blob: f47c1528eb9a69aeba5f2cd0740333a05aa9238d [file] [log] [blame]
David Sterbac1d7c512018-04-03 19:23:33 +02001// SPDX-License-Identifier: GPL-2.0
Jan Schmidta542ad12011-06-13 19:52:59 +02002/*
3 * Copyright (C) 2011 STRATO. All rights reserved.
Jan Schmidta542ad12011-06-13 19:52:59 +02004 */
5
David Sterbaf54de062017-05-31 19:32:09 +02006#include <linux/mm.h>
Lu Fengqiafce7722016-06-13 09:36:46 +08007#include <linux/rbtree.h>
Jeff Mahoney00142752017-07-12 16:20:08 -06008#include <trace/events/btrfs.h>
Jan Schmidta542ad12011-06-13 19:52:59 +02009#include "ctree.h"
10#include "disk-io.h"
11#include "backref.h"
Jan Schmidt8da6d582011-11-23 18:55:04 +010012#include "ulist.h"
13#include "transaction.h"
14#include "delayed-ref.h"
Jan Schmidtb916a592012-04-13 12:28:08 +020015#include "locking.h"
Qu Wenruo1b60d2e2020-03-23 16:08:34 +080016#include "misc.h"
Jan Schmidta542ad12011-06-13 19:52:59 +020017
Josef Bacikdc046b12014-09-10 16:20:45 -040018/* Just an arbitrary number so we can be sure this happened */
19#define BACKREF_FOUND_SHARED 6
20
Jan Schmidt976b1902012-05-17 16:43:03 +020021struct extent_inode_elem {
22 u64 inum;
23 u64 offset;
24 struct extent_inode_elem *next;
25};
26
Jeff Mahoney73980be2017-06-28 21:56:55 -060027static int check_extent_in_eb(const struct btrfs_key *key,
28 const struct extent_buffer *eb,
29 const struct btrfs_file_extent_item *fi,
30 u64 extent_item_pos,
Zygo Blaxellc995ab32017-09-22 13:58:45 -040031 struct extent_inode_elem **eie,
32 bool ignore_offset)
Jan Schmidt976b1902012-05-17 16:43:03 +020033{
Josef Bacik8ca15e02013-07-05 13:58:19 -040034 u64 offset = 0;
Jan Schmidt976b1902012-05-17 16:43:03 +020035 struct extent_inode_elem *e;
36
Zygo Blaxellc995ab32017-09-22 13:58:45 -040037 if (!ignore_offset &&
38 !btrfs_file_extent_compression(eb, fi) &&
Josef Bacik8ca15e02013-07-05 13:58:19 -040039 !btrfs_file_extent_encryption(eb, fi) &&
40 !btrfs_file_extent_other_encoding(eb, fi)) {
41 u64 data_offset;
42 u64 data_len;
Jan Schmidt976b1902012-05-17 16:43:03 +020043
Josef Bacik8ca15e02013-07-05 13:58:19 -040044 data_offset = btrfs_file_extent_offset(eb, fi);
45 data_len = btrfs_file_extent_num_bytes(eb, fi);
46
47 if (extent_item_pos < data_offset ||
48 extent_item_pos >= data_offset + data_len)
49 return 1;
50 offset = extent_item_pos - data_offset;
51 }
Jan Schmidt976b1902012-05-17 16:43:03 +020052
53 e = kmalloc(sizeof(*e), GFP_NOFS);
54 if (!e)
55 return -ENOMEM;
56
57 e->next = *eie;
58 e->inum = key->objectid;
Josef Bacik8ca15e02013-07-05 13:58:19 -040059 e->offset = key->offset + offset;
Jan Schmidt976b1902012-05-17 16:43:03 +020060 *eie = e;
61
62 return 0;
63}
64
Wang Shilongf05c4742014-01-28 19:13:38 +080065static void free_inode_elem_list(struct extent_inode_elem *eie)
66{
67 struct extent_inode_elem *eie_next;
68
69 for (; eie; eie = eie_next) {
70 eie_next = eie->next;
71 kfree(eie);
72 }
73}
74
Jeff Mahoney73980be2017-06-28 21:56:55 -060075static int find_extent_in_eb(const struct extent_buffer *eb,
76 u64 wanted_disk_byte, u64 extent_item_pos,
Zygo Blaxellc995ab32017-09-22 13:58:45 -040077 struct extent_inode_elem **eie,
78 bool ignore_offset)
Jan Schmidt976b1902012-05-17 16:43:03 +020079{
80 u64 disk_byte;
81 struct btrfs_key key;
82 struct btrfs_file_extent_item *fi;
83 int slot;
84 int nritems;
85 int extent_type;
86 int ret;
87
88 /*
89 * from the shared data ref, we only have the leaf but we need
90 * the key. thus, we must look into all items and see that we
91 * find one (some) with a reference to our extent item.
92 */
93 nritems = btrfs_header_nritems(eb);
94 for (slot = 0; slot < nritems; ++slot) {
95 btrfs_item_key_to_cpu(eb, &key, slot);
96 if (key.type != BTRFS_EXTENT_DATA_KEY)
97 continue;
98 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
99 extent_type = btrfs_file_extent_type(eb, fi);
100 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
101 continue;
102 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
103 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
104 if (disk_byte != wanted_disk_byte)
105 continue;
106
Zygo Blaxellc995ab32017-09-22 13:58:45 -0400107 ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie, ignore_offset);
Jan Schmidt976b1902012-05-17 16:43:03 +0200108 if (ret < 0)
109 return ret;
110 }
111
112 return 0;
113}
114
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600115struct preftree {
Liu Boecf160b2018-08-23 03:51:53 +0800116 struct rb_root_cached root;
Jeff Mahoney6c336b22017-07-12 16:20:07 -0600117 unsigned int count;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600118};
119
Liu Boecf160b2018-08-23 03:51:53 +0800120#define PREFTREE_INIT { .root = RB_ROOT_CACHED, .count = 0 }
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600121
122struct preftrees {
123 struct preftree direct; /* BTRFS_SHARED_[DATA|BLOCK]_REF_KEY */
124 struct preftree indirect; /* BTRFS_[TREE_BLOCK|EXTENT_DATA]_REF_KEY */
125 struct preftree indirect_missing_keys;
126};
127
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600128/*
129 * Checks for a shared extent during backref search.
130 *
131 * The share_count tracks prelim_refs (direct and indirect) having a
132 * ref->count >0:
133 * - incremented when a ref->count transitions to >0
134 * - decremented when a ref->count transitions to <1
135 */
136struct share_check {
137 u64 root_objectid;
138 u64 inum;
139 int share_count;
140};
141
142static inline int extent_is_shared(struct share_check *sc)
143{
144 return (sc && sc->share_count > 1) ? BACKREF_FOUND_SHARED : 0;
145}
146
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800147static struct kmem_cache *btrfs_prelim_ref_cache;
148
149int __init btrfs_prelim_ref_init(void)
150{
151 btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600152 sizeof(struct prelim_ref),
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800153 0,
Nikolay Borisovfba4b692016-06-23 21:17:08 +0300154 SLAB_MEM_SPREAD,
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800155 NULL);
156 if (!btrfs_prelim_ref_cache)
157 return -ENOMEM;
158 return 0;
159}
160
David Sterbae67c7182018-02-19 17:24:18 +0100161void __cold btrfs_prelim_ref_exit(void)
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800162{
Kinglong Mee5598e902016-01-29 21:36:35 +0800163 kmem_cache_destroy(btrfs_prelim_ref_cache);
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800164}
165
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600166static void free_pref(struct prelim_ref *ref)
167{
168 kmem_cache_free(btrfs_prelim_ref_cache, ref);
169}
170
171/*
172 * Return 0 when both refs are for the same block (and can be merged).
173 * A -1 return indicates ref1 is a 'lower' block than ref2, while 1
174 * indicates a 'higher' block.
175 */
176static int prelim_ref_compare(struct prelim_ref *ref1,
177 struct prelim_ref *ref2)
178{
179 if (ref1->level < ref2->level)
180 return -1;
181 if (ref1->level > ref2->level)
182 return 1;
183 if (ref1->root_id < ref2->root_id)
184 return -1;
185 if (ref1->root_id > ref2->root_id)
186 return 1;
187 if (ref1->key_for_search.type < ref2->key_for_search.type)
188 return -1;
189 if (ref1->key_for_search.type > ref2->key_for_search.type)
190 return 1;
191 if (ref1->key_for_search.objectid < ref2->key_for_search.objectid)
192 return -1;
193 if (ref1->key_for_search.objectid > ref2->key_for_search.objectid)
194 return 1;
195 if (ref1->key_for_search.offset < ref2->key_for_search.offset)
196 return -1;
197 if (ref1->key_for_search.offset > ref2->key_for_search.offset)
198 return 1;
199 if (ref1->parent < ref2->parent)
200 return -1;
201 if (ref1->parent > ref2->parent)
202 return 1;
203
204 return 0;
205}
206
Colin Ian Kingccc8dc72017-11-30 12:14:47 +0000207static void update_share_count(struct share_check *sc, int oldcount,
208 int newcount)
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600209{
210 if ((!sc) || (oldcount == 0 && newcount < 1))
211 return;
212
213 if (oldcount > 0 && newcount < 1)
214 sc->share_count--;
215 else if (oldcount < 1 && newcount > 0)
216 sc->share_count++;
217}
218
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600219/*
220 * Add @newref to the @root rbtree, merging identical refs.
221 *
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600222 * Callers should assume that newref has been freed after calling.
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600223 */
Jeff Mahoney00142752017-07-12 16:20:08 -0600224static void prelim_ref_insert(const struct btrfs_fs_info *fs_info,
225 struct preftree *preftree,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600226 struct prelim_ref *newref,
227 struct share_check *sc)
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600228{
Liu Boecf160b2018-08-23 03:51:53 +0800229 struct rb_root_cached *root;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600230 struct rb_node **p;
231 struct rb_node *parent = NULL;
232 struct prelim_ref *ref;
233 int result;
Liu Boecf160b2018-08-23 03:51:53 +0800234 bool leftmost = true;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600235
236 root = &preftree->root;
Liu Boecf160b2018-08-23 03:51:53 +0800237 p = &root->rb_root.rb_node;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600238
239 while (*p) {
240 parent = *p;
241 ref = rb_entry(parent, struct prelim_ref, rbnode);
242 result = prelim_ref_compare(ref, newref);
243 if (result < 0) {
244 p = &(*p)->rb_left;
245 } else if (result > 0) {
246 p = &(*p)->rb_right;
Liu Boecf160b2018-08-23 03:51:53 +0800247 leftmost = false;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600248 } else {
249 /* Identical refs, merge them and free @newref */
250 struct extent_inode_elem *eie = ref->inode_list;
251
252 while (eie && eie->next)
253 eie = eie->next;
254
255 if (!eie)
256 ref->inode_list = newref->inode_list;
257 else
258 eie->next = newref->inode_list;
Jeff Mahoney00142752017-07-12 16:20:08 -0600259 trace_btrfs_prelim_ref_merge(fs_info, ref, newref,
260 preftree->count);
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600261 /*
262 * A delayed ref can have newref->count < 0.
263 * The ref->count is updated to follow any
264 * BTRFS_[ADD|DROP]_DELAYED_REF actions.
265 */
266 update_share_count(sc, ref->count,
267 ref->count + newref->count);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600268 ref->count += newref->count;
269 free_pref(newref);
270 return;
271 }
272 }
273
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600274 update_share_count(sc, 0, newref->count);
Jeff Mahoney6c336b22017-07-12 16:20:07 -0600275 preftree->count++;
Jeff Mahoney00142752017-07-12 16:20:08 -0600276 trace_btrfs_prelim_ref_insert(fs_info, newref, NULL, preftree->count);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600277 rb_link_node(&newref->rbnode, parent, p);
Liu Boecf160b2018-08-23 03:51:53 +0800278 rb_insert_color_cached(&newref->rbnode, root, leftmost);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600279}
280
281/*
282 * Release the entire tree. We don't care about internal consistency so
283 * just free everything and then reset the tree root.
284 */
285static void prelim_release(struct preftree *preftree)
286{
287 struct prelim_ref *ref, *next_ref;
288
Liu Boecf160b2018-08-23 03:51:53 +0800289 rbtree_postorder_for_each_entry_safe(ref, next_ref,
290 &preftree->root.rb_root, rbnode)
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600291 free_pref(ref);
292
Liu Boecf160b2018-08-23 03:51:53 +0800293 preftree->root = RB_ROOT_CACHED;
Jeff Mahoney6c336b22017-07-12 16:20:07 -0600294 preftree->count = 0;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600295}
296
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200297/*
298 * the rules for all callers of this function are:
299 * - obtaining the parent is the goal
300 * - if you add a key, you must know that it is a correct key
301 * - if you cannot add the parent or a correct key, then we will look into the
302 * block later to set a correct key
303 *
304 * delayed refs
305 * ============
306 * backref type | shared | indirect | shared | indirect
307 * information | tree | tree | data | data
308 * --------------------+--------+----------+--------+----------
309 * parent logical | y | - | - | -
310 * key to resolve | - | y | y | y
311 * tree block logical | - | - | - | -
312 * root for resolving | y | y | y | y
313 *
314 * - column 1: we've the parent -> done
315 * - column 2, 3, 4: we use the key to find the parent
316 *
317 * on disk refs (inline or keyed)
318 * ==============================
319 * backref type | shared | indirect | shared | indirect
320 * information | tree | tree | data | data
321 * --------------------+--------+----------+--------+----------
322 * parent logical | y | - | y | -
323 * key to resolve | - | - | - | y
324 * tree block logical | y | y | y | y
325 * root for resolving | - | y | y | y
326 *
327 * - column 1, 3: we've the parent -> done
328 * - column 2: we take the first key from the block to find the parent
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600329 * (see add_missing_keys)
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200330 * - column 4: we use the key to find the parent
331 *
332 * additional information that's available but not required to find the parent
333 * block might help in merging entries to gain some speed.
334 */
Jeff Mahoney00142752017-07-12 16:20:08 -0600335static int add_prelim_ref(const struct btrfs_fs_info *fs_info,
336 struct preftree *preftree, u64 root_id,
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600337 const struct btrfs_key *key, int level, u64 parent,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600338 u64 wanted_disk_byte, int count,
339 struct share_check *sc, gfp_t gfp_mask)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100340{
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600341 struct prelim_ref *ref;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100342
Liu Bo48ec4732013-10-30 13:25:24 +0800343 if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
344 return 0;
345
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800346 ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100347 if (!ref)
348 return -ENOMEM;
349
350 ref->root_id = root_id;
ethanwu7ac8b882020-02-07 17:38:15 +0800351 if (key)
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200352 ref->key_for_search = *key;
ethanwu7ac8b882020-02-07 17:38:15 +0800353 else
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200354 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
Jan Schmidt8da6d582011-11-23 18:55:04 +0100355
Jan Schmidt33019582012-05-30 18:05:21 +0200356 ref->inode_list = NULL;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100357 ref->level = level;
358 ref->count = count;
359 ref->parent = parent;
360 ref->wanted_disk_byte = wanted_disk_byte;
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600361 prelim_ref_insert(fs_info, preftree, ref, sc);
362 return extent_is_shared(sc);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100363}
364
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600365/* direct refs use root == 0, key == NULL */
Jeff Mahoney00142752017-07-12 16:20:08 -0600366static int add_direct_ref(const struct btrfs_fs_info *fs_info,
367 struct preftrees *preftrees, int level, u64 parent,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600368 u64 wanted_disk_byte, int count,
369 struct share_check *sc, gfp_t gfp_mask)
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600370{
Jeff Mahoney00142752017-07-12 16:20:08 -0600371 return add_prelim_ref(fs_info, &preftrees->direct, 0, NULL, level,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600372 parent, wanted_disk_byte, count, sc, gfp_mask);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600373}
374
375/* indirect refs use parent == 0 */
Jeff Mahoney00142752017-07-12 16:20:08 -0600376static int add_indirect_ref(const struct btrfs_fs_info *fs_info,
377 struct preftrees *preftrees, u64 root_id,
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600378 const struct btrfs_key *key, int level,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600379 u64 wanted_disk_byte, int count,
380 struct share_check *sc, gfp_t gfp_mask)
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600381{
382 struct preftree *tree = &preftrees->indirect;
383
384 if (!key)
385 tree = &preftrees->indirect_missing_keys;
Jeff Mahoney00142752017-07-12 16:20:08 -0600386 return add_prelim_ref(fs_info, tree, root_id, key, level, 0,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600387 wanted_disk_byte, count, sc, gfp_mask);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600388}
389
ethanwued58f2e2020-02-07 17:38:16 +0800390static int is_shared_data_backref(struct preftrees *preftrees, u64 bytenr)
391{
392 struct rb_node **p = &preftrees->direct.root.rb_root.rb_node;
393 struct rb_node *parent = NULL;
394 struct prelim_ref *ref = NULL;
Arnd Bergmann9c6c7232020-04-29 15:27:32 +0200395 struct prelim_ref target = {};
ethanwued58f2e2020-02-07 17:38:16 +0800396 int result;
397
398 target.parent = bytenr;
399
400 while (*p) {
401 parent = *p;
402 ref = rb_entry(parent, struct prelim_ref, rbnode);
403 result = prelim_ref_compare(ref, &target);
404
405 if (result < 0)
406 p = &(*p)->rb_left;
407 else if (result > 0)
408 p = &(*p)->rb_right;
409 else
410 return 1;
411 }
412 return 0;
413}
414
Jan Schmidt8da6d582011-11-23 18:55:04 +0100415static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
ethanwued58f2e2020-02-07 17:38:16 +0800416 struct ulist *parents,
417 struct preftrees *preftrees, struct prelim_ref *ref,
Josef Bacik44853862014-03-19 13:35:14 -0400418 int level, u64 time_seq, const u64 *extent_item_pos,
ethanwub25b0b82020-02-07 17:38:18 +0800419 bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100420{
Alexander Block69bca402012-06-19 07:42:26 -0600421 int ret = 0;
422 int slot;
423 struct extent_buffer *eb;
424 struct btrfs_key key;
Josef Bacik7ef81ac2014-01-24 14:05:42 -0500425 struct btrfs_key *key_for_search = &ref->key_for_search;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100426 struct btrfs_file_extent_item *fi;
Josef Baciked8c4912013-07-05 14:03:47 -0400427 struct extent_inode_elem *eie = NULL, *old = NULL;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100428 u64 disk_byte;
Josef Bacik7ef81ac2014-01-24 14:05:42 -0500429 u64 wanted_disk_byte = ref->wanted_disk_byte;
430 u64 count = 0;
ethanwu7ac8b882020-02-07 17:38:15 +0800431 u64 data_offset;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100432
Alexander Block69bca402012-06-19 07:42:26 -0600433 if (level != 0) {
434 eb = path->nodes[level];
435 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
Jan Schmidt33019582012-05-30 18:05:21 +0200436 if (ret < 0)
437 return ret;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100438 return 0;
Alexander Block69bca402012-06-19 07:42:26 -0600439 }
Jan Schmidt8da6d582011-11-23 18:55:04 +0100440
441 /*
ethanwued58f2e2020-02-07 17:38:16 +0800442 * 1. We normally enter this function with the path already pointing to
443 * the first item to check. But sometimes, we may enter it with
444 * slot == nritems.
445 * 2. We are searching for normal backref but bytenr of this leaf
446 * matches shared data backref
ethanwucfc0eed2020-02-07 17:38:17 +0800447 * 3. The leaf owner is not equal to the root we are searching
448 *
ethanwued58f2e2020-02-07 17:38:16 +0800449 * For these cases, go to the next leaf before we continue.
Jan Schmidt8da6d582011-11-23 18:55:04 +0100450 */
ethanwued58f2e2020-02-07 17:38:16 +0800451 eb = path->nodes[0];
452 if (path->slots[0] >= btrfs_header_nritems(eb) ||
ethanwucfc0eed2020-02-07 17:38:17 +0800453 is_shared_data_backref(preftrees, eb->start) ||
454 ref->root_id != btrfs_header_owner(eb)) {
Edmund Nadolskide47c9d2017-03-16 10:04:34 -0600455 if (time_seq == SEQ_LAST)
Qu Wenruo21633fc2015-04-16 14:54:50 +0800456 ret = btrfs_next_leaf(root, path);
457 else
458 ret = btrfs_next_old_leaf(root, path, time_seq);
459 }
Jan Schmidt8da6d582011-11-23 18:55:04 +0100460
ethanwub25b0b82020-02-07 17:38:18 +0800461 while (!ret && count < ref->count) {
Jan Schmidt8da6d582011-11-23 18:55:04 +0100462 eb = path->nodes[0];
Alexander Block69bca402012-06-19 07:42:26 -0600463 slot = path->slots[0];
464
465 btrfs_item_key_to_cpu(eb, &key, slot);
466
467 if (key.objectid != key_for_search->objectid ||
468 key.type != BTRFS_EXTENT_DATA_KEY)
469 break;
470
ethanwued58f2e2020-02-07 17:38:16 +0800471 /*
472 * We are searching for normal backref but bytenr of this leaf
ethanwucfc0eed2020-02-07 17:38:17 +0800473 * matches shared data backref, OR
474 * the leaf owner is not equal to the root we are searching for
ethanwued58f2e2020-02-07 17:38:16 +0800475 */
ethanwucfc0eed2020-02-07 17:38:17 +0800476 if (slot == 0 &&
477 (is_shared_data_backref(preftrees, eb->start) ||
478 ref->root_id != btrfs_header_owner(eb))) {
ethanwued58f2e2020-02-07 17:38:16 +0800479 if (time_seq == SEQ_LAST)
480 ret = btrfs_next_leaf(root, path);
481 else
482 ret = btrfs_next_old_leaf(root, path, time_seq);
483 continue;
484 }
Alexander Block69bca402012-06-19 07:42:26 -0600485 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
486 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
ethanwu7ac8b882020-02-07 17:38:15 +0800487 data_offset = btrfs_file_extent_offset(eb, fi);
Alexander Block69bca402012-06-19 07:42:26 -0600488
489 if (disk_byte == wanted_disk_byte) {
490 eie = NULL;
Josef Baciked8c4912013-07-05 14:03:47 -0400491 old = NULL;
ethanwu7ac8b882020-02-07 17:38:15 +0800492 if (ref->key_for_search.offset == key.offset - data_offset)
493 count++;
494 else
495 goto next;
Alexander Block69bca402012-06-19 07:42:26 -0600496 if (extent_item_pos) {
497 ret = check_extent_in_eb(&key, eb, fi,
498 *extent_item_pos,
Zygo Blaxellc995ab32017-09-22 13:58:45 -0400499 &eie, ignore_offset);
Alexander Block69bca402012-06-19 07:42:26 -0600500 if (ret < 0)
501 break;
502 }
Josef Baciked8c4912013-07-05 14:03:47 -0400503 if (ret > 0)
504 goto next;
Takashi Iwai4eb1f662014-07-28 10:57:04 +0200505 ret = ulist_add_merge_ptr(parents, eb->start,
506 eie, (void **)&old, GFP_NOFS);
Josef Baciked8c4912013-07-05 14:03:47 -0400507 if (ret < 0)
508 break;
509 if (!ret && extent_item_pos) {
510 while (old->next)
511 old = old->next;
512 old->next = eie;
Alexander Block69bca402012-06-19 07:42:26 -0600513 }
Wang Shilongf05c4742014-01-28 19:13:38 +0800514 eie = NULL;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100515 }
Josef Baciked8c4912013-07-05 14:03:47 -0400516next:
Edmund Nadolskide47c9d2017-03-16 10:04:34 -0600517 if (time_seq == SEQ_LAST)
Qu Wenruo21633fc2015-04-16 14:54:50 +0800518 ret = btrfs_next_item(root, path);
519 else
520 ret = btrfs_next_old_item(root, path, time_seq);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100521 }
522
Alexander Block69bca402012-06-19 07:42:26 -0600523 if (ret > 0)
524 ret = 0;
Wang Shilongf05c4742014-01-28 19:13:38 +0800525 else if (ret < 0)
526 free_inode_elem_list(eie);
Alexander Block69bca402012-06-19 07:42:26 -0600527 return ret;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100528}
529
530/*
531 * resolve an indirect backref in the form (root_id, key, level)
532 * to a logical address
533 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600534static int resolve_indirect_ref(struct btrfs_fs_info *fs_info,
535 struct btrfs_path *path, u64 time_seq,
ethanwued58f2e2020-02-07 17:38:16 +0800536 struct preftrees *preftrees,
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600537 struct prelim_ref *ref, struct ulist *parents,
ethanwub25b0b82020-02-07 17:38:18 +0800538 const u64 *extent_item_pos, bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100539{
Jan Schmidt8da6d582011-11-23 18:55:04 +0100540 struct btrfs_root *root;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100541 struct extent_buffer *eb;
542 int ret = 0;
543 int root_level;
544 int level = ref->level;
ethanwu7ac8b882020-02-07 17:38:15 +0800545 struct btrfs_key search_key = ref->key_for_search;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100546
Josef Bacik49d11be2020-10-19 16:02:31 -0400547 /*
548 * If we're search_commit_root we could possibly be holding locks on
549 * other tree nodes. This happens when qgroups does backref walks when
550 * adding new delayed refs. To deal with this we need to look in cache
551 * for the root, and if we don't find it then we need to search the
552 * tree_root's commit root, thus the btrfs_get_fs_root_commit_root usage
553 * here.
554 */
555 if (path->search_commit_root)
556 root = btrfs_get_fs_root_commit_root(fs_info, path, ref->root_id);
557 else
558 root = btrfs_get_fs_root(fs_info, ref->root_id, false);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100559 if (IS_ERR(root)) {
560 ret = PTR_ERR(root);
Josef Bacik9326f762020-01-24 09:32:28 -0500561 goto out_free;
562 }
563
Josef Bacik39dba872020-03-13 17:17:09 -0400564 if (!path->search_commit_root &&
565 test_bit(BTRFS_ROOT_DELETING, &root->state)) {
566 ret = -ENOENT;
567 goto out;
568 }
569
Jeff Mahoneyf5ee5c92016-06-21 09:52:41 -0400570 if (btrfs_is_testing(fs_info)) {
Josef Bacikd9ee5222015-10-05 10:35:29 -0400571 ret = -ENOENT;
572 goto out;
573 }
574
Josef Bacik9e351cc2014-03-13 15:42:13 -0400575 if (path->search_commit_root)
576 root_level = btrfs_header_level(root->commit_root);
Edmund Nadolskide47c9d2017-03-16 10:04:34 -0600577 else if (time_seq == SEQ_LAST)
Qu Wenruo21633fc2015-04-16 14:54:50 +0800578 root_level = btrfs_header_level(root->node);
Josef Bacik9e351cc2014-03-13 15:42:13 -0400579 else
580 root_level = btrfs_old_root_level(root, time_seq);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100581
Josef Bacikc75e8392020-02-14 16:11:47 -0500582 if (root_level + 1 == level)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100583 goto out;
584
ethanwu7ac8b882020-02-07 17:38:15 +0800585 /*
586 * We can often find data backrefs with an offset that is too large
587 * (>= LLONG_MAX, maximum allowed file offset) due to underflows when
588 * subtracting a file's offset with the data offset of its
589 * corresponding extent data item. This can happen for example in the
590 * clone ioctl.
591 *
592 * So if we detect such case we set the search key's offset to zero to
593 * make sure we will find the matching file extent item at
594 * add_all_parents(), otherwise we will miss it because the offset
595 * taken form the backref is much larger then the offset of the file
596 * extent item. This can make us scan a very large number of file
597 * extent items, but at least it will not make us miss any.
598 *
599 * This is an ugly workaround for a behaviour that should have never
600 * existed, but it does and a fix for the clone ioctl would touch a lot
601 * of places, cause backwards incompatibility and would not fix the
602 * problem for extents cloned with older kernels.
603 */
604 if (search_key.type == BTRFS_EXTENT_DATA_KEY &&
605 search_key.offset >= LLONG_MAX)
606 search_key.offset = 0;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100607 path->lowest_level = level;
Edmund Nadolskide47c9d2017-03-16 10:04:34 -0600608 if (time_seq == SEQ_LAST)
ethanwu7ac8b882020-02-07 17:38:15 +0800609 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
Qu Wenruo21633fc2015-04-16 14:54:50 +0800610 else
ethanwu7ac8b882020-02-07 17:38:15 +0800611 ret = btrfs_search_old_slot(root, &search_key, path, time_seq);
Wang Shilong538f72cd2014-01-23 13:47:48 +0800612
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -0400613 btrfs_debug(fs_info,
614 "search slot in root %llu (level %d, ref count %d) returned %d for key (%llu %u %llu)",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200615 ref->root_id, level, ref->count, ret,
616 ref->key_for_search.objectid, ref->key_for_search.type,
617 ref->key_for_search.offset);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100618 if (ret < 0)
619 goto out;
620
621 eb = path->nodes[level];
Jan Schmidt93454572012-06-27 15:23:09 +0200622 while (!eb) {
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530623 if (WARN_ON(!level)) {
Jan Schmidt93454572012-06-27 15:23:09 +0200624 ret = 1;
625 goto out;
626 }
627 level--;
628 eb = path->nodes[level];
Jan Schmidt8da6d582011-11-23 18:55:04 +0100629 }
630
ethanwued58f2e2020-02-07 17:38:16 +0800631 ret = add_all_parents(root, path, parents, preftrees, ref, level,
ethanwub25b0b82020-02-07 17:38:18 +0800632 time_seq, extent_item_pos, ignore_offset);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100633out:
Josef Bacik00246522020-01-24 09:33:01 -0500634 btrfs_put_root(root);
Josef Bacik9326f762020-01-24 09:32:28 -0500635out_free:
Josef Bacikda61d312013-06-12 16:20:08 -0400636 path->lowest_level = 0;
637 btrfs_release_path(path);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100638 return ret;
639}
640
Jeff Mahoney4dae0772017-06-28 21:56:56 -0600641static struct extent_inode_elem *
642unode_aux_to_inode_list(struct ulist_node *node)
643{
644 if (!node)
645 return NULL;
646 return (struct extent_inode_elem *)(uintptr_t)node->aux;
647}
648
Jan Schmidt8da6d582011-11-23 18:55:04 +0100649/*
Andrea Gelmini52042d82018-11-28 12:05:13 +0100650 * We maintain three separate rbtrees: one for direct refs, one for
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600651 * indirect refs which have a key, and one for indirect refs which do not
652 * have a key. Each tree does merge on insertion.
653 *
654 * Once all of the references are located, we iterate over the tree of
655 * indirect refs with missing keys. An appropriate key is located and
656 * the ref is moved onto the tree for indirect refs. After all missing
657 * keys are thus located, we iterate over the indirect ref tree, resolve
658 * each reference, and then insert the resolved reference onto the
659 * direct tree (merging there too).
660 *
661 * New backrefs (i.e., for parent nodes) are added to the appropriate
662 * rbtree as they are encountered. The new backrefs are subsequently
663 * resolved as above.
Jan Schmidt8da6d582011-11-23 18:55:04 +0100664 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600665static int resolve_indirect_refs(struct btrfs_fs_info *fs_info,
666 struct btrfs_path *path, u64 time_seq,
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600667 struct preftrees *preftrees,
ethanwub25b0b82020-02-07 17:38:18 +0800668 const u64 *extent_item_pos,
Zygo Blaxellc995ab32017-09-22 13:58:45 -0400669 struct share_check *sc, bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100670{
671 int err;
672 int ret = 0;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100673 struct ulist *parents;
674 struct ulist_node *node;
Jan Schmidtcd1b4132012-05-22 14:56:50 +0200675 struct ulist_iterator uiter;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600676 struct rb_node *rnode;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100677
678 parents = ulist_alloc(GFP_NOFS);
679 if (!parents)
680 return -ENOMEM;
681
682 /*
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600683 * We could trade memory usage for performance here by iterating
684 * the tree, allocating new refs for each insertion, and then
685 * freeing the entire indirect tree when we're done. In some test
686 * cases, the tree can grow quite large (~200k objects).
Jan Schmidt8da6d582011-11-23 18:55:04 +0100687 */
Liu Boecf160b2018-08-23 03:51:53 +0800688 while ((rnode = rb_first_cached(&preftrees->indirect.root))) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600689 struct prelim_ref *ref;
690
691 ref = rb_entry(rnode, struct prelim_ref, rbnode);
692 if (WARN(ref->parent,
693 "BUG: direct ref found in indirect tree")) {
694 ret = -EINVAL;
695 goto out;
696 }
697
Liu Boecf160b2018-08-23 03:51:53 +0800698 rb_erase_cached(&ref->rbnode, &preftrees->indirect.root);
Jeff Mahoney6c336b22017-07-12 16:20:07 -0600699 preftrees->indirect.count--;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600700
701 if (ref->count == 0) {
702 free_pref(ref);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100703 continue;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600704 }
705
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600706 if (sc && sc->root_objectid &&
707 ref->root_id != sc->root_objectid) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600708 free_pref(ref);
Josef Bacikdc046b12014-09-10 16:20:45 -0400709 ret = BACKREF_FOUND_SHARED;
710 goto out;
711 }
ethanwued58f2e2020-02-07 17:38:16 +0800712 err = resolve_indirect_ref(fs_info, path, time_seq, preftrees,
713 ref, parents, extent_item_pos,
ethanwub25b0b82020-02-07 17:38:18 +0800714 ignore_offset);
Wang Shilong95def2ed2014-01-23 13:47:49 +0800715 /*
716 * we can only tolerate ENOENT,otherwise,we should catch error
717 * and return directly.
718 */
719 if (err == -ENOENT) {
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600720 prelim_ref_insert(fs_info, &preftrees->direct, ref,
721 NULL);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100722 continue;
Wang Shilong95def2ed2014-01-23 13:47:49 +0800723 } else if (err) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600724 free_pref(ref);
Wang Shilong95def2ed2014-01-23 13:47:49 +0800725 ret = err;
726 goto out;
727 }
Jan Schmidt8da6d582011-11-23 18:55:04 +0100728
729 /* we put the first parent into the ref at hand */
Jan Schmidtcd1b4132012-05-22 14:56:50 +0200730 ULIST_ITER_INIT(&uiter);
731 node = ulist_next(parents, &uiter);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100732 ref->parent = node ? node->val : 0;
Jeff Mahoney4dae0772017-06-28 21:56:56 -0600733 ref->inode_list = unode_aux_to_inode_list(node);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100734
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600735 /* Add a prelim_ref(s) for any other parent(s). */
Jan Schmidtcd1b4132012-05-22 14:56:50 +0200736 while ((node = ulist_next(parents, &uiter))) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600737 struct prelim_ref *new_ref;
738
Wang Shilongb9e9a6c2013-08-09 13:25:36 +0800739 new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
740 GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100741 if (!new_ref) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600742 free_pref(ref);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100743 ret = -ENOMEM;
Wang Shilonge36902d2013-04-15 10:26:38 +0000744 goto out;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100745 }
746 memcpy(new_ref, ref, sizeof(*ref));
747 new_ref->parent = node->val;
Jeff Mahoney4dae0772017-06-28 21:56:56 -0600748 new_ref->inode_list = unode_aux_to_inode_list(node);
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600749 prelim_ref_insert(fs_info, &preftrees->direct,
750 new_ref, NULL);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100751 }
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600752
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600753 /*
Andrea Gelmini52042d82018-11-28 12:05:13 +0100754 * Now it's a direct ref, put it in the direct tree. We must
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600755 * do this last because the ref could be merged/freed here.
756 */
757 prelim_ref_insert(fs_info, &preftrees->direct, ref, NULL);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600758
Jan Schmidt8da6d582011-11-23 18:55:04 +0100759 ulist_reinit(parents);
Edmund Nadolski9dd14fd2017-07-12 16:20:09 -0600760 cond_resched();
Jan Schmidt8da6d582011-11-23 18:55:04 +0100761 }
Wang Shilonge36902d2013-04-15 10:26:38 +0000762out:
Jan Schmidt8da6d582011-11-23 18:55:04 +0100763 ulist_free(parents);
764 return ret;
765}
766
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200767/*
768 * read tree blocks and add keys where required.
769 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600770static int add_missing_keys(struct btrfs_fs_info *fs_info,
Josef Bacik38e3eeb2019-01-16 11:00:57 -0500771 struct preftrees *preftrees, bool lock)
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200772{
Jeff Mahoneye0c476b2017-06-28 21:56:57 -0600773 struct prelim_ref *ref;
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200774 struct extent_buffer *eb;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600775 struct preftree *tree = &preftrees->indirect_missing_keys;
776 struct rb_node *node;
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200777
Liu Boecf160b2018-08-23 03:51:53 +0800778 while ((node = rb_first_cached(&tree->root))) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600779 ref = rb_entry(node, struct prelim_ref, rbnode);
Liu Boecf160b2018-08-23 03:51:53 +0800780 rb_erase_cached(node, &tree->root);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600781
782 BUG_ON(ref->parent); /* should not be a direct ref */
783 BUG_ON(ref->key_for_search.type);
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200784 BUG_ON(!ref->wanted_disk_byte);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600785
Josef Bacik1b7ec852020-11-05 10:45:18 -0500786 eb = read_tree_block(fs_info, ref->wanted_disk_byte,
787 ref->root_id, 0, ref->level - 1, NULL);
Liu Bo64c043d2015-05-25 17:30:15 +0800788 if (IS_ERR(eb)) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600789 free_pref(ref);
Liu Bo64c043d2015-05-25 17:30:15 +0800790 return PTR_ERR(eb);
791 } else if (!extent_buffer_uptodate(eb)) {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600792 free_pref(ref);
Josef Bacik416bc652013-04-23 14:17:42 -0400793 free_extent_buffer(eb);
794 return -EIO;
795 }
Josef Bacik38e3eeb2019-01-16 11:00:57 -0500796 if (lock)
797 btrfs_tree_read_lock(eb);
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200798 if (btrfs_header_level(eb) == 0)
799 btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
800 else
801 btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
Josef Bacik38e3eeb2019-01-16 11:00:57 -0500802 if (lock)
803 btrfs_tree_read_unlock(eb);
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200804 free_extent_buffer(eb);
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600805 prelim_ref_insert(fs_info, &preftrees->indirect, ref, NULL);
Edmund Nadolski9dd14fd2017-07-12 16:20:09 -0600806 cond_resched();
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200807 }
808 return 0;
809}
810
Jan Schmidt8da6d582011-11-23 18:55:04 +0100811/*
Jan Schmidt8da6d582011-11-23 18:55:04 +0100812 * add all currently queued delayed refs from this head whose seq nr is
813 * smaller or equal that seq to the list
814 */
Jeff Mahoney00142752017-07-12 16:20:08 -0600815static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
816 struct btrfs_delayed_ref_head *head, u64 seq,
ethanwub25b0b82020-02-07 17:38:18 +0800817 struct preftrees *preftrees, struct share_check *sc)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100818{
Qu Wenruoc6fc2452015-03-30 17:03:00 +0800819 struct btrfs_delayed_ref_node *node;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100820 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
Jan Schmidtd5c88b72012-05-15 17:55:51 +0200821 struct btrfs_key key;
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600822 struct btrfs_key tmp_op_key;
Josef Bacik0e0adbc2017-10-19 14:16:00 -0400823 struct rb_node *n;
Edmund Nadolski01747e92017-07-12 16:20:11 -0600824 int count;
Jan Schmidtb1375d62012-01-26 15:01:11 -0500825 int ret = 0;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100826
Nikolay Borisova6dbcea2018-03-15 14:36:22 +0200827 if (extent_op && extent_op->update_key)
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600828 btrfs_disk_key_to_cpu(&tmp_op_key, &extent_op->key);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100829
Josef Bacikd7df2c72014-01-23 09:21:38 -0500830 spin_lock(&head->lock);
Liu Boe3d03962018-08-23 03:51:50 +0800831 for (n = rb_first_cached(&head->ref_tree); n; n = rb_next(n)) {
Josef Bacik0e0adbc2017-10-19 14:16:00 -0400832 node = rb_entry(n, struct btrfs_delayed_ref_node,
833 ref_node);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100834 if (node->seq > seq)
835 continue;
836
837 switch (node->action) {
838 case BTRFS_ADD_DELAYED_EXTENT:
839 case BTRFS_UPDATE_DELAYED_HEAD:
840 WARN_ON(1);
841 continue;
842 case BTRFS_ADD_DELAYED_REF:
Edmund Nadolski01747e92017-07-12 16:20:11 -0600843 count = node->ref_mod;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100844 break;
845 case BTRFS_DROP_DELAYED_REF:
Edmund Nadolski01747e92017-07-12 16:20:11 -0600846 count = node->ref_mod * -1;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100847 break;
848 default:
Arnd Bergmann290342f2019-03-25 14:02:25 +0100849 BUG();
Jan Schmidt8da6d582011-11-23 18:55:04 +0100850 }
851 switch (node->type) {
852 case BTRFS_TREE_BLOCK_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600853 /* NORMAL INDIRECT METADATA backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +0100854 struct btrfs_delayed_tree_ref *ref;
855
856 ref = btrfs_delayed_node_to_tree_ref(node);
Jeff Mahoney00142752017-07-12 16:20:08 -0600857 ret = add_indirect_ref(fs_info, preftrees, ref->root,
858 &tmp_op_key, ref->level + 1,
Edmund Nadolski01747e92017-07-12 16:20:11 -0600859 node->bytenr, count, sc,
860 GFP_ATOMIC);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100861 break;
862 }
863 case BTRFS_SHARED_BLOCK_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600864 /* SHARED DIRECT METADATA backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +0100865 struct btrfs_delayed_tree_ref *ref;
866
867 ref = btrfs_delayed_node_to_tree_ref(node);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600868
Edmund Nadolski01747e92017-07-12 16:20:11 -0600869 ret = add_direct_ref(fs_info, preftrees, ref->level + 1,
870 ref->parent, node->bytenr, count,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600871 sc, GFP_ATOMIC);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100872 break;
873 }
874 case BTRFS_EXTENT_DATA_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600875 /* NORMAL INDIRECT DATA backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +0100876 struct btrfs_delayed_data_ref *ref;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100877 ref = btrfs_delayed_node_to_data_ref(node);
878
879 key.objectid = ref->objectid;
880 key.type = BTRFS_EXTENT_DATA_KEY;
881 key.offset = ref->offset;
Josef Bacikdc046b12014-09-10 16:20:45 -0400882
883 /*
884 * Found a inum that doesn't match our known inum, we
885 * know it's shared.
886 */
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600887 if (sc && sc->inum && ref->objectid != sc->inum) {
Josef Bacikdc046b12014-09-10 16:20:45 -0400888 ret = BACKREF_FOUND_SHARED;
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600889 goto out;
Josef Bacikdc046b12014-09-10 16:20:45 -0400890 }
891
Jeff Mahoney00142752017-07-12 16:20:08 -0600892 ret = add_indirect_ref(fs_info, preftrees, ref->root,
Edmund Nadolski01747e92017-07-12 16:20:11 -0600893 &key, 0, node->bytenr, count, sc,
894 GFP_ATOMIC);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100895 break;
896 }
897 case BTRFS_SHARED_DATA_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600898 /* SHARED DIRECT FULL backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +0100899 struct btrfs_delayed_data_ref *ref;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100900
901 ref = btrfs_delayed_node_to_data_ref(node);
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600902
Edmund Nadolski01747e92017-07-12 16:20:11 -0600903 ret = add_direct_ref(fs_info, preftrees, 0, ref->parent,
904 node->bytenr, count, sc,
905 GFP_ATOMIC);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100906 break;
907 }
908 default:
909 WARN_ON(1);
910 }
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600911 /*
912 * We must ignore BACKREF_FOUND_SHARED until all delayed
913 * refs have been checked.
914 */
915 if (ret && (ret != BACKREF_FOUND_SHARED))
Josef Bacikd7df2c72014-01-23 09:21:38 -0500916 break;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100917 }
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600918 if (!ret)
919 ret = extent_is_shared(sc);
920out:
Josef Bacikd7df2c72014-01-23 09:21:38 -0500921 spin_unlock(&head->lock);
922 return ret;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100923}
924
925/*
926 * add all inline backrefs for bytenr to the list
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600927 *
928 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
Jan Schmidt8da6d582011-11-23 18:55:04 +0100929 */
Jeff Mahoney00142752017-07-12 16:20:08 -0600930static int add_inline_refs(const struct btrfs_fs_info *fs_info,
931 struct btrfs_path *path, u64 bytenr,
Edmund Nadolski86d5f992017-07-12 16:20:06 -0600932 int *info_level, struct preftrees *preftrees,
ethanwub25b0b82020-02-07 17:38:18 +0800933 struct share_check *sc)
Jan Schmidt8da6d582011-11-23 18:55:04 +0100934{
Jan Schmidtb1375d62012-01-26 15:01:11 -0500935 int ret = 0;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100936 int slot;
937 struct extent_buffer *leaf;
938 struct btrfs_key key;
Josef Bacik261c84b2013-06-28 13:11:22 -0400939 struct btrfs_key found_key;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100940 unsigned long ptr;
941 unsigned long end;
942 struct btrfs_extent_item *ei;
943 u64 flags;
944 u64 item_size;
945
946 /*
947 * enumerate all inline refs
948 */
949 leaf = path->nodes[0];
Jan Schmidtdadcaf72012-05-22 13:43:25 +0200950 slot = path->slots[0];
Jan Schmidt8da6d582011-11-23 18:55:04 +0100951
952 item_size = btrfs_item_size_nr(leaf, slot);
953 BUG_ON(item_size < sizeof(*ei));
954
955 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
956 flags = btrfs_extent_flags(leaf, ei);
Josef Bacik261c84b2013-06-28 13:11:22 -0400957 btrfs_item_key_to_cpu(leaf, &found_key, slot);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100958
959 ptr = (unsigned long)(ei + 1);
960 end = (unsigned long)ei + item_size;
961
Josef Bacik261c84b2013-06-28 13:11:22 -0400962 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
963 flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
Jan Schmidt8da6d582011-11-23 18:55:04 +0100964 struct btrfs_tree_block_info *info;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100965
966 info = (struct btrfs_tree_block_info *)ptr;
967 *info_level = btrfs_tree_block_level(leaf, info);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100968 ptr += sizeof(struct btrfs_tree_block_info);
969 BUG_ON(ptr > end);
Josef Bacik261c84b2013-06-28 13:11:22 -0400970 } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
971 *info_level = found_key.offset;
Jan Schmidt8da6d582011-11-23 18:55:04 +0100972 } else {
973 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
974 }
975
976 while (ptr < end) {
977 struct btrfs_extent_inline_ref *iref;
978 u64 offset;
979 int type;
980
981 iref = (struct btrfs_extent_inline_ref *)ptr;
Liu Bo3de28d52017-08-18 15:15:19 -0600982 type = btrfs_get_extent_inline_ref_type(leaf, iref,
983 BTRFS_REF_TYPE_ANY);
984 if (type == BTRFS_REF_TYPE_INVALID)
Su Yueaf431dc2018-06-22 16:18:01 +0800985 return -EUCLEAN;
Liu Bo3de28d52017-08-18 15:15:19 -0600986
Jan Schmidt8da6d582011-11-23 18:55:04 +0100987 offset = btrfs_extent_inline_ref_offset(leaf, iref);
988
989 switch (type) {
990 case BTRFS_SHARED_BLOCK_REF_KEY:
Jeff Mahoney00142752017-07-12 16:20:08 -0600991 ret = add_direct_ref(fs_info, preftrees,
992 *info_level + 1, offset,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -0600993 bytenr, 1, NULL, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +0100994 break;
995 case BTRFS_SHARED_DATA_REF_KEY: {
996 struct btrfs_shared_data_ref *sdref;
997 int count;
998
999 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
1000 count = btrfs_shared_data_ref_count(leaf, sdref);
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001001
Jeff Mahoney00142752017-07-12 16:20:08 -06001002 ret = add_direct_ref(fs_info, preftrees, 0, offset,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001003 bytenr, count, sc, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001004 break;
1005 }
1006 case BTRFS_TREE_BLOCK_REF_KEY:
Jeff Mahoney00142752017-07-12 16:20:08 -06001007 ret = add_indirect_ref(fs_info, preftrees, offset,
1008 NULL, *info_level + 1,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001009 bytenr, 1, NULL, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001010 break;
1011 case BTRFS_EXTENT_DATA_REF_KEY: {
1012 struct btrfs_extent_data_ref *dref;
1013 int count;
1014 u64 root;
1015
1016 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1017 count = btrfs_extent_data_ref_count(leaf, dref);
1018 key.objectid = btrfs_extent_data_ref_objectid(leaf,
1019 dref);
1020 key.type = BTRFS_EXTENT_DATA_KEY;
1021 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
Josef Bacikdc046b12014-09-10 16:20:45 -04001022
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001023 if (sc && sc->inum && key.objectid != sc->inum) {
Josef Bacikdc046b12014-09-10 16:20:45 -04001024 ret = BACKREF_FOUND_SHARED;
1025 break;
1026 }
1027
Jan Schmidt8da6d582011-11-23 18:55:04 +01001028 root = btrfs_extent_data_ref_root(leaf, dref);
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001029
Jeff Mahoney00142752017-07-12 16:20:08 -06001030 ret = add_indirect_ref(fs_info, preftrees, root,
1031 &key, 0, bytenr, count,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001032 sc, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001033 break;
1034 }
1035 default:
1036 WARN_ON(1);
1037 }
Wang Shilong1149ab62013-04-10 11:22:50 +00001038 if (ret)
1039 return ret;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001040 ptr += btrfs_extent_inline_ref_size(type);
1041 }
1042
1043 return 0;
1044}
1045
1046/*
1047 * add all non-inline backrefs for bytenr to the list
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001048 *
1049 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
Jan Schmidt8da6d582011-11-23 18:55:04 +01001050 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001051static int add_keyed_refs(struct btrfs_fs_info *fs_info,
1052 struct btrfs_path *path, u64 bytenr,
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001053 int info_level, struct preftrees *preftrees,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001054 struct share_check *sc)
Jan Schmidt8da6d582011-11-23 18:55:04 +01001055{
1056 struct btrfs_root *extent_root = fs_info->extent_root;
1057 int ret;
1058 int slot;
1059 struct extent_buffer *leaf;
1060 struct btrfs_key key;
1061
1062 while (1) {
1063 ret = btrfs_next_item(extent_root, path);
1064 if (ret < 0)
1065 break;
1066 if (ret) {
1067 ret = 0;
1068 break;
1069 }
1070
1071 slot = path->slots[0];
1072 leaf = path->nodes[0];
1073 btrfs_item_key_to_cpu(leaf, &key, slot);
1074
1075 if (key.objectid != bytenr)
1076 break;
1077 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
1078 continue;
1079 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
1080 break;
1081
1082 switch (key.type) {
1083 case BTRFS_SHARED_BLOCK_REF_KEY:
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001084 /* SHARED DIRECT METADATA backref */
Jeff Mahoney00142752017-07-12 16:20:08 -06001085 ret = add_direct_ref(fs_info, preftrees,
1086 info_level + 1, key.offset,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001087 bytenr, 1, NULL, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001088 break;
1089 case BTRFS_SHARED_DATA_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001090 /* SHARED DIRECT FULL backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +01001091 struct btrfs_shared_data_ref *sdref;
1092 int count;
1093
1094 sdref = btrfs_item_ptr(leaf, slot,
1095 struct btrfs_shared_data_ref);
1096 count = btrfs_shared_data_ref_count(leaf, sdref);
Jeff Mahoney00142752017-07-12 16:20:08 -06001097 ret = add_direct_ref(fs_info, preftrees, 0,
1098 key.offset, bytenr, count,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001099 sc, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001100 break;
1101 }
1102 case BTRFS_TREE_BLOCK_REF_KEY:
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001103 /* NORMAL INDIRECT METADATA backref */
Jeff Mahoney00142752017-07-12 16:20:08 -06001104 ret = add_indirect_ref(fs_info, preftrees, key.offset,
1105 NULL, info_level + 1, bytenr,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001106 1, NULL, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001107 break;
1108 case BTRFS_EXTENT_DATA_REF_KEY: {
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001109 /* NORMAL INDIRECT DATA backref */
Jan Schmidt8da6d582011-11-23 18:55:04 +01001110 struct btrfs_extent_data_ref *dref;
1111 int count;
1112 u64 root;
1113
1114 dref = btrfs_item_ptr(leaf, slot,
1115 struct btrfs_extent_data_ref);
1116 count = btrfs_extent_data_ref_count(leaf, dref);
1117 key.objectid = btrfs_extent_data_ref_objectid(leaf,
1118 dref);
1119 key.type = BTRFS_EXTENT_DATA_KEY;
1120 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
Josef Bacikdc046b12014-09-10 16:20:45 -04001121
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001122 if (sc && sc->inum && key.objectid != sc->inum) {
Josef Bacikdc046b12014-09-10 16:20:45 -04001123 ret = BACKREF_FOUND_SHARED;
1124 break;
1125 }
1126
Jan Schmidt8da6d582011-11-23 18:55:04 +01001127 root = btrfs_extent_data_ref_root(leaf, dref);
Jeff Mahoney00142752017-07-12 16:20:08 -06001128 ret = add_indirect_ref(fs_info, preftrees, root,
1129 &key, 0, bytenr, count,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001130 sc, GFP_NOFS);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001131 break;
1132 }
1133 default:
1134 WARN_ON(1);
1135 }
Wang Shilong1149ab62013-04-10 11:22:50 +00001136 if (ret)
1137 return ret;
1138
Jan Schmidt8da6d582011-11-23 18:55:04 +01001139 }
1140
1141 return ret;
1142}
1143
1144/*
1145 * this adds all existing backrefs (inline backrefs, backrefs and delayed
1146 * refs) for the given bytenr to the refs list, merges duplicates and resolves
1147 * indirect refs to their parent bytenr.
1148 * When roots are found, they're added to the roots list
1149 *
Edmund Nadolskide47c9d2017-03-16 10:04:34 -06001150 * If time_seq is set to SEQ_LAST, it will not search delayed_refs, and behave
Qu Wenruo21633fc2015-04-16 14:54:50 +08001151 * much like trans == NULL case, the difference only lies in it will not
1152 * commit root.
1153 * The special case is for qgroup to search roots in commit_transaction().
1154 *
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001155 * @sc - if !NULL, then immediately return BACKREF_FOUND_SHARED when a
1156 * shared extent is detected.
1157 *
1158 * Otherwise this returns 0 for success and <0 for an error.
1159 *
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001160 * If ignore_offset is set to false, only extent refs whose offsets match
1161 * extent_item_pos are returned. If true, every extent ref is returned
1162 * and extent_item_pos is ignored.
1163 *
Jan Schmidt8da6d582011-11-23 18:55:04 +01001164 * FIXME some caching might speed things up
1165 */
1166static int find_parent_nodes(struct btrfs_trans_handle *trans,
1167 struct btrfs_fs_info *fs_info, u64 bytenr,
Jan Schmidt097b8a72012-06-21 11:08:04 +02001168 u64 time_seq, struct ulist *refs,
Josef Bacikdc046b12014-09-10 16:20:45 -04001169 struct ulist *roots, const u64 *extent_item_pos,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001170 struct share_check *sc, bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +01001171{
1172 struct btrfs_key key;
1173 struct btrfs_path *path;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001174 struct btrfs_delayed_ref_root *delayed_refs = NULL;
Li Zefand3b01062012-03-03 07:41:15 -05001175 struct btrfs_delayed_ref_head *head;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001176 int info_level = 0;
1177 int ret;
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001178 struct prelim_ref *ref;
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001179 struct rb_node *node;
Wang Shilongf05c4742014-01-28 19:13:38 +08001180 struct extent_inode_elem *eie = NULL;
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001181 struct preftrees preftrees = {
1182 .direct = PREFTREE_INIT,
1183 .indirect = PREFTREE_INIT,
1184 .indirect_missing_keys = PREFTREE_INIT
1185 };
Jan Schmidt8da6d582011-11-23 18:55:04 +01001186
1187 key.objectid = bytenr;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001188 key.offset = (u64)-1;
Josef Bacik261c84b2013-06-28 13:11:22 -04001189 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1190 key.type = BTRFS_METADATA_ITEM_KEY;
1191 else
1192 key.type = BTRFS_EXTENT_ITEM_KEY;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001193
1194 path = btrfs_alloc_path();
1195 if (!path)
1196 return -ENOMEM;
Wang Shilonge84752d2014-02-13 11:19:47 +08001197 if (!trans) {
Josef Bacikda61d312013-06-12 16:20:08 -04001198 path->search_commit_root = 1;
Wang Shilonge84752d2014-02-13 11:19:47 +08001199 path->skip_locking = 1;
1200 }
Jan Schmidt8da6d582011-11-23 18:55:04 +01001201
Edmund Nadolskide47c9d2017-03-16 10:04:34 -06001202 if (time_seq == SEQ_LAST)
Qu Wenruo21633fc2015-04-16 14:54:50 +08001203 path->skip_locking = 1;
1204
Jan Schmidt8da6d582011-11-23 18:55:04 +01001205 /*
1206 * grab both a lock on the path and a lock on the delayed ref head.
1207 * We need both to get a consistent picture of how the refs look
1208 * at a specified point in time
1209 */
1210again:
Li Zefand3b01062012-03-03 07:41:15 -05001211 head = NULL;
1212
Jan Schmidt8da6d582011-11-23 18:55:04 +01001213 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
1214 if (ret < 0)
1215 goto out;
1216 BUG_ON(ret == 0);
1217
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001218#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
Qu Wenruo21633fc2015-04-16 14:54:50 +08001219 if (trans && likely(trans->type != __TRANS_DUMMY) &&
Edmund Nadolskide47c9d2017-03-16 10:04:34 -06001220 time_seq != SEQ_LAST) {
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001221#else
Edmund Nadolskide47c9d2017-03-16 10:04:34 -06001222 if (trans && time_seq != SEQ_LAST) {
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001223#endif
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001224 /*
1225 * look if there are updates for this ref queued and lock the
1226 * head
1227 */
1228 delayed_refs = &trans->transaction->delayed_refs;
1229 spin_lock(&delayed_refs->lock);
Liu Bof72ad18e2017-01-30 12:24:37 -08001230 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001231 if (head) {
1232 if (!mutex_trylock(&head->mutex)) {
Josef Bacikd2788502017-09-29 15:43:57 -04001233 refcount_inc(&head->refs);
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001234 spin_unlock(&delayed_refs->lock);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001235
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001236 btrfs_release_path(path);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001237
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001238 /*
1239 * Mutex was contended, block until it's
1240 * released and try again
1241 */
1242 mutex_lock(&head->mutex);
1243 mutex_unlock(&head->mutex);
Josef Bacikd2788502017-09-29 15:43:57 -04001244 btrfs_put_delayed_ref_head(head);
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001245 goto again;
1246 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05001247 spin_unlock(&delayed_refs->lock);
Jeff Mahoney00142752017-07-12 16:20:08 -06001248 ret = add_delayed_refs(fs_info, head, time_seq,
ethanwub25b0b82020-02-07 17:38:18 +08001249 &preftrees, sc);
Jan Schmidt155725c2012-06-22 14:01:00 +02001250 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05001251 if (ret)
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001252 goto out;
Josef Bacikd7df2c72014-01-23 09:21:38 -05001253 } else {
1254 spin_unlock(&delayed_refs->lock);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001255 }
Jan Schmidt8da6d582011-11-23 18:55:04 +01001256 }
Jan Schmidt8da6d582011-11-23 18:55:04 +01001257
1258 if (path->slots[0]) {
1259 struct extent_buffer *leaf;
1260 int slot;
1261
Jan Schmidtdadcaf72012-05-22 13:43:25 +02001262 path->slots[0]--;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001263 leaf = path->nodes[0];
Jan Schmidtdadcaf72012-05-22 13:43:25 +02001264 slot = path->slots[0];
Jan Schmidt8da6d582011-11-23 18:55:04 +01001265 btrfs_item_key_to_cpu(leaf, &key, slot);
1266 if (key.objectid == bytenr &&
Josef Bacik261c84b2013-06-28 13:11:22 -04001267 (key.type == BTRFS_EXTENT_ITEM_KEY ||
1268 key.type == BTRFS_METADATA_ITEM_KEY)) {
Jeff Mahoney00142752017-07-12 16:20:08 -06001269 ret = add_inline_refs(fs_info, path, bytenr,
ethanwub25b0b82020-02-07 17:38:18 +08001270 &info_level, &preftrees, sc);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001271 if (ret)
1272 goto out;
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001273 ret = add_keyed_refs(fs_info, path, bytenr, info_level,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001274 &preftrees, sc);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001275 if (ret)
1276 goto out;
1277 }
1278 }
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001279
Jan Schmidt8da6d582011-11-23 18:55:04 +01001280 btrfs_release_path(path);
1281
Josef Bacik38e3eeb2019-01-16 11:00:57 -05001282 ret = add_missing_keys(fs_info, &preftrees, path->skip_locking == 0);
Jan Schmidtd5c88b72012-05-15 17:55:51 +02001283 if (ret)
1284 goto out;
1285
Liu Boecf160b2018-08-23 03:51:53 +08001286 WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect_missing_keys.root.rb_root));
Jan Schmidt8da6d582011-11-23 18:55:04 +01001287
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001288 ret = resolve_indirect_refs(fs_info, path, time_seq, &preftrees,
ethanwub25b0b82020-02-07 17:38:18 +08001289 extent_item_pos, sc, ignore_offset);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001290 if (ret)
1291 goto out;
1292
Liu Boecf160b2018-08-23 03:51:53 +08001293 WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect.root.rb_root));
Jan Schmidt8da6d582011-11-23 18:55:04 +01001294
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001295 /*
1296 * This walks the tree of merged and resolved refs. Tree blocks are
1297 * read in as needed. Unique entries are added to the ulist, and
1298 * the list of found roots is updated.
1299 *
1300 * We release the entire tree in one go before returning.
1301 */
Liu Boecf160b2018-08-23 03:51:53 +08001302 node = rb_first_cached(&preftrees.direct.root);
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001303 while (node) {
1304 ref = rb_entry(node, struct prelim_ref, rbnode);
1305 node = rb_next(&ref->rbnode);
Zygo Blaxellc8195a72018-01-23 22:22:09 -05001306 /*
1307 * ref->count < 0 can happen here if there are delayed
1308 * refs with a node->action of BTRFS_DROP_DELAYED_REF.
1309 * prelim_ref_insert() relies on this when merging
1310 * identical refs to keep the overall count correct.
1311 * prelim_ref_insert() will merge only those refs
1312 * which compare identically. Any refs having
1313 * e.g. different offsets would not be merged,
1314 * and would retain their original ref->count < 0.
1315 */
Wang Shilong98cfee212014-02-01 00:42:05 +08001316 if (roots && ref->count && ref->root_id && ref->parent == 0) {
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001317 if (sc && sc->root_objectid &&
1318 ref->root_id != sc->root_objectid) {
Josef Bacikdc046b12014-09-10 16:20:45 -04001319 ret = BACKREF_FOUND_SHARED;
1320 goto out;
1321 }
1322
Jan Schmidt8da6d582011-11-23 18:55:04 +01001323 /* no parent == root of tree */
1324 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
Wang Shilongf1723932013-03-29 23:03:21 +00001325 if (ret < 0)
1326 goto out;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001327 }
1328 if (ref->count && ref->parent) {
Josef Bacik8a564572014-06-05 16:08:45 -04001329 if (extent_item_pos && !ref->inode_list &&
1330 ref->level == 0) {
Jan Schmidt976b1902012-05-17 16:43:03 +02001331 struct extent_buffer *eb;
David Sterba707e8a02014-06-04 19:22:26 +02001332
Qu Wenruo581c1762018-03-29 09:08:11 +08001333 eb = read_tree_block(fs_info, ref->parent, 0,
Josef Bacik1b7ec852020-11-05 10:45:18 -05001334 0, ref->level, NULL);
Liu Bo64c043d2015-05-25 17:30:15 +08001335 if (IS_ERR(eb)) {
1336 ret = PTR_ERR(eb);
1337 goto out;
1338 } else if (!extent_buffer_uptodate(eb)) {
Josef Bacik416bc652013-04-23 14:17:42 -04001339 free_extent_buffer(eb);
Wang Shilongc16c2e22013-05-08 08:10:25 +00001340 ret = -EIO;
1341 goto out;
Josef Bacik416bc652013-04-23 14:17:42 -04001342 }
Josef Bacik38e3eeb2019-01-16 11:00:57 -05001343
Josef Bacikac5887c2020-08-20 11:46:10 -04001344 if (!path->skip_locking)
Josef Bacik38e3eeb2019-01-16 11:00:57 -05001345 btrfs_tree_read_lock(eb);
Jan Schmidt976b1902012-05-17 16:43:03 +02001346 ret = find_extent_in_eb(eb, bytenr,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001347 *extent_item_pos, &eie, ignore_offset);
Josef Bacik38e3eeb2019-01-16 11:00:57 -05001348 if (!path->skip_locking)
Josef Bacikac5887c2020-08-20 11:46:10 -04001349 btrfs_tree_read_unlock(eb);
Jan Schmidt976b1902012-05-17 16:43:03 +02001350 free_extent_buffer(eb);
Filipe David Borba Mananaf5929cd2013-07-31 00:26:35 +01001351 if (ret < 0)
1352 goto out;
1353 ref->inode_list = eie;
Jan Schmidt976b1902012-05-17 16:43:03 +02001354 }
Takashi Iwai4eb1f662014-07-28 10:57:04 +02001355 ret = ulist_add_merge_ptr(refs, ref->parent,
1356 ref->inode_list,
1357 (void **)&eie, GFP_NOFS);
Wang Shilongf1723932013-03-29 23:03:21 +00001358 if (ret < 0)
1359 goto out;
Jan Schmidt33019582012-05-30 18:05:21 +02001360 if (!ret && extent_item_pos) {
1361 /*
1362 * we've recorded that parent, so we must extend
1363 * its inode list here
1364 */
1365 BUG_ON(!eie);
1366 while (eie->next)
1367 eie = eie->next;
1368 eie->next = ref->inode_list;
1369 }
Wang Shilongf05c4742014-01-28 19:13:38 +08001370 eie = NULL;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001371 }
Edmund Nadolski9dd14fd2017-07-12 16:20:09 -06001372 cond_resched();
Jan Schmidt8da6d582011-11-23 18:55:04 +01001373 }
1374
1375out:
Jan Schmidt8da6d582011-11-23 18:55:04 +01001376 btrfs_free_path(path);
Edmund Nadolski86d5f992017-07-12 16:20:06 -06001377
1378 prelim_release(&preftrees.direct);
1379 prelim_release(&preftrees.indirect);
1380 prelim_release(&preftrees.indirect_missing_keys);
1381
Wang Shilongf05c4742014-01-28 19:13:38 +08001382 if (ret < 0)
1383 free_inode_elem_list(eie);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001384 return ret;
1385}
1386
Jan Schmidt976b1902012-05-17 16:43:03 +02001387static void free_leaf_list(struct ulist *blocks)
1388{
1389 struct ulist_node *node = NULL;
1390 struct extent_inode_elem *eie;
Jan Schmidt976b1902012-05-17 16:43:03 +02001391 struct ulist_iterator uiter;
1392
1393 ULIST_ITER_INIT(&uiter);
1394 while ((node = ulist_next(blocks, &uiter))) {
1395 if (!node->aux)
1396 continue;
Jeff Mahoney4dae0772017-06-28 21:56:56 -06001397 eie = unode_aux_to_inode_list(node);
Wang Shilongf05c4742014-01-28 19:13:38 +08001398 free_inode_elem_list(eie);
Jan Schmidt976b1902012-05-17 16:43:03 +02001399 node->aux = 0;
1400 }
1401
1402 ulist_free(blocks);
1403}
1404
Jan Schmidt8da6d582011-11-23 18:55:04 +01001405/*
1406 * Finds all leafs with a reference to the specified combination of bytenr and
1407 * offset. key_list_head will point to a list of corresponding keys (caller must
1408 * free each list element). The leafs will be stored in the leafs ulist, which
1409 * must be freed with ulist_free.
1410 *
1411 * returns 0 on success, <0 on error
1412 */
Qu Wenruo19b546d2020-03-10 16:14:15 +08001413int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
1414 struct btrfs_fs_info *fs_info, u64 bytenr,
1415 u64 time_seq, struct ulist **leafs,
1416 const u64 *extent_item_pos, bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +01001417{
Jan Schmidt8da6d582011-11-23 18:55:04 +01001418 int ret;
1419
Jan Schmidt8da6d582011-11-23 18:55:04 +01001420 *leafs = ulist_alloc(GFP_NOFS);
Wang Shilong98cfee212014-02-01 00:42:05 +08001421 if (!*leafs)
Jan Schmidt8da6d582011-11-23 18:55:04 +01001422 return -ENOMEM;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001423
Lu Fengqiafce7722016-06-13 09:36:46 +08001424 ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001425 *leafs, NULL, extent_item_pos, NULL, ignore_offset);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001426 if (ret < 0 && ret != -ENOENT) {
Jan Schmidt976b1902012-05-17 16:43:03 +02001427 free_leaf_list(*leafs);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001428 return ret;
1429 }
1430
1431 return 0;
1432}
1433
1434/*
1435 * walk all backrefs for a given extent to find all roots that reference this
1436 * extent. Walking a backref means finding all extents that reference this
1437 * extent and in turn walk the backrefs of those, too. Naturally this is a
1438 * recursive process, but here it is implemented in an iterative fashion: We
1439 * find all referencing extents for the extent in question and put them on a
1440 * list. In turn, we find all referencing extents for those, further appending
1441 * to the list. The way we iterate the list allows adding more elements after
1442 * the current while iterating. The process stops when we reach the end of the
1443 * list. Found roots are added to the roots list.
1444 *
1445 * returns 0 on success, < 0 on error.
1446 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001447static int btrfs_find_all_roots_safe(struct btrfs_trans_handle *trans,
1448 struct btrfs_fs_info *fs_info, u64 bytenr,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001449 u64 time_seq, struct ulist **roots,
1450 bool ignore_offset)
Jan Schmidt8da6d582011-11-23 18:55:04 +01001451{
1452 struct ulist *tmp;
1453 struct ulist_node *node = NULL;
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001454 struct ulist_iterator uiter;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001455 int ret;
1456
1457 tmp = ulist_alloc(GFP_NOFS);
1458 if (!tmp)
1459 return -ENOMEM;
1460 *roots = ulist_alloc(GFP_NOFS);
1461 if (!*roots) {
1462 ulist_free(tmp);
1463 return -ENOMEM;
1464 }
1465
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001466 ULIST_ITER_INIT(&uiter);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001467 while (1) {
Lu Fengqiafce7722016-06-13 09:36:46 +08001468 ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001469 tmp, *roots, NULL, NULL, ignore_offset);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001470 if (ret < 0 && ret != -ENOENT) {
1471 ulist_free(tmp);
1472 ulist_free(*roots);
Filipe Manana580c0792020-07-13 15:11:56 +01001473 *roots = NULL;
Jan Schmidt8da6d582011-11-23 18:55:04 +01001474 return ret;
1475 }
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001476 node = ulist_next(tmp, &uiter);
Jan Schmidt8da6d582011-11-23 18:55:04 +01001477 if (!node)
1478 break;
1479 bytenr = node->val;
Wang Shilongbca1a292014-01-26 22:32:18 +08001480 cond_resched();
Jan Schmidt8da6d582011-11-23 18:55:04 +01001481 }
1482
1483 ulist_free(tmp);
1484 return 0;
1485}
1486
Josef Bacik9e351cc2014-03-13 15:42:13 -04001487int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1488 struct btrfs_fs_info *fs_info, u64 bytenr,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001489 u64 time_seq, struct ulist **roots,
1490 bool ignore_offset)
Josef Bacik9e351cc2014-03-13 15:42:13 -04001491{
1492 int ret;
1493
1494 if (!trans)
1495 down_read(&fs_info->commit_root_sem);
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001496 ret = btrfs_find_all_roots_safe(trans, fs_info, bytenr,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001497 time_seq, roots, ignore_offset);
Josef Bacik9e351cc2014-03-13 15:42:13 -04001498 if (!trans)
1499 up_read(&fs_info->commit_root_sem);
1500 return ret;
1501}
1502
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001503/**
Nikolay Borisov6e353e32021-01-22 11:58:00 +02001504 * Check if an extent is shared or not
1505 *
1506 * @root: root inode belongs to
1507 * @inum: inode number of the inode whose extent we are checking
1508 * @bytenr: logical bytenr of the extent we are checking
1509 * @roots: list of roots this extent is shared among
1510 * @tmp: temporary list used for iteration
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001511 *
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001512 * btrfs_check_shared uses the backref walking code but will short
1513 * circuit as soon as it finds a root or inode that doesn't match the
1514 * one passed in. This provides a significant performance benefit for
1515 * callers (such as fiemap) which want to know whether the extent is
1516 * shared but do not need a ref count.
1517 *
Filipe Manana03628cd2019-04-15 14:50:51 +01001518 * This attempts to attach to the running transaction in order to account for
1519 * delayed refs, but continues on even when no running transaction exists.
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001520 *
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001521 * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
1522 */
David Sterba5911c8f2019-05-15 15:31:04 +02001523int btrfs_check_shared(struct btrfs_root *root, u64 inum, u64 bytenr,
1524 struct ulist *roots, struct ulist *tmp)
Josef Bacikdc046b12014-09-10 16:20:45 -04001525{
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001526 struct btrfs_fs_info *fs_info = root->fs_info;
1527 struct btrfs_trans_handle *trans;
Josef Bacikdc046b12014-09-10 16:20:45 -04001528 struct ulist_iterator uiter;
1529 struct ulist_node *node;
David Sterba3284da72015-02-25 15:47:32 +01001530 struct seq_list elem = SEQ_LIST_INIT(elem);
Josef Bacikdc046b12014-09-10 16:20:45 -04001531 int ret = 0;
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001532 struct share_check shared = {
Misono Tomohiro4fd786e2018-08-06 14:25:24 +09001533 .root_objectid = root->root_key.objectid,
Edmund Nadolski3ec4d322017-07-12 16:20:10 -06001534 .inum = inum,
1535 .share_count = 0,
1536 };
Josef Bacikdc046b12014-09-10 16:20:45 -04001537
David Sterba5911c8f2019-05-15 15:31:04 +02001538 ulist_init(roots);
1539 ulist_init(tmp);
Josef Bacikdc046b12014-09-10 16:20:45 -04001540
Filipe Mananaa6d155d2019-07-29 09:37:10 +01001541 trans = btrfs_join_transaction_nostart(root);
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001542 if (IS_ERR(trans)) {
Filipe Manana03628cd2019-04-15 14:50:51 +01001543 if (PTR_ERR(trans) != -ENOENT && PTR_ERR(trans) != -EROFS) {
1544 ret = PTR_ERR(trans);
1545 goto out;
1546 }
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001547 trans = NULL;
Josef Bacikdc046b12014-09-10 16:20:45 -04001548 down_read(&fs_info->commit_root_sem);
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001549 } else {
1550 btrfs_get_tree_mod_seq(fs_info, &elem);
1551 }
1552
Josef Bacikdc046b12014-09-10 16:20:45 -04001553 ULIST_ITER_INIT(&uiter);
1554 while (1) {
1555 ret = find_parent_nodes(trans, fs_info, bytenr, elem.seq, tmp,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001556 roots, NULL, &shared, false);
Josef Bacikdc046b12014-09-10 16:20:45 -04001557 if (ret == BACKREF_FOUND_SHARED) {
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001558 /* this is the only condition under which we return 1 */
Josef Bacikdc046b12014-09-10 16:20:45 -04001559 ret = 1;
1560 break;
1561 }
1562 if (ret < 0 && ret != -ENOENT)
1563 break;
Mark Fasheh2c2ed5a2015-05-19 12:49:50 -07001564 ret = 0;
Josef Bacikdc046b12014-09-10 16:20:45 -04001565 node = ulist_next(tmp, &uiter);
1566 if (!node)
1567 break;
1568 bytenr = node->val;
Edmund Nadolski18bf5912018-03-14 09:03:11 -06001569 shared.share_count = 0;
Josef Bacikdc046b12014-09-10 16:20:45 -04001570 cond_resched();
1571 }
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001572
1573 if (trans) {
Josef Bacikdc046b12014-09-10 16:20:45 -04001574 btrfs_put_tree_mod_seq(fs_info, &elem);
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001575 btrfs_end_transaction(trans);
1576 } else {
Josef Bacikdc046b12014-09-10 16:20:45 -04001577 up_read(&fs_info->commit_root_sem);
Edmund Nadolskibb739cf2017-06-28 21:56:58 -06001578 }
Filipe Manana03628cd2019-04-15 14:50:51 +01001579out:
David Sterba5911c8f2019-05-15 15:31:04 +02001580 ulist_release(roots);
1581 ulist_release(tmp);
Josef Bacikdc046b12014-09-10 16:20:45 -04001582 return ret;
1583}
1584
Mark Fashehf1863732012-08-08 11:32:27 -07001585int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1586 u64 start_off, struct btrfs_path *path,
1587 struct btrfs_inode_extref **ret_extref,
1588 u64 *found_off)
1589{
1590 int ret, slot;
1591 struct btrfs_key key;
1592 struct btrfs_key found_key;
1593 struct btrfs_inode_extref *extref;
Jeff Mahoney73980be2017-06-28 21:56:55 -06001594 const struct extent_buffer *leaf;
Mark Fashehf1863732012-08-08 11:32:27 -07001595 unsigned long ptr;
1596
1597 key.objectid = inode_objectid;
David Sterba962a2982014-06-04 18:41:45 +02001598 key.type = BTRFS_INODE_EXTREF_KEY;
Mark Fashehf1863732012-08-08 11:32:27 -07001599 key.offset = start_off;
1600
1601 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1602 if (ret < 0)
1603 return ret;
1604
1605 while (1) {
1606 leaf = path->nodes[0];
1607 slot = path->slots[0];
1608 if (slot >= btrfs_header_nritems(leaf)) {
1609 /*
1610 * If the item at offset is not found,
1611 * btrfs_search_slot will point us to the slot
1612 * where it should be inserted. In our case
1613 * that will be the slot directly before the
1614 * next INODE_REF_KEY_V2 item. In the case
1615 * that we're pointing to the last slot in a
1616 * leaf, we must move one leaf over.
1617 */
1618 ret = btrfs_next_leaf(root, path);
1619 if (ret) {
1620 if (ret >= 1)
1621 ret = -ENOENT;
1622 break;
1623 }
1624 continue;
1625 }
1626
1627 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1628
1629 /*
1630 * Check that we're still looking at an extended ref key for
1631 * this particular objectid. If we have different
1632 * objectid or type then there are no more to be found
1633 * in the tree and we can exit.
1634 */
1635 ret = -ENOENT;
1636 if (found_key.objectid != inode_objectid)
1637 break;
David Sterba962a2982014-06-04 18:41:45 +02001638 if (found_key.type != BTRFS_INODE_EXTREF_KEY)
Mark Fashehf1863732012-08-08 11:32:27 -07001639 break;
1640
1641 ret = 0;
1642 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1643 extref = (struct btrfs_inode_extref *)ptr;
1644 *ret_extref = extref;
1645 if (found_off)
1646 *found_off = found_key.offset;
1647 break;
1648 }
1649
1650 return ret;
1651}
1652
Eric Sandeen48a3b632013-04-25 20:41:01 +00001653/*
1654 * this iterates to turn a name (from iref/extref) into a full filesystem path.
1655 * Elements of the path are separated by '/' and the path is guaranteed to be
1656 * 0-terminated. the path is only given within the current file system.
1657 * Therefore, it never starts with a '/'. the caller is responsible to provide
1658 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1659 * the start point of the resulting string is returned. this pointer is within
1660 * dest, normally.
1661 * in case the path buffer would overflow, the pointer is decremented further
1662 * as if output was written to the buffer, though no more output is actually
1663 * generated. that way, the caller can determine how much space would be
1664 * required for the path to fit into the buffer. in that case, the returned
1665 * value will be smaller than dest. callers must check this!
1666 */
Jan Schmidt96b5bd72012-10-15 08:30:45 +00001667char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1668 u32 name_len, unsigned long name_off,
1669 struct extent_buffer *eb_in, u64 parent,
1670 char *dest, u32 size)
Jan Schmidta542ad12011-06-13 19:52:59 +02001671{
Jan Schmidta542ad12011-06-13 19:52:59 +02001672 int slot;
1673 u64 next_inum;
1674 int ret;
Gabriel de Perthuis661bec62012-10-10 08:50:47 -06001675 s64 bytes_left = ((s64)size) - 1;
Jan Schmidta542ad12011-06-13 19:52:59 +02001676 struct extent_buffer *eb = eb_in;
1677 struct btrfs_key found_key;
Mark Fashehd24bec32012-08-08 11:33:54 -07001678 struct btrfs_inode_ref *iref;
Jan Schmidta542ad12011-06-13 19:52:59 +02001679
1680 if (bytes_left >= 0)
1681 dest[bytes_left] = '\0';
1682
1683 while (1) {
Mark Fashehd24bec32012-08-08 11:33:54 -07001684 bytes_left -= name_len;
Jan Schmidta542ad12011-06-13 19:52:59 +02001685 if (bytes_left >= 0)
1686 read_extent_buffer(eb, dest + bytes_left,
Mark Fashehd24bec32012-08-08 11:33:54 -07001687 name_off, name_len);
Jan Schmidtb916a592012-04-13 12:28:08 +02001688 if (eb != eb_in) {
Filipe Manana0c0fe3b2016-02-03 19:17:27 +00001689 if (!path->skip_locking)
Josef Bacikac5887c2020-08-20 11:46:10 -04001690 btrfs_tree_read_unlock(eb);
Jan Schmidta542ad12011-06-13 19:52:59 +02001691 free_extent_buffer(eb);
Jan Schmidtb916a592012-04-13 12:28:08 +02001692 }
David Sterbac234a242015-01-02 19:03:17 +01001693 ret = btrfs_find_item(fs_root, path, parent, 0,
1694 BTRFS_INODE_REF_KEY, &found_key);
Jan Schmidt8f24b492012-02-08 16:01:01 +01001695 if (ret > 0)
1696 ret = -ENOENT;
Jan Schmidta542ad12011-06-13 19:52:59 +02001697 if (ret)
1698 break;
Mark Fashehd24bec32012-08-08 11:33:54 -07001699
Jan Schmidta542ad12011-06-13 19:52:59 +02001700 next_inum = found_key.offset;
1701
1702 /* regular exit ahead */
1703 if (parent == next_inum)
1704 break;
1705
1706 slot = path->slots[0];
1707 eb = path->nodes[0];
1708 /* make sure we can use eb after releasing the path */
Jan Schmidtb916a592012-04-13 12:28:08 +02001709 if (eb != eb_in) {
Filipe Manana0c0fe3b2016-02-03 19:17:27 +00001710 path->nodes[0] = NULL;
1711 path->locks[0] = 0;
Jan Schmidtb916a592012-04-13 12:28:08 +02001712 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001713 btrfs_release_path(path);
Jan Schmidta542ad12011-06-13 19:52:59 +02001714 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
Mark Fashehd24bec32012-08-08 11:33:54 -07001715
1716 name_len = btrfs_inode_ref_name_len(eb, iref);
1717 name_off = (unsigned long)(iref + 1);
1718
Jan Schmidta542ad12011-06-13 19:52:59 +02001719 parent = next_inum;
1720 --bytes_left;
1721 if (bytes_left >= 0)
1722 dest[bytes_left] = '/';
1723 }
1724
1725 btrfs_release_path(path);
1726
1727 if (ret)
1728 return ERR_PTR(ret);
1729
1730 return dest + bytes_left;
1731}
1732
1733/*
1734 * this makes the path point to (logical EXTENT_ITEM *)
1735 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1736 * tree blocks and <0 on error.
1737 */
1738int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
Liu Bo69917e42012-09-07 20:01:28 -06001739 struct btrfs_path *path, struct btrfs_key *found_key,
1740 u64 *flags_ret)
Jan Schmidta542ad12011-06-13 19:52:59 +02001741{
1742 int ret;
1743 u64 flags;
Josef Bacik261c84b2013-06-28 13:11:22 -04001744 u64 size = 0;
Jan Schmidta542ad12011-06-13 19:52:59 +02001745 u32 item_size;
Jeff Mahoney73980be2017-06-28 21:56:55 -06001746 const struct extent_buffer *eb;
Jan Schmidta542ad12011-06-13 19:52:59 +02001747 struct btrfs_extent_item *ei;
1748 struct btrfs_key key;
1749
Josef Bacik261c84b2013-06-28 13:11:22 -04001750 if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1751 key.type = BTRFS_METADATA_ITEM_KEY;
1752 else
1753 key.type = BTRFS_EXTENT_ITEM_KEY;
Jan Schmidta542ad12011-06-13 19:52:59 +02001754 key.objectid = logical;
1755 key.offset = (u64)-1;
1756
1757 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1758 if (ret < 0)
1759 return ret;
Jan Schmidta542ad12011-06-13 19:52:59 +02001760
Wang Shilong850a8cd2014-02-06 20:02:29 +08001761 ret = btrfs_previous_extent_item(fs_info->extent_root, path, 0);
1762 if (ret) {
1763 if (ret > 0)
1764 ret = -ENOENT;
1765 return ret;
Josef Bacik580f0a62014-01-23 16:03:45 -05001766 }
Wang Shilong850a8cd2014-02-06 20:02:29 +08001767 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
Josef Bacik261c84b2013-06-28 13:11:22 -04001768 if (found_key->type == BTRFS_METADATA_ITEM_KEY)
Jeff Mahoneyda170662016-06-15 09:22:56 -04001769 size = fs_info->nodesize;
Josef Bacik261c84b2013-06-28 13:11:22 -04001770 else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
1771 size = found_key->offset;
1772
Josef Bacik580f0a62014-01-23 16:03:45 -05001773 if (found_key->objectid > logical ||
Josef Bacik261c84b2013-06-28 13:11:22 -04001774 found_key->objectid + size <= logical) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001775 btrfs_debug(fs_info,
1776 "logical %llu is not within any extent", logical);
Jan Schmidta542ad12011-06-13 19:52:59 +02001777 return -ENOENT;
Jan Schmidt4692cf52011-12-02 14:56:41 +01001778 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001779
1780 eb = path->nodes[0];
1781 item_size = btrfs_item_size_nr(eb, path->slots[0]);
1782 BUG_ON(item_size < sizeof(*ei));
1783
1784 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1785 flags = btrfs_extent_flags(eb, ei);
1786
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001787 btrfs_debug(fs_info,
1788 "logical %llu is at position %llu within the extent (%llu EXTENT_ITEM %llu) flags %#llx size %u",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001789 logical, logical - found_key->objectid, found_key->objectid,
1790 found_key->offset, flags, item_size);
Liu Bo69917e42012-09-07 20:01:28 -06001791
1792 WARN_ON(!flags_ret);
1793 if (flags_ret) {
1794 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1795 *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1796 else if (flags & BTRFS_EXTENT_FLAG_DATA)
1797 *flags_ret = BTRFS_EXTENT_FLAG_DATA;
1798 else
Arnd Bergmann290342f2019-03-25 14:02:25 +01001799 BUG();
Liu Bo69917e42012-09-07 20:01:28 -06001800 return 0;
1801 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001802
1803 return -EIO;
1804}
1805
1806/*
1807 * helper function to iterate extent inline refs. ptr must point to a 0 value
1808 * for the first call and may be modified. it is used to track state.
1809 * if more refs exist, 0 is returned and the next call to
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001810 * get_extent_inline_ref must pass the modified ptr parameter to get the
Jan Schmidta542ad12011-06-13 19:52:59 +02001811 * next ref. after the last ref was processed, 1 is returned.
1812 * returns <0 on error
1813 */
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001814static int get_extent_inline_ref(unsigned long *ptr,
1815 const struct extent_buffer *eb,
1816 const struct btrfs_key *key,
1817 const struct btrfs_extent_item *ei,
1818 u32 item_size,
1819 struct btrfs_extent_inline_ref **out_eiref,
1820 int *out_type)
Jan Schmidta542ad12011-06-13 19:52:59 +02001821{
1822 unsigned long end;
1823 u64 flags;
1824 struct btrfs_tree_block_info *info;
1825
1826 if (!*ptr) {
1827 /* first call */
1828 flags = btrfs_extent_flags(eb, ei);
1829 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
Liu Bo6eda71d2014-06-09 10:54:07 +08001830 if (key->type == BTRFS_METADATA_ITEM_KEY) {
1831 /* a skinny metadata extent */
1832 *out_eiref =
1833 (struct btrfs_extent_inline_ref *)(ei + 1);
1834 } else {
1835 WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
1836 info = (struct btrfs_tree_block_info *)(ei + 1);
1837 *out_eiref =
1838 (struct btrfs_extent_inline_ref *)(info + 1);
1839 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001840 } else {
1841 *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1842 }
1843 *ptr = (unsigned long)*out_eiref;
Liu Bocd857dd2014-06-08 19:04:13 +08001844 if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
Jan Schmidta542ad12011-06-13 19:52:59 +02001845 return -ENOENT;
1846 }
1847
1848 end = (unsigned long)ei + item_size;
Liu Bo6eda71d2014-06-09 10:54:07 +08001849 *out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
Liu Bo3de28d52017-08-18 15:15:19 -06001850 *out_type = btrfs_get_extent_inline_ref_type(eb, *out_eiref,
1851 BTRFS_REF_TYPE_ANY);
1852 if (*out_type == BTRFS_REF_TYPE_INVALID)
Su Yueaf431dc2018-06-22 16:18:01 +08001853 return -EUCLEAN;
Jan Schmidta542ad12011-06-13 19:52:59 +02001854
1855 *ptr += btrfs_extent_inline_ref_size(*out_type);
1856 WARN_ON(*ptr > end);
1857 if (*ptr == end)
1858 return 1; /* last */
1859
1860 return 0;
1861}
1862
1863/*
1864 * reads the tree block backref for an extent. tree level and root are returned
1865 * through out_level and out_root. ptr must point to a 0 value for the first
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001866 * call and may be modified (see get_extent_inline_ref comment).
Jan Schmidta542ad12011-06-13 19:52:59 +02001867 * returns 0 if data was provided, 1 if there was no more data to provide or
1868 * <0 on error.
1869 */
1870int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
Liu Bo6eda71d2014-06-09 10:54:07 +08001871 struct btrfs_key *key, struct btrfs_extent_item *ei,
1872 u32 item_size, u64 *out_root, u8 *out_level)
Jan Schmidta542ad12011-06-13 19:52:59 +02001873{
1874 int ret;
1875 int type;
Jan Schmidta542ad12011-06-13 19:52:59 +02001876 struct btrfs_extent_inline_ref *eiref;
1877
1878 if (*ptr == (unsigned long)-1)
1879 return 1;
1880
1881 while (1) {
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001882 ret = get_extent_inline_ref(ptr, eb, key, ei, item_size,
Liu Bo6eda71d2014-06-09 10:54:07 +08001883 &eiref, &type);
Jan Schmidta542ad12011-06-13 19:52:59 +02001884 if (ret < 0)
1885 return ret;
1886
1887 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1888 type == BTRFS_SHARED_BLOCK_REF_KEY)
1889 break;
1890
1891 if (ret == 1)
1892 return 1;
1893 }
1894
1895 /* we can treat both ref types equally here */
Jan Schmidta542ad12011-06-13 19:52:59 +02001896 *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
Filipe Mananaa1317f42014-12-15 16:04:42 +00001897
1898 if (key->type == BTRFS_EXTENT_ITEM_KEY) {
1899 struct btrfs_tree_block_info *info;
1900
1901 info = (struct btrfs_tree_block_info *)(ei + 1);
1902 *out_level = btrfs_tree_block_level(eb, info);
1903 } else {
1904 ASSERT(key->type == BTRFS_METADATA_ITEM_KEY);
1905 *out_level = (u8)key->offset;
1906 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001907
1908 if (ret == 1)
1909 *ptr = (unsigned long)-1;
1910
1911 return 0;
1912}
1913
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001914static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
1915 struct extent_inode_elem *inode_list,
1916 u64 root, u64 extent_item_objectid,
1917 iterate_extent_inodes_t *iterate, void *ctx)
Jan Schmidta542ad12011-06-13 19:52:59 +02001918{
Jan Schmidt976b1902012-05-17 16:43:03 +02001919 struct extent_inode_elem *eie;
Jan Schmidt4692cf52011-12-02 14:56:41 +01001920 int ret = 0;
Jan Schmidta542ad12011-06-13 19:52:59 +02001921
Jan Schmidt976b1902012-05-17 16:43:03 +02001922 for (eie = inode_list; eie; eie = eie->next) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001923 btrfs_debug(fs_info,
1924 "ref for %llu resolved, key (%llu EXTEND_DATA %llu), root %llu",
1925 extent_item_objectid, eie->inum,
1926 eie->offset, root);
Jan Schmidt976b1902012-05-17 16:43:03 +02001927 ret = iterate(eie->inum, eie->offset, root, ctx);
Jan Schmidt4692cf52011-12-02 14:56:41 +01001928 if (ret) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001929 btrfs_debug(fs_info,
1930 "stopping iteration for %llu due to ret=%d",
1931 extent_item_objectid, ret);
Jan Schmidt4692cf52011-12-02 14:56:41 +01001932 break;
1933 }
Jan Schmidta542ad12011-06-13 19:52:59 +02001934 }
1935
Jan Schmidta542ad12011-06-13 19:52:59 +02001936 return ret;
1937}
1938
1939/*
1940 * calls iterate() for every inode that references the extent identified by
Jan Schmidt4692cf52011-12-02 14:56:41 +01001941 * the given parameters.
Jan Schmidta542ad12011-06-13 19:52:59 +02001942 * when the iterator function returns a non-zero value, iteration stops.
1943 */
1944int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
Jan Schmidt4692cf52011-12-02 14:56:41 +01001945 u64 extent_item_objectid, u64 extent_item_pos,
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001946 int search_commit_root,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001947 iterate_extent_inodes_t *iterate, void *ctx,
1948 bool ignore_offset)
Jan Schmidta542ad12011-06-13 19:52:59 +02001949{
Jan Schmidta542ad12011-06-13 19:52:59 +02001950 int ret;
Josef Bacikda61d312013-06-12 16:20:08 -04001951 struct btrfs_trans_handle *trans = NULL;
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001952 struct ulist *refs = NULL;
1953 struct ulist *roots = NULL;
Jan Schmidt4692cf52011-12-02 14:56:41 +01001954 struct ulist_node *ref_node = NULL;
1955 struct ulist_node *root_node = NULL;
David Sterba3284da72015-02-25 15:47:32 +01001956 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001957 struct ulist_iterator ref_uiter;
1958 struct ulist_iterator root_uiter;
Jan Schmidta542ad12011-06-13 19:52:59 +02001959
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001960 btrfs_debug(fs_info, "resolving all inodes for extent %llu",
Jan Schmidt4692cf52011-12-02 14:56:41 +01001961 extent_item_objectid);
1962
Josef Bacikda61d312013-06-12 16:20:08 -04001963 if (!search_commit_root) {
Filipe Mananabfc61c32019-04-17 11:30:30 +01001964 trans = btrfs_attach_transaction(fs_info->extent_root);
1965 if (IS_ERR(trans)) {
1966 if (PTR_ERR(trans) != -ENOENT &&
1967 PTR_ERR(trans) != -EROFS)
1968 return PTR_ERR(trans);
1969 trans = NULL;
1970 }
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01001971 }
Jan Schmidt4692cf52011-12-02 14:56:41 +01001972
Filipe Mananabfc61c32019-04-17 11:30:30 +01001973 if (trans)
1974 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1975 else
1976 down_read(&fs_info->commit_root_sem);
1977
Jan Schmidt4692cf52011-12-02 14:56:41 +01001978 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
Jan Schmidt097b8a72012-06-21 11:08:04 +02001979 tree_mod_seq_elem.seq, &refs,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001980 &extent_item_pos, ignore_offset);
Jan Schmidt4692cf52011-12-02 14:56:41 +01001981 if (ret)
1982 goto out;
1983
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001984 ULIST_ITER_INIT(&ref_uiter);
1985 while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
Jeff Mahoneye0c476b2017-06-28 21:56:57 -06001986 ret = btrfs_find_all_roots_safe(trans, fs_info, ref_node->val,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04001987 tree_mod_seq_elem.seq, &roots,
1988 ignore_offset);
Jan Schmidt4692cf52011-12-02 14:56:41 +01001989 if (ret)
Jan Schmidta542ad12011-06-13 19:52:59 +02001990 break;
Jan Schmidtcd1b4132012-05-22 14:56:50 +02001991 ULIST_ITER_INIT(&root_uiter);
1992 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04001993 btrfs_debug(fs_info,
1994 "root %llu references leaf %llu, data list %#llx",
1995 root_node->val, ref_node->val,
1996 ref_node->aux);
1997 ret = iterate_leaf_refs(fs_info,
1998 (struct extent_inode_elem *)
Jan Schmidt995e01b2012-08-13 02:52:38 -06001999 (uintptr_t)ref_node->aux,
2000 root_node->val,
2001 extent_item_objectid,
2002 iterate, ctx);
Jan Schmidta542ad12011-06-13 19:52:59 +02002003 }
Jan Schmidt976b1902012-05-17 16:43:03 +02002004 ulist_free(roots);
Jan Schmidta542ad12011-06-13 19:52:59 +02002005 }
2006
Jan Schmidt976b1902012-05-17 16:43:03 +02002007 free_leaf_list(refs);
Jan Schmidt4692cf52011-12-02 14:56:41 +01002008out:
Filipe Mananabfc61c32019-04-17 11:30:30 +01002009 if (trans) {
Jan Schmidt8445f612012-05-16 18:36:03 +02002010 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
Jeff Mahoney3a45bb22016-09-09 21:39:03 -04002011 btrfs_end_transaction(trans);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002012 } else {
2013 up_read(&fs_info->commit_root_sem);
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01002014 }
2015
Jan Schmidta542ad12011-06-13 19:52:59 +02002016 return ret;
2017}
2018
2019int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
2020 struct btrfs_path *path,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04002021 iterate_extent_inodes_t *iterate, void *ctx,
2022 bool ignore_offset)
Jan Schmidta542ad12011-06-13 19:52:59 +02002023{
2024 int ret;
Jan Schmidt4692cf52011-12-02 14:56:41 +01002025 u64 extent_item_pos;
Liu Bo69917e42012-09-07 20:01:28 -06002026 u64 flags = 0;
Jan Schmidta542ad12011-06-13 19:52:59 +02002027 struct btrfs_key found_key;
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01002028 int search_commit_root = path->search_commit_root;
Jan Schmidta542ad12011-06-13 19:52:59 +02002029
Liu Bo69917e42012-09-07 20:01:28 -06002030 ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
Jan Schmidt4692cf52011-12-02 14:56:41 +01002031 btrfs_release_path(path);
Jan Schmidta542ad12011-06-13 19:52:59 +02002032 if (ret < 0)
2033 return ret;
Liu Bo69917e42012-09-07 20:01:28 -06002034 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
Stefan Behrens3627bf42012-08-01 04:28:01 -06002035 return -EINVAL;
Jan Schmidta542ad12011-06-13 19:52:59 +02002036
Jan Schmidt4692cf52011-12-02 14:56:41 +01002037 extent_item_pos = logical - found_key.objectid;
Jan Schmidt7a3ae2f2012-03-23 17:32:28 +01002038 ret = iterate_extent_inodes(fs_info, found_key.objectid,
2039 extent_item_pos, search_commit_root,
Zygo Blaxellc995ab32017-09-22 13:58:45 -04002040 iterate, ctx, ignore_offset);
Jan Schmidta542ad12011-06-13 19:52:59 +02002041
2042 return ret;
2043}
2044
Mark Fashehd24bec32012-08-08 11:33:54 -07002045typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
2046 struct extent_buffer *eb, void *ctx);
2047
2048static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
2049 struct btrfs_path *path,
2050 iterate_irefs_t *iterate, void *ctx)
Jan Schmidta542ad12011-06-13 19:52:59 +02002051{
Jan Schmidtaefc1eb2012-04-13 12:28:00 +02002052 int ret = 0;
Jan Schmidta542ad12011-06-13 19:52:59 +02002053 int slot;
2054 u32 cur;
2055 u32 len;
2056 u32 name_len;
2057 u64 parent = 0;
2058 int found = 0;
2059 struct extent_buffer *eb;
2060 struct btrfs_item *item;
2061 struct btrfs_inode_ref *iref;
2062 struct btrfs_key found_key;
2063
Jan Schmidtaefc1eb2012-04-13 12:28:00 +02002064 while (!ret) {
David Sterbac234a242015-01-02 19:03:17 +01002065 ret = btrfs_find_item(fs_root, path, inum,
2066 parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
2067 &found_key);
2068
Jan Schmidta542ad12011-06-13 19:52:59 +02002069 if (ret < 0)
2070 break;
2071 if (ret) {
2072 ret = found ? 0 : -ENOENT;
2073 break;
2074 }
2075 ++found;
2076
2077 parent = found_key.offset;
2078 slot = path->slots[0];
Filipe David Borba Manana3fe81ce2013-12-15 12:43:58 +00002079 eb = btrfs_clone_extent_buffer(path->nodes[0]);
2080 if (!eb) {
2081 ret = -ENOMEM;
2082 break;
2083 }
Jan Schmidta542ad12011-06-13 19:52:59 +02002084 btrfs_release_path(path);
2085
Ross Kirkdd3cc162013-09-16 15:58:09 +01002086 item = btrfs_item_nr(slot);
Jan Schmidta542ad12011-06-13 19:52:59 +02002087 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
2088
2089 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
2090 name_len = btrfs_inode_ref_name_len(eb, iref);
2091 /* path must be released before calling iterate()! */
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04002092 btrfs_debug(fs_root->fs_info,
2093 "following ref at offset %u for inode %llu in tree %llu",
Misono Tomohiro4fd786e2018-08-06 14:25:24 +09002094 cur, found_key.objectid,
2095 fs_root->root_key.objectid);
Mark Fashehd24bec32012-08-08 11:33:54 -07002096 ret = iterate(parent, name_len,
2097 (unsigned long)(iref + 1), eb, ctx);
Jan Schmidtaefc1eb2012-04-13 12:28:00 +02002098 if (ret)
Jan Schmidta542ad12011-06-13 19:52:59 +02002099 break;
Jan Schmidta542ad12011-06-13 19:52:59 +02002100 len = sizeof(*iref) + name_len;
2101 iref = (struct btrfs_inode_ref *)((char *)iref + len);
2102 }
2103 free_extent_buffer(eb);
2104 }
2105
2106 btrfs_release_path(path);
2107
2108 return ret;
2109}
2110
Mark Fashehd24bec32012-08-08 11:33:54 -07002111static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
2112 struct btrfs_path *path,
2113 iterate_irefs_t *iterate, void *ctx)
2114{
2115 int ret;
2116 int slot;
2117 u64 offset = 0;
2118 u64 parent;
2119 int found = 0;
2120 struct extent_buffer *eb;
2121 struct btrfs_inode_extref *extref;
Mark Fashehd24bec32012-08-08 11:33:54 -07002122 u32 item_size;
2123 u32 cur_offset;
2124 unsigned long ptr;
2125
2126 while (1) {
2127 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
2128 &offset);
2129 if (ret < 0)
2130 break;
2131 if (ret) {
2132 ret = found ? 0 : -ENOENT;
2133 break;
2134 }
2135 ++found;
2136
2137 slot = path->slots[0];
Filipe David Borba Manana3fe81ce2013-12-15 12:43:58 +00002138 eb = btrfs_clone_extent_buffer(path->nodes[0]);
2139 if (!eb) {
2140 ret = -ENOMEM;
2141 break;
2142 }
Mark Fashehd24bec32012-08-08 11:33:54 -07002143 btrfs_release_path(path);
2144
Chris Mason2849a852015-10-13 14:06:48 -04002145 item_size = btrfs_item_size_nr(eb, slot);
2146 ptr = btrfs_item_ptr_offset(eb, slot);
Mark Fashehd24bec32012-08-08 11:33:54 -07002147 cur_offset = 0;
2148
2149 while (cur_offset < item_size) {
2150 u32 name_len;
2151
2152 extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
2153 parent = btrfs_inode_extref_parent(eb, extref);
2154 name_len = btrfs_inode_extref_name_len(eb, extref);
2155 ret = iterate(parent, name_len,
2156 (unsigned long)&extref->name, eb, ctx);
2157 if (ret)
2158 break;
2159
Chris Mason2849a852015-10-13 14:06:48 -04002160 cur_offset += btrfs_inode_extref_name_len(eb, extref);
Mark Fashehd24bec32012-08-08 11:33:54 -07002161 cur_offset += sizeof(*extref);
2162 }
Mark Fashehd24bec32012-08-08 11:33:54 -07002163 free_extent_buffer(eb);
2164
2165 offset++;
2166 }
2167
2168 btrfs_release_path(path);
2169
2170 return ret;
2171}
2172
2173static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
2174 struct btrfs_path *path, iterate_irefs_t *iterate,
2175 void *ctx)
2176{
2177 int ret;
2178 int found_refs = 0;
2179
2180 ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
2181 if (!ret)
2182 ++found_refs;
2183 else if (ret != -ENOENT)
2184 return ret;
2185
2186 ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
2187 if (ret == -ENOENT && found_refs)
2188 return 0;
2189
2190 return ret;
2191}
2192
Jan Schmidta542ad12011-06-13 19:52:59 +02002193/*
2194 * returns 0 if the path could be dumped (probably truncated)
2195 * returns <0 in case of an error
2196 */
Mark Fashehd24bec32012-08-08 11:33:54 -07002197static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
2198 struct extent_buffer *eb, void *ctx)
Jan Schmidta542ad12011-06-13 19:52:59 +02002199{
2200 struct inode_fs_paths *ipath = ctx;
2201 char *fspath;
2202 char *fspath_min;
2203 int i = ipath->fspath->elem_cnt;
2204 const int s_ptr = sizeof(char *);
2205 u32 bytes_left;
2206
2207 bytes_left = ipath->fspath->bytes_left > s_ptr ?
2208 ipath->fspath->bytes_left - s_ptr : 0;
2209
Chris Mason740c3d22011-11-02 15:48:34 -04002210 fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
Jan Schmidt96b5bd72012-10-15 08:30:45 +00002211 fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
2212 name_off, eb, inum, fspath_min, bytes_left);
Jan Schmidta542ad12011-06-13 19:52:59 +02002213 if (IS_ERR(fspath))
2214 return PTR_ERR(fspath);
2215
2216 if (fspath > fspath_min) {
Jeff Mahoney745c4d82011-11-20 07:31:57 -05002217 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
Jan Schmidta542ad12011-06-13 19:52:59 +02002218 ++ipath->fspath->elem_cnt;
2219 ipath->fspath->bytes_left = fspath - fspath_min;
2220 } else {
2221 ++ipath->fspath->elem_missed;
2222 ipath->fspath->bytes_missing += fspath_min - fspath;
2223 ipath->fspath->bytes_left = 0;
2224 }
2225
2226 return 0;
2227}
2228
2229/*
2230 * this dumps all file system paths to the inode into the ipath struct, provided
2231 * is has been created large enough. each path is zero-terminated and accessed
Chris Mason740c3d22011-11-02 15:48:34 -04002232 * from ipath->fspath->val[i].
Jan Schmidta542ad12011-06-13 19:52:59 +02002233 * when it returns, there are ipath->fspath->elem_cnt number of paths available
Chris Mason740c3d22011-11-02 15:48:34 -04002234 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
Nicholas D Steeves01327612016-05-19 21:18:45 -04002235 * number of missed paths is recorded in ipath->fspath->elem_missed, otherwise,
Jan Schmidta542ad12011-06-13 19:52:59 +02002236 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
2237 * have been needed to return all paths.
2238 */
2239int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
2240{
2241 return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
Mark Fashehd24bec32012-08-08 11:33:54 -07002242 inode_to_path, ipath);
Jan Schmidta542ad12011-06-13 19:52:59 +02002243}
2244
Jan Schmidta542ad12011-06-13 19:52:59 +02002245struct btrfs_data_container *init_data_container(u32 total_bytes)
2246{
2247 struct btrfs_data_container *data;
2248 size_t alloc_bytes;
2249
2250 alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
David Sterbaf54de062017-05-31 19:32:09 +02002251 data = kvmalloc(alloc_bytes, GFP_KERNEL);
Jan Schmidta542ad12011-06-13 19:52:59 +02002252 if (!data)
2253 return ERR_PTR(-ENOMEM);
2254
2255 if (total_bytes >= sizeof(*data)) {
2256 data->bytes_left = total_bytes - sizeof(*data);
2257 data->bytes_missing = 0;
2258 } else {
2259 data->bytes_missing = sizeof(*data) - total_bytes;
2260 data->bytes_left = 0;
2261 }
2262
2263 data->elem_cnt = 0;
2264 data->elem_missed = 0;
2265
2266 return data;
2267}
2268
2269/*
2270 * allocates space to return multiple file system paths for an inode.
2271 * total_bytes to allocate are passed, note that space usable for actual path
2272 * information will be total_bytes - sizeof(struct inode_fs_paths).
2273 * the returned pointer must be freed with free_ipath() in the end.
2274 */
2275struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
2276 struct btrfs_path *path)
2277{
2278 struct inode_fs_paths *ifp;
2279 struct btrfs_data_container *fspath;
2280
2281 fspath = init_data_container(total_bytes);
2282 if (IS_ERR(fspath))
Misono Tomohiroafc69612018-07-26 10:22:58 +09002283 return ERR_CAST(fspath);
Jan Schmidta542ad12011-06-13 19:52:59 +02002284
David Sterbaf54de062017-05-31 19:32:09 +02002285 ifp = kmalloc(sizeof(*ifp), GFP_KERNEL);
Jan Schmidta542ad12011-06-13 19:52:59 +02002286 if (!ifp) {
David Sterbaf54de062017-05-31 19:32:09 +02002287 kvfree(fspath);
Jan Schmidta542ad12011-06-13 19:52:59 +02002288 return ERR_PTR(-ENOMEM);
2289 }
2290
2291 ifp->btrfs_path = path;
2292 ifp->fspath = fspath;
2293 ifp->fs_root = fs_root;
2294
2295 return ifp;
2296}
2297
2298void free_ipath(struct inode_fs_paths *ipath)
2299{
Jesper Juhl4735fb22012-04-12 22:47:52 +02002300 if (!ipath)
2301 return;
David Sterbaf54de062017-05-31 19:32:09 +02002302 kvfree(ipath->fspath);
Jan Schmidta542ad12011-06-13 19:52:59 +02002303 kfree(ipath);
2304}
Qu Wenruoa37f2322020-02-13 14:11:04 +08002305
2306struct btrfs_backref_iter *btrfs_backref_iter_alloc(
2307 struct btrfs_fs_info *fs_info, gfp_t gfp_flag)
2308{
2309 struct btrfs_backref_iter *ret;
2310
2311 ret = kzalloc(sizeof(*ret), gfp_flag);
2312 if (!ret)
2313 return NULL;
2314
2315 ret->path = btrfs_alloc_path();
Boleyn Suc15c2ec2020-08-06 15:31:44 +09002316 if (!ret->path) {
Qu Wenruoa37f2322020-02-13 14:11:04 +08002317 kfree(ret);
2318 return NULL;
2319 }
2320
2321 /* Current backref iterator only supports iteration in commit root */
2322 ret->path->search_commit_root = 1;
2323 ret->path->skip_locking = 1;
2324 ret->fs_info = fs_info;
2325
2326 return ret;
2327}
2328
2329int btrfs_backref_iter_start(struct btrfs_backref_iter *iter, u64 bytenr)
2330{
2331 struct btrfs_fs_info *fs_info = iter->fs_info;
2332 struct btrfs_path *path = iter->path;
2333 struct btrfs_extent_item *ei;
2334 struct btrfs_key key;
2335 int ret;
2336
2337 key.objectid = bytenr;
2338 key.type = BTRFS_METADATA_ITEM_KEY;
2339 key.offset = (u64)-1;
2340 iter->bytenr = bytenr;
2341
2342 ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
2343 if (ret < 0)
2344 return ret;
2345 if (ret == 0) {
2346 ret = -EUCLEAN;
2347 goto release;
2348 }
2349 if (path->slots[0] == 0) {
2350 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
2351 ret = -EUCLEAN;
2352 goto release;
2353 }
2354 path->slots[0]--;
2355
2356 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2357 if ((key.type != BTRFS_EXTENT_ITEM_KEY &&
2358 key.type != BTRFS_METADATA_ITEM_KEY) || key.objectid != bytenr) {
2359 ret = -ENOENT;
2360 goto release;
2361 }
2362 memcpy(&iter->cur_key, &key, sizeof(key));
2363 iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
2364 path->slots[0]);
2365 iter->end_ptr = (u32)(iter->item_ptr +
2366 btrfs_item_size_nr(path->nodes[0], path->slots[0]));
2367 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
2368 struct btrfs_extent_item);
2369
2370 /*
2371 * Only support iteration on tree backref yet.
2372 *
2373 * This is an extra precaution for non skinny-metadata, where
2374 * EXTENT_ITEM is also used for tree blocks, that we can only use
2375 * extent flags to determine if it's a tree block.
2376 */
2377 if (btrfs_extent_flags(path->nodes[0], ei) & BTRFS_EXTENT_FLAG_DATA) {
2378 ret = -ENOTSUPP;
2379 goto release;
2380 }
2381 iter->cur_ptr = (u32)(iter->item_ptr + sizeof(*ei));
2382
2383 /* If there is no inline backref, go search for keyed backref */
2384 if (iter->cur_ptr >= iter->end_ptr) {
2385 ret = btrfs_next_item(fs_info->extent_root, path);
2386
2387 /* No inline nor keyed ref */
2388 if (ret > 0) {
2389 ret = -ENOENT;
2390 goto release;
2391 }
2392 if (ret < 0)
2393 goto release;
2394
2395 btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key,
2396 path->slots[0]);
2397 if (iter->cur_key.objectid != bytenr ||
2398 (iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY &&
2399 iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY)) {
2400 ret = -ENOENT;
2401 goto release;
2402 }
2403 iter->cur_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
2404 path->slots[0]);
2405 iter->item_ptr = iter->cur_ptr;
2406 iter->end_ptr = (u32)(iter->item_ptr + btrfs_item_size_nr(
2407 path->nodes[0], path->slots[0]));
2408 }
2409
2410 return 0;
2411release:
2412 btrfs_backref_iter_release(iter);
2413 return ret;
2414}
Qu Wenruoc39c2dd2020-02-13 15:04:04 +08002415
2416/*
2417 * Go to the next backref item of current bytenr, can be either inlined or
2418 * keyed.
2419 *
2420 * Caller needs to check whether it's inline ref or not by iter->cur_key.
2421 *
2422 * Return 0 if we get next backref without problem.
2423 * Return >0 if there is no extra backref for this bytenr.
2424 * Return <0 if there is something wrong happened.
2425 */
2426int btrfs_backref_iter_next(struct btrfs_backref_iter *iter)
2427{
2428 struct extent_buffer *eb = btrfs_backref_get_eb(iter);
2429 struct btrfs_path *path = iter->path;
2430 struct btrfs_extent_inline_ref *iref;
2431 int ret;
2432 u32 size;
2433
2434 if (btrfs_backref_iter_is_inline_ref(iter)) {
2435 /* We're still inside the inline refs */
2436 ASSERT(iter->cur_ptr < iter->end_ptr);
2437
2438 if (btrfs_backref_has_tree_block_info(iter)) {
2439 /* First tree block info */
2440 size = sizeof(struct btrfs_tree_block_info);
2441 } else {
2442 /* Use inline ref type to determine the size */
2443 int type;
2444
2445 iref = (struct btrfs_extent_inline_ref *)
2446 ((unsigned long)iter->cur_ptr);
2447 type = btrfs_extent_inline_ref_type(eb, iref);
2448
2449 size = btrfs_extent_inline_ref_size(type);
2450 }
2451 iter->cur_ptr += size;
2452 if (iter->cur_ptr < iter->end_ptr)
2453 return 0;
2454
2455 /* All inline items iterated, fall through */
2456 }
2457
2458 /* We're at keyed items, there is no inline item, go to the next one */
2459 ret = btrfs_next_item(iter->fs_info->extent_root, iter->path);
2460 if (ret)
2461 return ret;
2462
2463 btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key, path->slots[0]);
2464 if (iter->cur_key.objectid != iter->bytenr ||
2465 (iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY &&
2466 iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY))
2467 return 1;
2468 iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
2469 path->slots[0]);
2470 iter->cur_ptr = iter->item_ptr;
2471 iter->end_ptr = iter->item_ptr + (u32)btrfs_item_size_nr(path->nodes[0],
2472 path->slots[0]);
2473 return 0;
2474}
Qu Wenruo584fb122020-03-03 13:14:41 +08002475
2476void btrfs_backref_init_cache(struct btrfs_fs_info *fs_info,
2477 struct btrfs_backref_cache *cache, int is_reloc)
2478{
2479 int i;
2480
2481 cache->rb_root = RB_ROOT;
2482 for (i = 0; i < BTRFS_MAX_LEVEL; i++)
2483 INIT_LIST_HEAD(&cache->pending[i]);
2484 INIT_LIST_HEAD(&cache->changed);
2485 INIT_LIST_HEAD(&cache->detached);
2486 INIT_LIST_HEAD(&cache->leaves);
2487 INIT_LIST_HEAD(&cache->pending_edge);
2488 INIT_LIST_HEAD(&cache->useless_node);
2489 cache->fs_info = fs_info;
2490 cache->is_reloc = is_reloc;
2491}
Qu Wenruob1818da2020-03-03 13:21:30 +08002492
2493struct btrfs_backref_node *btrfs_backref_alloc_node(
2494 struct btrfs_backref_cache *cache, u64 bytenr, int level)
2495{
2496 struct btrfs_backref_node *node;
2497
2498 ASSERT(level >= 0 && level < BTRFS_MAX_LEVEL);
2499 node = kzalloc(sizeof(*node), GFP_NOFS);
2500 if (!node)
2501 return node;
2502
2503 INIT_LIST_HEAD(&node->list);
2504 INIT_LIST_HEAD(&node->upper);
2505 INIT_LIST_HEAD(&node->lower);
2506 RB_CLEAR_NODE(&node->rb_node);
2507 cache->nr_nodes++;
2508 node->level = level;
2509 node->bytenr = bytenr;
2510
2511 return node;
2512}
Qu Wenruo47254d02020-03-03 13:22:57 +08002513
2514struct btrfs_backref_edge *btrfs_backref_alloc_edge(
2515 struct btrfs_backref_cache *cache)
2516{
2517 struct btrfs_backref_edge *edge;
2518
2519 edge = kzalloc(sizeof(*edge), GFP_NOFS);
2520 if (edge)
2521 cache->nr_edges++;
2522 return edge;
2523}
Qu Wenruo023acb02020-03-23 15:42:25 +08002524
2525/*
2526 * Drop the backref node from cache, also cleaning up all its
2527 * upper edges and any uncached nodes in the path.
2528 *
2529 * This cleanup happens bottom up, thus the node should either
2530 * be the lowest node in the cache or a detached node.
2531 */
2532void btrfs_backref_cleanup_node(struct btrfs_backref_cache *cache,
2533 struct btrfs_backref_node *node)
2534{
2535 struct btrfs_backref_node *upper;
2536 struct btrfs_backref_edge *edge;
2537
2538 if (!node)
2539 return;
2540
2541 BUG_ON(!node->lowest && !node->detached);
2542 while (!list_empty(&node->upper)) {
2543 edge = list_entry(node->upper.next, struct btrfs_backref_edge,
2544 list[LOWER]);
2545 upper = edge->node[UPPER];
2546 list_del(&edge->list[LOWER]);
2547 list_del(&edge->list[UPPER]);
2548 btrfs_backref_free_edge(cache, edge);
2549
Qu Wenruo023acb02020-03-23 15:42:25 +08002550 /*
2551 * Add the node to leaf node list if no other child block
2552 * cached.
2553 */
2554 if (list_empty(&upper->lower)) {
2555 list_add_tail(&upper->lower, &cache->leaves);
2556 upper->lowest = 1;
2557 }
2558 }
2559
2560 btrfs_backref_drop_node(cache, node);
2561}
Qu Wenruo13fe1bd2020-03-03 13:55:12 +08002562
2563/*
2564 * Release all nodes/edges from current cache
2565 */
2566void btrfs_backref_release_cache(struct btrfs_backref_cache *cache)
2567{
2568 struct btrfs_backref_node *node;
2569 int i;
2570
2571 while (!list_empty(&cache->detached)) {
2572 node = list_entry(cache->detached.next,
2573 struct btrfs_backref_node, list);
2574 btrfs_backref_cleanup_node(cache, node);
2575 }
2576
2577 while (!list_empty(&cache->leaves)) {
2578 node = list_entry(cache->leaves.next,
2579 struct btrfs_backref_node, lower);
2580 btrfs_backref_cleanup_node(cache, node);
2581 }
2582
2583 cache->last_trans = 0;
2584
2585 for (i = 0; i < BTRFS_MAX_LEVEL; i++)
2586 ASSERT(list_empty(&cache->pending[i]));
2587 ASSERT(list_empty(&cache->pending_edge));
2588 ASSERT(list_empty(&cache->useless_node));
2589 ASSERT(list_empty(&cache->changed));
2590 ASSERT(list_empty(&cache->detached));
2591 ASSERT(RB_EMPTY_ROOT(&cache->rb_root));
2592 ASSERT(!cache->nr_nodes);
2593 ASSERT(!cache->nr_edges);
2594}
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002595
2596/*
2597 * Handle direct tree backref
2598 *
2599 * Direct tree backref means, the backref item shows its parent bytenr
2600 * directly. This is for SHARED_BLOCK_REF backref (keyed or inlined).
2601 *
2602 * @ref_key: The converted backref key.
2603 * For keyed backref, it's the item key.
2604 * For inlined backref, objectid is the bytenr,
2605 * type is btrfs_inline_ref_type, offset is
2606 * btrfs_inline_ref_offset.
2607 */
2608static int handle_direct_tree_backref(struct btrfs_backref_cache *cache,
2609 struct btrfs_key *ref_key,
2610 struct btrfs_backref_node *cur)
2611{
2612 struct btrfs_backref_edge *edge;
2613 struct btrfs_backref_node *upper;
2614 struct rb_node *rb_node;
2615
2616 ASSERT(ref_key->type == BTRFS_SHARED_BLOCK_REF_KEY);
2617
2618 /* Only reloc root uses backref pointing to itself */
2619 if (ref_key->objectid == ref_key->offset) {
2620 struct btrfs_root *root;
2621
2622 cur->is_reloc_root = 1;
2623 /* Only reloc backref cache cares about a specific root */
2624 if (cache->is_reloc) {
2625 root = find_reloc_root(cache->fs_info, cur->bytenr);
Josef Bacikf78743f2021-01-14 14:02:44 -05002626 if (!root)
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002627 return -ENOENT;
2628 cur->root = root;
2629 } else {
2630 /*
2631 * For generic purpose backref cache, reloc root node
2632 * is useless.
2633 */
2634 list_add(&cur->list, &cache->useless_node);
2635 }
2636 return 0;
2637 }
2638
2639 edge = btrfs_backref_alloc_edge(cache);
2640 if (!edge)
2641 return -ENOMEM;
2642
2643 rb_node = rb_simple_search(&cache->rb_root, ref_key->offset);
2644 if (!rb_node) {
2645 /* Parent node not yet cached */
2646 upper = btrfs_backref_alloc_node(cache, ref_key->offset,
2647 cur->level + 1);
2648 if (!upper) {
2649 btrfs_backref_free_edge(cache, edge);
2650 return -ENOMEM;
2651 }
2652
2653 /*
2654 * Backrefs for the upper level block isn't cached, add the
2655 * block to pending list
2656 */
2657 list_add_tail(&edge->list[UPPER], &cache->pending_edge);
2658 } else {
2659 /* Parent node already cached */
2660 upper = rb_entry(rb_node, struct btrfs_backref_node, rb_node);
2661 ASSERT(upper->checked);
2662 INIT_LIST_HEAD(&edge->list[UPPER]);
2663 }
2664 btrfs_backref_link_edge(edge, cur, upper, LINK_LOWER);
2665 return 0;
2666}
2667
2668/*
2669 * Handle indirect tree backref
2670 *
2671 * Indirect tree backref means, we only know which tree the node belongs to.
2672 * We still need to do a tree search to find out the parents. This is for
2673 * TREE_BLOCK_REF backref (keyed or inlined).
2674 *
2675 * @ref_key: The same as @ref_key in handle_direct_tree_backref()
2676 * @tree_key: The first key of this tree block.
2677 * @path: A clean (released) path, to avoid allocating path everytime
2678 * the function get called.
2679 */
2680static int handle_indirect_tree_backref(struct btrfs_backref_cache *cache,
2681 struct btrfs_path *path,
2682 struct btrfs_key *ref_key,
2683 struct btrfs_key *tree_key,
2684 struct btrfs_backref_node *cur)
2685{
2686 struct btrfs_fs_info *fs_info = cache->fs_info;
2687 struct btrfs_backref_node *upper;
2688 struct btrfs_backref_node *lower;
2689 struct btrfs_backref_edge *edge;
2690 struct extent_buffer *eb;
2691 struct btrfs_root *root;
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002692 struct rb_node *rb_node;
2693 int level;
2694 bool need_check = true;
2695 int ret;
2696
David Sterba56e93572020-05-15 19:35:55 +02002697 root = btrfs_get_fs_root(fs_info, ref_key->offset, false);
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002698 if (IS_ERR(root))
2699 return PTR_ERR(root);
Qu Wenruo92a7cc42020-05-15 14:01:40 +08002700 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002701 cur->cowonly = 1;
2702
2703 if (btrfs_root_level(&root->root_item) == cur->level) {
2704 /* Tree root */
2705 ASSERT(btrfs_root_bytenr(&root->root_item) == cur->bytenr);
Qu Wenruo876de782020-03-16 14:10:01 +08002706 /*
2707 * For reloc backref cache, we may ignore reloc root. But for
2708 * general purpose backref cache, we can't rely on
2709 * btrfs_should_ignore_reloc_root() as it may conflict with
2710 * current running relocation and lead to missing root.
2711 *
2712 * For general purpose backref cache, reloc root detection is
2713 * completely relying on direct backref (key->offset is parent
2714 * bytenr), thus only do such check for reloc cache.
2715 */
2716 if (btrfs_should_ignore_reloc_root(root) && cache->is_reloc) {
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002717 btrfs_put_root(root);
2718 list_add(&cur->list, &cache->useless_node);
2719 } else {
2720 cur->root = root;
2721 }
2722 return 0;
2723 }
2724
2725 level = cur->level + 1;
2726
2727 /* Search the tree to find parent blocks referring to the block */
2728 path->search_commit_root = 1;
2729 path->skip_locking = 1;
2730 path->lowest_level = level;
2731 ret = btrfs_search_slot(NULL, root, tree_key, path, 0, 0);
2732 path->lowest_level = 0;
2733 if (ret < 0) {
2734 btrfs_put_root(root);
2735 return ret;
2736 }
2737 if (ret > 0 && path->slots[level] > 0)
2738 path->slots[level]--;
2739
2740 eb = path->nodes[level];
2741 if (btrfs_node_blockptr(eb, path->slots[level]) != cur->bytenr) {
2742 btrfs_err(fs_info,
2743"couldn't find block (%llu) (level %d) in tree (%llu) with key (%llu %u %llu)",
2744 cur->bytenr, level - 1, root->root_key.objectid,
2745 tree_key->objectid, tree_key->type, tree_key->offset);
2746 btrfs_put_root(root);
2747 ret = -ENOENT;
2748 goto out;
2749 }
2750 lower = cur;
2751
2752 /* Add all nodes and edges in the path */
2753 for (; level < BTRFS_MAX_LEVEL; level++) {
2754 if (!path->nodes[level]) {
2755 ASSERT(btrfs_root_bytenr(&root->root_item) ==
2756 lower->bytenr);
Qu Wenruo876de782020-03-16 14:10:01 +08002757 /* Same as previous should_ignore_reloc_root() call */
2758 if (btrfs_should_ignore_reloc_root(root) &&
2759 cache->is_reloc) {
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002760 btrfs_put_root(root);
2761 list_add(&lower->list, &cache->useless_node);
2762 } else {
2763 lower->root = root;
2764 }
2765 break;
2766 }
2767
2768 edge = btrfs_backref_alloc_edge(cache);
2769 if (!edge) {
2770 btrfs_put_root(root);
2771 ret = -ENOMEM;
2772 goto out;
2773 }
2774
2775 eb = path->nodes[level];
2776 rb_node = rb_simple_search(&cache->rb_root, eb->start);
2777 if (!rb_node) {
2778 upper = btrfs_backref_alloc_node(cache, eb->start,
2779 lower->level + 1);
2780 if (!upper) {
2781 btrfs_put_root(root);
2782 btrfs_backref_free_edge(cache, edge);
2783 ret = -ENOMEM;
2784 goto out;
2785 }
2786 upper->owner = btrfs_header_owner(eb);
Qu Wenruo92a7cc42020-05-15 14:01:40 +08002787 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002788 upper->cowonly = 1;
2789
2790 /*
2791 * If we know the block isn't shared we can avoid
2792 * checking its backrefs.
2793 */
2794 if (btrfs_block_can_be_shared(root, eb))
2795 upper->checked = 0;
2796 else
2797 upper->checked = 1;
2798
2799 /*
2800 * Add the block to pending list if we need to check its
2801 * backrefs, we only do this once while walking up a
2802 * tree as we will catch anything else later on.
2803 */
2804 if (!upper->checked && need_check) {
2805 need_check = false;
2806 list_add_tail(&edge->list[UPPER],
2807 &cache->pending_edge);
2808 } else {
2809 if (upper->checked)
2810 need_check = true;
2811 INIT_LIST_HEAD(&edge->list[UPPER]);
2812 }
2813 } else {
2814 upper = rb_entry(rb_node, struct btrfs_backref_node,
2815 rb_node);
2816 ASSERT(upper->checked);
2817 INIT_LIST_HEAD(&edge->list[UPPER]);
2818 if (!upper->owner)
2819 upper->owner = btrfs_header_owner(eb);
2820 }
2821 btrfs_backref_link_edge(edge, lower, upper, LINK_LOWER);
2822
2823 if (rb_node) {
2824 btrfs_put_root(root);
2825 break;
2826 }
2827 lower = upper;
2828 upper = NULL;
2829 }
2830out:
2831 btrfs_release_path(path);
2832 return ret;
2833}
2834
2835/*
2836 * Add backref node @cur into @cache.
2837 *
2838 * NOTE: Even if the function returned 0, @cur is not yet cached as its upper
2839 * links aren't yet bi-directional. Needs to finish such links.
Qu Wenruofc997ed2020-03-23 16:14:08 +08002840 * Use btrfs_backref_finish_upper_links() to finish such linkage.
Qu Wenruo1b60d2e2020-03-23 16:08:34 +08002841 *
2842 * @path: Released path for indirect tree backref lookup
2843 * @iter: Released backref iter for extent tree search
2844 * @node_key: The first key of the tree block
2845 */
2846int btrfs_backref_add_tree_node(struct btrfs_backref_cache *cache,
2847 struct btrfs_path *path,
2848 struct btrfs_backref_iter *iter,
2849 struct btrfs_key *node_key,
2850 struct btrfs_backref_node *cur)
2851{
2852 struct btrfs_fs_info *fs_info = cache->fs_info;
2853 struct btrfs_backref_edge *edge;
2854 struct btrfs_backref_node *exist;
2855 int ret;
2856
2857 ret = btrfs_backref_iter_start(iter, cur->bytenr);
2858 if (ret < 0)
2859 return ret;
2860 /*
2861 * We skip the first btrfs_tree_block_info, as we don't use the key
2862 * stored in it, but fetch it from the tree block
2863 */
2864 if (btrfs_backref_has_tree_block_info(iter)) {
2865 ret = btrfs_backref_iter_next(iter);
2866 if (ret < 0)
2867 goto out;
2868 /* No extra backref? This means the tree block is corrupted */
2869 if (ret > 0) {
2870 ret = -EUCLEAN;
2871 goto out;
2872 }
2873 }
2874 WARN_ON(cur->checked);
2875 if (!list_empty(&cur->upper)) {
2876 /*
2877 * The backref was added previously when processing backref of
2878 * type BTRFS_TREE_BLOCK_REF_KEY
2879 */
2880 ASSERT(list_is_singular(&cur->upper));
2881 edge = list_entry(cur->upper.next, struct btrfs_backref_edge,
2882 list[LOWER]);
2883 ASSERT(list_empty(&edge->list[UPPER]));
2884 exist = edge->node[UPPER];
2885 /*
2886 * Add the upper level block to pending list if we need check
2887 * its backrefs
2888 */
2889 if (!exist->checked)
2890 list_add_tail(&edge->list[UPPER], &cache->pending_edge);
2891 } else {
2892 exist = NULL;
2893 }
2894
2895 for (; ret == 0; ret = btrfs_backref_iter_next(iter)) {
2896 struct extent_buffer *eb;
2897 struct btrfs_key key;
2898 int type;
2899
2900 cond_resched();
2901 eb = btrfs_backref_get_eb(iter);
2902
2903 key.objectid = iter->bytenr;
2904 if (btrfs_backref_iter_is_inline_ref(iter)) {
2905 struct btrfs_extent_inline_ref *iref;
2906
2907 /* Update key for inline backref */
2908 iref = (struct btrfs_extent_inline_ref *)
2909 ((unsigned long)iter->cur_ptr);
2910 type = btrfs_get_extent_inline_ref_type(eb, iref,
2911 BTRFS_REF_TYPE_BLOCK);
2912 if (type == BTRFS_REF_TYPE_INVALID) {
2913 ret = -EUCLEAN;
2914 goto out;
2915 }
2916 key.type = type;
2917 key.offset = btrfs_extent_inline_ref_offset(eb, iref);
2918 } else {
2919 key.type = iter->cur_key.type;
2920 key.offset = iter->cur_key.offset;
2921 }
2922
2923 /*
2924 * Parent node found and matches current inline ref, no need to
2925 * rebuild this node for this inline ref
2926 */
2927 if (exist &&
2928 ((key.type == BTRFS_TREE_BLOCK_REF_KEY &&
2929 exist->owner == key.offset) ||
2930 (key.type == BTRFS_SHARED_BLOCK_REF_KEY &&
2931 exist->bytenr == key.offset))) {
2932 exist = NULL;
2933 continue;
2934 }
2935
2936 /* SHARED_BLOCK_REF means key.offset is the parent bytenr */
2937 if (key.type == BTRFS_SHARED_BLOCK_REF_KEY) {
2938 ret = handle_direct_tree_backref(cache, &key, cur);
2939 if (ret < 0)
2940 goto out;
2941 continue;
2942 } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
2943 ret = -EINVAL;
2944 btrfs_print_v0_err(fs_info);
2945 btrfs_handle_fs_error(fs_info, ret, NULL);
2946 goto out;
2947 } else if (key.type != BTRFS_TREE_BLOCK_REF_KEY) {
2948 continue;
2949 }
2950
2951 /*
2952 * key.type == BTRFS_TREE_BLOCK_REF_KEY, inline ref offset
2953 * means the root objectid. We need to search the tree to get
2954 * its parent bytenr.
2955 */
2956 ret = handle_indirect_tree_backref(cache, path, &key, node_key,
2957 cur);
2958 if (ret < 0)
2959 goto out;
2960 }
2961 ret = 0;
2962 cur->checked = 1;
2963 WARN_ON(exist);
2964out:
2965 btrfs_backref_iter_release(iter);
2966 return ret;
2967}
Qu Wenruofc997ed2020-03-23 16:14:08 +08002968
2969/*
2970 * Finish the upwards linkage created by btrfs_backref_add_tree_node()
2971 */
2972int btrfs_backref_finish_upper_links(struct btrfs_backref_cache *cache,
2973 struct btrfs_backref_node *start)
2974{
2975 struct list_head *useless_node = &cache->useless_node;
2976 struct btrfs_backref_edge *edge;
2977 struct rb_node *rb_node;
2978 LIST_HEAD(pending_edge);
2979
2980 ASSERT(start->checked);
2981
2982 /* Insert this node to cache if it's not COW-only */
2983 if (!start->cowonly) {
2984 rb_node = rb_simple_insert(&cache->rb_root, start->bytenr,
2985 &start->rb_node);
2986 if (rb_node)
2987 btrfs_backref_panic(cache->fs_info, start->bytenr,
2988 -EEXIST);
2989 list_add_tail(&start->lower, &cache->leaves);
2990 }
2991
2992 /*
2993 * Use breadth first search to iterate all related edges.
2994 *
2995 * The starting points are all the edges of this node
2996 */
2997 list_for_each_entry(edge, &start->upper, list[LOWER])
2998 list_add_tail(&edge->list[UPPER], &pending_edge);
2999
3000 while (!list_empty(&pending_edge)) {
3001 struct btrfs_backref_node *upper;
3002 struct btrfs_backref_node *lower;
Qu Wenruofc997ed2020-03-23 16:14:08 +08003003
3004 edge = list_first_entry(&pending_edge,
3005 struct btrfs_backref_edge, list[UPPER]);
3006 list_del_init(&edge->list[UPPER]);
3007 upper = edge->node[UPPER];
3008 lower = edge->node[LOWER];
3009
3010 /* Parent is detached, no need to keep any edges */
3011 if (upper->detached) {
3012 list_del(&edge->list[LOWER]);
3013 btrfs_backref_free_edge(cache, edge);
3014
3015 /* Lower node is orphan, queue for cleanup */
3016 if (list_empty(&lower->upper))
3017 list_add(&lower->list, useless_node);
3018 continue;
3019 }
3020
3021 /*
3022 * All new nodes added in current build_backref_tree() haven't
3023 * been linked to the cache rb tree.
3024 * So if we have upper->rb_node populated, this means a cache
3025 * hit. We only need to link the edge, as @upper and all its
3026 * parents have already been linked.
3027 */
3028 if (!RB_EMPTY_NODE(&upper->rb_node)) {
3029 if (upper->lowest) {
3030 list_del_init(&upper->lower);
3031 upper->lowest = 0;
3032 }
3033
3034 list_add_tail(&edge->list[UPPER], &upper->lower);
3035 continue;
3036 }
3037
3038 /* Sanity check, we shouldn't have any unchecked nodes */
3039 if (!upper->checked) {
3040 ASSERT(0);
3041 return -EUCLEAN;
3042 }
3043
3044 /* Sanity check, COW-only node has non-COW-only parent */
3045 if (start->cowonly != upper->cowonly) {
3046 ASSERT(0);
3047 return -EUCLEAN;
3048 }
3049
3050 /* Only cache non-COW-only (subvolume trees) tree blocks */
3051 if (!upper->cowonly) {
3052 rb_node = rb_simple_insert(&cache->rb_root, upper->bytenr,
3053 &upper->rb_node);
3054 if (rb_node) {
3055 btrfs_backref_panic(cache->fs_info,
3056 upper->bytenr, -EEXIST);
3057 return -EUCLEAN;
3058 }
3059 }
3060
3061 list_add_tail(&edge->list[UPPER], &upper->lower);
3062
3063 /*
3064 * Also queue all the parent edges of this uncached node
3065 * to finish the upper linkage
3066 */
3067 list_for_each_entry(edge, &upper->upper, list[LOWER])
3068 list_add_tail(&edge->list[UPPER], &pending_edge);
3069 }
3070 return 0;
3071}
Qu Wenruo1b23ea12020-03-23 16:57:15 +08003072
3073void btrfs_backref_error_cleanup(struct btrfs_backref_cache *cache,
3074 struct btrfs_backref_node *node)
3075{
3076 struct btrfs_backref_node *lower;
3077 struct btrfs_backref_node *upper;
3078 struct btrfs_backref_edge *edge;
3079
3080 while (!list_empty(&cache->useless_node)) {
3081 lower = list_first_entry(&cache->useless_node,
3082 struct btrfs_backref_node, list);
3083 list_del_init(&lower->list);
3084 }
3085 while (!list_empty(&cache->pending_edge)) {
3086 edge = list_first_entry(&cache->pending_edge,
3087 struct btrfs_backref_edge, list[UPPER]);
3088 list_del(&edge->list[UPPER]);
3089 list_del(&edge->list[LOWER]);
3090 lower = edge->node[LOWER];
3091 upper = edge->node[UPPER];
3092 btrfs_backref_free_edge(cache, edge);
3093
3094 /*
3095 * Lower is no longer linked to any upper backref nodes and
3096 * isn't in the cache, we can free it ourselves.
3097 */
3098 if (list_empty(&lower->upper) &&
3099 RB_EMPTY_NODE(&lower->rb_node))
3100 list_add(&lower->list, &cache->useless_node);
3101
3102 if (!RB_EMPTY_NODE(&upper->rb_node))
3103 continue;
3104
3105 /* Add this guy's upper edges to the list to process */
3106 list_for_each_entry(edge, &upper->upper, list[LOWER])
3107 list_add_tail(&edge->list[UPPER],
3108 &cache->pending_edge);
3109 if (list_empty(&upper->upper))
3110 list_add(&upper->list, &cache->useless_node);
3111 }
3112
3113 while (!list_empty(&cache->useless_node)) {
3114 lower = list_first_entry(&cache->useless_node,
3115 struct btrfs_backref_node, list);
3116 list_del_init(&lower->list);
3117 if (lower == node)
3118 node = NULL;
Josef Bacik49ecc672020-12-16 11:22:11 -05003119 btrfs_backref_drop_node(cache, lower);
Qu Wenruo1b23ea12020-03-23 16:57:15 +08003120 }
3121
3122 btrfs_backref_cleanup_node(cache, node);
3123 ASSERT(list_empty(&cache->useless_node) &&
3124 list_empty(&cache->pending_edge));
3125}