blob: 57917932212eb33d68a0f626766c79fb608c442a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/time.h>
6#include <linux/fs.h>
7#include <linux/reiserfs_fs.h>
8#include <linux/reiserfs_acl.h>
9#include <linux/reiserfs_xattr.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070010#include <linux/exportfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/smp_lock.h>
12#include <linux/pagemap.h>
13#include <linux/highmem.h>
14#include <asm/uaccess.h>
15#include <asm/unaligned.h>
16#include <linux/buffer_head.h>
17#include <linux/mpage.h>
18#include <linux/writeback.h>
19#include <linux/quotaops.h>
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -070020#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -070022int reiserfs_commit_write(struct file *f, struct page *page,
23 unsigned from, unsigned to);
24int reiserfs_prepare_write(struct file *f, struct page *page,
25 unsigned from, unsigned to);
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070027void reiserfs_delete_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -070028{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070029 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
30 int jbegin_count =
31 JOURNAL_PER_BALANCE_CNT * 2 +
32 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
33 struct reiserfs_transaction_handle th;
Jeff Mahoney24996042005-12-14 14:38:05 -050034 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Mark Fashehfef26652005-09-09 13:01:31 -070036 truncate_inode_pages(&inode->i_data, 0);
37
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070038 reiserfs_write_lock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070040 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
41 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070042 reiserfs_delete_xattrs(inode);
43
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070044 if (journal_begin(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070045 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070046 reiserfs_update_inode_transaction(inode);
47
Jeff Mahoney24996042005-12-14 14:38:05 -050048 err = reiserfs_delete_object(&th, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070049
50 /* Do quota update inside a transaction for journaled quotas. We must do that
51 * after delete_object so that quota updates go into the same transaction as
52 * stat data deletion */
Jeff Mahoney24996042005-12-14 14:38:05 -050053 if (!err)
54 DQUOT_FREE_INODE(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070055
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070056 if (journal_end(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070057 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070058
Jeff Mahoney24996042005-12-14 14:38:05 -050059 /* check return value from reiserfs_delete_object after
60 * ending the transaction
61 */
62 if (err)
63 goto out;
64
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070065 /* all items of file are deleted, so we can remove "save" link */
66 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
67 * about an error here */
68 } else {
69 /* no object items are in the tree */
70 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070072 out:
73 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
74 inode->i_blocks = 0;
75 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070076}
77
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070078static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
79 __u32 objectid, loff_t offset, int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070080{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070081 key->version = version;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070083 key->on_disk_key.k_dir_id = dirid;
84 key->on_disk_key.k_objectid = objectid;
85 set_cpu_key_k_offset(key, offset);
86 set_cpu_key_k_type(key, type);
87 key->key_length = length;
Linus Torvalds1da177e2005-04-16 15:20:36 -070088}
89
Linus Torvalds1da177e2005-04-16 15:20:36 -070090/* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
91 offset and type of key */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070092void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
93 int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070094{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070095 _make_cpu_key(key, get_inode_item_key_version(inode),
96 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
97 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
98 length);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099}
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101//
102// when key is 0, do not set version and short key
103//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700104inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
105 int version,
106 loff_t offset, int type, int length,
107 int entry_count /*or ih_free_space */ )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700109 if (key) {
110 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
111 ih->ih_key.k_objectid =
112 cpu_to_le32(key->on_disk_key.k_objectid);
113 }
114 put_ih_version(ih, version);
115 set_le_ih_k_offset(ih, offset);
116 set_le_ih_k_type(ih, type);
117 put_ih_item_len(ih, length);
118 /* set_ih_free_space (ih, 0); */
119 // for directory items it is entry count, for directs and stat
120 // datas - 0xffff, for indirects - 0
121 put_ih_entry_count(ih, entry_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122}
123
124//
125// FIXME: we might cache recently accessed indirect item
126
127// Ugh. Not too eager for that....
128// I cut the code until such time as I see a convincing argument (benchmark).
129// I don't want a bloated inode struct..., and I don't like code complexity....
130
131/* cutting the code is fine, since it really isn't in use yet and is easy
132** to add back in. But, Vladimir has a really good idea here. Think
133** about what happens for reading a file. For each page,
134** The VFS layer calls reiserfs_readpage, who searches the tree to find
135** an indirect item. This indirect item has X number of pointers, where
136** X is a big number if we've done the block allocation right. But,
137** we only use one or two of these pointers during each call to readpage,
138** needlessly researching again later on.
139**
140** The size of the cache could be dynamic based on the size of the file.
141**
142** I'd also like to see us cache the location the stat data item, since
143** we are needlessly researching for that frequently.
144**
145** --chris
146*/
147
148/* If this page has a file tail in it, and
149** it was read in by get_block_create_0, the page data is valid,
150** but tail is still sitting in a direct item, and we can't write to
151** it. So, look through this page, and check all the mapped buffers
152** to make sure they have valid block numbers. Any that don't need
153** to be unmapped, so that block_prepare_write will correctly call
154** reiserfs_get_block to convert the tail into an unformatted node
155*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700156static inline void fix_tail_page_for_writing(struct page *page)
157{
158 struct buffer_head *head, *next, *bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700160 if (page && page_has_buffers(page)) {
161 head = page_buffers(page);
162 bh = head;
163 do {
164 next = bh->b_this_page;
165 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
166 reiserfs_unmap_buffer(bh);
167 }
168 bh = next;
169 } while (bh != head);
170 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171}
172
173/* reiserfs_get_block does not need to allocate a block only if it has been
174 done already or non-hole position has been found in the indirect item */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700175static inline int allocation_needed(int retval, b_blocknr_t allocated,
176 struct item_head *ih,
177 __le32 * item, int pos_in_item)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700179 if (allocated)
180 return 0;
181 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
182 get_block_num(item, pos_in_item))
183 return 0;
184 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185}
186
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700187static inline int indirect_item_found(int retval, struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700189 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190}
191
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700192static inline void set_block_dev_mapped(struct buffer_head *bh,
193 b_blocknr_t block, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194{
195 map_bh(bh, inode->i_sb, block);
196}
197
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198//
199// files which were created in the earlier version can not be longer,
200// than 2 gb
201//
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700202static int file_capable(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700204 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
205 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
206 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700208 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209}
210
Adrian Bunkdeba0f42007-10-16 23:26:03 -0700211static int restart_transaction(struct reiserfs_transaction_handle *th,
212 struct inode *inode, struct treepath *path)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700213{
214 struct super_block *s = th->t_super;
215 int len = th->t_blocks_allocated;
216 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700218 BUG_ON(!th->t_trans_id);
219 BUG_ON(!th->t_refcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Suzuki K P87b41262006-12-06 20:36:10 -0800221 pathrelse(path);
222
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700223 /* we cannot restart while nested */
224 if (th->t_refcount > 1) {
225 return 0;
226 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700227 reiserfs_update_sd(th, inode);
228 err = journal_end(th, s, len);
229 if (!err) {
230 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
231 if (!err)
232 reiserfs_update_inode_transaction(inode);
233 }
234 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235}
236
237// it is called by get_block when create == 0. Returns block number
238// for 'block'-th logical block of file. When it hits direct item it
239// returns 0 (being called from bmap) or read direct item into piece
240// of page (bh_result)
241
242// Please improve the english/clarity in the comment above, as it is
243// hard to understand.
244
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700245static int _get_block_create_0(struct inode *inode, sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700246 struct buffer_head *bh_result, int args)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700248 INITIALIZE_PATH(path);
249 struct cpu_key key;
250 struct buffer_head *bh;
251 struct item_head *ih, tmp_ih;
252 int fs_gen;
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700253 b_blocknr_t blocknr;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700254 char *p = NULL;
255 int chars;
256 int ret;
257 int result;
258 int done = 0;
259 unsigned long offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700261 // prepare the key to look for the 'block'-th block of file
262 make_cpu_key(&key, inode,
263 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
264 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700266 research:
267 result = search_for_position_by_key(inode->i_sb, &key, &path);
268 if (result != POSITION_FOUND) {
269 pathrelse(&path);
270 if (p)
271 kunmap(bh_result->b_page);
272 if (result == IO_ERROR)
273 return -EIO;
274 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
275 // That there is some MMAPED data associated with it that is yet to be written to disk.
276 if ((args & GET_BLOCK_NO_HOLE)
277 && !PageUptodate(bh_result->b_page)) {
278 return -ENOENT;
279 }
280 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700282 //
283 bh = get_last_bh(&path);
284 ih = get_ih(&path);
285 if (is_indirect_le_ih(ih)) {
286 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700288 /* FIXME: here we could cache indirect item or part of it in
289 the inode to avoid search_by_key in case of subsequent
290 access to file */
291 blocknr = get_block_num(ind_item, path.pos_in_item);
292 ret = 0;
293 if (blocknr) {
294 map_bh(bh_result, inode->i_sb, blocknr);
295 if (path.pos_in_item ==
296 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
297 set_buffer_boundary(bh_result);
298 }
299 } else
300 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
301 // That there is some MMAPED data associated with it that is yet to be written to disk.
302 if ((args & GET_BLOCK_NO_HOLE)
303 && !PageUptodate(bh_result->b_page)) {
304 ret = -ENOENT;
305 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700307 pathrelse(&path);
308 if (p)
309 kunmap(bh_result->b_page);
310 return ret;
311 }
312 // requested data are in direct item(s)
313 if (!(args & GET_BLOCK_READ_DIRECT)) {
314 // we are called by bmap. FIXME: we can not map block of file
315 // when it is stored in direct item(s)
316 pathrelse(&path);
317 if (p)
318 kunmap(bh_result->b_page);
319 return -ENOENT;
320 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700322 /* if we've got a direct item, and the buffer or page was uptodate,
323 ** we don't want to pull data off disk again. skip to the
324 ** end, where we map the buffer and return
325 */
326 if (buffer_uptodate(bh_result)) {
327 goto finished;
328 } else
329 /*
330 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
331 ** pages without any buffers. If the page is up to date, we don't want
332 ** read old data off disk. Set the up to date bit on the buffer instead
333 ** and jump to the end
334 */
335 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 set_buffer_uptodate(bh_result);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700337 goto finished;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700339 // read file tail into part of page
340 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
341 fs_gen = get_generation(inode->i_sb);
342 copy_item_head(&tmp_ih, ih);
343
344 /* we only want to kmap if we are reading the tail into the page.
345 ** this is not the common case, so we don't kmap until we are
346 ** sure we need to. But, this means the item might move if
347 ** kmap schedules
348 */
349 if (!p) {
350 p = (char *)kmap(bh_result->b_page);
351 if (fs_changed(fs_gen, inode->i_sb)
352 && item_moved(&tmp_ih, &path)) {
353 goto research;
354 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700356 p += offset;
357 memset(p, 0, inode->i_sb->s_blocksize);
358 do {
359 if (!is_direct_le_ih(ih)) {
360 BUG();
361 }
362 /* make sure we don't read more bytes than actually exist in
363 ** the file. This can happen in odd cases where i_size isn't
364 ** correct, and when direct item padding results in a few
365 ** extra bytes at the end of the direct item
366 */
367 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
368 break;
369 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
370 chars =
371 inode->i_size - (le_ih_k_offset(ih) - 1) -
372 path.pos_in_item;
373 done = 1;
374 } else {
375 chars = ih_item_len(ih) - path.pos_in_item;
376 }
377 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700379 if (done)
380 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700382 p += chars;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700384 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
385 // we done, if read direct item is not the last item of
386 // node FIXME: we could try to check right delimiting key
387 // to see whether direct item continues in the right
388 // neighbor or rely on i_size
389 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700391 // update key to look for the next piece
392 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
393 result = search_for_position_by_key(inode->i_sb, &key, &path);
394 if (result != POSITION_FOUND)
395 // i/o error most likely
396 break;
397 bh = get_last_bh(&path);
398 ih = get_ih(&path);
399 } while (1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700401 flush_dcache_page(bh_result->b_page);
402 kunmap(bh_result->b_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700404 finished:
405 pathrelse(&path);
Qu Fuping6283d582005-06-25 14:55:44 -0700406
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700407 if (result == IO_ERROR)
408 return -EIO;
Qu Fuping6283d582005-06-25 14:55:44 -0700409
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700410 /* this buffer has valid data, but isn't valid for io. mapping it to
411 * block #0 tells the rest of reiserfs it just has a tail in it
412 */
413 map_bh(bh_result, inode->i_sb, 0);
414 set_buffer_uptodate(bh_result);
415 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416}
417
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418// this is called to create file map. So, _get_block_create_0 will not
419// read direct item
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700420static int reiserfs_bmap(struct inode *inode, sector_t block,
421 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700423 if (!file_capable(inode, block))
424 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700426 reiserfs_write_lock(inode->i_sb);
427 /* do not read the direct item */
428 _get_block_create_0(inode, block, bh_result, 0);
429 reiserfs_write_unlock(inode->i_sb);
430 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431}
432
433/* special version of get_block that is only used by grab_tail_page right
434** now. It is sent to block_prepare_write, and when you try to get a
435** block past the end of the file (or a block from a hole) it returns
436** -ENOENT instead of a valid buffer. block_prepare_write expects to
437** be able to do i/o on the buffers returned, unless an error value
438** is also returned.
439**
440** So, this allows block_prepare_write to be used for reading a single block
441** in a page. Where it does not produce a valid page for holes, or past the
442** end of the file. This turns out to be exactly what we need for reading
443** tails for conversion.
444**
445** The point of the wrapper is forcing a certain value for create, even
446** though the VFS layer is calling this function with create==1. If you
447** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
448** don't use this function.
449*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700450static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
451 struct buffer_head *bh_result,
452 int create)
453{
454 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455}
456
457/* This is special helper for reiserfs_get_block in case we are executing
458 direct_IO request. */
459static int reiserfs_get_blocks_direct_io(struct inode *inode,
460 sector_t iblock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 struct buffer_head *bh_result,
462 int create)
463{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700464 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700466 bh_result->b_page = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700468 /* We set the b_size before reiserfs_get_block call since it is
469 referenced in convert_tail_for_hole() that may be called from
470 reiserfs_get_block() */
471 bh_result->b_size = (1 << inode->i_blkbits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700473 ret = reiserfs_get_block(inode, iblock, bh_result,
474 create | GET_BLOCK_NO_DANGLE);
475 if (ret)
476 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700478 /* don't allow direct io onto tail pages */
479 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
480 /* make sure future calls to the direct io funcs for this offset
481 ** in the file fail by unmapping the buffer
482 */
483 clear_buffer_mapped(bh_result);
484 ret = -EINVAL;
485 }
486 /* Possible unpacked tail. Flush the data before pages have
487 disappeared */
488 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
489 int err;
490 lock_kernel();
491 err = reiserfs_commit_for_inode(inode);
492 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
493 unlock_kernel();
494 if (err < 0)
495 ret = err;
496 }
497 out:
498 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499}
500
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501/*
502** helper function for when reiserfs_get_block is called for a hole
503** but the file tail is still in a direct item
504** bh_result is the buffer head for the hole
505** tail_offset is the offset of the start of the tail in the file
506**
507** This calls prepare_write, which will start a new transaction
508** you should not be in a transaction, or have any paths held when you
509** call this.
510*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700511static int convert_tail_for_hole(struct inode *inode,
512 struct buffer_head *bh_result,
513 loff_t tail_offset)
514{
515 unsigned long index;
516 unsigned long tail_end;
517 unsigned long tail_start;
518 struct page *tail_page;
519 struct page *hole_page = bh_result->b_page;
520 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700522 if ((tail_offset & (bh_result->b_size - 1)) != 1)
523 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700525 /* always try to read until the end of the block */
526 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
527 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700529 index = tail_offset >> PAGE_CACHE_SHIFT;
530 /* hole_page can be zero in case of direct_io, we are sure
531 that we cannot get here if we write with O_DIRECT into
532 tail page */
533 if (!hole_page || index != hole_page->index) {
534 tail_page = grab_cache_page(inode->i_mapping, index);
535 retval = -ENOMEM;
536 if (!tail_page) {
537 goto out;
538 }
539 } else {
540 tail_page = hole_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700543 /* we don't have to make sure the conversion did not happen while
544 ** we were locking the page because anyone that could convert
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800545 ** must first take i_mutex.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700546 **
547 ** We must fix the tail page for writing because it might have buffers
548 ** that are mapped, but have a block number of 0. This indicates tail
549 ** data that has been read directly into the page, and block_prepare_write
550 ** won't trigger a get_block in this case.
551 */
552 fix_tail_page_for_writing(tail_page);
553 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
554 if (retval)
555 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700557 /* tail conversion might change the data in the page */
558 flush_dcache_page(tail_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700560 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700562 unlock:
563 if (tail_page != hole_page) {
564 unlock_page(tail_page);
565 page_cache_release(tail_page);
566 }
567 out:
568 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569}
570
571static inline int _allocate_block(struct reiserfs_transaction_handle *th,
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700572 sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700573 struct inode *inode,
574 b_blocknr_t * allocated_block_nr,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -0800575 struct treepath *path, int flags)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700576{
577 BUG_ON(!th->t_trans_id);
578
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579#ifdef REISERFS_PREALLOCATE
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800580 if (!(flags & GET_BLOCK_NO_IMUX)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700581 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
582 path, block);
583 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700585 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
586 block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587}
588
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700589int reiserfs_get_block(struct inode *inode, sector_t block,
590 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700592 int repeat, retval = 0;
593 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
594 INITIALIZE_PATH(path);
595 int pos_in_item;
596 struct cpu_key key;
597 struct buffer_head *bh, *unbh = NULL;
598 struct item_head *ih, tmp_ih;
599 __le32 *item;
600 int done;
601 int fs_gen;
602 struct reiserfs_transaction_handle *th = NULL;
603 /* space reserved in transaction batch:
604 . 3 balancings in direct->indirect conversion
605 . 1 block involved into reiserfs_update_sd()
606 XXX in practically impossible worst case direct2indirect()
607 can incur (much) more than 3 balancings.
608 quota update for user, group */
609 int jbegin_count =
610 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
611 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
612 int version;
613 int dangle = 1;
614 loff_t new_offset =
615 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700617 /* bad.... */
618 reiserfs_write_lock(inode->i_sb);
619 version = get_inode_item_key_version(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700621 if (!file_capable(inode, block)) {
622 reiserfs_write_unlock(inode->i_sb);
623 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 }
625
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700626 /* if !create, we aren't changing the FS, so we don't need to
627 ** log anything, so we don't need to start a transaction
628 */
629 if (!(create & GET_BLOCK_CREATE)) {
630 int ret;
631 /* find number of block-th logical block of the file */
632 ret = _get_block_create_0(inode, block, bh_result,
633 create | GET_BLOCK_READ_DIRECT);
634 reiserfs_write_unlock(inode->i_sb);
635 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700637 /*
638 * if we're already in a transaction, make sure to close
639 * any new transactions we start in this func
640 */
641 if ((create & GET_BLOCK_NO_DANGLE) ||
642 reiserfs_transaction_running(inode->i_sb))
643 dangle = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700645 /* If file is of such a size, that it might have a tail and tails are enabled
646 ** we should mark it as possibly needing tail packing on close
647 */
648 if ((have_large_tails(inode->i_sb)
649 && inode->i_size < i_block_size(inode) * 4)
650 || (have_small_tails(inode->i_sb)
651 && inode->i_size < i_block_size(inode)))
652 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
653
654 /* set the key of the first byte in the 'block'-th block of file */
655 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
656 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
657 start_trans:
658 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
659 if (!th) {
660 retval = -ENOMEM;
661 goto failure;
662 }
663 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700665 research:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700667 retval = search_for_position_by_key(inode->i_sb, &key, &path);
668 if (retval == IO_ERROR) {
669 retval = -EIO;
670 goto failure;
671 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700673 bh = get_last_bh(&path);
674 ih = get_ih(&path);
675 item = get_item(&path);
676 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700678 fs_gen = get_generation(inode->i_sb);
679 copy_item_head(&tmp_ih, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700681 if (allocation_needed
682 (retval, allocated_block_nr, ih, item, pos_in_item)) {
683 /* we have to allocate block for the unformatted node */
684 if (!th) {
685 pathrelse(&path);
686 goto start_trans;
687 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700689 repeat =
690 _allocate_block(th, block, inode, &allocated_block_nr,
691 &path, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700693 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
694 /* restart the transaction to give the journal a chance to free
695 ** some blocks. releases the path, so we have to go back to
696 ** research if we succeed on the second try
697 */
698 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
699 retval = restart_transaction(th, inode, &path);
700 if (retval)
701 goto failure;
702 repeat =
703 _allocate_block(th, block, inode,
704 &allocated_block_nr, NULL, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700706 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
707 goto research;
708 }
709 if (repeat == QUOTA_EXCEEDED)
710 retval = -EDQUOT;
711 else
712 retval = -ENOSPC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 goto failure;
714 }
715
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700716 if (fs_changed(fs_gen, inode->i_sb)
717 && item_moved(&tmp_ih, &path)) {
718 goto research;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700722 if (indirect_item_found(retval, ih)) {
723 b_blocknr_t unfm_ptr;
724 /* 'block'-th block is in the file already (there is
725 corresponding cell in some indirect item). But it may be
726 zero unformatted node pointer (hole) */
727 unfm_ptr = get_block_num(item, pos_in_item);
728 if (unfm_ptr == 0) {
729 /* use allocated block to plug the hole */
730 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
731 if (fs_changed(fs_gen, inode->i_sb)
732 && item_moved(&tmp_ih, &path)) {
733 reiserfs_restore_prepared_buffer(inode->i_sb,
734 bh);
735 goto research;
736 }
737 set_buffer_new(bh_result);
738 if (buffer_dirty(bh_result)
739 && reiserfs_data_ordered(inode->i_sb))
740 reiserfs_add_ordered_list(inode, bh_result);
741 put_block_num(item, pos_in_item, allocated_block_nr);
742 unfm_ptr = allocated_block_nr;
743 journal_mark_dirty(th, inode->i_sb, bh);
744 reiserfs_update_sd(th, inode);
745 }
746 set_block_dev_mapped(bh_result, unfm_ptr, inode);
747 pathrelse(&path);
748 retval = 0;
749 if (!dangle && th)
750 retval = reiserfs_end_persistent_transaction(th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700752 reiserfs_write_unlock(inode->i_sb);
753
754 /* the item was found, so new blocks were not added to the file
755 ** there is no need to make sure the inode is updated with this
756 ** transaction
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700758 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 }
760
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700761 if (!th) {
762 pathrelse(&path);
763 goto start_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700766 /* desired position is not found or is in the direct item. We have
767 to append file with holes up to 'block'-th block converting
768 direct items to indirect one if necessary */
769 done = 0;
770 do {
771 if (is_statdata_le_ih(ih)) {
772 __le32 unp = 0;
773 struct cpu_key tmp_key;
774
775 /* indirect item has to be inserted */
776 make_le_item_head(&tmp_ih, &key, version, 1,
777 TYPE_INDIRECT, UNFM_P_SIZE,
778 0 /* free_space */ );
779
780 if (cpu_key_k_offset(&key) == 1) {
781 /* we are going to add 'block'-th block to the file. Use
782 allocated block for that */
783 unp = cpu_to_le32(allocated_block_nr);
784 set_block_dev_mapped(bh_result,
785 allocated_block_nr, inode);
786 set_buffer_new(bh_result);
787 done = 1;
788 }
789 tmp_key = key; // ;)
790 set_cpu_key_k_offset(&tmp_key, 1);
791 PATH_LAST_POSITION(&path)++;
792
793 retval =
794 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
795 inode, (char *)&unp);
796 if (retval) {
797 reiserfs_free_block(th, inode,
798 allocated_block_nr, 1);
799 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
800 }
801 //mark_tail_converted (inode);
802 } else if (is_direct_le_ih(ih)) {
803 /* direct item has to be converted */
804 loff_t tail_offset;
805
806 tail_offset =
807 ((le_ih_k_offset(ih) -
808 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
809 if (tail_offset == cpu_key_k_offset(&key)) {
810 /* direct item we just found fits into block we have
811 to map. Convert it into unformatted node: use
812 bh_result for the conversion */
813 set_block_dev_mapped(bh_result,
814 allocated_block_nr, inode);
815 unbh = bh_result;
816 done = 1;
817 } else {
818 /* we have to padd file tail stored in direct item(s)
819 up to block size and convert it to unformatted
820 node. FIXME: this should also get into page cache */
821
822 pathrelse(&path);
823 /*
824 * ugly, but we can only end the transaction if
825 * we aren't nested
826 */
827 BUG_ON(!th->t_refcount);
828 if (th->t_refcount == 1) {
829 retval =
830 reiserfs_end_persistent_transaction
831 (th);
832 th = NULL;
833 if (retval)
834 goto failure;
835 }
836
837 retval =
838 convert_tail_for_hole(inode, bh_result,
839 tail_offset);
840 if (retval) {
841 if (retval != -ENOSPC)
842 reiserfs_warning(inode->i_sb,
843 "clm-6004: convert tail failed inode %lu, error %d",
844 inode->i_ino,
845 retval);
846 if (allocated_block_nr) {
847 /* the bitmap, the super, and the stat data == 3 */
848 if (!th)
849 th = reiserfs_persistent_transaction(inode->i_sb, 3);
850 if (th)
851 reiserfs_free_block(th,
852 inode,
853 allocated_block_nr,
854 1);
855 }
856 goto failure;
857 }
858 goto research;
859 }
860 retval =
861 direct2indirect(th, inode, &path, unbh,
862 tail_offset);
863 if (retval) {
864 reiserfs_unmap_buffer(unbh);
865 reiserfs_free_block(th, inode,
866 allocated_block_nr, 1);
867 goto failure;
868 }
869 /* it is important the set_buffer_uptodate is done after
870 ** the direct2indirect. The buffer might contain valid
871 ** data newer than the data on disk (read by readpage, changed,
872 ** and then sent here by writepage). direct2indirect needs
873 ** to know if unbh was already up to date, so it can decide
874 ** if the data in unbh needs to be replaced with data from
875 ** the disk
876 */
877 set_buffer_uptodate(unbh);
878
879 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
880 buffer will disappear shortly, so it should not be added to
881 */
882 if (unbh->b_page) {
883 /* we've converted the tail, so we must
884 ** flush unbh before the transaction commits
885 */
886 reiserfs_add_tail_list(inode, unbh);
887
888 /* mark it dirty now to prevent commit_write from adding
889 ** this buffer to the inode's dirty buffer list
890 */
891 /*
892 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
893 * It's still atomic, but it sets the page dirty too,
894 * which makes it eligible for writeback at any time by the
895 * VM (which was also the case with __mark_buffer_dirty())
896 */
897 mark_buffer_dirty(unbh);
898 }
899 } else {
900 /* append indirect item with holes if needed, when appending
901 pointer to 'block'-th block use block, which is already
902 allocated */
903 struct cpu_key tmp_key;
904 unp_t unf_single = 0; // We use this in case we need to allocate only
905 // one block which is a fastpath
906 unp_t *un;
907 __u64 max_to_insert =
908 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
909 UNFM_P_SIZE;
910 __u64 blocks_needed;
911
912 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
913 "vs-804: invalid position for append");
914 /* indirect item has to be appended, set up key of that position */
915 make_cpu_key(&tmp_key, inode,
916 le_key_k_offset(version,
917 &(ih->ih_key)) +
918 op_bytes_number(ih,
919 inode->i_sb->s_blocksize),
920 //pos_in_item * inode->i_sb->s_blocksize,
921 TYPE_INDIRECT, 3); // key type is unimportant
922
Vladimir V. Savelievc499ec22006-03-02 02:54:39 -0800923 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
924 "green-805: invalid offset");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700925 blocks_needed =
926 1 +
927 ((cpu_key_k_offset(&key) -
928 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
929 s_blocksize_bits);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700930
931 if (blocks_needed == 1) {
932 un = &unf_single;
933 } else {
Yan Burman01afb212006-12-06 20:39:01 -0800934 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_ATOMIC); // We need to avoid scheduling.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700935 if (!un) {
936 un = &unf_single;
937 blocks_needed = 1;
938 max_to_insert = 0;
Yan Burman01afb212006-12-06 20:39:01 -0800939 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700940 }
941 if (blocks_needed <= max_to_insert) {
942 /* we are going to add target block to the file. Use allocated
943 block for that */
944 un[blocks_needed - 1] =
945 cpu_to_le32(allocated_block_nr);
946 set_block_dev_mapped(bh_result,
947 allocated_block_nr, inode);
948 set_buffer_new(bh_result);
949 done = 1;
950 } else {
951 /* paste hole to the indirect item */
952 /* If kmalloc failed, max_to_insert becomes zero and it means we
953 only have space for one block */
954 blocks_needed =
955 max_to_insert ? max_to_insert : 1;
956 }
957 retval =
958 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
959 (char *)un,
960 UNFM_P_SIZE *
961 blocks_needed);
962
963 if (blocks_needed != 1)
964 kfree(un);
965
966 if (retval) {
967 reiserfs_free_block(th, inode,
968 allocated_block_nr, 1);
969 goto failure;
970 }
971 if (!done) {
972 /* We need to mark new file size in case this function will be
973 interrupted/aborted later on. And we may do this only for
974 holes. */
975 inode->i_size +=
976 inode->i_sb->s_blocksize * blocks_needed;
977 }
978 }
979
980 if (done == 1)
981 break;
982
983 /* this loop could log more blocks than we had originally asked
984 ** for. So, we have to allow the transaction to end if it is
985 ** too big or too full. Update the inode so things are
986 ** consistent if we crash before the function returns
987 **
988 ** release the path so that anybody waiting on the path before
989 ** ending their transaction will be able to continue.
990 */
991 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
992 retval = restart_transaction(th, inode, &path);
993 if (retval)
994 goto failure;
995 }
996 /* inserting indirect pointers for a hole can take a
997 ** long time. reschedule if needed
998 */
999 cond_resched();
1000
1001 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1002 if (retval == IO_ERROR) {
1003 retval = -EIO;
1004 goto failure;
1005 }
1006 if (retval == POSITION_FOUND) {
1007 reiserfs_warning(inode->i_sb,
1008 "vs-825: reiserfs_get_block: "
1009 "%K should not be found", &key);
1010 retval = -EEXIST;
1011 if (allocated_block_nr)
1012 reiserfs_free_block(th, inode,
1013 allocated_block_nr, 1);
1014 pathrelse(&path);
1015 goto failure;
1016 }
1017 bh = get_last_bh(&path);
1018 ih = get_ih(&path);
1019 item = get_item(&path);
1020 pos_in_item = path.pos_in_item;
1021 } while (1);
1022
1023 retval = 0;
1024
1025 failure:
1026 if (th && (!dangle || (retval && !th->t_trans_id))) {
1027 int err;
1028 if (th->t_trans_id)
1029 reiserfs_update_sd(th, inode);
1030 err = reiserfs_end_persistent_transaction(th);
1031 if (err)
1032 retval = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001035 reiserfs_write_unlock(inode->i_sb);
1036 reiserfs_check_path(&path);
1037 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038}
1039
1040static int
1041reiserfs_readpages(struct file *file, struct address_space *mapping,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001042 struct list_head *pages, unsigned nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001044 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045}
1046
1047/* Compute real number of used bytes by file
1048 * Following three functions can go away when we'll have enough space in stat item
1049 */
1050static int real_space_diff(struct inode *inode, int sd_size)
1051{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001052 int bytes;
1053 loff_t blocksize = inode->i_sb->s_blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001055 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1056 return sd_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001058 /* End of file is also in full block with indirect reference, so round
1059 ** up to the next block.
1060 **
1061 ** there is just no way to know if the tail is actually packed
1062 ** on the file, so we have to assume it isn't. When we pack the
1063 ** tail, we add 4 bytes to pretend there really is an unformatted
1064 ** node pointer
1065 */
1066 bytes =
1067 ((inode->i_size +
1068 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1069 sd_size;
1070 return bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071}
1072
1073static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001074 int sd_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001076 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1077 return inode->i_size +
1078 (loff_t) (real_space_diff(inode, sd_size));
1079 }
1080 return ((loff_t) real_space_diff(inode, sd_size)) +
1081 (((loff_t) blocks) << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082}
1083
1084/* Compute number of blocks used by file in ReiserFS counting */
1085static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1086{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001087 loff_t bytes = inode_get_bytes(inode);
1088 loff_t real_space = real_space_diff(inode, sd_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001090 /* keeps fsck and non-quota versions of reiserfs happy */
1091 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1092 bytes += (loff_t) 511;
1093 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001095 /* files from before the quota patch might i_blocks such that
1096 ** bytes < real_space. Deal with that here to prevent it from
1097 ** going negative.
1098 */
1099 if (bytes < real_space)
1100 return 0;
1101 return (bytes - real_space) >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102}
1103
1104//
1105// BAD: new directories have stat data of new type and all other items
1106// of old type. Version stored in the inode says about body items, so
1107// in update_stat_data we can not rely on inode, but have to check
1108// item version directly
1109//
1110
1111// called by read_locked_inode
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001112static void init_inode(struct inode *inode, struct treepath *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001114 struct buffer_head *bh;
1115 struct item_head *ih;
1116 __u32 rdev;
1117 //int version = ITEM_VERSION_1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001119 bh = PATH_PLAST_BUFFER(path);
1120 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001122 copy_key(INODE_PKEY(inode), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001124 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1125 REISERFS_I(inode)->i_flags = 0;
1126 REISERFS_I(inode)->i_prealloc_block = 0;
1127 REISERFS_I(inode)->i_prealloc_count = 0;
1128 REISERFS_I(inode)->i_trans_id = 0;
1129 REISERFS_I(inode)->i_jl = NULL;
Vladimir Savelievde145692007-01-22 20:40:46 -08001130 mutex_init(&(REISERFS_I(inode)->i_mmap));
Alexey Dobriyancfe14672006-09-29 02:00:00 -07001131 reiserfs_init_acl_access(inode);
1132 reiserfs_init_acl_default(inode);
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001133 reiserfs_init_xattr_rwsem(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001135 if (stat_data_v1(ih)) {
1136 struct stat_data_v1 *sd =
1137 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1138 unsigned long blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001140 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1141 set_inode_sd_version(inode, STAT_DATA_V1);
1142 inode->i_mode = sd_v1_mode(sd);
1143 inode->i_nlink = sd_v1_nlink(sd);
1144 inode->i_uid = sd_v1_uid(sd);
1145 inode->i_gid = sd_v1_gid(sd);
1146 inode->i_size = sd_v1_size(sd);
1147 inode->i_atime.tv_sec = sd_v1_atime(sd);
1148 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1149 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1150 inode->i_atime.tv_nsec = 0;
1151 inode->i_ctime.tv_nsec = 0;
1152 inode->i_mtime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001154 inode->i_blocks = sd_v1_blocks(sd);
1155 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1156 blocks = (inode->i_size + 511) >> 9;
1157 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1158 if (inode->i_blocks > blocks) {
1159 // there was a bug in <=3.5.23 when i_blocks could take negative
1160 // values. Starting from 3.5.17 this value could even be stored in
1161 // stat data. For such files we set i_blocks based on file
1162 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1163 // only updated if file's inode will ever change
1164 inode->i_blocks = blocks;
1165 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001167 rdev = sd_v1_rdev(sd);
1168 REISERFS_I(inode)->i_first_direct_byte =
1169 sd_v1_first_direct_byte(sd);
1170 /* an early bug in the quota code can give us an odd number for the
1171 ** block count. This is incorrect, fix it here.
1172 */
1173 if (inode->i_blocks & 1) {
1174 inode->i_blocks++;
1175 }
1176 inode_set_bytes(inode,
1177 to_real_used_space(inode, inode->i_blocks,
1178 SD_V1_SIZE));
1179 /* nopack is initially zero for v1 objects. For v2 objects,
1180 nopack is initialised from sd_attrs */
1181 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1182 } else {
1183 // new stat data found, but object may have old items
1184 // (directories and symlinks)
1185 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1186
1187 inode->i_mode = sd_v2_mode(sd);
1188 inode->i_nlink = sd_v2_nlink(sd);
1189 inode->i_uid = sd_v2_uid(sd);
1190 inode->i_size = sd_v2_size(sd);
1191 inode->i_gid = sd_v2_gid(sd);
1192 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1193 inode->i_atime.tv_sec = sd_v2_atime(sd);
1194 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1195 inode->i_ctime.tv_nsec = 0;
1196 inode->i_mtime.tv_nsec = 0;
1197 inode->i_atime.tv_nsec = 0;
1198 inode->i_blocks = sd_v2_blocks(sd);
1199 rdev = sd_v2_rdev(sd);
1200 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1201 inode->i_generation =
1202 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1203 else
1204 inode->i_generation = sd_v2_generation(sd);
1205
1206 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1207 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1208 else
1209 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1210 REISERFS_I(inode)->i_first_direct_byte = 0;
1211 set_inode_sd_version(inode, STAT_DATA_V2);
1212 inode_set_bytes(inode,
1213 to_real_used_space(inode, inode->i_blocks,
1214 SD_V2_SIZE));
1215 /* read persistent inode attributes from sd and initalise
1216 generic inode flags from them */
1217 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1218 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 }
1220
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001221 pathrelse(path);
1222 if (S_ISREG(inode->i_mode)) {
1223 inode->i_op = &reiserfs_file_inode_operations;
1224 inode->i_fop = &reiserfs_file_operations;
1225 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1226 } else if (S_ISDIR(inode->i_mode)) {
1227 inode->i_op = &reiserfs_dir_inode_operations;
1228 inode->i_fop = &reiserfs_dir_operations;
1229 } else if (S_ISLNK(inode->i_mode)) {
1230 inode->i_op = &reiserfs_symlink_inode_operations;
1231 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1232 } else {
1233 inode->i_blocks = 0;
1234 inode->i_op = &reiserfs_special_inode_operations;
1235 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237}
1238
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239// update new stat data with inode fields
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001240static void inode2sd(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001242 struct stat_data *sd_v2 = (struct stat_data *)sd;
1243 __u16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001245 set_sd_v2_mode(sd_v2, inode->i_mode);
1246 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1247 set_sd_v2_uid(sd_v2, inode->i_uid);
1248 set_sd_v2_size(sd_v2, size);
1249 set_sd_v2_gid(sd_v2, inode->i_gid);
1250 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1251 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1252 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1253 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1254 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1255 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1256 else
1257 set_sd_v2_generation(sd_v2, inode->i_generation);
1258 flags = REISERFS_I(inode)->i_attrs;
1259 i_attrs_to_sd_attrs(inode, &flags);
1260 set_sd_v2_attrs(sd_v2, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261}
1262
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263// used to copy inode's fields to old stat data
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001264static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001266 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001268 set_sd_v1_mode(sd_v1, inode->i_mode);
1269 set_sd_v1_uid(sd_v1, inode->i_uid);
1270 set_sd_v1_gid(sd_v1, inode->i_gid);
1271 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1272 set_sd_v1_size(sd_v1, size);
1273 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1274 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1275 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001277 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1278 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1279 else
1280 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001282 // Sigh. i_first_direct_byte is back
1283 set_sd_v1_first_direct_byte(sd_v1,
1284 REISERFS_I(inode)->i_first_direct_byte);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285}
1286
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287/* NOTE, you must prepare the buffer head before sending it here,
1288** and then log it after the call
1289*/
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001290static void update_stat_data(struct treepath *path, struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001291 loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001293 struct buffer_head *bh;
1294 struct item_head *ih;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001296 bh = PATH_PLAST_BUFFER(path);
1297 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001299 if (!is_statdata_le_ih(ih))
1300 reiserfs_panic(inode->i_sb,
1301 "vs-13065: update_stat_data: key %k, found item %h",
1302 INODE_PKEY(inode), ih);
1303
1304 if (stat_data_v1(ih)) {
1305 // path points to old stat data
1306 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1307 } else {
1308 inode2sd(B_I_PITEM(bh, ih), inode, size);
1309 }
1310
1311 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312}
1313
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001314void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1315 struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001317 struct cpu_key key;
1318 INITIALIZE_PATH(path);
1319 struct buffer_head *bh;
1320 int fs_gen;
1321 struct item_head *ih, tmp_ih;
1322 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001324 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001326 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1327
1328 for (;;) {
1329 int pos;
1330 /* look for the object's stat data */
1331 retval = search_item(inode->i_sb, &key, &path);
1332 if (retval == IO_ERROR) {
1333 reiserfs_warning(inode->i_sb,
1334 "vs-13050: reiserfs_update_sd: "
1335 "i/o failure occurred trying to update %K stat data",
1336 &key);
1337 return;
1338 }
1339 if (retval == ITEM_NOT_FOUND) {
1340 pos = PATH_LAST_POSITION(&path);
1341 pathrelse(&path);
1342 if (inode->i_nlink == 0) {
1343 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1344 return;
1345 }
1346 reiserfs_warning(inode->i_sb,
1347 "vs-13060: reiserfs_update_sd: "
1348 "stat data of object %k (nlink == %d) not found (pos %d)",
1349 INODE_PKEY(inode), inode->i_nlink,
1350 pos);
1351 reiserfs_check_path(&path);
1352 return;
1353 }
1354
1355 /* sigh, prepare_for_journal might schedule. When it schedules the
1356 ** FS might change. We have to detect that, and loop back to the
1357 ** search if the stat data item has moved
1358 */
1359 bh = get_last_bh(&path);
1360 ih = get_ih(&path);
1361 copy_item_head(&tmp_ih, ih);
1362 fs_gen = get_generation(inode->i_sb);
1363 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1364 if (fs_changed(fs_gen, inode->i_sb)
1365 && item_moved(&tmp_ih, &path)) {
1366 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1367 continue; /* Stat_data item has been moved after scheduling. */
1368 }
1369 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001371 update_stat_data(&path, inode, size);
1372 journal_mark_dirty(th, th->t_super, bh);
1373 pathrelse(&path);
1374 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375}
1376
1377/* reiserfs_read_locked_inode is called to read the inode off disk, and it
1378** does a make_bad_inode when things go wrong. But, we need to make sure
1379** and clear the key in the private portion of the inode, otherwise a
1380** corresponding iput might try to delete whatever object the inode last
1381** represented.
1382*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001383static void reiserfs_make_bad_inode(struct inode *inode)
1384{
1385 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1386 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
1389//
1390// initially this function was derived from minix or ext2's analog and
1391// evolved as the prototype did
1392//
1393
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001394int reiserfs_init_locked_inode(struct inode *inode, void *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001396 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1397 inode->i_ino = args->objectid;
1398 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1399 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400}
1401
1402/* looks for stat data in the tree, and fills up the fields of in-core
1403 inode stat data fields */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001404void reiserfs_read_locked_inode(struct inode *inode,
1405 struct reiserfs_iget_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001407 INITIALIZE_PATH(path_to_sd);
1408 struct cpu_key key;
1409 unsigned long dirino;
1410 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001412 dirino = args->dirid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001414 /* set version 1, version 2 could be used too, because stat data
1415 key is the same in both versions */
1416 key.version = KEY_FORMAT_3_5;
1417 key.on_disk_key.k_dir_id = dirino;
1418 key.on_disk_key.k_objectid = inode->i_ino;
1419 key.on_disk_key.k_offset = 0;
1420 key.on_disk_key.k_type = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001422 /* look for the object's stat data */
1423 retval = search_item(inode->i_sb, &key, &path_to_sd);
1424 if (retval == IO_ERROR) {
1425 reiserfs_warning(inode->i_sb,
1426 "vs-13070: reiserfs_read_locked_inode: "
1427 "i/o failure occurred trying to find stat data of %K",
1428 &key);
1429 reiserfs_make_bad_inode(inode);
1430 return;
1431 }
1432 if (retval != ITEM_FOUND) {
1433 /* a stale NFS handle can trigger this without it being an error */
1434 pathrelse(&path_to_sd);
1435 reiserfs_make_bad_inode(inode);
1436 inode->i_nlink = 0;
1437 return;
1438 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001440 init_inode(inode, &path_to_sd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001442 /* It is possible that knfsd is trying to access inode of a file
1443 that is being removed from the disk by some other thread. As we
1444 update sd on unlink all that is required is to check for nlink
1445 here. This bug was first found by Sizif when debugging
1446 SquidNG/Butterfly, forgotten, and found again after Philippe
1447 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001449 More logical fix would require changes in fs/inode.c:iput() to
1450 remove inode from hash-table _after_ fs cleaned disk stuff up and
1451 in iget() to return NULL if I_FREEING inode is found in
1452 hash-table. */
1453 /* Currently there is one place where it's ok to meet inode with
1454 nlink==0: processing of open-unlinked and half-truncated files
1455 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1456 if ((inode->i_nlink == 0) &&
1457 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1458 reiserfs_warning(inode->i_sb,
1459 "vs-13075: reiserfs_read_locked_inode: "
1460 "dead inode read from disk %K. "
1461 "This is likely to be race with knfsd. Ignore",
1462 &key);
1463 reiserfs_make_bad_inode(inode);
1464 }
1465
1466 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467
1468}
1469
1470/**
1471 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1472 *
1473 * @inode: inode from hash table to check
1474 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1475 *
1476 * This function is called by iget5_locked() to distinguish reiserfs inodes
1477 * having the same inode numbers. Such inodes can only exist due to some
1478 * error condition. One of them should be bad. Inodes with identical
1479 * inode numbers (objectids) are distinguished by parent directory ids.
1480 *
1481 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001482int reiserfs_find_actor(struct inode *inode, void *opaque)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001484 struct reiserfs_iget_args *args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001486 args = opaque;
1487 /* args is already in CPU order */
1488 return (inode->i_ino == args->objectid) &&
1489 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490}
1491
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001492struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001494 struct inode *inode;
1495 struct reiserfs_iget_args args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001497 args.objectid = key->on_disk_key.k_objectid;
1498 args.dirid = key->on_disk_key.k_dir_id;
1499 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1500 reiserfs_find_actor, reiserfs_init_locked_inode,
1501 (void *)(&args));
1502 if (!inode)
1503 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001505 if (inode->i_state & I_NEW) {
1506 reiserfs_read_locked_inode(inode, &args);
1507 unlock_new_inode(inode);
1508 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001510 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1511 /* either due to i/o error or a stale NFS handle */
1512 iput(inode);
1513 inode = NULL;
1514 }
1515 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516}
1517
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001518static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1519 u32 objectid, u32 dir_id, u32 generation)
1520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001522 struct cpu_key key;
1523 struct dentry *result;
1524 struct inode *inode;
1525
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001526 key.on_disk_key.k_objectid = objectid;
1527 key.on_disk_key.k_dir_id = dir_id;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001528 reiserfs_write_lock(sb);
1529 inode = reiserfs_iget(sb, &key);
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001530 if (inode && !IS_ERR(inode) && generation != 0 &&
1531 generation != inode->i_generation) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001532 iput(inode);
1533 inode = NULL;
1534 }
1535 reiserfs_write_unlock(sb);
1536 if (!inode)
1537 inode = ERR_PTR(-ESTALE);
1538 if (IS_ERR(inode))
David Howellse231c2e2008-02-07 00:15:26 -08001539 return ERR_CAST(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001540 result = d_alloc_anon(inode);
1541 if (!result) {
1542 iput(inode);
1543 return ERR_PTR(-ENOMEM);
1544 }
1545 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546}
1547
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001548struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1549 int fh_len, int fh_type)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001550{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001551 /* fhtype happens to reflect the number of u32s encoded.
1552 * due to a bug in earlier code, fhtype might indicate there
1553 * are more u32s then actually fitted.
1554 * so if fhtype seems to be more than len, reduce fhtype.
1555 * Valid types are:
1556 * 2 - objectid + dir_id - legacy support
1557 * 3 - objectid + dir_id + generation
1558 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1559 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1560 * 6 - as above plus generation of directory
1561 * 6 does not fit in NFSv2 handles
1562 */
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001563 if (fh_type > fh_len) {
1564 if (fh_type != 6 || fh_len != 5)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001565 reiserfs_warning(sb,
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001566 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1567 fh_type, fh_len);
1568 fh_type = 5;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001569 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001571 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1572 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1573}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001575struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1576 int fh_len, int fh_type)
1577{
1578 if (fh_type < 4)
1579 return NULL;
1580
1581 return reiserfs_get_dentry(sb,
1582 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1583 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1584 (fh_type == 6) ? fid->raw[5] : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585}
1586
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001587int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1588 int need_parent)
1589{
1590 struct inode *inode = dentry->d_inode;
1591 int maxlen = *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001593 if (maxlen < 3)
1594 return 255;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001596 data[0] = inode->i_ino;
1597 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1598 data[2] = inode->i_generation;
1599 *lenp = 3;
1600 /* no room for directory info? return what we've stored so far */
1601 if (maxlen < 5 || !need_parent)
1602 return 3;
1603
1604 spin_lock(&dentry->d_lock);
1605 inode = dentry->d_parent->d_inode;
1606 data[3] = inode->i_ino;
1607 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1608 *lenp = 5;
1609 if (maxlen >= 6) {
1610 data[5] = inode->i_generation;
1611 *lenp = 6;
1612 }
1613 spin_unlock(&dentry->d_lock);
1614 return *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615}
1616
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617/* looks for stat data, then copies fields to it, marks the buffer
1618 containing stat data as dirty */
1619/* reiserfs inodes are never really dirty, since the dirty inode call
1620** always logs them. This call allows the VFS inode marking routines
1621** to properly mark inodes for datasync and such, but only actually
1622** does something when called for a synchronous update.
1623*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001624int reiserfs_write_inode(struct inode *inode, int do_sync)
1625{
1626 struct reiserfs_transaction_handle th;
1627 int jbegin_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001629 if (inode->i_sb->s_flags & MS_RDONLY)
1630 return -EROFS;
1631 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1632 ** these cases are just when the system needs ram, not when the
1633 ** inode needs to reach disk for safety, and they can safely be
1634 ** ignored because the altered inode has already been logged.
1635 */
1636 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1637 reiserfs_write_lock(inode->i_sb);
1638 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1639 reiserfs_update_sd(&th, inode);
1640 journal_end_sync(&th, inode->i_sb, jbegin_count);
1641 }
1642 reiserfs_write_unlock(inode->i_sb);
1643 }
1644 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645}
1646
1647/* stat data of new object is inserted already, this inserts the item
1648 containing "." and ".." entries */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001649static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1650 struct inode *inode,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001651 struct item_head *ih, struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001652 struct inode *dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001654 struct super_block *sb = th->t_super;
1655 char empty_dir[EMPTY_DIR_SIZE];
1656 char *body = empty_dir;
1657 struct cpu_key key;
1658 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001660 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001662 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1663 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1664 TYPE_DIRENTRY, 3 /*key length */ );
1665
1666 /* compose item head for new item. Directories consist of items of
1667 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1668 is done by reiserfs_new_inode */
1669 if (old_format_only(sb)) {
1670 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1671 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1672
1673 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1674 ih->ih_key.k_objectid,
1675 INODE_PKEY(dir)->k_dir_id,
1676 INODE_PKEY(dir)->k_objectid);
1677 } else {
1678 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1679 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1680
1681 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1682 ih->ih_key.k_objectid,
1683 INODE_PKEY(dir)->k_dir_id,
1684 INODE_PKEY(dir)->k_objectid);
1685 }
1686
1687 /* look for place in the tree for new item */
1688 retval = search_item(sb, &key, path);
1689 if (retval == IO_ERROR) {
1690 reiserfs_warning(sb, "vs-13080: reiserfs_new_directory: "
1691 "i/o failure occurred creating new directory");
1692 return -EIO;
1693 }
1694 if (retval == ITEM_FOUND) {
1695 pathrelse(path);
1696 reiserfs_warning(sb, "vs-13070: reiserfs_new_directory: "
1697 "object with this key exists (%k)",
1698 &(ih->ih_key));
1699 return -EEXIST;
1700 }
1701
1702 /* insert item, that is empty directory item */
1703 return reiserfs_insert_item(th, path, &key, ih, inode, body);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704}
1705
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706/* stat data of object has been inserted, this inserts the item
1707 containing the body of symlink */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001708static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1709 struct item_head *ih,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001710 struct treepath *path, const char *symname,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001711 int item_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001713 struct super_block *sb = th->t_super;
1714 struct cpu_key key;
1715 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001717 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001719 _make_cpu_key(&key, KEY_FORMAT_3_5,
1720 le32_to_cpu(ih->ih_key.k_dir_id),
1721 le32_to_cpu(ih->ih_key.k_objectid),
1722 1, TYPE_DIRECT, 3 /*key length */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001724 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1725 0 /*free_space */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001727 /* look for place in the tree for new item */
1728 retval = search_item(sb, &key, path);
1729 if (retval == IO_ERROR) {
1730 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlinik: "
1731 "i/o failure occurred creating new symlink");
1732 return -EIO;
1733 }
1734 if (retval == ITEM_FOUND) {
1735 pathrelse(path);
1736 reiserfs_warning(sb, "vs-13080: reiserfs_new_symlink: "
1737 "object with this key exists (%k)",
1738 &(ih->ih_key));
1739 return -EEXIST;
1740 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001742 /* insert item, that is body of symlink */
1743 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744}
1745
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746/* inserts the stat data into the tree, and then calls
1747 reiserfs_new_directory (to insert ".", ".." item if new object is
1748 directory) or reiserfs_new_symlink (to insert symlink body if new
1749 object is symlink) or nothing (if new object is regular file)
1750
1751 NOTE! uid and gid must already be set in the inode. If we return
1752 non-zero due to an error, we have to drop the quota previously allocated
1753 for the fresh inode. This can only be done outside a transaction, so
1754 if we return non-zero, we also end the transaction. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001755int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1756 struct inode *dir, int mode, const char *symname,
1757 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1758 strlen (symname) for symlinks) */
1759 loff_t i_size, struct dentry *dentry,
1760 struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001762 struct super_block *sb;
1763 INITIALIZE_PATH(path_to_key);
1764 struct cpu_key key;
1765 struct item_head ih;
1766 struct stat_data sd;
1767 int retval;
1768 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001770 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001772 if (DQUOT_ALLOC_INODE(inode)) {
1773 err = -EDQUOT;
1774 goto out_end_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 }
Eric Sesterhenn585b7742006-10-04 02:15:30 -07001776 if (!dir->i_nlink) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001777 err = -EPERM;
1778 goto out_bad_inode;
1779 }
1780
1781 sb = dir->i_sb;
1782
1783 /* item head of new item */
1784 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1785 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1786 if (!ih.ih_key.k_objectid) {
1787 err = -ENOMEM;
1788 goto out_bad_inode;
1789 }
1790 if (old_format_only(sb))
1791 /* not a perfect generation count, as object ids can be reused, but
1792 ** this is as good as reiserfs can do right now.
1793 ** note that the private part of inode isn't filled in yet, we have
1794 ** to use the directory.
1795 */
1796 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1797 else
1798#if defined( USE_INODE_GENERATION_COUNTER )
1799 inode->i_generation =
1800 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1801#else
1802 inode->i_generation = ++event;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001805 /* fill stat data */
1806 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1807
1808 /* uid and gid must already be set by the caller for quota init */
1809
1810 /* symlink cannot be immutable or append only, right? */
1811 if (S_ISLNK(inode->i_mode))
1812 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1813
1814 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1815 inode->i_size = i_size;
1816 inode->i_blocks = 0;
1817 inode->i_bytes = 0;
1818 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1819 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1820
1821 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1822 REISERFS_I(inode)->i_flags = 0;
1823 REISERFS_I(inode)->i_prealloc_block = 0;
1824 REISERFS_I(inode)->i_prealloc_count = 0;
1825 REISERFS_I(inode)->i_trans_id = 0;
1826 REISERFS_I(inode)->i_jl = NULL;
1827 REISERFS_I(inode)->i_attrs =
1828 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1829 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
Vladimir Savelievde145692007-01-22 20:40:46 -08001830 mutex_init(&(REISERFS_I(inode)->i_mmap));
Alexey Dobriyancfe14672006-09-29 02:00:00 -07001831 reiserfs_init_acl_access(inode);
1832 reiserfs_init_acl_default(inode);
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001833 reiserfs_init_xattr_rwsem(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001834
1835 if (old_format_only(sb))
1836 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1837 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1838 else
1839 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1840 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1841
1842 /* key to search for correct place for new stat data */
1843 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1844 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1845 TYPE_STAT_DATA, 3 /*key length */ );
1846
1847 /* find proper place for inserting of stat data */
1848 retval = search_item(sb, &key, &path_to_key);
1849 if (retval == IO_ERROR) {
1850 err = -EIO;
1851 goto out_bad_inode;
1852 }
1853 if (retval == ITEM_FOUND) {
1854 pathrelse(&path_to_key);
1855 err = -EEXIST;
1856 goto out_bad_inode;
1857 }
1858 if (old_format_only(sb)) {
1859 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1860 pathrelse(&path_to_key);
1861 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1862 err = -EINVAL;
1863 goto out_bad_inode;
1864 }
1865 inode2sd_v1(&sd, inode, inode->i_size);
1866 } else {
1867 inode2sd(&sd, inode, inode->i_size);
1868 }
1869 // these do not go to on-disk stat data
1870 inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001871
1872 // store in in-core inode the key of stat data and version all
1873 // object items will have (directory items will have old offset
1874 // format, other new objects will consist of new items)
1875 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1876 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1877 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1878 else
1879 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1880 if (old_format_only(sb))
1881 set_inode_sd_version(inode, STAT_DATA_V1);
1882 else
1883 set_inode_sd_version(inode, STAT_DATA_V2);
1884
1885 /* insert the stat data into the tree */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001887 if (REISERFS_I(dir)->new_packing_locality)
1888 th->displace_new_blocks = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001890 retval =
1891 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1892 (char *)(&sd));
1893 if (retval) {
1894 err = retval;
1895 reiserfs_check_path(&path_to_key);
1896 goto out_bad_inode;
1897 }
1898#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1899 if (!th->displace_new_blocks)
1900 REISERFS_I(dir)->new_packing_locality = 0;
1901#endif
1902 if (S_ISDIR(mode)) {
1903 /* insert item with "." and ".." */
1904 retval =
1905 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1906 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001908 if (S_ISLNK(mode)) {
1909 /* insert body of symlink */
1910 if (!old_format_only(sb))
1911 i_size = ROUND_UP(i_size);
1912 retval =
1913 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1914 i_size);
1915 }
1916 if (retval) {
1917 err = retval;
1918 reiserfs_check_path(&path_to_key);
1919 journal_end(th, th->t_super, th->t_blocks_allocated);
1920 goto out_inserted_sd;
1921 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001923 /* XXX CHECK THIS */
1924 if (reiserfs_posixacl(inode->i_sb)) {
1925 retval = reiserfs_inherit_default_acl(dir, dentry, inode);
1926 if (retval) {
1927 err = retval;
1928 reiserfs_check_path(&path_to_key);
1929 journal_end(th, th->t_super, th->t_blocks_allocated);
1930 goto out_inserted_sd;
1931 }
1932 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1933 reiserfs_warning(inode->i_sb, "ACLs aren't enabled in the fs, "
1934 "but vfs thinks they are!");
1935 } else if (is_reiserfs_priv_object(dir)) {
1936 reiserfs_mark_inode_private(inode);
1937 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001939 insert_inode_hash(inode);
1940 reiserfs_update_sd(th, inode);
1941 reiserfs_check_path(&path_to_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001943 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944
1945/* it looks like you can easily compress these two goto targets into
1946 * one. Keeping it like this doesn't actually hurt anything, and they
1947 * are place holders for what the quota code actually needs.
1948 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001949 out_bad_inode:
1950 /* Invalidate the object, nothing was inserted yet */
1951 INODE_PKEY(inode)->k_objectid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001953 /* Quota change must be inside a transaction for journaling */
1954 DQUOT_FREE_INODE(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001956 out_end_trans:
1957 journal_end(th, th->t_super, th->t_blocks_allocated);
1958 /* Drop can be outside and it needs more credits so it's better to have it outside */
1959 DQUOT_DROP(inode);
1960 inode->i_flags |= S_NOQUOTA;
1961 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001963 out_inserted_sd:
1964 inode->i_nlink = 0;
1965 th->t_trans_id = 0; /* so the caller can't use this handle later */
Jeff Mahoneyb3bb8af2005-07-27 11:43:38 -07001966
1967 /* If we were inheriting an ACL, we need to release the lock so that
1968 * iput doesn't deadlock in reiserfs_delete_xattrs. The locking
1969 * code really needs to be reworked, but this will take care of it
1970 * for now. -jeffm */
Alexey Dobriyancfe14672006-09-29 02:00:00 -07001971#ifdef CONFIG_REISERFS_FS_POSIX_ACL
Jan Karad86c3902005-08-18 11:24:17 -07001972 if (REISERFS_I(dir)->i_acl_default && !IS_ERR(REISERFS_I(dir)->i_acl_default)) {
Jeff Mahoneyb3bb8af2005-07-27 11:43:38 -07001973 reiserfs_write_unlock_xattrs(dir->i_sb);
1974 iput(inode);
1975 reiserfs_write_lock_xattrs(dir->i_sb);
1976 } else
Alexey Dobriyancfe14672006-09-29 02:00:00 -07001977#endif
Jeff Mahoneyb3bb8af2005-07-27 11:43:38 -07001978 iput(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001979 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980}
1981
1982/*
1983** finds the tail page in the page cache,
1984** reads the last block in.
1985**
1986** On success, page_result is set to a locked, pinned page, and bh_result
1987** is set to an up to date buffer for the last block in the file. returns 0.
1988**
1989** tail conversion is not done, so bh_result might not be valid for writing
1990** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1991** trying to write the block.
1992**
1993** on failure, nonzero is returned, page_result and bh_result are untouched.
1994*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001995static int grab_tail_page(struct inode *p_s_inode,
1996 struct page **page_result,
1997 struct buffer_head **bh_result)
1998{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002000 /* we want the page with the last byte in the file,
2001 ** not the page that will hold the next byte for appending
2002 */
2003 unsigned long index = (p_s_inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2004 unsigned long pos = 0;
2005 unsigned long start = 0;
2006 unsigned long blocksize = p_s_inode->i_sb->s_blocksize;
2007 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1);
2008 struct buffer_head *bh;
2009 struct buffer_head *head;
2010 struct page *page;
2011 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002013 /* we know that we are only called with inode->i_size > 0.
2014 ** we also know that a file tail can never be as big as a block
2015 ** If i_size % blocksize == 0, our file is currently block aligned
2016 ** and it won't need converting or zeroing after a truncate.
2017 */
2018 if ((offset & (blocksize - 1)) == 0) {
2019 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002021 page = grab_cache_page(p_s_inode->i_mapping, index);
2022 error = -ENOMEM;
2023 if (!page) {
2024 goto out;
2025 }
2026 /* start within the page of the last block in the file */
2027 start = (offset / blocksize) * blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002029 error = block_prepare_write(page, start, offset,
2030 reiserfs_get_block_create_0);
2031 if (error)
2032 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002034 head = page_buffers(page);
2035 bh = head;
2036 do {
2037 if (pos >= start) {
2038 break;
2039 }
2040 bh = bh->b_this_page;
2041 pos += blocksize;
2042 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002044 if (!buffer_uptodate(bh)) {
2045 /* note, this should never happen, prepare_write should
2046 ** be taking care of this for us. If the buffer isn't up to date,
2047 ** I've screwed up the code to find the buffer, or the code to
2048 ** call prepare_write
2049 */
2050 reiserfs_warning(p_s_inode->i_sb,
2051 "clm-6000: error reading block %lu on dev %s",
2052 bh->b_blocknr,
2053 reiserfs_bdevname(p_s_inode->i_sb));
2054 error = -EIO;
2055 goto unlock;
2056 }
2057 *bh_result = bh;
2058 *page_result = page;
2059
2060 out:
2061 return error;
2062
2063 unlock:
2064 unlock_page(page);
2065 page_cache_release(page);
2066 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067}
2068
2069/*
2070** vfs version of truncate file. Must NOT be called with
2071** a transaction already started.
2072**
2073** some code taken from block_truncate_page
2074*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002075int reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps)
2076{
2077 struct reiserfs_transaction_handle th;
2078 /* we want the offset for the first byte after the end of the file */
2079 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1);
2080 unsigned blocksize = p_s_inode->i_sb->s_blocksize;
2081 unsigned length;
2082 struct page *page = NULL;
2083 int error;
2084 struct buffer_head *bh = NULL;
Jeff Mahoney24996042005-12-14 14:38:05 -05002085 int err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002087 reiserfs_write_lock(p_s_inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002089 if (p_s_inode->i_size > 0) {
2090 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
2091 // -ENOENT means we truncated past the end of the file,
2092 // and get_block_create_0 could not find a block to read in,
2093 // which is ok.
2094 if (error != -ENOENT)
2095 reiserfs_warning(p_s_inode->i_sb,
2096 "clm-6001: grab_tail_page failed %d",
2097 error);
2098 page = NULL;
2099 bh = NULL;
2100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002103 /* so, if page != NULL, we have a buffer head for the offset at
2104 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2105 ** then we have an unformatted node. Otherwise, we have a direct item,
2106 ** and no zeroing is required on disk. We zero after the truncate,
2107 ** because the truncate might pack the item anyway
2108 ** (it will unmap bh if it packs).
2109 */
2110 /* it is enough to reserve space in transaction for 2 balancings:
2111 one for "save" link adding and another for the first
2112 cut_from_item. 1 is for update_sd */
2113 error = journal_begin(&th, p_s_inode->i_sb,
2114 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2115 if (error)
2116 goto out;
2117 reiserfs_update_inode_transaction(p_s_inode);
2118 if (update_timestamps)
2119 /* we are doing real truncate: if the system crashes before the last
2120 transaction of truncating gets committed - on reboot the file
2121 either appears truncated properly or not truncated at all */
2122 add_save_link(&th, p_s_inode, 1);
Jeff Mahoney24996042005-12-14 14:38:05 -05002123 err2 = reiserfs_do_truncate(&th, p_s_inode, page, update_timestamps);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002124 error =
2125 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2126 if (error)
2127 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128
Jeff Mahoney24996042005-12-14 14:38:05 -05002129 /* check reiserfs_do_truncate after ending the transaction */
2130 if (err2) {
2131 error = err2;
2132 goto out;
2133 }
2134
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002135 if (update_timestamps) {
2136 error = remove_save_link(p_s_inode, 1 /* truncate */ );
2137 if (error)
2138 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002141 if (page) {
2142 length = offset & (blocksize - 1);
2143 /* if we are not on a block boundary */
2144 if (length) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002145 length = blocksize - length;
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002146 zero_user(page, offset, length);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002147 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2148 mark_buffer_dirty(bh);
2149 }
2150 }
2151 unlock_page(page);
2152 page_cache_release(page);
2153 }
2154
2155 reiserfs_write_unlock(p_s_inode->i_sb);
2156 return 0;
2157 out:
2158 if (page) {
2159 unlock_page(page);
2160 page_cache_release(page);
2161 }
2162 reiserfs_write_unlock(p_s_inode->i_sb);
2163 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164}
2165
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002166static int map_block_for_writepage(struct inode *inode,
2167 struct buffer_head *bh_result,
2168 unsigned long block)
2169{
2170 struct reiserfs_transaction_handle th;
2171 int fs_gen;
2172 struct item_head tmp_ih;
2173 struct item_head *ih;
2174 struct buffer_head *bh;
2175 __le32 *item;
2176 struct cpu_key key;
2177 INITIALIZE_PATH(path);
2178 int pos_in_item;
2179 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
Oleg Drokin7729ac52005-11-28 13:43:53 -08002180 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002181 int retval;
2182 int use_get_block = 0;
2183 int bytes_copied = 0;
2184 int copy_size;
2185 int trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002187 /* catch places below that try to log something without starting a trans */
2188 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002190 if (!buffer_uptodate(bh_result)) {
2191 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 }
2193
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002194 kmap(bh_result->b_page);
2195 start_over:
2196 reiserfs_write_lock(inode->i_sb);
2197 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002199 research:
2200 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2201 if (retval != POSITION_FOUND) {
2202 use_get_block = 1;
2203 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204 }
2205
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002206 bh = get_last_bh(&path);
2207 ih = get_ih(&path);
2208 item = get_item(&path);
2209 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002211 /* we've found an unformatted node */
2212 if (indirect_item_found(retval, ih)) {
2213 if (bytes_copied > 0) {
2214 reiserfs_warning(inode->i_sb,
2215 "clm-6002: bytes_copied %d",
2216 bytes_copied);
2217 }
2218 if (!get_block_num(item, pos_in_item)) {
2219 /* crap, we are writing to a hole */
2220 use_get_block = 1;
2221 goto out;
2222 }
2223 set_block_dev_mapped(bh_result,
2224 get_block_num(item, pos_in_item), inode);
2225 } else if (is_direct_le_ih(ih)) {
2226 char *p;
2227 p = page_address(bh_result->b_page);
2228 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2229 copy_size = ih_item_len(ih) - pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002231 fs_gen = get_generation(inode->i_sb);
2232 copy_item_head(&tmp_ih, ih);
2233
2234 if (!trans_running) {
2235 /* vs-3050 is gone, no need to drop the path */
2236 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2237 if (retval)
2238 goto out;
2239 reiserfs_update_inode_transaction(inode);
2240 trans_running = 1;
2241 if (fs_changed(fs_gen, inode->i_sb)
2242 && item_moved(&tmp_ih, &path)) {
2243 reiserfs_restore_prepared_buffer(inode->i_sb,
2244 bh);
2245 goto research;
2246 }
2247 }
2248
2249 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2250
2251 if (fs_changed(fs_gen, inode->i_sb)
2252 && item_moved(&tmp_ih, &path)) {
2253 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2254 goto research;
2255 }
2256
2257 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2258 copy_size);
2259
2260 journal_mark_dirty(&th, inode->i_sb, bh);
2261 bytes_copied += copy_size;
2262 set_block_dev_mapped(bh_result, 0, inode);
2263
2264 /* are there still bytes left? */
2265 if (bytes_copied < bh_result->b_size &&
2266 (byte_offset + bytes_copied) < inode->i_size) {
2267 set_cpu_key_k_offset(&key,
2268 cpu_key_k_offset(&key) +
2269 copy_size);
2270 goto research;
2271 }
2272 } else {
2273 reiserfs_warning(inode->i_sb,
2274 "clm-6003: bad item inode %lu, device %s",
2275 inode->i_ino, reiserfs_bdevname(inode->i_sb));
2276 retval = -EIO;
2277 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002279 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002281 out:
2282 pathrelse(&path);
2283 if (trans_running) {
2284 int err = journal_end(&th, inode->i_sb, jbegin_count);
2285 if (err)
2286 retval = err;
2287 trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002289 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002291 /* this is where we fill in holes in the file. */
2292 if (use_get_block) {
2293 retval = reiserfs_get_block(inode, block, bh_result,
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002294 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002295 | GET_BLOCK_NO_DANGLE);
2296 if (!retval) {
2297 if (!buffer_mapped(bh_result)
2298 || bh_result->b_blocknr == 0) {
2299 /* get_block failed to find a mapped unformatted node. */
2300 use_get_block = 0;
2301 goto start_over;
2302 }
2303 }
2304 }
2305 kunmap(bh_result->b_page);
2306
2307 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2308 /* we've copied data from the page into the direct item, so the
2309 * buffer in the page is now clean, mark it to reflect that.
2310 */
2311 lock_buffer(bh_result);
2312 clear_buffer_dirty(bh_result);
2313 unlock_buffer(bh_result);
2314 }
2315 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316}
2317
2318/*
2319 * mason@suse.com: updated in 2.5.54 to follow the same general io
2320 * start/recovery path as __block_write_full_page, along with special
2321 * code to handle reiserfs tails.
2322 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002323static int reiserfs_write_full_page(struct page *page,
2324 struct writeback_control *wbc)
2325{
2326 struct inode *inode = page->mapping->host;
2327 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2328 int error = 0;
2329 unsigned long block;
Chris Masonb4c76fa2006-08-05 12:15:10 -07002330 sector_t last_block;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002331 struct buffer_head *head, *bh;
2332 int partial = 0;
2333 int nr = 0;
2334 int checked = PageChecked(page);
2335 struct reiserfs_transaction_handle th;
2336 struct super_block *s = inode->i_sb;
2337 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2338 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339
Chris Masone0e851c2006-02-01 03:06:49 -08002340 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2341 if (checked && (current->flags & PF_MEMALLOC)) {
2342 redirty_page_for_writepage(wbc, page);
2343 unlock_page(page);
2344 return 0;
2345 }
2346
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002347 /* The page dirty bit is cleared before writepage is called, which
2348 * means we have to tell create_empty_buffers to make dirty buffers
2349 * The page really should be up to date at this point, so tossing
2350 * in the BH_Uptodate is just a sanity check.
2351 */
2352 if (!page_has_buffers(page)) {
2353 create_empty_buffers(page, s->s_blocksize,
2354 (1 << BH_Dirty) | (1 << BH_Uptodate));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002356 head = page_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002358 /* last page in the file, zero out any contents past the
2359 ** last byte in the file
2360 */
2361 if (page->index >= end_index) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002362 unsigned last_offset;
2363
2364 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2365 /* no file contents in this page */
2366 if (page->index >= end_index + 1 || !last_offset) {
2367 unlock_page(page);
2368 return 0;
2369 }
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002370 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002372 bh = head;
2373 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
Chris Masonb4c76fa2006-08-05 12:15:10 -07002374 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002375 /* first map all the buffers, logging any direct items we find */
2376 do {
Chris Masonb4c76fa2006-08-05 12:15:10 -07002377 if (block > last_block) {
2378 /*
2379 * This can happen when the block size is less than
2380 * the page size. The corresponding bytes in the page
2381 * were zero filled above
2382 */
2383 clear_buffer_dirty(bh);
2384 set_buffer_uptodate(bh);
2385 } else if ((checked || buffer_dirty(bh)) &&
2386 (!buffer_mapped(bh) || (buffer_mapped(bh)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002387 && bh->b_blocknr ==
2388 0))) {
2389 /* not mapped yet, or it points to a direct item, search
2390 * the btree for the mapping info, and log any direct
2391 * items found
2392 */
2393 if ((error = map_block_for_writepage(inode, bh, block))) {
2394 goto fail;
2395 }
2396 }
2397 bh = bh->b_this_page;
2398 block++;
2399 } while (bh != head);
2400
2401 /*
2402 * we start the transaction after map_block_for_writepage,
2403 * because it can create holes in the file (an unbounded operation).
2404 * starting it here, we can make a reliable estimate for how many
2405 * blocks we're going to log
2406 */
2407 if (checked) {
2408 ClearPageChecked(page);
2409 reiserfs_write_lock(s);
2410 error = journal_begin(&th, s, bh_per_page + 1);
2411 if (error) {
2412 reiserfs_write_unlock(s);
2413 goto fail;
2414 }
2415 reiserfs_update_inode_transaction(inode);
2416 }
2417 /* now go through and lock any dirty buffers on the page */
2418 do {
2419 get_bh(bh);
2420 if (!buffer_mapped(bh))
2421 continue;
2422 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2423 continue;
2424
2425 if (checked) {
2426 reiserfs_prepare_for_journal(s, bh, 1);
2427 journal_mark_dirty(&th, s, bh);
2428 continue;
2429 }
2430 /* from this point on, we know the buffer is mapped to a
2431 * real block and not a direct item
2432 */
2433 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2434 lock_buffer(bh);
2435 } else {
2436 if (test_set_buffer_locked(bh)) {
2437 redirty_page_for_writepage(wbc, page);
2438 continue;
2439 }
2440 }
2441 if (test_clear_buffer_dirty(bh)) {
2442 mark_buffer_async_write(bh);
2443 } else {
2444 unlock_buffer(bh);
2445 }
2446 } while ((bh = bh->b_this_page) != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447
2448 if (checked) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002449 error = journal_end(&th, s, bh_per_page + 1);
2450 reiserfs_write_unlock(s);
2451 if (error)
2452 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002454 BUG_ON(PageWriteback(page));
2455 set_page_writeback(page);
2456 unlock_page(page);
2457
2458 /*
2459 * since any buffer might be the only dirty buffer on the page,
2460 * the first submit_bh can bring the page out of writeback.
2461 * be careful with the buffers.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 do {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002464 struct buffer_head *next = bh->b_this_page;
2465 if (buffer_async_write(bh)) {
2466 submit_bh(WRITE, bh);
2467 nr++;
2468 }
2469 put_bh(bh);
2470 bh = next;
2471 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002473 error = 0;
2474 done:
2475 if (nr == 0) {
2476 /*
2477 * if this page only had a direct item, it is very possible for
2478 * no io to be required without there being an error. Or,
2479 * someone else could have locked them and sent them down the
2480 * pipe without locking the page
2481 */
2482 bh = head;
2483 do {
2484 if (!buffer_uptodate(bh)) {
2485 partial = 1;
2486 break;
2487 }
2488 bh = bh->b_this_page;
2489 } while (bh != head);
2490 if (!partial)
2491 SetPageUptodate(page);
2492 end_page_writeback(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002494 return error;
2495
2496 fail:
2497 /* catches various errors, we need to make sure any valid dirty blocks
2498 * get to the media. The page is currently locked and not marked for
2499 * writeback
2500 */
2501 ClearPageUptodate(page);
2502 bh = head;
2503 do {
2504 get_bh(bh);
2505 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2506 lock_buffer(bh);
2507 mark_buffer_async_write(bh);
2508 } else {
2509 /*
2510 * clear any dirty bits that might have come from getting
2511 * attached to a dirty page
2512 */
2513 clear_buffer_dirty(bh);
2514 }
2515 bh = bh->b_this_page;
2516 } while (bh != head);
2517 SetPageError(page);
2518 BUG_ON(PageWriteback(page));
2519 set_page_writeback(page);
2520 unlock_page(page);
2521 do {
2522 struct buffer_head *next = bh->b_this_page;
2523 if (buffer_async_write(bh)) {
2524 clear_buffer_dirty(bh);
2525 submit_bh(WRITE, bh);
2526 nr++;
2527 }
2528 put_bh(bh);
2529 bh = next;
2530 } while (bh != head);
2531 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002532}
2533
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002534static int reiserfs_readpage(struct file *f, struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002536 return block_read_full_page(page, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537}
2538
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002539static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002541 struct inode *inode = page->mapping->host;
2542 reiserfs_wait_on_write_block(inode->i_sb);
2543 return reiserfs_write_full_page(page, wbc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002544}
2545
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002546static int reiserfs_write_begin(struct file *file,
2547 struct address_space *mapping,
2548 loff_t pos, unsigned len, unsigned flags,
2549 struct page **pagep, void **fsdata)
2550{
2551 struct inode *inode;
2552 struct page *page;
2553 pgoff_t index;
2554 int ret;
2555 int old_ref = 0;
2556
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002557 inode = mapping->host;
2558 *fsdata = 0;
2559 if (flags & AOP_FLAG_CONT_EXPAND &&
2560 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2561 pos ++;
2562 *fsdata = (void *)(unsigned long)flags;
2563 }
2564
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002565 index = pos >> PAGE_CACHE_SHIFT;
2566 page = __grab_cache_page(mapping, index);
2567 if (!page)
2568 return -ENOMEM;
2569 *pagep = page;
2570
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002571 reiserfs_wait_on_write_block(inode->i_sb);
2572 fix_tail_page_for_writing(page);
2573 if (reiserfs_transaction_running(inode->i_sb)) {
2574 struct reiserfs_transaction_handle *th;
2575 th = (struct reiserfs_transaction_handle *)current->
2576 journal_info;
2577 BUG_ON(!th->t_refcount);
2578 BUG_ON(!th->t_trans_id);
2579 old_ref = th->t_refcount;
2580 th->t_refcount++;
2581 }
2582 ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
2583 reiserfs_get_block);
2584 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2585 struct reiserfs_transaction_handle *th = current->journal_info;
2586 /* this gets a little ugly. If reiserfs_get_block returned an
2587 * error and left a transacstion running, we've got to close it,
2588 * and we've got to free handle if it was a persistent transaction.
2589 *
2590 * But, if we had nested into an existing transaction, we need
2591 * to just drop the ref count on the handle.
2592 *
2593 * If old_ref == 0, the transaction is from reiserfs_get_block,
2594 * and it was a persistent trans. Otherwise, it was nested above.
2595 */
2596 if (th->t_refcount > old_ref) {
2597 if (old_ref)
2598 th->t_refcount--;
2599 else {
2600 int err;
2601 reiserfs_write_lock(inode->i_sb);
2602 err = reiserfs_end_persistent_transaction(th);
2603 reiserfs_write_unlock(inode->i_sb);
2604 if (err)
2605 ret = err;
2606 }
2607 }
2608 }
2609 if (ret) {
2610 unlock_page(page);
2611 page_cache_release(page);
2612 }
2613 return ret;
2614}
2615
2616int reiserfs_prepare_write(struct file *f, struct page *page,
2617 unsigned from, unsigned to)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002619 struct inode *inode = page->mapping->host;
2620 int ret;
2621 int old_ref = 0;
2622
2623 reiserfs_wait_on_write_block(inode->i_sb);
2624 fix_tail_page_for_writing(page);
2625 if (reiserfs_transaction_running(inode->i_sb)) {
2626 struct reiserfs_transaction_handle *th;
2627 th = (struct reiserfs_transaction_handle *)current->
2628 journal_info;
2629 BUG_ON(!th->t_refcount);
2630 BUG_ON(!th->t_trans_id);
2631 old_ref = th->t_refcount;
2632 th->t_refcount++;
2633 }
2634
2635 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2636 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2637 struct reiserfs_transaction_handle *th = current->journal_info;
2638 /* this gets a little ugly. If reiserfs_get_block returned an
2639 * error and left a transacstion running, we've got to close it,
2640 * and we've got to free handle if it was a persistent transaction.
2641 *
2642 * But, if we had nested into an existing transaction, we need
2643 * to just drop the ref count on the handle.
2644 *
2645 * If old_ref == 0, the transaction is from reiserfs_get_block,
2646 * and it was a persistent trans. Otherwise, it was nested above.
2647 */
2648 if (th->t_refcount > old_ref) {
2649 if (old_ref)
2650 th->t_refcount--;
2651 else {
2652 int err;
2653 reiserfs_write_lock(inode->i_sb);
2654 err = reiserfs_end_persistent_transaction(th);
2655 reiserfs_write_unlock(inode->i_sb);
2656 if (err)
2657 ret = err;
2658 }
2659 }
2660 }
2661 return ret;
2662
2663}
2664
2665static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2666{
2667 return generic_block_bmap(as, block, reiserfs_bmap);
2668}
2669
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002670static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2671 loff_t pos, unsigned len, unsigned copied,
2672 struct page *page, void *fsdata)
2673{
2674 struct inode *inode = page->mapping->host;
2675 int ret = 0;
2676 int update_sd = 0;
2677 struct reiserfs_transaction_handle *th;
2678 unsigned start;
2679
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002680 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2681 pos ++;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002682
2683 reiserfs_wait_on_write_block(inode->i_sb);
2684 if (reiserfs_transaction_running(inode->i_sb))
2685 th = current->journal_info;
2686 else
2687 th = NULL;
2688
2689 start = pos & (PAGE_CACHE_SIZE - 1);
2690 if (unlikely(copied < len)) {
2691 if (!PageUptodate(page))
2692 copied = 0;
2693
2694 page_zero_new_buffers(page, start + copied, start + len);
2695 }
2696 flush_dcache_page(page);
2697
2698 reiserfs_commit_page(inode, page, start, start + copied);
2699
2700 /* generic_commit_write does this for us, but does not update the
2701 ** transaction tracking stuff when the size changes. So, we have
2702 ** to do the i_size updates here.
2703 */
2704 pos += copied;
2705 if (pos > inode->i_size) {
2706 struct reiserfs_transaction_handle myth;
2707 reiserfs_write_lock(inode->i_sb);
2708 /* If the file have grown beyond the border where it
2709 can have a tail, unmark it as needing a tail
2710 packing */
2711 if ((have_large_tails(inode->i_sb)
2712 && inode->i_size > i_block_size(inode) * 4)
2713 || (have_small_tails(inode->i_sb)
2714 && inode->i_size > i_block_size(inode)))
2715 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2716
2717 ret = journal_begin(&myth, inode->i_sb, 1);
2718 if (ret) {
2719 reiserfs_write_unlock(inode->i_sb);
2720 goto journal_error;
2721 }
2722 reiserfs_update_inode_transaction(inode);
2723 inode->i_size = pos;
2724 /*
2725 * this will just nest into our transaction. It's important
2726 * to use mark_inode_dirty so the inode gets pushed around on the
2727 * dirty lists, and so that O_SYNC works as expected
2728 */
2729 mark_inode_dirty(inode);
2730 reiserfs_update_sd(&myth, inode);
2731 update_sd = 1;
2732 ret = journal_end(&myth, inode->i_sb, 1);
2733 reiserfs_write_unlock(inode->i_sb);
2734 if (ret)
2735 goto journal_error;
2736 }
2737 if (th) {
2738 reiserfs_write_lock(inode->i_sb);
2739 if (!update_sd)
2740 mark_inode_dirty(inode);
2741 ret = reiserfs_end_persistent_transaction(th);
2742 reiserfs_write_unlock(inode->i_sb);
2743 if (ret)
2744 goto out;
2745 }
2746
2747 out:
2748 unlock_page(page);
2749 page_cache_release(page);
2750 return ret == 0 ? copied : ret;
2751
2752 journal_error:
2753 if (th) {
2754 reiserfs_write_lock(inode->i_sb);
2755 if (!update_sd)
2756 reiserfs_update_sd(th, inode);
2757 ret = reiserfs_end_persistent_transaction(th);
2758 reiserfs_write_unlock(inode->i_sb);
2759 }
2760
2761 goto out;
2762}
2763
2764int reiserfs_commit_write(struct file *f, struct page *page,
2765 unsigned from, unsigned to)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002766{
2767 struct inode *inode = page->mapping->host;
2768 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2769 int ret = 0;
2770 int update_sd = 0;
2771 struct reiserfs_transaction_handle *th = NULL;
2772
2773 reiserfs_wait_on_write_block(inode->i_sb);
2774 if (reiserfs_transaction_running(inode->i_sb)) {
2775 th = current->journal_info;
2776 }
2777 reiserfs_commit_page(inode, page, from, to);
2778
2779 /* generic_commit_write does this for us, but does not update the
2780 ** transaction tracking stuff when the size changes. So, we have
2781 ** to do the i_size updates here.
2782 */
2783 if (pos > inode->i_size) {
2784 struct reiserfs_transaction_handle myth;
2785 reiserfs_write_lock(inode->i_sb);
2786 /* If the file have grown beyond the border where it
2787 can have a tail, unmark it as needing a tail
2788 packing */
2789 if ((have_large_tails(inode->i_sb)
2790 && inode->i_size > i_block_size(inode) * 4)
2791 || (have_small_tails(inode->i_sb)
2792 && inode->i_size > i_block_size(inode)))
2793 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2794
2795 ret = journal_begin(&myth, inode->i_sb, 1);
2796 if (ret) {
2797 reiserfs_write_unlock(inode->i_sb);
2798 goto journal_error;
2799 }
2800 reiserfs_update_inode_transaction(inode);
2801 inode->i_size = pos;
Chris Mason9f037832005-09-13 01:25:17 -07002802 /*
2803 * this will just nest into our transaction. It's important
2804 * to use mark_inode_dirty so the inode gets pushed around on the
2805 * dirty lists, and so that O_SYNC works as expected
2806 */
2807 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002808 reiserfs_update_sd(&myth, inode);
2809 update_sd = 1;
2810 ret = journal_end(&myth, inode->i_sb, 1);
2811 reiserfs_write_unlock(inode->i_sb);
2812 if (ret)
2813 goto journal_error;
2814 }
2815 if (th) {
2816 reiserfs_write_lock(inode->i_sb);
2817 if (!update_sd)
Chris Mason9f037832005-09-13 01:25:17 -07002818 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002819 ret = reiserfs_end_persistent_transaction(th);
2820 reiserfs_write_unlock(inode->i_sb);
2821 if (ret)
2822 goto out;
2823 }
2824
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002825 out:
2826 return ret;
2827
2828 journal_error:
2829 if (th) {
2830 reiserfs_write_lock(inode->i_sb);
2831 if (!update_sd)
2832 reiserfs_update_sd(th, inode);
2833 ret = reiserfs_end_persistent_transaction(th);
2834 reiserfs_write_unlock(inode->i_sb);
2835 }
2836
2837 return ret;
2838}
2839
2840void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2841{
2842 if (reiserfs_attrs(inode->i_sb)) {
2843 if (sd_attrs & REISERFS_SYNC_FL)
2844 inode->i_flags |= S_SYNC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002846 inode->i_flags &= ~S_SYNC;
2847 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2848 inode->i_flags |= S_IMMUTABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002849 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002850 inode->i_flags &= ~S_IMMUTABLE;
2851 if (sd_attrs & REISERFS_APPEND_FL)
2852 inode->i_flags |= S_APPEND;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002853 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002854 inode->i_flags &= ~S_APPEND;
2855 if (sd_attrs & REISERFS_NOATIME_FL)
2856 inode->i_flags |= S_NOATIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002858 inode->i_flags &= ~S_NOATIME;
2859 if (sd_attrs & REISERFS_NOTAIL_FL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002860 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2861 else
2862 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2863 }
2864}
2865
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002866void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002868 if (reiserfs_attrs(inode->i_sb)) {
2869 if (inode->i_flags & S_IMMUTABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002870 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2871 else
2872 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002873 if (inode->i_flags & S_SYNC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874 *sd_attrs |= REISERFS_SYNC_FL;
2875 else
2876 *sd_attrs &= ~REISERFS_SYNC_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002877 if (inode->i_flags & S_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878 *sd_attrs |= REISERFS_NOATIME_FL;
2879 else
2880 *sd_attrs &= ~REISERFS_NOATIME_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002881 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882 *sd_attrs |= REISERFS_NOTAIL_FL;
2883 else
2884 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2885 }
2886}
2887
2888/* decide if this buffer needs to stay around for data logging or ordered
2889** write purposes
2890*/
2891static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2892{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002893 int ret = 1;
2894 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002895
Chris Masond62b1b82006-02-01 03:06:47 -08002896 lock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002897 spin_lock(&j->j_dirty_buffers_lock);
2898 if (!buffer_mapped(bh)) {
2899 goto free_jh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002901 /* the page is locked, and the only places that log a data buffer
2902 * also lock the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002903 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002904 if (reiserfs_file_data_log(inode)) {
2905 /*
2906 * very conservative, leave the buffer pinned if
2907 * anyone might need it.
2908 */
2909 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2910 ret = 0;
2911 }
Chris Masond62b1b82006-02-01 03:06:47 -08002912 } else if (buffer_dirty(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002913 struct reiserfs_journal_list *jl;
2914 struct reiserfs_jh *jh = bh->b_private;
2915
2916 /* why is this safe?
2917 * reiserfs_setattr updates i_size in the on disk
2918 * stat data before allowing vmtruncate to be called.
2919 *
2920 * If buffer was put onto the ordered list for this
2921 * transaction, we know for sure either this transaction
2922 * or an older one already has updated i_size on disk,
2923 * and this ordered data won't be referenced in the file
2924 * if we crash.
2925 *
2926 * if the buffer was put onto the ordered list for an older
2927 * transaction, we need to leave it around
2928 */
2929 if (jh && (jl = jh->jl)
2930 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2931 ret = 0;
2932 }
2933 free_jh:
2934 if (ret && bh->b_private) {
2935 reiserfs_free_jh(bh);
2936 }
2937 spin_unlock(&j->j_dirty_buffers_lock);
Chris Masond62b1b82006-02-01 03:06:47 -08002938 unlock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002939 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002940}
2941
2942/* clm -- taken from fs/buffer.c:block_invalidate_page */
NeilBrown2ff28e22006-03-26 01:37:18 -08002943static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002944{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002945 struct buffer_head *head, *bh, *next;
2946 struct inode *inode = page->mapping->host;
2947 unsigned int curr_off = 0;
2948 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002949
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002950 BUG_ON(!PageLocked(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002951
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002952 if (offset == 0)
2953 ClearPageChecked(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002954
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002955 if (!page_has_buffers(page))
2956 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002957
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002958 head = page_buffers(page);
2959 bh = head;
2960 do {
2961 unsigned int next_off = curr_off + bh->b_size;
2962 next = bh->b_this_page;
2963
2964 /*
2965 * is this block fully invalidated?
2966 */
2967 if (offset <= curr_off) {
2968 if (invalidatepage_can_drop(inode, bh))
2969 reiserfs_unmap_buffer(bh);
2970 else
2971 ret = 0;
2972 }
2973 curr_off = next_off;
2974 bh = next;
2975 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002976
2977 /*
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002978 * We release buffers only if the entire page is being invalidated.
2979 * The get_block cached value has been unconditionally invalidated,
2980 * so real IO is not possible anymore.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002981 */
NeilBrown2ff28e22006-03-26 01:37:18 -08002982 if (!offset && ret) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002983 ret = try_to_release_page(page, 0);
NeilBrown2ff28e22006-03-26 01:37:18 -08002984 /* maybe should BUG_ON(!ret); - neilb */
2985 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002986 out:
NeilBrown2ff28e22006-03-26 01:37:18 -08002987 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002988}
2989
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002990static int reiserfs_set_page_dirty(struct page *page)
2991{
2992 struct inode *inode = page->mapping->host;
2993 if (reiserfs_file_data_log(inode)) {
2994 SetPageChecked(page);
2995 return __set_page_dirty_nobuffers(page);
2996 }
2997 return __set_page_dirty_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002998}
2999
3000/*
3001 * Returns 1 if the page's buffers were dropped. The page is locked.
3002 *
3003 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3004 * in the buffers at page_buffers(page).
3005 *
3006 * even in -o notail mode, we can't be sure an old mount without -o notail
3007 * didn't create files with tails.
3008 */
Al Viro27496a82005-10-21 03:20:48 -04003009static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003010{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003011 struct inode *inode = page->mapping->host;
3012 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3013 struct buffer_head *head;
3014 struct buffer_head *bh;
3015 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003016
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003017 WARN_ON(PageChecked(page));
3018 spin_lock(&j->j_dirty_buffers_lock);
3019 head = page_buffers(page);
3020 bh = head;
3021 do {
3022 if (bh->b_private) {
3023 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3024 reiserfs_free_jh(bh);
3025 } else {
3026 ret = 0;
3027 break;
3028 }
3029 }
3030 bh = bh->b_this_page;
3031 } while (bh != head);
3032 if (ret)
3033 ret = try_to_free_buffers(page);
3034 spin_unlock(&j->j_dirty_buffers_lock);
3035 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003036}
3037
3038/* We thank Mingming Cao for helping us understand in great detail what
3039 to do in this section of the code. */
3040static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003041 const struct iovec *iov, loff_t offset,
3042 unsigned long nr_segs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003043{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003044 struct file *file = iocb->ki_filp;
3045 struct inode *inode = file->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003046
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003047 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
3048 offset, nr_segs,
3049 reiserfs_get_blocks_direct_io, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003050}
3051
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003052int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3053{
3054 struct inode *inode = dentry->d_inode;
3055 int error;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003056 unsigned int ia_valid;
3057
3058 /* must be turned off for recursive notify_change calls */
3059 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3060
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003061 reiserfs_write_lock(inode->i_sb);
3062 if (attr->ia_valid & ATTR_SIZE) {
3063 /* version 2 items will be caught by the s_maxbytes check
3064 ** done for us in vmtruncate
3065 */
3066 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3067 attr->ia_size > MAX_NON_LFS) {
3068 error = -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003069 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003071 /* fill in hole pointers in the expanding truncate case. */
3072 if (attr->ia_size > inode->i_size) {
Vladimir Savelievf7557e82007-10-16 01:25:14 -07003073 error = generic_cont_expand_simple(inode, attr->ia_size);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003074 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3075 int err;
3076 struct reiserfs_transaction_handle th;
3077 /* we're changing at most 2 bitmaps, inode + super */
3078 err = journal_begin(&th, inode->i_sb, 4);
3079 if (!err) {
3080 reiserfs_discard_prealloc(&th, inode);
3081 err = journal_end(&th, inode->i_sb, 4);
3082 }
3083 if (err)
3084 error = err;
3085 }
3086 if (error)
3087 goto out;
Vladimir Savelievdd535a52006-07-01 04:36:32 -07003088 /*
3089 * file size is changed, ctime and mtime are
3090 * to be updated
3091 */
3092 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003093 }
3094 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003095
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003096 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
3097 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
3098 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3099 /* stat data of format v3.5 has 16 bit uid and gid */
3100 error = -EINVAL;
3101 goto out;
3102 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003103
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003104 error = inode_change_ok(inode, attr);
3105 if (!error) {
3106 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3107 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
3108 error = reiserfs_chown_xattrs(inode, attr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003109
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003110 if (!error) {
3111 struct reiserfs_transaction_handle th;
3112 int jbegin_count =
3113 2 *
3114 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3115 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3116 2;
3117
3118 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3119 error =
3120 journal_begin(&th, inode->i_sb,
3121 jbegin_count);
3122 if (error)
3123 goto out;
3124 error =
3125 DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
3126 if (error) {
3127 journal_end(&th, inode->i_sb,
3128 jbegin_count);
3129 goto out;
3130 }
3131 /* Update corresponding info in inode so that everything is in
3132 * one transaction */
3133 if (attr->ia_valid & ATTR_UID)
3134 inode->i_uid = attr->ia_uid;
3135 if (attr->ia_valid & ATTR_GID)
3136 inode->i_gid = attr->ia_gid;
3137 mark_inode_dirty(inode);
3138 error =
3139 journal_end(&th, inode->i_sb, jbegin_count);
3140 }
3141 }
3142 if (!error)
3143 error = inode_setattr(inode, attr);
3144 }
3145
3146 if (!error && reiserfs_posixacl(inode->i_sb)) {
3147 if (attr->ia_valid & ATTR_MODE)
3148 error = reiserfs_acl_chmod(inode);
3149 }
3150
3151 out:
3152 reiserfs_write_unlock(inode->i_sb);
3153 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003154}
3155
Christoph Hellwigf5e54d62006-06-28 04:26:44 -07003156const struct address_space_operations reiserfs_address_space_operations = {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003157 .writepage = reiserfs_writepage,
3158 .readpage = reiserfs_readpage,
3159 .readpages = reiserfs_readpages,
3160 .releasepage = reiserfs_releasepage,
3161 .invalidatepage = reiserfs_invalidatepage,
3162 .sync_page = block_sync_page,
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07003163 .write_begin = reiserfs_write_begin,
3164 .write_end = reiserfs_write_end,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003165 .bmap = reiserfs_aop_bmap,
3166 .direct_IO = reiserfs_direct_IO,
3167 .set_page_dirty = reiserfs_set_page_dirty,
3168};