blob: 4cefcb6a47a582f504e5515ca8f30067ab9a0ee8 [file] [log] [blame]
Mark Fashehccd979b2005-12-15 14:31:24 -08001/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * alloc.c
5 *
6 * Extent allocs and frees
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/fs.h>
27#include <linux/types.h>
28#include <linux/slab.h>
29#include <linux/highmem.h>
Mark Fasheh60b11392007-02-16 11:46:50 -080030#include <linux/swap.h>
Mark Fashehccd979b2005-12-15 14:31:24 -080031
32#define MLOG_MASK_PREFIX ML_DISK_ALLOC
33#include <cluster/masklog.h>
34
35#include "ocfs2.h"
36
37#include "alloc.h"
Mark Fasheh60b11392007-02-16 11:46:50 -080038#include "aops.h"
Mark Fashehccd979b2005-12-15 14:31:24 -080039#include "dlmglue.h"
40#include "extent_map.h"
41#include "inode.h"
42#include "journal.h"
43#include "localalloc.h"
44#include "suballoc.h"
45#include "sysfile.h"
46#include "file.h"
47#include "super.h"
48#include "uptodate.h"
49
50#include "buffer_head_io.h"
51
Tao Mae7d4cb62008-08-18 17:38:44 +080052/*
53 * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract
54 * the b-tree operations in ocfs2. Now all the b-tree operations are not
55 * limited to ocfs2_dinode only. Any data which need to allocate clusters
56 * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree
57 * and operation.
58 *
59 * ocfs2_extent_tree contains info for the root of the b-tree, it must have a
60 * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree
61 * functions.
62 * ocfs2_extent_tree_operations abstract the normal operations we do for
63 * the root of extent b-tree.
64 */
65struct ocfs2_extent_tree;
66
67struct ocfs2_extent_tree_operations {
Joel Becker35dc0aa2008-08-20 16:25:06 -070068 void (*eo_set_last_eb_blk)(struct ocfs2_extent_tree *et,
69 u64 blkno);
70 u64 (*eo_get_last_eb_blk)(struct ocfs2_extent_tree *et);
71 void (*eo_update_clusters)(struct inode *inode,
72 struct ocfs2_extent_tree *et,
73 u32 new_clusters);
74 int (*eo_sanity_check)(struct inode *inode, struct ocfs2_extent_tree *et);
Tao Mae7d4cb62008-08-18 17:38:44 +080075};
76
77struct ocfs2_extent_tree {
Joel Beckerce1d9ea2008-08-20 16:30:07 -070078 enum ocfs2_extent_tree_type et_type;
79 struct ocfs2_extent_tree_operations *et_ops;
80 struct buffer_head *et_root_bh;
81 struct ocfs2_extent_list *et_root_el;
Joel Beckerea5efa12008-08-20 16:57:27 -070082 void *et_object;
Joel Beckerce1d9ea2008-08-20 16:30:07 -070083 unsigned int et_max_leaf_clusters;
Tao Mae7d4cb62008-08-18 17:38:44 +080084};
85
86static void ocfs2_dinode_set_last_eb_blk(struct ocfs2_extent_tree *et,
87 u64 blkno)
88{
Joel Beckerea5efa12008-08-20 16:57:27 -070089 struct ocfs2_dinode *di = et->et_object;
Tao Mae7d4cb62008-08-18 17:38:44 +080090
Joel Beckerce1d9ea2008-08-20 16:30:07 -070091 BUG_ON(et->et_type != OCFS2_DINODE_EXTENT);
Tao Mae7d4cb62008-08-18 17:38:44 +080092 di->i_last_eb_blk = cpu_to_le64(blkno);
93}
94
95static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et)
96{
Joel Beckerea5efa12008-08-20 16:57:27 -070097 struct ocfs2_dinode *di = et->et_object;
Tao Mae7d4cb62008-08-18 17:38:44 +080098
Joel Beckerce1d9ea2008-08-20 16:30:07 -070099 BUG_ON(et->et_type != OCFS2_DINODE_EXTENT);
Tao Mae7d4cb62008-08-18 17:38:44 +0800100 return le64_to_cpu(di->i_last_eb_blk);
101}
102
103static void ocfs2_dinode_update_clusters(struct inode *inode,
104 struct ocfs2_extent_tree *et,
105 u32 clusters)
106{
Joel Beckerea5efa12008-08-20 16:57:27 -0700107 struct ocfs2_dinode *di = et->et_object;
Tao Mae7d4cb62008-08-18 17:38:44 +0800108
109 le32_add_cpu(&di->i_clusters, clusters);
110 spin_lock(&OCFS2_I(inode)->ip_lock);
111 OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters);
112 spin_unlock(&OCFS2_I(inode)->ip_lock);
113}
114
115static int ocfs2_dinode_sanity_check(struct inode *inode,
116 struct ocfs2_extent_tree *et)
117{
118 int ret = 0;
119 struct ocfs2_dinode *di;
120
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700121 BUG_ON(et->et_type != OCFS2_DINODE_EXTENT);
Tao Mae7d4cb62008-08-18 17:38:44 +0800122
Joel Beckerea5efa12008-08-20 16:57:27 -0700123 di = et->et_object;
Tao Mae7d4cb62008-08-18 17:38:44 +0800124 if (!OCFS2_IS_VALID_DINODE(di)) {
125 ret = -EIO;
126 ocfs2_error(inode->i_sb,
127 "Inode %llu has invalid path root",
128 (unsigned long long)OCFS2_I(inode)->ip_blkno);
129 }
130
131 return ret;
132}
133
134static struct ocfs2_extent_tree_operations ocfs2_dinode_et_ops = {
Joel Becker35dc0aa2008-08-20 16:25:06 -0700135 .eo_set_last_eb_blk = ocfs2_dinode_set_last_eb_blk,
136 .eo_get_last_eb_blk = ocfs2_dinode_get_last_eb_blk,
137 .eo_update_clusters = ocfs2_dinode_update_clusters,
138 .eo_sanity_check = ocfs2_dinode_sanity_check,
Tao Mae7d4cb62008-08-18 17:38:44 +0800139};
140
Tao Maf56654c2008-08-18 17:38:48 +0800141static void ocfs2_xattr_value_set_last_eb_blk(struct ocfs2_extent_tree *et,
142 u64 blkno)
143{
144 struct ocfs2_xattr_value_root *xv =
Joel Beckerea5efa12008-08-20 16:57:27 -0700145 (struct ocfs2_xattr_value_root *)et->et_object;
Tao Maf56654c2008-08-18 17:38:48 +0800146
147 xv->xr_last_eb_blk = cpu_to_le64(blkno);
148}
149
150static u64 ocfs2_xattr_value_get_last_eb_blk(struct ocfs2_extent_tree *et)
151{
152 struct ocfs2_xattr_value_root *xv =
Joel Beckerea5efa12008-08-20 16:57:27 -0700153 (struct ocfs2_xattr_value_root *) et->et_object;
Tao Maf56654c2008-08-18 17:38:48 +0800154
155 return le64_to_cpu(xv->xr_last_eb_blk);
156}
157
158static void ocfs2_xattr_value_update_clusters(struct inode *inode,
159 struct ocfs2_extent_tree *et,
160 u32 clusters)
161{
162 struct ocfs2_xattr_value_root *xv =
Joel Beckerea5efa12008-08-20 16:57:27 -0700163 (struct ocfs2_xattr_value_root *)et->et_object;
Tao Maf56654c2008-08-18 17:38:48 +0800164
165 le32_add_cpu(&xv->xr_clusters, clusters);
166}
167
168static int ocfs2_xattr_value_sanity_check(struct inode *inode,
169 struct ocfs2_extent_tree *et)
170{
171 return 0;
172}
173
174static struct ocfs2_extent_tree_operations ocfs2_xattr_et_ops = {
Joel Becker35dc0aa2008-08-20 16:25:06 -0700175 .eo_set_last_eb_blk = ocfs2_xattr_value_set_last_eb_blk,
176 .eo_get_last_eb_blk = ocfs2_xattr_value_get_last_eb_blk,
177 .eo_update_clusters = ocfs2_xattr_value_update_clusters,
178 .eo_sanity_check = ocfs2_xattr_value_sanity_check,
Tao Maf56654c2008-08-18 17:38:48 +0800179};
180
Tao Maba492612008-08-18 17:38:49 +0800181static void ocfs2_xattr_tree_set_last_eb_blk(struct ocfs2_extent_tree *et,
182 u64 blkno)
183{
Joel Beckerea5efa12008-08-20 16:57:27 -0700184 struct ocfs2_xattr_block *xb = et->et_object;
Tao Maba492612008-08-18 17:38:49 +0800185 struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root;
186
187 xt->xt_last_eb_blk = cpu_to_le64(blkno);
188}
189
190static u64 ocfs2_xattr_tree_get_last_eb_blk(struct ocfs2_extent_tree *et)
191{
Joel Beckerea5efa12008-08-20 16:57:27 -0700192 struct ocfs2_xattr_block *xb = et->et_object;
Tao Maba492612008-08-18 17:38:49 +0800193 struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root;
194
195 return le64_to_cpu(xt->xt_last_eb_blk);
196}
197
198static void ocfs2_xattr_tree_update_clusters(struct inode *inode,
199 struct ocfs2_extent_tree *et,
200 u32 clusters)
201{
Joel Beckerea5efa12008-08-20 16:57:27 -0700202 struct ocfs2_xattr_block *xb = et->et_object;
Tao Maba492612008-08-18 17:38:49 +0800203
204 le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, clusters);
205}
206
207static int ocfs2_xattr_tree_sanity_check(struct inode *inode,
208 struct ocfs2_extent_tree *et)
209{
210 return 0;
211}
212
213static struct ocfs2_extent_tree_operations ocfs2_xattr_tree_et_ops = {
Joel Becker35dc0aa2008-08-20 16:25:06 -0700214 .eo_set_last_eb_blk = ocfs2_xattr_tree_set_last_eb_blk,
215 .eo_get_last_eb_blk = ocfs2_xattr_tree_get_last_eb_blk,
216 .eo_update_clusters = ocfs2_xattr_tree_update_clusters,
217 .eo_sanity_check = ocfs2_xattr_tree_sanity_check,
Tao Maba492612008-08-18 17:38:49 +0800218};
219
Joel Beckerdc0ce612008-08-20 16:48:35 -0700220static void ocfs2_get_extent_tree(struct ocfs2_extent_tree *et,
221 struct inode *inode,
222 struct buffer_head *bh,
223 enum ocfs2_extent_tree_type et_type,
Joel Beckerea5efa12008-08-20 16:57:27 -0700224 void *obj)
Tao Mae7d4cb62008-08-18 17:38:44 +0800225{
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700226 et->et_type = et_type;
Tao Mae7d4cb62008-08-18 17:38:44 +0800227 get_bh(bh);
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700228 et->et_root_bh = bh;
Joel Beckerdc0ce612008-08-20 16:48:35 -0700229 et->et_max_leaf_clusters = 0;
Joel Beckerea5efa12008-08-20 16:57:27 -0700230 if (!obj)
231 obj = (void *)bh->b_data;
232 et->et_object = obj;
Tao Mae7d4cb62008-08-18 17:38:44 +0800233
Tao Mae7d4cb62008-08-18 17:38:44 +0800234 if (et_type == OCFS2_DINODE_EXTENT) {
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700235 et->et_root_el =
Joel Beckerea5efa12008-08-20 16:57:27 -0700236 &((struct ocfs2_dinode *)obj)->id2.i_list;
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700237 et->et_ops = &ocfs2_dinode_et_ops;
Tao Maf56654c2008-08-18 17:38:48 +0800238 } else if (et_type == OCFS2_XATTR_VALUE_EXTENT) {
239 struct ocfs2_xattr_value_root *xv =
Joel Beckerea5efa12008-08-20 16:57:27 -0700240 (struct ocfs2_xattr_value_root *)obj;
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700241 et->et_root_el = &xv->xr_list;
242 et->et_ops = &ocfs2_xattr_et_ops;
Tao Maba492612008-08-18 17:38:49 +0800243 } else if (et_type == OCFS2_XATTR_TREE_EXTENT) {
244 struct ocfs2_xattr_block *xb =
Joel Beckerea5efa12008-08-20 16:57:27 -0700245 (struct ocfs2_xattr_block *)obj;
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700246 et->et_root_el = &xb->xb_attrs.xb_root.xt_list;
247 et->et_ops = &ocfs2_xattr_tree_et_ops;
248 et->et_max_leaf_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
Tao Maca12b7c2008-08-18 17:38:52 +0800249 OCFS2_MAX_XATTR_TREE_LEAF_SIZE);
Tao Mae7d4cb62008-08-18 17:38:44 +0800250 }
Tao Mae7d4cb62008-08-18 17:38:44 +0800251}
252
Joel Beckerdc0ce612008-08-20 16:48:35 -0700253static void ocfs2_put_extent_tree(struct ocfs2_extent_tree *et)
Tao Mae7d4cb62008-08-18 17:38:44 +0800254{
Joel Beckerdc0ce612008-08-20 16:48:35 -0700255 brelse(et->et_root_bh);
Tao Mae7d4cb62008-08-18 17:38:44 +0800256}
257
Joel Becker35dc0aa2008-08-20 16:25:06 -0700258static inline void ocfs2_et_set_last_eb_blk(struct ocfs2_extent_tree *et,
259 u64 new_last_eb_blk)
Tao Mae7d4cb62008-08-18 17:38:44 +0800260{
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700261 et->et_ops->eo_set_last_eb_blk(et, new_last_eb_blk);
Tao Mae7d4cb62008-08-18 17:38:44 +0800262}
263
Joel Becker35dc0aa2008-08-20 16:25:06 -0700264static inline u64 ocfs2_et_get_last_eb_blk(struct ocfs2_extent_tree *et)
Tao Mae7d4cb62008-08-18 17:38:44 +0800265{
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700266 return et->et_ops->eo_get_last_eb_blk(et);
Tao Mae7d4cb62008-08-18 17:38:44 +0800267}
268
Joel Becker35dc0aa2008-08-20 16:25:06 -0700269static inline void ocfs2_et_update_clusters(struct inode *inode,
270 struct ocfs2_extent_tree *et,
271 u32 clusters)
Tao Mae7d4cb62008-08-18 17:38:44 +0800272{
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700273 et->et_ops->eo_update_clusters(inode, et, clusters);
Joel Becker35dc0aa2008-08-20 16:25:06 -0700274}
275
276static inline int ocfs2_et_sanity_check(struct inode *inode,
277 struct ocfs2_extent_tree *et)
278{
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700279 return et->et_ops->eo_sanity_check(inode, et);
Tao Mae7d4cb62008-08-18 17:38:44 +0800280}
281
Mark Fashehccd979b2005-12-15 14:31:24 -0800282static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
Mark Fasheh59a5e412007-06-22 15:52:36 -0700283static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
284 struct ocfs2_extent_block *eb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800285
Mark Fashehdcd05382007-01-16 11:32:23 -0800286/*
287 * Structures which describe a path through a btree, and functions to
288 * manipulate them.
289 *
290 * The idea here is to be as generic as possible with the tree
291 * manipulation code.
292 */
293struct ocfs2_path_item {
294 struct buffer_head *bh;
295 struct ocfs2_extent_list *el;
296};
297
298#define OCFS2_MAX_PATH_DEPTH 5
299
300struct ocfs2_path {
301 int p_tree_depth;
302 struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
303};
304
305#define path_root_bh(_path) ((_path)->p_node[0].bh)
306#define path_root_el(_path) ((_path)->p_node[0].el)
307#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
308#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
309#define path_num_items(_path) ((_path)->p_tree_depth + 1)
310
311/*
312 * Reset the actual path elements so that we can re-use the structure
313 * to build another path. Generally, this involves freeing the buffer
314 * heads.
315 */
316static void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root)
317{
318 int i, start = 0, depth = 0;
319 struct ocfs2_path_item *node;
320
321 if (keep_root)
322 start = 1;
323
324 for(i = start; i < path_num_items(path); i++) {
325 node = &path->p_node[i];
326
327 brelse(node->bh);
328 node->bh = NULL;
329 node->el = NULL;
330 }
331
332 /*
333 * Tree depth may change during truncate, or insert. If we're
334 * keeping the root extent list, then make sure that our path
335 * structure reflects the proper depth.
336 */
337 if (keep_root)
338 depth = le16_to_cpu(path_root_el(path)->l_tree_depth);
339
340 path->p_tree_depth = depth;
341}
342
343static void ocfs2_free_path(struct ocfs2_path *path)
344{
345 if (path) {
346 ocfs2_reinit_path(path, 0);
347 kfree(path);
348 }
349}
350
351/*
Mark Fasheh328d5752007-06-18 10:48:04 -0700352 * All the elements of src into dest. After this call, src could be freed
353 * without affecting dest.
354 *
355 * Both paths should have the same root. Any non-root elements of dest
356 * will be freed.
357 */
358static void ocfs2_cp_path(struct ocfs2_path *dest, struct ocfs2_path *src)
359{
360 int i;
361
362 BUG_ON(path_root_bh(dest) != path_root_bh(src));
363 BUG_ON(path_root_el(dest) != path_root_el(src));
364
365 ocfs2_reinit_path(dest, 1);
366
367 for(i = 1; i < OCFS2_MAX_PATH_DEPTH; i++) {
368 dest->p_node[i].bh = src->p_node[i].bh;
369 dest->p_node[i].el = src->p_node[i].el;
370
371 if (dest->p_node[i].bh)
372 get_bh(dest->p_node[i].bh);
373 }
374}
375
376/*
Mark Fashehdcd05382007-01-16 11:32:23 -0800377 * Make the *dest path the same as src and re-initialize src path to
378 * have a root only.
379 */
380static void ocfs2_mv_path(struct ocfs2_path *dest, struct ocfs2_path *src)
381{
382 int i;
383
384 BUG_ON(path_root_bh(dest) != path_root_bh(src));
385
386 for(i = 1; i < OCFS2_MAX_PATH_DEPTH; i++) {
387 brelse(dest->p_node[i].bh);
388
389 dest->p_node[i].bh = src->p_node[i].bh;
390 dest->p_node[i].el = src->p_node[i].el;
391
392 src->p_node[i].bh = NULL;
393 src->p_node[i].el = NULL;
394 }
395}
396
397/*
398 * Insert an extent block at given index.
399 *
400 * This will not take an additional reference on eb_bh.
401 */
402static inline void ocfs2_path_insert_eb(struct ocfs2_path *path, int index,
403 struct buffer_head *eb_bh)
404{
405 struct ocfs2_extent_block *eb = (struct ocfs2_extent_block *)eb_bh->b_data;
406
407 /*
408 * Right now, no root bh is an extent block, so this helps
409 * catch code errors with dinode trees. The assertion can be
410 * safely removed if we ever need to insert extent block
411 * structures at the root.
412 */
413 BUG_ON(index == 0);
414
415 path->p_node[index].bh = eb_bh;
416 path->p_node[index].el = &eb->h_list;
417}
418
419static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh,
420 struct ocfs2_extent_list *root_el)
421{
422 struct ocfs2_path *path;
423
424 BUG_ON(le16_to_cpu(root_el->l_tree_depth) >= OCFS2_MAX_PATH_DEPTH);
425
426 path = kzalloc(sizeof(*path), GFP_NOFS);
427 if (path) {
428 path->p_tree_depth = le16_to_cpu(root_el->l_tree_depth);
429 get_bh(root_bh);
430 path_root_bh(path) = root_bh;
431 path_root_el(path) = root_el;
432 }
433
434 return path;
435}
436
437/*
Mark Fashehdcd05382007-01-16 11:32:23 -0800438 * Convenience function to journal all components in a path.
439 */
440static int ocfs2_journal_access_path(struct inode *inode, handle_t *handle,
441 struct ocfs2_path *path)
442{
443 int i, ret = 0;
444
445 if (!path)
446 goto out;
447
448 for(i = 0; i < path_num_items(path); i++) {
449 ret = ocfs2_journal_access(handle, inode, path->p_node[i].bh,
450 OCFS2_JOURNAL_ACCESS_WRITE);
451 if (ret < 0) {
452 mlog_errno(ret);
453 goto out;
454 }
455 }
456
457out:
458 return ret;
459}
460
Mark Fasheh328d5752007-06-18 10:48:04 -0700461/*
462 * Return the index of the extent record which contains cluster #v_cluster.
463 * -1 is returned if it was not found.
464 *
465 * Should work fine on interior and exterior nodes.
466 */
467int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster)
468{
469 int ret = -1;
470 int i;
471 struct ocfs2_extent_rec *rec;
472 u32 rec_end, rec_start, clusters;
473
474 for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
475 rec = &el->l_recs[i];
476
477 rec_start = le32_to_cpu(rec->e_cpos);
478 clusters = ocfs2_rec_clusters(el, rec);
479
480 rec_end = rec_start + clusters;
481
482 if (v_cluster >= rec_start && v_cluster < rec_end) {
483 ret = i;
484 break;
485 }
486 }
487
488 return ret;
489}
490
Mark Fashehdcd05382007-01-16 11:32:23 -0800491enum ocfs2_contig_type {
492 CONTIG_NONE = 0,
493 CONTIG_LEFT,
Mark Fasheh328d5752007-06-18 10:48:04 -0700494 CONTIG_RIGHT,
495 CONTIG_LEFTRIGHT,
Mark Fashehdcd05382007-01-16 11:32:23 -0800496};
497
Mark Fashehe48edee2007-03-07 16:46:57 -0800498
499/*
500 * NOTE: ocfs2_block_extent_contig(), ocfs2_extents_adjacent() and
501 * ocfs2_extent_contig only work properly against leaf nodes!
502 */
Mark Fashehdcd05382007-01-16 11:32:23 -0800503static int ocfs2_block_extent_contig(struct super_block *sb,
504 struct ocfs2_extent_rec *ext,
505 u64 blkno)
Mark Fashehccd979b2005-12-15 14:31:24 -0800506{
Mark Fashehe48edee2007-03-07 16:46:57 -0800507 u64 blk_end = le64_to_cpu(ext->e_blkno);
508
509 blk_end += ocfs2_clusters_to_blocks(sb,
510 le16_to_cpu(ext->e_leaf_clusters));
511
512 return blkno == blk_end;
Mark Fashehccd979b2005-12-15 14:31:24 -0800513}
514
Mark Fashehdcd05382007-01-16 11:32:23 -0800515static int ocfs2_extents_adjacent(struct ocfs2_extent_rec *left,
516 struct ocfs2_extent_rec *right)
517{
Mark Fashehe48edee2007-03-07 16:46:57 -0800518 u32 left_range;
519
520 left_range = le32_to_cpu(left->e_cpos) +
521 le16_to_cpu(left->e_leaf_clusters);
522
523 return (left_range == le32_to_cpu(right->e_cpos));
Mark Fashehdcd05382007-01-16 11:32:23 -0800524}
525
526static enum ocfs2_contig_type
527 ocfs2_extent_contig(struct inode *inode,
528 struct ocfs2_extent_rec *ext,
529 struct ocfs2_extent_rec *insert_rec)
530{
531 u64 blkno = le64_to_cpu(insert_rec->e_blkno);
532
Mark Fasheh328d5752007-06-18 10:48:04 -0700533 /*
534 * Refuse to coalesce extent records with different flag
535 * fields - we don't want to mix unwritten extents with user
536 * data.
537 */
538 if (ext->e_flags != insert_rec->e_flags)
539 return CONTIG_NONE;
540
Mark Fashehdcd05382007-01-16 11:32:23 -0800541 if (ocfs2_extents_adjacent(ext, insert_rec) &&
542 ocfs2_block_extent_contig(inode->i_sb, ext, blkno))
543 return CONTIG_RIGHT;
544
545 blkno = le64_to_cpu(ext->e_blkno);
546 if (ocfs2_extents_adjacent(insert_rec, ext) &&
547 ocfs2_block_extent_contig(inode->i_sb, insert_rec, blkno))
548 return CONTIG_LEFT;
549
550 return CONTIG_NONE;
551}
552
553/*
554 * NOTE: We can have pretty much any combination of contiguousness and
555 * appending.
556 *
557 * The usefulness of APPEND_TAIL is more in that it lets us know that
558 * we'll have to update the path to that leaf.
559 */
560enum ocfs2_append_type {
561 APPEND_NONE = 0,
562 APPEND_TAIL,
563};
564
Mark Fasheh328d5752007-06-18 10:48:04 -0700565enum ocfs2_split_type {
566 SPLIT_NONE = 0,
567 SPLIT_LEFT,
568 SPLIT_RIGHT,
569};
570
Mark Fashehdcd05382007-01-16 11:32:23 -0800571struct ocfs2_insert_type {
Mark Fasheh328d5752007-06-18 10:48:04 -0700572 enum ocfs2_split_type ins_split;
Mark Fashehdcd05382007-01-16 11:32:23 -0800573 enum ocfs2_append_type ins_appending;
574 enum ocfs2_contig_type ins_contig;
575 int ins_contig_index;
Mark Fashehdcd05382007-01-16 11:32:23 -0800576 int ins_tree_depth;
577};
578
Mark Fasheh328d5752007-06-18 10:48:04 -0700579struct ocfs2_merge_ctxt {
580 enum ocfs2_contig_type c_contig_type;
581 int c_has_empty_extent;
582 int c_split_covers_rec;
Mark Fasheh328d5752007-06-18 10:48:04 -0700583};
584
Mark Fashehccd979b2005-12-15 14:31:24 -0800585/*
586 * How many free extents have we got before we need more meta data?
587 */
588int ocfs2_num_free_extents(struct ocfs2_super *osb,
589 struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +0800590 struct buffer_head *root_bh,
Tao Maf56654c2008-08-18 17:38:48 +0800591 enum ocfs2_extent_tree_type type,
Joel Beckerea5efa12008-08-20 16:57:27 -0700592 void *obj)
Mark Fashehccd979b2005-12-15 14:31:24 -0800593{
594 int retval;
Tao Mae7d4cb62008-08-18 17:38:44 +0800595 struct ocfs2_extent_list *el = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -0800596 struct ocfs2_extent_block *eb;
597 struct buffer_head *eb_bh = NULL;
Tao Mae7d4cb62008-08-18 17:38:44 +0800598 u64 last_eb_blk = 0;
Mark Fashehccd979b2005-12-15 14:31:24 -0800599
600 mlog_entry_void();
601
Tao Mae7d4cb62008-08-18 17:38:44 +0800602 if (type == OCFS2_DINODE_EXTENT) {
603 struct ocfs2_dinode *fe =
604 (struct ocfs2_dinode *)root_bh->b_data;
605 if (!OCFS2_IS_VALID_DINODE(fe)) {
606 OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
607 retval = -EIO;
608 goto bail;
609 }
610
611 if (fe->i_last_eb_blk)
612 last_eb_blk = le64_to_cpu(fe->i_last_eb_blk);
613 el = &fe->id2.i_list;
Tao Maf56654c2008-08-18 17:38:48 +0800614 } else if (type == OCFS2_XATTR_VALUE_EXTENT) {
615 struct ocfs2_xattr_value_root *xv =
Joel Beckerea5efa12008-08-20 16:57:27 -0700616 (struct ocfs2_xattr_value_root *) obj;
Tao Maf56654c2008-08-18 17:38:48 +0800617
618 last_eb_blk = le64_to_cpu(xv->xr_last_eb_blk);
619 el = &xv->xr_list;
Tao Maba492612008-08-18 17:38:49 +0800620 } else if (type == OCFS2_XATTR_TREE_EXTENT) {
621 struct ocfs2_xattr_block *xb =
622 (struct ocfs2_xattr_block *)root_bh->b_data;
623
624 last_eb_blk = le64_to_cpu(xb->xb_attrs.xb_root.xt_last_eb_blk);
625 el = &xb->xb_attrs.xb_root.xt_list;
Mark Fashehccd979b2005-12-15 14:31:24 -0800626 }
627
Tao Mae7d4cb62008-08-18 17:38:44 +0800628 if (last_eb_blk) {
629 retval = ocfs2_read_block(osb, last_eb_blk,
Mark Fashehccd979b2005-12-15 14:31:24 -0800630 &eb_bh, OCFS2_BH_CACHED, inode);
631 if (retval < 0) {
632 mlog_errno(retval);
633 goto bail;
634 }
635 eb = (struct ocfs2_extent_block *) eb_bh->b_data;
636 el = &eb->h_list;
Tao Mae7d4cb62008-08-18 17:38:44 +0800637 }
Mark Fashehccd979b2005-12-15 14:31:24 -0800638
639 BUG_ON(el->l_tree_depth != 0);
640
641 retval = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec);
642bail:
643 if (eb_bh)
644 brelse(eb_bh);
645
646 mlog_exit(retval);
647 return retval;
648}
649
650/* expects array to already be allocated
651 *
652 * sets h_signature, h_blkno, h_suballoc_bit, h_suballoc_slot, and
653 * l_count for you
654 */
655static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
Mark Fasheh1fabe142006-10-09 18:11:45 -0700656 handle_t *handle,
Mark Fashehccd979b2005-12-15 14:31:24 -0800657 struct inode *inode,
658 int wanted,
659 struct ocfs2_alloc_context *meta_ac,
660 struct buffer_head *bhs[])
661{
662 int count, status, i;
663 u16 suballoc_bit_start;
664 u32 num_got;
665 u64 first_blkno;
666 struct ocfs2_extent_block *eb;
667
668 mlog_entry_void();
669
670 count = 0;
671 while (count < wanted) {
672 status = ocfs2_claim_metadata(osb,
673 handle,
674 meta_ac,
675 wanted - count,
676 &suballoc_bit_start,
677 &num_got,
678 &first_blkno);
679 if (status < 0) {
680 mlog_errno(status);
681 goto bail;
682 }
683
684 for(i = count; i < (num_got + count); i++) {
685 bhs[i] = sb_getblk(osb->sb, first_blkno);
686 if (bhs[i] == NULL) {
687 status = -EIO;
688 mlog_errno(status);
689 goto bail;
690 }
691 ocfs2_set_new_buffer_uptodate(inode, bhs[i]);
692
693 status = ocfs2_journal_access(handle, inode, bhs[i],
694 OCFS2_JOURNAL_ACCESS_CREATE);
695 if (status < 0) {
696 mlog_errno(status);
697 goto bail;
698 }
699
700 memset(bhs[i]->b_data, 0, osb->sb->s_blocksize);
701 eb = (struct ocfs2_extent_block *) bhs[i]->b_data;
702 /* Ok, setup the minimal stuff here. */
703 strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE);
704 eb->h_blkno = cpu_to_le64(first_blkno);
705 eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
Mark Fashehccd979b2005-12-15 14:31:24 -0800706 eb->h_suballoc_slot = cpu_to_le16(osb->slot_num);
Mark Fashehccd979b2005-12-15 14:31:24 -0800707 eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
708 eb->h_list.l_count =
709 cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
710
711 suballoc_bit_start++;
712 first_blkno++;
713
714 /* We'll also be dirtied by the caller, so
715 * this isn't absolutely necessary. */
716 status = ocfs2_journal_dirty(handle, bhs[i]);
717 if (status < 0) {
718 mlog_errno(status);
719 goto bail;
720 }
721 }
722
723 count += num_got;
724 }
725
726 status = 0;
727bail:
728 if (status < 0) {
729 for(i = 0; i < wanted; i++) {
730 if (bhs[i])
731 brelse(bhs[i]);
732 bhs[i] = NULL;
733 }
734 }
735 mlog_exit(status);
736 return status;
737}
738
739/*
Mark Fashehdcd05382007-01-16 11:32:23 -0800740 * Helper function for ocfs2_add_branch() and ocfs2_shift_tree_depth().
741 *
742 * Returns the sum of the rightmost extent rec logical offset and
743 * cluster count.
744 *
745 * ocfs2_add_branch() uses this to determine what logical cluster
746 * value should be populated into the leftmost new branch records.
747 *
748 * ocfs2_shift_tree_depth() uses this to determine the # clusters
749 * value for the new topmost tree record.
750 */
751static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list *el)
752{
753 int i;
754
755 i = le16_to_cpu(el->l_next_free_rec) - 1;
756
757 return le32_to_cpu(el->l_recs[i].e_cpos) +
Mark Fashehe48edee2007-03-07 16:46:57 -0800758 ocfs2_rec_clusters(el, &el->l_recs[i]);
Mark Fashehdcd05382007-01-16 11:32:23 -0800759}
760
761/*
Mark Fashehccd979b2005-12-15 14:31:24 -0800762 * Add an entire tree branch to our inode. eb_bh is the extent block
763 * to start at, if we don't want to start the branch at the dinode
764 * structure.
765 *
766 * last_eb_bh is required as we have to update it's next_leaf pointer
767 * for the new last extent block.
768 *
769 * the new branch will be 'empty' in the sense that every block will
Mark Fashehe48edee2007-03-07 16:46:57 -0800770 * contain a single record with cluster count == 0.
Mark Fashehccd979b2005-12-15 14:31:24 -0800771 */
772static int ocfs2_add_branch(struct ocfs2_super *osb,
Mark Fasheh1fabe142006-10-09 18:11:45 -0700773 handle_t *handle,
Mark Fashehccd979b2005-12-15 14:31:24 -0800774 struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +0800775 struct ocfs2_extent_tree *et,
Mark Fashehccd979b2005-12-15 14:31:24 -0800776 struct buffer_head *eb_bh,
Mark Fasheh328d5752007-06-18 10:48:04 -0700777 struct buffer_head **last_eb_bh,
Mark Fashehccd979b2005-12-15 14:31:24 -0800778 struct ocfs2_alloc_context *meta_ac)
779{
780 int status, new_blocks, i;
781 u64 next_blkno, new_last_eb_blk;
782 struct buffer_head *bh;
783 struct buffer_head **new_eb_bhs = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -0800784 struct ocfs2_extent_block *eb;
785 struct ocfs2_extent_list *eb_el;
786 struct ocfs2_extent_list *el;
Mark Fashehdcd05382007-01-16 11:32:23 -0800787 u32 new_cpos;
Mark Fashehccd979b2005-12-15 14:31:24 -0800788
789 mlog_entry_void();
790
Mark Fasheh328d5752007-06-18 10:48:04 -0700791 BUG_ON(!last_eb_bh || !*last_eb_bh);
Mark Fashehccd979b2005-12-15 14:31:24 -0800792
Mark Fashehccd979b2005-12-15 14:31:24 -0800793 if (eb_bh) {
794 eb = (struct ocfs2_extent_block *) eb_bh->b_data;
795 el = &eb->h_list;
796 } else
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700797 el = et->et_root_el;
Mark Fashehccd979b2005-12-15 14:31:24 -0800798
799 /* we never add a branch to a leaf. */
800 BUG_ON(!el->l_tree_depth);
801
802 new_blocks = le16_to_cpu(el->l_tree_depth);
803
804 /* allocate the number of new eb blocks we need */
805 new_eb_bhs = kcalloc(new_blocks, sizeof(struct buffer_head *),
806 GFP_KERNEL);
807 if (!new_eb_bhs) {
808 status = -ENOMEM;
809 mlog_errno(status);
810 goto bail;
811 }
812
813 status = ocfs2_create_new_meta_bhs(osb, handle, inode, new_blocks,
814 meta_ac, new_eb_bhs);
815 if (status < 0) {
816 mlog_errno(status);
817 goto bail;
818 }
819
Mark Fasheh328d5752007-06-18 10:48:04 -0700820 eb = (struct ocfs2_extent_block *)(*last_eb_bh)->b_data;
Mark Fashehdcd05382007-01-16 11:32:23 -0800821 new_cpos = ocfs2_sum_rightmost_rec(&eb->h_list);
822
Mark Fashehccd979b2005-12-15 14:31:24 -0800823 /* Note: new_eb_bhs[new_blocks - 1] is the guy which will be
824 * linked with the rest of the tree.
825 * conversly, new_eb_bhs[0] is the new bottommost leaf.
826 *
827 * when we leave the loop, new_last_eb_blk will point to the
828 * newest leaf, and next_blkno will point to the topmost extent
829 * block. */
830 next_blkno = new_last_eb_blk = 0;
831 for(i = 0; i < new_blocks; i++) {
832 bh = new_eb_bhs[i];
833 eb = (struct ocfs2_extent_block *) bh->b_data;
834 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
835 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
836 status = -EIO;
837 goto bail;
838 }
839 eb_el = &eb->h_list;
840
841 status = ocfs2_journal_access(handle, inode, bh,
842 OCFS2_JOURNAL_ACCESS_CREATE);
843 if (status < 0) {
844 mlog_errno(status);
845 goto bail;
846 }
847
848 eb->h_next_leaf_blk = 0;
849 eb_el->l_tree_depth = cpu_to_le16(i);
850 eb_el->l_next_free_rec = cpu_to_le16(1);
Mark Fashehdcd05382007-01-16 11:32:23 -0800851 /*
852 * This actually counts as an empty extent as
853 * c_clusters == 0
854 */
855 eb_el->l_recs[0].e_cpos = cpu_to_le32(new_cpos);
Mark Fashehccd979b2005-12-15 14:31:24 -0800856 eb_el->l_recs[0].e_blkno = cpu_to_le64(next_blkno);
Mark Fashehe48edee2007-03-07 16:46:57 -0800857 /*
858 * eb_el isn't always an interior node, but even leaf
859 * nodes want a zero'd flags and reserved field so
860 * this gets the whole 32 bits regardless of use.
861 */
862 eb_el->l_recs[0].e_int_clusters = cpu_to_le32(0);
Mark Fashehccd979b2005-12-15 14:31:24 -0800863 if (!eb_el->l_tree_depth)
864 new_last_eb_blk = le64_to_cpu(eb->h_blkno);
865
866 status = ocfs2_journal_dirty(handle, bh);
867 if (status < 0) {
868 mlog_errno(status);
869 goto bail;
870 }
871
872 next_blkno = le64_to_cpu(eb->h_blkno);
873 }
874
875 /* This is a bit hairy. We want to update up to three blocks
876 * here without leaving any of them in an inconsistent state
877 * in case of error. We don't have to worry about
878 * journal_dirty erroring as it won't unless we've aborted the
879 * handle (in which case we would never be here) so reserving
880 * the write with journal_access is all we need to do. */
Mark Fasheh328d5752007-06-18 10:48:04 -0700881 status = ocfs2_journal_access(handle, inode, *last_eb_bh,
Mark Fashehccd979b2005-12-15 14:31:24 -0800882 OCFS2_JOURNAL_ACCESS_WRITE);
883 if (status < 0) {
884 mlog_errno(status);
885 goto bail;
886 }
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700887 status = ocfs2_journal_access(handle, inode, et->et_root_bh,
Mark Fashehccd979b2005-12-15 14:31:24 -0800888 OCFS2_JOURNAL_ACCESS_WRITE);
889 if (status < 0) {
890 mlog_errno(status);
891 goto bail;
892 }
893 if (eb_bh) {
894 status = ocfs2_journal_access(handle, inode, eb_bh,
895 OCFS2_JOURNAL_ACCESS_WRITE);
896 if (status < 0) {
897 mlog_errno(status);
898 goto bail;
899 }
900 }
901
902 /* Link the new branch into the rest of the tree (el will
Tao Mae7d4cb62008-08-18 17:38:44 +0800903 * either be on the root_bh, or the extent block passed in. */
Mark Fashehccd979b2005-12-15 14:31:24 -0800904 i = le16_to_cpu(el->l_next_free_rec);
905 el->l_recs[i].e_blkno = cpu_to_le64(next_blkno);
Mark Fashehdcd05382007-01-16 11:32:23 -0800906 el->l_recs[i].e_cpos = cpu_to_le32(new_cpos);
Mark Fashehe48edee2007-03-07 16:46:57 -0800907 el->l_recs[i].e_int_clusters = 0;
Mark Fashehccd979b2005-12-15 14:31:24 -0800908 le16_add_cpu(&el->l_next_free_rec, 1);
909
910 /* fe needs a new last extent block pointer, as does the
911 * next_leaf on the previously last-extent-block. */
Joel Becker35dc0aa2008-08-20 16:25:06 -0700912 ocfs2_et_set_last_eb_blk(et, new_last_eb_blk);
Mark Fashehccd979b2005-12-15 14:31:24 -0800913
Mark Fasheh328d5752007-06-18 10:48:04 -0700914 eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
Mark Fashehccd979b2005-12-15 14:31:24 -0800915 eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
916
Mark Fasheh328d5752007-06-18 10:48:04 -0700917 status = ocfs2_journal_dirty(handle, *last_eb_bh);
Mark Fashehccd979b2005-12-15 14:31:24 -0800918 if (status < 0)
919 mlog_errno(status);
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700920 status = ocfs2_journal_dirty(handle, et->et_root_bh);
Mark Fashehccd979b2005-12-15 14:31:24 -0800921 if (status < 0)
922 mlog_errno(status);
923 if (eb_bh) {
924 status = ocfs2_journal_dirty(handle, eb_bh);
925 if (status < 0)
926 mlog_errno(status);
927 }
928
Mark Fasheh328d5752007-06-18 10:48:04 -0700929 /*
930 * Some callers want to track the rightmost leaf so pass it
931 * back here.
932 */
933 brelse(*last_eb_bh);
934 get_bh(new_eb_bhs[0]);
935 *last_eb_bh = new_eb_bhs[0];
936
Mark Fashehccd979b2005-12-15 14:31:24 -0800937 status = 0;
938bail:
939 if (new_eb_bhs) {
940 for (i = 0; i < new_blocks; i++)
941 if (new_eb_bhs[i])
942 brelse(new_eb_bhs[i]);
943 kfree(new_eb_bhs);
944 }
945
946 mlog_exit(status);
947 return status;
948}
949
950/*
951 * adds another level to the allocation tree.
952 * returns back the new extent block so you can add a branch to it
953 * after this call.
954 */
955static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
Mark Fasheh1fabe142006-10-09 18:11:45 -0700956 handle_t *handle,
Mark Fashehccd979b2005-12-15 14:31:24 -0800957 struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +0800958 struct ocfs2_extent_tree *et,
Mark Fashehccd979b2005-12-15 14:31:24 -0800959 struct ocfs2_alloc_context *meta_ac,
960 struct buffer_head **ret_new_eb_bh)
961{
962 int status, i;
Mark Fashehdcd05382007-01-16 11:32:23 -0800963 u32 new_clusters;
Mark Fashehccd979b2005-12-15 14:31:24 -0800964 struct buffer_head *new_eb_bh = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -0800965 struct ocfs2_extent_block *eb;
Tao Mae7d4cb62008-08-18 17:38:44 +0800966 struct ocfs2_extent_list *root_el;
Mark Fashehccd979b2005-12-15 14:31:24 -0800967 struct ocfs2_extent_list *eb_el;
968
969 mlog_entry_void();
970
971 status = ocfs2_create_new_meta_bhs(osb, handle, inode, 1, meta_ac,
972 &new_eb_bh);
973 if (status < 0) {
974 mlog_errno(status);
975 goto bail;
976 }
977
978 eb = (struct ocfs2_extent_block *) new_eb_bh->b_data;
979 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
980 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
981 status = -EIO;
982 goto bail;
983 }
984
985 eb_el = &eb->h_list;
Joel Beckerce1d9ea2008-08-20 16:30:07 -0700986 root_el = et->et_root_el;
Mark Fashehccd979b2005-12-15 14:31:24 -0800987
988 status = ocfs2_journal_access(handle, inode, new_eb_bh,
989 OCFS2_JOURNAL_ACCESS_CREATE);
990 if (status < 0) {
991 mlog_errno(status);
992 goto bail;
993 }
994
Tao Mae7d4cb62008-08-18 17:38:44 +0800995 /* copy the root extent list data into the new extent block */
996 eb_el->l_tree_depth = root_el->l_tree_depth;
997 eb_el->l_next_free_rec = root_el->l_next_free_rec;
998 for (i = 0; i < le16_to_cpu(root_el->l_next_free_rec); i++)
999 eb_el->l_recs[i] = root_el->l_recs[i];
Mark Fashehccd979b2005-12-15 14:31:24 -08001000
1001 status = ocfs2_journal_dirty(handle, new_eb_bh);
1002 if (status < 0) {
1003 mlog_errno(status);
1004 goto bail;
1005 }
1006
Joel Beckerce1d9ea2008-08-20 16:30:07 -07001007 status = ocfs2_journal_access(handle, inode, et->et_root_bh,
Mark Fashehccd979b2005-12-15 14:31:24 -08001008 OCFS2_JOURNAL_ACCESS_WRITE);
1009 if (status < 0) {
1010 mlog_errno(status);
1011 goto bail;
1012 }
1013
Mark Fashehdcd05382007-01-16 11:32:23 -08001014 new_clusters = ocfs2_sum_rightmost_rec(eb_el);
1015
Tao Mae7d4cb62008-08-18 17:38:44 +08001016 /* update root_bh now */
1017 le16_add_cpu(&root_el->l_tree_depth, 1);
1018 root_el->l_recs[0].e_cpos = 0;
1019 root_el->l_recs[0].e_blkno = eb->h_blkno;
1020 root_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters);
1021 for (i = 1; i < le16_to_cpu(root_el->l_next_free_rec); i++)
1022 memset(&root_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
1023 root_el->l_next_free_rec = cpu_to_le16(1);
Mark Fashehccd979b2005-12-15 14:31:24 -08001024
1025 /* If this is our 1st tree depth shift, then last_eb_blk
1026 * becomes the allocated extent block */
Tao Mae7d4cb62008-08-18 17:38:44 +08001027 if (root_el->l_tree_depth == cpu_to_le16(1))
Joel Becker35dc0aa2008-08-20 16:25:06 -07001028 ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
Mark Fashehccd979b2005-12-15 14:31:24 -08001029
Joel Beckerce1d9ea2008-08-20 16:30:07 -07001030 status = ocfs2_journal_dirty(handle, et->et_root_bh);
Mark Fashehccd979b2005-12-15 14:31:24 -08001031 if (status < 0) {
1032 mlog_errno(status);
1033 goto bail;
1034 }
1035
1036 *ret_new_eb_bh = new_eb_bh;
1037 new_eb_bh = NULL;
1038 status = 0;
1039bail:
1040 if (new_eb_bh)
1041 brelse(new_eb_bh);
1042
1043 mlog_exit(status);
1044 return status;
1045}
1046
1047/*
Mark Fashehccd979b2005-12-15 14:31:24 -08001048 * Should only be called when there is no space left in any of the
1049 * leaf nodes. What we want to do is find the lowest tree depth
1050 * non-leaf extent block with room for new records. There are three
1051 * valid results of this search:
1052 *
1053 * 1) a lowest extent block is found, then we pass it back in
1054 * *lowest_eb_bh and return '0'
1055 *
Tao Mae7d4cb62008-08-18 17:38:44 +08001056 * 2) the search fails to find anything, but the root_el has room. We
Mark Fashehccd979b2005-12-15 14:31:24 -08001057 * pass NULL back in *lowest_eb_bh, but still return '0'
1058 *
Tao Mae7d4cb62008-08-18 17:38:44 +08001059 * 3) the search fails to find anything AND the root_el is full, in
Mark Fashehccd979b2005-12-15 14:31:24 -08001060 * which case we return > 0
1061 *
1062 * return status < 0 indicates an error.
1063 */
1064static int ocfs2_find_branch_target(struct ocfs2_super *osb,
1065 struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +08001066 struct ocfs2_extent_tree *et,
Mark Fashehccd979b2005-12-15 14:31:24 -08001067 struct buffer_head **target_bh)
1068{
1069 int status = 0, i;
1070 u64 blkno;
Mark Fashehccd979b2005-12-15 14:31:24 -08001071 struct ocfs2_extent_block *eb;
1072 struct ocfs2_extent_list *el;
1073 struct buffer_head *bh = NULL;
1074 struct buffer_head *lowest_bh = NULL;
1075
1076 mlog_entry_void();
1077
1078 *target_bh = NULL;
1079
Joel Beckerce1d9ea2008-08-20 16:30:07 -07001080 el = et->et_root_el;
Mark Fashehccd979b2005-12-15 14:31:24 -08001081
1082 while(le16_to_cpu(el->l_tree_depth) > 1) {
1083 if (le16_to_cpu(el->l_next_free_rec) == 0) {
Mark Fashehb06970532006-03-03 10:24:33 -08001084 ocfs2_error(inode->i_sb, "Dinode %llu has empty "
Mark Fashehccd979b2005-12-15 14:31:24 -08001085 "extent list (next_free_rec == 0)",
Mark Fashehb06970532006-03-03 10:24:33 -08001086 (unsigned long long)OCFS2_I(inode)->ip_blkno);
Mark Fashehccd979b2005-12-15 14:31:24 -08001087 status = -EIO;
1088 goto bail;
1089 }
1090 i = le16_to_cpu(el->l_next_free_rec) - 1;
1091 blkno = le64_to_cpu(el->l_recs[i].e_blkno);
1092 if (!blkno) {
Mark Fashehb06970532006-03-03 10:24:33 -08001093 ocfs2_error(inode->i_sb, "Dinode %llu has extent "
Mark Fashehccd979b2005-12-15 14:31:24 -08001094 "list where extent # %d has no physical "
1095 "block start",
Mark Fashehb06970532006-03-03 10:24:33 -08001096 (unsigned long long)OCFS2_I(inode)->ip_blkno, i);
Mark Fashehccd979b2005-12-15 14:31:24 -08001097 status = -EIO;
1098 goto bail;
1099 }
1100
1101 if (bh) {
1102 brelse(bh);
1103 bh = NULL;
1104 }
1105
1106 status = ocfs2_read_block(osb, blkno, &bh, OCFS2_BH_CACHED,
1107 inode);
1108 if (status < 0) {
1109 mlog_errno(status);
1110 goto bail;
1111 }
1112
1113 eb = (struct ocfs2_extent_block *) bh->b_data;
1114 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
1115 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
1116 status = -EIO;
1117 goto bail;
1118 }
1119 el = &eb->h_list;
1120
1121 if (le16_to_cpu(el->l_next_free_rec) <
1122 le16_to_cpu(el->l_count)) {
1123 if (lowest_bh)
1124 brelse(lowest_bh);
1125 lowest_bh = bh;
1126 get_bh(lowest_bh);
1127 }
1128 }
1129
1130 /* If we didn't find one and the fe doesn't have any room,
1131 * then return '1' */
Joel Beckerce1d9ea2008-08-20 16:30:07 -07001132 el = et->et_root_el;
Tao Mae7d4cb62008-08-18 17:38:44 +08001133 if (!lowest_bh && (el->l_next_free_rec == el->l_count))
Mark Fashehccd979b2005-12-15 14:31:24 -08001134 status = 1;
1135
1136 *target_bh = lowest_bh;
1137bail:
1138 if (bh)
1139 brelse(bh);
1140
1141 mlog_exit(status);
1142 return status;
1143}
1144
Mark Fashehe48edee2007-03-07 16:46:57 -08001145/*
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001146 * Grow a b-tree so that it has more records.
1147 *
1148 * We might shift the tree depth in which case existing paths should
1149 * be considered invalid.
1150 *
1151 * Tree depth after the grow is returned via *final_depth.
Mark Fasheh328d5752007-06-18 10:48:04 -07001152 *
1153 * *last_eb_bh will be updated by ocfs2_add_branch().
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001154 */
1155static int ocfs2_grow_tree(struct inode *inode, handle_t *handle,
Tao Mae7d4cb62008-08-18 17:38:44 +08001156 struct ocfs2_extent_tree *et, int *final_depth,
Mark Fasheh328d5752007-06-18 10:48:04 -07001157 struct buffer_head **last_eb_bh,
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001158 struct ocfs2_alloc_context *meta_ac)
1159{
1160 int ret, shift;
Joel Beckerce1d9ea2008-08-20 16:30:07 -07001161 struct ocfs2_extent_list *el = et->et_root_el;
Tao Mae7d4cb62008-08-18 17:38:44 +08001162 int depth = le16_to_cpu(el->l_tree_depth);
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001163 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1164 struct buffer_head *bh = NULL;
1165
1166 BUG_ON(meta_ac == NULL);
1167
Tao Mae7d4cb62008-08-18 17:38:44 +08001168 shift = ocfs2_find_branch_target(osb, inode, et, &bh);
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001169 if (shift < 0) {
1170 ret = shift;
1171 mlog_errno(ret);
1172 goto out;
1173 }
1174
1175 /* We traveled all the way to the bottom of the allocation tree
1176 * and didn't find room for any more extents - we need to add
1177 * another tree level */
1178 if (shift) {
1179 BUG_ON(bh);
1180 mlog(0, "need to shift tree depth (current = %d)\n", depth);
1181
1182 /* ocfs2_shift_tree_depth will return us a buffer with
1183 * the new extent block (so we can pass that to
1184 * ocfs2_add_branch). */
Tao Mae7d4cb62008-08-18 17:38:44 +08001185 ret = ocfs2_shift_tree_depth(osb, handle, inode, et,
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001186 meta_ac, &bh);
1187 if (ret < 0) {
1188 mlog_errno(ret);
1189 goto out;
1190 }
1191 depth++;
Mark Fasheh328d5752007-06-18 10:48:04 -07001192 if (depth == 1) {
1193 /*
1194 * Special case: we have room now if we shifted from
1195 * tree_depth 0, so no more work needs to be done.
1196 *
1197 * We won't be calling add_branch, so pass
1198 * back *last_eb_bh as the new leaf. At depth
1199 * zero, it should always be null so there's
1200 * no reason to brelse.
1201 */
1202 BUG_ON(*last_eb_bh);
1203 get_bh(bh);
1204 *last_eb_bh = bh;
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001205 goto out;
Mark Fasheh328d5752007-06-18 10:48:04 -07001206 }
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001207 }
1208
1209 /* call ocfs2_add_branch to add the final part of the tree with
1210 * the new data. */
1211 mlog(0, "add branch. bh = %p\n", bh);
Tao Mae7d4cb62008-08-18 17:38:44 +08001212 ret = ocfs2_add_branch(osb, handle, inode, et, bh, last_eb_bh,
Mark Fashehc3afcbb2007-05-29 14:28:51 -07001213 meta_ac);
1214 if (ret < 0) {
1215 mlog_errno(ret);
1216 goto out;
1217 }
1218
1219out:
1220 if (final_depth)
1221 *final_depth = depth;
1222 brelse(bh);
1223 return ret;
1224}
1225
1226/*
Mark Fashehdcd05382007-01-16 11:32:23 -08001227 * This function will discard the rightmost extent record.
1228 */
1229static void ocfs2_shift_records_right(struct ocfs2_extent_list *el)
1230{
1231 int next_free = le16_to_cpu(el->l_next_free_rec);
1232 int count = le16_to_cpu(el->l_count);
1233 unsigned int num_bytes;
1234
1235 BUG_ON(!next_free);
1236 /* This will cause us to go off the end of our extent list. */
1237 BUG_ON(next_free >= count);
1238
1239 num_bytes = sizeof(struct ocfs2_extent_rec) * next_free;
1240
1241 memmove(&el->l_recs[1], &el->l_recs[0], num_bytes);
1242}
1243
1244static void ocfs2_rotate_leaf(struct ocfs2_extent_list *el,
1245 struct ocfs2_extent_rec *insert_rec)
1246{
1247 int i, insert_index, next_free, has_empty, num_bytes;
1248 u32 insert_cpos = le32_to_cpu(insert_rec->e_cpos);
1249 struct ocfs2_extent_rec *rec;
1250
1251 next_free = le16_to_cpu(el->l_next_free_rec);
1252 has_empty = ocfs2_is_empty_extent(&el->l_recs[0]);
1253
1254 BUG_ON(!next_free);
1255
1256 /* The tree code before us didn't allow enough room in the leaf. */
Julia Lawallb1f35502008-03-04 15:21:05 -08001257 BUG_ON(el->l_next_free_rec == el->l_count && !has_empty);
Mark Fashehdcd05382007-01-16 11:32:23 -08001258
1259 /*
1260 * The easiest way to approach this is to just remove the
1261 * empty extent and temporarily decrement next_free.
1262 */
1263 if (has_empty) {
1264 /*
1265 * If next_free was 1 (only an empty extent), this
1266 * loop won't execute, which is fine. We still want
1267 * the decrement above to happen.
1268 */
1269 for(i = 0; i < (next_free - 1); i++)
1270 el->l_recs[i] = el->l_recs[i+1];
1271
1272 next_free--;
1273 }
1274
1275 /*
1276 * Figure out what the new record index should be.
1277 */
1278 for(i = 0; i < next_free; i++) {
1279 rec = &el->l_recs[i];
1280
1281 if (insert_cpos < le32_to_cpu(rec->e_cpos))
1282 break;
1283 }
1284 insert_index = i;
1285
1286 mlog(0, "ins %u: index %d, has_empty %d, next_free %d, count %d\n",
1287 insert_cpos, insert_index, has_empty, next_free, le16_to_cpu(el->l_count));
1288
1289 BUG_ON(insert_index < 0);
1290 BUG_ON(insert_index >= le16_to_cpu(el->l_count));
1291 BUG_ON(insert_index > next_free);
1292
1293 /*
1294 * No need to memmove if we're just adding to the tail.
1295 */
1296 if (insert_index != next_free) {
1297 BUG_ON(next_free >= le16_to_cpu(el->l_count));
1298
1299 num_bytes = next_free - insert_index;
1300 num_bytes *= sizeof(struct ocfs2_extent_rec);
1301 memmove(&el->l_recs[insert_index + 1],
1302 &el->l_recs[insert_index],
1303 num_bytes);
1304 }
1305
1306 /*
1307 * Either we had an empty extent, and need to re-increment or
1308 * there was no empty extent on a non full rightmost leaf node,
1309 * in which case we still need to increment.
1310 */
1311 next_free++;
1312 el->l_next_free_rec = cpu_to_le16(next_free);
1313 /*
1314 * Make sure none of the math above just messed up our tree.
1315 */
1316 BUG_ON(le16_to_cpu(el->l_next_free_rec) > le16_to_cpu(el->l_count));
1317
1318 el->l_recs[insert_index] = *insert_rec;
1319
1320}
1321
Mark Fasheh328d5752007-06-18 10:48:04 -07001322static void ocfs2_remove_empty_extent(struct ocfs2_extent_list *el)
1323{
1324 int size, num_recs = le16_to_cpu(el->l_next_free_rec);
1325
1326 BUG_ON(num_recs == 0);
1327
1328 if (ocfs2_is_empty_extent(&el->l_recs[0])) {
1329 num_recs--;
1330 size = num_recs * sizeof(struct ocfs2_extent_rec);
1331 memmove(&el->l_recs[0], &el->l_recs[1], size);
1332 memset(&el->l_recs[num_recs], 0,
1333 sizeof(struct ocfs2_extent_rec));
1334 el->l_next_free_rec = cpu_to_le16(num_recs);
1335 }
1336}
1337
Mark Fashehdcd05382007-01-16 11:32:23 -08001338/*
1339 * Create an empty extent record .
1340 *
1341 * l_next_free_rec may be updated.
1342 *
1343 * If an empty extent already exists do nothing.
1344 */
1345static void ocfs2_create_empty_extent(struct ocfs2_extent_list *el)
1346{
1347 int next_free = le16_to_cpu(el->l_next_free_rec);
1348
Mark Fashehe48edee2007-03-07 16:46:57 -08001349 BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
1350
Mark Fashehdcd05382007-01-16 11:32:23 -08001351 if (next_free == 0)
1352 goto set_and_inc;
1353
1354 if (ocfs2_is_empty_extent(&el->l_recs[0]))
1355 return;
1356
1357 mlog_bug_on_msg(el->l_count == el->l_next_free_rec,
1358 "Asked to create an empty extent in a full list:\n"
1359 "count = %u, tree depth = %u",
1360 le16_to_cpu(el->l_count),
1361 le16_to_cpu(el->l_tree_depth));
1362
1363 ocfs2_shift_records_right(el);
1364
1365set_and_inc:
1366 le16_add_cpu(&el->l_next_free_rec, 1);
1367 memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
1368}
1369
1370/*
1371 * For a rotation which involves two leaf nodes, the "root node" is
1372 * the lowest level tree node which contains a path to both leafs. This
1373 * resulting set of information can be used to form a complete "subtree"
1374 *
1375 * This function is passed two full paths from the dinode down to a
1376 * pair of adjacent leaves. It's task is to figure out which path
1377 * index contains the subtree root - this can be the root index itself
1378 * in a worst-case rotation.
1379 *
1380 * The array index of the subtree root is passed back.
1381 */
1382static int ocfs2_find_subtree_root(struct inode *inode,
1383 struct ocfs2_path *left,
1384 struct ocfs2_path *right)
1385{
1386 int i = 0;
1387
1388 /*
1389 * Check that the caller passed in two paths from the same tree.
1390 */
1391 BUG_ON(path_root_bh(left) != path_root_bh(right));
1392
1393 do {
1394 i++;
1395
1396 /*
1397 * The caller didn't pass two adjacent paths.
1398 */
1399 mlog_bug_on_msg(i > left->p_tree_depth,
1400 "Inode %lu, left depth %u, right depth %u\n"
1401 "left leaf blk %llu, right leaf blk %llu\n",
1402 inode->i_ino, left->p_tree_depth,
1403 right->p_tree_depth,
1404 (unsigned long long)path_leaf_bh(left)->b_blocknr,
1405 (unsigned long long)path_leaf_bh(right)->b_blocknr);
1406 } while (left->p_node[i].bh->b_blocknr ==
1407 right->p_node[i].bh->b_blocknr);
1408
1409 return i - 1;
1410}
1411
1412typedef void (path_insert_t)(void *, struct buffer_head *);
1413
1414/*
1415 * Traverse a btree path in search of cpos, starting at root_el.
1416 *
1417 * This code can be called with a cpos larger than the tree, in which
1418 * case it will return the rightmost path.
1419 */
1420static int __ocfs2_find_path(struct inode *inode,
1421 struct ocfs2_extent_list *root_el, u32 cpos,
1422 path_insert_t *func, void *data)
1423{
1424 int i, ret = 0;
1425 u32 range;
1426 u64 blkno;
1427 struct buffer_head *bh = NULL;
1428 struct ocfs2_extent_block *eb;
1429 struct ocfs2_extent_list *el;
1430 struct ocfs2_extent_rec *rec;
1431 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1432
1433 el = root_el;
1434 while (el->l_tree_depth) {
1435 if (le16_to_cpu(el->l_next_free_rec) == 0) {
1436 ocfs2_error(inode->i_sb,
1437 "Inode %llu has empty extent list at "
1438 "depth %u\n",
1439 (unsigned long long)oi->ip_blkno,
1440 le16_to_cpu(el->l_tree_depth));
1441 ret = -EROFS;
1442 goto out;
1443
1444 }
1445
1446 for(i = 0; i < le16_to_cpu(el->l_next_free_rec) - 1; i++) {
1447 rec = &el->l_recs[i];
1448
1449 /*
1450 * In the case that cpos is off the allocation
1451 * tree, this should just wind up returning the
1452 * rightmost record.
1453 */
1454 range = le32_to_cpu(rec->e_cpos) +
Mark Fashehe48edee2007-03-07 16:46:57 -08001455 ocfs2_rec_clusters(el, rec);
Mark Fashehdcd05382007-01-16 11:32:23 -08001456 if (cpos >= le32_to_cpu(rec->e_cpos) && cpos < range)
1457 break;
1458 }
1459
1460 blkno = le64_to_cpu(el->l_recs[i].e_blkno);
1461 if (blkno == 0) {
1462 ocfs2_error(inode->i_sb,
1463 "Inode %llu has bad blkno in extent list "
1464 "at depth %u (index %d)\n",
1465 (unsigned long long)oi->ip_blkno,
1466 le16_to_cpu(el->l_tree_depth), i);
1467 ret = -EROFS;
1468 goto out;
1469 }
1470
1471 brelse(bh);
1472 bh = NULL;
1473 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno,
1474 &bh, OCFS2_BH_CACHED, inode);
1475 if (ret) {
1476 mlog_errno(ret);
1477 goto out;
1478 }
1479
1480 eb = (struct ocfs2_extent_block *) bh->b_data;
1481 el = &eb->h_list;
1482 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
1483 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
1484 ret = -EIO;
1485 goto out;
1486 }
1487
1488 if (le16_to_cpu(el->l_next_free_rec) >
1489 le16_to_cpu(el->l_count)) {
1490 ocfs2_error(inode->i_sb,
1491 "Inode %llu has bad count in extent list "
1492 "at block %llu (next free=%u, count=%u)\n",
1493 (unsigned long long)oi->ip_blkno,
1494 (unsigned long long)bh->b_blocknr,
1495 le16_to_cpu(el->l_next_free_rec),
1496 le16_to_cpu(el->l_count));
1497 ret = -EROFS;
1498 goto out;
1499 }
1500
1501 if (func)
1502 func(data, bh);
1503 }
1504
1505out:
1506 /*
1507 * Catch any trailing bh that the loop didn't handle.
1508 */
1509 brelse(bh);
1510
1511 return ret;
1512}
1513
1514/*
1515 * Given an initialized path (that is, it has a valid root extent
1516 * list), this function will traverse the btree in search of the path
1517 * which would contain cpos.
1518 *
1519 * The path traveled is recorded in the path structure.
1520 *
1521 * Note that this will not do any comparisons on leaf node extent
1522 * records, so it will work fine in the case that we just added a tree
1523 * branch.
1524 */
1525struct find_path_data {
1526 int index;
1527 struct ocfs2_path *path;
1528};
1529static void find_path_ins(void *data, struct buffer_head *bh)
1530{
1531 struct find_path_data *fp = data;
1532
1533 get_bh(bh);
1534 ocfs2_path_insert_eb(fp->path, fp->index, bh);
1535 fp->index++;
1536}
1537static int ocfs2_find_path(struct inode *inode, struct ocfs2_path *path,
1538 u32 cpos)
1539{
1540 struct find_path_data data;
1541
1542 data.index = 1;
1543 data.path = path;
1544 return __ocfs2_find_path(inode, path_root_el(path), cpos,
1545 find_path_ins, &data);
1546}
1547
1548static void find_leaf_ins(void *data, struct buffer_head *bh)
1549{
1550 struct ocfs2_extent_block *eb =(struct ocfs2_extent_block *)bh->b_data;
1551 struct ocfs2_extent_list *el = &eb->h_list;
1552 struct buffer_head **ret = data;
1553
1554 /* We want to retain only the leaf block. */
1555 if (le16_to_cpu(el->l_tree_depth) == 0) {
1556 get_bh(bh);
1557 *ret = bh;
1558 }
1559}
1560/*
1561 * Find the leaf block in the tree which would contain cpos. No
1562 * checking of the actual leaf is done.
1563 *
1564 * Some paths want to call this instead of allocating a path structure
1565 * and calling ocfs2_find_path().
1566 *
1567 * This function doesn't handle non btree extent lists.
1568 */
Mark Fasheh363041a2007-01-17 12:31:35 -08001569int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el,
1570 u32 cpos, struct buffer_head **leaf_bh)
Mark Fashehdcd05382007-01-16 11:32:23 -08001571{
1572 int ret;
1573 struct buffer_head *bh = NULL;
1574
1575 ret = __ocfs2_find_path(inode, root_el, cpos, find_leaf_ins, &bh);
1576 if (ret) {
1577 mlog_errno(ret);
1578 goto out;
1579 }
1580
1581 *leaf_bh = bh;
1582out:
1583 return ret;
1584}
1585
1586/*
1587 * Adjust the adjacent records (left_rec, right_rec) involved in a rotation.
1588 *
1589 * Basically, we've moved stuff around at the bottom of the tree and
1590 * we need to fix up the extent records above the changes to reflect
1591 * the new changes.
1592 *
1593 * left_rec: the record on the left.
1594 * left_child_el: is the child list pointed to by left_rec
1595 * right_rec: the record to the right of left_rec
1596 * right_child_el: is the child list pointed to by right_rec
1597 *
1598 * By definition, this only works on interior nodes.
1599 */
1600static void ocfs2_adjust_adjacent_records(struct ocfs2_extent_rec *left_rec,
1601 struct ocfs2_extent_list *left_child_el,
1602 struct ocfs2_extent_rec *right_rec,
1603 struct ocfs2_extent_list *right_child_el)
1604{
1605 u32 left_clusters, right_end;
1606
1607 /*
1608 * Interior nodes never have holes. Their cpos is the cpos of
1609 * the leftmost record in their child list. Their cluster
1610 * count covers the full theoretical range of their child list
1611 * - the range between their cpos and the cpos of the record
1612 * immediately to their right.
1613 */
1614 left_clusters = le32_to_cpu(right_child_el->l_recs[0].e_cpos);
Mark Fasheh328d5752007-06-18 10:48:04 -07001615 if (ocfs2_is_empty_extent(&right_child_el->l_recs[0])) {
1616 BUG_ON(le16_to_cpu(right_child_el->l_next_free_rec) <= 1);
1617 left_clusters = le32_to_cpu(right_child_el->l_recs[1].e_cpos);
1618 }
Mark Fashehdcd05382007-01-16 11:32:23 -08001619 left_clusters -= le32_to_cpu(left_rec->e_cpos);
Mark Fashehe48edee2007-03-07 16:46:57 -08001620 left_rec->e_int_clusters = cpu_to_le32(left_clusters);
Mark Fashehdcd05382007-01-16 11:32:23 -08001621
1622 /*
1623 * Calculate the rightmost cluster count boundary before
Mark Fashehe48edee2007-03-07 16:46:57 -08001624 * moving cpos - we will need to adjust clusters after
Mark Fashehdcd05382007-01-16 11:32:23 -08001625 * updating e_cpos to keep the same highest cluster count.
1626 */
1627 right_end = le32_to_cpu(right_rec->e_cpos);
Mark Fashehe48edee2007-03-07 16:46:57 -08001628 right_end += le32_to_cpu(right_rec->e_int_clusters);
Mark Fashehdcd05382007-01-16 11:32:23 -08001629
1630 right_rec->e_cpos = left_rec->e_cpos;
1631 le32_add_cpu(&right_rec->e_cpos, left_clusters);
1632
1633 right_end -= le32_to_cpu(right_rec->e_cpos);
Mark Fashehe48edee2007-03-07 16:46:57 -08001634 right_rec->e_int_clusters = cpu_to_le32(right_end);
Mark Fashehdcd05382007-01-16 11:32:23 -08001635}
1636
1637/*
1638 * Adjust the adjacent root node records involved in a
1639 * rotation. left_el_blkno is passed in as a key so that we can easily
1640 * find it's index in the root list.
1641 */
1642static void ocfs2_adjust_root_records(struct ocfs2_extent_list *root_el,
1643 struct ocfs2_extent_list *left_el,
1644 struct ocfs2_extent_list *right_el,
1645 u64 left_el_blkno)
1646{
1647 int i;
1648
1649 BUG_ON(le16_to_cpu(root_el->l_tree_depth) <=
1650 le16_to_cpu(left_el->l_tree_depth));
1651
1652 for(i = 0; i < le16_to_cpu(root_el->l_next_free_rec) - 1; i++) {
1653 if (le64_to_cpu(root_el->l_recs[i].e_blkno) == left_el_blkno)
1654 break;
1655 }
1656
1657 /*
1658 * The path walking code should have never returned a root and
1659 * two paths which are not adjacent.
1660 */
1661 BUG_ON(i >= (le16_to_cpu(root_el->l_next_free_rec) - 1));
1662
1663 ocfs2_adjust_adjacent_records(&root_el->l_recs[i], left_el,
1664 &root_el->l_recs[i + 1], right_el);
1665}
1666
1667/*
1668 * We've changed a leaf block (in right_path) and need to reflect that
1669 * change back up the subtree.
1670 *
1671 * This happens in multiple places:
1672 * - When we've moved an extent record from the left path leaf to the right
1673 * path leaf to make room for an empty extent in the left path leaf.
1674 * - When our insert into the right path leaf is at the leftmost edge
1675 * and requires an update of the path immediately to it's left. This
1676 * can occur at the end of some types of rotation and appending inserts.
Tao Ma677b9752008-01-30 14:21:05 +08001677 * - When we've adjusted the last extent record in the left path leaf and the
1678 * 1st extent record in the right path leaf during cross extent block merge.
Mark Fashehdcd05382007-01-16 11:32:23 -08001679 */
1680static void ocfs2_complete_edge_insert(struct inode *inode, handle_t *handle,
1681 struct ocfs2_path *left_path,
1682 struct ocfs2_path *right_path,
1683 int subtree_index)
1684{
1685 int ret, i, idx;
1686 struct ocfs2_extent_list *el, *left_el, *right_el;
1687 struct ocfs2_extent_rec *left_rec, *right_rec;
1688 struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
1689
1690 /*
1691 * Update the counts and position values within all the
1692 * interior nodes to reflect the leaf rotation we just did.
1693 *
1694 * The root node is handled below the loop.
1695 *
1696 * We begin the loop with right_el and left_el pointing to the
1697 * leaf lists and work our way up.
1698 *
1699 * NOTE: within this loop, left_el and right_el always refer
1700 * to the *child* lists.
1701 */
1702 left_el = path_leaf_el(left_path);
1703 right_el = path_leaf_el(right_path);
1704 for(i = left_path->p_tree_depth - 1; i > subtree_index; i--) {
1705 mlog(0, "Adjust records at index %u\n", i);
1706
1707 /*
1708 * One nice property of knowing that all of these
1709 * nodes are below the root is that we only deal with
1710 * the leftmost right node record and the rightmost
1711 * left node record.
1712 */
1713 el = left_path->p_node[i].el;
1714 idx = le16_to_cpu(left_el->l_next_free_rec) - 1;
1715 left_rec = &el->l_recs[idx];
1716
1717 el = right_path->p_node[i].el;
1718 right_rec = &el->l_recs[0];
1719
1720 ocfs2_adjust_adjacent_records(left_rec, left_el, right_rec,
1721 right_el);
1722
1723 ret = ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
1724 if (ret)
1725 mlog_errno(ret);
1726
1727 ret = ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
1728 if (ret)
1729 mlog_errno(ret);
1730
1731 /*
1732 * Setup our list pointers now so that the current
1733 * parents become children in the next iteration.
1734 */
1735 left_el = left_path->p_node[i].el;
1736 right_el = right_path->p_node[i].el;
1737 }
1738
1739 /*
1740 * At the root node, adjust the two adjacent records which
1741 * begin our path to the leaves.
1742 */
1743
1744 el = left_path->p_node[subtree_index].el;
1745 left_el = left_path->p_node[subtree_index + 1].el;
1746 right_el = right_path->p_node[subtree_index + 1].el;
1747
1748 ocfs2_adjust_root_records(el, left_el, right_el,
1749 left_path->p_node[subtree_index + 1].bh->b_blocknr);
1750
1751 root_bh = left_path->p_node[subtree_index].bh;
1752
1753 ret = ocfs2_journal_dirty(handle, root_bh);
1754 if (ret)
1755 mlog_errno(ret);
1756}
1757
1758static int ocfs2_rotate_subtree_right(struct inode *inode,
1759 handle_t *handle,
1760 struct ocfs2_path *left_path,
1761 struct ocfs2_path *right_path,
1762 int subtree_index)
1763{
1764 int ret, i;
1765 struct buffer_head *right_leaf_bh;
1766 struct buffer_head *left_leaf_bh = NULL;
1767 struct buffer_head *root_bh;
1768 struct ocfs2_extent_list *right_el, *left_el;
1769 struct ocfs2_extent_rec move_rec;
1770
1771 left_leaf_bh = path_leaf_bh(left_path);
1772 left_el = path_leaf_el(left_path);
1773
1774 if (left_el->l_next_free_rec != left_el->l_count) {
1775 ocfs2_error(inode->i_sb,
1776 "Inode %llu has non-full interior leaf node %llu"
1777 "(next free = %u)",
1778 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1779 (unsigned long long)left_leaf_bh->b_blocknr,
1780 le16_to_cpu(left_el->l_next_free_rec));
1781 return -EROFS;
1782 }
1783
1784 /*
1785 * This extent block may already have an empty record, so we
1786 * return early if so.
1787 */
1788 if (ocfs2_is_empty_extent(&left_el->l_recs[0]))
1789 return 0;
1790
1791 root_bh = left_path->p_node[subtree_index].bh;
1792 BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
1793
1794 ret = ocfs2_journal_access(handle, inode, root_bh,
1795 OCFS2_JOURNAL_ACCESS_WRITE);
1796 if (ret) {
1797 mlog_errno(ret);
1798 goto out;
1799 }
1800
1801 for(i = subtree_index + 1; i < path_num_items(right_path); i++) {
1802 ret = ocfs2_journal_access(handle, inode,
1803 right_path->p_node[i].bh,
1804 OCFS2_JOURNAL_ACCESS_WRITE);
1805 if (ret) {
1806 mlog_errno(ret);
1807 goto out;
1808 }
1809
1810 ret = ocfs2_journal_access(handle, inode,
1811 left_path->p_node[i].bh,
1812 OCFS2_JOURNAL_ACCESS_WRITE);
1813 if (ret) {
1814 mlog_errno(ret);
1815 goto out;
1816 }
1817 }
1818
1819 right_leaf_bh = path_leaf_bh(right_path);
1820 right_el = path_leaf_el(right_path);
1821
1822 /* This is a code error, not a disk corruption. */
1823 mlog_bug_on_msg(!right_el->l_next_free_rec, "Inode %llu: Rotate fails "
1824 "because rightmost leaf block %llu is empty\n",
1825 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1826 (unsigned long long)right_leaf_bh->b_blocknr);
1827
1828 ocfs2_create_empty_extent(right_el);
1829
1830 ret = ocfs2_journal_dirty(handle, right_leaf_bh);
1831 if (ret) {
1832 mlog_errno(ret);
1833 goto out;
1834 }
1835
1836 /* Do the copy now. */
1837 i = le16_to_cpu(left_el->l_next_free_rec) - 1;
1838 move_rec = left_el->l_recs[i];
1839 right_el->l_recs[0] = move_rec;
1840
1841 /*
1842 * Clear out the record we just copied and shift everything
1843 * over, leaving an empty extent in the left leaf.
1844 *
1845 * We temporarily subtract from next_free_rec so that the
1846 * shift will lose the tail record (which is now defunct).
1847 */
1848 le16_add_cpu(&left_el->l_next_free_rec, -1);
1849 ocfs2_shift_records_right(left_el);
1850 memset(&left_el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
1851 le16_add_cpu(&left_el->l_next_free_rec, 1);
1852
1853 ret = ocfs2_journal_dirty(handle, left_leaf_bh);
1854 if (ret) {
1855 mlog_errno(ret);
1856 goto out;
1857 }
1858
1859 ocfs2_complete_edge_insert(inode, handle, left_path, right_path,
1860 subtree_index);
1861
1862out:
1863 return ret;
1864}
1865
1866/*
1867 * Given a full path, determine what cpos value would return us a path
1868 * containing the leaf immediately to the left of the current one.
1869 *
1870 * Will return zero if the path passed in is already the leftmost path.
1871 */
1872static int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
1873 struct ocfs2_path *path, u32 *cpos)
1874{
1875 int i, j, ret = 0;
1876 u64 blkno;
1877 struct ocfs2_extent_list *el;
1878
Mark Fashehe48edee2007-03-07 16:46:57 -08001879 BUG_ON(path->p_tree_depth == 0);
1880
Mark Fashehdcd05382007-01-16 11:32:23 -08001881 *cpos = 0;
1882
1883 blkno = path_leaf_bh(path)->b_blocknr;
1884
1885 /* Start at the tree node just above the leaf and work our way up. */
1886 i = path->p_tree_depth - 1;
1887 while (i >= 0) {
1888 el = path->p_node[i].el;
1889
1890 /*
1891 * Find the extent record just before the one in our
1892 * path.
1893 */
1894 for(j = 0; j < le16_to_cpu(el->l_next_free_rec); j++) {
1895 if (le64_to_cpu(el->l_recs[j].e_blkno) == blkno) {
1896 if (j == 0) {
1897 if (i == 0) {
1898 /*
1899 * We've determined that the
1900 * path specified is already
1901 * the leftmost one - return a
1902 * cpos of zero.
1903 */
1904 goto out;
1905 }
1906 /*
1907 * The leftmost record points to our
1908 * leaf - we need to travel up the
1909 * tree one level.
1910 */
1911 goto next_node;
1912 }
1913
1914 *cpos = le32_to_cpu(el->l_recs[j - 1].e_cpos);
Mark Fashehe48edee2007-03-07 16:46:57 -08001915 *cpos = *cpos + ocfs2_rec_clusters(el,
1916 &el->l_recs[j - 1]);
1917 *cpos = *cpos - 1;
Mark Fashehdcd05382007-01-16 11:32:23 -08001918 goto out;
1919 }
1920 }
1921
1922 /*
1923 * If we got here, we never found a valid node where
1924 * the tree indicated one should be.
1925 */
1926 ocfs2_error(sb,
1927 "Invalid extent tree at extent block %llu\n",
1928 (unsigned long long)blkno);
1929 ret = -EROFS;
1930 goto out;
1931
1932next_node:
1933 blkno = path->p_node[i].bh->b_blocknr;
1934 i--;
1935 }
1936
1937out:
1938 return ret;
1939}
1940
Mark Fasheh328d5752007-06-18 10:48:04 -07001941/*
1942 * Extend the transaction by enough credits to complete the rotation,
1943 * and still leave at least the original number of credits allocated
1944 * to this transaction.
1945 */
Mark Fashehdcd05382007-01-16 11:32:23 -08001946static int ocfs2_extend_rotate_transaction(handle_t *handle, int subtree_depth,
Mark Fasheh328d5752007-06-18 10:48:04 -07001947 int op_credits,
Mark Fashehdcd05382007-01-16 11:32:23 -08001948 struct ocfs2_path *path)
1949{
Mark Fasheh328d5752007-06-18 10:48:04 -07001950 int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits;
Mark Fashehdcd05382007-01-16 11:32:23 -08001951
1952 if (handle->h_buffer_credits < credits)
1953 return ocfs2_extend_trans(handle, credits);
1954
1955 return 0;
1956}
1957
1958/*
1959 * Trap the case where we're inserting into the theoretical range past
1960 * the _actual_ left leaf range. Otherwise, we'll rotate a record
1961 * whose cpos is less than ours into the right leaf.
1962 *
1963 * It's only necessary to look at the rightmost record of the left
1964 * leaf because the logic that calls us should ensure that the
1965 * theoretical ranges in the path components above the leaves are
1966 * correct.
1967 */
1968static int ocfs2_rotate_requires_path_adjustment(struct ocfs2_path *left_path,
1969 u32 insert_cpos)
1970{
1971 struct ocfs2_extent_list *left_el;
1972 struct ocfs2_extent_rec *rec;
1973 int next_free;
1974
1975 left_el = path_leaf_el(left_path);
1976 next_free = le16_to_cpu(left_el->l_next_free_rec);
1977 rec = &left_el->l_recs[next_free - 1];
1978
1979 if (insert_cpos > le32_to_cpu(rec->e_cpos))
1980 return 1;
1981 return 0;
1982}
1983
Mark Fasheh328d5752007-06-18 10:48:04 -07001984static int ocfs2_leftmost_rec_contains(struct ocfs2_extent_list *el, u32 cpos)
1985{
1986 int next_free = le16_to_cpu(el->l_next_free_rec);
1987 unsigned int range;
1988 struct ocfs2_extent_rec *rec;
1989
1990 if (next_free == 0)
1991 return 0;
1992
1993 rec = &el->l_recs[0];
1994 if (ocfs2_is_empty_extent(rec)) {
1995 /* Empty list. */
1996 if (next_free == 1)
1997 return 0;
1998 rec = &el->l_recs[1];
1999 }
2000
2001 range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
2002 if (cpos >= le32_to_cpu(rec->e_cpos) && cpos < range)
2003 return 1;
2004 return 0;
2005}
2006
Mark Fashehdcd05382007-01-16 11:32:23 -08002007/*
2008 * Rotate all the records in a btree right one record, starting at insert_cpos.
2009 *
2010 * The path to the rightmost leaf should be passed in.
2011 *
2012 * The array is assumed to be large enough to hold an entire path (tree depth).
2013 *
2014 * Upon succesful return from this function:
2015 *
2016 * - The 'right_path' array will contain a path to the leaf block
2017 * whose range contains e_cpos.
2018 * - That leaf block will have a single empty extent in list index 0.
2019 * - In the case that the rotation requires a post-insert update,
2020 * *ret_left_path will contain a valid path which can be passed to
2021 * ocfs2_insert_path().
2022 */
2023static int ocfs2_rotate_tree_right(struct inode *inode,
2024 handle_t *handle,
Mark Fasheh328d5752007-06-18 10:48:04 -07002025 enum ocfs2_split_type split,
Mark Fashehdcd05382007-01-16 11:32:23 -08002026 u32 insert_cpos,
2027 struct ocfs2_path *right_path,
2028 struct ocfs2_path **ret_left_path)
2029{
Mark Fasheh328d5752007-06-18 10:48:04 -07002030 int ret, start, orig_credits = handle->h_buffer_credits;
Mark Fashehdcd05382007-01-16 11:32:23 -08002031 u32 cpos;
2032 struct ocfs2_path *left_path = NULL;
2033
2034 *ret_left_path = NULL;
2035
2036 left_path = ocfs2_new_path(path_root_bh(right_path),
2037 path_root_el(right_path));
2038 if (!left_path) {
2039 ret = -ENOMEM;
2040 mlog_errno(ret);
2041 goto out;
2042 }
2043
2044 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path, &cpos);
2045 if (ret) {
2046 mlog_errno(ret);
2047 goto out;
2048 }
2049
2050 mlog(0, "Insert: %u, first left path cpos: %u\n", insert_cpos, cpos);
2051
2052 /*
2053 * What we want to do here is:
2054 *
2055 * 1) Start with the rightmost path.
2056 *
2057 * 2) Determine a path to the leaf block directly to the left
2058 * of that leaf.
2059 *
2060 * 3) Determine the 'subtree root' - the lowest level tree node
2061 * which contains a path to both leaves.
2062 *
2063 * 4) Rotate the subtree.
2064 *
2065 * 5) Find the next subtree by considering the left path to be
2066 * the new right path.
2067 *
2068 * The check at the top of this while loop also accepts
2069 * insert_cpos == cpos because cpos is only a _theoretical_
2070 * value to get us the left path - insert_cpos might very well
2071 * be filling that hole.
2072 *
2073 * Stop at a cpos of '0' because we either started at the
2074 * leftmost branch (i.e., a tree with one branch and a
2075 * rotation inside of it), or we've gone as far as we can in
2076 * rotating subtrees.
2077 */
2078 while (cpos && insert_cpos <= cpos) {
2079 mlog(0, "Rotating a tree: ins. cpos: %u, left path cpos: %u\n",
2080 insert_cpos, cpos);
2081
2082 ret = ocfs2_find_path(inode, left_path, cpos);
2083 if (ret) {
2084 mlog_errno(ret);
2085 goto out;
2086 }
2087
2088 mlog_bug_on_msg(path_leaf_bh(left_path) ==
2089 path_leaf_bh(right_path),
2090 "Inode %lu: error during insert of %u "
2091 "(left path cpos %u) results in two identical "
2092 "paths ending at %llu\n",
2093 inode->i_ino, insert_cpos, cpos,
2094 (unsigned long long)
2095 path_leaf_bh(left_path)->b_blocknr);
2096
Mark Fasheh328d5752007-06-18 10:48:04 -07002097 if (split == SPLIT_NONE &&
2098 ocfs2_rotate_requires_path_adjustment(left_path,
Mark Fashehdcd05382007-01-16 11:32:23 -08002099 insert_cpos)) {
Mark Fashehdcd05382007-01-16 11:32:23 -08002100
2101 /*
2102 * We've rotated the tree as much as we
2103 * should. The rest is up to
2104 * ocfs2_insert_path() to complete, after the
2105 * record insertion. We indicate this
2106 * situation by returning the left path.
2107 *
2108 * The reason we don't adjust the records here
2109 * before the record insert is that an error
2110 * later might break the rule where a parent
2111 * record e_cpos will reflect the actual
2112 * e_cpos of the 1st nonempty record of the
2113 * child list.
2114 */
2115 *ret_left_path = left_path;
2116 goto out_ret_path;
2117 }
2118
2119 start = ocfs2_find_subtree_root(inode, left_path, right_path);
2120
2121 mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n",
2122 start,
2123 (unsigned long long) right_path->p_node[start].bh->b_blocknr,
2124 right_path->p_tree_depth);
2125
2126 ret = ocfs2_extend_rotate_transaction(handle, start,
Mark Fasheh328d5752007-06-18 10:48:04 -07002127 orig_credits, right_path);
Mark Fashehdcd05382007-01-16 11:32:23 -08002128 if (ret) {
2129 mlog_errno(ret);
2130 goto out;
2131 }
2132
2133 ret = ocfs2_rotate_subtree_right(inode, handle, left_path,
2134 right_path, start);
2135 if (ret) {
2136 mlog_errno(ret);
2137 goto out;
2138 }
2139
Mark Fasheh328d5752007-06-18 10:48:04 -07002140 if (split != SPLIT_NONE &&
2141 ocfs2_leftmost_rec_contains(path_leaf_el(right_path),
2142 insert_cpos)) {
2143 /*
2144 * A rotate moves the rightmost left leaf
2145 * record over to the leftmost right leaf
2146 * slot. If we're doing an extent split
2147 * instead of a real insert, then we have to
2148 * check that the extent to be split wasn't
2149 * just moved over. If it was, then we can
2150 * exit here, passing left_path back -
2151 * ocfs2_split_extent() is smart enough to
2152 * search both leaves.
2153 */
2154 *ret_left_path = left_path;
2155 goto out_ret_path;
2156 }
2157
Mark Fashehdcd05382007-01-16 11:32:23 -08002158 /*
2159 * There is no need to re-read the next right path
2160 * as we know that it'll be our current left
2161 * path. Optimize by copying values instead.
2162 */
2163 ocfs2_mv_path(right_path, left_path);
2164
2165 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path,
2166 &cpos);
2167 if (ret) {
2168 mlog_errno(ret);
2169 goto out;
2170 }
2171 }
2172
2173out:
2174 ocfs2_free_path(left_path);
2175
2176out_ret_path:
2177 return ret;
2178}
2179
Mark Fasheh328d5752007-06-18 10:48:04 -07002180static void ocfs2_update_edge_lengths(struct inode *inode, handle_t *handle,
2181 struct ocfs2_path *path)
2182{
2183 int i, idx;
2184 struct ocfs2_extent_rec *rec;
2185 struct ocfs2_extent_list *el;
2186 struct ocfs2_extent_block *eb;
2187 u32 range;
2188
2189 /* Path should always be rightmost. */
2190 eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
2191 BUG_ON(eb->h_next_leaf_blk != 0ULL);
2192
2193 el = &eb->h_list;
2194 BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
2195 idx = le16_to_cpu(el->l_next_free_rec) - 1;
2196 rec = &el->l_recs[idx];
2197 range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
2198
2199 for (i = 0; i < path->p_tree_depth; i++) {
2200 el = path->p_node[i].el;
2201 idx = le16_to_cpu(el->l_next_free_rec) - 1;
2202 rec = &el->l_recs[idx];
2203
2204 rec->e_int_clusters = cpu_to_le32(range);
2205 le32_add_cpu(&rec->e_int_clusters, -le32_to_cpu(rec->e_cpos));
2206
2207 ocfs2_journal_dirty(handle, path->p_node[i].bh);
2208 }
2209}
2210
2211static void ocfs2_unlink_path(struct inode *inode, handle_t *handle,
2212 struct ocfs2_cached_dealloc_ctxt *dealloc,
2213 struct ocfs2_path *path, int unlink_start)
2214{
2215 int ret, i;
2216 struct ocfs2_extent_block *eb;
2217 struct ocfs2_extent_list *el;
2218 struct buffer_head *bh;
2219
2220 for(i = unlink_start; i < path_num_items(path); i++) {
2221 bh = path->p_node[i].bh;
2222
2223 eb = (struct ocfs2_extent_block *)bh->b_data;
2224 /*
2225 * Not all nodes might have had their final count
2226 * decremented by the caller - handle this here.
2227 */
2228 el = &eb->h_list;
2229 if (le16_to_cpu(el->l_next_free_rec) > 1) {
2230 mlog(ML_ERROR,
2231 "Inode %llu, attempted to remove extent block "
2232 "%llu with %u records\n",
2233 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2234 (unsigned long long)le64_to_cpu(eb->h_blkno),
2235 le16_to_cpu(el->l_next_free_rec));
2236
2237 ocfs2_journal_dirty(handle, bh);
2238 ocfs2_remove_from_cache(inode, bh);
2239 continue;
2240 }
2241
2242 el->l_next_free_rec = 0;
2243 memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2244
2245 ocfs2_journal_dirty(handle, bh);
2246
2247 ret = ocfs2_cache_extent_block_free(dealloc, eb);
2248 if (ret)
2249 mlog_errno(ret);
2250
2251 ocfs2_remove_from_cache(inode, bh);
2252 }
2253}
2254
2255static void ocfs2_unlink_subtree(struct inode *inode, handle_t *handle,
2256 struct ocfs2_path *left_path,
2257 struct ocfs2_path *right_path,
2258 int subtree_index,
2259 struct ocfs2_cached_dealloc_ctxt *dealloc)
2260{
2261 int i;
2262 struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
2263 struct ocfs2_extent_list *root_el = left_path->p_node[subtree_index].el;
2264 struct ocfs2_extent_list *el;
2265 struct ocfs2_extent_block *eb;
2266
2267 el = path_leaf_el(left_path);
2268
2269 eb = (struct ocfs2_extent_block *)right_path->p_node[subtree_index + 1].bh->b_data;
2270
2271 for(i = 1; i < le16_to_cpu(root_el->l_next_free_rec); i++)
2272 if (root_el->l_recs[i].e_blkno == eb->h_blkno)
2273 break;
2274
2275 BUG_ON(i >= le16_to_cpu(root_el->l_next_free_rec));
2276
2277 memset(&root_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec));
2278 le16_add_cpu(&root_el->l_next_free_rec, -1);
2279
2280 eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
2281 eb->h_next_leaf_blk = 0;
2282
2283 ocfs2_journal_dirty(handle, root_bh);
2284 ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
2285
2286 ocfs2_unlink_path(inode, handle, dealloc, right_path,
2287 subtree_index + 1);
2288}
2289
2290static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle,
2291 struct ocfs2_path *left_path,
2292 struct ocfs2_path *right_path,
2293 int subtree_index,
2294 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08002295 int *deleted,
2296 struct ocfs2_extent_tree *et)
Mark Fasheh328d5752007-06-18 10:48:04 -07002297{
2298 int ret, i, del_right_subtree = 0, right_has_empty = 0;
Tao Mae7d4cb62008-08-18 17:38:44 +08002299 struct buffer_head *root_bh, *et_root_bh = path_root_bh(right_path);
Mark Fasheh328d5752007-06-18 10:48:04 -07002300 struct ocfs2_extent_list *right_leaf_el, *left_leaf_el;
2301 struct ocfs2_extent_block *eb;
2302
2303 *deleted = 0;
2304
2305 right_leaf_el = path_leaf_el(right_path);
2306 left_leaf_el = path_leaf_el(left_path);
2307 root_bh = left_path->p_node[subtree_index].bh;
2308 BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
2309
2310 if (!ocfs2_is_empty_extent(&left_leaf_el->l_recs[0]))
2311 return 0;
2312
2313 eb = (struct ocfs2_extent_block *)path_leaf_bh(right_path)->b_data;
2314 if (ocfs2_is_empty_extent(&right_leaf_el->l_recs[0])) {
2315 /*
2316 * It's legal for us to proceed if the right leaf is
2317 * the rightmost one and it has an empty extent. There
2318 * are two cases to handle - whether the leaf will be
2319 * empty after removal or not. If the leaf isn't empty
2320 * then just remove the empty extent up front. The
2321 * next block will handle empty leaves by flagging
2322 * them for unlink.
2323 *
2324 * Non rightmost leaves will throw -EAGAIN and the
2325 * caller can manually move the subtree and retry.
2326 */
2327
2328 if (eb->h_next_leaf_blk != 0ULL)
2329 return -EAGAIN;
2330
2331 if (le16_to_cpu(right_leaf_el->l_next_free_rec) > 1) {
2332 ret = ocfs2_journal_access(handle, inode,
2333 path_leaf_bh(right_path),
2334 OCFS2_JOURNAL_ACCESS_WRITE);
2335 if (ret) {
2336 mlog_errno(ret);
2337 goto out;
2338 }
2339
2340 ocfs2_remove_empty_extent(right_leaf_el);
2341 } else
2342 right_has_empty = 1;
2343 }
2344
2345 if (eb->h_next_leaf_blk == 0ULL &&
2346 le16_to_cpu(right_leaf_el->l_next_free_rec) == 1) {
2347 /*
2348 * We have to update i_last_eb_blk during the meta
2349 * data delete.
2350 */
Tao Mae7d4cb62008-08-18 17:38:44 +08002351 ret = ocfs2_journal_access(handle, inode, et_root_bh,
Mark Fasheh328d5752007-06-18 10:48:04 -07002352 OCFS2_JOURNAL_ACCESS_WRITE);
2353 if (ret) {
2354 mlog_errno(ret);
2355 goto out;
2356 }
2357
2358 del_right_subtree = 1;
2359 }
2360
2361 /*
2362 * Getting here with an empty extent in the right path implies
2363 * that it's the rightmost path and will be deleted.
2364 */
2365 BUG_ON(right_has_empty && !del_right_subtree);
2366
2367 ret = ocfs2_journal_access(handle, inode, root_bh,
2368 OCFS2_JOURNAL_ACCESS_WRITE);
2369 if (ret) {
2370 mlog_errno(ret);
2371 goto out;
2372 }
2373
2374 for(i = subtree_index + 1; i < path_num_items(right_path); i++) {
2375 ret = ocfs2_journal_access(handle, inode,
2376 right_path->p_node[i].bh,
2377 OCFS2_JOURNAL_ACCESS_WRITE);
2378 if (ret) {
2379 mlog_errno(ret);
2380 goto out;
2381 }
2382
2383 ret = ocfs2_journal_access(handle, inode,
2384 left_path->p_node[i].bh,
2385 OCFS2_JOURNAL_ACCESS_WRITE);
2386 if (ret) {
2387 mlog_errno(ret);
2388 goto out;
2389 }
2390 }
2391
2392 if (!right_has_empty) {
2393 /*
2394 * Only do this if we're moving a real
2395 * record. Otherwise, the action is delayed until
2396 * after removal of the right path in which case we
2397 * can do a simple shift to remove the empty extent.
2398 */
2399 ocfs2_rotate_leaf(left_leaf_el, &right_leaf_el->l_recs[0]);
2400 memset(&right_leaf_el->l_recs[0], 0,
2401 sizeof(struct ocfs2_extent_rec));
2402 }
2403 if (eb->h_next_leaf_blk == 0ULL) {
2404 /*
2405 * Move recs over to get rid of empty extent, decrease
2406 * next_free. This is allowed to remove the last
2407 * extent in our leaf (setting l_next_free_rec to
2408 * zero) - the delete code below won't care.
2409 */
2410 ocfs2_remove_empty_extent(right_leaf_el);
2411 }
2412
2413 ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
2414 if (ret)
2415 mlog_errno(ret);
2416 ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
2417 if (ret)
2418 mlog_errno(ret);
2419
2420 if (del_right_subtree) {
2421 ocfs2_unlink_subtree(inode, handle, left_path, right_path,
2422 subtree_index, dealloc);
2423 ocfs2_update_edge_lengths(inode, handle, left_path);
2424
2425 eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
Joel Becker35dc0aa2008-08-20 16:25:06 -07002426 ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
Mark Fasheh328d5752007-06-18 10:48:04 -07002427
2428 /*
2429 * Removal of the extent in the left leaf was skipped
2430 * above so we could delete the right path
2431 * 1st.
2432 */
2433 if (right_has_empty)
2434 ocfs2_remove_empty_extent(left_leaf_el);
2435
Tao Mae7d4cb62008-08-18 17:38:44 +08002436 ret = ocfs2_journal_dirty(handle, et_root_bh);
Mark Fasheh328d5752007-06-18 10:48:04 -07002437 if (ret)
2438 mlog_errno(ret);
2439
2440 *deleted = 1;
2441 } else
2442 ocfs2_complete_edge_insert(inode, handle, left_path, right_path,
2443 subtree_index);
2444
2445out:
2446 return ret;
2447}
2448
2449/*
2450 * Given a full path, determine what cpos value would return us a path
2451 * containing the leaf immediately to the right of the current one.
2452 *
2453 * Will return zero if the path passed in is already the rightmost path.
2454 *
2455 * This looks similar, but is subtly different to
2456 * ocfs2_find_cpos_for_left_leaf().
2457 */
2458static int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
2459 struct ocfs2_path *path, u32 *cpos)
2460{
2461 int i, j, ret = 0;
2462 u64 blkno;
2463 struct ocfs2_extent_list *el;
2464
2465 *cpos = 0;
2466
2467 if (path->p_tree_depth == 0)
2468 return 0;
2469
2470 blkno = path_leaf_bh(path)->b_blocknr;
2471
2472 /* Start at the tree node just above the leaf and work our way up. */
2473 i = path->p_tree_depth - 1;
2474 while (i >= 0) {
2475 int next_free;
2476
2477 el = path->p_node[i].el;
2478
2479 /*
2480 * Find the extent record just after the one in our
2481 * path.
2482 */
2483 next_free = le16_to_cpu(el->l_next_free_rec);
2484 for(j = 0; j < le16_to_cpu(el->l_next_free_rec); j++) {
2485 if (le64_to_cpu(el->l_recs[j].e_blkno) == blkno) {
2486 if (j == (next_free - 1)) {
2487 if (i == 0) {
2488 /*
2489 * We've determined that the
2490 * path specified is already
2491 * the rightmost one - return a
2492 * cpos of zero.
2493 */
2494 goto out;
2495 }
2496 /*
2497 * The rightmost record points to our
2498 * leaf - we need to travel up the
2499 * tree one level.
2500 */
2501 goto next_node;
2502 }
2503
2504 *cpos = le32_to_cpu(el->l_recs[j + 1].e_cpos);
2505 goto out;
2506 }
2507 }
2508
2509 /*
2510 * If we got here, we never found a valid node where
2511 * the tree indicated one should be.
2512 */
2513 ocfs2_error(sb,
2514 "Invalid extent tree at extent block %llu\n",
2515 (unsigned long long)blkno);
2516 ret = -EROFS;
2517 goto out;
2518
2519next_node:
2520 blkno = path->p_node[i].bh->b_blocknr;
2521 i--;
2522 }
2523
2524out:
2525 return ret;
2526}
2527
2528static int ocfs2_rotate_rightmost_leaf_left(struct inode *inode,
2529 handle_t *handle,
2530 struct buffer_head *bh,
2531 struct ocfs2_extent_list *el)
2532{
2533 int ret;
2534
2535 if (!ocfs2_is_empty_extent(&el->l_recs[0]))
2536 return 0;
2537
2538 ret = ocfs2_journal_access(handle, inode, bh,
2539 OCFS2_JOURNAL_ACCESS_WRITE);
2540 if (ret) {
2541 mlog_errno(ret);
2542 goto out;
2543 }
2544
2545 ocfs2_remove_empty_extent(el);
2546
2547 ret = ocfs2_journal_dirty(handle, bh);
2548 if (ret)
2549 mlog_errno(ret);
2550
2551out:
2552 return ret;
2553}
2554
2555static int __ocfs2_rotate_tree_left(struct inode *inode,
2556 handle_t *handle, int orig_credits,
2557 struct ocfs2_path *path,
2558 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08002559 struct ocfs2_path **empty_extent_path,
2560 struct ocfs2_extent_tree *et)
Mark Fasheh328d5752007-06-18 10:48:04 -07002561{
2562 int ret, subtree_root, deleted;
2563 u32 right_cpos;
2564 struct ocfs2_path *left_path = NULL;
2565 struct ocfs2_path *right_path = NULL;
2566
2567 BUG_ON(!ocfs2_is_empty_extent(&(path_leaf_el(path)->l_recs[0])));
2568
2569 *empty_extent_path = NULL;
2570
2571 ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, path,
2572 &right_cpos);
2573 if (ret) {
2574 mlog_errno(ret);
2575 goto out;
2576 }
2577
2578 left_path = ocfs2_new_path(path_root_bh(path),
2579 path_root_el(path));
2580 if (!left_path) {
2581 ret = -ENOMEM;
2582 mlog_errno(ret);
2583 goto out;
2584 }
2585
2586 ocfs2_cp_path(left_path, path);
2587
2588 right_path = ocfs2_new_path(path_root_bh(path),
2589 path_root_el(path));
2590 if (!right_path) {
2591 ret = -ENOMEM;
2592 mlog_errno(ret);
2593 goto out;
2594 }
2595
2596 while (right_cpos) {
2597 ret = ocfs2_find_path(inode, right_path, right_cpos);
2598 if (ret) {
2599 mlog_errno(ret);
2600 goto out;
2601 }
2602
2603 subtree_root = ocfs2_find_subtree_root(inode, left_path,
2604 right_path);
2605
2606 mlog(0, "Subtree root at index %d (blk %llu, depth %d)\n",
2607 subtree_root,
2608 (unsigned long long)
2609 right_path->p_node[subtree_root].bh->b_blocknr,
2610 right_path->p_tree_depth);
2611
2612 ret = ocfs2_extend_rotate_transaction(handle, subtree_root,
2613 orig_credits, left_path);
2614 if (ret) {
2615 mlog_errno(ret);
2616 goto out;
2617 }
2618
Mark Fashehe8aed342007-12-03 16:43:01 -08002619 /*
2620 * Caller might still want to make changes to the
2621 * tree root, so re-add it to the journal here.
2622 */
2623 ret = ocfs2_journal_access(handle, inode,
2624 path_root_bh(left_path),
2625 OCFS2_JOURNAL_ACCESS_WRITE);
2626 if (ret) {
2627 mlog_errno(ret);
2628 goto out;
2629 }
2630
Mark Fasheh328d5752007-06-18 10:48:04 -07002631 ret = ocfs2_rotate_subtree_left(inode, handle, left_path,
2632 right_path, subtree_root,
Tao Mae7d4cb62008-08-18 17:38:44 +08002633 dealloc, &deleted, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07002634 if (ret == -EAGAIN) {
2635 /*
2636 * The rotation has to temporarily stop due to
2637 * the right subtree having an empty
2638 * extent. Pass it back to the caller for a
2639 * fixup.
2640 */
2641 *empty_extent_path = right_path;
2642 right_path = NULL;
2643 goto out;
2644 }
2645 if (ret) {
2646 mlog_errno(ret);
2647 goto out;
2648 }
2649
2650 /*
2651 * The subtree rotate might have removed records on
2652 * the rightmost edge. If so, then rotation is
2653 * complete.
2654 */
2655 if (deleted)
2656 break;
2657
2658 ocfs2_mv_path(left_path, right_path);
2659
2660 ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, left_path,
2661 &right_cpos);
2662 if (ret) {
2663 mlog_errno(ret);
2664 goto out;
2665 }
2666 }
2667
2668out:
2669 ocfs2_free_path(right_path);
2670 ocfs2_free_path(left_path);
2671
2672 return ret;
2673}
2674
2675static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle,
Tao Mae7d4cb62008-08-18 17:38:44 +08002676 struct ocfs2_path *path,
2677 struct ocfs2_cached_dealloc_ctxt *dealloc,
2678 struct ocfs2_extent_tree *et)
Mark Fasheh328d5752007-06-18 10:48:04 -07002679{
2680 int ret, subtree_index;
2681 u32 cpos;
2682 struct ocfs2_path *left_path = NULL;
Mark Fasheh328d5752007-06-18 10:48:04 -07002683 struct ocfs2_extent_block *eb;
2684 struct ocfs2_extent_list *el;
2685
Mark Fasheh328d5752007-06-18 10:48:04 -07002686
Joel Becker35dc0aa2008-08-20 16:25:06 -07002687 ret = ocfs2_et_sanity_check(inode, et);
Tao Mae7d4cb62008-08-18 17:38:44 +08002688 if (ret)
2689 goto out;
Mark Fasheh328d5752007-06-18 10:48:04 -07002690 /*
2691 * There's two ways we handle this depending on
2692 * whether path is the only existing one.
2693 */
2694 ret = ocfs2_extend_rotate_transaction(handle, 0,
2695 handle->h_buffer_credits,
2696 path);
2697 if (ret) {
2698 mlog_errno(ret);
2699 goto out;
2700 }
2701
2702 ret = ocfs2_journal_access_path(inode, handle, path);
2703 if (ret) {
2704 mlog_errno(ret);
2705 goto out;
2706 }
2707
2708 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
2709 if (ret) {
2710 mlog_errno(ret);
2711 goto out;
2712 }
2713
2714 if (cpos) {
2715 /*
2716 * We have a path to the left of this one - it needs
2717 * an update too.
2718 */
2719 left_path = ocfs2_new_path(path_root_bh(path),
2720 path_root_el(path));
2721 if (!left_path) {
2722 ret = -ENOMEM;
2723 mlog_errno(ret);
2724 goto out;
2725 }
2726
2727 ret = ocfs2_find_path(inode, left_path, cpos);
2728 if (ret) {
2729 mlog_errno(ret);
2730 goto out;
2731 }
2732
2733 ret = ocfs2_journal_access_path(inode, handle, left_path);
2734 if (ret) {
2735 mlog_errno(ret);
2736 goto out;
2737 }
2738
2739 subtree_index = ocfs2_find_subtree_root(inode, left_path, path);
2740
2741 ocfs2_unlink_subtree(inode, handle, left_path, path,
2742 subtree_index, dealloc);
2743 ocfs2_update_edge_lengths(inode, handle, left_path);
2744
2745 eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data;
Joel Becker35dc0aa2008-08-20 16:25:06 -07002746 ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
Mark Fasheh328d5752007-06-18 10:48:04 -07002747 } else {
2748 /*
2749 * 'path' is also the leftmost path which
2750 * means it must be the only one. This gets
2751 * handled differently because we want to
2752 * revert the inode back to having extents
2753 * in-line.
2754 */
2755 ocfs2_unlink_path(inode, handle, dealloc, path, 1);
2756
Joel Beckerce1d9ea2008-08-20 16:30:07 -07002757 el = et->et_root_el;
Mark Fasheh328d5752007-06-18 10:48:04 -07002758 el->l_tree_depth = 0;
2759 el->l_next_free_rec = 0;
2760 memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2761
Joel Becker35dc0aa2008-08-20 16:25:06 -07002762 ocfs2_et_set_last_eb_blk(et, 0);
Mark Fasheh328d5752007-06-18 10:48:04 -07002763 }
2764
2765 ocfs2_journal_dirty(handle, path_root_bh(path));
2766
2767out:
2768 ocfs2_free_path(left_path);
2769 return ret;
2770}
2771
2772/*
2773 * Left rotation of btree records.
2774 *
2775 * In many ways, this is (unsurprisingly) the opposite of right
2776 * rotation. We start at some non-rightmost path containing an empty
2777 * extent in the leaf block. The code works its way to the rightmost
2778 * path by rotating records to the left in every subtree.
2779 *
2780 * This is used by any code which reduces the number of extent records
2781 * in a leaf. After removal, an empty record should be placed in the
2782 * leftmost list position.
2783 *
2784 * This won't handle a length update of the rightmost path records if
2785 * the rightmost tree leaf record is removed so the caller is
2786 * responsible for detecting and correcting that.
2787 */
2788static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle,
2789 struct ocfs2_path *path,
Tao Mae7d4cb62008-08-18 17:38:44 +08002790 struct ocfs2_cached_dealloc_ctxt *dealloc,
2791 struct ocfs2_extent_tree *et)
Mark Fasheh328d5752007-06-18 10:48:04 -07002792{
2793 int ret, orig_credits = handle->h_buffer_credits;
2794 struct ocfs2_path *tmp_path = NULL, *restart_path = NULL;
2795 struct ocfs2_extent_block *eb;
2796 struct ocfs2_extent_list *el;
2797
2798 el = path_leaf_el(path);
2799 if (!ocfs2_is_empty_extent(&el->l_recs[0]))
2800 return 0;
2801
2802 if (path->p_tree_depth == 0) {
2803rightmost_no_delete:
2804 /*
Tao Mae7d4cb62008-08-18 17:38:44 +08002805 * Inline extents. This is trivially handled, so do
Mark Fasheh328d5752007-06-18 10:48:04 -07002806 * it up front.
2807 */
2808 ret = ocfs2_rotate_rightmost_leaf_left(inode, handle,
2809 path_leaf_bh(path),
2810 path_leaf_el(path));
2811 if (ret)
2812 mlog_errno(ret);
2813 goto out;
2814 }
2815
2816 /*
2817 * Handle rightmost branch now. There's several cases:
2818 * 1) simple rotation leaving records in there. That's trivial.
2819 * 2) rotation requiring a branch delete - there's no more
2820 * records left. Two cases of this:
2821 * a) There are branches to the left.
2822 * b) This is also the leftmost (the only) branch.
2823 *
2824 * 1) is handled via ocfs2_rotate_rightmost_leaf_left()
2825 * 2a) we need the left branch so that we can update it with the unlink
2826 * 2b) we need to bring the inode back to inline extents.
2827 */
2828
2829 eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
2830 el = &eb->h_list;
2831 if (eb->h_next_leaf_blk == 0) {
2832 /*
2833 * This gets a bit tricky if we're going to delete the
2834 * rightmost path. Get the other cases out of the way
2835 * 1st.
2836 */
2837 if (le16_to_cpu(el->l_next_free_rec) > 1)
2838 goto rightmost_no_delete;
2839
2840 if (le16_to_cpu(el->l_next_free_rec) == 0) {
2841 ret = -EIO;
2842 ocfs2_error(inode->i_sb,
2843 "Inode %llu has empty extent block at %llu",
2844 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2845 (unsigned long long)le64_to_cpu(eb->h_blkno));
2846 goto out;
2847 }
2848
2849 /*
2850 * XXX: The caller can not trust "path" any more after
2851 * this as it will have been deleted. What do we do?
2852 *
2853 * In theory the rotate-for-merge code will never get
2854 * here because it'll always ask for a rotate in a
2855 * nonempty list.
2856 */
2857
2858 ret = ocfs2_remove_rightmost_path(inode, handle, path,
Tao Mae7d4cb62008-08-18 17:38:44 +08002859 dealloc, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07002860 if (ret)
2861 mlog_errno(ret);
2862 goto out;
2863 }
2864
2865 /*
2866 * Now we can loop, remembering the path we get from -EAGAIN
2867 * and restarting from there.
2868 */
2869try_rotate:
2870 ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, path,
Tao Mae7d4cb62008-08-18 17:38:44 +08002871 dealloc, &restart_path, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07002872 if (ret && ret != -EAGAIN) {
2873 mlog_errno(ret);
2874 goto out;
2875 }
2876
2877 while (ret == -EAGAIN) {
2878 tmp_path = restart_path;
2879 restart_path = NULL;
2880
2881 ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits,
2882 tmp_path, dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08002883 &restart_path, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07002884 if (ret && ret != -EAGAIN) {
2885 mlog_errno(ret);
2886 goto out;
2887 }
2888
2889 ocfs2_free_path(tmp_path);
2890 tmp_path = NULL;
2891
2892 if (ret == 0)
2893 goto try_rotate;
2894 }
2895
2896out:
2897 ocfs2_free_path(tmp_path);
2898 ocfs2_free_path(restart_path);
2899 return ret;
2900}
2901
2902static void ocfs2_cleanup_merge(struct ocfs2_extent_list *el,
2903 int index)
2904{
2905 struct ocfs2_extent_rec *rec = &el->l_recs[index];
2906 unsigned int size;
2907
2908 if (rec->e_leaf_clusters == 0) {
2909 /*
2910 * We consumed all of the merged-from record. An empty
2911 * extent cannot exist anywhere but the 1st array
2912 * position, so move things over if the merged-from
2913 * record doesn't occupy that position.
2914 *
2915 * This creates a new empty extent so the caller
2916 * should be smart enough to have removed any existing
2917 * ones.
2918 */
2919 if (index > 0) {
2920 BUG_ON(ocfs2_is_empty_extent(&el->l_recs[0]));
2921 size = index * sizeof(struct ocfs2_extent_rec);
2922 memmove(&el->l_recs[1], &el->l_recs[0], size);
2923 }
2924
2925 /*
2926 * Always memset - the caller doesn't check whether it
2927 * created an empty extent, so there could be junk in
2928 * the other fields.
2929 */
2930 memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
2931 }
2932}
2933
Tao Ma677b9752008-01-30 14:21:05 +08002934static int ocfs2_get_right_path(struct inode *inode,
2935 struct ocfs2_path *left_path,
2936 struct ocfs2_path **ret_right_path)
Mark Fasheh328d5752007-06-18 10:48:04 -07002937{
2938 int ret;
Tao Ma677b9752008-01-30 14:21:05 +08002939 u32 right_cpos;
2940 struct ocfs2_path *right_path = NULL;
2941 struct ocfs2_extent_list *left_el;
2942
2943 *ret_right_path = NULL;
2944
2945 /* This function shouldn't be called for non-trees. */
2946 BUG_ON(left_path->p_tree_depth == 0);
2947
2948 left_el = path_leaf_el(left_path);
2949 BUG_ON(left_el->l_next_free_rec != left_el->l_count);
2950
2951 ret = ocfs2_find_cpos_for_right_leaf(inode->i_sb, left_path,
2952 &right_cpos);
2953 if (ret) {
2954 mlog_errno(ret);
2955 goto out;
2956 }
2957
2958 /* This function shouldn't be called for the rightmost leaf. */
2959 BUG_ON(right_cpos == 0);
2960
2961 right_path = ocfs2_new_path(path_root_bh(left_path),
2962 path_root_el(left_path));
2963 if (!right_path) {
2964 ret = -ENOMEM;
2965 mlog_errno(ret);
2966 goto out;
2967 }
2968
2969 ret = ocfs2_find_path(inode, right_path, right_cpos);
2970 if (ret) {
2971 mlog_errno(ret);
2972 goto out;
2973 }
2974
2975 *ret_right_path = right_path;
2976out:
2977 if (ret)
2978 ocfs2_free_path(right_path);
2979 return ret;
2980}
2981
2982/*
2983 * Remove split_rec clusters from the record at index and merge them
2984 * onto the beginning of the record "next" to it.
2985 * For index < l_count - 1, the next means the extent rec at index + 1.
2986 * For index == l_count - 1, the "next" means the 1st extent rec of the
2987 * next extent block.
2988 */
2989static int ocfs2_merge_rec_right(struct inode *inode,
2990 struct ocfs2_path *left_path,
2991 handle_t *handle,
2992 struct ocfs2_extent_rec *split_rec,
2993 int index)
2994{
2995 int ret, next_free, i;
Mark Fasheh328d5752007-06-18 10:48:04 -07002996 unsigned int split_clusters = le16_to_cpu(split_rec->e_leaf_clusters);
2997 struct ocfs2_extent_rec *left_rec;
2998 struct ocfs2_extent_rec *right_rec;
Tao Ma677b9752008-01-30 14:21:05 +08002999 struct ocfs2_extent_list *right_el;
3000 struct ocfs2_path *right_path = NULL;
3001 int subtree_index = 0;
3002 struct ocfs2_extent_list *el = path_leaf_el(left_path);
3003 struct buffer_head *bh = path_leaf_bh(left_path);
3004 struct buffer_head *root_bh = NULL;
Mark Fasheh328d5752007-06-18 10:48:04 -07003005
3006 BUG_ON(index >= le16_to_cpu(el->l_next_free_rec));
Mark Fasheh328d5752007-06-18 10:48:04 -07003007 left_rec = &el->l_recs[index];
Tao Ma677b9752008-01-30 14:21:05 +08003008
Al Viro9d8df6a2008-05-21 06:32:11 +01003009 if (index == le16_to_cpu(el->l_next_free_rec) - 1 &&
Tao Ma677b9752008-01-30 14:21:05 +08003010 le16_to_cpu(el->l_next_free_rec) == le16_to_cpu(el->l_count)) {
3011 /* we meet with a cross extent block merge. */
3012 ret = ocfs2_get_right_path(inode, left_path, &right_path);
3013 if (ret) {
3014 mlog_errno(ret);
3015 goto out;
3016 }
3017
3018 right_el = path_leaf_el(right_path);
3019 next_free = le16_to_cpu(right_el->l_next_free_rec);
3020 BUG_ON(next_free <= 0);
3021 right_rec = &right_el->l_recs[0];
3022 if (ocfs2_is_empty_extent(right_rec)) {
Al Viro9d8df6a2008-05-21 06:32:11 +01003023 BUG_ON(next_free <= 1);
Tao Ma677b9752008-01-30 14:21:05 +08003024 right_rec = &right_el->l_recs[1];
3025 }
3026
3027 BUG_ON(le32_to_cpu(left_rec->e_cpos) +
3028 le16_to_cpu(left_rec->e_leaf_clusters) !=
3029 le32_to_cpu(right_rec->e_cpos));
3030
3031 subtree_index = ocfs2_find_subtree_root(inode,
3032 left_path, right_path);
3033
3034 ret = ocfs2_extend_rotate_transaction(handle, subtree_index,
3035 handle->h_buffer_credits,
3036 right_path);
3037 if (ret) {
3038 mlog_errno(ret);
3039 goto out;
3040 }
3041
3042 root_bh = left_path->p_node[subtree_index].bh;
3043 BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
3044
3045 ret = ocfs2_journal_access(handle, inode, root_bh,
3046 OCFS2_JOURNAL_ACCESS_WRITE);
3047 if (ret) {
3048 mlog_errno(ret);
3049 goto out;
3050 }
3051
3052 for (i = subtree_index + 1;
3053 i < path_num_items(right_path); i++) {
3054 ret = ocfs2_journal_access(handle, inode,
3055 right_path->p_node[i].bh,
3056 OCFS2_JOURNAL_ACCESS_WRITE);
3057 if (ret) {
3058 mlog_errno(ret);
3059 goto out;
3060 }
3061
3062 ret = ocfs2_journal_access(handle, inode,
3063 left_path->p_node[i].bh,
3064 OCFS2_JOURNAL_ACCESS_WRITE);
3065 if (ret) {
3066 mlog_errno(ret);
3067 goto out;
3068 }
3069 }
3070
3071 } else {
3072 BUG_ON(index == le16_to_cpu(el->l_next_free_rec) - 1);
3073 right_rec = &el->l_recs[index + 1];
3074 }
Mark Fasheh328d5752007-06-18 10:48:04 -07003075
3076 ret = ocfs2_journal_access(handle, inode, bh,
3077 OCFS2_JOURNAL_ACCESS_WRITE);
3078 if (ret) {
3079 mlog_errno(ret);
3080 goto out;
3081 }
3082
3083 le16_add_cpu(&left_rec->e_leaf_clusters, -split_clusters);
3084
3085 le32_add_cpu(&right_rec->e_cpos, -split_clusters);
3086 le64_add_cpu(&right_rec->e_blkno,
3087 -ocfs2_clusters_to_blocks(inode->i_sb, split_clusters));
3088 le16_add_cpu(&right_rec->e_leaf_clusters, split_clusters);
3089
3090 ocfs2_cleanup_merge(el, index);
3091
3092 ret = ocfs2_journal_dirty(handle, bh);
3093 if (ret)
3094 mlog_errno(ret);
3095
Tao Ma677b9752008-01-30 14:21:05 +08003096 if (right_path) {
3097 ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
3098 if (ret)
3099 mlog_errno(ret);
3100
3101 ocfs2_complete_edge_insert(inode, handle, left_path,
3102 right_path, subtree_index);
3103 }
Mark Fasheh328d5752007-06-18 10:48:04 -07003104out:
Tao Ma677b9752008-01-30 14:21:05 +08003105 if (right_path)
3106 ocfs2_free_path(right_path);
3107 return ret;
3108}
3109
3110static int ocfs2_get_left_path(struct inode *inode,
3111 struct ocfs2_path *right_path,
3112 struct ocfs2_path **ret_left_path)
3113{
3114 int ret;
3115 u32 left_cpos;
3116 struct ocfs2_path *left_path = NULL;
3117
3118 *ret_left_path = NULL;
3119
3120 /* This function shouldn't be called for non-trees. */
3121 BUG_ON(right_path->p_tree_depth == 0);
3122
3123 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
3124 right_path, &left_cpos);
3125 if (ret) {
3126 mlog_errno(ret);
3127 goto out;
3128 }
3129
3130 /* This function shouldn't be called for the leftmost leaf. */
3131 BUG_ON(left_cpos == 0);
3132
3133 left_path = ocfs2_new_path(path_root_bh(right_path),
3134 path_root_el(right_path));
3135 if (!left_path) {
3136 ret = -ENOMEM;
3137 mlog_errno(ret);
3138 goto out;
3139 }
3140
3141 ret = ocfs2_find_path(inode, left_path, left_cpos);
3142 if (ret) {
3143 mlog_errno(ret);
3144 goto out;
3145 }
3146
3147 *ret_left_path = left_path;
3148out:
3149 if (ret)
3150 ocfs2_free_path(left_path);
Mark Fasheh328d5752007-06-18 10:48:04 -07003151 return ret;
3152}
3153
3154/*
3155 * Remove split_rec clusters from the record at index and merge them
Tao Ma677b9752008-01-30 14:21:05 +08003156 * onto the tail of the record "before" it.
3157 * For index > 0, the "before" means the extent rec at index - 1.
3158 *
3159 * For index == 0, the "before" means the last record of the previous
3160 * extent block. And there is also a situation that we may need to
3161 * remove the rightmost leaf extent block in the right_path and change
3162 * the right path to indicate the new rightmost path.
Mark Fasheh328d5752007-06-18 10:48:04 -07003163 */
Tao Ma677b9752008-01-30 14:21:05 +08003164static int ocfs2_merge_rec_left(struct inode *inode,
3165 struct ocfs2_path *right_path,
Mark Fasheh328d5752007-06-18 10:48:04 -07003166 handle_t *handle,
3167 struct ocfs2_extent_rec *split_rec,
Tao Ma677b9752008-01-30 14:21:05 +08003168 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08003169 struct ocfs2_extent_tree *et,
Tao Ma677b9752008-01-30 14:21:05 +08003170 int index)
Mark Fasheh328d5752007-06-18 10:48:04 -07003171{
Tao Ma677b9752008-01-30 14:21:05 +08003172 int ret, i, subtree_index = 0, has_empty_extent = 0;
Mark Fasheh328d5752007-06-18 10:48:04 -07003173 unsigned int split_clusters = le16_to_cpu(split_rec->e_leaf_clusters);
3174 struct ocfs2_extent_rec *left_rec;
3175 struct ocfs2_extent_rec *right_rec;
Tao Ma677b9752008-01-30 14:21:05 +08003176 struct ocfs2_extent_list *el = path_leaf_el(right_path);
3177 struct buffer_head *bh = path_leaf_bh(right_path);
3178 struct buffer_head *root_bh = NULL;
3179 struct ocfs2_path *left_path = NULL;
3180 struct ocfs2_extent_list *left_el;
Mark Fasheh328d5752007-06-18 10:48:04 -07003181
Tao Ma677b9752008-01-30 14:21:05 +08003182 BUG_ON(index < 0);
Mark Fasheh328d5752007-06-18 10:48:04 -07003183
Mark Fasheh328d5752007-06-18 10:48:04 -07003184 right_rec = &el->l_recs[index];
Tao Ma677b9752008-01-30 14:21:05 +08003185 if (index == 0) {
3186 /* we meet with a cross extent block merge. */
3187 ret = ocfs2_get_left_path(inode, right_path, &left_path);
3188 if (ret) {
3189 mlog_errno(ret);
3190 goto out;
3191 }
3192
3193 left_el = path_leaf_el(left_path);
3194 BUG_ON(le16_to_cpu(left_el->l_next_free_rec) !=
3195 le16_to_cpu(left_el->l_count));
3196
3197 left_rec = &left_el->l_recs[
3198 le16_to_cpu(left_el->l_next_free_rec) - 1];
3199 BUG_ON(le32_to_cpu(left_rec->e_cpos) +
3200 le16_to_cpu(left_rec->e_leaf_clusters) !=
3201 le32_to_cpu(split_rec->e_cpos));
3202
3203 subtree_index = ocfs2_find_subtree_root(inode,
3204 left_path, right_path);
3205
3206 ret = ocfs2_extend_rotate_transaction(handle, subtree_index,
3207 handle->h_buffer_credits,
3208 left_path);
3209 if (ret) {
3210 mlog_errno(ret);
3211 goto out;
3212 }
3213
3214 root_bh = left_path->p_node[subtree_index].bh;
3215 BUG_ON(root_bh != right_path->p_node[subtree_index].bh);
3216
3217 ret = ocfs2_journal_access(handle, inode, root_bh,
3218 OCFS2_JOURNAL_ACCESS_WRITE);
3219 if (ret) {
3220 mlog_errno(ret);
3221 goto out;
3222 }
3223
3224 for (i = subtree_index + 1;
3225 i < path_num_items(right_path); i++) {
3226 ret = ocfs2_journal_access(handle, inode,
3227 right_path->p_node[i].bh,
3228 OCFS2_JOURNAL_ACCESS_WRITE);
3229 if (ret) {
3230 mlog_errno(ret);
3231 goto out;
3232 }
3233
3234 ret = ocfs2_journal_access(handle, inode,
3235 left_path->p_node[i].bh,
3236 OCFS2_JOURNAL_ACCESS_WRITE);
3237 if (ret) {
3238 mlog_errno(ret);
3239 goto out;
3240 }
3241 }
3242 } else {
3243 left_rec = &el->l_recs[index - 1];
3244 if (ocfs2_is_empty_extent(&el->l_recs[0]))
3245 has_empty_extent = 1;
3246 }
Mark Fasheh328d5752007-06-18 10:48:04 -07003247
3248 ret = ocfs2_journal_access(handle, inode, bh,
3249 OCFS2_JOURNAL_ACCESS_WRITE);
3250 if (ret) {
3251 mlog_errno(ret);
3252 goto out;
3253 }
3254
3255 if (has_empty_extent && index == 1) {
3256 /*
3257 * The easy case - we can just plop the record right in.
3258 */
3259 *left_rec = *split_rec;
3260
3261 has_empty_extent = 0;
Tao Ma677b9752008-01-30 14:21:05 +08003262 } else
Mark Fasheh328d5752007-06-18 10:48:04 -07003263 le16_add_cpu(&left_rec->e_leaf_clusters, split_clusters);
Mark Fasheh328d5752007-06-18 10:48:04 -07003264
3265 le32_add_cpu(&right_rec->e_cpos, split_clusters);
3266 le64_add_cpu(&right_rec->e_blkno,
3267 ocfs2_clusters_to_blocks(inode->i_sb, split_clusters));
3268 le16_add_cpu(&right_rec->e_leaf_clusters, -split_clusters);
3269
3270 ocfs2_cleanup_merge(el, index);
3271
3272 ret = ocfs2_journal_dirty(handle, bh);
3273 if (ret)
3274 mlog_errno(ret);
3275
Tao Ma677b9752008-01-30 14:21:05 +08003276 if (left_path) {
3277 ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
3278 if (ret)
3279 mlog_errno(ret);
3280
3281 /*
3282 * In the situation that the right_rec is empty and the extent
3283 * block is empty also, ocfs2_complete_edge_insert can't handle
3284 * it and we need to delete the right extent block.
3285 */
3286 if (le16_to_cpu(right_rec->e_leaf_clusters) == 0 &&
3287 le16_to_cpu(el->l_next_free_rec) == 1) {
3288
3289 ret = ocfs2_remove_rightmost_path(inode, handle,
Tao Mae7d4cb62008-08-18 17:38:44 +08003290 right_path,
3291 dealloc, et);
Tao Ma677b9752008-01-30 14:21:05 +08003292 if (ret) {
3293 mlog_errno(ret);
3294 goto out;
3295 }
3296
3297 /* Now the rightmost extent block has been deleted.
3298 * So we use the new rightmost path.
3299 */
3300 ocfs2_mv_path(right_path, left_path);
3301 left_path = NULL;
3302 } else
3303 ocfs2_complete_edge_insert(inode, handle, left_path,
3304 right_path, subtree_index);
3305 }
Mark Fasheh328d5752007-06-18 10:48:04 -07003306out:
Tao Ma677b9752008-01-30 14:21:05 +08003307 if (left_path)
3308 ocfs2_free_path(left_path);
Mark Fasheh328d5752007-06-18 10:48:04 -07003309 return ret;
3310}
3311
3312static int ocfs2_try_to_merge_extent(struct inode *inode,
3313 handle_t *handle,
Tao Ma677b9752008-01-30 14:21:05 +08003314 struct ocfs2_path *path,
Mark Fasheh328d5752007-06-18 10:48:04 -07003315 int split_index,
3316 struct ocfs2_extent_rec *split_rec,
3317 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08003318 struct ocfs2_merge_ctxt *ctxt,
3319 struct ocfs2_extent_tree *et)
Mark Fasheh328d5752007-06-18 10:48:04 -07003320
3321{
Tao Mao518d7262007-08-28 17:25:35 -07003322 int ret = 0;
Tao Ma677b9752008-01-30 14:21:05 +08003323 struct ocfs2_extent_list *el = path_leaf_el(path);
Mark Fasheh328d5752007-06-18 10:48:04 -07003324 struct ocfs2_extent_rec *rec = &el->l_recs[split_index];
3325
3326 BUG_ON(ctxt->c_contig_type == CONTIG_NONE);
3327
Tao Mao518d7262007-08-28 17:25:35 -07003328 if (ctxt->c_split_covers_rec && ctxt->c_has_empty_extent) {
3329 /*
3330 * The merge code will need to create an empty
3331 * extent to take the place of the newly
3332 * emptied slot. Remove any pre-existing empty
3333 * extents - having more than one in a leaf is
3334 * illegal.
3335 */
Tao Ma677b9752008-01-30 14:21:05 +08003336 ret = ocfs2_rotate_tree_left(inode, handle, path,
Tao Mae7d4cb62008-08-18 17:38:44 +08003337 dealloc, et);
Tao Mao518d7262007-08-28 17:25:35 -07003338 if (ret) {
3339 mlog_errno(ret);
3340 goto out;
Mark Fasheh328d5752007-06-18 10:48:04 -07003341 }
Tao Mao518d7262007-08-28 17:25:35 -07003342 split_index--;
3343 rec = &el->l_recs[split_index];
Mark Fasheh328d5752007-06-18 10:48:04 -07003344 }
3345
3346 if (ctxt->c_contig_type == CONTIG_LEFTRIGHT) {
3347 /*
3348 * Left-right contig implies this.
3349 */
3350 BUG_ON(!ctxt->c_split_covers_rec);
Mark Fasheh328d5752007-06-18 10:48:04 -07003351
3352 /*
3353 * Since the leftright insert always covers the entire
3354 * extent, this call will delete the insert record
3355 * entirely, resulting in an empty extent record added to
3356 * the extent block.
3357 *
3358 * Since the adding of an empty extent shifts
3359 * everything back to the right, there's no need to
3360 * update split_index here.
Tao Ma677b9752008-01-30 14:21:05 +08003361 *
3362 * When the split_index is zero, we need to merge it to the
3363 * prevoius extent block. It is more efficient and easier
3364 * if we do merge_right first and merge_left later.
Mark Fasheh328d5752007-06-18 10:48:04 -07003365 */
Tao Ma677b9752008-01-30 14:21:05 +08003366 ret = ocfs2_merge_rec_right(inode, path,
3367 handle, split_rec,
3368 split_index);
Mark Fasheh328d5752007-06-18 10:48:04 -07003369 if (ret) {
3370 mlog_errno(ret);
3371 goto out;
3372 }
3373
3374 /*
3375 * We can only get this from logic error above.
3376 */
3377 BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0]));
3378
Tao Ma677b9752008-01-30 14:21:05 +08003379 /* The merge left us with an empty extent, remove it. */
Tao Mae7d4cb62008-08-18 17:38:44 +08003380 ret = ocfs2_rotate_tree_left(inode, handle, path,
3381 dealloc, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07003382 if (ret) {
3383 mlog_errno(ret);
3384 goto out;
3385 }
Tao Ma677b9752008-01-30 14:21:05 +08003386
Mark Fasheh328d5752007-06-18 10:48:04 -07003387 rec = &el->l_recs[split_index];
3388
3389 /*
3390 * Note that we don't pass split_rec here on purpose -
Tao Ma677b9752008-01-30 14:21:05 +08003391 * we've merged it into the rec already.
Mark Fasheh328d5752007-06-18 10:48:04 -07003392 */
Tao Ma677b9752008-01-30 14:21:05 +08003393 ret = ocfs2_merge_rec_left(inode, path,
3394 handle, rec,
Tao Mae7d4cb62008-08-18 17:38:44 +08003395 dealloc, et,
Tao Ma677b9752008-01-30 14:21:05 +08003396 split_index);
3397
Mark Fasheh328d5752007-06-18 10:48:04 -07003398 if (ret) {
3399 mlog_errno(ret);
3400 goto out;
3401 }
3402
Tao Ma677b9752008-01-30 14:21:05 +08003403 ret = ocfs2_rotate_tree_left(inode, handle, path,
Tao Mae7d4cb62008-08-18 17:38:44 +08003404 dealloc, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07003405 /*
3406 * Error from this last rotate is not critical, so
3407 * print but don't bubble it up.
3408 */
3409 if (ret)
3410 mlog_errno(ret);
3411 ret = 0;
3412 } else {
3413 /*
3414 * Merge a record to the left or right.
3415 *
3416 * 'contig_type' is relative to the existing record,
3417 * so for example, if we're "right contig", it's to
3418 * the record on the left (hence the left merge).
3419 */
3420 if (ctxt->c_contig_type == CONTIG_RIGHT) {
3421 ret = ocfs2_merge_rec_left(inode,
Tao Ma677b9752008-01-30 14:21:05 +08003422 path,
3423 handle, split_rec,
Tao Mae7d4cb62008-08-18 17:38:44 +08003424 dealloc, et,
Mark Fasheh328d5752007-06-18 10:48:04 -07003425 split_index);
3426 if (ret) {
3427 mlog_errno(ret);
3428 goto out;
3429 }
3430 } else {
3431 ret = ocfs2_merge_rec_right(inode,
Tao Ma677b9752008-01-30 14:21:05 +08003432 path,
3433 handle, split_rec,
Mark Fasheh328d5752007-06-18 10:48:04 -07003434 split_index);
3435 if (ret) {
3436 mlog_errno(ret);
3437 goto out;
3438 }
3439 }
3440
3441 if (ctxt->c_split_covers_rec) {
3442 /*
3443 * The merge may have left an empty extent in
3444 * our leaf. Try to rotate it away.
3445 */
Tao Ma677b9752008-01-30 14:21:05 +08003446 ret = ocfs2_rotate_tree_left(inode, handle, path,
Tao Mae7d4cb62008-08-18 17:38:44 +08003447 dealloc, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07003448 if (ret)
3449 mlog_errno(ret);
3450 ret = 0;
3451 }
3452 }
3453
3454out:
3455 return ret;
3456}
3457
3458static void ocfs2_subtract_from_rec(struct super_block *sb,
3459 enum ocfs2_split_type split,
3460 struct ocfs2_extent_rec *rec,
3461 struct ocfs2_extent_rec *split_rec)
3462{
3463 u64 len_blocks;
3464
3465 len_blocks = ocfs2_clusters_to_blocks(sb,
3466 le16_to_cpu(split_rec->e_leaf_clusters));
3467
3468 if (split == SPLIT_LEFT) {
3469 /*
3470 * Region is on the left edge of the existing
3471 * record.
3472 */
3473 le32_add_cpu(&rec->e_cpos,
3474 le16_to_cpu(split_rec->e_leaf_clusters));
3475 le64_add_cpu(&rec->e_blkno, len_blocks);
3476 le16_add_cpu(&rec->e_leaf_clusters,
3477 -le16_to_cpu(split_rec->e_leaf_clusters));
3478 } else {
3479 /*
3480 * Region is on the right edge of the existing
3481 * record.
3482 */
3483 le16_add_cpu(&rec->e_leaf_clusters,
3484 -le16_to_cpu(split_rec->e_leaf_clusters));
3485 }
3486}
3487
Mark Fashehdcd05382007-01-16 11:32:23 -08003488/*
3489 * Do the final bits of extent record insertion at the target leaf
3490 * list. If this leaf is part of an allocation tree, it is assumed
3491 * that the tree above has been prepared.
3492 */
3493static void ocfs2_insert_at_leaf(struct ocfs2_extent_rec *insert_rec,
3494 struct ocfs2_extent_list *el,
3495 struct ocfs2_insert_type *insert,
3496 struct inode *inode)
3497{
3498 int i = insert->ins_contig_index;
3499 unsigned int range;
3500 struct ocfs2_extent_rec *rec;
3501
Mark Fashehe48edee2007-03-07 16:46:57 -08003502 BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
Mark Fashehdcd05382007-01-16 11:32:23 -08003503
Mark Fasheh328d5752007-06-18 10:48:04 -07003504 if (insert->ins_split != SPLIT_NONE) {
3505 i = ocfs2_search_extent_list(el, le32_to_cpu(insert_rec->e_cpos));
3506 BUG_ON(i == -1);
3507 rec = &el->l_recs[i];
3508 ocfs2_subtract_from_rec(inode->i_sb, insert->ins_split, rec,
3509 insert_rec);
3510 goto rotate;
3511 }
3512
Mark Fashehdcd05382007-01-16 11:32:23 -08003513 /*
3514 * Contiguous insert - either left or right.
3515 */
3516 if (insert->ins_contig != CONTIG_NONE) {
3517 rec = &el->l_recs[i];
3518 if (insert->ins_contig == CONTIG_LEFT) {
3519 rec->e_blkno = insert_rec->e_blkno;
3520 rec->e_cpos = insert_rec->e_cpos;
3521 }
Mark Fashehe48edee2007-03-07 16:46:57 -08003522 le16_add_cpu(&rec->e_leaf_clusters,
3523 le16_to_cpu(insert_rec->e_leaf_clusters));
Mark Fashehdcd05382007-01-16 11:32:23 -08003524 return;
3525 }
3526
3527 /*
3528 * Handle insert into an empty leaf.
3529 */
3530 if (le16_to_cpu(el->l_next_free_rec) == 0 ||
3531 ((le16_to_cpu(el->l_next_free_rec) == 1) &&
3532 ocfs2_is_empty_extent(&el->l_recs[0]))) {
3533 el->l_recs[0] = *insert_rec;
3534 el->l_next_free_rec = cpu_to_le16(1);
3535 return;
3536 }
3537
3538 /*
3539 * Appending insert.
3540 */
3541 if (insert->ins_appending == APPEND_TAIL) {
3542 i = le16_to_cpu(el->l_next_free_rec) - 1;
3543 rec = &el->l_recs[i];
Mark Fashehe48edee2007-03-07 16:46:57 -08003544 range = le32_to_cpu(rec->e_cpos)
3545 + le16_to_cpu(rec->e_leaf_clusters);
Mark Fashehdcd05382007-01-16 11:32:23 -08003546 BUG_ON(le32_to_cpu(insert_rec->e_cpos) < range);
3547
3548 mlog_bug_on_msg(le16_to_cpu(el->l_next_free_rec) >=
3549 le16_to_cpu(el->l_count),
3550 "inode %lu, depth %u, count %u, next free %u, "
3551 "rec.cpos %u, rec.clusters %u, "
3552 "insert.cpos %u, insert.clusters %u\n",
3553 inode->i_ino,
3554 le16_to_cpu(el->l_tree_depth),
3555 le16_to_cpu(el->l_count),
3556 le16_to_cpu(el->l_next_free_rec),
3557 le32_to_cpu(el->l_recs[i].e_cpos),
Mark Fashehe48edee2007-03-07 16:46:57 -08003558 le16_to_cpu(el->l_recs[i].e_leaf_clusters),
Mark Fashehdcd05382007-01-16 11:32:23 -08003559 le32_to_cpu(insert_rec->e_cpos),
Mark Fashehe48edee2007-03-07 16:46:57 -08003560 le16_to_cpu(insert_rec->e_leaf_clusters));
Mark Fashehdcd05382007-01-16 11:32:23 -08003561 i++;
3562 el->l_recs[i] = *insert_rec;
3563 le16_add_cpu(&el->l_next_free_rec, 1);
3564 return;
3565 }
3566
Mark Fasheh328d5752007-06-18 10:48:04 -07003567rotate:
Mark Fashehdcd05382007-01-16 11:32:23 -08003568 /*
3569 * Ok, we have to rotate.
3570 *
3571 * At this point, it is safe to assume that inserting into an
3572 * empty leaf and appending to a leaf have both been handled
3573 * above.
3574 *
3575 * This leaf needs to have space, either by the empty 1st
3576 * extent record, or by virtue of an l_next_rec < l_count.
3577 */
3578 ocfs2_rotate_leaf(el, insert_rec);
3579}
3580
Mark Fasheh328d5752007-06-18 10:48:04 -07003581static void ocfs2_adjust_rightmost_records(struct inode *inode,
3582 handle_t *handle,
3583 struct ocfs2_path *path,
3584 struct ocfs2_extent_rec *insert_rec)
3585{
3586 int ret, i, next_free;
3587 struct buffer_head *bh;
3588 struct ocfs2_extent_list *el;
3589 struct ocfs2_extent_rec *rec;
3590
3591 /*
3592 * Update everything except the leaf block.
3593 */
3594 for (i = 0; i < path->p_tree_depth; i++) {
3595 bh = path->p_node[i].bh;
3596 el = path->p_node[i].el;
3597
3598 next_free = le16_to_cpu(el->l_next_free_rec);
3599 if (next_free == 0) {
3600 ocfs2_error(inode->i_sb,
3601 "Dinode %llu has a bad extent list",
3602 (unsigned long long)OCFS2_I(inode)->ip_blkno);
3603 ret = -EIO;
3604 return;
3605 }
3606
3607 rec = &el->l_recs[next_free - 1];
3608
3609 rec->e_int_clusters = insert_rec->e_cpos;
3610 le32_add_cpu(&rec->e_int_clusters,
3611 le16_to_cpu(insert_rec->e_leaf_clusters));
3612 le32_add_cpu(&rec->e_int_clusters,
3613 -le32_to_cpu(rec->e_cpos));
3614
3615 ret = ocfs2_journal_dirty(handle, bh);
3616 if (ret)
3617 mlog_errno(ret);
3618
3619 }
3620}
3621
Mark Fashehdcd05382007-01-16 11:32:23 -08003622static int ocfs2_append_rec_to_path(struct inode *inode, handle_t *handle,
3623 struct ocfs2_extent_rec *insert_rec,
3624 struct ocfs2_path *right_path,
3625 struct ocfs2_path **ret_left_path)
3626{
Mark Fasheh328d5752007-06-18 10:48:04 -07003627 int ret, next_free;
Mark Fashehdcd05382007-01-16 11:32:23 -08003628 struct ocfs2_extent_list *el;
3629 struct ocfs2_path *left_path = NULL;
3630
3631 *ret_left_path = NULL;
3632
3633 /*
Mark Fashehe48edee2007-03-07 16:46:57 -08003634 * This shouldn't happen for non-trees. The extent rec cluster
3635 * count manipulation below only works for interior nodes.
3636 */
3637 BUG_ON(right_path->p_tree_depth == 0);
3638
3639 /*
Mark Fashehdcd05382007-01-16 11:32:23 -08003640 * If our appending insert is at the leftmost edge of a leaf,
3641 * then we might need to update the rightmost records of the
3642 * neighboring path.
3643 */
3644 el = path_leaf_el(right_path);
3645 next_free = le16_to_cpu(el->l_next_free_rec);
3646 if (next_free == 0 ||
3647 (next_free == 1 && ocfs2_is_empty_extent(&el->l_recs[0]))) {
3648 u32 left_cpos;
3649
3650 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, right_path,
3651 &left_cpos);
3652 if (ret) {
3653 mlog_errno(ret);
3654 goto out;
3655 }
3656
3657 mlog(0, "Append may need a left path update. cpos: %u, "
3658 "left_cpos: %u\n", le32_to_cpu(insert_rec->e_cpos),
3659 left_cpos);
3660
3661 /*
3662 * No need to worry if the append is already in the
3663 * leftmost leaf.
3664 */
3665 if (left_cpos) {
3666 left_path = ocfs2_new_path(path_root_bh(right_path),
3667 path_root_el(right_path));
3668 if (!left_path) {
3669 ret = -ENOMEM;
3670 mlog_errno(ret);
3671 goto out;
3672 }
3673
3674 ret = ocfs2_find_path(inode, left_path, left_cpos);
3675 if (ret) {
3676 mlog_errno(ret);
3677 goto out;
3678 }
3679
3680 /*
3681 * ocfs2_insert_path() will pass the left_path to the
3682 * journal for us.
3683 */
3684 }
3685 }
3686
3687 ret = ocfs2_journal_access_path(inode, handle, right_path);
3688 if (ret) {
3689 mlog_errno(ret);
3690 goto out;
3691 }
3692
Mark Fasheh328d5752007-06-18 10:48:04 -07003693 ocfs2_adjust_rightmost_records(inode, handle, right_path, insert_rec);
Mark Fashehdcd05382007-01-16 11:32:23 -08003694
3695 *ret_left_path = left_path;
3696 ret = 0;
3697out:
3698 if (ret != 0)
3699 ocfs2_free_path(left_path);
3700
3701 return ret;
3702}
3703
Mark Fasheh328d5752007-06-18 10:48:04 -07003704static void ocfs2_split_record(struct inode *inode,
3705 struct ocfs2_path *left_path,
3706 struct ocfs2_path *right_path,
3707 struct ocfs2_extent_rec *split_rec,
3708 enum ocfs2_split_type split)
3709{
3710 int index;
3711 u32 cpos = le32_to_cpu(split_rec->e_cpos);
3712 struct ocfs2_extent_list *left_el = NULL, *right_el, *insert_el, *el;
3713 struct ocfs2_extent_rec *rec, *tmprec;
3714
3715 right_el = path_leaf_el(right_path);;
3716 if (left_path)
3717 left_el = path_leaf_el(left_path);
3718
3719 el = right_el;
3720 insert_el = right_el;
3721 index = ocfs2_search_extent_list(el, cpos);
3722 if (index != -1) {
3723 if (index == 0 && left_path) {
3724 BUG_ON(ocfs2_is_empty_extent(&el->l_recs[0]));
3725
3726 /*
3727 * This typically means that the record
3728 * started in the left path but moved to the
3729 * right as a result of rotation. We either
3730 * move the existing record to the left, or we
3731 * do the later insert there.
3732 *
3733 * In this case, the left path should always
3734 * exist as the rotate code will have passed
3735 * it back for a post-insert update.
3736 */
3737
3738 if (split == SPLIT_LEFT) {
3739 /*
3740 * It's a left split. Since we know
3741 * that the rotate code gave us an
3742 * empty extent in the left path, we
3743 * can just do the insert there.
3744 */
3745 insert_el = left_el;
3746 } else {
3747 /*
3748 * Right split - we have to move the
3749 * existing record over to the left
3750 * leaf. The insert will be into the
3751 * newly created empty extent in the
3752 * right leaf.
3753 */
3754 tmprec = &right_el->l_recs[index];
3755 ocfs2_rotate_leaf(left_el, tmprec);
3756 el = left_el;
3757
3758 memset(tmprec, 0, sizeof(*tmprec));
3759 index = ocfs2_search_extent_list(left_el, cpos);
3760 BUG_ON(index == -1);
3761 }
3762 }
3763 } else {
3764 BUG_ON(!left_path);
3765 BUG_ON(!ocfs2_is_empty_extent(&left_el->l_recs[0]));
3766 /*
3767 * Left path is easy - we can just allow the insert to
3768 * happen.
3769 */
3770 el = left_el;
3771 insert_el = left_el;
3772 index = ocfs2_search_extent_list(el, cpos);
3773 BUG_ON(index == -1);
3774 }
3775
3776 rec = &el->l_recs[index];
3777 ocfs2_subtract_from_rec(inode->i_sb, split, rec, split_rec);
3778 ocfs2_rotate_leaf(insert_el, split_rec);
3779}
3780
Mark Fashehdcd05382007-01-16 11:32:23 -08003781/*
Tao Mae7d4cb62008-08-18 17:38:44 +08003782 * This function only does inserts on an allocation b-tree. For tree
3783 * depth = 0, ocfs2_insert_at_leaf() is called directly.
Mark Fashehdcd05382007-01-16 11:32:23 -08003784 *
3785 * right_path is the path we want to do the actual insert
3786 * in. left_path should only be passed in if we need to update that
3787 * portion of the tree after an edge insert.
3788 */
3789static int ocfs2_insert_path(struct inode *inode,
3790 handle_t *handle,
3791 struct ocfs2_path *left_path,
3792 struct ocfs2_path *right_path,
3793 struct ocfs2_extent_rec *insert_rec,
3794 struct ocfs2_insert_type *insert)
3795{
3796 int ret, subtree_index;
3797 struct buffer_head *leaf_bh = path_leaf_bh(right_path);
Mark Fashehdcd05382007-01-16 11:32:23 -08003798
Mark Fashehdcd05382007-01-16 11:32:23 -08003799 if (left_path) {
3800 int credits = handle->h_buffer_credits;
3801
3802 /*
3803 * There's a chance that left_path got passed back to
3804 * us without being accounted for in the
3805 * journal. Extend our transaction here to be sure we
3806 * can change those blocks.
3807 */
3808 credits += left_path->p_tree_depth;
3809
3810 ret = ocfs2_extend_trans(handle, credits);
3811 if (ret < 0) {
3812 mlog_errno(ret);
3813 goto out;
3814 }
3815
3816 ret = ocfs2_journal_access_path(inode, handle, left_path);
3817 if (ret < 0) {
3818 mlog_errno(ret);
3819 goto out;
3820 }
3821 }
3822
Mark Fashehe8aed342007-12-03 16:43:01 -08003823 /*
3824 * Pass both paths to the journal. The majority of inserts
3825 * will be touching all components anyway.
3826 */
3827 ret = ocfs2_journal_access_path(inode, handle, right_path);
3828 if (ret < 0) {
3829 mlog_errno(ret);
3830 goto out;
3831 }
3832
Mark Fasheh328d5752007-06-18 10:48:04 -07003833 if (insert->ins_split != SPLIT_NONE) {
3834 /*
3835 * We could call ocfs2_insert_at_leaf() for some types
Joe Perchesc78bad12008-02-03 17:33:42 +02003836 * of splits, but it's easier to just let one separate
Mark Fasheh328d5752007-06-18 10:48:04 -07003837 * function sort it all out.
3838 */
3839 ocfs2_split_record(inode, left_path, right_path,
3840 insert_rec, insert->ins_split);
Mark Fashehe8aed342007-12-03 16:43:01 -08003841
3842 /*
3843 * Split might have modified either leaf and we don't
3844 * have a guarantee that the later edge insert will
3845 * dirty this for us.
3846 */
3847 if (left_path)
3848 ret = ocfs2_journal_dirty(handle,
3849 path_leaf_bh(left_path));
3850 if (ret)
3851 mlog_errno(ret);
Mark Fasheh328d5752007-06-18 10:48:04 -07003852 } else
3853 ocfs2_insert_at_leaf(insert_rec, path_leaf_el(right_path),
3854 insert, inode);
Mark Fashehdcd05382007-01-16 11:32:23 -08003855
Mark Fashehdcd05382007-01-16 11:32:23 -08003856 ret = ocfs2_journal_dirty(handle, leaf_bh);
3857 if (ret)
3858 mlog_errno(ret);
3859
3860 if (left_path) {
3861 /*
3862 * The rotate code has indicated that we need to fix
3863 * up portions of the tree after the insert.
3864 *
3865 * XXX: Should we extend the transaction here?
3866 */
3867 subtree_index = ocfs2_find_subtree_root(inode, left_path,
3868 right_path);
3869 ocfs2_complete_edge_insert(inode, handle, left_path,
3870 right_path, subtree_index);
3871 }
3872
3873 ret = 0;
3874out:
3875 return ret;
3876}
3877
3878static int ocfs2_do_insert_extent(struct inode *inode,
3879 handle_t *handle,
Tao Mae7d4cb62008-08-18 17:38:44 +08003880 struct ocfs2_extent_tree *et,
Mark Fashehdcd05382007-01-16 11:32:23 -08003881 struct ocfs2_extent_rec *insert_rec,
3882 struct ocfs2_insert_type *type)
3883{
3884 int ret, rotate = 0;
3885 u32 cpos;
3886 struct ocfs2_path *right_path = NULL;
3887 struct ocfs2_path *left_path = NULL;
Mark Fashehdcd05382007-01-16 11:32:23 -08003888 struct ocfs2_extent_list *el;
3889
Joel Beckerce1d9ea2008-08-20 16:30:07 -07003890 el = et->et_root_el;
Mark Fashehdcd05382007-01-16 11:32:23 -08003891
Joel Beckerce1d9ea2008-08-20 16:30:07 -07003892 ret = ocfs2_journal_access(handle, inode, et->et_root_bh,
Mark Fashehdcd05382007-01-16 11:32:23 -08003893 OCFS2_JOURNAL_ACCESS_WRITE);
3894 if (ret) {
3895 mlog_errno(ret);
3896 goto out;
3897 }
3898
3899 if (le16_to_cpu(el->l_tree_depth) == 0) {
3900 ocfs2_insert_at_leaf(insert_rec, el, type, inode);
3901 goto out_update_clusters;
3902 }
3903
Joel Beckerce1d9ea2008-08-20 16:30:07 -07003904 right_path = ocfs2_new_path(et->et_root_bh, et->et_root_el);
Mark Fashehdcd05382007-01-16 11:32:23 -08003905 if (!right_path) {
3906 ret = -ENOMEM;
3907 mlog_errno(ret);
3908 goto out;
3909 }
3910
3911 /*
3912 * Determine the path to start with. Rotations need the
3913 * rightmost path, everything else can go directly to the
3914 * target leaf.
3915 */
3916 cpos = le32_to_cpu(insert_rec->e_cpos);
3917 if (type->ins_appending == APPEND_NONE &&
3918 type->ins_contig == CONTIG_NONE) {
3919 rotate = 1;
3920 cpos = UINT_MAX;
3921 }
3922
3923 ret = ocfs2_find_path(inode, right_path, cpos);
3924 if (ret) {
3925 mlog_errno(ret);
3926 goto out;
3927 }
3928
3929 /*
3930 * Rotations and appends need special treatment - they modify
3931 * parts of the tree's above them.
3932 *
3933 * Both might pass back a path immediate to the left of the
3934 * one being inserted to. This will be cause
3935 * ocfs2_insert_path() to modify the rightmost records of
3936 * left_path to account for an edge insert.
3937 *
3938 * XXX: When modifying this code, keep in mind that an insert
3939 * can wind up skipping both of these two special cases...
3940 */
3941 if (rotate) {
Mark Fasheh328d5752007-06-18 10:48:04 -07003942 ret = ocfs2_rotate_tree_right(inode, handle, type->ins_split,
Mark Fashehdcd05382007-01-16 11:32:23 -08003943 le32_to_cpu(insert_rec->e_cpos),
3944 right_path, &left_path);
3945 if (ret) {
3946 mlog_errno(ret);
3947 goto out;
3948 }
Mark Fashehe8aed342007-12-03 16:43:01 -08003949
3950 /*
3951 * ocfs2_rotate_tree_right() might have extended the
3952 * transaction without re-journaling our tree root.
3953 */
Joel Beckerce1d9ea2008-08-20 16:30:07 -07003954 ret = ocfs2_journal_access(handle, inode, et->et_root_bh,
Mark Fashehe8aed342007-12-03 16:43:01 -08003955 OCFS2_JOURNAL_ACCESS_WRITE);
3956 if (ret) {
3957 mlog_errno(ret);
3958 goto out;
3959 }
Mark Fashehdcd05382007-01-16 11:32:23 -08003960 } else if (type->ins_appending == APPEND_TAIL
3961 && type->ins_contig != CONTIG_LEFT) {
3962 ret = ocfs2_append_rec_to_path(inode, handle, insert_rec,
3963 right_path, &left_path);
3964 if (ret) {
3965 mlog_errno(ret);
3966 goto out;
3967 }
3968 }
3969
3970 ret = ocfs2_insert_path(inode, handle, left_path, right_path,
3971 insert_rec, type);
3972 if (ret) {
3973 mlog_errno(ret);
3974 goto out;
3975 }
3976
3977out_update_clusters:
Mark Fasheh328d5752007-06-18 10:48:04 -07003978 if (type->ins_split == SPLIT_NONE)
Joel Becker35dc0aa2008-08-20 16:25:06 -07003979 ocfs2_et_update_clusters(inode, et,
3980 le16_to_cpu(insert_rec->e_leaf_clusters));
Mark Fashehdcd05382007-01-16 11:32:23 -08003981
Joel Beckerce1d9ea2008-08-20 16:30:07 -07003982 ret = ocfs2_journal_dirty(handle, et->et_root_bh);
Mark Fashehdcd05382007-01-16 11:32:23 -08003983 if (ret)
3984 mlog_errno(ret);
3985
3986out:
3987 ocfs2_free_path(left_path);
3988 ocfs2_free_path(right_path);
3989
3990 return ret;
3991}
3992
Mark Fasheh328d5752007-06-18 10:48:04 -07003993static enum ocfs2_contig_type
Tao Maad5a4d72008-01-30 14:21:32 +08003994ocfs2_figure_merge_contig_type(struct inode *inode, struct ocfs2_path *path,
Mark Fasheh328d5752007-06-18 10:48:04 -07003995 struct ocfs2_extent_list *el, int index,
3996 struct ocfs2_extent_rec *split_rec)
3997{
Tao Maad5a4d72008-01-30 14:21:32 +08003998 int status;
Mark Fasheh328d5752007-06-18 10:48:04 -07003999 enum ocfs2_contig_type ret = CONTIG_NONE;
Tao Maad5a4d72008-01-30 14:21:32 +08004000 u32 left_cpos, right_cpos;
4001 struct ocfs2_extent_rec *rec = NULL;
4002 struct ocfs2_extent_list *new_el;
4003 struct ocfs2_path *left_path = NULL, *right_path = NULL;
4004 struct buffer_head *bh;
4005 struct ocfs2_extent_block *eb;
4006
4007 if (index > 0) {
4008 rec = &el->l_recs[index - 1];
4009 } else if (path->p_tree_depth > 0) {
4010 status = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
4011 path, &left_cpos);
4012 if (status)
4013 goto out;
4014
4015 if (left_cpos != 0) {
4016 left_path = ocfs2_new_path(path_root_bh(path),
4017 path_root_el(path));
4018 if (!left_path)
4019 goto out;
4020
4021 status = ocfs2_find_path(inode, left_path, left_cpos);
4022 if (status)
4023 goto out;
4024
4025 new_el = path_leaf_el(left_path);
4026
4027 if (le16_to_cpu(new_el->l_next_free_rec) !=
4028 le16_to_cpu(new_el->l_count)) {
4029 bh = path_leaf_bh(left_path);
4030 eb = (struct ocfs2_extent_block *)bh->b_data;
4031 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb,
4032 eb);
4033 goto out;
4034 }
4035 rec = &new_el->l_recs[
4036 le16_to_cpu(new_el->l_next_free_rec) - 1];
4037 }
4038 }
Mark Fasheh328d5752007-06-18 10:48:04 -07004039
4040 /*
4041 * We're careful to check for an empty extent record here -
4042 * the merge code will know what to do if it sees one.
4043 */
Tao Maad5a4d72008-01-30 14:21:32 +08004044 if (rec) {
Mark Fasheh328d5752007-06-18 10:48:04 -07004045 if (index == 1 && ocfs2_is_empty_extent(rec)) {
4046 if (split_rec->e_cpos == el->l_recs[index].e_cpos)
4047 ret = CONTIG_RIGHT;
4048 } else {
4049 ret = ocfs2_extent_contig(inode, rec, split_rec);
4050 }
4051 }
4052
Tao Maad5a4d72008-01-30 14:21:32 +08004053 rec = NULL;
4054 if (index < (le16_to_cpu(el->l_next_free_rec) - 1))
4055 rec = &el->l_recs[index + 1];
4056 else if (le16_to_cpu(el->l_next_free_rec) == le16_to_cpu(el->l_count) &&
4057 path->p_tree_depth > 0) {
4058 status = ocfs2_find_cpos_for_right_leaf(inode->i_sb,
4059 path, &right_cpos);
4060 if (status)
4061 goto out;
4062
4063 if (right_cpos == 0)
4064 goto out;
4065
4066 right_path = ocfs2_new_path(path_root_bh(path),
4067 path_root_el(path));
4068 if (!right_path)
4069 goto out;
4070
4071 status = ocfs2_find_path(inode, right_path, right_cpos);
4072 if (status)
4073 goto out;
4074
4075 new_el = path_leaf_el(right_path);
4076 rec = &new_el->l_recs[0];
4077 if (ocfs2_is_empty_extent(rec)) {
4078 if (le16_to_cpu(new_el->l_next_free_rec) <= 1) {
4079 bh = path_leaf_bh(right_path);
4080 eb = (struct ocfs2_extent_block *)bh->b_data;
4081 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb,
4082 eb);
4083 goto out;
4084 }
4085 rec = &new_el->l_recs[1];
4086 }
4087 }
4088
4089 if (rec) {
Mark Fasheh328d5752007-06-18 10:48:04 -07004090 enum ocfs2_contig_type contig_type;
4091
Mark Fasheh328d5752007-06-18 10:48:04 -07004092 contig_type = ocfs2_extent_contig(inode, rec, split_rec);
4093
4094 if (contig_type == CONTIG_LEFT && ret == CONTIG_RIGHT)
4095 ret = CONTIG_LEFTRIGHT;
4096 else if (ret == CONTIG_NONE)
4097 ret = contig_type;
4098 }
4099
Tao Maad5a4d72008-01-30 14:21:32 +08004100out:
4101 if (left_path)
4102 ocfs2_free_path(left_path);
4103 if (right_path)
4104 ocfs2_free_path(right_path);
4105
Mark Fasheh328d5752007-06-18 10:48:04 -07004106 return ret;
4107}
4108
Mark Fashehdcd05382007-01-16 11:32:23 -08004109static void ocfs2_figure_contig_type(struct inode *inode,
4110 struct ocfs2_insert_type *insert,
4111 struct ocfs2_extent_list *el,
Tao Maca12b7c2008-08-18 17:38:52 +08004112 struct ocfs2_extent_rec *insert_rec,
4113 struct ocfs2_extent_tree *et)
Mark Fashehdcd05382007-01-16 11:32:23 -08004114{
4115 int i;
4116 enum ocfs2_contig_type contig_type = CONTIG_NONE;
4117
Mark Fashehe48edee2007-03-07 16:46:57 -08004118 BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
4119
Mark Fashehdcd05382007-01-16 11:32:23 -08004120 for(i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
4121 contig_type = ocfs2_extent_contig(inode, &el->l_recs[i],
4122 insert_rec);
4123 if (contig_type != CONTIG_NONE) {
4124 insert->ins_contig_index = i;
4125 break;
4126 }
4127 }
4128 insert->ins_contig = contig_type;
Tao Maca12b7c2008-08-18 17:38:52 +08004129
4130 if (insert->ins_contig != CONTIG_NONE) {
4131 struct ocfs2_extent_rec *rec =
4132 &el->l_recs[insert->ins_contig_index];
4133 unsigned int len = le16_to_cpu(rec->e_leaf_clusters) +
4134 le16_to_cpu(insert_rec->e_leaf_clusters);
4135
4136 /*
4137 * Caller might want us to limit the size of extents, don't
4138 * calculate contiguousness if we might exceed that limit.
4139 */
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004140 if (et->et_max_leaf_clusters &&
4141 (len > et->et_max_leaf_clusters))
Tao Maca12b7c2008-08-18 17:38:52 +08004142 insert->ins_contig = CONTIG_NONE;
4143 }
Mark Fashehdcd05382007-01-16 11:32:23 -08004144}
4145
4146/*
4147 * This should only be called against the righmost leaf extent list.
4148 *
4149 * ocfs2_figure_appending_type() will figure out whether we'll have to
4150 * insert at the tail of the rightmost leaf.
4151 *
Tao Mae7d4cb62008-08-18 17:38:44 +08004152 * This should also work against the root extent list for tree's with 0
4153 * depth. If we consider the root extent list to be the rightmost leaf node
Mark Fashehdcd05382007-01-16 11:32:23 -08004154 * then the logic here makes sense.
4155 */
4156static void ocfs2_figure_appending_type(struct ocfs2_insert_type *insert,
4157 struct ocfs2_extent_list *el,
4158 struct ocfs2_extent_rec *insert_rec)
4159{
4160 int i;
4161 u32 cpos = le32_to_cpu(insert_rec->e_cpos);
4162 struct ocfs2_extent_rec *rec;
4163
4164 insert->ins_appending = APPEND_NONE;
4165
Mark Fashehe48edee2007-03-07 16:46:57 -08004166 BUG_ON(le16_to_cpu(el->l_tree_depth) != 0);
Mark Fashehdcd05382007-01-16 11:32:23 -08004167
4168 if (!el->l_next_free_rec)
4169 goto set_tail_append;
4170
4171 if (ocfs2_is_empty_extent(&el->l_recs[0])) {
4172 /* Were all records empty? */
4173 if (le16_to_cpu(el->l_next_free_rec) == 1)
4174 goto set_tail_append;
4175 }
4176
4177 i = le16_to_cpu(el->l_next_free_rec) - 1;
4178 rec = &el->l_recs[i];
4179
Mark Fashehe48edee2007-03-07 16:46:57 -08004180 if (cpos >=
4181 (le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)))
Mark Fashehdcd05382007-01-16 11:32:23 -08004182 goto set_tail_append;
4183
4184 return;
4185
4186set_tail_append:
4187 insert->ins_appending = APPEND_TAIL;
4188}
4189
4190/*
4191 * Helper function called at the begining of an insert.
4192 *
4193 * This computes a few things that are commonly used in the process of
4194 * inserting into the btree:
4195 * - Whether the new extent is contiguous with an existing one.
4196 * - The current tree depth.
4197 * - Whether the insert is an appending one.
4198 * - The total # of free records in the tree.
4199 *
4200 * All of the information is stored on the ocfs2_insert_type
4201 * structure.
4202 */
4203static int ocfs2_figure_insert_type(struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +08004204 struct ocfs2_extent_tree *et,
Mark Fashehdcd05382007-01-16 11:32:23 -08004205 struct buffer_head **last_eb_bh,
4206 struct ocfs2_extent_rec *insert_rec,
Tao Maoc77534f2007-08-28 17:22:33 -07004207 int *free_records,
Mark Fashehdcd05382007-01-16 11:32:23 -08004208 struct ocfs2_insert_type *insert)
4209{
4210 int ret;
Mark Fashehdcd05382007-01-16 11:32:23 -08004211 struct ocfs2_extent_block *eb;
4212 struct ocfs2_extent_list *el;
4213 struct ocfs2_path *path = NULL;
4214 struct buffer_head *bh = NULL;
4215
Mark Fasheh328d5752007-06-18 10:48:04 -07004216 insert->ins_split = SPLIT_NONE;
4217
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004218 el = et->et_root_el;
Mark Fashehdcd05382007-01-16 11:32:23 -08004219 insert->ins_tree_depth = le16_to_cpu(el->l_tree_depth);
4220
4221 if (el->l_tree_depth) {
4222 /*
4223 * If we have tree depth, we read in the
4224 * rightmost extent block ahead of time as
4225 * ocfs2_figure_insert_type() and ocfs2_add_branch()
4226 * may want it later.
4227 */
4228 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
Joel Becker35dc0aa2008-08-20 16:25:06 -07004229 ocfs2_et_get_last_eb_blk(et), &bh,
Mark Fashehdcd05382007-01-16 11:32:23 -08004230 OCFS2_BH_CACHED, inode);
4231 if (ret) {
4232 mlog_exit(ret);
4233 goto out;
4234 }
4235 eb = (struct ocfs2_extent_block *) bh->b_data;
4236 el = &eb->h_list;
4237 }
4238
4239 /*
4240 * Unless we have a contiguous insert, we'll need to know if
4241 * there is room left in our allocation tree for another
4242 * extent record.
4243 *
4244 * XXX: This test is simplistic, we can search for empty
4245 * extent records too.
4246 */
Tao Maoc77534f2007-08-28 17:22:33 -07004247 *free_records = le16_to_cpu(el->l_count) -
Mark Fashehdcd05382007-01-16 11:32:23 -08004248 le16_to_cpu(el->l_next_free_rec);
4249
4250 if (!insert->ins_tree_depth) {
Tao Maca12b7c2008-08-18 17:38:52 +08004251 ocfs2_figure_contig_type(inode, insert, el, insert_rec, et);
Mark Fashehdcd05382007-01-16 11:32:23 -08004252 ocfs2_figure_appending_type(insert, el, insert_rec);
4253 return 0;
4254 }
4255
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004256 path = ocfs2_new_path(et->et_root_bh, et->et_root_el);
Mark Fashehdcd05382007-01-16 11:32:23 -08004257 if (!path) {
4258 ret = -ENOMEM;
4259 mlog_errno(ret);
4260 goto out;
4261 }
4262
4263 /*
4264 * In the case that we're inserting past what the tree
4265 * currently accounts for, ocfs2_find_path() will return for
4266 * us the rightmost tree path. This is accounted for below in
4267 * the appending code.
4268 */
4269 ret = ocfs2_find_path(inode, path, le32_to_cpu(insert_rec->e_cpos));
4270 if (ret) {
4271 mlog_errno(ret);
4272 goto out;
4273 }
4274
4275 el = path_leaf_el(path);
4276
4277 /*
4278 * Now that we have the path, there's two things we want to determine:
4279 * 1) Contiguousness (also set contig_index if this is so)
4280 *
4281 * 2) Are we doing an append? We can trivially break this up
4282 * into two types of appends: simple record append, or a
4283 * rotate inside the tail leaf.
4284 */
Tao Maca12b7c2008-08-18 17:38:52 +08004285 ocfs2_figure_contig_type(inode, insert, el, insert_rec, et);
Mark Fashehdcd05382007-01-16 11:32:23 -08004286
4287 /*
4288 * The insert code isn't quite ready to deal with all cases of
4289 * left contiguousness. Specifically, if it's an insert into
4290 * the 1st record in a leaf, it will require the adjustment of
Mark Fashehe48edee2007-03-07 16:46:57 -08004291 * cluster count on the last record of the path directly to it's
Mark Fashehdcd05382007-01-16 11:32:23 -08004292 * left. For now, just catch that case and fool the layers
4293 * above us. This works just fine for tree_depth == 0, which
4294 * is why we allow that above.
4295 */
4296 if (insert->ins_contig == CONTIG_LEFT &&
4297 insert->ins_contig_index == 0)
4298 insert->ins_contig = CONTIG_NONE;
4299
4300 /*
4301 * Ok, so we can simply compare against last_eb to figure out
4302 * whether the path doesn't exist. This will only happen in
4303 * the case that we're doing a tail append, so maybe we can
4304 * take advantage of that information somehow.
4305 */
Joel Becker35dc0aa2008-08-20 16:25:06 -07004306 if (ocfs2_et_get_last_eb_blk(et) ==
Tao Mae7d4cb62008-08-18 17:38:44 +08004307 path_leaf_bh(path)->b_blocknr) {
Mark Fashehdcd05382007-01-16 11:32:23 -08004308 /*
4309 * Ok, ocfs2_find_path() returned us the rightmost
4310 * tree path. This might be an appending insert. There are
4311 * two cases:
4312 * 1) We're doing a true append at the tail:
4313 * -This might even be off the end of the leaf
4314 * 2) We're "appending" by rotating in the tail
4315 */
4316 ocfs2_figure_appending_type(insert, el, insert_rec);
4317 }
4318
4319out:
4320 ocfs2_free_path(path);
4321
4322 if (ret == 0)
4323 *last_eb_bh = bh;
4324 else
4325 brelse(bh);
4326 return ret;
4327}
4328
4329/*
4330 * Insert an extent into an inode btree.
4331 *
4332 * The caller needs to update fe->i_clusters
4333 */
Tao Maf56654c2008-08-18 17:38:48 +08004334static int ocfs2_insert_extent(struct ocfs2_super *osb,
4335 handle_t *handle,
4336 struct inode *inode,
4337 struct buffer_head *root_bh,
4338 u32 cpos,
4339 u64 start_blk,
4340 u32 new_clusters,
4341 u8 flags,
4342 struct ocfs2_alloc_context *meta_ac,
4343 struct ocfs2_extent_tree *et)
Mark Fashehccd979b2005-12-15 14:31:24 -08004344{
Mark Fashehc3afcbb2007-05-29 14:28:51 -07004345 int status;
Tao Maoc77534f2007-08-28 17:22:33 -07004346 int uninitialized_var(free_records);
Mark Fashehccd979b2005-12-15 14:31:24 -08004347 struct buffer_head *last_eb_bh = NULL;
Mark Fashehdcd05382007-01-16 11:32:23 -08004348 struct ocfs2_insert_type insert = {0, };
4349 struct ocfs2_extent_rec rec;
Mark Fashehccd979b2005-12-15 14:31:24 -08004350
Mark Fasheh1afc32b2007-09-07 14:46:51 -07004351 BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL);
4352
Mark Fashehdcd05382007-01-16 11:32:23 -08004353 mlog(0, "add %u clusters at position %u to inode %llu\n",
4354 new_clusters, cpos, (unsigned long long)OCFS2_I(inode)->ip_blkno);
Mark Fashehccd979b2005-12-15 14:31:24 -08004355
Mark Fashehdcd05382007-01-16 11:32:23 -08004356 mlog_bug_on_msg(!ocfs2_sparse_alloc(osb) &&
4357 (OCFS2_I(inode)->ip_clusters != cpos),
4358 "Device %s, asking for sparse allocation: inode %llu, "
4359 "cpos %u, clusters %u\n",
4360 osb->dev_str,
4361 (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos,
4362 OCFS2_I(inode)->ip_clusters);
Mark Fashehccd979b2005-12-15 14:31:24 -08004363
Mark Fashehe48edee2007-03-07 16:46:57 -08004364 memset(&rec, 0, sizeof(rec));
Mark Fashehdcd05382007-01-16 11:32:23 -08004365 rec.e_cpos = cpu_to_le32(cpos);
4366 rec.e_blkno = cpu_to_le64(start_blk);
Mark Fashehe48edee2007-03-07 16:46:57 -08004367 rec.e_leaf_clusters = cpu_to_le16(new_clusters);
Mark Fasheh2ae99a62007-03-09 16:43:28 -08004368 rec.e_flags = flags;
Mark Fashehccd979b2005-12-15 14:31:24 -08004369
Tao Mae7d4cb62008-08-18 17:38:44 +08004370 status = ocfs2_figure_insert_type(inode, et, &last_eb_bh, &rec,
Tao Maoc77534f2007-08-28 17:22:33 -07004371 &free_records, &insert);
Mark Fashehdcd05382007-01-16 11:32:23 -08004372 if (status < 0) {
4373 mlog_errno(status);
4374 goto bail;
Mark Fashehccd979b2005-12-15 14:31:24 -08004375 }
4376
Mark Fashehdcd05382007-01-16 11:32:23 -08004377 mlog(0, "Insert.appending: %u, Insert.Contig: %u, "
4378 "Insert.contig_index: %d, Insert.free_records: %d, "
4379 "Insert.tree_depth: %d\n",
4380 insert.ins_appending, insert.ins_contig, insert.ins_contig_index,
Tao Maoc77534f2007-08-28 17:22:33 -07004381 free_records, insert.ins_tree_depth);
Mark Fashehccd979b2005-12-15 14:31:24 -08004382
Tao Maoc77534f2007-08-28 17:22:33 -07004383 if (insert.ins_contig == CONTIG_NONE && free_records == 0) {
Tao Mae7d4cb62008-08-18 17:38:44 +08004384 status = ocfs2_grow_tree(inode, handle, et,
Mark Fasheh328d5752007-06-18 10:48:04 -07004385 &insert.ins_tree_depth, &last_eb_bh,
Mark Fashehc3afcbb2007-05-29 14:28:51 -07004386 meta_ac);
4387 if (status) {
Mark Fashehccd979b2005-12-15 14:31:24 -08004388 mlog_errno(status);
4389 goto bail;
4390 }
Mark Fashehccd979b2005-12-15 14:31:24 -08004391 }
4392
Mark Fashehdcd05382007-01-16 11:32:23 -08004393 /* Finally, we can add clusters. This might rotate the tree for us. */
Tao Mae7d4cb62008-08-18 17:38:44 +08004394 status = ocfs2_do_insert_extent(inode, handle, et, &rec, &insert);
Mark Fashehccd979b2005-12-15 14:31:24 -08004395 if (status < 0)
4396 mlog_errno(status);
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004397 else if (et->et_type == OCFS2_DINODE_EXTENT)
Mark Fasheh83418972007-04-23 18:53:12 -07004398 ocfs2_extent_map_insert_rec(inode, &rec);
Mark Fashehccd979b2005-12-15 14:31:24 -08004399
4400bail:
Mark Fashehccd979b2005-12-15 14:31:24 -08004401 if (last_eb_bh)
4402 brelse(last_eb_bh);
4403
Tao Maf56654c2008-08-18 17:38:48 +08004404 mlog_exit(status);
4405 return status;
4406}
4407
4408int ocfs2_dinode_insert_extent(struct ocfs2_super *osb,
4409 handle_t *handle,
4410 struct inode *inode,
4411 struct buffer_head *root_bh,
4412 u32 cpos,
4413 u64 start_blk,
4414 u32 new_clusters,
4415 u8 flags,
4416 struct ocfs2_alloc_context *meta_ac)
4417{
4418 int status;
Joel Beckerdc0ce612008-08-20 16:48:35 -07004419 struct ocfs2_extent_tree et;
Tao Maf56654c2008-08-18 17:38:48 +08004420
Joel Beckerdc0ce612008-08-20 16:48:35 -07004421 ocfs2_get_extent_tree(&et, inode, root_bh, OCFS2_DINODE_EXTENT,
4422 NULL);
Tao Maf56654c2008-08-18 17:38:48 +08004423 status = ocfs2_insert_extent(osb, handle, inode, root_bh,
4424 cpos, start_blk, new_clusters,
Joel Beckerdc0ce612008-08-20 16:48:35 -07004425 flags, meta_ac, &et);
4426 ocfs2_put_extent_tree(&et);
Tao Maf56654c2008-08-18 17:38:48 +08004427
Tao Maf56654c2008-08-18 17:38:48 +08004428 return status;
4429}
4430
4431int ocfs2_xattr_value_insert_extent(struct ocfs2_super *osb,
4432 handle_t *handle,
4433 struct inode *inode,
4434 struct buffer_head *root_bh,
4435 u32 cpos,
4436 u64 start_blk,
4437 u32 new_clusters,
4438 u8 flags,
4439 struct ocfs2_alloc_context *meta_ac,
Joel Beckerea5efa12008-08-20 16:57:27 -07004440 void *obj)
Tao Maf56654c2008-08-18 17:38:48 +08004441{
4442 int status;
Joel Beckerdc0ce612008-08-20 16:48:35 -07004443 struct ocfs2_extent_tree et;
Tao Maf56654c2008-08-18 17:38:48 +08004444
Joel Beckerdc0ce612008-08-20 16:48:35 -07004445 ocfs2_get_extent_tree(&et, inode, root_bh,
Joel Beckerea5efa12008-08-20 16:57:27 -07004446 OCFS2_XATTR_VALUE_EXTENT, obj);
Tao Maf56654c2008-08-18 17:38:48 +08004447 status = ocfs2_insert_extent(osb, handle, inode, root_bh,
4448 cpos, start_blk, new_clusters,
Joel Beckerdc0ce612008-08-20 16:48:35 -07004449 flags, meta_ac, &et);
4450 ocfs2_put_extent_tree(&et);
Tao Maf56654c2008-08-18 17:38:48 +08004451
Mark Fashehccd979b2005-12-15 14:31:24 -08004452 return status;
4453}
4454
Tao Maba492612008-08-18 17:38:49 +08004455int ocfs2_xattr_tree_insert_extent(struct ocfs2_super *osb,
4456 handle_t *handle,
4457 struct inode *inode,
4458 struct buffer_head *root_bh,
4459 u32 cpos,
4460 u64 start_blk,
4461 u32 new_clusters,
4462 u8 flags,
4463 struct ocfs2_alloc_context *meta_ac)
4464{
4465 int status;
Joel Beckerdc0ce612008-08-20 16:48:35 -07004466 struct ocfs2_extent_tree et;
Tao Maba492612008-08-18 17:38:49 +08004467
Joel Beckerdc0ce612008-08-20 16:48:35 -07004468 ocfs2_get_extent_tree(&et, inode, root_bh, OCFS2_XATTR_TREE_EXTENT,
4469 NULL);
Tao Maba492612008-08-18 17:38:49 +08004470 status = ocfs2_insert_extent(osb, handle, inode, root_bh,
4471 cpos, start_blk, new_clusters,
Joel Beckerdc0ce612008-08-20 16:48:35 -07004472 flags, meta_ac, &et);
4473 ocfs2_put_extent_tree(&et);
Tao Maba492612008-08-18 17:38:49 +08004474
Tao Maba492612008-08-18 17:38:49 +08004475 return status;
4476}
4477
Tao Ma0eb8d472008-08-18 17:38:45 +08004478/*
4479 * Allcate and add clusters into the extent b-tree.
4480 * The new clusters(clusters_to_add) will be inserted at logical_offset.
4481 * The extent b-tree's root is root_el and it should be in root_bh, and
4482 * it is not limited to the file storage. Any extent tree can use this
4483 * function if it implements the proper ocfs2_extent_tree.
4484 */
4485int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb,
4486 struct inode *inode,
4487 u32 *logical_offset,
4488 u32 clusters_to_add,
4489 int mark_unwritten,
4490 struct buffer_head *root_bh,
4491 struct ocfs2_extent_list *root_el,
4492 handle_t *handle,
4493 struct ocfs2_alloc_context *data_ac,
4494 struct ocfs2_alloc_context *meta_ac,
4495 enum ocfs2_alloc_restarted *reason_ret,
Tao Maf56654c2008-08-18 17:38:48 +08004496 enum ocfs2_extent_tree_type type,
Joel Beckerea5efa12008-08-20 16:57:27 -07004497 void *obj)
Tao Ma0eb8d472008-08-18 17:38:45 +08004498{
4499 int status = 0;
4500 int free_extents;
4501 enum ocfs2_alloc_restarted reason = RESTART_NONE;
4502 u32 bit_off, num_bits;
4503 u64 block;
4504 u8 flags = 0;
4505
4506 BUG_ON(!clusters_to_add);
4507
4508 if (mark_unwritten)
4509 flags = OCFS2_EXT_UNWRITTEN;
4510
Tao Maf56654c2008-08-18 17:38:48 +08004511 free_extents = ocfs2_num_free_extents(osb, inode, root_bh, type,
Joel Beckerea5efa12008-08-20 16:57:27 -07004512 obj);
Tao Ma0eb8d472008-08-18 17:38:45 +08004513 if (free_extents < 0) {
4514 status = free_extents;
4515 mlog_errno(status);
4516 goto leave;
4517 }
4518
4519 /* there are two cases which could cause us to EAGAIN in the
4520 * we-need-more-metadata case:
4521 * 1) we haven't reserved *any*
4522 * 2) we are so fragmented, we've needed to add metadata too
4523 * many times. */
4524 if (!free_extents && !meta_ac) {
4525 mlog(0, "we haven't reserved any metadata!\n");
4526 status = -EAGAIN;
4527 reason = RESTART_META;
4528 goto leave;
4529 } else if ((!free_extents)
4530 && (ocfs2_alloc_context_bits_left(meta_ac)
4531 < ocfs2_extend_meta_needed(root_el))) {
4532 mlog(0, "filesystem is really fragmented...\n");
4533 status = -EAGAIN;
4534 reason = RESTART_META;
4535 goto leave;
4536 }
4537
4538 status = __ocfs2_claim_clusters(osb, handle, data_ac, 1,
4539 clusters_to_add, &bit_off, &num_bits);
4540 if (status < 0) {
4541 if (status != -ENOSPC)
4542 mlog_errno(status);
4543 goto leave;
4544 }
4545
4546 BUG_ON(num_bits > clusters_to_add);
4547
4548 /* reserve our write early -- insert_extent may update the inode */
4549 status = ocfs2_journal_access(handle, inode, root_bh,
4550 OCFS2_JOURNAL_ACCESS_WRITE);
4551 if (status < 0) {
4552 mlog_errno(status);
4553 goto leave;
4554 }
4555
4556 block = ocfs2_clusters_to_blocks(osb->sb, bit_off);
4557 mlog(0, "Allocating %u clusters at block %u for inode %llu\n",
4558 num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno);
Tao Maf56654c2008-08-18 17:38:48 +08004559 if (type == OCFS2_DINODE_EXTENT)
4560 status = ocfs2_dinode_insert_extent(osb, handle, inode, root_bh,
4561 *logical_offset, block,
4562 num_bits, flags, meta_ac);
Tao Maba492612008-08-18 17:38:49 +08004563 else if (type == OCFS2_XATTR_TREE_EXTENT)
4564 status = ocfs2_xattr_tree_insert_extent(osb, handle,
4565 inode, root_bh,
4566 *logical_offset,
4567 block, num_bits, flags,
4568 meta_ac);
Tao Maf56654c2008-08-18 17:38:48 +08004569 else
4570 status = ocfs2_xattr_value_insert_extent(osb, handle,
4571 inode, root_bh,
4572 *logical_offset,
4573 block, num_bits, flags,
Joel Beckerea5efa12008-08-20 16:57:27 -07004574 meta_ac, obj);
Tao Ma0eb8d472008-08-18 17:38:45 +08004575 if (status < 0) {
4576 mlog_errno(status);
4577 goto leave;
4578 }
4579
4580 status = ocfs2_journal_dirty(handle, root_bh);
4581 if (status < 0) {
4582 mlog_errno(status);
4583 goto leave;
4584 }
4585
4586 clusters_to_add -= num_bits;
4587 *logical_offset += num_bits;
4588
4589 if (clusters_to_add) {
4590 mlog(0, "need to alloc once more, wanted = %u\n",
4591 clusters_to_add);
4592 status = -EAGAIN;
4593 reason = RESTART_TRANS;
4594 }
4595
4596leave:
4597 mlog_exit(status);
4598 if (reason_ret)
4599 *reason_ret = reason;
4600 return status;
4601}
4602
Mark Fasheh328d5752007-06-18 10:48:04 -07004603static void ocfs2_make_right_split_rec(struct super_block *sb,
4604 struct ocfs2_extent_rec *split_rec,
4605 u32 cpos,
4606 struct ocfs2_extent_rec *rec)
4607{
4608 u32 rec_cpos = le32_to_cpu(rec->e_cpos);
4609 u32 rec_range = rec_cpos + le16_to_cpu(rec->e_leaf_clusters);
4610
4611 memset(split_rec, 0, sizeof(struct ocfs2_extent_rec));
4612
4613 split_rec->e_cpos = cpu_to_le32(cpos);
4614 split_rec->e_leaf_clusters = cpu_to_le16(rec_range - cpos);
4615
4616 split_rec->e_blkno = rec->e_blkno;
4617 le64_add_cpu(&split_rec->e_blkno,
4618 ocfs2_clusters_to_blocks(sb, cpos - rec_cpos));
4619
4620 split_rec->e_flags = rec->e_flags;
4621}
4622
4623static int ocfs2_split_and_insert(struct inode *inode,
4624 handle_t *handle,
4625 struct ocfs2_path *path,
Tao Mae7d4cb62008-08-18 17:38:44 +08004626 struct ocfs2_extent_tree *et,
Mark Fasheh328d5752007-06-18 10:48:04 -07004627 struct buffer_head **last_eb_bh,
4628 int split_index,
4629 struct ocfs2_extent_rec *orig_split_rec,
4630 struct ocfs2_alloc_context *meta_ac)
4631{
4632 int ret = 0, depth;
4633 unsigned int insert_range, rec_range, do_leftright = 0;
4634 struct ocfs2_extent_rec tmprec;
4635 struct ocfs2_extent_list *rightmost_el;
4636 struct ocfs2_extent_rec rec;
4637 struct ocfs2_extent_rec split_rec = *orig_split_rec;
4638 struct ocfs2_insert_type insert;
4639 struct ocfs2_extent_block *eb;
Mark Fasheh328d5752007-06-18 10:48:04 -07004640
4641leftright:
4642 /*
4643 * Store a copy of the record on the stack - it might move
4644 * around as the tree is manipulated below.
4645 */
4646 rec = path_leaf_el(path)->l_recs[split_index];
4647
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004648 rightmost_el = et->et_root_el;
Mark Fasheh328d5752007-06-18 10:48:04 -07004649
4650 depth = le16_to_cpu(rightmost_el->l_tree_depth);
4651 if (depth) {
4652 BUG_ON(!(*last_eb_bh));
4653 eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
4654 rightmost_el = &eb->h_list;
4655 }
4656
4657 if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
4658 le16_to_cpu(rightmost_el->l_count)) {
Tao Mae7d4cb62008-08-18 17:38:44 +08004659 ret = ocfs2_grow_tree(inode, handle, et,
4660 &depth, last_eb_bh, meta_ac);
Mark Fasheh328d5752007-06-18 10:48:04 -07004661 if (ret) {
4662 mlog_errno(ret);
4663 goto out;
4664 }
Mark Fasheh328d5752007-06-18 10:48:04 -07004665 }
4666
4667 memset(&insert, 0, sizeof(struct ocfs2_insert_type));
4668 insert.ins_appending = APPEND_NONE;
4669 insert.ins_contig = CONTIG_NONE;
Mark Fasheh328d5752007-06-18 10:48:04 -07004670 insert.ins_tree_depth = depth;
4671
4672 insert_range = le32_to_cpu(split_rec.e_cpos) +
4673 le16_to_cpu(split_rec.e_leaf_clusters);
4674 rec_range = le32_to_cpu(rec.e_cpos) +
4675 le16_to_cpu(rec.e_leaf_clusters);
4676
4677 if (split_rec.e_cpos == rec.e_cpos) {
4678 insert.ins_split = SPLIT_LEFT;
4679 } else if (insert_range == rec_range) {
4680 insert.ins_split = SPLIT_RIGHT;
4681 } else {
4682 /*
4683 * Left/right split. We fake this as a right split
4684 * first and then make a second pass as a left split.
4685 */
4686 insert.ins_split = SPLIT_RIGHT;
4687
4688 ocfs2_make_right_split_rec(inode->i_sb, &tmprec, insert_range,
4689 &rec);
4690
4691 split_rec = tmprec;
4692
4693 BUG_ON(do_leftright);
4694 do_leftright = 1;
4695 }
4696
Tao Mae7d4cb62008-08-18 17:38:44 +08004697 ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert);
Mark Fasheh328d5752007-06-18 10:48:04 -07004698 if (ret) {
4699 mlog_errno(ret);
4700 goto out;
4701 }
4702
4703 if (do_leftright == 1) {
4704 u32 cpos;
4705 struct ocfs2_extent_list *el;
4706
4707 do_leftright++;
4708 split_rec = *orig_split_rec;
4709
4710 ocfs2_reinit_path(path, 1);
4711
4712 cpos = le32_to_cpu(split_rec.e_cpos);
4713 ret = ocfs2_find_path(inode, path, cpos);
4714 if (ret) {
4715 mlog_errno(ret);
4716 goto out;
4717 }
4718
4719 el = path_leaf_el(path);
4720 split_index = ocfs2_search_extent_list(el, cpos);
4721 goto leftright;
4722 }
4723out:
4724
4725 return ret;
4726}
4727
4728/*
4729 * Mark part or all of the extent record at split_index in the leaf
4730 * pointed to by path as written. This removes the unwritten
4731 * extent flag.
4732 *
4733 * Care is taken to handle contiguousness so as to not grow the tree.
4734 *
4735 * meta_ac is not strictly necessary - we only truly need it if growth
4736 * of the tree is required. All other cases will degrade into a less
4737 * optimal tree layout.
4738 *
Tao Mae7d4cb62008-08-18 17:38:44 +08004739 * last_eb_bh should be the rightmost leaf block for any extent
4740 * btree. Since a split may grow the tree or a merge might shrink it,
4741 * the caller cannot trust the contents of that buffer after this call.
Mark Fasheh328d5752007-06-18 10:48:04 -07004742 *
4743 * This code is optimized for readability - several passes might be
4744 * made over certain portions of the tree. All of those blocks will
4745 * have been brought into cache (and pinned via the journal), so the
4746 * extra overhead is not expressed in terms of disk reads.
4747 */
4748static int __ocfs2_mark_extent_written(struct inode *inode,
Tao Mae7d4cb62008-08-18 17:38:44 +08004749 struct ocfs2_extent_tree *et,
Mark Fasheh328d5752007-06-18 10:48:04 -07004750 handle_t *handle,
4751 struct ocfs2_path *path,
4752 int split_index,
4753 struct ocfs2_extent_rec *split_rec,
4754 struct ocfs2_alloc_context *meta_ac,
4755 struct ocfs2_cached_dealloc_ctxt *dealloc)
4756{
4757 int ret = 0;
4758 struct ocfs2_extent_list *el = path_leaf_el(path);
Mark Fashehe8aed342007-12-03 16:43:01 -08004759 struct buffer_head *last_eb_bh = NULL;
Mark Fasheh328d5752007-06-18 10:48:04 -07004760 struct ocfs2_extent_rec *rec = &el->l_recs[split_index];
4761 struct ocfs2_merge_ctxt ctxt;
4762 struct ocfs2_extent_list *rightmost_el;
4763
Roel Kluin3cf0c502007-10-27 00:20:36 +02004764 if (!(rec->e_flags & OCFS2_EXT_UNWRITTEN)) {
Mark Fasheh328d5752007-06-18 10:48:04 -07004765 ret = -EIO;
4766 mlog_errno(ret);
4767 goto out;
4768 }
4769
4770 if (le32_to_cpu(rec->e_cpos) > le32_to_cpu(split_rec->e_cpos) ||
4771 ((le32_to_cpu(rec->e_cpos) + le16_to_cpu(rec->e_leaf_clusters)) <
4772 (le32_to_cpu(split_rec->e_cpos) + le16_to_cpu(split_rec->e_leaf_clusters)))) {
4773 ret = -EIO;
4774 mlog_errno(ret);
4775 goto out;
4776 }
4777
Tao Maad5a4d72008-01-30 14:21:32 +08004778 ctxt.c_contig_type = ocfs2_figure_merge_contig_type(inode, path, el,
Mark Fasheh328d5752007-06-18 10:48:04 -07004779 split_index,
4780 split_rec);
4781
4782 /*
4783 * The core merge / split code wants to know how much room is
4784 * left in this inodes allocation tree, so we pass the
4785 * rightmost extent list.
4786 */
4787 if (path->p_tree_depth) {
4788 struct ocfs2_extent_block *eb;
Mark Fasheh328d5752007-06-18 10:48:04 -07004789
4790 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
Joel Becker35dc0aa2008-08-20 16:25:06 -07004791 ocfs2_et_get_last_eb_blk(et),
Mark Fasheh328d5752007-06-18 10:48:04 -07004792 &last_eb_bh, OCFS2_BH_CACHED, inode);
4793 if (ret) {
4794 mlog_exit(ret);
4795 goto out;
4796 }
4797
4798 eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
4799 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
4800 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
4801 ret = -EROFS;
4802 goto out;
4803 }
4804
4805 rightmost_el = &eb->h_list;
4806 } else
4807 rightmost_el = path_root_el(path);
4808
Mark Fasheh328d5752007-06-18 10:48:04 -07004809 if (rec->e_cpos == split_rec->e_cpos &&
4810 rec->e_leaf_clusters == split_rec->e_leaf_clusters)
4811 ctxt.c_split_covers_rec = 1;
4812 else
4813 ctxt.c_split_covers_rec = 0;
4814
4815 ctxt.c_has_empty_extent = ocfs2_is_empty_extent(&el->l_recs[0]);
4816
Mark Fasheh015452b2007-09-12 10:21:22 -07004817 mlog(0, "index: %d, contig: %u, has_empty: %u, split_covers: %u\n",
4818 split_index, ctxt.c_contig_type, ctxt.c_has_empty_extent,
4819 ctxt.c_split_covers_rec);
Mark Fasheh328d5752007-06-18 10:48:04 -07004820
4821 if (ctxt.c_contig_type == CONTIG_NONE) {
4822 if (ctxt.c_split_covers_rec)
4823 el->l_recs[split_index] = *split_rec;
4824 else
Tao Mae7d4cb62008-08-18 17:38:44 +08004825 ret = ocfs2_split_and_insert(inode, handle, path, et,
Mark Fasheh328d5752007-06-18 10:48:04 -07004826 &last_eb_bh, split_index,
4827 split_rec, meta_ac);
4828 if (ret)
4829 mlog_errno(ret);
4830 } else {
4831 ret = ocfs2_try_to_merge_extent(inode, handle, path,
4832 split_index, split_rec,
Tao Mae7d4cb62008-08-18 17:38:44 +08004833 dealloc, &ctxt, et);
Mark Fasheh328d5752007-06-18 10:48:04 -07004834 if (ret)
4835 mlog_errno(ret);
4836 }
4837
Mark Fasheh328d5752007-06-18 10:48:04 -07004838out:
4839 brelse(last_eb_bh);
4840 return ret;
4841}
4842
4843/*
4844 * Mark the already-existing extent at cpos as written for len clusters.
4845 *
4846 * If the existing extent is larger than the request, initiate a
4847 * split. An attempt will be made at merging with adjacent extents.
4848 *
4849 * The caller is responsible for passing down meta_ac if we'll need it.
4850 */
Tao Mae7d4cb62008-08-18 17:38:44 +08004851int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *root_bh,
Mark Fasheh328d5752007-06-18 10:48:04 -07004852 handle_t *handle, u32 cpos, u32 len, u32 phys,
4853 struct ocfs2_alloc_context *meta_ac,
Tao Mae7d4cb62008-08-18 17:38:44 +08004854 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Maf56654c2008-08-18 17:38:48 +08004855 enum ocfs2_extent_tree_type et_type,
Joel Beckerea5efa12008-08-20 16:57:27 -07004856 void *obj)
Mark Fasheh328d5752007-06-18 10:48:04 -07004857{
4858 int ret, index;
4859 u64 start_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys);
4860 struct ocfs2_extent_rec split_rec;
4861 struct ocfs2_path *left_path = NULL;
4862 struct ocfs2_extent_list *el;
Joel Beckerdc0ce612008-08-20 16:48:35 -07004863 struct ocfs2_extent_tree et;
Mark Fasheh328d5752007-06-18 10:48:04 -07004864
4865 mlog(0, "Inode %lu cpos %u, len %u, phys %u (%llu)\n",
4866 inode->i_ino, cpos, len, phys, (unsigned long long)start_blkno);
4867
Joel Beckerea5efa12008-08-20 16:57:27 -07004868 ocfs2_get_extent_tree(&et, inode, root_bh, et_type, obj);
Joel Beckerdc0ce612008-08-20 16:48:35 -07004869
Mark Fasheh328d5752007-06-18 10:48:04 -07004870 if (!ocfs2_writes_unwritten_extents(OCFS2_SB(inode->i_sb))) {
4871 ocfs2_error(inode->i_sb, "Inode %llu has unwritten extents "
4872 "that are being written to, but the feature bit "
4873 "is not set in the super block.",
4874 (unsigned long long)OCFS2_I(inode)->ip_blkno);
4875 ret = -EROFS;
4876 goto out;
4877 }
4878
4879 /*
4880 * XXX: This should be fixed up so that we just re-insert the
4881 * next extent records.
4882 */
Tao Mae7d4cb62008-08-18 17:38:44 +08004883 if (et_type == OCFS2_DINODE_EXTENT)
4884 ocfs2_extent_map_trunc(inode, 0);
Mark Fasheh328d5752007-06-18 10:48:04 -07004885
Joel Beckerdc0ce612008-08-20 16:48:35 -07004886 left_path = ocfs2_new_path(et.et_root_bh, et.et_root_el);
Mark Fasheh328d5752007-06-18 10:48:04 -07004887 if (!left_path) {
4888 ret = -ENOMEM;
4889 mlog_errno(ret);
4890 goto out;
4891 }
4892
4893 ret = ocfs2_find_path(inode, left_path, cpos);
4894 if (ret) {
4895 mlog_errno(ret);
4896 goto out;
4897 }
4898 el = path_leaf_el(left_path);
4899
4900 index = ocfs2_search_extent_list(el, cpos);
4901 if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
4902 ocfs2_error(inode->i_sb,
4903 "Inode %llu has an extent at cpos %u which can no "
4904 "longer be found.\n",
4905 (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
4906 ret = -EROFS;
4907 goto out;
4908 }
4909
4910 memset(&split_rec, 0, sizeof(struct ocfs2_extent_rec));
4911 split_rec.e_cpos = cpu_to_le32(cpos);
4912 split_rec.e_leaf_clusters = cpu_to_le16(len);
4913 split_rec.e_blkno = cpu_to_le64(start_blkno);
4914 split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags;
4915 split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN;
4916
Joel Beckerdc0ce612008-08-20 16:48:35 -07004917 ret = __ocfs2_mark_extent_written(inode, &et, handle, left_path,
Tao Mae7d4cb62008-08-18 17:38:44 +08004918 index, &split_rec, meta_ac,
4919 dealloc);
Mark Fasheh328d5752007-06-18 10:48:04 -07004920 if (ret)
4921 mlog_errno(ret);
4922
4923out:
4924 ocfs2_free_path(left_path);
Joel Beckerdc0ce612008-08-20 16:48:35 -07004925 ocfs2_put_extent_tree(&et);
Mark Fasheh328d5752007-06-18 10:48:04 -07004926 return ret;
4927}
4928
Tao Mae7d4cb62008-08-18 17:38:44 +08004929static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et,
Mark Fashehd0c7d702007-07-03 13:27:22 -07004930 handle_t *handle, struct ocfs2_path *path,
4931 int index, u32 new_range,
4932 struct ocfs2_alloc_context *meta_ac)
4933{
4934 int ret, depth, credits = handle->h_buffer_credits;
Mark Fashehd0c7d702007-07-03 13:27:22 -07004935 struct buffer_head *last_eb_bh = NULL;
4936 struct ocfs2_extent_block *eb;
4937 struct ocfs2_extent_list *rightmost_el, *el;
4938 struct ocfs2_extent_rec split_rec;
4939 struct ocfs2_extent_rec *rec;
4940 struct ocfs2_insert_type insert;
4941
4942 /*
4943 * Setup the record to split before we grow the tree.
4944 */
4945 el = path_leaf_el(path);
4946 rec = &el->l_recs[index];
4947 ocfs2_make_right_split_rec(inode->i_sb, &split_rec, new_range, rec);
4948
4949 depth = path->p_tree_depth;
4950 if (depth > 0) {
4951 ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
Joel Becker35dc0aa2008-08-20 16:25:06 -07004952 ocfs2_et_get_last_eb_blk(et),
Mark Fashehd0c7d702007-07-03 13:27:22 -07004953 &last_eb_bh, OCFS2_BH_CACHED, inode);
4954 if (ret < 0) {
4955 mlog_errno(ret);
4956 goto out;
4957 }
4958
4959 eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
4960 rightmost_el = &eb->h_list;
4961 } else
4962 rightmost_el = path_leaf_el(path);
4963
Tao Ma811f9332008-08-18 17:38:43 +08004964 credits += path->p_tree_depth +
Joel Beckerce1d9ea2008-08-20 16:30:07 -07004965 ocfs2_extend_meta_needed(et->et_root_el);
Mark Fashehd0c7d702007-07-03 13:27:22 -07004966 ret = ocfs2_extend_trans(handle, credits);
4967 if (ret) {
4968 mlog_errno(ret);
4969 goto out;
4970 }
4971
4972 if (le16_to_cpu(rightmost_el->l_next_free_rec) ==
4973 le16_to_cpu(rightmost_el->l_count)) {
Tao Mae7d4cb62008-08-18 17:38:44 +08004974 ret = ocfs2_grow_tree(inode, handle, et, &depth, &last_eb_bh,
Mark Fashehd0c7d702007-07-03 13:27:22 -07004975 meta_ac);
4976 if (ret) {
4977 mlog_errno(ret);
4978 goto out;
4979 }
Mark Fashehd0c7d702007-07-03 13:27:22 -07004980 }
4981
4982 memset(&insert, 0, sizeof(struct ocfs2_insert_type));
4983 insert.ins_appending = APPEND_NONE;
4984 insert.ins_contig = CONTIG_NONE;
4985 insert.ins_split = SPLIT_RIGHT;
Mark Fashehd0c7d702007-07-03 13:27:22 -07004986 insert.ins_tree_depth = depth;
4987
Tao Mae7d4cb62008-08-18 17:38:44 +08004988 ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert);
Mark Fashehd0c7d702007-07-03 13:27:22 -07004989 if (ret)
4990 mlog_errno(ret);
4991
4992out:
4993 brelse(last_eb_bh);
4994 return ret;
4995}
4996
4997static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle,
4998 struct ocfs2_path *path, int index,
4999 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Mae7d4cb62008-08-18 17:38:44 +08005000 u32 cpos, u32 len,
5001 struct ocfs2_extent_tree *et)
Mark Fashehd0c7d702007-07-03 13:27:22 -07005002{
5003 int ret;
5004 u32 left_cpos, rec_range, trunc_range;
5005 int wants_rotate = 0, is_rightmost_tree_rec = 0;
5006 struct super_block *sb = inode->i_sb;
5007 struct ocfs2_path *left_path = NULL;
5008 struct ocfs2_extent_list *el = path_leaf_el(path);
5009 struct ocfs2_extent_rec *rec;
5010 struct ocfs2_extent_block *eb;
5011
5012 if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) {
Tao Mae7d4cb62008-08-18 17:38:44 +08005013 ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005014 if (ret) {
5015 mlog_errno(ret);
5016 goto out;
5017 }
5018
5019 index--;
5020 }
5021
5022 if (index == (le16_to_cpu(el->l_next_free_rec) - 1) &&
5023 path->p_tree_depth) {
5024 /*
5025 * Check whether this is the rightmost tree record. If
5026 * we remove all of this record or part of its right
5027 * edge then an update of the record lengths above it
5028 * will be required.
5029 */
5030 eb = (struct ocfs2_extent_block *)path_leaf_bh(path)->b_data;
5031 if (eb->h_next_leaf_blk == 0)
5032 is_rightmost_tree_rec = 1;
5033 }
5034
5035 rec = &el->l_recs[index];
5036 if (index == 0 && path->p_tree_depth &&
5037 le32_to_cpu(rec->e_cpos) == cpos) {
5038 /*
5039 * Changing the leftmost offset (via partial or whole
5040 * record truncate) of an interior (or rightmost) path
5041 * means we have to update the subtree that is formed
5042 * by this leaf and the one to it's left.
5043 *
5044 * There are two cases we can skip:
5045 * 1) Path is the leftmost one in our inode tree.
5046 * 2) The leaf is rightmost and will be empty after
5047 * we remove the extent record - the rotate code
5048 * knows how to update the newly formed edge.
5049 */
5050
5051 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path,
5052 &left_cpos);
5053 if (ret) {
5054 mlog_errno(ret);
5055 goto out;
5056 }
5057
5058 if (left_cpos && le16_to_cpu(el->l_next_free_rec) > 1) {
5059 left_path = ocfs2_new_path(path_root_bh(path),
5060 path_root_el(path));
5061 if (!left_path) {
5062 ret = -ENOMEM;
5063 mlog_errno(ret);
5064 goto out;
5065 }
5066
5067 ret = ocfs2_find_path(inode, left_path, left_cpos);
5068 if (ret) {
5069 mlog_errno(ret);
5070 goto out;
5071 }
5072 }
5073 }
5074
5075 ret = ocfs2_extend_rotate_transaction(handle, 0,
5076 handle->h_buffer_credits,
5077 path);
5078 if (ret) {
5079 mlog_errno(ret);
5080 goto out;
5081 }
5082
5083 ret = ocfs2_journal_access_path(inode, handle, path);
5084 if (ret) {
5085 mlog_errno(ret);
5086 goto out;
5087 }
5088
5089 ret = ocfs2_journal_access_path(inode, handle, left_path);
5090 if (ret) {
5091 mlog_errno(ret);
5092 goto out;
5093 }
5094
5095 rec_range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
5096 trunc_range = cpos + len;
5097
5098 if (le32_to_cpu(rec->e_cpos) == cpos && rec_range == trunc_range) {
5099 int next_free;
5100
5101 memset(rec, 0, sizeof(*rec));
5102 ocfs2_cleanup_merge(el, index);
5103 wants_rotate = 1;
5104
5105 next_free = le16_to_cpu(el->l_next_free_rec);
5106 if (is_rightmost_tree_rec && next_free > 1) {
5107 /*
5108 * We skip the edge update if this path will
5109 * be deleted by the rotate code.
5110 */
5111 rec = &el->l_recs[next_free - 1];
5112 ocfs2_adjust_rightmost_records(inode, handle, path,
5113 rec);
5114 }
5115 } else if (le32_to_cpu(rec->e_cpos) == cpos) {
5116 /* Remove leftmost portion of the record. */
5117 le32_add_cpu(&rec->e_cpos, len);
5118 le64_add_cpu(&rec->e_blkno, ocfs2_clusters_to_blocks(sb, len));
5119 le16_add_cpu(&rec->e_leaf_clusters, -len);
5120 } else if (rec_range == trunc_range) {
5121 /* Remove rightmost portion of the record */
5122 le16_add_cpu(&rec->e_leaf_clusters, -len);
5123 if (is_rightmost_tree_rec)
5124 ocfs2_adjust_rightmost_records(inode, handle, path, rec);
5125 } else {
5126 /* Caller should have trapped this. */
5127 mlog(ML_ERROR, "Inode %llu: Invalid record truncate: (%u, %u) "
5128 "(%u, %u)\n", (unsigned long long)OCFS2_I(inode)->ip_blkno,
5129 le32_to_cpu(rec->e_cpos),
5130 le16_to_cpu(rec->e_leaf_clusters), cpos, len);
5131 BUG();
5132 }
5133
5134 if (left_path) {
5135 int subtree_index;
5136
5137 subtree_index = ocfs2_find_subtree_root(inode, left_path, path);
5138 ocfs2_complete_edge_insert(inode, handle, left_path, path,
5139 subtree_index);
5140 }
5141
5142 ocfs2_journal_dirty(handle, path_leaf_bh(path));
5143
Tao Mae7d4cb62008-08-18 17:38:44 +08005144 ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005145 if (ret) {
5146 mlog_errno(ret);
5147 goto out;
5148 }
5149
5150out:
5151 ocfs2_free_path(left_path);
5152 return ret;
5153}
5154
Tao Mae7d4cb62008-08-18 17:38:44 +08005155int ocfs2_remove_extent(struct inode *inode, struct buffer_head *root_bh,
Mark Fasheh063c4562007-07-03 13:34:11 -07005156 u32 cpos, u32 len, handle_t *handle,
5157 struct ocfs2_alloc_context *meta_ac,
Tao Mae7d4cb62008-08-18 17:38:44 +08005158 struct ocfs2_cached_dealloc_ctxt *dealloc,
Tao Maf56654c2008-08-18 17:38:48 +08005159 enum ocfs2_extent_tree_type et_type,
Joel Beckerea5efa12008-08-20 16:57:27 -07005160 void *obj)
Mark Fashehd0c7d702007-07-03 13:27:22 -07005161{
5162 int ret, index;
5163 u32 rec_range, trunc_range;
5164 struct ocfs2_extent_rec *rec;
5165 struct ocfs2_extent_list *el;
Tao Mae7d4cb62008-08-18 17:38:44 +08005166 struct ocfs2_path *path = NULL;
Joel Beckerdc0ce612008-08-20 16:48:35 -07005167 struct ocfs2_extent_tree et;
Tao Mae7d4cb62008-08-18 17:38:44 +08005168
Joel Beckerea5efa12008-08-20 16:57:27 -07005169 ocfs2_get_extent_tree(&et, inode, root_bh, et_type, obj);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005170
5171 ocfs2_extent_map_trunc(inode, 0);
5172
Joel Beckerdc0ce612008-08-20 16:48:35 -07005173 path = ocfs2_new_path(et.et_root_bh, et.et_root_el);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005174 if (!path) {
5175 ret = -ENOMEM;
5176 mlog_errno(ret);
5177 goto out;
5178 }
5179
5180 ret = ocfs2_find_path(inode, path, cpos);
5181 if (ret) {
5182 mlog_errno(ret);
5183 goto out;
5184 }
5185
5186 el = path_leaf_el(path);
5187 index = ocfs2_search_extent_list(el, cpos);
5188 if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
5189 ocfs2_error(inode->i_sb,
5190 "Inode %llu has an extent at cpos %u which can no "
5191 "longer be found.\n",
5192 (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
5193 ret = -EROFS;
5194 goto out;
5195 }
5196
5197 /*
5198 * We have 3 cases of extent removal:
5199 * 1) Range covers the entire extent rec
5200 * 2) Range begins or ends on one edge of the extent rec
5201 * 3) Range is in the middle of the extent rec (no shared edges)
5202 *
5203 * For case 1 we remove the extent rec and left rotate to
5204 * fill the hole.
5205 *
5206 * For case 2 we just shrink the existing extent rec, with a
5207 * tree update if the shrinking edge is also the edge of an
5208 * extent block.
5209 *
5210 * For case 3 we do a right split to turn the extent rec into
5211 * something case 2 can handle.
5212 */
5213 rec = &el->l_recs[index];
5214 rec_range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
5215 trunc_range = cpos + len;
5216
5217 BUG_ON(cpos < le32_to_cpu(rec->e_cpos) || trunc_range > rec_range);
5218
5219 mlog(0, "Inode %llu, remove (cpos %u, len %u). Existing index %d "
5220 "(cpos %u, len %u)\n",
5221 (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos, len, index,
5222 le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec));
5223
5224 if (le32_to_cpu(rec->e_cpos) == cpos || rec_range == trunc_range) {
5225 ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
Joel Beckerdc0ce612008-08-20 16:48:35 -07005226 cpos, len, &et);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005227 if (ret) {
5228 mlog_errno(ret);
5229 goto out;
5230 }
5231 } else {
Joel Beckerdc0ce612008-08-20 16:48:35 -07005232 ret = ocfs2_split_tree(inode, &et, handle, path, index,
Mark Fashehd0c7d702007-07-03 13:27:22 -07005233 trunc_range, meta_ac);
5234 if (ret) {
5235 mlog_errno(ret);
5236 goto out;
5237 }
5238
5239 /*
5240 * The split could have manipulated the tree enough to
5241 * move the record location, so we have to look for it again.
5242 */
5243 ocfs2_reinit_path(path, 1);
5244
5245 ret = ocfs2_find_path(inode, path, cpos);
5246 if (ret) {
5247 mlog_errno(ret);
5248 goto out;
5249 }
5250
5251 el = path_leaf_el(path);
5252 index = ocfs2_search_extent_list(el, cpos);
5253 if (index == -1 || index >= le16_to_cpu(el->l_next_free_rec)) {
5254 ocfs2_error(inode->i_sb,
5255 "Inode %llu: split at cpos %u lost record.",
5256 (unsigned long long)OCFS2_I(inode)->ip_blkno,
5257 cpos);
5258 ret = -EROFS;
5259 goto out;
5260 }
5261
5262 /*
5263 * Double check our values here. If anything is fishy,
5264 * it's easier to catch it at the top level.
5265 */
5266 rec = &el->l_recs[index];
5267 rec_range = le32_to_cpu(rec->e_cpos) +
5268 ocfs2_rec_clusters(el, rec);
5269 if (rec_range != trunc_range) {
5270 ocfs2_error(inode->i_sb,
5271 "Inode %llu: error after split at cpos %u"
5272 "trunc len %u, existing record is (%u,%u)",
5273 (unsigned long long)OCFS2_I(inode)->ip_blkno,
5274 cpos, len, le32_to_cpu(rec->e_cpos),
5275 ocfs2_rec_clusters(el, rec));
5276 ret = -EROFS;
5277 goto out;
5278 }
5279
5280 ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc,
Joel Beckerdc0ce612008-08-20 16:48:35 -07005281 cpos, len, &et);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005282 if (ret) {
5283 mlog_errno(ret);
5284 goto out;
5285 }
5286 }
5287
5288out:
5289 ocfs2_free_path(path);
Joel Beckerdc0ce612008-08-20 16:48:35 -07005290 ocfs2_put_extent_tree(&et);
Mark Fashehd0c7d702007-07-03 13:27:22 -07005291 return ret;
5292}
5293
Mark Fasheh063c4562007-07-03 13:34:11 -07005294int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb)
Mark Fashehccd979b2005-12-15 14:31:24 -08005295{
5296 struct buffer_head *tl_bh = osb->osb_tl_bh;
5297 struct ocfs2_dinode *di;
5298 struct ocfs2_truncate_log *tl;
5299
5300 di = (struct ocfs2_dinode *) tl_bh->b_data;
5301 tl = &di->id2.i_dealloc;
5302
5303 mlog_bug_on_msg(le16_to_cpu(tl->tl_used) > le16_to_cpu(tl->tl_count),
5304 "slot %d, invalid truncate log parameters: used = "
5305 "%u, count = %u\n", osb->slot_num,
5306 le16_to_cpu(tl->tl_used), le16_to_cpu(tl->tl_count));
5307 return le16_to_cpu(tl->tl_used) == le16_to_cpu(tl->tl_count);
5308}
5309
5310static int ocfs2_truncate_log_can_coalesce(struct ocfs2_truncate_log *tl,
5311 unsigned int new_start)
5312{
5313 unsigned int tail_index;
5314 unsigned int current_tail;
5315
5316 /* No records, nothing to coalesce */
5317 if (!le16_to_cpu(tl->tl_used))
5318 return 0;
5319
5320 tail_index = le16_to_cpu(tl->tl_used) - 1;
5321 current_tail = le32_to_cpu(tl->tl_recs[tail_index].t_start);
5322 current_tail += le32_to_cpu(tl->tl_recs[tail_index].t_clusters);
5323
5324 return current_tail == new_start;
5325}
5326
Mark Fasheh063c4562007-07-03 13:34:11 -07005327int ocfs2_truncate_log_append(struct ocfs2_super *osb,
5328 handle_t *handle,
5329 u64 start_blk,
5330 unsigned int num_clusters)
Mark Fashehccd979b2005-12-15 14:31:24 -08005331{
5332 int status, index;
5333 unsigned int start_cluster, tl_count;
5334 struct inode *tl_inode = osb->osb_tl_inode;
5335 struct buffer_head *tl_bh = osb->osb_tl_bh;
5336 struct ocfs2_dinode *di;
5337 struct ocfs2_truncate_log *tl;
5338
Mark Fashehb06970532006-03-03 10:24:33 -08005339 mlog_entry("start_blk = %llu, num_clusters = %u\n",
5340 (unsigned long long)start_blk, num_clusters);
Mark Fashehccd979b2005-12-15 14:31:24 -08005341
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005342 BUG_ON(mutex_trylock(&tl_inode->i_mutex));
Mark Fashehccd979b2005-12-15 14:31:24 -08005343
5344 start_cluster = ocfs2_blocks_to_clusters(osb->sb, start_blk);
5345
5346 di = (struct ocfs2_dinode *) tl_bh->b_data;
5347 tl = &di->id2.i_dealloc;
5348 if (!OCFS2_IS_VALID_DINODE(di)) {
5349 OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
5350 status = -EIO;
5351 goto bail;
5352 }
5353
5354 tl_count = le16_to_cpu(tl->tl_count);
5355 mlog_bug_on_msg(tl_count > ocfs2_truncate_recs_per_inode(osb->sb) ||
5356 tl_count == 0,
Mark Fashehb06970532006-03-03 10:24:33 -08005357 "Truncate record count on #%llu invalid "
5358 "wanted %u, actual %u\n",
5359 (unsigned long long)OCFS2_I(tl_inode)->ip_blkno,
Mark Fashehccd979b2005-12-15 14:31:24 -08005360 ocfs2_truncate_recs_per_inode(osb->sb),
5361 le16_to_cpu(tl->tl_count));
5362
5363 /* Caller should have known to flush before calling us. */
5364 index = le16_to_cpu(tl->tl_used);
5365 if (index >= tl_count) {
5366 status = -ENOSPC;
5367 mlog_errno(status);
5368 goto bail;
5369 }
5370
5371 status = ocfs2_journal_access(handle, tl_inode, tl_bh,
5372 OCFS2_JOURNAL_ACCESS_WRITE);
5373 if (status < 0) {
5374 mlog_errno(status);
5375 goto bail;
5376 }
5377
5378 mlog(0, "Log truncate of %u clusters starting at cluster %u to "
Mark Fashehb06970532006-03-03 10:24:33 -08005379 "%llu (index = %d)\n", num_clusters, start_cluster,
5380 (unsigned long long)OCFS2_I(tl_inode)->ip_blkno, index);
Mark Fashehccd979b2005-12-15 14:31:24 -08005381
5382 if (ocfs2_truncate_log_can_coalesce(tl, start_cluster)) {
5383 /*
5384 * Move index back to the record we are coalescing with.
5385 * ocfs2_truncate_log_can_coalesce() guarantees nonzero
5386 */
5387 index--;
5388
5389 num_clusters += le32_to_cpu(tl->tl_recs[index].t_clusters);
5390 mlog(0, "Coalesce with index %u (start = %u, clusters = %u)\n",
5391 index, le32_to_cpu(tl->tl_recs[index].t_start),
5392 num_clusters);
5393 } else {
5394 tl->tl_recs[index].t_start = cpu_to_le32(start_cluster);
5395 tl->tl_used = cpu_to_le16(index + 1);
5396 }
5397 tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters);
5398
5399 status = ocfs2_journal_dirty(handle, tl_bh);
5400 if (status < 0) {
5401 mlog_errno(status);
5402 goto bail;
5403 }
5404
5405bail:
5406 mlog_exit(status);
5407 return status;
5408}
5409
5410static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
Mark Fasheh1fabe142006-10-09 18:11:45 -07005411 handle_t *handle,
Mark Fashehccd979b2005-12-15 14:31:24 -08005412 struct inode *data_alloc_inode,
5413 struct buffer_head *data_alloc_bh)
5414{
5415 int status = 0;
5416 int i;
5417 unsigned int num_clusters;
5418 u64 start_blk;
5419 struct ocfs2_truncate_rec rec;
5420 struct ocfs2_dinode *di;
5421 struct ocfs2_truncate_log *tl;
5422 struct inode *tl_inode = osb->osb_tl_inode;
5423 struct buffer_head *tl_bh = osb->osb_tl_bh;
5424
5425 mlog_entry_void();
5426
5427 di = (struct ocfs2_dinode *) tl_bh->b_data;
5428 tl = &di->id2.i_dealloc;
5429 i = le16_to_cpu(tl->tl_used) - 1;
5430 while (i >= 0) {
5431 /* Caller has given us at least enough credits to
5432 * update the truncate log dinode */
5433 status = ocfs2_journal_access(handle, tl_inode, tl_bh,
5434 OCFS2_JOURNAL_ACCESS_WRITE);
5435 if (status < 0) {
5436 mlog_errno(status);
5437 goto bail;
5438 }
5439
5440 tl->tl_used = cpu_to_le16(i);
5441
5442 status = ocfs2_journal_dirty(handle, tl_bh);
5443 if (status < 0) {
5444 mlog_errno(status);
5445 goto bail;
5446 }
5447
5448 /* TODO: Perhaps we can calculate the bulk of the
5449 * credits up front rather than extending like
5450 * this. */
5451 status = ocfs2_extend_trans(handle,
5452 OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
5453 if (status < 0) {
5454 mlog_errno(status);
5455 goto bail;
5456 }
5457
5458 rec = tl->tl_recs[i];
5459 start_blk = ocfs2_clusters_to_blocks(data_alloc_inode->i_sb,
5460 le32_to_cpu(rec.t_start));
5461 num_clusters = le32_to_cpu(rec.t_clusters);
5462
5463 /* if start_blk is not set, we ignore the record as
5464 * invalid. */
5465 if (start_blk) {
5466 mlog(0, "free record %d, start = %u, clusters = %u\n",
5467 i, le32_to_cpu(rec.t_start), num_clusters);
5468
5469 status = ocfs2_free_clusters(handle, data_alloc_inode,
5470 data_alloc_bh, start_blk,
5471 num_clusters);
5472 if (status < 0) {
5473 mlog_errno(status);
5474 goto bail;
5475 }
5476 }
5477 i--;
5478 }
5479
5480bail:
5481 mlog_exit(status);
5482 return status;
5483}
5484
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005485/* Expects you to already be holding tl_inode->i_mutex */
Mark Fasheh063c4562007-07-03 13:34:11 -07005486int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
Mark Fashehccd979b2005-12-15 14:31:24 -08005487{
5488 int status;
5489 unsigned int num_to_flush;
Mark Fasheh1fabe142006-10-09 18:11:45 -07005490 handle_t *handle;
Mark Fashehccd979b2005-12-15 14:31:24 -08005491 struct inode *tl_inode = osb->osb_tl_inode;
5492 struct inode *data_alloc_inode = NULL;
5493 struct buffer_head *tl_bh = osb->osb_tl_bh;
5494 struct buffer_head *data_alloc_bh = NULL;
5495 struct ocfs2_dinode *di;
5496 struct ocfs2_truncate_log *tl;
5497
5498 mlog_entry_void();
5499
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005500 BUG_ON(mutex_trylock(&tl_inode->i_mutex));
Mark Fashehccd979b2005-12-15 14:31:24 -08005501
5502 di = (struct ocfs2_dinode *) tl_bh->b_data;
5503 tl = &di->id2.i_dealloc;
5504 if (!OCFS2_IS_VALID_DINODE(di)) {
5505 OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
5506 status = -EIO;
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005507 goto out;
Mark Fashehccd979b2005-12-15 14:31:24 -08005508 }
5509
5510 num_to_flush = le16_to_cpu(tl->tl_used);
Mark Fashehb06970532006-03-03 10:24:33 -08005511 mlog(0, "Flush %u records from truncate log #%llu\n",
5512 num_to_flush, (unsigned long long)OCFS2_I(tl_inode)->ip_blkno);
Mark Fashehccd979b2005-12-15 14:31:24 -08005513 if (!num_to_flush) {
5514 status = 0;
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005515 goto out;
Mark Fashehccd979b2005-12-15 14:31:24 -08005516 }
5517
5518 data_alloc_inode = ocfs2_get_system_file_inode(osb,
5519 GLOBAL_BITMAP_SYSTEM_INODE,
5520 OCFS2_INVALID_SLOT);
5521 if (!data_alloc_inode) {
5522 status = -EINVAL;
5523 mlog(ML_ERROR, "Could not get bitmap inode!\n");
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005524 goto out;
Mark Fashehccd979b2005-12-15 14:31:24 -08005525 }
5526
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005527 mutex_lock(&data_alloc_inode->i_mutex);
5528
Mark Fashehe63aecb62007-10-18 15:30:42 -07005529 status = ocfs2_inode_lock(data_alloc_inode, &data_alloc_bh, 1);
Mark Fashehccd979b2005-12-15 14:31:24 -08005530 if (status < 0) {
5531 mlog_errno(status);
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005532 goto out_mutex;
Mark Fashehccd979b2005-12-15 14:31:24 -08005533 }
5534
Mark Fasheh65eff9c2006-10-09 17:26:22 -07005535 handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
Mark Fashehccd979b2005-12-15 14:31:24 -08005536 if (IS_ERR(handle)) {
5537 status = PTR_ERR(handle);
Mark Fashehccd979b2005-12-15 14:31:24 -08005538 mlog_errno(status);
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005539 goto out_unlock;
Mark Fashehccd979b2005-12-15 14:31:24 -08005540 }
5541
5542 status = ocfs2_replay_truncate_records(osb, handle, data_alloc_inode,
5543 data_alloc_bh);
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005544 if (status < 0)
Mark Fashehccd979b2005-12-15 14:31:24 -08005545 mlog_errno(status);
Mark Fashehccd979b2005-12-15 14:31:24 -08005546
Mark Fasheh02dc1af2006-10-09 16:48:10 -07005547 ocfs2_commit_trans(osb, handle);
Mark Fashehccd979b2005-12-15 14:31:24 -08005548
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005549out_unlock:
5550 brelse(data_alloc_bh);
Mark Fashehe63aecb62007-10-18 15:30:42 -07005551 ocfs2_inode_unlock(data_alloc_inode, 1);
Mark Fashehccd979b2005-12-15 14:31:24 -08005552
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005553out_mutex:
5554 mutex_unlock(&data_alloc_inode->i_mutex);
5555 iput(data_alloc_inode);
Mark Fashehccd979b2005-12-15 14:31:24 -08005556
Mark Fashehe08dc8b2006-10-05 15:58:48 -07005557out:
Mark Fashehccd979b2005-12-15 14:31:24 -08005558 mlog_exit(status);
5559 return status;
5560}
5561
5562int ocfs2_flush_truncate_log(struct ocfs2_super *osb)
5563{
5564 int status;
5565 struct inode *tl_inode = osb->osb_tl_inode;
5566
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005567 mutex_lock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08005568 status = __ocfs2_flush_truncate_log(osb);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005569 mutex_unlock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08005570
5571 return status;
5572}
5573
David Howellsc4028952006-11-22 14:57:56 +00005574static void ocfs2_truncate_log_worker(struct work_struct *work)
Mark Fashehccd979b2005-12-15 14:31:24 -08005575{
5576 int status;
David Howellsc4028952006-11-22 14:57:56 +00005577 struct ocfs2_super *osb =
5578 container_of(work, struct ocfs2_super,
5579 osb_truncate_log_wq.work);
Mark Fashehccd979b2005-12-15 14:31:24 -08005580
5581 mlog_entry_void();
5582
5583 status = ocfs2_flush_truncate_log(osb);
5584 if (status < 0)
5585 mlog_errno(status);
Tao Ma4d0ddb22008-03-05 16:11:46 +08005586 else
5587 ocfs2_init_inode_steal_slot(osb);
Mark Fashehccd979b2005-12-15 14:31:24 -08005588
5589 mlog_exit(status);
5590}
5591
5592#define OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL (2 * HZ)
5593void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
5594 int cancel)
5595{
5596 if (osb->osb_tl_inode) {
5597 /* We want to push off log flushes while truncates are
5598 * still running. */
5599 if (cancel)
5600 cancel_delayed_work(&osb->osb_truncate_log_wq);
5601
5602 queue_delayed_work(ocfs2_wq, &osb->osb_truncate_log_wq,
5603 OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL);
5604 }
5605}
5606
5607static int ocfs2_get_truncate_log_info(struct ocfs2_super *osb,
5608 int slot_num,
5609 struct inode **tl_inode,
5610 struct buffer_head **tl_bh)
5611{
5612 int status;
5613 struct inode *inode = NULL;
5614 struct buffer_head *bh = NULL;
5615
5616 inode = ocfs2_get_system_file_inode(osb,
5617 TRUNCATE_LOG_SYSTEM_INODE,
5618 slot_num);
5619 if (!inode) {
5620 status = -EINVAL;
5621 mlog(ML_ERROR, "Could not get load truncate log inode!\n");
5622 goto bail;
5623 }
5624
5625 status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh,
5626 OCFS2_BH_CACHED, inode);
5627 if (status < 0) {
5628 iput(inode);
5629 mlog_errno(status);
5630 goto bail;
5631 }
5632
5633 *tl_inode = inode;
5634 *tl_bh = bh;
5635bail:
5636 mlog_exit(status);
5637 return status;
5638}
5639
5640/* called during the 1st stage of node recovery. we stamp a clean
5641 * truncate log and pass back a copy for processing later. if the
5642 * truncate log does not require processing, a *tl_copy is set to
5643 * NULL. */
5644int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
5645 int slot_num,
5646 struct ocfs2_dinode **tl_copy)
5647{
5648 int status;
5649 struct inode *tl_inode = NULL;
5650 struct buffer_head *tl_bh = NULL;
5651 struct ocfs2_dinode *di;
5652 struct ocfs2_truncate_log *tl;
5653
5654 *tl_copy = NULL;
5655
5656 mlog(0, "recover truncate log from slot %d\n", slot_num);
5657
5658 status = ocfs2_get_truncate_log_info(osb, slot_num, &tl_inode, &tl_bh);
5659 if (status < 0) {
5660 mlog_errno(status);
5661 goto bail;
5662 }
5663
5664 di = (struct ocfs2_dinode *) tl_bh->b_data;
5665 tl = &di->id2.i_dealloc;
5666 if (!OCFS2_IS_VALID_DINODE(di)) {
5667 OCFS2_RO_ON_INVALID_DINODE(tl_inode->i_sb, di);
5668 status = -EIO;
5669 goto bail;
5670 }
5671
5672 if (le16_to_cpu(tl->tl_used)) {
5673 mlog(0, "We'll have %u logs to recover\n",
5674 le16_to_cpu(tl->tl_used));
5675
5676 *tl_copy = kmalloc(tl_bh->b_size, GFP_KERNEL);
5677 if (!(*tl_copy)) {
5678 status = -ENOMEM;
5679 mlog_errno(status);
5680 goto bail;
5681 }
5682
5683 /* Assuming the write-out below goes well, this copy
5684 * will be passed back to recovery for processing. */
5685 memcpy(*tl_copy, tl_bh->b_data, tl_bh->b_size);
5686
5687 /* All we need to do to clear the truncate log is set
5688 * tl_used. */
5689 tl->tl_used = 0;
5690
5691 status = ocfs2_write_block(osb, tl_bh, tl_inode);
5692 if (status < 0) {
5693 mlog_errno(status);
5694 goto bail;
5695 }
5696 }
5697
5698bail:
5699 if (tl_inode)
5700 iput(tl_inode);
5701 if (tl_bh)
5702 brelse(tl_bh);
5703
5704 if (status < 0 && (*tl_copy)) {
5705 kfree(*tl_copy);
5706 *tl_copy = NULL;
5707 }
5708
5709 mlog_exit(status);
5710 return status;
5711}
5712
5713int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
5714 struct ocfs2_dinode *tl_copy)
5715{
5716 int status = 0;
5717 int i;
5718 unsigned int clusters, num_recs, start_cluster;
5719 u64 start_blk;
Mark Fasheh1fabe142006-10-09 18:11:45 -07005720 handle_t *handle;
Mark Fashehccd979b2005-12-15 14:31:24 -08005721 struct inode *tl_inode = osb->osb_tl_inode;
5722 struct ocfs2_truncate_log *tl;
5723
5724 mlog_entry_void();
5725
5726 if (OCFS2_I(tl_inode)->ip_blkno == le64_to_cpu(tl_copy->i_blkno)) {
5727 mlog(ML_ERROR, "Asked to recover my own truncate log!\n");
5728 return -EINVAL;
5729 }
5730
5731 tl = &tl_copy->id2.i_dealloc;
5732 num_recs = le16_to_cpu(tl->tl_used);
Mark Fashehb06970532006-03-03 10:24:33 -08005733 mlog(0, "cleanup %u records from %llu\n", num_recs,
Mark Fasheh1ca1a112007-04-27 16:01:25 -07005734 (unsigned long long)le64_to_cpu(tl_copy->i_blkno));
Mark Fashehccd979b2005-12-15 14:31:24 -08005735
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005736 mutex_lock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08005737 for(i = 0; i < num_recs; i++) {
5738 if (ocfs2_truncate_log_needs_flush(osb)) {
5739 status = __ocfs2_flush_truncate_log(osb);
5740 if (status < 0) {
5741 mlog_errno(status);
5742 goto bail_up;
5743 }
5744 }
5745
Mark Fasheh65eff9c2006-10-09 17:26:22 -07005746 handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
Mark Fashehccd979b2005-12-15 14:31:24 -08005747 if (IS_ERR(handle)) {
5748 status = PTR_ERR(handle);
5749 mlog_errno(status);
5750 goto bail_up;
5751 }
5752
5753 clusters = le32_to_cpu(tl->tl_recs[i].t_clusters);
5754 start_cluster = le32_to_cpu(tl->tl_recs[i].t_start);
5755 start_blk = ocfs2_clusters_to_blocks(osb->sb, start_cluster);
5756
5757 status = ocfs2_truncate_log_append(osb, handle,
5758 start_blk, clusters);
Mark Fasheh02dc1af2006-10-09 16:48:10 -07005759 ocfs2_commit_trans(osb, handle);
Mark Fashehccd979b2005-12-15 14:31:24 -08005760 if (status < 0) {
5761 mlog_errno(status);
5762 goto bail_up;
5763 }
5764 }
5765
5766bail_up:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08005767 mutex_unlock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08005768
5769 mlog_exit(status);
5770 return status;
5771}
5772
5773void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb)
5774{
5775 int status;
5776 struct inode *tl_inode = osb->osb_tl_inode;
5777
5778 mlog_entry_void();
5779
5780 if (tl_inode) {
5781 cancel_delayed_work(&osb->osb_truncate_log_wq);
5782 flush_workqueue(ocfs2_wq);
5783
5784 status = ocfs2_flush_truncate_log(osb);
5785 if (status < 0)
5786 mlog_errno(status);
5787
5788 brelse(osb->osb_tl_bh);
5789 iput(osb->osb_tl_inode);
5790 }
5791
5792 mlog_exit_void();
5793}
5794
5795int ocfs2_truncate_log_init(struct ocfs2_super *osb)
5796{
5797 int status;
5798 struct inode *tl_inode = NULL;
5799 struct buffer_head *tl_bh = NULL;
5800
5801 mlog_entry_void();
5802
5803 status = ocfs2_get_truncate_log_info(osb,
5804 osb->slot_num,
5805 &tl_inode,
5806 &tl_bh);
5807 if (status < 0)
5808 mlog_errno(status);
5809
5810 /* ocfs2_truncate_log_shutdown keys on the existence of
5811 * osb->osb_tl_inode so we don't set any of the osb variables
5812 * until we're sure all is well. */
David Howellsc4028952006-11-22 14:57:56 +00005813 INIT_DELAYED_WORK(&osb->osb_truncate_log_wq,
5814 ocfs2_truncate_log_worker);
Mark Fashehccd979b2005-12-15 14:31:24 -08005815 osb->osb_tl_bh = tl_bh;
5816 osb->osb_tl_inode = tl_inode;
5817
5818 mlog_exit(status);
5819 return status;
5820}
5821
Mark Fasheh2b604352007-06-22 15:45:27 -07005822/*
5823 * Delayed de-allocation of suballocator blocks.
5824 *
5825 * Some sets of block de-allocations might involve multiple suballocator inodes.
5826 *
5827 * The locking for this can get extremely complicated, especially when
5828 * the suballocator inodes to delete from aren't known until deep
5829 * within an unrelated codepath.
5830 *
5831 * ocfs2_extent_block structures are a good example of this - an inode
5832 * btree could have been grown by any number of nodes each allocating
5833 * out of their own suballoc inode.
5834 *
5835 * These structures allow the delay of block de-allocation until a
5836 * later time, when locking of multiple cluster inodes won't cause
5837 * deadlock.
5838 */
5839
5840/*
5841 * Describes a single block free from a suballocator
5842 */
5843struct ocfs2_cached_block_free {
5844 struct ocfs2_cached_block_free *free_next;
5845 u64 free_blk;
5846 unsigned int free_bit;
5847};
5848
5849struct ocfs2_per_slot_free_list {
5850 struct ocfs2_per_slot_free_list *f_next_suballocator;
5851 int f_inode_type;
5852 int f_slot;
5853 struct ocfs2_cached_block_free *f_first;
5854};
5855
5856static int ocfs2_free_cached_items(struct ocfs2_super *osb,
5857 int sysfile_type,
5858 int slot,
5859 struct ocfs2_cached_block_free *head)
5860{
5861 int ret;
5862 u64 bg_blkno;
5863 handle_t *handle;
5864 struct inode *inode;
5865 struct buffer_head *di_bh = NULL;
5866 struct ocfs2_cached_block_free *tmp;
5867
5868 inode = ocfs2_get_system_file_inode(osb, sysfile_type, slot);
5869 if (!inode) {
5870 ret = -EINVAL;
5871 mlog_errno(ret);
5872 goto out;
5873 }
5874
5875 mutex_lock(&inode->i_mutex);
5876
Mark Fashehe63aecb62007-10-18 15:30:42 -07005877 ret = ocfs2_inode_lock(inode, &di_bh, 1);
Mark Fasheh2b604352007-06-22 15:45:27 -07005878 if (ret) {
5879 mlog_errno(ret);
5880 goto out_mutex;
5881 }
5882
5883 handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE);
5884 if (IS_ERR(handle)) {
5885 ret = PTR_ERR(handle);
5886 mlog_errno(ret);
5887 goto out_unlock;
5888 }
5889
5890 while (head) {
5891 bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
5892 head->free_bit);
5893 mlog(0, "Free bit: (bit %u, blkno %llu)\n",
5894 head->free_bit, (unsigned long long)head->free_blk);
5895
5896 ret = ocfs2_free_suballoc_bits(handle, inode, di_bh,
5897 head->free_bit, bg_blkno, 1);
5898 if (ret) {
5899 mlog_errno(ret);
5900 goto out_journal;
5901 }
5902
5903 ret = ocfs2_extend_trans(handle, OCFS2_SUBALLOC_FREE);
5904 if (ret) {
5905 mlog_errno(ret);
5906 goto out_journal;
5907 }
5908
5909 tmp = head;
5910 head = head->free_next;
5911 kfree(tmp);
5912 }
5913
5914out_journal:
5915 ocfs2_commit_trans(osb, handle);
5916
5917out_unlock:
Mark Fashehe63aecb62007-10-18 15:30:42 -07005918 ocfs2_inode_unlock(inode, 1);
Mark Fasheh2b604352007-06-22 15:45:27 -07005919 brelse(di_bh);
5920out_mutex:
5921 mutex_unlock(&inode->i_mutex);
5922 iput(inode);
5923out:
5924 while(head) {
5925 /* Premature exit may have left some dangling items. */
5926 tmp = head;
5927 head = head->free_next;
5928 kfree(tmp);
5929 }
5930
5931 return ret;
5932}
5933
5934int ocfs2_run_deallocs(struct ocfs2_super *osb,
5935 struct ocfs2_cached_dealloc_ctxt *ctxt)
5936{
5937 int ret = 0, ret2;
5938 struct ocfs2_per_slot_free_list *fl;
5939
5940 if (!ctxt)
5941 return 0;
5942
5943 while (ctxt->c_first_suballocator) {
5944 fl = ctxt->c_first_suballocator;
5945
5946 if (fl->f_first) {
5947 mlog(0, "Free items: (type %u, slot %d)\n",
5948 fl->f_inode_type, fl->f_slot);
5949 ret2 = ocfs2_free_cached_items(osb, fl->f_inode_type,
5950 fl->f_slot, fl->f_first);
5951 if (ret2)
5952 mlog_errno(ret2);
5953 if (!ret)
5954 ret = ret2;
5955 }
5956
5957 ctxt->c_first_suballocator = fl->f_next_suballocator;
5958 kfree(fl);
5959 }
5960
5961 return ret;
5962}
5963
5964static struct ocfs2_per_slot_free_list *
5965ocfs2_find_per_slot_free_list(int type,
5966 int slot,
5967 struct ocfs2_cached_dealloc_ctxt *ctxt)
5968{
5969 struct ocfs2_per_slot_free_list *fl = ctxt->c_first_suballocator;
5970
5971 while (fl) {
5972 if (fl->f_inode_type == type && fl->f_slot == slot)
5973 return fl;
5974
5975 fl = fl->f_next_suballocator;
5976 }
5977
5978 fl = kmalloc(sizeof(*fl), GFP_NOFS);
5979 if (fl) {
5980 fl->f_inode_type = type;
5981 fl->f_slot = slot;
5982 fl->f_first = NULL;
5983 fl->f_next_suballocator = ctxt->c_first_suballocator;
5984
5985 ctxt->c_first_suballocator = fl;
5986 }
5987 return fl;
5988}
5989
5990static int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
5991 int type, int slot, u64 blkno,
5992 unsigned int bit)
5993{
5994 int ret;
5995 struct ocfs2_per_slot_free_list *fl;
5996 struct ocfs2_cached_block_free *item;
5997
5998 fl = ocfs2_find_per_slot_free_list(type, slot, ctxt);
5999 if (fl == NULL) {
6000 ret = -ENOMEM;
6001 mlog_errno(ret);
6002 goto out;
6003 }
6004
6005 item = kmalloc(sizeof(*item), GFP_NOFS);
6006 if (item == NULL) {
6007 ret = -ENOMEM;
6008 mlog_errno(ret);
6009 goto out;
6010 }
6011
6012 mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n",
6013 type, slot, bit, (unsigned long long)blkno);
6014
6015 item->free_blk = blkno;
6016 item->free_bit = bit;
6017 item->free_next = fl->f_first;
6018
6019 fl->f_first = item;
6020
6021 ret = 0;
6022out:
6023 return ret;
6024}
6025
Mark Fasheh59a5e412007-06-22 15:52:36 -07006026static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
6027 struct ocfs2_extent_block *eb)
6028{
6029 return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE,
6030 le16_to_cpu(eb->h_suballoc_slot),
6031 le64_to_cpu(eb->h_blkno),
6032 le16_to_cpu(eb->h_suballoc_bit));
6033}
6034
Mark Fashehccd979b2005-12-15 14:31:24 -08006035/* This function will figure out whether the currently last extent
6036 * block will be deleted, and if it will, what the new last extent
6037 * block will be so we can update his h_next_leaf_blk field, as well
6038 * as the dinodes i_last_eb_blk */
Mark Fashehdcd05382007-01-16 11:32:23 -08006039static int ocfs2_find_new_last_ext_blk(struct inode *inode,
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006040 unsigned int clusters_to_del,
Mark Fashehdcd05382007-01-16 11:32:23 -08006041 struct ocfs2_path *path,
Mark Fashehccd979b2005-12-15 14:31:24 -08006042 struct buffer_head **new_last_eb)
6043{
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006044 int next_free, ret = 0;
Mark Fashehdcd05382007-01-16 11:32:23 -08006045 u32 cpos;
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006046 struct ocfs2_extent_rec *rec;
Mark Fashehccd979b2005-12-15 14:31:24 -08006047 struct ocfs2_extent_block *eb;
6048 struct ocfs2_extent_list *el;
6049 struct buffer_head *bh = NULL;
6050
6051 *new_last_eb = NULL;
6052
Mark Fashehccd979b2005-12-15 14:31:24 -08006053 /* we have no tree, so of course, no last_eb. */
Mark Fashehdcd05382007-01-16 11:32:23 -08006054 if (!path->p_tree_depth)
6055 goto out;
Mark Fashehccd979b2005-12-15 14:31:24 -08006056
6057 /* trunc to zero special case - this makes tree_depth = 0
6058 * regardless of what it is. */
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006059 if (OCFS2_I(inode)->ip_clusters == clusters_to_del)
Mark Fashehdcd05382007-01-16 11:32:23 -08006060 goto out;
Mark Fashehccd979b2005-12-15 14:31:24 -08006061
Mark Fashehdcd05382007-01-16 11:32:23 -08006062 el = path_leaf_el(path);
Mark Fashehccd979b2005-12-15 14:31:24 -08006063 BUG_ON(!el->l_next_free_rec);
6064
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006065 /*
6066 * Make sure that this extent list will actually be empty
6067 * after we clear away the data. We can shortcut out if
6068 * there's more than one non-empty extent in the
6069 * list. Otherwise, a check of the remaining extent is
6070 * necessary.
6071 */
6072 next_free = le16_to_cpu(el->l_next_free_rec);
6073 rec = NULL;
Mark Fashehdcd05382007-01-16 11:32:23 -08006074 if (ocfs2_is_empty_extent(&el->l_recs[0])) {
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006075 if (next_free > 2)
Mark Fashehdcd05382007-01-16 11:32:23 -08006076 goto out;
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006077
6078 /* We may have a valid extent in index 1, check it. */
6079 if (next_free == 2)
6080 rec = &el->l_recs[1];
6081
6082 /*
6083 * Fall through - no more nonempty extents, so we want
6084 * to delete this leaf.
6085 */
6086 } else {
6087 if (next_free > 1)
6088 goto out;
6089
6090 rec = &el->l_recs[0];
6091 }
6092
6093 if (rec) {
6094 /*
6095 * Check it we'll only be trimming off the end of this
6096 * cluster.
6097 */
Mark Fashehe48edee2007-03-07 16:46:57 -08006098 if (le16_to_cpu(rec->e_leaf_clusters) > clusters_to_del)
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006099 goto out;
6100 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006101
Mark Fashehdcd05382007-01-16 11:32:23 -08006102 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
6103 if (ret) {
6104 mlog_errno(ret);
6105 goto out;
6106 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006107
Mark Fashehdcd05382007-01-16 11:32:23 -08006108 ret = ocfs2_find_leaf(inode, path_root_el(path), cpos, &bh);
6109 if (ret) {
6110 mlog_errno(ret);
6111 goto out;
6112 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006113
Mark Fashehdcd05382007-01-16 11:32:23 -08006114 eb = (struct ocfs2_extent_block *) bh->b_data;
6115 el = &eb->h_list;
6116 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
6117 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
6118 ret = -EROFS;
6119 goto out;
6120 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006121
6122 *new_last_eb = bh;
6123 get_bh(*new_last_eb);
Mark Fashehdcd05382007-01-16 11:32:23 -08006124 mlog(0, "returning block %llu, (cpos: %u)\n",
6125 (unsigned long long)le64_to_cpu(eb->h_blkno), cpos);
6126out:
6127 brelse(bh);
Mark Fashehccd979b2005-12-15 14:31:24 -08006128
Mark Fashehdcd05382007-01-16 11:32:23 -08006129 return ret;
Mark Fashehccd979b2005-12-15 14:31:24 -08006130}
6131
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006132/*
6133 * Trim some clusters off the rightmost edge of a tree. Only called
6134 * during truncate.
6135 *
6136 * The caller needs to:
6137 * - start journaling of each path component.
6138 * - compute and fully set up any new last ext block
6139 */
6140static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path,
6141 handle_t *handle, struct ocfs2_truncate_context *tc,
6142 u32 clusters_to_del, u64 *delete_start)
6143{
6144 int ret, i, index = path->p_tree_depth;
6145 u32 new_edge = 0;
6146 u64 deleted_eb = 0;
6147 struct buffer_head *bh;
6148 struct ocfs2_extent_list *el;
6149 struct ocfs2_extent_rec *rec;
6150
6151 *delete_start = 0;
6152
6153 while (index >= 0) {
6154 bh = path->p_node[index].bh;
6155 el = path->p_node[index].el;
6156
6157 mlog(0, "traveling tree (index = %d, block = %llu)\n",
6158 index, (unsigned long long)bh->b_blocknr);
6159
6160 BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
6161
6162 if (index !=
6163 (path->p_tree_depth - le16_to_cpu(el->l_tree_depth))) {
6164 ocfs2_error(inode->i_sb,
6165 "Inode %lu has invalid ext. block %llu",
6166 inode->i_ino,
6167 (unsigned long long)bh->b_blocknr);
6168 ret = -EROFS;
6169 goto out;
6170 }
6171
6172find_tail_record:
6173 i = le16_to_cpu(el->l_next_free_rec) - 1;
6174 rec = &el->l_recs[i];
6175
6176 mlog(0, "Extent list before: record %d: (%u, %u, %llu), "
6177 "next = %u\n", i, le32_to_cpu(rec->e_cpos),
Mark Fashehe48edee2007-03-07 16:46:57 -08006178 ocfs2_rec_clusters(el, rec),
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006179 (unsigned long long)le64_to_cpu(rec->e_blkno),
6180 le16_to_cpu(el->l_next_free_rec));
6181
Mark Fashehe48edee2007-03-07 16:46:57 -08006182 BUG_ON(ocfs2_rec_clusters(el, rec) < clusters_to_del);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006183
6184 if (le16_to_cpu(el->l_tree_depth) == 0) {
6185 /*
6186 * If the leaf block contains a single empty
6187 * extent and no records, we can just remove
6188 * the block.
6189 */
6190 if (i == 0 && ocfs2_is_empty_extent(rec)) {
6191 memset(rec, 0,
6192 sizeof(struct ocfs2_extent_rec));
6193 el->l_next_free_rec = cpu_to_le16(0);
6194
6195 goto delete;
6196 }
6197
6198 /*
6199 * Remove any empty extents by shifting things
6200 * left. That should make life much easier on
6201 * the code below. This condition is rare
6202 * enough that we shouldn't see a performance
6203 * hit.
6204 */
6205 if (ocfs2_is_empty_extent(&el->l_recs[0])) {
6206 le16_add_cpu(&el->l_next_free_rec, -1);
6207
6208 for(i = 0;
6209 i < le16_to_cpu(el->l_next_free_rec); i++)
6210 el->l_recs[i] = el->l_recs[i + 1];
6211
6212 memset(&el->l_recs[i], 0,
6213 sizeof(struct ocfs2_extent_rec));
6214
6215 /*
6216 * We've modified our extent list. The
6217 * simplest way to handle this change
6218 * is to being the search from the
6219 * start again.
6220 */
6221 goto find_tail_record;
6222 }
6223
Mark Fashehe48edee2007-03-07 16:46:57 -08006224 le16_add_cpu(&rec->e_leaf_clusters, -clusters_to_del);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006225
6226 /*
6227 * We'll use "new_edge" on our way back up the
6228 * tree to know what our rightmost cpos is.
6229 */
Mark Fashehe48edee2007-03-07 16:46:57 -08006230 new_edge = le16_to_cpu(rec->e_leaf_clusters);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006231 new_edge += le32_to_cpu(rec->e_cpos);
6232
6233 /*
6234 * The caller will use this to delete data blocks.
6235 */
6236 *delete_start = le64_to_cpu(rec->e_blkno)
6237 + ocfs2_clusters_to_blocks(inode->i_sb,
Mark Fashehe48edee2007-03-07 16:46:57 -08006238 le16_to_cpu(rec->e_leaf_clusters));
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006239
6240 /*
6241 * If it's now empty, remove this record.
6242 */
Mark Fashehe48edee2007-03-07 16:46:57 -08006243 if (le16_to_cpu(rec->e_leaf_clusters) == 0) {
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006244 memset(rec, 0,
6245 sizeof(struct ocfs2_extent_rec));
6246 le16_add_cpu(&el->l_next_free_rec, -1);
6247 }
6248 } else {
6249 if (le64_to_cpu(rec->e_blkno) == deleted_eb) {
6250 memset(rec, 0,
6251 sizeof(struct ocfs2_extent_rec));
6252 le16_add_cpu(&el->l_next_free_rec, -1);
6253
6254 goto delete;
6255 }
6256
6257 /* Can this actually happen? */
6258 if (le16_to_cpu(el->l_next_free_rec) == 0)
6259 goto delete;
6260
6261 /*
6262 * We never actually deleted any clusters
6263 * because our leaf was empty. There's no
6264 * reason to adjust the rightmost edge then.
6265 */
6266 if (new_edge == 0)
6267 goto delete;
6268
Mark Fashehe48edee2007-03-07 16:46:57 -08006269 rec->e_int_clusters = cpu_to_le32(new_edge);
6270 le32_add_cpu(&rec->e_int_clusters,
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006271 -le32_to_cpu(rec->e_cpos));
6272
6273 /*
6274 * A deleted child record should have been
6275 * caught above.
6276 */
Mark Fashehe48edee2007-03-07 16:46:57 -08006277 BUG_ON(le32_to_cpu(rec->e_int_clusters) == 0);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006278 }
6279
6280delete:
6281 ret = ocfs2_journal_dirty(handle, bh);
6282 if (ret) {
6283 mlog_errno(ret);
6284 goto out;
6285 }
6286
6287 mlog(0, "extent list container %llu, after: record %d: "
6288 "(%u, %u, %llu), next = %u.\n",
6289 (unsigned long long)bh->b_blocknr, i,
Mark Fashehe48edee2007-03-07 16:46:57 -08006290 le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec),
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006291 (unsigned long long)le64_to_cpu(rec->e_blkno),
6292 le16_to_cpu(el->l_next_free_rec));
6293
6294 /*
6295 * We must be careful to only attempt delete of an
6296 * extent block (and not the root inode block).
6297 */
6298 if (index > 0 && le16_to_cpu(el->l_next_free_rec) == 0) {
6299 struct ocfs2_extent_block *eb =
6300 (struct ocfs2_extent_block *)bh->b_data;
6301
6302 /*
6303 * Save this for use when processing the
6304 * parent block.
6305 */
6306 deleted_eb = le64_to_cpu(eb->h_blkno);
6307
6308 mlog(0, "deleting this extent block.\n");
6309
6310 ocfs2_remove_from_cache(inode, bh);
6311
Mark Fashehe48edee2007-03-07 16:46:57 -08006312 BUG_ON(ocfs2_rec_clusters(el, &el->l_recs[0]));
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006313 BUG_ON(le32_to_cpu(el->l_recs[0].e_cpos));
6314 BUG_ON(le64_to_cpu(el->l_recs[0].e_blkno));
6315
Mark Fasheh59a5e412007-06-22 15:52:36 -07006316 ret = ocfs2_cache_extent_block_free(&tc->tc_dealloc, eb);
6317 /* An error here is not fatal. */
6318 if (ret < 0)
6319 mlog_errno(ret);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006320 } else {
6321 deleted_eb = 0;
6322 }
6323
6324 index--;
6325 }
6326
6327 ret = 0;
6328out:
6329 return ret;
6330}
6331
Mark Fashehccd979b2005-12-15 14:31:24 -08006332static int ocfs2_do_truncate(struct ocfs2_super *osb,
6333 unsigned int clusters_to_del,
6334 struct inode *inode,
6335 struct buffer_head *fe_bh,
Mark Fasheh1fabe142006-10-09 18:11:45 -07006336 handle_t *handle,
Mark Fashehdcd05382007-01-16 11:32:23 -08006337 struct ocfs2_truncate_context *tc,
6338 struct ocfs2_path *path)
Mark Fashehccd979b2005-12-15 14:31:24 -08006339{
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006340 int status;
Mark Fashehccd979b2005-12-15 14:31:24 -08006341 struct ocfs2_dinode *fe;
Mark Fashehccd979b2005-12-15 14:31:24 -08006342 struct ocfs2_extent_block *last_eb = NULL;
6343 struct ocfs2_extent_list *el;
Mark Fashehccd979b2005-12-15 14:31:24 -08006344 struct buffer_head *last_eb_bh = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -08006345 u64 delete_blk = 0;
6346
6347 fe = (struct ocfs2_dinode *) fe_bh->b_data;
6348
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006349 status = ocfs2_find_new_last_ext_blk(inode, clusters_to_del,
Mark Fashehdcd05382007-01-16 11:32:23 -08006350 path, &last_eb_bh);
Mark Fashehccd979b2005-12-15 14:31:24 -08006351 if (status < 0) {
6352 mlog_errno(status);
6353 goto bail;
6354 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006355
Mark Fashehdcd05382007-01-16 11:32:23 -08006356 /*
6357 * Each component will be touched, so we might as well journal
6358 * here to avoid having to handle errors later.
6359 */
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006360 status = ocfs2_journal_access_path(inode, handle, path);
6361 if (status < 0) {
6362 mlog_errno(status);
6363 goto bail;
Mark Fashehdcd05382007-01-16 11:32:23 -08006364 }
6365
6366 if (last_eb_bh) {
6367 status = ocfs2_journal_access(handle, inode, last_eb_bh,
6368 OCFS2_JOURNAL_ACCESS_WRITE);
6369 if (status < 0) {
6370 mlog_errno(status);
6371 goto bail;
6372 }
6373
6374 last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
6375 }
6376
6377 el = &(fe->id2.i_list);
6378
6379 /*
6380 * Lower levels depend on this never happening, but it's best
6381 * to check it up here before changing the tree.
6382 */
Mark Fashehe48edee2007-03-07 16:46:57 -08006383 if (el->l_tree_depth && el->l_recs[0].e_int_clusters == 0) {
Mark Fashehdcd05382007-01-16 11:32:23 -08006384 ocfs2_error(inode->i_sb,
6385 "Inode %lu has an empty extent record, depth %u\n",
6386 inode->i_ino, le16_to_cpu(el->l_tree_depth));
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006387 status = -EROFS;
Mark Fashehccd979b2005-12-15 14:31:24 -08006388 goto bail;
6389 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006390
6391 spin_lock(&OCFS2_I(inode)->ip_lock);
6392 OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) -
6393 clusters_to_del;
6394 spin_unlock(&OCFS2_I(inode)->ip_lock);
6395 le32_add_cpu(&fe->i_clusters, -clusters_to_del);
Mark Fashehe535e2e2007-08-31 10:23:41 -07006396 inode->i_blocks = ocfs2_inode_sector_count(inode);
Mark Fashehccd979b2005-12-15 14:31:24 -08006397
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006398 status = ocfs2_trim_tree(inode, path, handle, tc,
6399 clusters_to_del, &delete_blk);
6400 if (status) {
6401 mlog_errno(status);
6402 goto bail;
Mark Fashehccd979b2005-12-15 14:31:24 -08006403 }
6404
Mark Fashehdcd05382007-01-16 11:32:23 -08006405 if (le32_to_cpu(fe->i_clusters) == 0) {
Mark Fashehccd979b2005-12-15 14:31:24 -08006406 /* trunc to zero is a special case. */
6407 el->l_tree_depth = 0;
6408 fe->i_last_eb_blk = 0;
6409 } else if (last_eb)
6410 fe->i_last_eb_blk = last_eb->h_blkno;
6411
6412 status = ocfs2_journal_dirty(handle, fe_bh);
6413 if (status < 0) {
6414 mlog_errno(status);
6415 goto bail;
6416 }
6417
6418 if (last_eb) {
6419 /* If there will be a new last extent block, then by
6420 * definition, there cannot be any leaves to the right of
6421 * him. */
Mark Fashehccd979b2005-12-15 14:31:24 -08006422 last_eb->h_next_leaf_blk = 0;
6423 status = ocfs2_journal_dirty(handle, last_eb_bh);
6424 if (status < 0) {
6425 mlog_errno(status);
6426 goto bail;
6427 }
6428 }
6429
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006430 if (delete_blk) {
6431 status = ocfs2_truncate_log_append(osb, handle, delete_blk,
6432 clusters_to_del);
Mark Fashehccd979b2005-12-15 14:31:24 -08006433 if (status < 0) {
6434 mlog_errno(status);
6435 goto bail;
6436 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006437 }
6438 status = 0;
6439bail:
Mark Fashehdcd05382007-01-16 11:32:23 -08006440
Mark Fashehccd979b2005-12-15 14:31:24 -08006441 mlog_exit(status);
6442 return status;
6443}
6444
Mark Fasheh60b11392007-02-16 11:46:50 -08006445static int ocfs2_writeback_zero_func(handle_t *handle, struct buffer_head *bh)
6446{
6447 set_buffer_uptodate(bh);
6448 mark_buffer_dirty(bh);
6449 return 0;
6450}
6451
6452static int ocfs2_ordered_zero_func(handle_t *handle, struct buffer_head *bh)
6453{
6454 set_buffer_uptodate(bh);
6455 mark_buffer_dirty(bh);
6456 return ocfs2_journal_dirty_data(handle, bh);
6457}
6458
Mark Fasheh1d410a62007-09-07 14:20:45 -07006459static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
6460 unsigned int from, unsigned int to,
6461 struct page *page, int zero, u64 *phys)
6462{
6463 int ret, partial = 0;
6464
6465 ret = ocfs2_map_page_blocks(page, phys, inode, from, to, 0);
6466 if (ret)
6467 mlog_errno(ret);
6468
6469 if (zero)
Christoph Lametereebd2aa2008-02-04 22:28:29 -08006470 zero_user_segment(page, from, to);
Mark Fasheh1d410a62007-09-07 14:20:45 -07006471
6472 /*
6473 * Need to set the buffers we zero'd into uptodate
6474 * here if they aren't - ocfs2_map_page_blocks()
6475 * might've skipped some
6476 */
6477 if (ocfs2_should_order_data(inode)) {
6478 ret = walk_page_buffers(handle,
6479 page_buffers(page),
6480 from, to, &partial,
6481 ocfs2_ordered_zero_func);
6482 if (ret < 0)
6483 mlog_errno(ret);
6484 } else {
6485 ret = walk_page_buffers(handle, page_buffers(page),
6486 from, to, &partial,
6487 ocfs2_writeback_zero_func);
6488 if (ret < 0)
6489 mlog_errno(ret);
6490 }
6491
6492 if (!partial)
6493 SetPageUptodate(page);
6494
6495 flush_dcache_page(page);
6496}
6497
Mark Fasheh35edec12007-07-06 14:41:18 -07006498static void ocfs2_zero_cluster_pages(struct inode *inode, loff_t start,
6499 loff_t end, struct page **pages,
6500 int numpages, u64 phys, handle_t *handle)
Mark Fasheh60b11392007-02-16 11:46:50 -08006501{
Mark Fasheh1d410a62007-09-07 14:20:45 -07006502 int i;
Mark Fasheh60b11392007-02-16 11:46:50 -08006503 struct page *page;
6504 unsigned int from, to = PAGE_CACHE_SIZE;
6505 struct super_block *sb = inode->i_sb;
6506
6507 BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(sb)));
6508
6509 if (numpages == 0)
6510 goto out;
6511
Mark Fasheh35edec12007-07-06 14:41:18 -07006512 to = PAGE_CACHE_SIZE;
Mark Fasheh60b11392007-02-16 11:46:50 -08006513 for(i = 0; i < numpages; i++) {
6514 page = pages[i];
6515
Mark Fasheh35edec12007-07-06 14:41:18 -07006516 from = start & (PAGE_CACHE_SIZE - 1);
6517 if ((end >> PAGE_CACHE_SHIFT) == page->index)
6518 to = end & (PAGE_CACHE_SIZE - 1);
6519
Mark Fasheh60b11392007-02-16 11:46:50 -08006520 BUG_ON(from > PAGE_CACHE_SIZE);
6521 BUG_ON(to > PAGE_CACHE_SIZE);
6522
Mark Fasheh1d410a62007-09-07 14:20:45 -07006523 ocfs2_map_and_dirty_page(inode, handle, from, to, page, 1,
6524 &phys);
Mark Fasheh60b11392007-02-16 11:46:50 -08006525
Mark Fasheh35edec12007-07-06 14:41:18 -07006526 start = (page->index + 1) << PAGE_CACHE_SHIFT;
Mark Fasheh60b11392007-02-16 11:46:50 -08006527 }
6528out:
Mark Fasheh1d410a62007-09-07 14:20:45 -07006529 if (pages)
6530 ocfs2_unlock_and_free_pages(pages, numpages);
Mark Fasheh60b11392007-02-16 11:46:50 -08006531}
6532
Mark Fasheh35edec12007-07-06 14:41:18 -07006533static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end,
Mark Fasheh1d410a62007-09-07 14:20:45 -07006534 struct page **pages, int *num)
Mark Fasheh60b11392007-02-16 11:46:50 -08006535{
Mark Fasheh1d410a62007-09-07 14:20:45 -07006536 int numpages, ret = 0;
Mark Fasheh60b11392007-02-16 11:46:50 -08006537 struct super_block *sb = inode->i_sb;
6538 struct address_space *mapping = inode->i_mapping;
6539 unsigned long index;
Mark Fasheh35edec12007-07-06 14:41:18 -07006540 loff_t last_page_bytes;
Mark Fasheh60b11392007-02-16 11:46:50 -08006541
Mark Fasheh35edec12007-07-06 14:41:18 -07006542 BUG_ON(start > end);
Mark Fasheh60b11392007-02-16 11:46:50 -08006543
Mark Fasheh35edec12007-07-06 14:41:18 -07006544 BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits !=
6545 (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits);
6546
Mark Fasheh1d410a62007-09-07 14:20:45 -07006547 numpages = 0;
Mark Fasheh35edec12007-07-06 14:41:18 -07006548 last_page_bytes = PAGE_ALIGN(end);
6549 index = start >> PAGE_CACHE_SHIFT;
Mark Fasheh60b11392007-02-16 11:46:50 -08006550 do {
6551 pages[numpages] = grab_cache_page(mapping, index);
6552 if (!pages[numpages]) {
6553 ret = -ENOMEM;
6554 mlog_errno(ret);
6555 goto out;
6556 }
6557
6558 numpages++;
6559 index++;
Mark Fasheh35edec12007-07-06 14:41:18 -07006560 } while (index < (last_page_bytes >> PAGE_CACHE_SHIFT));
Mark Fasheh60b11392007-02-16 11:46:50 -08006561
6562out:
6563 if (ret != 0) {
Mark Fasheh1d410a62007-09-07 14:20:45 -07006564 if (pages)
6565 ocfs2_unlock_and_free_pages(pages, numpages);
Mark Fasheh60b11392007-02-16 11:46:50 -08006566 numpages = 0;
6567 }
6568
6569 *num = numpages;
6570
6571 return ret;
6572}
6573
6574/*
6575 * Zero the area past i_size but still within an allocated
6576 * cluster. This avoids exposing nonzero data on subsequent file
6577 * extends.
6578 *
6579 * We need to call this before i_size is updated on the inode because
6580 * otherwise block_write_full_page() will skip writeout of pages past
6581 * i_size. The new_i_size parameter is passed for this reason.
6582 */
Mark Fasheh35edec12007-07-06 14:41:18 -07006583int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
6584 u64 range_start, u64 range_end)
Mark Fasheh60b11392007-02-16 11:46:50 -08006585{
Mark Fasheh1d410a62007-09-07 14:20:45 -07006586 int ret = 0, numpages;
Mark Fasheh60b11392007-02-16 11:46:50 -08006587 struct page **pages = NULL;
6588 u64 phys;
Mark Fasheh1d410a62007-09-07 14:20:45 -07006589 unsigned int ext_flags;
6590 struct super_block *sb = inode->i_sb;
Mark Fasheh60b11392007-02-16 11:46:50 -08006591
6592 /*
6593 * File systems which don't support sparse files zero on every
6594 * extend.
6595 */
Mark Fasheh1d410a62007-09-07 14:20:45 -07006596 if (!ocfs2_sparse_alloc(OCFS2_SB(sb)))
Mark Fasheh60b11392007-02-16 11:46:50 -08006597 return 0;
6598
Mark Fasheh1d410a62007-09-07 14:20:45 -07006599 pages = kcalloc(ocfs2_pages_per_cluster(sb),
Mark Fasheh60b11392007-02-16 11:46:50 -08006600 sizeof(struct page *), GFP_NOFS);
6601 if (pages == NULL) {
6602 ret = -ENOMEM;
6603 mlog_errno(ret);
6604 goto out;
6605 }
6606
Mark Fasheh1d410a62007-09-07 14:20:45 -07006607 if (range_start == range_end)
6608 goto out;
6609
6610 ret = ocfs2_extent_map_get_blocks(inode,
6611 range_start >> sb->s_blocksize_bits,
6612 &phys, NULL, &ext_flags);
Mark Fasheh60b11392007-02-16 11:46:50 -08006613 if (ret) {
6614 mlog_errno(ret);
6615 goto out;
6616 }
6617
Mark Fasheh1d410a62007-09-07 14:20:45 -07006618 /*
6619 * Tail is a hole, or is marked unwritten. In either case, we
6620 * can count on read and write to return/push zero's.
6621 */
6622 if (phys == 0 || ext_flags & OCFS2_EXT_UNWRITTEN)
Mark Fasheh60b11392007-02-16 11:46:50 -08006623 goto out;
6624
Mark Fasheh1d410a62007-09-07 14:20:45 -07006625 ret = ocfs2_grab_eof_pages(inode, range_start, range_end, pages,
6626 &numpages);
6627 if (ret) {
6628 mlog_errno(ret);
6629 goto out;
6630 }
6631
Mark Fasheh35edec12007-07-06 14:41:18 -07006632 ocfs2_zero_cluster_pages(inode, range_start, range_end, pages,
6633 numpages, phys, handle);
Mark Fasheh60b11392007-02-16 11:46:50 -08006634
6635 /*
6636 * Initiate writeout of the pages we zero'd here. We don't
6637 * wait on them - the truncate_inode_pages() call later will
6638 * do that for us.
6639 */
Mark Fasheh35edec12007-07-06 14:41:18 -07006640 ret = do_sync_mapping_range(inode->i_mapping, range_start,
6641 range_end - 1, SYNC_FILE_RANGE_WRITE);
Mark Fasheh60b11392007-02-16 11:46:50 -08006642 if (ret)
6643 mlog_errno(ret);
6644
6645out:
6646 if (pages)
6647 kfree(pages);
6648
6649 return ret;
6650}
6651
Tiger Yangfdd77702008-08-18 17:08:55 +08006652static void ocfs2_zero_dinode_id2_with_xattr(struct inode *inode,
6653 struct ocfs2_dinode *di)
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006654{
6655 unsigned int blocksize = 1 << inode->i_sb->s_blocksize_bits;
Tiger Yangfdd77702008-08-18 17:08:55 +08006656 unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006657
Tiger Yangfdd77702008-08-18 17:08:55 +08006658 if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
6659 memset(&di->id2, 0, blocksize -
6660 offsetof(struct ocfs2_dinode, id2) -
6661 xattrsize);
6662 else
6663 memset(&di->id2, 0, blocksize -
6664 offsetof(struct ocfs2_dinode, id2));
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006665}
6666
Mark Fasheh5b6a3a22007-09-13 16:33:54 -07006667void ocfs2_dinode_new_extent_list(struct inode *inode,
6668 struct ocfs2_dinode *di)
6669{
Tiger Yangfdd77702008-08-18 17:08:55 +08006670 ocfs2_zero_dinode_id2_with_xattr(inode, di);
Mark Fasheh5b6a3a22007-09-13 16:33:54 -07006671 di->id2.i_list.l_tree_depth = 0;
6672 di->id2.i_list.l_next_free_rec = 0;
Tiger Yangfdd77702008-08-18 17:08:55 +08006673 di->id2.i_list.l_count = cpu_to_le16(
6674 ocfs2_extent_recs_per_inode_with_xattr(inode->i_sb, di));
Mark Fasheh5b6a3a22007-09-13 16:33:54 -07006675}
6676
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006677void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di)
6678{
6679 struct ocfs2_inode_info *oi = OCFS2_I(inode);
6680 struct ocfs2_inline_data *idata = &di->id2.i_data;
6681
6682 spin_lock(&oi->ip_lock);
6683 oi->ip_dyn_features |= OCFS2_INLINE_DATA_FL;
6684 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
6685 spin_unlock(&oi->ip_lock);
6686
6687 /*
6688 * We clear the entire i_data structure here so that all
6689 * fields can be properly initialized.
6690 */
Tiger Yangfdd77702008-08-18 17:08:55 +08006691 ocfs2_zero_dinode_id2_with_xattr(inode, di);
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006692
Tiger Yangfdd77702008-08-18 17:08:55 +08006693 idata->id_count = cpu_to_le16(
6694 ocfs2_max_inline_data_with_xattr(inode->i_sb, di));
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006695}
6696
6697int ocfs2_convert_inline_data_to_extents(struct inode *inode,
6698 struct buffer_head *di_bh)
6699{
6700 int ret, i, has_data, num_pages = 0;
6701 handle_t *handle;
6702 u64 uninitialized_var(block);
6703 struct ocfs2_inode_info *oi = OCFS2_I(inode);
6704 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
6705 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006706 struct ocfs2_alloc_context *data_ac = NULL;
6707 struct page **pages = NULL;
6708 loff_t end = osb->s_clustersize;
6709
6710 has_data = i_size_read(inode) ? 1 : 0;
6711
6712 if (has_data) {
6713 pages = kcalloc(ocfs2_pages_per_cluster(osb->sb),
6714 sizeof(struct page *), GFP_NOFS);
6715 if (pages == NULL) {
6716 ret = -ENOMEM;
6717 mlog_errno(ret);
6718 goto out;
6719 }
6720
6721 ret = ocfs2_reserve_clusters(osb, 1, &data_ac);
6722 if (ret) {
6723 mlog_errno(ret);
6724 goto out;
6725 }
6726 }
6727
6728 handle = ocfs2_start_trans(osb, OCFS2_INLINE_TO_EXTENTS_CREDITS);
6729 if (IS_ERR(handle)) {
6730 ret = PTR_ERR(handle);
6731 mlog_errno(ret);
6732 goto out_unlock;
6733 }
6734
6735 ret = ocfs2_journal_access(handle, inode, di_bh,
6736 OCFS2_JOURNAL_ACCESS_WRITE);
6737 if (ret) {
6738 mlog_errno(ret);
6739 goto out_commit;
6740 }
6741
6742 if (has_data) {
6743 u32 bit_off, num;
6744 unsigned int page_end;
6745 u64 phys;
6746
6747 ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
6748 &num);
6749 if (ret) {
6750 mlog_errno(ret);
6751 goto out_commit;
6752 }
6753
6754 /*
6755 * Save two copies, one for insert, and one that can
6756 * be changed by ocfs2_map_and_dirty_page() below.
6757 */
6758 block = phys = ocfs2_clusters_to_blocks(inode->i_sb, bit_off);
6759
6760 /*
6761 * Non sparse file systems zero on extend, so no need
6762 * to do that now.
6763 */
6764 if (!ocfs2_sparse_alloc(osb) &&
6765 PAGE_CACHE_SIZE < osb->s_clustersize)
6766 end = PAGE_CACHE_SIZE;
6767
6768 ret = ocfs2_grab_eof_pages(inode, 0, end, pages, &num_pages);
6769 if (ret) {
6770 mlog_errno(ret);
6771 goto out_commit;
6772 }
6773
6774 /*
6775 * This should populate the 1st page for us and mark
6776 * it up to date.
6777 */
6778 ret = ocfs2_read_inline_data(inode, pages[0], di_bh);
6779 if (ret) {
6780 mlog_errno(ret);
6781 goto out_commit;
6782 }
6783
6784 page_end = PAGE_CACHE_SIZE;
6785 if (PAGE_CACHE_SIZE > osb->s_clustersize)
6786 page_end = osb->s_clustersize;
6787
6788 for (i = 0; i < num_pages; i++)
6789 ocfs2_map_and_dirty_page(inode, handle, 0, page_end,
6790 pages[i], i > 0, &phys);
6791 }
6792
6793 spin_lock(&oi->ip_lock);
6794 oi->ip_dyn_features &= ~OCFS2_INLINE_DATA_FL;
6795 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
6796 spin_unlock(&oi->ip_lock);
6797
Mark Fasheh5b6a3a22007-09-13 16:33:54 -07006798 ocfs2_dinode_new_extent_list(inode, di);
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006799
6800 ocfs2_journal_dirty(handle, di_bh);
6801
6802 if (has_data) {
6803 /*
6804 * An error at this point should be extremely rare. If
6805 * this proves to be false, we could always re-build
6806 * the in-inode data from our pages.
6807 */
Tao Maf56654c2008-08-18 17:38:48 +08006808 ret = ocfs2_dinode_insert_extent(osb, handle, inode, di_bh,
6809 0, block, 1, 0, NULL);
Mark Fasheh1afc32b2007-09-07 14:46:51 -07006810 if (ret) {
6811 mlog_errno(ret);
6812 goto out_commit;
6813 }
6814
6815 inode->i_blocks = ocfs2_inode_sector_count(inode);
6816 }
6817
6818out_commit:
6819 ocfs2_commit_trans(osb, handle);
6820
6821out_unlock:
6822 if (data_ac)
6823 ocfs2_free_alloc_context(data_ac);
6824
6825out:
6826 if (pages) {
6827 ocfs2_unlock_and_free_pages(pages, num_pages);
6828 kfree(pages);
6829 }
6830
6831 return ret;
6832}
6833
Mark Fashehccd979b2005-12-15 14:31:24 -08006834/*
6835 * It is expected, that by the time you call this function,
6836 * inode->i_size and fe->i_size have been adjusted.
6837 *
6838 * WARNING: This will kfree the truncate context
6839 */
6840int ocfs2_commit_truncate(struct ocfs2_super *osb,
6841 struct inode *inode,
6842 struct buffer_head *fe_bh,
6843 struct ocfs2_truncate_context *tc)
6844{
6845 int status, i, credits, tl_sem = 0;
Mark Fashehdcd05382007-01-16 11:32:23 -08006846 u32 clusters_to_del, new_highest_cpos, range;
Mark Fashehccd979b2005-12-15 14:31:24 -08006847 struct ocfs2_extent_list *el;
Mark Fasheh1fabe142006-10-09 18:11:45 -07006848 handle_t *handle = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -08006849 struct inode *tl_inode = osb->osb_tl_inode;
Mark Fashehdcd05382007-01-16 11:32:23 -08006850 struct ocfs2_path *path = NULL;
Tao Mae7d4cb62008-08-18 17:38:44 +08006851 struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data;
Mark Fashehccd979b2005-12-15 14:31:24 -08006852
6853 mlog_entry_void();
6854
Mark Fashehdcd05382007-01-16 11:32:23 -08006855 new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb,
Mark Fashehccd979b2005-12-15 14:31:24 -08006856 i_size_read(inode));
6857
Tao Mae7d4cb62008-08-18 17:38:44 +08006858 path = ocfs2_new_path(fe_bh, &di->id2.i_list);
Mark Fashehdcd05382007-01-16 11:32:23 -08006859 if (!path) {
6860 status = -ENOMEM;
6861 mlog_errno(status);
6862 goto bail;
6863 }
Mark Fasheh83418972007-04-23 18:53:12 -07006864
6865 ocfs2_extent_map_trunc(inode, new_highest_cpos);
6866
Mark Fashehccd979b2005-12-15 14:31:24 -08006867start:
Mark Fashehdcd05382007-01-16 11:32:23 -08006868 /*
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006869 * Check that we still have allocation to delete.
6870 */
6871 if (OCFS2_I(inode)->ip_clusters == 0) {
6872 status = 0;
6873 goto bail;
6874 }
6875
6876 /*
Mark Fashehdcd05382007-01-16 11:32:23 -08006877 * Truncate always works against the rightmost tree branch.
6878 */
6879 status = ocfs2_find_path(inode, path, UINT_MAX);
6880 if (status) {
6881 mlog_errno(status);
6882 goto bail;
Mark Fashehccd979b2005-12-15 14:31:24 -08006883 }
6884
Mark Fashehdcd05382007-01-16 11:32:23 -08006885 mlog(0, "inode->ip_clusters = %u, tree_depth = %u\n",
6886 OCFS2_I(inode)->ip_clusters, path->p_tree_depth);
6887
6888 /*
6889 * By now, el will point to the extent list on the bottom most
6890 * portion of this tree. Only the tail record is considered in
6891 * each pass.
6892 *
6893 * We handle the following cases, in order:
6894 * - empty extent: delete the remaining branch
6895 * - remove the entire record
6896 * - remove a partial record
6897 * - no record needs to be removed (truncate has completed)
6898 */
6899 el = path_leaf_el(path);
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006900 if (le16_to_cpu(el->l_next_free_rec) == 0) {
6901 ocfs2_error(inode->i_sb,
6902 "Inode %llu has empty extent block at %llu\n",
6903 (unsigned long long)OCFS2_I(inode)->ip_blkno,
6904 (unsigned long long)path_leaf_bh(path)->b_blocknr);
6905 status = -EROFS;
6906 goto bail;
6907 }
6908
Mark Fashehccd979b2005-12-15 14:31:24 -08006909 i = le16_to_cpu(el->l_next_free_rec) - 1;
Mark Fashehdcd05382007-01-16 11:32:23 -08006910 range = le32_to_cpu(el->l_recs[i].e_cpos) +
Mark Fashehe48edee2007-03-07 16:46:57 -08006911 ocfs2_rec_clusters(el, &el->l_recs[i]);
Mark Fashehdcd05382007-01-16 11:32:23 -08006912 if (i == 0 && ocfs2_is_empty_extent(&el->l_recs[i])) {
6913 clusters_to_del = 0;
6914 } else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) {
Mark Fashehe48edee2007-03-07 16:46:57 -08006915 clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]);
Mark Fashehdcd05382007-01-16 11:32:23 -08006916 } else if (range > new_highest_cpos) {
Mark Fashehe48edee2007-03-07 16:46:57 -08006917 clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) +
Mark Fashehccd979b2005-12-15 14:31:24 -08006918 le32_to_cpu(el->l_recs[i].e_cpos)) -
Mark Fashehdcd05382007-01-16 11:32:23 -08006919 new_highest_cpos;
6920 } else {
6921 status = 0;
6922 goto bail;
6923 }
Mark Fashehccd979b2005-12-15 14:31:24 -08006924
Mark Fashehdcd05382007-01-16 11:32:23 -08006925 mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n",
6926 clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr);
6927
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08006928 mutex_lock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08006929 tl_sem = 1;
6930 /* ocfs2_truncate_log_needs_flush guarantees us at least one
6931 * record is free for use. If there isn't any, we flush to get
6932 * an empty truncate log. */
6933 if (ocfs2_truncate_log_needs_flush(osb)) {
6934 status = __ocfs2_flush_truncate_log(osb);
6935 if (status < 0) {
6936 mlog_errno(status);
6937 goto bail;
6938 }
6939 }
6940
6941 credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
Mark Fashehdcd05382007-01-16 11:32:23 -08006942 (struct ocfs2_dinode *)fe_bh->b_data,
6943 el);
Mark Fasheh65eff9c2006-10-09 17:26:22 -07006944 handle = ocfs2_start_trans(osb, credits);
Mark Fashehccd979b2005-12-15 14:31:24 -08006945 if (IS_ERR(handle)) {
6946 status = PTR_ERR(handle);
6947 handle = NULL;
6948 mlog_errno(status);
6949 goto bail;
6950 }
6951
Mark Fashehdcd05382007-01-16 11:32:23 -08006952 status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle,
6953 tc, path);
Mark Fashehccd979b2005-12-15 14:31:24 -08006954 if (status < 0) {
6955 mlog_errno(status);
6956 goto bail;
6957 }
6958
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08006959 mutex_unlock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08006960 tl_sem = 0;
6961
Mark Fasheh02dc1af2006-10-09 16:48:10 -07006962 ocfs2_commit_trans(osb, handle);
Mark Fashehccd979b2005-12-15 14:31:24 -08006963 handle = NULL;
6964
Mark Fashehdcd05382007-01-16 11:32:23 -08006965 ocfs2_reinit_path(path, 1);
6966
6967 /*
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006968 * The check above will catch the case where we've truncated
6969 * away all allocation.
Mark Fashehdcd05382007-01-16 11:32:23 -08006970 */
Mark Fasheh3a0782d2007-01-17 12:53:31 -08006971 goto start;
6972
Mark Fashehccd979b2005-12-15 14:31:24 -08006973bail:
Mark Fashehccd979b2005-12-15 14:31:24 -08006974
6975 ocfs2_schedule_truncate_log_flush(osb, 1);
6976
6977 if (tl_sem)
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08006978 mutex_unlock(&tl_inode->i_mutex);
Mark Fashehccd979b2005-12-15 14:31:24 -08006979
6980 if (handle)
Mark Fasheh02dc1af2006-10-09 16:48:10 -07006981 ocfs2_commit_trans(osb, handle);
Mark Fashehccd979b2005-12-15 14:31:24 -08006982
Mark Fasheh59a5e412007-06-22 15:52:36 -07006983 ocfs2_run_deallocs(osb, &tc->tc_dealloc);
6984
Mark Fashehdcd05382007-01-16 11:32:23 -08006985 ocfs2_free_path(path);
Mark Fashehccd979b2005-12-15 14:31:24 -08006986
6987 /* This will drop the ext_alloc cluster lock for us */
6988 ocfs2_free_truncate_context(tc);
6989
6990 mlog_exit(status);
6991 return status;
6992}
6993
Mark Fashehccd979b2005-12-15 14:31:24 -08006994/*
Mark Fasheh59a5e412007-06-22 15:52:36 -07006995 * Expects the inode to already be locked.
Mark Fashehccd979b2005-12-15 14:31:24 -08006996 */
6997int ocfs2_prepare_truncate(struct ocfs2_super *osb,
6998 struct inode *inode,
6999 struct buffer_head *fe_bh,
7000 struct ocfs2_truncate_context **tc)
7001{
Mark Fasheh59a5e412007-06-22 15:52:36 -07007002 int status;
Mark Fashehccd979b2005-12-15 14:31:24 -08007003 unsigned int new_i_clusters;
7004 struct ocfs2_dinode *fe;
7005 struct ocfs2_extent_block *eb;
Mark Fashehccd979b2005-12-15 14:31:24 -08007006 struct buffer_head *last_eb_bh = NULL;
Mark Fashehccd979b2005-12-15 14:31:24 -08007007
7008 mlog_entry_void();
7009
7010 *tc = NULL;
7011
7012 new_i_clusters = ocfs2_clusters_for_bytes(osb->sb,
7013 i_size_read(inode));
7014 fe = (struct ocfs2_dinode *) fe_bh->b_data;
7015
7016 mlog(0, "fe->i_clusters = %u, new_i_clusters = %u, fe->i_size ="
Mark Fasheh1ca1a112007-04-27 16:01:25 -07007017 "%llu\n", le32_to_cpu(fe->i_clusters), new_i_clusters,
7018 (unsigned long long)le64_to_cpu(fe->i_size));
Mark Fashehccd979b2005-12-15 14:31:24 -08007019
Robert P. J. Daycd861282006-12-13 00:34:52 -08007020 *tc = kzalloc(sizeof(struct ocfs2_truncate_context), GFP_KERNEL);
Mark Fashehccd979b2005-12-15 14:31:24 -08007021 if (!(*tc)) {
7022 status = -ENOMEM;
7023 mlog_errno(status);
7024 goto bail;
7025 }
Mark Fasheh59a5e412007-06-22 15:52:36 -07007026 ocfs2_init_dealloc_ctxt(&(*tc)->tc_dealloc);
Mark Fashehccd979b2005-12-15 14:31:24 -08007027
Mark Fashehccd979b2005-12-15 14:31:24 -08007028 if (fe->id2.i_list.l_tree_depth) {
Mark Fashehccd979b2005-12-15 14:31:24 -08007029 status = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
7030 &last_eb_bh, OCFS2_BH_CACHED, inode);
7031 if (status < 0) {
7032 mlog_errno(status);
7033 goto bail;
7034 }
7035 eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
7036 if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
7037 OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
7038
7039 brelse(last_eb_bh);
7040 status = -EIO;
7041 goto bail;
7042 }
Mark Fashehccd979b2005-12-15 14:31:24 -08007043 }
7044
7045 (*tc)->tc_last_eb_bh = last_eb_bh;
7046
Mark Fashehccd979b2005-12-15 14:31:24 -08007047 status = 0;
7048bail:
7049 if (status < 0) {
7050 if (*tc)
7051 ocfs2_free_truncate_context(*tc);
7052 *tc = NULL;
7053 }
7054 mlog_exit_void();
7055 return status;
7056}
7057
Mark Fasheh1afc32b2007-09-07 14:46:51 -07007058/*
7059 * 'start' is inclusive, 'end' is not.
7060 */
7061int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
7062 unsigned int start, unsigned int end, int trunc)
7063{
7064 int ret;
7065 unsigned int numbytes;
7066 handle_t *handle;
7067 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
7068 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
7069 struct ocfs2_inline_data *idata = &di->id2.i_data;
7070
7071 if (end > i_size_read(inode))
7072 end = i_size_read(inode);
7073
7074 BUG_ON(start >= end);
7075
7076 if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) ||
7077 !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) ||
7078 !ocfs2_supports_inline_data(osb)) {
7079 ocfs2_error(inode->i_sb,
7080 "Inline data flags for inode %llu don't agree! "
7081 "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x\n",
7082 (unsigned long long)OCFS2_I(inode)->ip_blkno,
7083 le16_to_cpu(di->i_dyn_features),
7084 OCFS2_I(inode)->ip_dyn_features,
7085 osb->s_feature_incompat);
7086 ret = -EROFS;
7087 goto out;
7088 }
7089
7090 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
7091 if (IS_ERR(handle)) {
7092 ret = PTR_ERR(handle);
7093 mlog_errno(ret);
7094 goto out;
7095 }
7096
7097 ret = ocfs2_journal_access(handle, inode, di_bh,
7098 OCFS2_JOURNAL_ACCESS_WRITE);
7099 if (ret) {
7100 mlog_errno(ret);
7101 goto out_commit;
7102 }
7103
7104 numbytes = end - start;
7105 memset(idata->id_data + start, 0, numbytes);
7106
7107 /*
7108 * No need to worry about the data page here - it's been
7109 * truncated already and inline data doesn't need it for
7110 * pushing zero's to disk, so we'll let readpage pick it up
7111 * later.
7112 */
7113 if (trunc) {
7114 i_size_write(inode, start);
7115 di->i_size = cpu_to_le64(start);
7116 }
7117
7118 inode->i_blocks = ocfs2_inode_sector_count(inode);
7119 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
7120
7121 di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
7122 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
7123
7124 ocfs2_journal_dirty(handle, di_bh);
7125
7126out_commit:
7127 ocfs2_commit_trans(osb, handle);
7128
7129out:
7130 return ret;
7131}
7132
Mark Fashehccd979b2005-12-15 14:31:24 -08007133static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc)
7134{
Mark Fasheh59a5e412007-06-22 15:52:36 -07007135 /*
7136 * The caller is responsible for completing deallocation
7137 * before freeing the context.
7138 */
7139 if (tc->tc_dealloc.c_first_suballocator != NULL)
7140 mlog(ML_NOTICE,
7141 "Truncate completion has non-empty dealloc context\n");
Mark Fashehccd979b2005-12-15 14:31:24 -08007142
7143 if (tc->tc_last_eb_bh)
7144 brelse(tc->tc_last_eb_bh);
7145
7146 kfree(tc);
7147}