xref: /dragonfly/sys/vfs/hammer/hammer_ioctl.c (revision b0d289c2)
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 static int hammer_ioc_get_data(hammer_transaction_t trans, hammer_inode_t ip,
62 				struct hammer_ioc_data *data);
63 
64 int
65 hammer_ioctl(hammer_inode_t ip, u_long com, caddr_t data, int fflag,
66 	     struct ucred *cred)
67 {
68 	struct hammer_transaction trans;
69 	int error;
70 
71 	error = priv_check_cred(cred, PRIV_HAMMER_IOCTL, 0);
72 
73 	hammer_start_transaction(&trans, ip->hmp);
74 
75 	switch(com) {
76 	case HAMMERIOC_PRUNE:
77 		if (error == 0) {
78 			error = hammer_ioc_prune(&trans, ip,
79 					(struct hammer_ioc_prune *)data);
80 		}
81 		break;
82 	case HAMMERIOC_GETHISTORY:
83 		error = hammer_ioc_gethistory(&trans, ip,
84 					(struct hammer_ioc_history *)data);
85 		break;
86 	case HAMMERIOC_REBLOCK:
87 		if (error == 0) {
88 			error = hammer_ioc_reblock(&trans, ip,
89 					(struct hammer_ioc_reblock *)data);
90 		}
91 		break;
92 	case HAMMERIOC_REBALANCE:
93 		/*
94 		 * Rebalancing needs to lock a lot of B-Tree nodes.  The
95 		 * children and children's children.  Systems with very
96 		 * little memory will not be able to do it.
97 		 */
98 		if (error == 0 && nbuf < HAMMER_REBALANCE_MIN_BUFS) {
99 			kprintf("hammer: System has insufficient buffers "
100 				"to rebalance the tree.  nbuf < %d\n",
101 				HAMMER_REBALANCE_MIN_BUFS);
102 			error = ENOSPC;
103 		}
104 		if (error == 0) {
105 			error = hammer_ioc_rebalance(&trans, ip,
106 					(struct hammer_ioc_rebalance *)data);
107 		}
108 		break;
109 	case HAMMERIOC_SYNCTID:
110 		error = hammer_ioc_synctid(&trans, ip,
111 					(struct hammer_ioc_synctid *)data);
112 		break;
113 	case HAMMERIOC_GET_PSEUDOFS:
114 		error = hammer_ioc_get_pseudofs(&trans, ip,
115 				    (struct hammer_ioc_pseudofs_rw *)data);
116 		break;
117 	case HAMMERIOC_SET_PSEUDOFS:
118 		if (error == 0) {
119 			error = hammer_ioc_set_pseudofs(&trans, ip, cred,
120 				    (struct hammer_ioc_pseudofs_rw *)data);
121 		}
122 		break;
123 	case HAMMERIOC_UPG_PSEUDOFS:
124 		if (error == 0) {
125 			error = hammer_ioc_upgrade_pseudofs(&trans, ip,
126 				    (struct hammer_ioc_pseudofs_rw *)data);
127 		}
128 		break;
129 	case HAMMERIOC_DGD_PSEUDOFS:
130 		if (error == 0) {
131 			error = hammer_ioc_downgrade_pseudofs(&trans, ip,
132 				    (struct hammer_ioc_pseudofs_rw *)data);
133 		}
134 		break;
135 	case HAMMERIOC_RMR_PSEUDOFS:
136 		if (error == 0) {
137 			error = hammer_ioc_destroy_pseudofs(&trans, ip,
138 				    (struct hammer_ioc_pseudofs_rw *)data);
139 		}
140 		break;
141 	case HAMMERIOC_WAI_PSEUDOFS:
142 		if (error == 0) {
143 			error = hammer_ioc_wait_pseudofs(&trans, ip,
144 				    (struct hammer_ioc_pseudofs_rw *)data);
145 		}
146 		break;
147 	case HAMMERIOC_MIRROR_READ:
148 		if (error == 0) {
149 			error = hammer_ioc_mirror_read(&trans, ip,
150 				    (struct hammer_ioc_mirror_rw *)data);
151 		}
152 		break;
153 	case HAMMERIOC_MIRROR_WRITE:
154 		if (error == 0) {
155 			error = hammer_ioc_mirror_write(&trans, ip,
156 				    (struct hammer_ioc_mirror_rw *)data);
157 		}
158 		break;
159 	case HAMMERIOC_GET_VERSION:
160 		error = hammer_ioc_get_version(&trans, ip,
161 				    (struct hammer_ioc_version *)data);
162 		break;
163 	case HAMMERIOC_GET_INFO:
164 		error = hammer_ioc_get_info(&trans,
165 				    (struct hammer_ioc_info *)data);
166 		break;
167 	case HAMMERIOC_SET_VERSION:
168 		if (error == 0) {
169 			error = hammer_ioc_set_version(&trans, ip,
170 					    (struct hammer_ioc_version *)data);
171 		}
172 		break;
173 	case HAMMERIOC_ADD_VOLUME:
174 		if (error == 0) {
175 			error = priv_check_cred(cred, PRIV_HAMMER_VOLUME, 0);
176 			if (error == 0)
177 				error = hammer_ioc_volume_add(&trans, ip,
178 					    (struct hammer_ioc_volume *)data);
179 		}
180 		break;
181 	case HAMMERIOC_DEL_VOLUME:
182 		if (error == 0) {
183 			error = priv_check_cred(cred, PRIV_HAMMER_VOLUME, 0);
184 			if (error == 0)
185 				error = hammer_ioc_volume_del(&trans, ip,
186 					    (struct hammer_ioc_volume *)data);
187 		}
188 		break;
189 	case HAMMERIOC_LIST_VOLUMES:
190 		error = hammer_ioc_volume_list(&trans, ip,
191 		    (struct hammer_ioc_volume_list *)data);
192 		break;
193 	case HAMMERIOC_ADD_SNAPSHOT:
194 		if (error == 0) {
195 			error = hammer_ioc_add_snapshot(
196 					&trans, ip, (struct hammer_ioc_snapshot *)data);
197 		}
198 		break;
199 	case HAMMERIOC_DEL_SNAPSHOT:
200 		if (error == 0) {
201 			error = hammer_ioc_del_snapshot(
202 					&trans, ip, (struct hammer_ioc_snapshot *)data);
203 		}
204 		break;
205 	case HAMMERIOC_GET_SNAPSHOT:
206 		error = hammer_ioc_get_snapshot(
207 					&trans, ip, (struct hammer_ioc_snapshot *)data);
208 		break;
209 	case HAMMERIOC_GET_CONFIG:
210 		error = hammer_ioc_get_config(
211 					&trans, ip, (struct hammer_ioc_config *)data);
212 		break;
213 	case HAMMERIOC_SET_CONFIG:
214 		if (error == 0) {
215 			error = hammer_ioc_set_config(
216 					&trans, ip, (struct hammer_ioc_config *)data);
217 		}
218 		break;
219 	case HAMMERIOC_DEDUP:
220 		if (error == 0) {
221 			error = hammer_ioc_dedup(
222 					&trans, ip, (struct hammer_ioc_dedup *)data);
223 		}
224 		break;
225 	case HAMMERIOC_GET_DATA:
226 		if (error == 0) {
227 			error = hammer_ioc_get_data(
228 					&trans, ip, (struct hammer_ioc_data *)data);
229 		}
230 		break;
231 	case HAMMERIOC_PFS_ITERATE:
232 		error = hammer_ioc_iterate_pseudofs(
233 				&trans, ip, (struct hammer_ioc_pfs_iterate *)data);
234 		break;
235 	default:
236 		error = EOPNOTSUPP;
237 		break;
238 	}
239 	hammer_done_transaction(&trans);
240 	return (error);
241 }
242 
243 /*
244  * Iterate through an object's inode or an object's records and record
245  * modification TIDs.
246  */
247 static void add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
248 			hammer_btree_elm_t elm);
249 
250 static
251 int
252 hammer_ioc_gethistory(hammer_transaction_t trans, hammer_inode_t ip,
253 		      struct hammer_ioc_history *hist)
254 {
255 	struct hammer_cursor cursor;
256 	hammer_btree_elm_t elm;
257 	int error;
258 
259 	/*
260 	 * Validate the structure and initialize for return.
261 	 */
262 	if (hist->beg_tid > hist->end_tid)
263 		return(EINVAL);
264 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
265 		if (hist->key > hist->nxt_key)
266 			return(EINVAL);
267 	}
268 
269 	hist->obj_id = ip->obj_id;
270 	hist->count = 0;
271 	hist->nxt_tid = hist->end_tid;
272 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_TID;
273 	hist->head.flags &= ~HAMMER_IOC_HISTORY_NEXT_KEY;
274 	hist->head.flags &= ~HAMMER_IOC_HISTORY_EOF;
275 	hist->head.flags &= ~HAMMER_IOC_HISTORY_UNSYNCED;
276 	if ((ip->flags & HAMMER_INODE_MODMASK) &
277 	    ~(HAMMER_INODE_ATIME | HAMMER_INODE_MTIME)) {
278 		hist->head.flags |= HAMMER_IOC_HISTORY_UNSYNCED;
279 	}
280 
281 	/*
282 	 * Setup the cursor.  We can't handle undeletable records
283 	 * (create_tid of 0) at the moment.  A create_tid of 0 has
284 	 * a special meaning and cannot be specified in the cursor.
285 	 */
286 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
287 	if (error) {
288 		hammer_done_cursor(&cursor);
289 		return(error);
290 	}
291 
292 	cursor.key_beg.obj_id = hist->obj_id;
293 	cursor.key_beg.create_tid = hist->beg_tid;
294 	cursor.key_beg.delete_tid = 0;
295 	cursor.key_beg.obj_type = 0;
296 	if (cursor.key_beg.create_tid == HAMMER_MIN_TID)
297 		cursor.key_beg.create_tid = 1;
298 
299 	cursor.key_end.obj_id = hist->obj_id;
300 	cursor.key_end.create_tid = hist->end_tid;
301 	cursor.key_end.delete_tid = 0;
302 	cursor.key_end.obj_type = 0;
303 
304 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE;
305 
306 	if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
307 		/*
308 		 * key-range within the file.  For a regular file the
309 		 * on-disk key represents BASE+LEN, not BASE, so the
310 		 * first possible record containing the offset 'key'
311 		 * has an on-disk key of (key + 1).
312 		 */
313 		cursor.key_beg.key = hist->key;
314 		cursor.key_end.key = HAMMER_MAX_KEY;
315 		cursor.key_beg.localization = ip->obj_localization +
316 					      HAMMER_LOCALIZE_MISC;
317 		cursor.key_end.localization = ip->obj_localization +
318 					      HAMMER_LOCALIZE_MISC;
319 
320 		switch(ip->ino_data.obj_type) {
321 		case HAMMER_OBJTYPE_REGFILE:
322 			++cursor.key_beg.key;
323 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DATA;
324 			break;
325 		case HAMMER_OBJTYPE_DIRECTORY:
326 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DIRENTRY;
327 			cursor.key_beg.localization = ip->obj_localization +
328 						hammer_dir_localization(ip);
329 			cursor.key_end.localization = ip->obj_localization +
330 						hammer_dir_localization(ip);
331 			break;
332 		case HAMMER_OBJTYPE_DBFILE:
333 			cursor.key_beg.rec_type = HAMMER_RECTYPE_DB;
334 			break;
335 		default:
336 			error = EINVAL;
337 			break;
338 		}
339 		cursor.key_end.rec_type = cursor.key_beg.rec_type;
340 	} else {
341 		/*
342 		 * The inode itself.
343 		 */
344 		cursor.key_beg.key = 0;
345 		cursor.key_end.key = 0;
346 		cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE;
347 		cursor.key_end.rec_type = HAMMER_RECTYPE_INODE;
348 		cursor.key_beg.localization = ip->obj_localization +
349 					      HAMMER_LOCALIZE_INODE;
350 		cursor.key_end.localization = ip->obj_localization +
351 					      HAMMER_LOCALIZE_INODE;
352 	}
353 
354 	error = hammer_btree_first(&cursor);
355 	while (error == 0) {
356 		elm = &cursor.node->ondisk->elms[cursor.index];
357 
358 		add_history(ip, hist, elm);
359 		if (hist->head.flags & (HAMMER_IOC_HISTORY_NEXT_TID |
360 				        HAMMER_IOC_HISTORY_NEXT_KEY |
361 				        HAMMER_IOC_HISTORY_EOF)) {
362 			break;
363 		}
364 		error = hammer_btree_iterate(&cursor);
365 	}
366 	if (error == ENOENT) {
367 		hist->head.flags |= HAMMER_IOC_HISTORY_EOF;
368 		error = 0;
369 	}
370 	hammer_done_cursor(&cursor);
371 	return(error);
372 }
373 
374 /*
375  * Add the scanned element to the ioctl return structure.  Some special
376  * casing is required for regular files to accomodate how data ranges are
377  * stored on-disk.
378  */
379 static void
380 add_history(hammer_inode_t ip, struct hammer_ioc_history *hist,
381 	    hammer_btree_elm_t elm)
382 {
383 	int i;
384 
385 	if (elm->base.btype != HAMMER_BTREE_TYPE_RECORD)
386 		return;
387 	if ((hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) &&
388 	    ip->ino_data.obj_type == HAMMER_OBJTYPE_REGFILE) {
389 		/*
390 		 * Adjust nxt_key
391 		 */
392 		if (hist->nxt_key > elm->leaf.base.key - elm->leaf.data_len &&
393 		    hist->key < elm->leaf.base.key - elm->leaf.data_len) {
394 			hist->nxt_key = elm->leaf.base.key - elm->leaf.data_len;
395 		}
396 		if (hist->nxt_key > elm->leaf.base.key)
397 			hist->nxt_key = elm->leaf.base.key;
398 
399 		/*
400 		 * Record is beyond MAXPHYS, there won't be any more records
401 		 * in the iteration covering the requested offset (key).
402 		 */
403 		if (elm->leaf.base.key >= MAXPHYS &&
404 		    elm->leaf.base.key - MAXPHYS > hist->key) {
405 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
406 		}
407 
408 		/*
409 		 * Data-range of record does not cover the key.
410 		 */
411 		if (elm->leaf.base.key - elm->leaf.data_len > hist->key)
412 			return;
413 
414 	} else if (hist->head.flags & HAMMER_IOC_HISTORY_ATKEY) {
415 		/*
416 		 * Adjust nxt_key
417 		 */
418 		if (hist->nxt_key > elm->leaf.base.key &&
419 		    hist->key < elm->leaf.base.key) {
420 			hist->nxt_key = elm->leaf.base.key;
421 		}
422 
423 		/*
424 		 * Record is beyond the requested key.
425 		 */
426 		if (elm->leaf.base.key > hist->key)
427 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_KEY;
428 	}
429 
430 	/*
431 	 * Add create_tid if it is in-bounds.
432 	 */
433 	i = hist->count;
434 	if ((i == 0 ||
435 	     elm->leaf.base.create_tid != hist->hist_ary[i - 1].tid) &&
436 	    elm->leaf.base.create_tid >= hist->beg_tid &&
437 	    elm->leaf.base.create_tid < hist->end_tid) {
438 		if (hist->count == HAMMER_MAX_HISTORY_ELMS) {
439 			hist->nxt_tid = elm->leaf.base.create_tid;
440 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
441 			return;
442 		}
443 		hist->hist_ary[i].tid = elm->leaf.base.create_tid;
444 		hist->hist_ary[i].time32 = elm->leaf.create_ts;
445 		++hist->count;
446 	}
447 
448 	/*
449 	 * Add delete_tid if it is in-bounds.  Note that different portions
450 	 * of the history may have overlapping data ranges with different
451 	 * delete_tid's.  If this case occurs the delete_tid may match the
452 	 * create_tid of a following record.  XXX
453 	 *
454 	 *	[        ]
455 	 *            [     ]
456 	 */
457 	i = hist->count;
458 	if (elm->leaf.base.delete_tid &&
459 	    elm->leaf.base.delete_tid >= hist->beg_tid &&
460 	    elm->leaf.base.delete_tid < hist->end_tid) {
461 		if (i == HAMMER_MAX_HISTORY_ELMS) {
462 			hist->nxt_tid = elm->leaf.base.delete_tid;
463 			hist->head.flags |= HAMMER_IOC_HISTORY_NEXT_TID;
464 			return;
465 		}
466 		hist->hist_ary[i].tid = elm->leaf.base.delete_tid;
467 		hist->hist_ary[i].time32 = elm->leaf.delete_ts;
468 		++hist->count;
469 	}
470 }
471 
472 /*
473  * Acquire synchronization TID
474  */
475 static
476 int
477 hammer_ioc_synctid(hammer_transaction_t trans, hammer_inode_t ip,
478 		   struct hammer_ioc_synctid *std)
479 {
480 	hammer_mount_t hmp = ip->hmp;
481 	int error = 0;
482 
483 	switch(std->op) {
484 	case HAMMER_SYNCTID_NONE:
485 		std->tid = hmp->flusher.tid;	/* inaccurate */
486 		break;
487 	case HAMMER_SYNCTID_ASYNC:
488 		hammer_queue_inodes_flusher(hmp, MNT_NOWAIT);
489 		hammer_flusher_async(hmp, NULL);
490 		std->tid = hmp->flusher.tid;	/* inaccurate */
491 		break;
492 	case HAMMER_SYNCTID_SYNC1:
493 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
494 		hammer_flusher_sync(hmp);
495 		std->tid = hmp->flusher.tid;
496 		break;
497 	case HAMMER_SYNCTID_SYNC2:
498 		hammer_queue_inodes_flusher(hmp, MNT_WAIT);
499 		hammer_flusher_sync(hmp);
500 		std->tid = hmp->flusher.tid;
501 		hammer_flusher_sync(hmp);
502 		break;
503 	default:
504 		error = EOPNOTSUPP;
505 		break;
506 	}
507 	return(error);
508 }
509 
510 /*
511  * Retrieve version info.
512  *
513  * Load min_version, wip_version, and max_versino.  If cur_version is passed
514  * as 0 then load the current version into cur_version.  Load the description
515  * for cur_version into the description array.
516  *
517  * Returns 0 on success, EINVAL if cur_version is non-zero and set to an
518  * unsupported value.
519  */
520 static
521 int
522 hammer_ioc_get_version(hammer_transaction_t trans, hammer_inode_t ip,
523 		   struct hammer_ioc_version *ver)
524 {
525 	int error = 0;
526 
527 	ver->min_version = HAMMER_VOL_VERSION_MIN;
528 	ver->wip_version = HAMMER_VOL_VERSION_WIP;
529 	ver->max_version = HAMMER_VOL_VERSION_MAX;
530 	if (ver->cur_version == 0)
531 		ver->cur_version = trans->hmp->version;
532 	switch(ver->cur_version) {
533 	case 1:
534 		ksnprintf(ver->description, sizeof(ver->description),
535 			 "First HAMMER release (DragonFly 2.0+)");
536 		break;
537 	case 2:
538 		ksnprintf(ver->description, sizeof(ver->description),
539 			 "New directory entry layout (DragonFly 2.3+)");
540 		break;
541 	case 3:
542 		ksnprintf(ver->description, sizeof(ver->description),
543 			 "New snapshot management (DragonFly 2.5+)");
544 		break;
545 	case 4:
546 		ksnprintf(ver->description, sizeof(ver->description),
547 			 "New undo/flush, faster flush/sync (DragonFly 2.5+)");
548 		break;
549 	case 5:
550 		ksnprintf(ver->description, sizeof(ver->description),
551 			 "Adjustments for dedup support (DragonFly 2.9+)");
552 		break;
553 	case 6:
554 		ksnprintf(ver->description, sizeof(ver->description),
555 			  "Directory Hash ALG1 (tmp/rename resistance)");
556 		break;
557 	default:
558 		ksnprintf(ver->description, sizeof(ver->description),
559 			 "Unknown");
560 		error = EINVAL;
561 		break;
562 	}
563 	return(error);
564 };
565 
566 /*
567  * Set version info
568  */
569 static
570 int
571 hammer_ioc_set_version(hammer_transaction_t trans, hammer_inode_t ip,
572 		   struct hammer_ioc_version *ver)
573 {
574 	hammer_mount_t hmp = trans->hmp;
575 	struct hammer_cursor cursor;
576 	hammer_volume_t volume;
577 	int error;
578 	int over = hmp->version;
579 
580 	/*
581 	 * Generally do not allow downgrades.  However, version 4 can
582 	 * be downgraded to version 3.
583 	 */
584 	if (ver->cur_version < hmp->version) {
585 		if (!(ver->cur_version == 3 && hmp->version == 4))
586 			return(EINVAL);
587 	}
588 	if (ver->cur_version == hmp->version)
589 		return(0);
590 	if (ver->cur_version > HAMMER_VOL_VERSION_MAX)
591 		return(EINVAL);
592 	if (hmp->ronly)
593 		return(EROFS);
594 
595 	/*
596 	 * Update the root volume header and the version cached in
597 	 * the hammer_mount structure.
598 	 */
599 	error = hammer_init_cursor(trans, &cursor, NULL, NULL);
600 	if (error)
601 		goto failed;
602 	hammer_lock_ex(&hmp->flusher.finalize_lock);
603 	hammer_sync_lock_ex(trans);
604 	hmp->version = ver->cur_version;
605 
606 	/*
607 	 * If upgrading from version < 4 to version >= 4 the UNDO FIFO
608 	 * must be reinitialized.
609 	 */
610 	if (over < HAMMER_VOL_VERSION_FOUR &&
611 	    ver->cur_version >= HAMMER_VOL_VERSION_FOUR) {
612 		kprintf("upgrade undo to version 4\n");
613 		error = hammer_upgrade_undo_4(trans);
614 		if (error)
615 			goto failed;
616 	}
617 
618 	/*
619 	 * Adjust the version in the volume header
620 	 */
621 	volume = hammer_get_root_volume(hmp, &error);
622 	KKASSERT(error == 0);
623 	hammer_modify_volume_field(cursor.trans, volume, vol_version);
624 	volume->ondisk->vol_version = ver->cur_version;
625 	hammer_modify_volume_done(volume);
626 	hammer_rel_volume(volume, 0);
627 
628 	hammer_sync_unlock(trans);
629 	hammer_unlock(&hmp->flusher.finalize_lock);
630 failed:
631 	ver->head.error = error;
632 	hammer_done_cursor(&cursor);
633 	return(0);
634 }
635 
636 /*
637  * Get information
638  */
639 static
640 int
641 hammer_ioc_get_info(hammer_transaction_t trans, struct hammer_ioc_info *info)
642 {
643 	struct hammer_volume_ondisk	*od = trans->hmp->rootvol->ondisk;
644 	struct hammer_mount 		*hm = trans->hmp;
645 
646 	/* Fill the structure with the necessary information */
647 	_hammer_checkspace(hm, HAMMER_CHKSPC_WRITE, &info->rsvbigblocks);
648 	info->rsvbigblocks = info->rsvbigblocks >> HAMMER_BIGBLOCK_BITS;
649 	strlcpy(info->vol_name, od->vol_name, sizeof(od->vol_name));
650 
651 	info->vol_fsid = hm->fsid;
652 	info->vol_fstype = od->vol_fstype;
653 	info->version = hm->version;
654 
655 	info->inodes = od->vol0_stat_inodes;
656 	info->bigblocks = od->vol0_stat_bigblocks;
657 	info->freebigblocks = od->vol0_stat_freebigblocks;
658 	info->nvolumes = hm->nvolumes;
659 
660 	return 0;
661 }
662 
663 /*
664  * Add a snapshot transction id(s) to the list of snapshots.
665  *
666  * NOTE: Records are created with an allocated TID.  If a flush cycle
667  *	 is in progress the record may be synced in the current flush
668  *	 cycle and the volume header will reflect the allocation of the
669  *	 TID, but the synchronization point may not catch up to the
670  *	 TID until the next flush cycle.
671  */
672 static
673 int
674 hammer_ioc_add_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
675 			struct hammer_ioc_snapshot *snap)
676 {
677 	hammer_mount_t hmp = ip->hmp;
678 	struct hammer_btree_leaf_elm leaf;
679 	struct hammer_cursor cursor;
680 	int error;
681 
682 	/*
683 	 * Validate structure
684 	 */
685 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
686 		return (EINVAL);
687 	if (snap->index >= snap->count)
688 		return (EINVAL);
689 
690 	hammer_lock_ex(&hmp->snapshot_lock);
691 again:
692 	/*
693 	 * Look for keys starting after the previous iteration, or at
694 	 * the beginning if snap->count is 0.
695 	 */
696 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
697 	if (error) {
698 		hammer_done_cursor(&cursor);
699 		return(error);
700 	}
701 
702 	cursor.asof = HAMMER_MAX_TID;
703 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
704 
705 	bzero(&leaf, sizeof(leaf));
706 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
707 	leaf.base.rec_type = HAMMER_RECTYPE_SNAPSHOT;
708 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
709 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
710 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
711 	leaf.data_len = sizeof(struct hammer_snapshot_data);
712 
713 	while (snap->index < snap->count) {
714 		leaf.base.key = (int64_t)snap->snaps[snap->index].tid;
715 		cursor.key_beg = leaf.base;
716 		error = hammer_btree_lookup(&cursor);
717 		if (error == 0) {
718 			error = EEXIST;
719 			break;
720 		}
721 
722 		/*
723 		 * NOTE: Must reload key_beg after an ASOF search because
724 		 *	 the create_tid may have been modified during the
725 		 *	 search.
726 		 */
727 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
728 		cursor.key_beg = leaf.base;
729 		error = hammer_create_at_cursor(&cursor, &leaf,
730 						&snap->snaps[snap->index],
731 						HAMMER_CREATE_MODE_SYS);
732 		if (error == EDEADLK) {
733 			hammer_done_cursor(&cursor);
734 			goto again;
735 		}
736 		cursor.flags |= HAMMER_CURSOR_ASOF;
737 		if (error)
738 			break;
739 		++snap->index;
740 	}
741 	snap->head.error = error;
742 	hammer_done_cursor(&cursor);
743 	hammer_unlock(&hmp->snapshot_lock);
744 	return(0);
745 }
746 
747 /*
748  * Delete snapshot transaction id(s) from the list of snapshots.
749  */
750 static
751 int
752 hammer_ioc_del_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
753 			struct hammer_ioc_snapshot *snap)
754 {
755 	hammer_mount_t hmp = ip->hmp;
756 	struct hammer_cursor cursor;
757 	int error;
758 
759 	/*
760 	 * Validate structure
761 	 */
762 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
763 		return (EINVAL);
764 	if (snap->index >= snap->count)
765 		return (EINVAL);
766 
767 	hammer_lock_ex(&hmp->snapshot_lock);
768 again:
769 	/*
770 	 * Look for keys starting after the previous iteration, or at
771 	 * the beginning if snap->count is 0.
772 	 */
773 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
774 	if (error) {
775 		hammer_done_cursor(&cursor);
776 		return(error);
777 	}
778 
779 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
780 	cursor.key_beg.create_tid = 0;
781 	cursor.key_beg.delete_tid = 0;
782 	cursor.key_beg.obj_type = 0;
783 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
784 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
785 	cursor.asof = HAMMER_MAX_TID;
786 	cursor.flags |= HAMMER_CURSOR_ASOF;
787 
788 	while (snap->index < snap->count) {
789 		cursor.key_beg.key = (int64_t)snap->snaps[snap->index].tid;
790 		error = hammer_btree_lookup(&cursor);
791 		if (error)
792 			break;
793 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
794 		if (error)
795 			break;
796 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
797 						0, 0, 0, NULL);
798 		if (error == EDEADLK) {
799 			hammer_done_cursor(&cursor);
800 			goto again;
801 		}
802 		if (error)
803 			break;
804 		++snap->index;
805 	}
806 	snap->head.error = error;
807 	hammer_done_cursor(&cursor);
808 	hammer_unlock(&hmp->snapshot_lock);
809 	return(0);
810 }
811 
812 /*
813  * Retrieve as many snapshot ids as possible or until the array is
814  * full, starting after the last transction id passed in.  If count
815  * is 0 we retrieve starting at the beginning.
816  *
817  * NOTE: Because the b-tree key field is signed but transaction ids
818  *       are unsigned the returned list will be signed-sorted instead
819  *	 of unsigned sorted.  The Caller must still sort the aggregate
820  *	 results.
821  */
822 static
823 int
824 hammer_ioc_get_snapshot(hammer_transaction_t trans, hammer_inode_t ip,
825 			struct hammer_ioc_snapshot *snap)
826 {
827 	struct hammer_cursor cursor;
828 	int error;
829 
830 	/*
831 	 * Validate structure
832 	 */
833 	if (snap->index != 0)
834 		return (EINVAL);
835 	if (snap->count > HAMMER_SNAPS_PER_IOCTL)
836 		return (EINVAL);
837 
838 	/*
839 	 * Look for keys starting after the previous iteration, or at
840 	 * the beginning if snap->count is 0.
841 	 */
842 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
843 	if (error) {
844 		hammer_done_cursor(&cursor);
845 		return(error);
846 	}
847 
848 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
849 	cursor.key_beg.create_tid = 0;
850 	cursor.key_beg.delete_tid = 0;
851 	cursor.key_beg.obj_type = 0;
852 	cursor.key_beg.rec_type = HAMMER_RECTYPE_SNAPSHOT;
853 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
854 	if (snap->count == 0)
855 		cursor.key_beg.key = HAMMER_MIN_KEY;
856 	else
857 		cursor.key_beg.key = (int64_t)snap->snaps[snap->count - 1].tid + 1;
858 
859 	cursor.key_end = cursor.key_beg;
860 	cursor.key_end.key = HAMMER_MAX_KEY;
861 	cursor.asof = HAMMER_MAX_TID;
862 	cursor.flags |= HAMMER_CURSOR_END_EXCLUSIVE | HAMMER_CURSOR_ASOF;
863 
864 	snap->count = 0;
865 
866 	error = hammer_btree_first(&cursor);
867 	while (error == 0 && snap->count < HAMMER_SNAPS_PER_IOCTL) {
868 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF);
869 		if (error)
870 			break;
871 		if (cursor.leaf->base.rec_type == HAMMER_RECTYPE_SNAPSHOT) {
872 			error = hammer_btree_extract(
873 					     &cursor, HAMMER_CURSOR_GET_LEAF |
874 						      HAMMER_CURSOR_GET_DATA);
875 			snap->snaps[snap->count] = cursor.data->snap;
876 
877 			/*
878 			 * The snap data tid should match the key but might
879 			 * not due to a bug in the HAMMER v3 conversion code.
880 			 *
881 			 * This error will work itself out over time but we
882 			 * have to force a match or the snapshot will not
883 			 * be deletable.
884 			 */
885 			if (cursor.data->snap.tid !=
886 			    (hammer_tid_t)cursor.leaf->base.key) {
887 				kprintf("HAMMER: lo=%08x snapshot key "
888 					"0x%016jx data mismatch 0x%016jx\n",
889 					cursor.key_beg.localization,
890 					(uintmax_t)cursor.data->snap.tid,
891 					cursor.leaf->base.key);
892 				kprintf("HAMMER: Probably left over from the "
893 					"original v3 conversion, hammer "
894 					"cleanup should get it eventually\n");
895 				snap->snaps[snap->count].tid =
896 					cursor.leaf->base.key;
897 			}
898 			++snap->count;
899 		}
900 		error = hammer_btree_iterate(&cursor);
901 	}
902 
903 	if (error == ENOENT) {
904 		snap->head.flags |= HAMMER_IOC_SNAPSHOT_EOF;
905 		error = 0;
906 	}
907 	snap->head.error = error;
908 	hammer_done_cursor(&cursor);
909 	return(0);
910 }
911 
912 /*
913  * Retrieve the PFS hammer cleanup utility config record.  This is
914  * different (newer than) the PFS config.
915  */
916 static
917 int
918 hammer_ioc_get_config(hammer_transaction_t trans, hammer_inode_t ip,
919 			struct hammer_ioc_config *config)
920 {
921 	struct hammer_cursor cursor;
922 	int error;
923 
924 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
925 	if (error) {
926 		hammer_done_cursor(&cursor);
927 		return(error);
928 	}
929 
930 	cursor.key_beg.obj_id = HAMMER_OBJID_ROOT;
931 	cursor.key_beg.create_tid = 0;
932 	cursor.key_beg.delete_tid = 0;
933 	cursor.key_beg.obj_type = 0;
934 	cursor.key_beg.rec_type = HAMMER_RECTYPE_CONFIG;
935 	cursor.key_beg.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
936 	cursor.key_beg.key = 0;		/* config space page 0 */
937 
938 	cursor.asof = HAMMER_MAX_TID;
939 	cursor.flags |= HAMMER_CURSOR_ASOF;
940 
941 	error = hammer_btree_lookup(&cursor);
942 	if (error == 0) {
943 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
944 						      HAMMER_CURSOR_GET_DATA);
945 		if (error == 0)
946 			config->config = cursor.data->config;
947 	}
948 	/* error can be ENOENT */
949 	config->head.error = error;
950 	hammer_done_cursor(&cursor);
951 	return(0);
952 }
953 
954 /*
955  * Retrieve the PFS hammer cleanup utility config record.  This is
956  * different (newer than) the PFS config.
957  *
958  * This is kinda a hack.
959  */
960 static
961 int
962 hammer_ioc_set_config(hammer_transaction_t trans, hammer_inode_t ip,
963 			struct hammer_ioc_config *config)
964 {
965 	struct hammer_btree_leaf_elm leaf;
966 	struct hammer_cursor cursor;
967 	hammer_mount_t hmp = ip->hmp;
968 	int error;
969 
970 again:
971 	error = hammer_init_cursor(trans, &cursor, &ip->cache[0], NULL);
972 	if (error) {
973 		hammer_done_cursor(&cursor);
974 		return(error);
975 	}
976 
977 	bzero(&leaf, sizeof(leaf));
978 	leaf.base.obj_id = HAMMER_OBJID_ROOT;
979 	leaf.base.rec_type = HAMMER_RECTYPE_CONFIG;
980 	leaf.base.create_tid = hammer_alloc_tid(hmp, 1);
981 	leaf.base.btype = HAMMER_BTREE_TYPE_RECORD;
982 	leaf.base.localization = ip->obj_localization + HAMMER_LOCALIZE_INODE;
983 	leaf.base.key = 0;	/* page 0 */
984 	leaf.data_len = sizeof(struct hammer_config_data);
985 
986 	cursor.key_beg = leaf.base;
987 
988 	cursor.asof = HAMMER_MAX_TID;
989 	cursor.flags |= HAMMER_CURSOR_BACKEND | HAMMER_CURSOR_ASOF;
990 
991 	error = hammer_btree_lookup(&cursor);
992 	if (error == 0) {
993 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
994 						      HAMMER_CURSOR_GET_DATA);
995 		error = hammer_delete_at_cursor(&cursor, HAMMER_DELETE_DESTROY,
996 						0, 0, 0, NULL);
997 		if (error == EDEADLK) {
998 			hammer_done_cursor(&cursor);
999 			goto again;
1000 		}
1001 	}
1002 	if (error == ENOENT)
1003 		error = 0;
1004 	if (error == 0) {
1005 		/*
1006 		 * NOTE: Must reload key_beg after an ASOF search because
1007 		 *	 the create_tid may have been modified during the
1008 		 *	 search.
1009 		 */
1010 		cursor.flags &= ~HAMMER_CURSOR_ASOF;
1011 		cursor.key_beg = leaf.base;
1012 		error = hammer_create_at_cursor(&cursor, &leaf,
1013 						&config->config,
1014 						HAMMER_CREATE_MODE_SYS);
1015 		if (error == EDEADLK) {
1016 			hammer_done_cursor(&cursor);
1017 			goto again;
1018 		}
1019 	}
1020 	config->head.error = error;
1021 	hammer_done_cursor(&cursor);
1022 	return(0);
1023 }
1024 
1025 static
1026 int
1027 hammer_ioc_get_data(hammer_transaction_t trans, hammer_inode_t ip,
1028 			struct hammer_ioc_data *data)
1029 {
1030 	struct hammer_cursor cursor;
1031 	int bytes;
1032 	int error;
1033 
1034 	/* XXX cached inode ? */
1035 	error = hammer_init_cursor(trans, &cursor, NULL, NULL);
1036 	if (error)
1037 		goto failed;
1038 
1039 	cursor.key_beg = data->elm;
1040 	cursor.flags |= HAMMER_CURSOR_BACKEND;
1041 
1042 	error = hammer_btree_lookup(&cursor);
1043 	if (error == 0) {
1044 		error = hammer_btree_extract(&cursor, HAMMER_CURSOR_GET_LEAF |
1045 						      HAMMER_CURSOR_GET_DATA);
1046 		if (error == 0) {
1047 			data->leaf = *cursor.leaf;
1048 			bytes = cursor.leaf->data_len;
1049 			if (bytes > data->size)
1050 				bytes = data->size;
1051 			error = copyout(cursor.data, data->ubuf, bytes);
1052 		}
1053 	}
1054 
1055 failed:
1056 	hammer_done_cursor(&cursor);
1057 	return (error);
1058 }
1059