Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
[dragonfly.git] / sys / vfs / hammer / hammer_prune.c
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_prune.c,v 1.19 2008/09/23 21:03:52 dillon Exp $
35  */
36
37 #include "hammer.h"
38
39 /*
40  * Iterate through the specified range of object ids and remove any
41  * deleted records that fall entirely within a prune modulo.
42  *
43  * A reverse iteration is used to prevent overlapping records from being
44  * created during the iteration due to alignments.  This also allows us
45  * to adjust alignments without blowing up the B-Tree.
46  */
47 static int prune_should_delete(struct hammer_ioc_prune *prune,
48                                hammer_btree_leaf_elm_t elm);
49 static void prune_check_nlinks(hammer_cursor_t cursor,
50                                hammer_btree_leaf_elm_t elm);
51
52 int
53 hammer_ioc_prune(hammer_transaction_t trans, hammer_inode_t ip,
54                  struct hammer_ioc_prune *prune)
55 {
56         struct hammer_cursor cursor;
57         hammer_btree_leaf_elm_t elm;
58         struct hammer_ioc_prune_elm *copy_elms;
59         struct hammer_ioc_prune_elm *user_elms;
60         int error;
61         int isdir;
62         int elm_array_size;
63         int seq;
64
65         if (prune->nelms < 0 || prune->nelms > HAMMER_MAX_PRUNE_ELMS)
66                 return(EINVAL);
67         if ((prune->key_beg.localization | prune->key_end.localization) &
68             HAMMER_LOCALIZE_PSEUDOFS_MASK) {
69                 return(EINVAL);
70         }
71         if (prune->key_beg.localization > prune->key_end.localization)
72                 return(EINVAL);
73         if (prune->key_beg.localization == prune->key_end.localization) {
74                 if (prune->key_beg.obj_id > prune->key_end.obj_id)
75                         return(EINVAL);
76                 /* key-space limitations - no check needed */
77         }
78         if ((prune->head.flags & HAMMER_IOC_PRUNE_ALL) && prune->nelms)
79                 return(EINVAL);
80
81         prune->key_cur.localization = (prune->key_end.localization &
82                                         HAMMER_LOCALIZE_MASK) +
83                                       ip->obj_localization;
84         prune->key_cur.obj_id = prune->key_end.obj_id;
85         prune->key_cur.key = HAMMER_MAX_KEY;
86
87         /*
88          * Copy element array from userland
89          */
90         elm_array_size = sizeof(*copy_elms) * prune->nelms;
91         user_elms = prune->elms;
92         copy_elms = kmalloc(elm_array_size, M_TEMP, M_WAITOK);
93         if ((error = copyin(user_elms, copy_elms, elm_array_size)) != 0)
94                 goto failed;
95         prune->elms = copy_elms;
96
97         seq = trans->hmp->flusher.act;
98
99         /*
100          * Scan backwards.  Retries typically occur if a deadlock is detected.
101          */
102 retry:
103         error = hammer_init_cursor(trans, &cursor, NULL, NULL);
104         if (error) {
105                 hammer_done_cursor(&cursor);
106                 goto failed;
107         }
108         cursor.key_beg.localization = (prune->key_beg.localization &
109                                         HAMMER_LOCALIZE_MASK) +
110                                       ip->obj_localization;
111         cursor.key_beg.obj_id = prune->key_beg.obj_id;
112         cursor.key_beg.key = HAMMER_MIN_KEY;
113         cursor.key_beg.create_tid = 1;
114         cursor.key_beg.delete_tid = 0;
115         cursor.key_beg.rec_type = HAMMER_MIN_RECTYPE;
116         cursor.key_beg.obj_type = 0;
117
118         cursor.key_end.localization = prune->key_cur.localization;
119         cursor.key_end.obj_id = prune->key_cur.obj_id;
120         cursor.key_end.key = prune->key_cur.key;
121         cursor.key_end.create_tid = HAMMER_MAX_TID - 1;
122         cursor.key_end.delete_tid = 0;
123         cursor.key_end.rec_type = HAMMER_MAX_RECTYPE;
124         cursor.key_end.obj_type = 0;
125
126         cursor.flags |= HAMMER_CURSOR_END_INCLUSIVE;
127         cursor.flags |= HAMMER_CURSOR_BACKEND;
128
129         /*
130          * This flag allows the B-Tree code to clean up loose ends.  At
131          * the moment (XXX) it also means we have to hold the sync lock
132          * through the iteration.
133          */
134         cursor.flags |= HAMMER_CURSOR_PRUNING;
135
136         hammer_sync_lock_sh(trans);
137         error = hammer_btree_last(&cursor);
138         hammer_sync_unlock(trans);
139
140         while (error == 0) {
141                 /*
142                  * Check for work
143                  */
144                 elm = &cursor.node->ondisk->elms[cursor.index].leaf;
145                 prune->key_cur = elm->base;
146
147                 /*
148                  * Yield to more important tasks
149                  */
150                 if ((error = hammer_signal_check(trans->hmp)) != 0)
151                         break;
152
153                 if (prune->stat_oldest_tid > elm->base.create_tid)
154                         prune->stat_oldest_tid = elm->base.create_tid;
155
156                 if (hammer_debug_general & 0x0200) {
157                         kprintf("check %016llx %016llx cre=%016llx del=%016llx\n",
158                                         (long long)elm->base.obj_id,
159                                         (long long)elm->base.key,
160                                         (long long)elm->base.create_tid,
161                                         (long long)elm->base.delete_tid);
162                 }
163                                 
164                 if (prune_should_delete(prune, elm)) {
165                         if (hammer_debug_general & 0x0200) {
166                                 kprintf("check %016llx %016llx: DELETE\n",
167                                         (long long)elm->base.obj_id,
168                                         (long long)elm->base.key);
169                         }
170
171                         /*
172                          * NOTE: This can return EDEADLK
173                          *
174                          * Acquiring the sync lock guarantees that the
175                          * operation will not cross a synchronization
176                          * boundary (see the flusher).
177                          *
178                          * We dont need to track inodes or next_tid when
179                          * we are destroying deleted records.
180                          */
181                         isdir = (elm->base.rec_type == HAMMER_RECTYPE_DIRENTRY);
182
183                         hammer_sync_lock_sh(trans);
184                         error = hammer_delete_at_cursor(&cursor,
185                                                         HAMMER_DELETE_DESTROY,
186                                                         cursor.trans->tid,
187                                                         cursor.trans->time32,
188                                                         0, &prune->stat_bytes);
189                         hammer_sync_unlock(trans);
190                         if (error)
191                                 break;
192
193                         if (isdir)
194                                 ++prune->stat_dirrecords;
195                         else
196                                 ++prune->stat_rawrecords;
197
198                         /*
199                          * The current record might now be the one after
200                          * the one we deleted, set ATEDISK to force us
201                          * to skip it (since we are iterating backwards).
202                          */
203                         cursor.flags |= HAMMER_CURSOR_ATEDISK;
204                 } else {
205                         /*
206                          * Nothing to delete, but we may have to check other
207                          * things.
208                          */
209                         prune_check_nlinks(&cursor, elm);
210                         cursor.flags |= HAMMER_CURSOR_ATEDISK;
211                         if (hammer_debug_general & 0x0100) {
212                                 kprintf("check %016llx %016llx: SKIP\n",
213                                         (long long)elm->base.obj_id,
214                                         (long long)elm->base.key);
215                         }
216                 }
217                 ++prune->stat_scanrecords;
218
219                 while (hammer_flusher_meta_halflimit(trans->hmp) ||
220                        hammer_flusher_undo_exhausted(trans, 2)) {
221                         hammer_unlock_cursor(&cursor);
222                         hammer_flusher_wait(trans->hmp, seq);
223                         hammer_lock_cursor(&cursor);
224                         seq = hammer_flusher_async_one(trans->hmp);
225                 }
226                 hammer_sync_lock_sh(trans);
227                 error = hammer_btree_iterate_reverse(&cursor);
228                 hammer_sync_unlock(trans);
229         }
230         if (error == ENOENT)
231                 error = 0;
232         hammer_done_cursor(&cursor);
233         if (error == EDEADLK)
234                 goto retry;
235         if (error == EINTR) {
236                 prune->head.flags |= HAMMER_IOC_HEAD_INTR;
237                 error = 0;
238         }
239 failed:
240         prune->key_cur.localization &= HAMMER_LOCALIZE_MASK;
241         prune->elms = user_elms;
242         kfree(copy_elms, M_TEMP);
243         return(error);
244 }
245
246 /*
247  * Check pruning list.  The list must be sorted in descending order.
248  *
249  * Return non-zero if the record should be deleted.
250  */
251 static int
252 prune_should_delete(struct hammer_ioc_prune *prune, hammer_btree_leaf_elm_t elm)
253 {
254         struct hammer_ioc_prune_elm *scan;
255         int i;
256
257         /*
258          * If pruning everything remove all records with a non-zero
259          * delete_tid.
260          */
261         if (prune->head.flags & HAMMER_IOC_PRUNE_ALL) {
262                 if (elm->base.delete_tid != 0)
263                         return(1);
264                 return(0);
265         }
266
267         for (i = 0; i < prune->nelms; ++i) {
268                 scan = &prune->elms[i];
269
270                 /*
271                  * Check for loop termination.
272                  */
273                 if (elm->base.create_tid >= scan->end_tid ||
274                     elm->base.delete_tid > scan->end_tid) {
275                         break;
276                 }
277
278                 /*
279                  * Determine if we can delete the record.
280                  */
281                 if (elm->base.delete_tid &&
282                     elm->base.create_tid >= scan->beg_tid &&
283                     elm->base.delete_tid <= scan->end_tid &&
284                     (elm->base.create_tid - scan->beg_tid) / scan->mod_tid ==
285                     (elm->base.delete_tid - scan->beg_tid) / scan->mod_tid) {
286                         return(1);
287                 }
288         }
289         return(0);
290 }
291
292 /*
293  * Dangling inodes can occur if processes are holding open descriptors on
294  * deleted files as-of when a machine crashes.  When we find one simply
295  * acquire the inode and release it.  The inode handling code will then
296  * do the right thing.
297  */
298 static
299 void
300 prune_check_nlinks(hammer_cursor_t cursor, hammer_btree_leaf_elm_t elm)
301 {
302         hammer_inode_t ip;
303         int error;
304
305         if (elm->base.rec_type != HAMMER_RECTYPE_INODE)
306                 return;
307         if (elm->base.delete_tid != 0)
308                 return;
309         if (hammer_btree_extract(cursor, HAMMER_CURSOR_GET_DATA))
310                 return;
311         if (cursor->data->inode.nlinks)
312                 return;
313         hammer_cursor_downgrade(cursor);
314         ip = hammer_get_inode(cursor->trans, NULL, elm->base.obj_id,
315                       HAMMER_MAX_TID,
316                       elm->base.localization & HAMMER_LOCALIZE_PSEUDOFS_MASK,
317                       0, &error);
318         if (ip) {
319                 if (hammer_debug_general & 0x0001) {
320                         kprintf("pruning disconnected inode %016llx\n",
321                                 (long long)elm->base.obj_id);
322                 }
323                 hammer_rel_inode(ip, 0);
324                 hammer_inode_waitreclaims(cursor->trans->hmp);
325         } else {
326                 kprintf("unable to prune disconnected inode %016llx\n",
327                         (long long)elm->base.obj_id);
328         }
329 }
330
331 #if 0
332
333 /*
334  * NOTE: THIS CODE HAS BEEN REMOVED!  Pruning no longer attempts to realign
335  *       adjacent records because it seriously interferes with every 
336  *       mirroring algorithm I could come up with.
337  *
338  *       This means that historical accesses beyond the first snapshot
339  *       softlink should be on snapshot boundaries only.  Historical
340  *       accesses from "now" to the first snapshot softlink continue to
341  *       be fine-grained.
342  *
343  * NOTE: It also looks like there's a bug in the removed code.  It is believed
344  *       that create_tid can sometimes get set to 0xffffffffffffffff.  Just as
345  *       well we no longer try to do this fancy shit.  Probably the attempt to
346  *       correct the rhb is blowing up the cursor's indexing or addressing mapping.
347  *
348  * Align the record to cover any gaps created through the deletion of
349  * records within the pruning space.  If we were to just delete the records
350  * there would be gaps which in turn would cause a snapshot that is NOT on
351  * a pruning boundary to appear corrupt to the user.  Forcing alignment
352  * of the create_tid and delete_tid for retained records 'reconnects'
353  * the previously contiguous space, making it contiguous again after the
354  * deletions.
355  *
356  * The use of a reverse iteration allows us to safely align the records and
357  * related elements without creating temporary overlaps.  XXX we should
358  * add ordering dependancies for record buffers to guarantee consistency
359  * during recovery.
360  */
361 static int
362 realign_prune(struct hammer_ioc_prune *prune,
363               hammer_cursor_t cursor, int realign_cre, int realign_del)
364 {
365         struct hammer_ioc_prune_elm *scan;
366         hammer_btree_elm_t elm;
367         hammer_tid_t delta;
368         hammer_tid_t tid;
369         int error;
370
371         hammer_cursor_downgrade(cursor);
372
373         elm = &cursor->node->ondisk->elms[cursor->index];
374         ++prune->stat_realignments;
375
376         /*
377          * Align the create_tid.  By doing a reverse iteration we guarantee
378          * that all records after our current record have already been
379          * aligned, allowing us to safely correct the right-hand-boundary
380          * (because no record to our right is otherwise exactly matching
381          * will have a create_tid to the left of our aligned create_tid).
382          */
383         error = 0;
384         if (realign_cre >= 0) {
385                 scan = &prune->elms[realign_cre];
386
387                 delta = (elm->leaf.base.create_tid - scan->beg_tid) % 
388                         scan->mod_tid;
389                 if (delta) {
390                         tid = elm->leaf.base.create_tid - delta + scan->mod_tid;
391
392                         /* can EDEADLK */
393                         error = hammer_btree_correct_rhb(cursor, tid + 1);
394                         if (error == 0) {
395                                 error = hammer_btree_extract(cursor,
396                                                      HAMMER_CURSOR_GET_LEAF);
397                         }
398                         if (error == 0) {
399                                 /* can EDEADLK */
400                                 error = hammer_cursor_upgrade(cursor);
401                         }
402                         if (error == 0) {
403                                 hammer_modify_node(cursor->trans, cursor->node,
404                                             &elm->leaf.base.create_tid,
405                                             sizeof(elm->leaf.base.create_tid));
406                                 elm->leaf.base.create_tid = tid;
407                                 hammer_modify_node_done(cursor->node);
408                         }
409                 }
410         }
411
412         /*
413          * Align the delete_tid.  This only occurs if the record is historical
414          * was deleted at some point.  Realigning the delete_tid does not
415          * move the record within the B-Tree but may cause it to temporarily
416          * overlap a record that has not yet been pruned.
417          */
418         if (error == 0 && realign_del >= 0) {
419                 scan = &prune->elms[realign_del];
420
421                 delta = (elm->leaf.base.delete_tid - scan->beg_tid) % 
422                         scan->mod_tid;
423                 if (delta) {
424                         error = hammer_btree_extract(cursor,
425                                                      HAMMER_CURSOR_GET_LEAF);
426                         if (error == 0) {
427                                 hammer_modify_node(cursor->trans, cursor->node,
428                                             &elm->leaf.base.delete_tid,
429                                             sizeof(elm->leaf.base.delete_tid));
430                                 elm->leaf.base.delete_tid =
431                                             elm->leaf.base.delete_tid -
432                                             delta + scan->mod_tid;
433                                 hammer_modify_node_done(cursor->node);
434                         }
435                 }
436         }
437         return (error);
438 }
439
440 #endif