1027 }
1028 
1029 static
1030 int
1031 hammer_ioc_pfs_iterate(hammer_transaction_t trans,
1032     struct hammer_ioc_pfs_iterate *pi)
1033 {
1034 	struct hammer_cursor cursor;
1035 	hammer_inode_t ip;
1036 	int error;
1037 
1038 	ip = hammer_get_inode(trans, NULL, HAMMER_OBJID_ROOT, HAMMER_MAX_TID,
1039 	    HAMMER_DEF_LOCALIZATION, 0, &error);
1040 
1041 	error = hammer_init_cursor(trans, &cursor,
1042 	    (ip ? &ip->cache[1] : NULL), ip);
1043 	if (error)
1044 		goto out;
1045 
1046 	pi->head.flags &= ~HAMMER_PFSD_DELETED;
1047 
1048 	cursor.key_beg.localization = HAMMER_DEF_LOCALIZATION +
1049 	    HAMMER_LOCALIZE_MISC;
1050 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
1051 	cursor.key_beg.create_tid = 0;
1052 	cursor.key_beg.delete_tid = 0;
1053 	cursor.key_beg.rec_type = HAMMER_RECTYPE_PFS;
1054 	cursor.key_beg.obj_type = 0;
1055 	cursor.key_end = cursor.key_beg;
1056 	cursor.key_end.key = HAMMER_MAX_KEY;
1057 	cursor.asof = HAMMER_MAX_TID;
1058 	cursor.flags |= HAMMER_CURSOR_ASOF;
1059 
1060 	if (pi->pos < 0)	/* Sanity check */
1061 		pi->pos = 0;
1062 
1063 	pi->pos <<= 16;
1064 	cursor.key_beg.key = pi->pos;
1065 	error = hammer_ip_lookup(&cursor);
1066 
1067 	if (error == 0) {
1068 		error = hammer_ip_resolve_data(&cursor);
1069 		if (error)
1070 			goto out;
1071 		if (cursor.data->pfsd.mirror_flags & HAMMER_PFSD_DELETED)
1072 			pi->head.flags |= HAMMER_PFSD_DELETED;
1073 		else
1074 			copyout(cursor.data, pi->ondisk, cursor.leaf->data_len);
1075 		pi->pos = (u_int32_t)(cursor.leaf->base.key >> 16);
1076 	}
1077 
1078 out:
1079 	hammer_done_cursor(&cursor);
1080 	if (ip)
1081 		hammer_rel_inode(ip, 0);
1082 
1083 	return (error);
1084 }
1085 
1086 static
1087 int
1088 hammer_ioc_get_data(hammer_transaction_t trans, hammer_inode_t ip,
1089 			struct hammer_ioc_data *data)
1090 {
1091 	struct hammer_cursor cursor;
1092 	int bytes;
1093 	int error;
1094 
1095 	/* XXX cached inode ? */
1096 	error = hammer_init_cursor(trans, &cursor, NULL, NULL);
1097 	if (error)
1098 		goto failed;
1099 
1100 	cursor.key_beg = data->elm;
1101 	cursor.flags |= HAMMER_CURSOR_BACKEND;
1102 
1103 	error = hammer_btree_lookup(&cursor);
1104 	if (error == 0) {
1105 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
1106 						      HAMMER_CURSOR_GET_DATA);
1107 		if (error == 0) {
1108 			data->leaf = *cursor.leaf;
1109 			bytes = cursor.leaf->data_len;
1110 			if (bytes > data->size)
1111 				bytes = data->size;
1112 			error = copyout(cursor.data, data->ubuf, bytes);
1113 		}
1114 	}
1115 
1116 failed:
1117 	hammer_done_cursor(&cursor);
1118 	return (error);
1119 }
1120