blob: a14d6cd9eeda2670251c0ec3377518912cb465a2 [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 Mahoneyeb35c212008-07-08 14:37:06 -040048 reiserfs_discard_prealloc(&th, inode);
49
Jeff Mahoney24996042005-12-14 14:38:05 -050050 err = reiserfs_delete_object(&th, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070051
52 /* Do quota update inside a transaction for journaled quotas. We must do that
53 * after delete_object so that quota updates go into the same transaction as
54 * stat data deletion */
Jeff Mahoney24996042005-12-14 14:38:05 -050055 if (!err)
Jan Kara77db4f22009-01-26 17:14:18 +010056 vfs_dq_free_inode(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070057
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070058 if (journal_end(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070059 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070060
Jeff Mahoney24996042005-12-14 14:38:05 -050061 /* check return value from reiserfs_delete_object after
62 * ending the transaction
63 */
64 if (err)
65 goto out;
66
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070067 /* all items of file are deleted, so we can remove "save" link */
68 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
69 * about an error here */
70 } else {
71 /* no object items are in the tree */
72 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070074 out:
75 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
76 inode->i_blocks = 0;
77 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078}
79
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070080static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
81 __u32 objectid, loff_t offset, int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070082{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070083 key->version = version;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070085 key->on_disk_key.k_dir_id = dirid;
86 key->on_disk_key.k_objectid = objectid;
87 set_cpu_key_k_offset(key, offset);
88 set_cpu_key_k_type(key, type);
89 key->key_length = length;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090}
91
Linus Torvalds1da177e2005-04-16 15:20:36 -070092/* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
93 offset and type of key */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070094void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
95 int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070096{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070097 _make_cpu_key(key, get_inode_item_key_version(inode),
98 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
99 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
100 length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101}
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103//
104// when key is 0, do not set version and short key
105//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700106inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
107 int version,
108 loff_t offset, int type, int length,
109 int entry_count /*or ih_free_space */ )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700111 if (key) {
112 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
113 ih->ih_key.k_objectid =
114 cpu_to_le32(key->on_disk_key.k_objectid);
115 }
116 put_ih_version(ih, version);
117 set_le_ih_k_offset(ih, offset);
118 set_le_ih_k_type(ih, type);
119 put_ih_item_len(ih, length);
120 /* set_ih_free_space (ih, 0); */
121 // for directory items it is entry count, for directs and stat
122 // datas - 0xffff, for indirects - 0
123 put_ih_entry_count(ih, entry_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124}
125
126//
127// FIXME: we might cache recently accessed indirect item
128
129// Ugh. Not too eager for that....
130// I cut the code until such time as I see a convincing argument (benchmark).
131// I don't want a bloated inode struct..., and I don't like code complexity....
132
133/* cutting the code is fine, since it really isn't in use yet and is easy
134** to add back in. But, Vladimir has a really good idea here. Think
135** about what happens for reading a file. For each page,
136** The VFS layer calls reiserfs_readpage, who searches the tree to find
137** an indirect item. This indirect item has X number of pointers, where
138** X is a big number if we've done the block allocation right. But,
139** we only use one or two of these pointers during each call to readpage,
140** needlessly researching again later on.
141**
142** The size of the cache could be dynamic based on the size of the file.
143**
144** I'd also like to see us cache the location the stat data item, since
145** we are needlessly researching for that frequently.
146**
147** --chris
148*/
149
150/* If this page has a file tail in it, and
151** it was read in by get_block_create_0, the page data is valid,
152** but tail is still sitting in a direct item, and we can't write to
153** it. So, look through this page, and check all the mapped buffers
154** to make sure they have valid block numbers. Any that don't need
155** to be unmapped, so that block_prepare_write will correctly call
156** reiserfs_get_block to convert the tail into an unformatted node
157*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700158static inline void fix_tail_page_for_writing(struct page *page)
159{
160 struct buffer_head *head, *next, *bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700162 if (page && page_has_buffers(page)) {
163 head = page_buffers(page);
164 bh = head;
165 do {
166 next = bh->b_this_page;
167 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
168 reiserfs_unmap_buffer(bh);
169 }
170 bh = next;
171 } while (bh != head);
172 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173}
174
175/* reiserfs_get_block does not need to allocate a block only if it has been
176 done already or non-hole position has been found in the indirect item */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700177static inline int allocation_needed(int retval, b_blocknr_t allocated,
178 struct item_head *ih,
179 __le32 * item, int pos_in_item)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700181 if (allocated)
182 return 0;
183 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
184 get_block_num(item, pos_in_item))
185 return 0;
186 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187}
188
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700189static inline int indirect_item_found(int retval, struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700191 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192}
193
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700194static inline void set_block_dev_mapped(struct buffer_head *bh,
195 b_blocknr_t block, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196{
197 map_bh(bh, inode->i_sb, block);
198}
199
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200//
201// files which were created in the earlier version can not be longer,
202// than 2 gb
203//
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700204static int file_capable(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700206 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
207 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
208 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700210 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211}
212
Adrian Bunkdeba0f42007-10-16 23:26:03 -0700213static int restart_transaction(struct reiserfs_transaction_handle *th,
214 struct inode *inode, struct treepath *path)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700215{
216 struct super_block *s = th->t_super;
217 int len = th->t_blocks_allocated;
218 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700220 BUG_ON(!th->t_trans_id);
221 BUG_ON(!th->t_refcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Suzuki K P87b41262006-12-06 20:36:10 -0800223 pathrelse(path);
224
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700225 /* we cannot restart while nested */
226 if (th->t_refcount > 1) {
227 return 0;
228 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700229 reiserfs_update_sd(th, inode);
230 err = journal_end(th, s, len);
231 if (!err) {
232 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
233 if (!err)
234 reiserfs_update_inode_transaction(inode);
235 }
236 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237}
238
239// it is called by get_block when create == 0. Returns block number
240// for 'block'-th logical block of file. When it hits direct item it
241// returns 0 (being called from bmap) or read direct item into piece
242// of page (bh_result)
243
244// Please improve the english/clarity in the comment above, as it is
245// hard to understand.
246
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700247static int _get_block_create_0(struct inode *inode, sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700248 struct buffer_head *bh_result, int args)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700250 INITIALIZE_PATH(path);
251 struct cpu_key key;
252 struct buffer_head *bh;
253 struct item_head *ih, tmp_ih;
254 int fs_gen;
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700255 b_blocknr_t blocknr;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700256 char *p = NULL;
257 int chars;
258 int ret;
259 int result;
260 int done = 0;
261 unsigned long offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700263 // prepare the key to look for the 'block'-th block of file
264 make_cpu_key(&key, inode,
265 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
266 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700268 research:
269 result = search_for_position_by_key(inode->i_sb, &key, &path);
270 if (result != POSITION_FOUND) {
271 pathrelse(&path);
272 if (p)
273 kunmap(bh_result->b_page);
274 if (result == IO_ERROR)
275 return -EIO;
276 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
277 // That there is some MMAPED data associated with it that is yet to be written to disk.
278 if ((args & GET_BLOCK_NO_HOLE)
279 && !PageUptodate(bh_result->b_page)) {
280 return -ENOENT;
281 }
282 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700284 //
285 bh = get_last_bh(&path);
286 ih = get_ih(&path);
287 if (is_indirect_le_ih(ih)) {
288 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700290 /* FIXME: here we could cache indirect item or part of it in
291 the inode to avoid search_by_key in case of subsequent
292 access to file */
293 blocknr = get_block_num(ind_item, path.pos_in_item);
294 ret = 0;
295 if (blocknr) {
296 map_bh(bh_result, inode->i_sb, blocknr);
297 if (path.pos_in_item ==
298 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
299 set_buffer_boundary(bh_result);
300 }
301 } else
302 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
303 // That there is some MMAPED data associated with it that is yet to be written to disk.
304 if ((args & GET_BLOCK_NO_HOLE)
305 && !PageUptodate(bh_result->b_page)) {
306 ret = -ENOENT;
307 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700309 pathrelse(&path);
310 if (p)
311 kunmap(bh_result->b_page);
312 return ret;
313 }
314 // requested data are in direct item(s)
315 if (!(args & GET_BLOCK_READ_DIRECT)) {
316 // we are called by bmap. FIXME: we can not map block of file
317 // when it is stored in direct item(s)
318 pathrelse(&path);
319 if (p)
320 kunmap(bh_result->b_page);
321 return -ENOENT;
322 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700324 /* if we've got a direct item, and the buffer or page was uptodate,
325 ** we don't want to pull data off disk again. skip to the
326 ** end, where we map the buffer and return
327 */
328 if (buffer_uptodate(bh_result)) {
329 goto finished;
330 } else
331 /*
332 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
333 ** pages without any buffers. If the page is up to date, we don't want
334 ** read old data off disk. Set the up to date bit on the buffer instead
335 ** and jump to the end
336 */
337 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 set_buffer_uptodate(bh_result);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700339 goto finished;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700341 // read file tail into part of page
342 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
343 fs_gen = get_generation(inode->i_sb);
344 copy_item_head(&tmp_ih, ih);
345
346 /* we only want to kmap if we are reading the tail into the page.
347 ** this is not the common case, so we don't kmap until we are
348 ** sure we need to. But, this means the item might move if
349 ** kmap schedules
350 */
351 if (!p) {
352 p = (char *)kmap(bh_result->b_page);
353 if (fs_changed(fs_gen, inode->i_sb)
354 && item_moved(&tmp_ih, &path)) {
355 goto research;
356 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700358 p += offset;
359 memset(p, 0, inode->i_sb->s_blocksize);
360 do {
361 if (!is_direct_le_ih(ih)) {
362 BUG();
363 }
364 /* make sure we don't read more bytes than actually exist in
365 ** the file. This can happen in odd cases where i_size isn't
Jeff Mahoney0222e652009-03-30 14:02:44 -0400366 ** correct, and when direct item padding results in a few
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700367 ** extra bytes at the end of the direct item
368 */
369 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
370 break;
371 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
372 chars =
373 inode->i_size - (le_ih_k_offset(ih) - 1) -
374 path.pos_in_item;
375 done = 1;
376 } else {
377 chars = ih_item_len(ih) - path.pos_in_item;
378 }
379 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700381 if (done)
382 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700384 p += chars;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700386 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
387 // we done, if read direct item is not the last item of
388 // node FIXME: we could try to check right delimiting key
389 // to see whether direct item continues in the right
390 // neighbor or rely on i_size
391 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700393 // update key to look for the next piece
394 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
395 result = search_for_position_by_key(inode->i_sb, &key, &path);
396 if (result != POSITION_FOUND)
397 // i/o error most likely
398 break;
399 bh = get_last_bh(&path);
400 ih = get_ih(&path);
401 } while (1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700403 flush_dcache_page(bh_result->b_page);
404 kunmap(bh_result->b_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700406 finished:
407 pathrelse(&path);
Qu Fuping6283d582005-06-25 14:55:44 -0700408
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700409 if (result == IO_ERROR)
410 return -EIO;
Qu Fuping6283d582005-06-25 14:55:44 -0700411
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700412 /* this buffer has valid data, but isn't valid for io. mapping it to
413 * block #0 tells the rest of reiserfs it just has a tail in it
414 */
415 map_bh(bh_result, inode->i_sb, 0);
416 set_buffer_uptodate(bh_result);
417 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418}
419
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420// this is called to create file map. So, _get_block_create_0 will not
421// read direct item
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700422static int reiserfs_bmap(struct inode *inode, sector_t block,
423 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700425 if (!file_capable(inode, block))
426 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700428 reiserfs_write_lock(inode->i_sb);
429 /* do not read the direct item */
430 _get_block_create_0(inode, block, bh_result, 0);
431 reiserfs_write_unlock(inode->i_sb);
432 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433}
434
435/* special version of get_block that is only used by grab_tail_page right
436** now. It is sent to block_prepare_write, and when you try to get a
437** block past the end of the file (or a block from a hole) it returns
438** -ENOENT instead of a valid buffer. block_prepare_write expects to
439** be able to do i/o on the buffers returned, unless an error value
440** is also returned.
Jeff Mahoney0222e652009-03-30 14:02:44 -0400441**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442** So, this allows block_prepare_write to be used for reading a single block
443** in a page. Where it does not produce a valid page for holes, or past the
444** end of the file. This turns out to be exactly what we need for reading
445** tails for conversion.
446**
447** The point of the wrapper is forcing a certain value for create, even
Jeff Mahoney0222e652009-03-30 14:02:44 -0400448** though the VFS layer is calling this function with create==1. If you
449** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450** don't use this function.
451*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700452static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
453 struct buffer_head *bh_result,
454 int create)
455{
456 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457}
458
459/* This is special helper for reiserfs_get_block in case we are executing
460 direct_IO request. */
461static int reiserfs_get_blocks_direct_io(struct inode *inode,
462 sector_t iblock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 struct buffer_head *bh_result,
464 int create)
465{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700466 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700468 bh_result->b_page = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700470 /* We set the b_size before reiserfs_get_block call since it is
471 referenced in convert_tail_for_hole() that may be called from
472 reiserfs_get_block() */
473 bh_result->b_size = (1 << inode->i_blkbits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700475 ret = reiserfs_get_block(inode, iblock, bh_result,
476 create | GET_BLOCK_NO_DANGLE);
477 if (ret)
478 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700480 /* don't allow direct io onto tail pages */
481 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
482 /* make sure future calls to the direct io funcs for this offset
483 ** in the file fail by unmapping the buffer
484 */
485 clear_buffer_mapped(bh_result);
486 ret = -EINVAL;
487 }
488 /* Possible unpacked tail. Flush the data before pages have
489 disappeared */
490 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
491 int err;
492 lock_kernel();
493 err = reiserfs_commit_for_inode(inode);
494 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
495 unlock_kernel();
496 if (err < 0)
497 ret = err;
498 }
499 out:
500 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501}
502
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503/*
504** helper function for when reiserfs_get_block is called for a hole
505** but the file tail is still in a direct item
506** bh_result is the buffer head for the hole
507** tail_offset is the offset of the start of the tail in the file
508**
509** This calls prepare_write, which will start a new transaction
510** you should not be in a transaction, or have any paths held when you
511** call this.
512*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700513static int convert_tail_for_hole(struct inode *inode,
514 struct buffer_head *bh_result,
515 loff_t tail_offset)
516{
517 unsigned long index;
518 unsigned long tail_end;
519 unsigned long tail_start;
520 struct page *tail_page;
521 struct page *hole_page = bh_result->b_page;
522 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700524 if ((tail_offset & (bh_result->b_size - 1)) != 1)
525 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700527 /* always try to read until the end of the block */
528 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
529 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700531 index = tail_offset >> PAGE_CACHE_SHIFT;
532 /* hole_page can be zero in case of direct_io, we are sure
533 that we cannot get here if we write with O_DIRECT into
534 tail page */
535 if (!hole_page || index != hole_page->index) {
536 tail_page = grab_cache_page(inode->i_mapping, index);
537 retval = -ENOMEM;
538 if (!tail_page) {
539 goto out;
540 }
541 } else {
542 tail_page = hole_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700545 /* we don't have to make sure the conversion did not happen while
546 ** we were locking the page because anyone that could convert
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800547 ** must first take i_mutex.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700548 **
549 ** We must fix the tail page for writing because it might have buffers
550 ** that are mapped, but have a block number of 0. This indicates tail
551 ** data that has been read directly into the page, and block_prepare_write
552 ** won't trigger a get_block in this case.
553 */
554 fix_tail_page_for_writing(tail_page);
555 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
556 if (retval)
557 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700559 /* tail conversion might change the data in the page */
560 flush_dcache_page(tail_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700562 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700564 unlock:
565 if (tail_page != hole_page) {
566 unlock_page(tail_page);
567 page_cache_release(tail_page);
568 }
569 out:
570 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571}
572
573static inline int _allocate_block(struct reiserfs_transaction_handle *th,
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700574 sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700575 struct inode *inode,
576 b_blocknr_t * allocated_block_nr,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -0800577 struct treepath *path, int flags)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700578{
579 BUG_ON(!th->t_trans_id);
580
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581#ifdef REISERFS_PREALLOCATE
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800582 if (!(flags & GET_BLOCK_NO_IMUX)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700583 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
584 path, block);
585 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700587 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
588 block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589}
590
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700591int reiserfs_get_block(struct inode *inode, sector_t block,
592 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700594 int repeat, retval = 0;
595 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
596 INITIALIZE_PATH(path);
597 int pos_in_item;
598 struct cpu_key key;
599 struct buffer_head *bh, *unbh = NULL;
600 struct item_head *ih, tmp_ih;
601 __le32 *item;
602 int done;
603 int fs_gen;
604 struct reiserfs_transaction_handle *th = NULL;
Jeff Mahoney0222e652009-03-30 14:02:44 -0400605 /* space reserved in transaction batch:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700606 . 3 balancings in direct->indirect conversion
607 . 1 block involved into reiserfs_update_sd()
608 XXX in practically impossible worst case direct2indirect()
609 can incur (much) more than 3 balancings.
610 quota update for user, group */
611 int jbegin_count =
612 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
613 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
614 int version;
615 int dangle = 1;
616 loff_t new_offset =
617 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700619 /* bad.... */
620 reiserfs_write_lock(inode->i_sb);
621 version = get_inode_item_key_version(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700623 if (!file_capable(inode, block)) {
624 reiserfs_write_unlock(inode->i_sb);
625 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
627
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700628 /* if !create, we aren't changing the FS, so we don't need to
629 ** log anything, so we don't need to start a transaction
630 */
631 if (!(create & GET_BLOCK_CREATE)) {
632 int ret;
633 /* find number of block-th logical block of the file */
634 ret = _get_block_create_0(inode, block, bh_result,
635 create | GET_BLOCK_READ_DIRECT);
636 reiserfs_write_unlock(inode->i_sb);
637 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700639 /*
640 * if we're already in a transaction, make sure to close
641 * any new transactions we start in this func
642 */
643 if ((create & GET_BLOCK_NO_DANGLE) ||
644 reiserfs_transaction_running(inode->i_sb))
645 dangle = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700647 /* If file is of such a size, that it might have a tail and tails are enabled
648 ** we should mark it as possibly needing tail packing on close
649 */
650 if ((have_large_tails(inode->i_sb)
651 && inode->i_size < i_block_size(inode) * 4)
652 || (have_small_tails(inode->i_sb)
653 && inode->i_size < i_block_size(inode)))
654 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
655
656 /* set the key of the first byte in the 'block'-th block of file */
657 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
658 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
659 start_trans:
660 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
661 if (!th) {
662 retval = -ENOMEM;
663 goto failure;
664 }
665 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700667 research:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700669 retval = search_for_position_by_key(inode->i_sb, &key, &path);
670 if (retval == IO_ERROR) {
671 retval = -EIO;
672 goto failure;
673 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700675 bh = get_last_bh(&path);
676 ih = get_ih(&path);
677 item = get_item(&path);
678 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700680 fs_gen = get_generation(inode->i_sb);
681 copy_item_head(&tmp_ih, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700683 if (allocation_needed
684 (retval, allocated_block_nr, ih, item, pos_in_item)) {
685 /* we have to allocate block for the unformatted node */
686 if (!th) {
687 pathrelse(&path);
688 goto start_trans;
689 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700691 repeat =
692 _allocate_block(th, block, inode, &allocated_block_nr,
693 &path, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700695 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
696 /* restart the transaction to give the journal a chance to free
697 ** some blocks. releases the path, so we have to go back to
698 ** research if we succeed on the second try
699 */
700 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
701 retval = restart_transaction(th, inode, &path);
702 if (retval)
703 goto failure;
704 repeat =
705 _allocate_block(th, block, inode,
706 &allocated_block_nr, NULL, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700708 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
709 goto research;
710 }
711 if (repeat == QUOTA_EXCEEDED)
712 retval = -EDQUOT;
713 else
714 retval = -ENOSPC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 goto failure;
716 }
717
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700718 if (fs_changed(fs_gen, inode->i_sb)
719 && item_moved(&tmp_ih, &path)) {
720 goto research;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700722 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700724 if (indirect_item_found(retval, ih)) {
725 b_blocknr_t unfm_ptr;
726 /* 'block'-th block is in the file already (there is
727 corresponding cell in some indirect item). But it may be
728 zero unformatted node pointer (hole) */
729 unfm_ptr = get_block_num(item, pos_in_item);
730 if (unfm_ptr == 0) {
731 /* use allocated block to plug the hole */
732 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
733 if (fs_changed(fs_gen, inode->i_sb)
734 && item_moved(&tmp_ih, &path)) {
735 reiserfs_restore_prepared_buffer(inode->i_sb,
736 bh);
737 goto research;
738 }
739 set_buffer_new(bh_result);
740 if (buffer_dirty(bh_result)
741 && reiserfs_data_ordered(inode->i_sb))
742 reiserfs_add_ordered_list(inode, bh_result);
743 put_block_num(item, pos_in_item, allocated_block_nr);
744 unfm_ptr = allocated_block_nr;
745 journal_mark_dirty(th, inode->i_sb, bh);
746 reiserfs_update_sd(th, inode);
747 }
748 set_block_dev_mapped(bh_result, unfm_ptr, inode);
749 pathrelse(&path);
750 retval = 0;
751 if (!dangle && th)
752 retval = reiserfs_end_persistent_transaction(th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700754 reiserfs_write_unlock(inode->i_sb);
755
756 /* the item was found, so new blocks were not added to the file
Jeff Mahoney0222e652009-03-30 14:02:44 -0400757 ** there is no need to make sure the inode is updated with this
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700758 ** transaction
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700760 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 }
762
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700763 if (!th) {
764 pathrelse(&path);
765 goto start_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700768 /* desired position is not found or is in the direct item. We have
769 to append file with holes up to 'block'-th block converting
770 direct items to indirect one if necessary */
771 done = 0;
772 do {
773 if (is_statdata_le_ih(ih)) {
774 __le32 unp = 0;
775 struct cpu_key tmp_key;
776
777 /* indirect item has to be inserted */
778 make_le_item_head(&tmp_ih, &key, version, 1,
779 TYPE_INDIRECT, UNFM_P_SIZE,
780 0 /* free_space */ );
781
782 if (cpu_key_k_offset(&key) == 1) {
783 /* we are going to add 'block'-th block to the file. Use
784 allocated block for that */
785 unp = cpu_to_le32(allocated_block_nr);
786 set_block_dev_mapped(bh_result,
787 allocated_block_nr, inode);
788 set_buffer_new(bh_result);
789 done = 1;
790 }
791 tmp_key = key; // ;)
792 set_cpu_key_k_offset(&tmp_key, 1);
793 PATH_LAST_POSITION(&path)++;
794
795 retval =
796 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
797 inode, (char *)&unp);
798 if (retval) {
799 reiserfs_free_block(th, inode,
800 allocated_block_nr, 1);
801 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
802 }
803 //mark_tail_converted (inode);
804 } else if (is_direct_le_ih(ih)) {
805 /* direct item has to be converted */
806 loff_t tail_offset;
807
808 tail_offset =
809 ((le_ih_k_offset(ih) -
810 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
811 if (tail_offset == cpu_key_k_offset(&key)) {
812 /* direct item we just found fits into block we have
813 to map. Convert it into unformatted node: use
814 bh_result for the conversion */
815 set_block_dev_mapped(bh_result,
816 allocated_block_nr, inode);
817 unbh = bh_result;
818 done = 1;
819 } else {
820 /* we have to padd file tail stored in direct item(s)
821 up to block size and convert it to unformatted
822 node. FIXME: this should also get into page cache */
823
824 pathrelse(&path);
825 /*
826 * ugly, but we can only end the transaction if
827 * we aren't nested
828 */
829 BUG_ON(!th->t_refcount);
830 if (th->t_refcount == 1) {
831 retval =
832 reiserfs_end_persistent_transaction
833 (th);
834 th = NULL;
835 if (retval)
836 goto failure;
837 }
838
839 retval =
840 convert_tail_for_hole(inode, bh_result,
841 tail_offset);
842 if (retval) {
843 if (retval != -ENOSPC)
Jeff Mahoney0030b642009-03-30 14:02:28 -0400844 reiserfs_error(inode->i_sb,
845 "clm-6004",
846 "convert tail failed "
847 "inode %lu, error %d",
848 inode->i_ino,
849 retval);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700850 if (allocated_block_nr) {
851 /* the bitmap, the super, and the stat data == 3 */
852 if (!th)
853 th = reiserfs_persistent_transaction(inode->i_sb, 3);
854 if (th)
855 reiserfs_free_block(th,
856 inode,
857 allocated_block_nr,
858 1);
859 }
860 goto failure;
861 }
862 goto research;
863 }
864 retval =
865 direct2indirect(th, inode, &path, unbh,
866 tail_offset);
867 if (retval) {
868 reiserfs_unmap_buffer(unbh);
869 reiserfs_free_block(th, inode,
870 allocated_block_nr, 1);
871 goto failure;
872 }
873 /* it is important the set_buffer_uptodate is done after
874 ** the direct2indirect. The buffer might contain valid
875 ** data newer than the data on disk (read by readpage, changed,
876 ** and then sent here by writepage). direct2indirect needs
877 ** to know if unbh was already up to date, so it can decide
878 ** if the data in unbh needs to be replaced with data from
879 ** the disk
880 */
881 set_buffer_uptodate(unbh);
882
883 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
884 buffer will disappear shortly, so it should not be added to
885 */
886 if (unbh->b_page) {
887 /* we've converted the tail, so we must
888 ** flush unbh before the transaction commits
889 */
890 reiserfs_add_tail_list(inode, unbh);
891
892 /* mark it dirty now to prevent commit_write from adding
893 ** this buffer to the inode's dirty buffer list
894 */
895 /*
896 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
897 * It's still atomic, but it sets the page dirty too,
898 * which makes it eligible for writeback at any time by the
899 * VM (which was also the case with __mark_buffer_dirty())
900 */
901 mark_buffer_dirty(unbh);
902 }
903 } else {
904 /* append indirect item with holes if needed, when appending
905 pointer to 'block'-th block use block, which is already
906 allocated */
907 struct cpu_key tmp_key;
908 unp_t unf_single = 0; // We use this in case we need to allocate only
909 // one block which is a fastpath
910 unp_t *un;
911 __u64 max_to_insert =
912 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
913 UNFM_P_SIZE;
914 __u64 blocks_needed;
915
916 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
917 "vs-804: invalid position for append");
918 /* indirect item has to be appended, set up key of that position */
919 make_cpu_key(&tmp_key, inode,
920 le_key_k_offset(version,
921 &(ih->ih_key)) +
922 op_bytes_number(ih,
923 inode->i_sb->s_blocksize),
924 //pos_in_item * inode->i_sb->s_blocksize,
925 TYPE_INDIRECT, 3); // key type is unimportant
926
Vladimir V. Savelievc499ec22006-03-02 02:54:39 -0800927 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
928 "green-805: invalid offset");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700929 blocks_needed =
930 1 +
931 ((cpu_key_k_offset(&key) -
932 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
933 s_blocksize_bits);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700934
935 if (blocks_needed == 1) {
936 un = &unf_single;
937 } else {
Yan Burman01afb212006-12-06 20:39:01 -0800938 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 -0700939 if (!un) {
940 un = &unf_single;
941 blocks_needed = 1;
942 max_to_insert = 0;
Yan Burman01afb212006-12-06 20:39:01 -0800943 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700944 }
945 if (blocks_needed <= max_to_insert) {
946 /* we are going to add target block to the file. Use allocated
947 block for that */
948 un[blocks_needed - 1] =
949 cpu_to_le32(allocated_block_nr);
950 set_block_dev_mapped(bh_result,
951 allocated_block_nr, inode);
952 set_buffer_new(bh_result);
953 done = 1;
954 } else {
955 /* paste hole to the indirect item */
956 /* If kmalloc failed, max_to_insert becomes zero and it means we
957 only have space for one block */
958 blocks_needed =
959 max_to_insert ? max_to_insert : 1;
960 }
961 retval =
962 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
963 (char *)un,
964 UNFM_P_SIZE *
965 blocks_needed);
966
967 if (blocks_needed != 1)
968 kfree(un);
969
970 if (retval) {
971 reiserfs_free_block(th, inode,
972 allocated_block_nr, 1);
973 goto failure;
974 }
975 if (!done) {
976 /* We need to mark new file size in case this function will be
977 interrupted/aborted later on. And we may do this only for
978 holes. */
979 inode->i_size +=
980 inode->i_sb->s_blocksize * blocks_needed;
981 }
982 }
983
984 if (done == 1)
985 break;
986
987 /* this loop could log more blocks than we had originally asked
988 ** for. So, we have to allow the transaction to end if it is
Jeff Mahoney0222e652009-03-30 14:02:44 -0400989 ** too big or too full. Update the inode so things are
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700990 ** consistent if we crash before the function returns
991 **
992 ** release the path so that anybody waiting on the path before
993 ** ending their transaction will be able to continue.
994 */
995 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
996 retval = restart_transaction(th, inode, &path);
997 if (retval)
998 goto failure;
999 }
Jeff Mahoney0222e652009-03-30 14:02:44 -04001000 /* inserting indirect pointers for a hole can take a
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001001 ** long time. reschedule if needed
1002 */
1003 cond_resched();
1004
1005 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1006 if (retval == IO_ERROR) {
1007 retval = -EIO;
1008 goto failure;
1009 }
1010 if (retval == POSITION_FOUND) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001011 reiserfs_warning(inode->i_sb, "vs-825",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001012 "%K should not be found", &key);
1013 retval = -EEXIST;
1014 if (allocated_block_nr)
1015 reiserfs_free_block(th, inode,
1016 allocated_block_nr, 1);
1017 pathrelse(&path);
1018 goto failure;
1019 }
1020 bh = get_last_bh(&path);
1021 ih = get_ih(&path);
1022 item = get_item(&path);
1023 pos_in_item = path.pos_in_item;
1024 } while (1);
1025
1026 retval = 0;
1027
1028 failure:
1029 if (th && (!dangle || (retval && !th->t_trans_id))) {
1030 int err;
1031 if (th->t_trans_id)
1032 reiserfs_update_sd(th, inode);
1033 err = reiserfs_end_persistent_transaction(th);
1034 if (err)
1035 retval = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001038 reiserfs_write_unlock(inode->i_sb);
1039 reiserfs_check_path(&path);
1040 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041}
1042
1043static int
1044reiserfs_readpages(struct file *file, struct address_space *mapping,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001045 struct list_head *pages, unsigned nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001047 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048}
1049
1050/* Compute real number of used bytes by file
1051 * Following three functions can go away when we'll have enough space in stat item
1052 */
1053static int real_space_diff(struct inode *inode, int sd_size)
1054{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001055 int bytes;
1056 loff_t blocksize = inode->i_sb->s_blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001058 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1059 return sd_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001061 /* End of file is also in full block with indirect reference, so round
1062 ** up to the next block.
1063 **
1064 ** there is just no way to know if the tail is actually packed
1065 ** on the file, so we have to assume it isn't. When we pack the
1066 ** tail, we add 4 bytes to pretend there really is an unformatted
1067 ** node pointer
1068 */
1069 bytes =
1070 ((inode->i_size +
1071 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1072 sd_size;
1073 return bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074}
1075
1076static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001077 int sd_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001079 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1080 return inode->i_size +
1081 (loff_t) (real_space_diff(inode, sd_size));
1082 }
1083 return ((loff_t) real_space_diff(inode, sd_size)) +
1084 (((loff_t) blocks) << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085}
1086
1087/* Compute number of blocks used by file in ReiserFS counting */
1088static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1089{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001090 loff_t bytes = inode_get_bytes(inode);
1091 loff_t real_space = real_space_diff(inode, sd_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001093 /* keeps fsck and non-quota versions of reiserfs happy */
1094 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1095 bytes += (loff_t) 511;
1096 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001098 /* files from before the quota patch might i_blocks such that
1099 ** bytes < real_space. Deal with that here to prevent it from
1100 ** going negative.
1101 */
1102 if (bytes < real_space)
1103 return 0;
1104 return (bytes - real_space) >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105}
1106
1107//
1108// BAD: new directories have stat data of new type and all other items
1109// of old type. Version stored in the inode says about body items, so
1110// in update_stat_data we can not rely on inode, but have to check
1111// item version directly
1112//
1113
1114// called by read_locked_inode
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001115static void init_inode(struct inode *inode, struct treepath *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001117 struct buffer_head *bh;
1118 struct item_head *ih;
1119 __u32 rdev;
1120 //int version = ITEM_VERSION_1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001122 bh = PATH_PLAST_BUFFER(path);
1123 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001125 copy_key(INODE_PKEY(inode), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001127 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1128 REISERFS_I(inode)->i_flags = 0;
1129 REISERFS_I(inode)->i_prealloc_block = 0;
1130 REISERFS_I(inode)->i_prealloc_count = 0;
1131 REISERFS_I(inode)->i_trans_id = 0;
1132 REISERFS_I(inode)->i_jl = NULL;
Vladimir Savelievde145692007-01-22 20:40:46 -08001133 mutex_init(&(REISERFS_I(inode)->i_mmap));
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001134 reiserfs_init_xattr_rwsem(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001136 if (stat_data_v1(ih)) {
1137 struct stat_data_v1 *sd =
1138 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1139 unsigned long blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001141 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1142 set_inode_sd_version(inode, STAT_DATA_V1);
1143 inode->i_mode = sd_v1_mode(sd);
1144 inode->i_nlink = sd_v1_nlink(sd);
1145 inode->i_uid = sd_v1_uid(sd);
1146 inode->i_gid = sd_v1_gid(sd);
1147 inode->i_size = sd_v1_size(sd);
1148 inode->i_atime.tv_sec = sd_v1_atime(sd);
1149 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1150 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1151 inode->i_atime.tv_nsec = 0;
1152 inode->i_ctime.tv_nsec = 0;
1153 inode->i_mtime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001155 inode->i_blocks = sd_v1_blocks(sd);
1156 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1157 blocks = (inode->i_size + 511) >> 9;
1158 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1159 if (inode->i_blocks > blocks) {
1160 // there was a bug in <=3.5.23 when i_blocks could take negative
1161 // values. Starting from 3.5.17 this value could even be stored in
1162 // stat data. For such files we set i_blocks based on file
1163 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1164 // only updated if file's inode will ever change
1165 inode->i_blocks = blocks;
1166 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001168 rdev = sd_v1_rdev(sd);
1169 REISERFS_I(inode)->i_first_direct_byte =
1170 sd_v1_first_direct_byte(sd);
1171 /* an early bug in the quota code can give us an odd number for the
1172 ** block count. This is incorrect, fix it here.
1173 */
1174 if (inode->i_blocks & 1) {
1175 inode->i_blocks++;
1176 }
1177 inode_set_bytes(inode,
1178 to_real_used_space(inode, inode->i_blocks,
1179 SD_V1_SIZE));
1180 /* nopack is initially zero for v1 objects. For v2 objects,
1181 nopack is initialised from sd_attrs */
1182 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1183 } else {
1184 // new stat data found, but object may have old items
1185 // (directories and symlinks)
1186 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1187
1188 inode->i_mode = sd_v2_mode(sd);
1189 inode->i_nlink = sd_v2_nlink(sd);
1190 inode->i_uid = sd_v2_uid(sd);
1191 inode->i_size = sd_v2_size(sd);
1192 inode->i_gid = sd_v2_gid(sd);
1193 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1194 inode->i_atime.tv_sec = sd_v2_atime(sd);
1195 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1196 inode->i_ctime.tv_nsec = 0;
1197 inode->i_mtime.tv_nsec = 0;
1198 inode->i_atime.tv_nsec = 0;
1199 inode->i_blocks = sd_v2_blocks(sd);
1200 rdev = sd_v2_rdev(sd);
1201 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1202 inode->i_generation =
1203 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1204 else
1205 inode->i_generation = sd_v2_generation(sd);
1206
1207 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1208 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1209 else
1210 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1211 REISERFS_I(inode)->i_first_direct_byte = 0;
1212 set_inode_sd_version(inode, STAT_DATA_V2);
1213 inode_set_bytes(inode,
1214 to_real_used_space(inode, inode->i_blocks,
1215 SD_V2_SIZE));
1216 /* read persistent inode attributes from sd and initalise
1217 generic inode flags from them */
1218 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1219 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 }
1221
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001222 pathrelse(path);
1223 if (S_ISREG(inode->i_mode)) {
1224 inode->i_op = &reiserfs_file_inode_operations;
1225 inode->i_fop = &reiserfs_file_operations;
1226 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1227 } else if (S_ISDIR(inode->i_mode)) {
1228 inode->i_op = &reiserfs_dir_inode_operations;
1229 inode->i_fop = &reiserfs_dir_operations;
1230 } else if (S_ISLNK(inode->i_mode)) {
1231 inode->i_op = &reiserfs_symlink_inode_operations;
1232 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1233 } else {
1234 inode->i_blocks = 0;
1235 inode->i_op = &reiserfs_special_inode_operations;
1236 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238}
1239
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240// update new stat data with inode fields
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001241static void inode2sd(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001243 struct stat_data *sd_v2 = (struct stat_data *)sd;
1244 __u16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001246 set_sd_v2_mode(sd_v2, inode->i_mode);
1247 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1248 set_sd_v2_uid(sd_v2, inode->i_uid);
1249 set_sd_v2_size(sd_v2, size);
1250 set_sd_v2_gid(sd_v2, inode->i_gid);
1251 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1252 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1253 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1254 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1255 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1256 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1257 else
1258 set_sd_v2_generation(sd_v2, inode->i_generation);
1259 flags = REISERFS_I(inode)->i_attrs;
1260 i_attrs_to_sd_attrs(inode, &flags);
1261 set_sd_v2_attrs(sd_v2, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262}
1263
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264// used to copy inode's fields to old stat data
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001265static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001267 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001269 set_sd_v1_mode(sd_v1, inode->i_mode);
1270 set_sd_v1_uid(sd_v1, inode->i_uid);
1271 set_sd_v1_gid(sd_v1, inode->i_gid);
1272 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1273 set_sd_v1_size(sd_v1, size);
1274 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1275 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1276 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001278 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1279 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1280 else
1281 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001283 // Sigh. i_first_direct_byte is back
1284 set_sd_v1_first_direct_byte(sd_v1,
1285 REISERFS_I(inode)->i_first_direct_byte);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286}
1287
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288/* NOTE, you must prepare the buffer head before sending it here,
1289** and then log it after the call
1290*/
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001291static void update_stat_data(struct treepath *path, struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001292 loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001294 struct buffer_head *bh;
1295 struct item_head *ih;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001297 bh = PATH_PLAST_BUFFER(path);
1298 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001300 if (!is_statdata_le_ih(ih))
Jeff Mahoneyc3a9c212009-03-30 14:02:25 -04001301 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001302 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) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001333 reiserfs_error(inode->i_sb, "vs-13050",
1334 "i/o failure occurred trying to "
1335 "update %K stat data", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001336 return;
1337 }
1338 if (retval == ITEM_NOT_FOUND) {
1339 pos = PATH_LAST_POSITION(&path);
1340 pathrelse(&path);
1341 if (inode->i_nlink == 0) {
1342 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1343 return;
1344 }
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001345 reiserfs_warning(inode->i_sb, "vs-13060",
1346 "stat data of object %k (nlink == %d) "
1347 "not found (pos %d)",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001348 INODE_PKEY(inode), inode->i_nlink,
1349 pos);
1350 reiserfs_check_path(&path);
1351 return;
1352 }
1353
1354 /* sigh, prepare_for_journal might schedule. When it schedules the
1355 ** FS might change. We have to detect that, and loop back to the
1356 ** search if the stat data item has moved
1357 */
1358 bh = get_last_bh(&path);
1359 ih = get_ih(&path);
1360 copy_item_head(&tmp_ih, ih);
1361 fs_gen = get_generation(inode->i_sb);
1362 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1363 if (fs_changed(fs_gen, inode->i_sb)
1364 && item_moved(&tmp_ih, &path)) {
1365 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1366 continue; /* Stat_data item has been moved after scheduling. */
1367 }
1368 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001370 update_stat_data(&path, inode, size);
1371 journal_mark_dirty(th, th->t_super, bh);
1372 pathrelse(&path);
1373 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374}
1375
1376/* reiserfs_read_locked_inode is called to read the inode off disk, and it
1377** does a make_bad_inode when things go wrong. But, we need to make sure
1378** and clear the key in the private portion of the inode, otherwise a
1379** corresponding iput might try to delete whatever object the inode last
1380** represented.
1381*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001382static void reiserfs_make_bad_inode(struct inode *inode)
1383{
1384 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1385 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386}
1387
1388//
1389// initially this function was derived from minix or ext2's analog and
1390// evolved as the prototype did
1391//
1392
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001393int reiserfs_init_locked_inode(struct inode *inode, void *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001395 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1396 inode->i_ino = args->objectid;
1397 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1398 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399}
1400
1401/* looks for stat data in the tree, and fills up the fields of in-core
1402 inode stat data fields */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001403void reiserfs_read_locked_inode(struct inode *inode,
1404 struct reiserfs_iget_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001406 INITIALIZE_PATH(path_to_sd);
1407 struct cpu_key key;
1408 unsigned long dirino;
1409 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001411 dirino = args->dirid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001413 /* set version 1, version 2 could be used too, because stat data
1414 key is the same in both versions */
1415 key.version = KEY_FORMAT_3_5;
1416 key.on_disk_key.k_dir_id = dirino;
1417 key.on_disk_key.k_objectid = inode->i_ino;
1418 key.on_disk_key.k_offset = 0;
1419 key.on_disk_key.k_type = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001421 /* look for the object's stat data */
1422 retval = search_item(inode->i_sb, &key, &path_to_sd);
1423 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001424 reiserfs_error(inode->i_sb, "vs-13070",
1425 "i/o failure occurred trying to find "
1426 "stat data of %K", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001427 reiserfs_make_bad_inode(inode);
1428 return;
1429 }
1430 if (retval != ITEM_FOUND) {
1431 /* a stale NFS handle can trigger this without it being an error */
1432 pathrelse(&path_to_sd);
1433 reiserfs_make_bad_inode(inode);
1434 inode->i_nlink = 0;
1435 return;
1436 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001438 init_inode(inode, &path_to_sd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001440 /* It is possible that knfsd is trying to access inode of a file
1441 that is being removed from the disk by some other thread. As we
1442 update sd on unlink all that is required is to check for nlink
1443 here. This bug was first found by Sizif when debugging
1444 SquidNG/Butterfly, forgotten, and found again after Philippe
Jeff Mahoney0222e652009-03-30 14:02:44 -04001445 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001447 More logical fix would require changes in fs/inode.c:iput() to
1448 remove inode from hash-table _after_ fs cleaned disk stuff up and
1449 in iget() to return NULL if I_FREEING inode is found in
1450 hash-table. */
1451 /* Currently there is one place where it's ok to meet inode with
1452 nlink==0: processing of open-unlinked and half-truncated files
1453 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1454 if ((inode->i_nlink == 0) &&
1455 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001456 reiserfs_warning(inode->i_sb, "vs-13075",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001457 "dead inode read from disk %K. "
1458 "This is likely to be race with knfsd. Ignore",
1459 &key);
1460 reiserfs_make_bad_inode(inode);
1461 }
1462
1463 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
1465}
1466
1467/**
1468 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1469 *
1470 * @inode: inode from hash table to check
1471 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1472 *
1473 * This function is called by iget5_locked() to distinguish reiserfs inodes
1474 * having the same inode numbers. Such inodes can only exist due to some
1475 * error condition. One of them should be bad. Inodes with identical
1476 * inode numbers (objectids) are distinguished by parent directory ids.
1477 *
1478 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001479int reiserfs_find_actor(struct inode *inode, void *opaque)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001481 struct reiserfs_iget_args *args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001483 args = opaque;
1484 /* args is already in CPU order */
1485 return (inode->i_ino == args->objectid) &&
1486 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
1488
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001489struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001491 struct inode *inode;
1492 struct reiserfs_iget_args args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001494 args.objectid = key->on_disk_key.k_objectid;
1495 args.dirid = key->on_disk_key.k_dir_id;
1496 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1497 reiserfs_find_actor, reiserfs_init_locked_inode,
1498 (void *)(&args));
1499 if (!inode)
1500 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001502 if (inode->i_state & I_NEW) {
1503 reiserfs_read_locked_inode(inode, &args);
1504 unlock_new_inode(inode);
1505 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001507 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1508 /* either due to i/o error or a stale NFS handle */
1509 iput(inode);
1510 inode = NULL;
1511 }
1512 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513}
1514
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001515static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1516 u32 objectid, u32 dir_id, u32 generation)
1517
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001519 struct cpu_key key;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001520 struct inode *inode;
1521
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001522 key.on_disk_key.k_objectid = objectid;
1523 key.on_disk_key.k_dir_id = dir_id;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001524 reiserfs_write_lock(sb);
1525 inode = reiserfs_iget(sb, &key);
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001526 if (inode && !IS_ERR(inode) && generation != 0 &&
1527 generation != inode->i_generation) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001528 iput(inode);
1529 inode = NULL;
1530 }
1531 reiserfs_write_unlock(sb);
Christoph Hellwig44003722008-08-11 15:49:04 +02001532
1533 return d_obtain_alias(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534}
1535
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001536struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1537 int fh_len, int fh_type)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001538{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001539 /* fhtype happens to reflect the number of u32s encoded.
1540 * due to a bug in earlier code, fhtype might indicate there
1541 * are more u32s then actually fitted.
1542 * so if fhtype seems to be more than len, reduce fhtype.
1543 * Valid types are:
1544 * 2 - objectid + dir_id - legacy support
1545 * 3 - objectid + dir_id + generation
1546 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1547 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1548 * 6 - as above plus generation of directory
1549 * 6 does not fit in NFSv2 handles
1550 */
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001551 if (fh_type > fh_len) {
1552 if (fh_type != 6 || fh_len != 5)
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001553 reiserfs_warning(sb, "reiserfs-13077",
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001554 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1555 fh_type, fh_len);
1556 fh_type = 5;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001557 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001559 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1560 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1561}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001563struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1564 int fh_len, int fh_type)
1565{
1566 if (fh_type < 4)
1567 return NULL;
1568
1569 return reiserfs_get_dentry(sb,
1570 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1571 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1572 (fh_type == 6) ? fid->raw[5] : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573}
1574
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001575int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1576 int need_parent)
1577{
1578 struct inode *inode = dentry->d_inode;
1579 int maxlen = *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001581 if (maxlen < 3)
1582 return 255;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001584 data[0] = inode->i_ino;
1585 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1586 data[2] = inode->i_generation;
1587 *lenp = 3;
1588 /* no room for directory info? return what we've stored so far */
1589 if (maxlen < 5 || !need_parent)
1590 return 3;
1591
1592 spin_lock(&dentry->d_lock);
1593 inode = dentry->d_parent->d_inode;
1594 data[3] = inode->i_ino;
1595 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1596 *lenp = 5;
1597 if (maxlen >= 6) {
1598 data[5] = inode->i_generation;
1599 *lenp = 6;
1600 }
1601 spin_unlock(&dentry->d_lock);
1602 return *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603}
1604
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605/* looks for stat data, then copies fields to it, marks the buffer
1606 containing stat data as dirty */
1607/* reiserfs inodes are never really dirty, since the dirty inode call
1608** always logs them. This call allows the VFS inode marking routines
1609** to properly mark inodes for datasync and such, but only actually
1610** does something when called for a synchronous update.
1611*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001612int reiserfs_write_inode(struct inode *inode, int do_sync)
1613{
1614 struct reiserfs_transaction_handle th;
1615 int jbegin_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001617 if (inode->i_sb->s_flags & MS_RDONLY)
1618 return -EROFS;
1619 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001620 ** these cases are just when the system needs ram, not when the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001621 ** inode needs to reach disk for safety, and they can safely be
1622 ** ignored because the altered inode has already been logged.
1623 */
1624 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1625 reiserfs_write_lock(inode->i_sb);
1626 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1627 reiserfs_update_sd(&th, inode);
1628 journal_end_sync(&th, inode->i_sb, jbegin_count);
1629 }
1630 reiserfs_write_unlock(inode->i_sb);
1631 }
1632 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633}
1634
1635/* stat data of new object is inserted already, this inserts the item
1636 containing "." and ".." entries */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001637static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1638 struct inode *inode,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001639 struct item_head *ih, struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001640 struct inode *dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001642 struct super_block *sb = th->t_super;
1643 char empty_dir[EMPTY_DIR_SIZE];
1644 char *body = empty_dir;
1645 struct cpu_key key;
1646 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001648 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001650 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1651 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1652 TYPE_DIRENTRY, 3 /*key length */ );
1653
1654 /* compose item head for new item. Directories consist of items of
1655 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1656 is done by reiserfs_new_inode */
1657 if (old_format_only(sb)) {
1658 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1659 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1660
1661 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1662 ih->ih_key.k_objectid,
1663 INODE_PKEY(dir)->k_dir_id,
1664 INODE_PKEY(dir)->k_objectid);
1665 } else {
1666 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1667 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1668
1669 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1670 ih->ih_key.k_objectid,
1671 INODE_PKEY(dir)->k_dir_id,
1672 INODE_PKEY(dir)->k_objectid);
1673 }
1674
1675 /* look for place in the tree for new item */
1676 retval = search_item(sb, &key, path);
1677 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001678 reiserfs_error(sb, "vs-13080",
1679 "i/o failure occurred creating new directory");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001680 return -EIO;
1681 }
1682 if (retval == ITEM_FOUND) {
1683 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001684 reiserfs_warning(sb, "vs-13070",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001685 "object with this key exists (%k)",
1686 &(ih->ih_key));
1687 return -EEXIST;
1688 }
1689
1690 /* insert item, that is empty directory item */
1691 return reiserfs_insert_item(th, path, &key, ih, inode, body);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692}
1693
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694/* stat data of object has been inserted, this inserts the item
1695 containing the body of symlink */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001696static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1697 struct item_head *ih,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001698 struct treepath *path, const char *symname,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001699 int item_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001701 struct super_block *sb = th->t_super;
1702 struct cpu_key key;
1703 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001705 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001707 _make_cpu_key(&key, KEY_FORMAT_3_5,
1708 le32_to_cpu(ih->ih_key.k_dir_id),
1709 le32_to_cpu(ih->ih_key.k_objectid),
1710 1, TYPE_DIRECT, 3 /*key length */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001712 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1713 0 /*free_space */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001715 /* look for place in the tree for new item */
1716 retval = search_item(sb, &key, path);
1717 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001718 reiserfs_error(sb, "vs-13080",
1719 "i/o failure occurred creating new symlink");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001720 return -EIO;
1721 }
1722 if (retval == ITEM_FOUND) {
1723 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001724 reiserfs_warning(sb, "vs-13080",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001725 "object with this key exists (%k)",
1726 &(ih->ih_key));
1727 return -EEXIST;
1728 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001730 /* insert item, that is body of symlink */
1731 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732}
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734/* inserts the stat data into the tree, and then calls
1735 reiserfs_new_directory (to insert ".", ".." item if new object is
1736 directory) or reiserfs_new_symlink (to insert symlink body if new
Jeff Mahoney0222e652009-03-30 14:02:44 -04001737 object is symlink) or nothing (if new object is regular file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
1739 NOTE! uid and gid must already be set in the inode. If we return
1740 non-zero due to an error, we have to drop the quota previously allocated
1741 for the fresh inode. This can only be done outside a transaction, so
1742 if we return non-zero, we also end the transaction. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001743int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1744 struct inode *dir, int mode, const char *symname,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001745 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001746 strlen (symname) for symlinks) */
1747 loff_t i_size, struct dentry *dentry,
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001748 struct inode *inode,
1749 struct reiserfs_security_handle *security)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001751 struct super_block *sb;
Al Viroc1eaa262008-12-30 02:03:58 -05001752 struct reiserfs_iget_args args;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001753 INITIALIZE_PATH(path_to_key);
1754 struct cpu_key key;
1755 struct item_head ih;
1756 struct stat_data sd;
1757 int retval;
1758 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001760 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
Jan Kara77db4f22009-01-26 17:14:18 +01001762 if (vfs_dq_alloc_inode(inode)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001763 err = -EDQUOT;
1764 goto out_end_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 }
Eric Sesterhenn585b7742006-10-04 02:15:30 -07001766 if (!dir->i_nlink) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001767 err = -EPERM;
1768 goto out_bad_inode;
1769 }
1770
1771 sb = dir->i_sb;
1772
1773 /* item head of new item */
1774 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1775 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1776 if (!ih.ih_key.k_objectid) {
1777 err = -ENOMEM;
1778 goto out_bad_inode;
1779 }
Al Viroc1eaa262008-12-30 02:03:58 -05001780 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
Al Viro2f1169e2009-01-02 08:16:51 -05001781 if (old_format_only(sb))
1782 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1783 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1784 else
1785 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1786 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
Al Viroc1eaa262008-12-30 02:03:58 -05001787 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1788 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1789 if (insert_inode_locked4(inode, args.objectid,
1790 reiserfs_find_actor, &args) < 0) {
1791 err = -EINVAL;
1792 goto out_bad_inode;
1793 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001794 if (old_format_only(sb))
Jeff Mahoney0222e652009-03-30 14:02:44 -04001795 /* not a perfect generation count, as object ids can be reused, but
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001796 ** this is as good as reiserfs can do right now.
1797 ** note that the private part of inode isn't filled in yet, we have
1798 ** to use the directory.
1799 */
1800 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1801 else
1802#if defined( USE_INODE_GENERATION_COUNTER )
1803 inode->i_generation =
1804 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1805#else
1806 inode->i_generation = ++event;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001809 /* fill stat data */
1810 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1811
1812 /* uid and gid must already be set by the caller for quota init */
1813
1814 /* symlink cannot be immutable or append only, right? */
1815 if (S_ISLNK(inode->i_mode))
1816 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1817
1818 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1819 inode->i_size = i_size;
1820 inode->i_blocks = 0;
1821 inode->i_bytes = 0;
1822 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1823 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1824
1825 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1826 REISERFS_I(inode)->i_flags = 0;
1827 REISERFS_I(inode)->i_prealloc_block = 0;
1828 REISERFS_I(inode)->i_prealloc_count = 0;
1829 REISERFS_I(inode)->i_trans_id = 0;
1830 REISERFS_I(inode)->i_jl = NULL;
1831 REISERFS_I(inode)->i_attrs =
1832 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1833 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
Vladimir Savelievde145692007-01-22 20:40:46 -08001834 mutex_init(&(REISERFS_I(inode)->i_mmap));
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001835 reiserfs_init_xattr_rwsem(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001836
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001837 /* key to search for correct place for new stat data */
1838 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1839 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1840 TYPE_STAT_DATA, 3 /*key length */ );
1841
1842 /* find proper place for inserting of stat data */
1843 retval = search_item(sb, &key, &path_to_key);
1844 if (retval == IO_ERROR) {
1845 err = -EIO;
1846 goto out_bad_inode;
1847 }
1848 if (retval == ITEM_FOUND) {
1849 pathrelse(&path_to_key);
1850 err = -EEXIST;
1851 goto out_bad_inode;
1852 }
1853 if (old_format_only(sb)) {
1854 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1855 pathrelse(&path_to_key);
1856 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1857 err = -EINVAL;
1858 goto out_bad_inode;
1859 }
1860 inode2sd_v1(&sd, inode, inode->i_size);
1861 } else {
1862 inode2sd(&sd, inode, inode->i_size);
1863 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001864 // store in in-core inode the key of stat data and version all
1865 // object items will have (directory items will have old offset
1866 // format, other new objects will consist of new items)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001867 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1868 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1869 else
1870 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1871 if (old_format_only(sb))
1872 set_inode_sd_version(inode, STAT_DATA_V1);
1873 else
1874 set_inode_sd_version(inode, STAT_DATA_V2);
1875
1876 /* insert the stat data into the tree */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001878 if (REISERFS_I(dir)->new_packing_locality)
1879 th->displace_new_blocks = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001881 retval =
1882 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1883 (char *)(&sd));
1884 if (retval) {
1885 err = retval;
1886 reiserfs_check_path(&path_to_key);
1887 goto out_bad_inode;
1888 }
1889#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1890 if (!th->displace_new_blocks)
1891 REISERFS_I(dir)->new_packing_locality = 0;
1892#endif
1893 if (S_ISDIR(mode)) {
1894 /* insert item with "." and ".." */
1895 retval =
1896 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1897 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001899 if (S_ISLNK(mode)) {
1900 /* insert body of symlink */
1901 if (!old_format_only(sb))
1902 i_size = ROUND_UP(i_size);
1903 retval =
1904 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1905 i_size);
1906 }
1907 if (retval) {
1908 err = retval;
1909 reiserfs_check_path(&path_to_key);
1910 journal_end(th, th->t_super, th->t_blocks_allocated);
1911 goto out_inserted_sd;
1912 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001914 if (reiserfs_posixacl(inode->i_sb)) {
Jeff Mahoney0ab26212009-03-30 14:02:39 -04001915 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001916 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 }
1922 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001923 reiserfs_warning(inode->i_sb, "jdm-13090",
1924 "ACLs aren't enabled in the fs, "
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001925 "but vfs thinks they are!");
Jeff Mahoney6dfede692009-03-30 14:02:32 -04001926 } else if (IS_PRIVATE(dir))
1927 inode->i_flags |= S_PRIVATE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001929 if (security->name) {
1930 retval = reiserfs_security_write(th, inode, security);
1931 if (retval) {
1932 err = retval;
1933 reiserfs_check_path(&path_to_key);
1934 retval = journal_end(th, th->t_super,
1935 th->t_blocks_allocated);
1936 if (retval)
1937 err = retval;
1938 goto out_inserted_sd;
1939 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 }
1941
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001942 reiserfs_update_sd(th, inode);
1943 reiserfs_check_path(&path_to_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001945 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946
1947/* it looks like you can easily compress these two goto targets into
1948 * one. Keeping it like this doesn't actually hurt anything, and they
1949 * are place holders for what the quota code actually needs.
1950 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001951 out_bad_inode:
1952 /* Invalidate the object, nothing was inserted yet */
1953 INODE_PKEY(inode)->k_objectid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001955 /* Quota change must be inside a transaction for journaling */
Jan Kara77db4f22009-01-26 17:14:18 +01001956 vfs_dq_free_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001958 out_end_trans:
1959 journal_end(th, th->t_super, th->t_blocks_allocated);
1960 /* Drop can be outside and it needs more credits so it's better to have it outside */
Jan Kara77db4f22009-01-26 17:14:18 +01001961 vfs_dq_drop(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001962 inode->i_flags |= S_NOQUOTA;
1963 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001965 out_inserted_sd:
1966 inode->i_nlink = 0;
1967 th->t_trans_id = 0; /* so the caller can't use this handle later */
Al Viroc1eaa262008-12-30 02:03:58 -05001968 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
Jeff Mahoneyd9845612009-03-30 14:02:35 -04001969 iput(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001970 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971}
1972
1973/*
1974** finds the tail page in the page cache,
1975** reads the last block in.
1976**
1977** On success, page_result is set to a locked, pinned page, and bh_result
1978** is set to an up to date buffer for the last block in the file. returns 0.
1979**
1980** tail conversion is not done, so bh_result might not be valid for writing
1981** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1982** trying to write the block.
1983**
1984** on failure, nonzero is returned, page_result and bh_result are untouched.
1985*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04001986static int grab_tail_page(struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001987 struct page **page_result,
1988 struct buffer_head **bh_result)
1989{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001991 /* we want the page with the last byte in the file,
1992 ** not the page that will hold the next byte for appending
1993 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04001994 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001995 unsigned long pos = 0;
1996 unsigned long start = 0;
Jeff Mahoney995c7622009-03-30 14:02:47 -04001997 unsigned long blocksize = inode->i_sb->s_blocksize;
1998 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001999 struct buffer_head *bh;
2000 struct buffer_head *head;
2001 struct page *page;
2002 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002004 /* we know that we are only called with inode->i_size > 0.
2005 ** we also know that a file tail can never be as big as a block
2006 ** If i_size % blocksize == 0, our file is currently block aligned
2007 ** and it won't need converting or zeroing after a truncate.
2008 */
2009 if ((offset & (blocksize - 1)) == 0) {
2010 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 }
Jeff Mahoney995c7622009-03-30 14:02:47 -04002012 page = grab_cache_page(inode->i_mapping, index);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002013 error = -ENOMEM;
2014 if (!page) {
2015 goto out;
2016 }
2017 /* start within the page of the last block in the file */
2018 start = (offset / blocksize) * blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002020 error = block_prepare_write(page, start, offset,
2021 reiserfs_get_block_create_0);
2022 if (error)
2023 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002025 head = page_buffers(page);
2026 bh = head;
2027 do {
2028 if (pos >= start) {
2029 break;
2030 }
2031 bh = bh->b_this_page;
2032 pos += blocksize;
2033 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002035 if (!buffer_uptodate(bh)) {
2036 /* note, this should never happen, prepare_write should
2037 ** be taking care of this for us. If the buffer isn't up to date,
2038 ** I've screwed up the code to find the buffer, or the code to
2039 ** call prepare_write
2040 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002041 reiserfs_error(inode->i_sb, "clm-6000",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002042 "error reading block %lu", bh->b_blocknr);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002043 error = -EIO;
2044 goto unlock;
2045 }
2046 *bh_result = bh;
2047 *page_result = page;
2048
2049 out:
2050 return error;
2051
2052 unlock:
2053 unlock_page(page);
2054 page_cache_release(page);
2055 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056}
2057
2058/*
2059** vfs version of truncate file. Must NOT be called with
2060** a transaction already started.
2061**
2062** some code taken from block_truncate_page
2063*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04002064int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002065{
2066 struct reiserfs_transaction_handle th;
2067 /* we want the offset for the first byte after the end of the file */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002068 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2069 unsigned blocksize = inode->i_sb->s_blocksize;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002070 unsigned length;
2071 struct page *page = NULL;
2072 int error;
2073 struct buffer_head *bh = NULL;
Jeff Mahoney24996042005-12-14 14:38:05 -05002074 int err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
Jeff Mahoney995c7622009-03-30 14:02:47 -04002076 reiserfs_write_lock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077
Jeff Mahoney995c7622009-03-30 14:02:47 -04002078 if (inode->i_size > 0) {
2079 error = grab_tail_page(inode, &page, &bh);
2080 if (error) {
Jeff Mahoney0222e652009-03-30 14:02:44 -04002081 // -ENOENT means we truncated past the end of the file,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002082 // and get_block_create_0 could not find a block to read in,
2083 // which is ok.
2084 if (error != -ENOENT)
Jeff Mahoney995c7622009-03-30 14:02:47 -04002085 reiserfs_error(inode->i_sb, "clm-6001",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002086 "grab_tail_page failed %d",
2087 error);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002088 page = NULL;
2089 bh = NULL;
2090 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
Jeff Mahoney0222e652009-03-30 14:02:44 -04002093 /* so, if page != NULL, we have a buffer head for the offset at
2094 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2095 ** then we have an unformatted node. Otherwise, we have a direct item,
2096 ** and no zeroing is required on disk. We zero after the truncate,
2097 ** because the truncate might pack the item anyway
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002098 ** (it will unmap bh if it packs).
2099 */
2100 /* it is enough to reserve space in transaction for 2 balancings:
2101 one for "save" link adding and another for the first
2102 cut_from_item. 1 is for update_sd */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002103 error = journal_begin(&th, inode->i_sb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002104 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2105 if (error)
2106 goto out;
Jeff Mahoney995c7622009-03-30 14:02:47 -04002107 reiserfs_update_inode_transaction(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002108 if (update_timestamps)
2109 /* we are doing real truncate: if the system crashes before the last
2110 transaction of truncating gets committed - on reboot the file
2111 either appears truncated properly or not truncated at all */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002112 add_save_link(&th, inode, 1);
2113 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002114 error =
Jeff Mahoney995c7622009-03-30 14:02:47 -04002115 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002116 if (error)
2117 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118
Jeff Mahoney24996042005-12-14 14:38:05 -05002119 /* check reiserfs_do_truncate after ending the transaction */
2120 if (err2) {
2121 error = err2;
2122 goto out;
2123 }
2124
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002125 if (update_timestamps) {
Jeff Mahoney995c7622009-03-30 14:02:47 -04002126 error = remove_save_link(inode, 1 /* truncate */);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002127 if (error)
2128 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002131 if (page) {
2132 length = offset & (blocksize - 1);
2133 /* if we are not on a block boundary */
2134 if (length) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002135 length = blocksize - length;
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002136 zero_user(page, offset, length);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002137 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2138 mark_buffer_dirty(bh);
2139 }
2140 }
2141 unlock_page(page);
2142 page_cache_release(page);
2143 }
2144
Jeff Mahoney995c7622009-03-30 14:02:47 -04002145 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002146 return 0;
2147 out:
2148 if (page) {
2149 unlock_page(page);
2150 page_cache_release(page);
2151 }
Jeff Mahoney995c7622009-03-30 14:02:47 -04002152 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002153 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154}
2155
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002156static int map_block_for_writepage(struct inode *inode,
2157 struct buffer_head *bh_result,
2158 unsigned long block)
2159{
2160 struct reiserfs_transaction_handle th;
2161 int fs_gen;
2162 struct item_head tmp_ih;
2163 struct item_head *ih;
2164 struct buffer_head *bh;
2165 __le32 *item;
2166 struct cpu_key key;
2167 INITIALIZE_PATH(path);
2168 int pos_in_item;
2169 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
Oleg Drokin7729ac52005-11-28 13:43:53 -08002170 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002171 int retval;
2172 int use_get_block = 0;
2173 int bytes_copied = 0;
2174 int copy_size;
2175 int trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002177 /* catch places below that try to log something without starting a trans */
2178 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002180 if (!buffer_uptodate(bh_result)) {
2181 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 }
2183
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002184 kmap(bh_result->b_page);
2185 start_over:
2186 reiserfs_write_lock(inode->i_sb);
2187 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002189 research:
2190 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2191 if (retval != POSITION_FOUND) {
2192 use_get_block = 1;
2193 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 }
2195
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002196 bh = get_last_bh(&path);
2197 ih = get_ih(&path);
2198 item = get_item(&path);
2199 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002201 /* we've found an unformatted node */
2202 if (indirect_item_found(retval, ih)) {
2203 if (bytes_copied > 0) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002204 reiserfs_warning(inode->i_sb, "clm-6002",
2205 "bytes_copied %d", bytes_copied);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002206 }
2207 if (!get_block_num(item, pos_in_item)) {
2208 /* crap, we are writing to a hole */
2209 use_get_block = 1;
2210 goto out;
2211 }
2212 set_block_dev_mapped(bh_result,
2213 get_block_num(item, pos_in_item), inode);
2214 } else if (is_direct_le_ih(ih)) {
2215 char *p;
2216 p = page_address(bh_result->b_page);
2217 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2218 copy_size = ih_item_len(ih) - pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002220 fs_gen = get_generation(inode->i_sb);
2221 copy_item_head(&tmp_ih, ih);
2222
2223 if (!trans_running) {
2224 /* vs-3050 is gone, no need to drop the path */
2225 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2226 if (retval)
2227 goto out;
2228 reiserfs_update_inode_transaction(inode);
2229 trans_running = 1;
2230 if (fs_changed(fs_gen, inode->i_sb)
2231 && item_moved(&tmp_ih, &path)) {
2232 reiserfs_restore_prepared_buffer(inode->i_sb,
2233 bh);
2234 goto research;
2235 }
2236 }
2237
2238 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2239
2240 if (fs_changed(fs_gen, inode->i_sb)
2241 && item_moved(&tmp_ih, &path)) {
2242 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2243 goto research;
2244 }
2245
2246 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2247 copy_size);
2248
2249 journal_mark_dirty(&th, inode->i_sb, bh);
2250 bytes_copied += copy_size;
2251 set_block_dev_mapped(bh_result, 0, inode);
2252
2253 /* are there still bytes left? */
2254 if (bytes_copied < bh_result->b_size &&
2255 (byte_offset + bytes_copied) < inode->i_size) {
2256 set_cpu_key_k_offset(&key,
2257 cpu_key_k_offset(&key) +
2258 copy_size);
2259 goto research;
2260 }
2261 } else {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002262 reiserfs_warning(inode->i_sb, "clm-6003",
2263 "bad item inode %lu", inode->i_ino);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002264 retval = -EIO;
2265 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002267 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002269 out:
2270 pathrelse(&path);
2271 if (trans_running) {
2272 int err = journal_end(&th, inode->i_sb, jbegin_count);
2273 if (err)
2274 retval = err;
2275 trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002277 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002279 /* this is where we fill in holes in the file. */
2280 if (use_get_block) {
2281 retval = reiserfs_get_block(inode, block, bh_result,
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002282 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002283 | GET_BLOCK_NO_DANGLE);
2284 if (!retval) {
2285 if (!buffer_mapped(bh_result)
2286 || bh_result->b_blocknr == 0) {
2287 /* get_block failed to find a mapped unformatted node. */
2288 use_get_block = 0;
2289 goto start_over;
2290 }
2291 }
2292 }
2293 kunmap(bh_result->b_page);
2294
2295 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2296 /* we've copied data from the page into the direct item, so the
2297 * buffer in the page is now clean, mark it to reflect that.
2298 */
2299 lock_buffer(bh_result);
2300 clear_buffer_dirty(bh_result);
2301 unlock_buffer(bh_result);
2302 }
2303 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304}
2305
Jeff Mahoney0222e652009-03-30 14:02:44 -04002306/*
2307 * mason@suse.com: updated in 2.5.54 to follow the same general io
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308 * start/recovery path as __block_write_full_page, along with special
2309 * code to handle reiserfs tails.
2310 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002311static int reiserfs_write_full_page(struct page *page,
2312 struct writeback_control *wbc)
2313{
2314 struct inode *inode = page->mapping->host;
2315 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2316 int error = 0;
2317 unsigned long block;
Chris Masonb4c76fa2006-08-05 12:15:10 -07002318 sector_t last_block;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002319 struct buffer_head *head, *bh;
2320 int partial = 0;
2321 int nr = 0;
2322 int checked = PageChecked(page);
2323 struct reiserfs_transaction_handle th;
2324 struct super_block *s = inode->i_sb;
2325 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2326 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327
Chris Masone0e851c2006-02-01 03:06:49 -08002328 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2329 if (checked && (current->flags & PF_MEMALLOC)) {
2330 redirty_page_for_writepage(wbc, page);
2331 unlock_page(page);
2332 return 0;
2333 }
2334
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002335 /* The page dirty bit is cleared before writepage is called, which
2336 * means we have to tell create_empty_buffers to make dirty buffers
2337 * The page really should be up to date at this point, so tossing
2338 * in the BH_Uptodate is just a sanity check.
2339 */
2340 if (!page_has_buffers(page)) {
2341 create_empty_buffers(page, s->s_blocksize,
2342 (1 << BH_Dirty) | (1 << BH_Uptodate));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002344 head = page_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002346 /* last page in the file, zero out any contents past the
2347 ** last byte in the file
2348 */
2349 if (page->index >= end_index) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002350 unsigned last_offset;
2351
2352 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2353 /* no file contents in this page */
2354 if (page->index >= end_index + 1 || !last_offset) {
2355 unlock_page(page);
2356 return 0;
2357 }
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002358 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002360 bh = head;
2361 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
Chris Masonb4c76fa2006-08-05 12:15:10 -07002362 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002363 /* first map all the buffers, logging any direct items we find */
2364 do {
Chris Masonb4c76fa2006-08-05 12:15:10 -07002365 if (block > last_block) {
2366 /*
2367 * This can happen when the block size is less than
2368 * the page size. The corresponding bytes in the page
2369 * were zero filled above
2370 */
2371 clear_buffer_dirty(bh);
2372 set_buffer_uptodate(bh);
2373 } else if ((checked || buffer_dirty(bh)) &&
2374 (!buffer_mapped(bh) || (buffer_mapped(bh)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002375 && bh->b_blocknr ==
2376 0))) {
2377 /* not mapped yet, or it points to a direct item, search
2378 * the btree for the mapping info, and log any direct
2379 * items found
2380 */
2381 if ((error = map_block_for_writepage(inode, bh, block))) {
2382 goto fail;
2383 }
2384 }
2385 bh = bh->b_this_page;
2386 block++;
2387 } while (bh != head);
2388
2389 /*
2390 * we start the transaction after map_block_for_writepage,
2391 * because it can create holes in the file (an unbounded operation).
2392 * starting it here, we can make a reliable estimate for how many
2393 * blocks we're going to log
2394 */
2395 if (checked) {
2396 ClearPageChecked(page);
2397 reiserfs_write_lock(s);
2398 error = journal_begin(&th, s, bh_per_page + 1);
2399 if (error) {
2400 reiserfs_write_unlock(s);
2401 goto fail;
2402 }
2403 reiserfs_update_inode_transaction(inode);
2404 }
2405 /* now go through and lock any dirty buffers on the page */
2406 do {
2407 get_bh(bh);
2408 if (!buffer_mapped(bh))
2409 continue;
2410 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2411 continue;
2412
2413 if (checked) {
2414 reiserfs_prepare_for_journal(s, bh, 1);
2415 journal_mark_dirty(&th, s, bh);
2416 continue;
2417 }
2418 /* from this point on, we know the buffer is mapped to a
2419 * real block and not a direct item
2420 */
2421 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2422 lock_buffer(bh);
2423 } else {
Nick Pigginca5de402008-08-02 12:02:13 +02002424 if (!trylock_buffer(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002425 redirty_page_for_writepage(wbc, page);
2426 continue;
2427 }
2428 }
2429 if (test_clear_buffer_dirty(bh)) {
2430 mark_buffer_async_write(bh);
2431 } else {
2432 unlock_buffer(bh);
2433 }
2434 } while ((bh = bh->b_this_page) != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435
2436 if (checked) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002437 error = journal_end(&th, s, bh_per_page + 1);
2438 reiserfs_write_unlock(s);
2439 if (error)
2440 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002442 BUG_ON(PageWriteback(page));
2443 set_page_writeback(page);
2444 unlock_page(page);
2445
2446 /*
Jeff Mahoney0222e652009-03-30 14:02:44 -04002447 * since any buffer might be the only dirty buffer on the page,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002448 * the first submit_bh can bring the page out of writeback.
2449 * be careful with the buffers.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 do {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002452 struct buffer_head *next = bh->b_this_page;
2453 if (buffer_async_write(bh)) {
2454 submit_bh(WRITE, bh);
2455 nr++;
2456 }
2457 put_bh(bh);
2458 bh = next;
2459 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002461 error = 0;
2462 done:
2463 if (nr == 0) {
2464 /*
2465 * if this page only had a direct item, it is very possible for
Jeff Mahoney0222e652009-03-30 14:02:44 -04002466 * no io to be required without there being an error. Or,
2467 * someone else could have locked them and sent them down the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002468 * pipe without locking the page
2469 */
2470 bh = head;
2471 do {
2472 if (!buffer_uptodate(bh)) {
2473 partial = 1;
2474 break;
2475 }
2476 bh = bh->b_this_page;
2477 } while (bh != head);
2478 if (!partial)
2479 SetPageUptodate(page);
2480 end_page_writeback(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002482 return error;
2483
2484 fail:
2485 /* catches various errors, we need to make sure any valid dirty blocks
Jeff Mahoney0222e652009-03-30 14:02:44 -04002486 * get to the media. The page is currently locked and not marked for
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002487 * writeback
2488 */
2489 ClearPageUptodate(page);
2490 bh = head;
2491 do {
2492 get_bh(bh);
2493 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2494 lock_buffer(bh);
2495 mark_buffer_async_write(bh);
2496 } else {
2497 /*
2498 * clear any dirty bits that might have come from getting
2499 * attached to a dirty page
2500 */
2501 clear_buffer_dirty(bh);
2502 }
2503 bh = bh->b_this_page;
2504 } while (bh != head);
2505 SetPageError(page);
2506 BUG_ON(PageWriteback(page));
2507 set_page_writeback(page);
2508 unlock_page(page);
2509 do {
2510 struct buffer_head *next = bh->b_this_page;
2511 if (buffer_async_write(bh)) {
2512 clear_buffer_dirty(bh);
2513 submit_bh(WRITE, bh);
2514 nr++;
2515 }
2516 put_bh(bh);
2517 bh = next;
2518 } while (bh != head);
2519 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002520}
2521
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002522static int reiserfs_readpage(struct file *f, struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002524 return block_read_full_page(page, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525}
2526
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002527static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002529 struct inode *inode = page->mapping->host;
2530 reiserfs_wait_on_write_block(inode->i_sb);
2531 return reiserfs_write_full_page(page, wbc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002532}
2533
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002534static int reiserfs_write_begin(struct file *file,
2535 struct address_space *mapping,
2536 loff_t pos, unsigned len, unsigned flags,
2537 struct page **pagep, void **fsdata)
2538{
2539 struct inode *inode;
2540 struct page *page;
2541 pgoff_t index;
2542 int ret;
2543 int old_ref = 0;
2544
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002545 inode = mapping->host;
2546 *fsdata = 0;
2547 if (flags & AOP_FLAG_CONT_EXPAND &&
2548 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2549 pos ++;
2550 *fsdata = (void *)(unsigned long)flags;
2551 }
2552
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002553 index = pos >> PAGE_CACHE_SHIFT;
Nick Piggin54566b22009-01-04 12:00:53 -08002554 page = grab_cache_page_write_begin(mapping, index, flags);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002555 if (!page)
2556 return -ENOMEM;
2557 *pagep = page;
2558
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002559 reiserfs_wait_on_write_block(inode->i_sb);
2560 fix_tail_page_for_writing(page);
2561 if (reiserfs_transaction_running(inode->i_sb)) {
2562 struct reiserfs_transaction_handle *th;
2563 th = (struct reiserfs_transaction_handle *)current->
2564 journal_info;
2565 BUG_ON(!th->t_refcount);
2566 BUG_ON(!th->t_trans_id);
2567 old_ref = th->t_refcount;
2568 th->t_refcount++;
2569 }
2570 ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata,
2571 reiserfs_get_block);
2572 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2573 struct reiserfs_transaction_handle *th = current->journal_info;
2574 /* this gets a little ugly. If reiserfs_get_block returned an
2575 * error and left a transacstion running, we've got to close it,
2576 * and we've got to free handle if it was a persistent transaction.
2577 *
2578 * But, if we had nested into an existing transaction, we need
2579 * to just drop the ref count on the handle.
2580 *
2581 * If old_ref == 0, the transaction is from reiserfs_get_block,
2582 * and it was a persistent trans. Otherwise, it was nested above.
2583 */
2584 if (th->t_refcount > old_ref) {
2585 if (old_ref)
2586 th->t_refcount--;
2587 else {
2588 int err;
2589 reiserfs_write_lock(inode->i_sb);
2590 err = reiserfs_end_persistent_transaction(th);
2591 reiserfs_write_unlock(inode->i_sb);
2592 if (err)
2593 ret = err;
2594 }
2595 }
2596 }
2597 if (ret) {
2598 unlock_page(page);
2599 page_cache_release(page);
2600 }
2601 return ret;
2602}
2603
2604int reiserfs_prepare_write(struct file *f, struct page *page,
2605 unsigned from, unsigned to)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002606{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002607 struct inode *inode = page->mapping->host;
2608 int ret;
2609 int old_ref = 0;
2610
2611 reiserfs_wait_on_write_block(inode->i_sb);
2612 fix_tail_page_for_writing(page);
2613 if (reiserfs_transaction_running(inode->i_sb)) {
2614 struct reiserfs_transaction_handle *th;
2615 th = (struct reiserfs_transaction_handle *)current->
2616 journal_info;
2617 BUG_ON(!th->t_refcount);
2618 BUG_ON(!th->t_trans_id);
2619 old_ref = th->t_refcount;
2620 th->t_refcount++;
2621 }
2622
2623 ret = block_prepare_write(page, from, to, reiserfs_get_block);
2624 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2625 struct reiserfs_transaction_handle *th = current->journal_info;
2626 /* this gets a little ugly. If reiserfs_get_block returned an
2627 * error and left a transacstion running, we've got to close it,
2628 * and we've got to free handle if it was a persistent transaction.
2629 *
2630 * But, if we had nested into an existing transaction, we need
2631 * to just drop the ref count on the handle.
2632 *
2633 * If old_ref == 0, the transaction is from reiserfs_get_block,
2634 * and it was a persistent trans. Otherwise, it was nested above.
2635 */
2636 if (th->t_refcount > old_ref) {
2637 if (old_ref)
2638 th->t_refcount--;
2639 else {
2640 int err;
2641 reiserfs_write_lock(inode->i_sb);
2642 err = reiserfs_end_persistent_transaction(th);
2643 reiserfs_write_unlock(inode->i_sb);
2644 if (err)
2645 ret = err;
2646 }
2647 }
2648 }
2649 return ret;
2650
2651}
2652
2653static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2654{
2655 return generic_block_bmap(as, block, reiserfs_bmap);
2656}
2657
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002658static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2659 loff_t pos, unsigned len, unsigned copied,
2660 struct page *page, void *fsdata)
2661{
2662 struct inode *inode = page->mapping->host;
2663 int ret = 0;
2664 int update_sd = 0;
2665 struct reiserfs_transaction_handle *th;
2666 unsigned start;
2667
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002668 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2669 pos ++;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002670
2671 reiserfs_wait_on_write_block(inode->i_sb);
2672 if (reiserfs_transaction_running(inode->i_sb))
2673 th = current->journal_info;
2674 else
2675 th = NULL;
2676
2677 start = pos & (PAGE_CACHE_SIZE - 1);
2678 if (unlikely(copied < len)) {
2679 if (!PageUptodate(page))
2680 copied = 0;
2681
2682 page_zero_new_buffers(page, start + copied, start + len);
2683 }
2684 flush_dcache_page(page);
2685
2686 reiserfs_commit_page(inode, page, start, start + copied);
2687
2688 /* generic_commit_write does this for us, but does not update the
2689 ** transaction tracking stuff when the size changes. So, we have
2690 ** to do the i_size updates here.
2691 */
2692 pos += copied;
2693 if (pos > inode->i_size) {
2694 struct reiserfs_transaction_handle myth;
2695 reiserfs_write_lock(inode->i_sb);
2696 /* If the file have grown beyond the border where it
2697 can have a tail, unmark it as needing a tail
2698 packing */
2699 if ((have_large_tails(inode->i_sb)
2700 && inode->i_size > i_block_size(inode) * 4)
2701 || (have_small_tails(inode->i_sb)
2702 && inode->i_size > i_block_size(inode)))
2703 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2704
2705 ret = journal_begin(&myth, inode->i_sb, 1);
2706 if (ret) {
2707 reiserfs_write_unlock(inode->i_sb);
2708 goto journal_error;
2709 }
2710 reiserfs_update_inode_transaction(inode);
2711 inode->i_size = pos;
2712 /*
2713 * this will just nest into our transaction. It's important
2714 * to use mark_inode_dirty so the inode gets pushed around on the
2715 * dirty lists, and so that O_SYNC works as expected
2716 */
2717 mark_inode_dirty(inode);
2718 reiserfs_update_sd(&myth, inode);
2719 update_sd = 1;
2720 ret = journal_end(&myth, inode->i_sb, 1);
2721 reiserfs_write_unlock(inode->i_sb);
2722 if (ret)
2723 goto journal_error;
2724 }
2725 if (th) {
2726 reiserfs_write_lock(inode->i_sb);
2727 if (!update_sd)
2728 mark_inode_dirty(inode);
2729 ret = reiserfs_end_persistent_transaction(th);
2730 reiserfs_write_unlock(inode->i_sb);
2731 if (ret)
2732 goto out;
2733 }
2734
2735 out:
2736 unlock_page(page);
2737 page_cache_release(page);
2738 return ret == 0 ? copied : ret;
2739
2740 journal_error:
2741 if (th) {
2742 reiserfs_write_lock(inode->i_sb);
2743 if (!update_sd)
2744 reiserfs_update_sd(th, inode);
2745 ret = reiserfs_end_persistent_transaction(th);
2746 reiserfs_write_unlock(inode->i_sb);
2747 }
2748
2749 goto out;
2750}
2751
2752int reiserfs_commit_write(struct file *f, struct page *page,
2753 unsigned from, unsigned to)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002754{
2755 struct inode *inode = page->mapping->host;
2756 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2757 int ret = 0;
2758 int update_sd = 0;
2759 struct reiserfs_transaction_handle *th = NULL;
2760
2761 reiserfs_wait_on_write_block(inode->i_sb);
2762 if (reiserfs_transaction_running(inode->i_sb)) {
2763 th = current->journal_info;
2764 }
2765 reiserfs_commit_page(inode, page, from, to);
2766
2767 /* generic_commit_write does this for us, but does not update the
2768 ** transaction tracking stuff when the size changes. So, we have
2769 ** to do the i_size updates here.
2770 */
2771 if (pos > inode->i_size) {
2772 struct reiserfs_transaction_handle myth;
2773 reiserfs_write_lock(inode->i_sb);
2774 /* If the file have grown beyond the border where it
2775 can have a tail, unmark it as needing a tail
2776 packing */
2777 if ((have_large_tails(inode->i_sb)
2778 && inode->i_size > i_block_size(inode) * 4)
2779 || (have_small_tails(inode->i_sb)
2780 && inode->i_size > i_block_size(inode)))
2781 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2782
2783 ret = journal_begin(&myth, inode->i_sb, 1);
2784 if (ret) {
2785 reiserfs_write_unlock(inode->i_sb);
2786 goto journal_error;
2787 }
2788 reiserfs_update_inode_transaction(inode);
2789 inode->i_size = pos;
Chris Mason9f037832005-09-13 01:25:17 -07002790 /*
2791 * this will just nest into our transaction. It's important
2792 * to use mark_inode_dirty so the inode gets pushed around on the
2793 * dirty lists, and so that O_SYNC works as expected
2794 */
2795 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002796 reiserfs_update_sd(&myth, inode);
2797 update_sd = 1;
2798 ret = journal_end(&myth, inode->i_sb, 1);
2799 reiserfs_write_unlock(inode->i_sb);
2800 if (ret)
2801 goto journal_error;
2802 }
2803 if (th) {
2804 reiserfs_write_lock(inode->i_sb);
2805 if (!update_sd)
Chris Mason9f037832005-09-13 01:25:17 -07002806 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002807 ret = reiserfs_end_persistent_transaction(th);
2808 reiserfs_write_unlock(inode->i_sb);
2809 if (ret)
2810 goto out;
2811 }
2812
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002813 out:
2814 return ret;
2815
2816 journal_error:
2817 if (th) {
2818 reiserfs_write_lock(inode->i_sb);
2819 if (!update_sd)
2820 reiserfs_update_sd(th, inode);
2821 ret = reiserfs_end_persistent_transaction(th);
2822 reiserfs_write_unlock(inode->i_sb);
2823 }
2824
2825 return ret;
2826}
2827
2828void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2829{
2830 if (reiserfs_attrs(inode->i_sb)) {
2831 if (sd_attrs & REISERFS_SYNC_FL)
2832 inode->i_flags |= S_SYNC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002834 inode->i_flags &= ~S_SYNC;
2835 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2836 inode->i_flags |= S_IMMUTABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002837 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002838 inode->i_flags &= ~S_IMMUTABLE;
2839 if (sd_attrs & REISERFS_APPEND_FL)
2840 inode->i_flags |= S_APPEND;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002842 inode->i_flags &= ~S_APPEND;
2843 if (sd_attrs & REISERFS_NOATIME_FL)
2844 inode->i_flags |= S_NOATIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002846 inode->i_flags &= ~S_NOATIME;
2847 if (sd_attrs & REISERFS_NOTAIL_FL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002848 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2849 else
2850 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2851 }
2852}
2853
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002854void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002856 if (reiserfs_attrs(inode->i_sb)) {
2857 if (inode->i_flags & S_IMMUTABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2859 else
2860 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002861 if (inode->i_flags & S_SYNC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862 *sd_attrs |= REISERFS_SYNC_FL;
2863 else
2864 *sd_attrs &= ~REISERFS_SYNC_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002865 if (inode->i_flags & S_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866 *sd_attrs |= REISERFS_NOATIME_FL;
2867 else
2868 *sd_attrs &= ~REISERFS_NOATIME_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002869 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002870 *sd_attrs |= REISERFS_NOTAIL_FL;
2871 else
2872 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2873 }
2874}
2875
2876/* decide if this buffer needs to stay around for data logging or ordered
2877** write purposes
2878*/
2879static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2880{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002881 int ret = 1;
2882 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002883
Chris Masond62b1b82006-02-01 03:06:47 -08002884 lock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002885 spin_lock(&j->j_dirty_buffers_lock);
2886 if (!buffer_mapped(bh)) {
2887 goto free_jh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002888 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002889 /* the page is locked, and the only places that log a data buffer
2890 * also lock the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002892 if (reiserfs_file_data_log(inode)) {
2893 /*
2894 * very conservative, leave the buffer pinned if
2895 * anyone might need it.
2896 */
2897 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2898 ret = 0;
2899 }
Chris Masond62b1b82006-02-01 03:06:47 -08002900 } else if (buffer_dirty(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002901 struct reiserfs_journal_list *jl;
2902 struct reiserfs_jh *jh = bh->b_private;
2903
2904 /* why is this safe?
2905 * reiserfs_setattr updates i_size in the on disk
2906 * stat data before allowing vmtruncate to be called.
2907 *
2908 * If buffer was put onto the ordered list for this
2909 * transaction, we know for sure either this transaction
2910 * or an older one already has updated i_size on disk,
2911 * and this ordered data won't be referenced in the file
2912 * if we crash.
2913 *
2914 * if the buffer was put onto the ordered list for an older
2915 * transaction, we need to leave it around
2916 */
2917 if (jh && (jl = jh->jl)
2918 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2919 ret = 0;
2920 }
2921 free_jh:
2922 if (ret && bh->b_private) {
2923 reiserfs_free_jh(bh);
2924 }
2925 spin_unlock(&j->j_dirty_buffers_lock);
Chris Masond62b1b82006-02-01 03:06:47 -08002926 unlock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002927 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002928}
2929
2930/* clm -- taken from fs/buffer.c:block_invalidate_page */
NeilBrown2ff28e22006-03-26 01:37:18 -08002931static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002932{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002933 struct buffer_head *head, *bh, *next;
2934 struct inode *inode = page->mapping->host;
2935 unsigned int curr_off = 0;
2936 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002937
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002938 BUG_ON(!PageLocked(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002939
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002940 if (offset == 0)
2941 ClearPageChecked(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002942
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002943 if (!page_has_buffers(page))
2944 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002945
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002946 head = page_buffers(page);
2947 bh = head;
2948 do {
2949 unsigned int next_off = curr_off + bh->b_size;
2950 next = bh->b_this_page;
2951
2952 /*
2953 * is this block fully invalidated?
2954 */
2955 if (offset <= curr_off) {
2956 if (invalidatepage_can_drop(inode, bh))
2957 reiserfs_unmap_buffer(bh);
2958 else
2959 ret = 0;
2960 }
2961 curr_off = next_off;
2962 bh = next;
2963 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002964
2965 /*
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002966 * We release buffers only if the entire page is being invalidated.
2967 * The get_block cached value has been unconditionally invalidated,
2968 * so real IO is not possible anymore.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002969 */
NeilBrown2ff28e22006-03-26 01:37:18 -08002970 if (!offset && ret) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002971 ret = try_to_release_page(page, 0);
NeilBrown2ff28e22006-03-26 01:37:18 -08002972 /* maybe should BUG_ON(!ret); - neilb */
2973 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002974 out:
NeilBrown2ff28e22006-03-26 01:37:18 -08002975 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002976}
2977
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002978static int reiserfs_set_page_dirty(struct page *page)
2979{
2980 struct inode *inode = page->mapping->host;
2981 if (reiserfs_file_data_log(inode)) {
2982 SetPageChecked(page);
2983 return __set_page_dirty_nobuffers(page);
2984 }
2985 return __set_page_dirty_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002986}
2987
2988/*
2989 * Returns 1 if the page's buffers were dropped. The page is locked.
2990 *
2991 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2992 * in the buffers at page_buffers(page).
2993 *
2994 * even in -o notail mode, we can't be sure an old mount without -o notail
2995 * didn't create files with tails.
2996 */
Al Viro27496a82005-10-21 03:20:48 -04002997static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002998{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002999 struct inode *inode = page->mapping->host;
3000 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3001 struct buffer_head *head;
3002 struct buffer_head *bh;
3003 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003004
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003005 WARN_ON(PageChecked(page));
3006 spin_lock(&j->j_dirty_buffers_lock);
3007 head = page_buffers(page);
3008 bh = head;
3009 do {
3010 if (bh->b_private) {
3011 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3012 reiserfs_free_jh(bh);
3013 } else {
3014 ret = 0;
3015 break;
3016 }
3017 }
3018 bh = bh->b_this_page;
3019 } while (bh != head);
3020 if (ret)
3021 ret = try_to_free_buffers(page);
3022 spin_unlock(&j->j_dirty_buffers_lock);
3023 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003024}
3025
3026/* We thank Mingming Cao for helping us understand in great detail what
3027 to do in this section of the code. */
3028static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003029 const struct iovec *iov, loff_t offset,
3030 unsigned long nr_segs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003031{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003032 struct file *file = iocb->ki_filp;
3033 struct inode *inode = file->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003034
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003035 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
3036 offset, nr_segs,
3037 reiserfs_get_blocks_direct_io, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003038}
3039
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003040int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3041{
3042 struct inode *inode = dentry->d_inode;
3043 int error;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003044 unsigned int ia_valid;
3045
3046 /* must be turned off for recursive notify_change calls */
3047 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3048
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003049 reiserfs_write_lock(inode->i_sb);
3050 if (attr->ia_valid & ATTR_SIZE) {
3051 /* version 2 items will be caught by the s_maxbytes check
3052 ** done for us in vmtruncate
3053 */
3054 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3055 attr->ia_size > MAX_NON_LFS) {
3056 error = -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003058 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003059 /* fill in hole pointers in the expanding truncate case. */
3060 if (attr->ia_size > inode->i_size) {
Vladimir Savelievf7557e82007-10-16 01:25:14 -07003061 error = generic_cont_expand_simple(inode, attr->ia_size);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003062 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3063 int err;
3064 struct reiserfs_transaction_handle th;
3065 /* we're changing at most 2 bitmaps, inode + super */
3066 err = journal_begin(&th, inode->i_sb, 4);
3067 if (!err) {
3068 reiserfs_discard_prealloc(&th, inode);
3069 err = journal_end(&th, inode->i_sb, 4);
3070 }
3071 if (err)
3072 error = err;
3073 }
3074 if (error)
3075 goto out;
Vladimir Savelievdd535a52006-07-01 04:36:32 -07003076 /*
3077 * file size is changed, ctime and mtime are
3078 * to be updated
3079 */
3080 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003081 }
3082 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003083
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003084 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
3085 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
3086 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3087 /* stat data of format v3.5 has 16 bit uid and gid */
3088 error = -EINVAL;
3089 goto out;
3090 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003091
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003092 error = inode_change_ok(inode, attr);
3093 if (!error) {
3094 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3095 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
3096 error = reiserfs_chown_xattrs(inode, attr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003097
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003098 if (!error) {
3099 struct reiserfs_transaction_handle th;
3100 int jbegin_count =
3101 2 *
3102 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3103 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3104 2;
3105
3106 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3107 error =
3108 journal_begin(&th, inode->i_sb,
3109 jbegin_count);
3110 if (error)
3111 goto out;
3112 error =
Jan Kara77db4f22009-01-26 17:14:18 +01003113 vfs_dq_transfer(inode, attr) ? -EDQUOT : 0;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003114 if (error) {
3115 journal_end(&th, inode->i_sb,
3116 jbegin_count);
3117 goto out;
3118 }
3119 /* Update corresponding info in inode so that everything is in
3120 * one transaction */
3121 if (attr->ia_valid & ATTR_UID)
3122 inode->i_uid = attr->ia_uid;
3123 if (attr->ia_valid & ATTR_GID)
3124 inode->i_gid = attr->ia_gid;
3125 mark_inode_dirty(inode);
3126 error =
3127 journal_end(&th, inode->i_sb, jbegin_count);
3128 }
3129 }
3130 if (!error)
3131 error = inode_setattr(inode, attr);
3132 }
3133
3134 if (!error && reiserfs_posixacl(inode->i_sb)) {
3135 if (attr->ia_valid & ATTR_MODE)
3136 error = reiserfs_acl_chmod(inode);
3137 }
3138
3139 out:
3140 reiserfs_write_unlock(inode->i_sb);
3141 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003142}
3143
Christoph Hellwigf5e54d62006-06-28 04:26:44 -07003144const struct address_space_operations reiserfs_address_space_operations = {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003145 .writepage = reiserfs_writepage,
3146 .readpage = reiserfs_readpage,
3147 .readpages = reiserfs_readpages,
3148 .releasepage = reiserfs_releasepage,
3149 .invalidatepage = reiserfs_invalidatepage,
3150 .sync_page = block_sync_page,
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07003151 .write_begin = reiserfs_write_begin,
3152 .write_end = reiserfs_write_end,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003153 .bmap = reiserfs_aop_bmap,
3154 .direct_IO = reiserfs_direct_IO,
3155 .set_page_dirty = reiserfs_set_page_dirty,
3156};