blob: 4fd5bb33dbb5e40f4969e38d7c4b52756d8bfab7 [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/pagemap.h>
12#include <linux/highmem.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090013#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#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);
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
Al Viro845a2cc2010-06-07 11:37:37 -040025void reiserfs_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -070026{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070027 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
28 int jbegin_count =
29 JOURNAL_PER_BALANCE_CNT * 2 +
30 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
31 struct reiserfs_transaction_handle th;
Frederic Weisbeckercb1c2e52009-12-13 23:32:06 +010032 int depth;
Jeff Mahoney24996042005-12-14 14:38:05 -050033 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Al Viro845a2cc2010-06-07 11:37:37 -040035 if (!inode->i_nlink && !is_bad_inode(inode))
Christoph Hellwig871a2932010-03-03 09:05:07 -050036 dquot_initialize(inode);
Christoph Hellwig907f4552010-03-03 09:05:06 -050037
Mark Fashehfef26652005-09-09 13:01:31 -070038 truncate_inode_pages(&inode->i_data, 0);
Al Viro845a2cc2010-06-07 11:37:37 -040039 if (inode->i_nlink)
40 goto no_delete;
Mark Fashehfef26652005-09-09 13:01:31 -070041
Frederic Weisbeckercb1c2e52009-12-13 23:32:06 +010042 depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070044 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
45 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070046 reiserfs_delete_xattrs(inode);
47
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070048 if (journal_begin(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070049 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070050 reiserfs_update_inode_transaction(inode);
51
Jeff Mahoneyeb35c212008-07-08 14:37:06 -040052 reiserfs_discard_prealloc(&th, inode);
53
Jeff Mahoney24996042005-12-14 14:38:05 -050054 err = reiserfs_delete_object(&th, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070055
56 /* Do quota update inside a transaction for journaled quotas. We must do that
57 * after delete_object so that quota updates go into the same transaction as
58 * stat data deletion */
Jeff Mahoney24996042005-12-14 14:38:05 -050059 if (!err)
Christoph Hellwig63936dd2010-03-03 09:05:01 -050060 dquot_free_inode(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070061
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070062 if (journal_end(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070063 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070064
Jeff Mahoney24996042005-12-14 14:38:05 -050065 /* check return value from reiserfs_delete_object after
66 * ending the transaction
67 */
68 if (err)
69 goto out;
70
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070071 /* all items of file are deleted, so we can remove "save" link */
72 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
73 * about an error here */
74 } else {
75 /* no object items are in the tree */
76 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070078 out:
Al Viro845a2cc2010-06-07 11:37:37 -040079 end_writeback(inode); /* note this must go after the journal_end to prevent deadlock */
80 dquot_drop(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070081 inode->i_blocks = 0;
Frederic Weisbeckercb1c2e52009-12-13 23:32:06 +010082 reiserfs_write_unlock_once(inode->i_sb, depth);
Sergey Senozhatskyf4ae2fa2010-08-11 14:07:01 +030083 return;
Al Viro845a2cc2010-06-07 11:37:37 -040084
85no_delete:
86 end_writeback(inode);
87 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088}
89
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070090static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
91 __u32 objectid, loff_t offset, int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070092{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070093 key->version = version;
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070095 key->on_disk_key.k_dir_id = dirid;
96 key->on_disk_key.k_objectid = objectid;
97 set_cpu_key_k_offset(key, offset);
98 set_cpu_key_k_type(key, type);
99 key->key_length = length;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100}
101
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102/* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
103 offset and type of key */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700104void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
105 int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700107 _make_cpu_key(key, get_inode_item_key_version(inode),
108 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
109 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
110 length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111}
112
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113//
114// when key is 0, do not set version and short key
115//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700116inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
117 int version,
118 loff_t offset, int type, int length,
119 int entry_count /*or ih_free_space */ )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700121 if (key) {
122 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
123 ih->ih_key.k_objectid =
124 cpu_to_le32(key->on_disk_key.k_objectid);
125 }
126 put_ih_version(ih, version);
127 set_le_ih_k_offset(ih, offset);
128 set_le_ih_k_type(ih, type);
129 put_ih_item_len(ih, length);
130 /* set_ih_free_space (ih, 0); */
131 // for directory items it is entry count, for directs and stat
132 // datas - 0xffff, for indirects - 0
133 put_ih_entry_count(ih, entry_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134}
135
136//
137// FIXME: we might cache recently accessed indirect item
138
139// Ugh. Not too eager for that....
140// I cut the code until such time as I see a convincing argument (benchmark).
141// I don't want a bloated inode struct..., and I don't like code complexity....
142
143/* cutting the code is fine, since it really isn't in use yet and is easy
144** to add back in. But, Vladimir has a really good idea here. Think
145** about what happens for reading a file. For each page,
146** The VFS layer calls reiserfs_readpage, who searches the tree to find
147** an indirect item. This indirect item has X number of pointers, where
148** X is a big number if we've done the block allocation right. But,
149** we only use one or two of these pointers during each call to readpage,
150** needlessly researching again later on.
151**
152** The size of the cache could be dynamic based on the size of the file.
153**
154** I'd also like to see us cache the location the stat data item, since
155** we are needlessly researching for that frequently.
156**
157** --chris
158*/
159
160/* If this page has a file tail in it, and
161** it was read in by get_block_create_0, the page data is valid,
162** but tail is still sitting in a direct item, and we can't write to
163** it. So, look through this page, and check all the mapped buffers
164** to make sure they have valid block numbers. Any that don't need
Christoph Hellwigebdec242010-10-06 10:47:23 +0200165** to be unmapped, so that __block_write_begin will correctly call
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166** reiserfs_get_block to convert the tail into an unformatted node
167*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700168static inline void fix_tail_page_for_writing(struct page *page)
169{
170 struct buffer_head *head, *next, *bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700172 if (page && page_has_buffers(page)) {
173 head = page_buffers(page);
174 bh = head;
175 do {
176 next = bh->b_this_page;
177 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
178 reiserfs_unmap_buffer(bh);
179 }
180 bh = next;
181 } while (bh != head);
182 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183}
184
185/* reiserfs_get_block does not need to allocate a block only if it has been
186 done already or non-hole position has been found in the indirect item */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700187static inline int allocation_needed(int retval, b_blocknr_t allocated,
188 struct item_head *ih,
189 __le32 * item, int pos_in_item)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700191 if (allocated)
192 return 0;
193 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
194 get_block_num(item, pos_in_item))
195 return 0;
196 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197}
198
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700199static inline int indirect_item_found(int retval, struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700201 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202}
203
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700204static inline void set_block_dev_mapped(struct buffer_head *bh,
205 b_blocknr_t block, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206{
207 map_bh(bh, inode->i_sb, block);
208}
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210//
211// files which were created in the earlier version can not be longer,
212// than 2 gb
213//
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700214static int file_capable(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700216 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
217 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
218 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700220 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221}
222
Adrian Bunkdeba0f42007-10-16 23:26:03 -0700223static int restart_transaction(struct reiserfs_transaction_handle *th,
224 struct inode *inode, struct treepath *path)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700225{
226 struct super_block *s = th->t_super;
227 int len = th->t_blocks_allocated;
228 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700230 BUG_ON(!th->t_trans_id);
231 BUG_ON(!th->t_refcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232
Suzuki K P87b41262006-12-06 20:36:10 -0800233 pathrelse(path);
234
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700235 /* we cannot restart while nested */
236 if (th->t_refcount > 1) {
237 return 0;
238 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700239 reiserfs_update_sd(th, inode);
240 err = journal_end(th, s, len);
241 if (!err) {
242 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
243 if (!err)
244 reiserfs_update_inode_transaction(inode);
245 }
246 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247}
248
249// it is called by get_block when create == 0. Returns block number
250// for 'block'-th logical block of file. When it hits direct item it
251// returns 0 (being called from bmap) or read direct item into piece
252// of page (bh_result)
253
254// Please improve the english/clarity in the comment above, as it is
255// hard to understand.
256
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700257static int _get_block_create_0(struct inode *inode, sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700258 struct buffer_head *bh_result, int args)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700260 INITIALIZE_PATH(path);
261 struct cpu_key key;
262 struct buffer_head *bh;
263 struct item_head *ih, tmp_ih;
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700264 b_blocknr_t blocknr;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700265 char *p = NULL;
266 int chars;
267 int ret;
268 int result;
269 int done = 0;
270 unsigned long offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700272 // prepare the key to look for the 'block'-th block of file
273 make_cpu_key(&key, inode,
274 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
275 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700277 result = search_for_position_by_key(inode->i_sb, &key, &path);
278 if (result != POSITION_FOUND) {
279 pathrelse(&path);
280 if (p)
281 kunmap(bh_result->b_page);
282 if (result == IO_ERROR)
283 return -EIO;
284 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
285 // That there is some MMAPED data associated with it that is yet to be written to disk.
286 if ((args & GET_BLOCK_NO_HOLE)
287 && !PageUptodate(bh_result->b_page)) {
288 return -ENOENT;
289 }
290 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700292 //
293 bh = get_last_bh(&path);
294 ih = get_ih(&path);
295 if (is_indirect_le_ih(ih)) {
296 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700298 /* FIXME: here we could cache indirect item or part of it in
299 the inode to avoid search_by_key in case of subsequent
300 access to file */
301 blocknr = get_block_num(ind_item, path.pos_in_item);
302 ret = 0;
303 if (blocknr) {
304 map_bh(bh_result, inode->i_sb, blocknr);
305 if (path.pos_in_item ==
306 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
307 set_buffer_boundary(bh_result);
308 }
309 } else
310 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
311 // That there is some MMAPED data associated with it that is yet to be written to disk.
312 if ((args & GET_BLOCK_NO_HOLE)
313 && !PageUptodate(bh_result->b_page)) {
314 ret = -ENOENT;
315 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700317 pathrelse(&path);
318 if (p)
319 kunmap(bh_result->b_page);
320 return ret;
321 }
322 // requested data are in direct item(s)
323 if (!(args & GET_BLOCK_READ_DIRECT)) {
324 // we are called by bmap. FIXME: we can not map block of file
325 // when it is stored in direct item(s)
326 pathrelse(&path);
327 if (p)
328 kunmap(bh_result->b_page);
329 return -ENOENT;
330 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700332 /* if we've got a direct item, and the buffer or page was uptodate,
333 ** we don't want to pull data off disk again. skip to the
334 ** end, where we map the buffer and return
335 */
336 if (buffer_uptodate(bh_result)) {
337 goto finished;
338 } else
339 /*
340 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
341 ** pages without any buffers. If the page is up to date, we don't want
342 ** read old data off disk. Set the up to date bit on the buffer instead
343 ** and jump to the end
344 */
345 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 set_buffer_uptodate(bh_result);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700347 goto finished;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700349 // read file tail into part of page
350 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700351 copy_item_head(&tmp_ih, ih);
352
353 /* we only want to kmap if we are reading the tail into the page.
354 ** this is not the common case, so we don't kmap until we are
355 ** sure we need to. But, this means the item might move if
356 ** kmap schedules
357 */
Frederic Weisbecker27b3a5c2009-10-14 23:34:31 +0200358 if (!p)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700359 p = (char *)kmap(bh_result->b_page);
Frederic Weisbecker27b3a5c2009-10-14 23:34:31 +0200360
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700361 p += offset;
362 memset(p, 0, inode->i_sb->s_blocksize);
363 do {
364 if (!is_direct_le_ih(ih)) {
365 BUG();
366 }
367 /* make sure we don't read more bytes than actually exist in
368 ** the file. This can happen in odd cases where i_size isn't
Jeff Mahoney0222e652009-03-30 14:02:44 -0400369 ** correct, and when direct item padding results in a few
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700370 ** extra bytes at the end of the direct item
371 */
372 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
373 break;
374 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
375 chars =
376 inode->i_size - (le_ih_k_offset(ih) - 1) -
377 path.pos_in_item;
378 done = 1;
379 } else {
380 chars = ih_item_len(ih) - path.pos_in_item;
381 }
382 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700384 if (done)
385 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700387 p += chars;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700389 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
390 // we done, if read direct item is not the last item of
391 // node FIXME: we could try to check right delimiting key
392 // to see whether direct item continues in the right
393 // neighbor or rely on i_size
394 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700396 // update key to look for the next piece
397 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
398 result = search_for_position_by_key(inode->i_sb, &key, &path);
399 if (result != POSITION_FOUND)
400 // i/o error most likely
401 break;
402 bh = get_last_bh(&path);
403 ih = get_ih(&path);
404 } while (1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700406 flush_dcache_page(bh_result->b_page);
407 kunmap(bh_result->b_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700409 finished:
410 pathrelse(&path);
Qu Fuping6283d582005-06-25 14:55:44 -0700411
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700412 if (result == IO_ERROR)
413 return -EIO;
Qu Fuping6283d582005-06-25 14:55:44 -0700414
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700415 /* this buffer has valid data, but isn't valid for io. mapping it to
416 * block #0 tells the rest of reiserfs it just has a tail in it
417 */
418 map_bh(bh_result, inode->i_sb, 0);
419 set_buffer_uptodate(bh_result);
420 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421}
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423// this is called to create file map. So, _get_block_create_0 will not
424// read direct item
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700425static int reiserfs_bmap(struct inode *inode, sector_t block,
426 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700428 if (!file_capable(inode, block))
429 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700431 reiserfs_write_lock(inode->i_sb);
432 /* do not read the direct item */
433 _get_block_create_0(inode, block, bh_result, 0);
434 reiserfs_write_unlock(inode->i_sb);
435 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436}
437
438/* special version of get_block that is only used by grab_tail_page right
Christoph Hellwigebdec242010-10-06 10:47:23 +0200439** now. It is sent to __block_write_begin, and when you try to get a
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440** block past the end of the file (or a block from a hole) it returns
Christoph Hellwigebdec242010-10-06 10:47:23 +0200441** -ENOENT instead of a valid buffer. __block_write_begin expects to
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442** be able to do i/o on the buffers returned, unless an error value
443** is also returned.
Jeff Mahoney0222e652009-03-30 14:02:44 -0400444**
Christoph Hellwigebdec242010-10-06 10:47:23 +0200445** So, this allows __block_write_begin to be used for reading a single block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446** in a page. Where it does not produce a valid page for holes, or past the
447** end of the file. This turns out to be exactly what we need for reading
448** tails for conversion.
449**
450** The point of the wrapper is forcing a certain value for create, even
Jeff Mahoney0222e652009-03-30 14:02:44 -0400451** though the VFS layer is calling this function with create==1. If you
452** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453** don't use this function.
454*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700455static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
456 struct buffer_head *bh_result,
457 int create)
458{
459 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460}
461
462/* This is special helper for reiserfs_get_block in case we are executing
463 direct_IO request. */
464static int reiserfs_get_blocks_direct_io(struct inode *inode,
465 sector_t iblock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 struct buffer_head *bh_result,
467 int create)
468{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700469 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700471 bh_result->b_page = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700473 /* We set the b_size before reiserfs_get_block call since it is
474 referenced in convert_tail_for_hole() that may be called from
475 reiserfs_get_block() */
476 bh_result->b_size = (1 << inode->i_blkbits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700478 ret = reiserfs_get_block(inode, iblock, bh_result,
479 create | GET_BLOCK_NO_DANGLE);
480 if (ret)
481 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700483 /* don't allow direct io onto tail pages */
484 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
485 /* make sure future calls to the direct io funcs for this offset
486 ** in the file fail by unmapping the buffer
487 */
488 clear_buffer_mapped(bh_result);
489 ret = -EINVAL;
490 }
491 /* Possible unpacked tail. Flush the data before pages have
492 disappeared */
493 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
494 int err;
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +0200495
496 reiserfs_write_lock(inode->i_sb);
497
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700498 err = reiserfs_commit_for_inode(inode);
499 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +0200500
501 reiserfs_write_unlock(inode->i_sb);
502
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700503 if (err < 0)
504 ret = err;
505 }
506 out:
507 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508}
509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510/*
511** helper function for when reiserfs_get_block is called for a hole
512** but the file tail is still in a direct item
513** bh_result is the buffer head for the hole
514** tail_offset is the offset of the start of the tail in the file
515**
516** This calls prepare_write, which will start a new transaction
517** you should not be in a transaction, or have any paths held when you
518** call this.
519*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700520static int convert_tail_for_hole(struct inode *inode,
521 struct buffer_head *bh_result,
522 loff_t tail_offset)
523{
524 unsigned long index;
525 unsigned long tail_end;
526 unsigned long tail_start;
527 struct page *tail_page;
528 struct page *hole_page = bh_result->b_page;
529 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700531 if ((tail_offset & (bh_result->b_size - 1)) != 1)
532 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700534 /* always try to read until the end of the block */
535 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
536 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700538 index = tail_offset >> PAGE_CACHE_SHIFT;
539 /* hole_page can be zero in case of direct_io, we are sure
540 that we cannot get here if we write with O_DIRECT into
541 tail page */
542 if (!hole_page || index != hole_page->index) {
543 tail_page = grab_cache_page(inode->i_mapping, index);
544 retval = -ENOMEM;
545 if (!tail_page) {
546 goto out;
547 }
548 } else {
549 tail_page = hole_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700552 /* we don't have to make sure the conversion did not happen while
553 ** we were locking the page because anyone that could convert
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800554 ** must first take i_mutex.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700555 **
556 ** We must fix the tail page for writing because it might have buffers
557 ** that are mapped, but have a block number of 0. This indicates tail
Christoph Hellwigebdec242010-10-06 10:47:23 +0200558 ** data that has been read directly into the page, and
559 ** __block_write_begin won't trigger a get_block in this case.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700560 */
561 fix_tail_page_for_writing(tail_page);
Christoph Hellwigebdec242010-10-06 10:47:23 +0200562 retval = __reiserfs_write_begin(tail_page, tail_start,
563 tail_end - tail_start);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700564 if (retval)
565 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700567 /* tail conversion might change the data in the page */
568 flush_dcache_page(tail_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700570 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700572 unlock:
573 if (tail_page != hole_page) {
574 unlock_page(tail_page);
575 page_cache_release(tail_page);
576 }
577 out:
578 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579}
580
581static inline int _allocate_block(struct reiserfs_transaction_handle *th,
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700582 sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700583 struct inode *inode,
584 b_blocknr_t * allocated_block_nr,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -0800585 struct treepath *path, int flags)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700586{
587 BUG_ON(!th->t_trans_id);
588
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589#ifdef REISERFS_PREALLOCATE
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800590 if (!(flags & GET_BLOCK_NO_IMUX)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700591 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
592 path, block);
593 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700595 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
596 block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597}
598
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700599int reiserfs_get_block(struct inode *inode, sector_t block,
600 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700602 int repeat, retval = 0;
603 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
604 INITIALIZE_PATH(path);
605 int pos_in_item;
606 struct cpu_key key;
607 struct buffer_head *bh, *unbh = NULL;
608 struct item_head *ih, tmp_ih;
609 __le32 *item;
610 int done;
611 int fs_gen;
Frederic Weisbecker26931302009-05-07 23:48:44 +0200612 int lock_depth;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700613 struct reiserfs_transaction_handle *th = NULL;
Jeff Mahoney0222e652009-03-30 14:02:44 -0400614 /* space reserved in transaction batch:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700615 . 3 balancings in direct->indirect conversion
616 . 1 block involved into reiserfs_update_sd()
617 XXX in practically impossible worst case direct2indirect()
618 can incur (much) more than 3 balancings.
619 quota update for user, group */
620 int jbegin_count =
621 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
622 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
623 int version;
624 int dangle = 1;
625 loff_t new_offset =
626 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Frederic Weisbecker26931302009-05-07 23:48:44 +0200628 lock_depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700629 version = get_inode_item_key_version(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700631 if (!file_capable(inode, block)) {
Frederic Weisbecker26931302009-05-07 23:48:44 +0200632 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700633 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 }
635
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700636 /* if !create, we aren't changing the FS, so we don't need to
637 ** log anything, so we don't need to start a transaction
638 */
639 if (!(create & GET_BLOCK_CREATE)) {
640 int ret;
641 /* find number of block-th logical block of the file */
642 ret = _get_block_create_0(inode, block, bh_result,
643 create | GET_BLOCK_READ_DIRECT);
Frederic Weisbecker26931302009-05-07 23:48:44 +0200644 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700645 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700647 /*
648 * if we're already in a transaction, make sure to close
649 * any new transactions we start in this func
650 */
651 if ((create & GET_BLOCK_NO_DANGLE) ||
652 reiserfs_transaction_running(inode->i_sb))
653 dangle = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700655 /* If file is of such a size, that it might have a tail and tails are enabled
656 ** we should mark it as possibly needing tail packing on close
657 */
658 if ((have_large_tails(inode->i_sb)
659 && inode->i_size < i_block_size(inode) * 4)
660 || (have_small_tails(inode->i_sb)
661 && inode->i_size < i_block_size(inode)))
662 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
663
664 /* set the key of the first byte in the 'block'-th block of file */
665 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
666 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
667 start_trans:
668 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
669 if (!th) {
670 retval = -ENOMEM;
671 goto failure;
672 }
673 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700675 research:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700677 retval = search_for_position_by_key(inode->i_sb, &key, &path);
678 if (retval == IO_ERROR) {
679 retval = -EIO;
680 goto failure;
681 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700683 bh = get_last_bh(&path);
684 ih = get_ih(&path);
685 item = get_item(&path);
686 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700688 fs_gen = get_generation(inode->i_sb);
689 copy_item_head(&tmp_ih, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700691 if (allocation_needed
692 (retval, allocated_block_nr, ih, item, pos_in_item)) {
693 /* we have to allocate block for the unformatted node */
694 if (!th) {
695 pathrelse(&path);
696 goto start_trans;
697 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700699 repeat =
700 _allocate_block(th, block, inode, &allocated_block_nr,
701 &path, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700703 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
704 /* restart the transaction to give the journal a chance to free
705 ** some blocks. releases the path, so we have to go back to
706 ** research if we succeed on the second try
707 */
708 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
709 retval = restart_transaction(th, inode, &path);
710 if (retval)
711 goto failure;
712 repeat =
713 _allocate_block(th, block, inode,
714 &allocated_block_nr, NULL, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700716 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
717 goto research;
718 }
719 if (repeat == QUOTA_EXCEEDED)
720 retval = -EDQUOT;
721 else
722 retval = -ENOSPC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 goto failure;
724 }
725
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700726 if (fs_changed(fs_gen, inode->i_sb)
727 && item_moved(&tmp_ih, &path)) {
728 goto research;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700730 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700732 if (indirect_item_found(retval, ih)) {
733 b_blocknr_t unfm_ptr;
734 /* 'block'-th block is in the file already (there is
735 corresponding cell in some indirect item). But it may be
736 zero unformatted node pointer (hole) */
737 unfm_ptr = get_block_num(item, pos_in_item);
738 if (unfm_ptr == 0) {
739 /* use allocated block to plug the hole */
740 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
741 if (fs_changed(fs_gen, inode->i_sb)
742 && item_moved(&tmp_ih, &path)) {
743 reiserfs_restore_prepared_buffer(inode->i_sb,
744 bh);
745 goto research;
746 }
747 set_buffer_new(bh_result);
748 if (buffer_dirty(bh_result)
749 && reiserfs_data_ordered(inode->i_sb))
750 reiserfs_add_ordered_list(inode, bh_result);
751 put_block_num(item, pos_in_item, allocated_block_nr);
752 unfm_ptr = allocated_block_nr;
753 journal_mark_dirty(th, inode->i_sb, bh);
754 reiserfs_update_sd(th, inode);
755 }
756 set_block_dev_mapped(bh_result, unfm_ptr, inode);
757 pathrelse(&path);
758 retval = 0;
759 if (!dangle && th)
760 retval = reiserfs_end_persistent_transaction(th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Frederic Weisbecker26931302009-05-07 23:48:44 +0200762 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700763
764 /* the item was found, so new blocks were not added to the file
Jeff Mahoney0222e652009-03-30 14:02:44 -0400765 ** there is no need to make sure the inode is updated with this
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700766 ** transaction
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700768 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 }
770
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700771 if (!th) {
772 pathrelse(&path);
773 goto start_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700776 /* desired position is not found or is in the direct item. We have
777 to append file with holes up to 'block'-th block converting
778 direct items to indirect one if necessary */
779 done = 0;
780 do {
781 if (is_statdata_le_ih(ih)) {
782 __le32 unp = 0;
783 struct cpu_key tmp_key;
784
785 /* indirect item has to be inserted */
786 make_le_item_head(&tmp_ih, &key, version, 1,
787 TYPE_INDIRECT, UNFM_P_SIZE,
788 0 /* free_space */ );
789
790 if (cpu_key_k_offset(&key) == 1) {
791 /* we are going to add 'block'-th block to the file. Use
792 allocated block for that */
793 unp = cpu_to_le32(allocated_block_nr);
794 set_block_dev_mapped(bh_result,
795 allocated_block_nr, inode);
796 set_buffer_new(bh_result);
797 done = 1;
798 }
799 tmp_key = key; // ;)
800 set_cpu_key_k_offset(&tmp_key, 1);
801 PATH_LAST_POSITION(&path)++;
802
803 retval =
804 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
805 inode, (char *)&unp);
806 if (retval) {
807 reiserfs_free_block(th, inode,
808 allocated_block_nr, 1);
809 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
810 }
811 //mark_tail_converted (inode);
812 } else if (is_direct_le_ih(ih)) {
813 /* direct item has to be converted */
814 loff_t tail_offset;
815
816 tail_offset =
817 ((le_ih_k_offset(ih) -
818 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
819 if (tail_offset == cpu_key_k_offset(&key)) {
820 /* direct item we just found fits into block we have
821 to map. Convert it into unformatted node: use
822 bh_result for the conversion */
823 set_block_dev_mapped(bh_result,
824 allocated_block_nr, inode);
825 unbh = bh_result;
826 done = 1;
827 } else {
828 /* we have to padd file tail stored in direct item(s)
829 up to block size and convert it to unformatted
830 node. FIXME: this should also get into page cache */
831
832 pathrelse(&path);
833 /*
834 * ugly, but we can only end the transaction if
835 * we aren't nested
836 */
837 BUG_ON(!th->t_refcount);
838 if (th->t_refcount == 1) {
839 retval =
840 reiserfs_end_persistent_transaction
841 (th);
842 th = NULL;
843 if (retval)
844 goto failure;
845 }
846
847 retval =
848 convert_tail_for_hole(inode, bh_result,
849 tail_offset);
850 if (retval) {
851 if (retval != -ENOSPC)
Jeff Mahoney0030b642009-03-30 14:02:28 -0400852 reiserfs_error(inode->i_sb,
853 "clm-6004",
854 "convert tail failed "
855 "inode %lu, error %d",
856 inode->i_ino,
857 retval);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700858 if (allocated_block_nr) {
859 /* the bitmap, the super, and the stat data == 3 */
860 if (!th)
861 th = reiserfs_persistent_transaction(inode->i_sb, 3);
862 if (th)
863 reiserfs_free_block(th,
864 inode,
865 allocated_block_nr,
866 1);
867 }
868 goto failure;
869 }
870 goto research;
871 }
872 retval =
873 direct2indirect(th, inode, &path, unbh,
874 tail_offset);
875 if (retval) {
876 reiserfs_unmap_buffer(unbh);
877 reiserfs_free_block(th, inode,
878 allocated_block_nr, 1);
879 goto failure;
880 }
881 /* it is important the set_buffer_uptodate is done after
882 ** the direct2indirect. The buffer might contain valid
883 ** data newer than the data on disk (read by readpage, changed,
884 ** and then sent here by writepage). direct2indirect needs
885 ** to know if unbh was already up to date, so it can decide
886 ** if the data in unbh needs to be replaced with data from
887 ** the disk
888 */
889 set_buffer_uptodate(unbh);
890
891 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
892 buffer will disappear shortly, so it should not be added to
893 */
894 if (unbh->b_page) {
895 /* we've converted the tail, so we must
896 ** flush unbh before the transaction commits
897 */
898 reiserfs_add_tail_list(inode, unbh);
899
900 /* mark it dirty now to prevent commit_write from adding
901 ** this buffer to the inode's dirty buffer list
902 */
903 /*
904 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
905 * It's still atomic, but it sets the page dirty too,
906 * which makes it eligible for writeback at any time by the
907 * VM (which was also the case with __mark_buffer_dirty())
908 */
909 mark_buffer_dirty(unbh);
910 }
911 } else {
912 /* append indirect item with holes if needed, when appending
913 pointer to 'block'-th block use block, which is already
914 allocated */
915 struct cpu_key tmp_key;
916 unp_t unf_single = 0; // We use this in case we need to allocate only
917 // one block which is a fastpath
918 unp_t *un;
919 __u64 max_to_insert =
920 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
921 UNFM_P_SIZE;
922 __u64 blocks_needed;
923
924 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
925 "vs-804: invalid position for append");
926 /* indirect item has to be appended, set up key of that position */
927 make_cpu_key(&tmp_key, inode,
928 le_key_k_offset(version,
929 &(ih->ih_key)) +
930 op_bytes_number(ih,
931 inode->i_sb->s_blocksize),
932 //pos_in_item * inode->i_sb->s_blocksize,
933 TYPE_INDIRECT, 3); // key type is unimportant
934
Vladimir V. Savelievc499ec22006-03-02 02:54:39 -0800935 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
936 "green-805: invalid offset");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700937 blocks_needed =
938 1 +
939 ((cpu_key_k_offset(&key) -
940 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
941 s_blocksize_bits);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700942
943 if (blocks_needed == 1) {
944 un = &unf_single;
945 } else {
Frederic Weisbecker1d2c6cf2009-11-20 04:11:30 +0100946 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_NOFS);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700947 if (!un) {
948 un = &unf_single;
949 blocks_needed = 1;
950 max_to_insert = 0;
Yan Burman01afb212006-12-06 20:39:01 -0800951 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700952 }
953 if (blocks_needed <= max_to_insert) {
954 /* we are going to add target block to the file. Use allocated
955 block for that */
956 un[blocks_needed - 1] =
957 cpu_to_le32(allocated_block_nr);
958 set_block_dev_mapped(bh_result,
959 allocated_block_nr, inode);
960 set_buffer_new(bh_result);
961 done = 1;
962 } else {
963 /* paste hole to the indirect item */
964 /* If kmalloc failed, max_to_insert becomes zero and it means we
965 only have space for one block */
966 blocks_needed =
967 max_to_insert ? max_to_insert : 1;
968 }
969 retval =
970 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
971 (char *)un,
972 UNFM_P_SIZE *
973 blocks_needed);
974
975 if (blocks_needed != 1)
976 kfree(un);
977
978 if (retval) {
979 reiserfs_free_block(th, inode,
980 allocated_block_nr, 1);
981 goto failure;
982 }
983 if (!done) {
984 /* We need to mark new file size in case this function will be
985 interrupted/aborted later on. And we may do this only for
986 holes. */
987 inode->i_size +=
988 inode->i_sb->s_blocksize * blocks_needed;
989 }
990 }
991
992 if (done == 1)
993 break;
994
995 /* this loop could log more blocks than we had originally asked
996 ** for. So, we have to allow the transaction to end if it is
Jeff Mahoney0222e652009-03-30 14:02:44 -0400997 ** too big or too full. Update the inode so things are
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700998 ** consistent if we crash before the function returns
999 **
1000 ** release the path so that anybody waiting on the path before
1001 ** ending their transaction will be able to continue.
1002 */
1003 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1004 retval = restart_transaction(th, inode, &path);
1005 if (retval)
1006 goto failure;
1007 }
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02001008 /*
1009 * inserting indirect pointers for a hole can take a
1010 * long time. reschedule if needed and also release the write
1011 * lock for others.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001012 */
Frederic Weisbecker26931302009-05-07 23:48:44 +02001013 if (need_resched()) {
1014 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
1015 schedule();
1016 lock_depth = reiserfs_write_lock_once(inode->i_sb);
1017 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001018
1019 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1020 if (retval == IO_ERROR) {
1021 retval = -EIO;
1022 goto failure;
1023 }
1024 if (retval == POSITION_FOUND) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001025 reiserfs_warning(inode->i_sb, "vs-825",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001026 "%K should not be found", &key);
1027 retval = -EEXIST;
1028 if (allocated_block_nr)
1029 reiserfs_free_block(th, inode,
1030 allocated_block_nr, 1);
1031 pathrelse(&path);
1032 goto failure;
1033 }
1034 bh = get_last_bh(&path);
1035 ih = get_ih(&path);
1036 item = get_item(&path);
1037 pos_in_item = path.pos_in_item;
1038 } while (1);
1039
1040 retval = 0;
1041
1042 failure:
1043 if (th && (!dangle || (retval && !th->t_trans_id))) {
1044 int err;
1045 if (th->t_trans_id)
1046 reiserfs_update_sd(th, inode);
1047 err = reiserfs_end_persistent_transaction(th);
1048 if (err)
1049 retval = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
Frederic Weisbecker26931302009-05-07 23:48:44 +02001052 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001053 reiserfs_check_path(&path);
1054 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055}
1056
1057static int
1058reiserfs_readpages(struct file *file, struct address_space *mapping,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001059 struct list_head *pages, unsigned nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001061 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062}
1063
1064/* Compute real number of used bytes by file
1065 * Following three functions can go away when we'll have enough space in stat item
1066 */
1067static int real_space_diff(struct inode *inode, int sd_size)
1068{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001069 int bytes;
1070 loff_t blocksize = inode->i_sb->s_blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001072 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1073 return sd_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001075 /* End of file is also in full block with indirect reference, so round
1076 ** up to the next block.
1077 **
1078 ** there is just no way to know if the tail is actually packed
1079 ** on the file, so we have to assume it isn't. When we pack the
1080 ** tail, we add 4 bytes to pretend there really is an unformatted
1081 ** node pointer
1082 */
1083 bytes =
1084 ((inode->i_size +
1085 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1086 sd_size;
1087 return bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088}
1089
1090static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001091 int sd_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001093 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1094 return inode->i_size +
1095 (loff_t) (real_space_diff(inode, sd_size));
1096 }
1097 return ((loff_t) real_space_diff(inode, sd_size)) +
1098 (((loff_t) blocks) << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099}
1100
1101/* Compute number of blocks used by file in ReiserFS counting */
1102static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1103{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001104 loff_t bytes = inode_get_bytes(inode);
1105 loff_t real_space = real_space_diff(inode, sd_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001107 /* keeps fsck and non-quota versions of reiserfs happy */
1108 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1109 bytes += (loff_t) 511;
1110 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001112 /* files from before the quota patch might i_blocks such that
1113 ** bytes < real_space. Deal with that here to prevent it from
1114 ** going negative.
1115 */
1116 if (bytes < real_space)
1117 return 0;
1118 return (bytes - real_space) >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119}
1120
1121//
1122// BAD: new directories have stat data of new type and all other items
1123// of old type. Version stored in the inode says about body items, so
1124// in update_stat_data we can not rely on inode, but have to check
1125// item version directly
1126//
1127
1128// called by read_locked_inode
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001129static void init_inode(struct inode *inode, struct treepath *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001131 struct buffer_head *bh;
1132 struct item_head *ih;
1133 __u32 rdev;
1134 //int version = ITEM_VERSION_1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001136 bh = PATH_PLAST_BUFFER(path);
1137 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001139 copy_key(INODE_PKEY(inode), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001141 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1142 REISERFS_I(inode)->i_flags = 0;
1143 REISERFS_I(inode)->i_prealloc_block = 0;
1144 REISERFS_I(inode)->i_prealloc_count = 0;
1145 REISERFS_I(inode)->i_trans_id = 0;
1146 REISERFS_I(inode)->i_jl = NULL;
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001147 reiserfs_init_xattr_rwsem(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001149 if (stat_data_v1(ih)) {
1150 struct stat_data_v1 *sd =
1151 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1152 unsigned long blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001154 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1155 set_inode_sd_version(inode, STAT_DATA_V1);
1156 inode->i_mode = sd_v1_mode(sd);
1157 inode->i_nlink = sd_v1_nlink(sd);
1158 inode->i_uid = sd_v1_uid(sd);
1159 inode->i_gid = sd_v1_gid(sd);
1160 inode->i_size = sd_v1_size(sd);
1161 inode->i_atime.tv_sec = sd_v1_atime(sd);
1162 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1163 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1164 inode->i_atime.tv_nsec = 0;
1165 inode->i_ctime.tv_nsec = 0;
1166 inode->i_mtime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001168 inode->i_blocks = sd_v1_blocks(sd);
1169 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1170 blocks = (inode->i_size + 511) >> 9;
1171 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1172 if (inode->i_blocks > blocks) {
1173 // there was a bug in <=3.5.23 when i_blocks could take negative
1174 // values. Starting from 3.5.17 this value could even be stored in
1175 // stat data. For such files we set i_blocks based on file
1176 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1177 // only updated if file's inode will ever change
1178 inode->i_blocks = blocks;
1179 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001181 rdev = sd_v1_rdev(sd);
1182 REISERFS_I(inode)->i_first_direct_byte =
1183 sd_v1_first_direct_byte(sd);
1184 /* an early bug in the quota code can give us an odd number for the
1185 ** block count. This is incorrect, fix it here.
1186 */
1187 if (inode->i_blocks & 1) {
1188 inode->i_blocks++;
1189 }
1190 inode_set_bytes(inode,
1191 to_real_used_space(inode, inode->i_blocks,
1192 SD_V1_SIZE));
1193 /* nopack is initially zero for v1 objects. For v2 objects,
1194 nopack is initialised from sd_attrs */
1195 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1196 } else {
1197 // new stat data found, but object may have old items
1198 // (directories and symlinks)
1199 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1200
1201 inode->i_mode = sd_v2_mode(sd);
1202 inode->i_nlink = sd_v2_nlink(sd);
1203 inode->i_uid = sd_v2_uid(sd);
1204 inode->i_size = sd_v2_size(sd);
1205 inode->i_gid = sd_v2_gid(sd);
1206 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1207 inode->i_atime.tv_sec = sd_v2_atime(sd);
1208 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1209 inode->i_ctime.tv_nsec = 0;
1210 inode->i_mtime.tv_nsec = 0;
1211 inode->i_atime.tv_nsec = 0;
1212 inode->i_blocks = sd_v2_blocks(sd);
1213 rdev = sd_v2_rdev(sd);
1214 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1215 inode->i_generation =
1216 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1217 else
1218 inode->i_generation = sd_v2_generation(sd);
1219
1220 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1221 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1222 else
1223 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1224 REISERFS_I(inode)->i_first_direct_byte = 0;
1225 set_inode_sd_version(inode, STAT_DATA_V2);
1226 inode_set_bytes(inode,
1227 to_real_used_space(inode, inode->i_blocks,
1228 SD_V2_SIZE));
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02001229 /* read persistent inode attributes from sd and initialise
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001230 generic inode flags from them */
1231 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1232 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 }
1234
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001235 pathrelse(path);
1236 if (S_ISREG(inode->i_mode)) {
1237 inode->i_op = &reiserfs_file_inode_operations;
1238 inode->i_fop = &reiserfs_file_operations;
1239 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1240 } else if (S_ISDIR(inode->i_mode)) {
1241 inode->i_op = &reiserfs_dir_inode_operations;
1242 inode->i_fop = &reiserfs_dir_operations;
1243 } else if (S_ISLNK(inode->i_mode)) {
1244 inode->i_op = &reiserfs_symlink_inode_operations;
1245 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1246 } else {
1247 inode->i_blocks = 0;
1248 inode->i_op = &reiserfs_special_inode_operations;
1249 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251}
1252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253// update new stat data with inode fields
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001254static void inode2sd(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001256 struct stat_data *sd_v2 = (struct stat_data *)sd;
1257 __u16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001259 set_sd_v2_mode(sd_v2, inode->i_mode);
1260 set_sd_v2_nlink(sd_v2, inode->i_nlink);
1261 set_sd_v2_uid(sd_v2, inode->i_uid);
1262 set_sd_v2_size(sd_v2, size);
1263 set_sd_v2_gid(sd_v2, inode->i_gid);
1264 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1265 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1266 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1267 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1268 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1269 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1270 else
1271 set_sd_v2_generation(sd_v2, inode->i_generation);
1272 flags = REISERFS_I(inode)->i_attrs;
1273 i_attrs_to_sd_attrs(inode, &flags);
1274 set_sd_v2_attrs(sd_v2, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275}
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277// used to copy inode's fields to old stat data
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001278static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001280 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001282 set_sd_v1_mode(sd_v1, inode->i_mode);
1283 set_sd_v1_uid(sd_v1, inode->i_uid);
1284 set_sd_v1_gid(sd_v1, inode->i_gid);
1285 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1286 set_sd_v1_size(sd_v1, size);
1287 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1288 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1289 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001291 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1292 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1293 else
1294 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001296 // Sigh. i_first_direct_byte is back
1297 set_sd_v1_first_direct_byte(sd_v1,
1298 REISERFS_I(inode)->i_first_direct_byte);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299}
1300
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301/* NOTE, you must prepare the buffer head before sending it here,
1302** and then log it after the call
1303*/
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001304static void update_stat_data(struct treepath *path, struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001305 loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001307 struct buffer_head *bh;
1308 struct item_head *ih;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001310 bh = PATH_PLAST_BUFFER(path);
1311 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001313 if (!is_statdata_le_ih(ih))
Jeff Mahoneyc3a9c212009-03-30 14:02:25 -04001314 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001315 INODE_PKEY(inode), ih);
1316
1317 if (stat_data_v1(ih)) {
1318 // path points to old stat data
1319 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1320 } else {
1321 inode2sd(B_I_PITEM(bh, ih), inode, size);
1322 }
1323
1324 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325}
1326
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001327void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1328 struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001330 struct cpu_key key;
1331 INITIALIZE_PATH(path);
1332 struct buffer_head *bh;
1333 int fs_gen;
1334 struct item_head *ih, tmp_ih;
1335 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001337 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001339 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1340
1341 for (;;) {
1342 int pos;
1343 /* look for the object's stat data */
1344 retval = search_item(inode->i_sb, &key, &path);
1345 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001346 reiserfs_error(inode->i_sb, "vs-13050",
1347 "i/o failure occurred trying to "
1348 "update %K stat data", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001349 return;
1350 }
1351 if (retval == ITEM_NOT_FOUND) {
1352 pos = PATH_LAST_POSITION(&path);
1353 pathrelse(&path);
1354 if (inode->i_nlink == 0) {
1355 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1356 return;
1357 }
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001358 reiserfs_warning(inode->i_sb, "vs-13060",
1359 "stat data of object %k (nlink == %d) "
1360 "not found (pos %d)",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001361 INODE_PKEY(inode), inode->i_nlink,
1362 pos);
1363 reiserfs_check_path(&path);
1364 return;
1365 }
1366
1367 /* sigh, prepare_for_journal might schedule. When it schedules the
1368 ** FS might change. We have to detect that, and loop back to the
1369 ** search if the stat data item has moved
1370 */
1371 bh = get_last_bh(&path);
1372 ih = get_ih(&path);
1373 copy_item_head(&tmp_ih, ih);
1374 fs_gen = get_generation(inode->i_sb);
1375 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1376 if (fs_changed(fs_gen, inode->i_sb)
1377 && item_moved(&tmp_ih, &path)) {
1378 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1379 continue; /* Stat_data item has been moved after scheduling. */
1380 }
1381 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001383 update_stat_data(&path, inode, size);
1384 journal_mark_dirty(th, th->t_super, bh);
1385 pathrelse(&path);
1386 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
1389/* reiserfs_read_locked_inode is called to read the inode off disk, and it
1390** does a make_bad_inode when things go wrong. But, we need to make sure
1391** and clear the key in the private portion of the inode, otherwise a
1392** corresponding iput might try to delete whatever object the inode last
1393** represented.
1394*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001395static void reiserfs_make_bad_inode(struct inode *inode)
1396{
1397 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1398 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399}
1400
1401//
1402// initially this function was derived from minix or ext2's analog and
1403// evolved as the prototype did
1404//
1405
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001406int reiserfs_init_locked_inode(struct inode *inode, void *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001408 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1409 inode->i_ino = args->objectid;
1410 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1411 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412}
1413
1414/* looks for stat data in the tree, and fills up the fields of in-core
1415 inode stat data fields */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001416void reiserfs_read_locked_inode(struct inode *inode,
1417 struct reiserfs_iget_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001419 INITIALIZE_PATH(path_to_sd);
1420 struct cpu_key key;
1421 unsigned long dirino;
1422 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001424 dirino = args->dirid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001426 /* set version 1, version 2 could be used too, because stat data
1427 key is the same in both versions */
1428 key.version = KEY_FORMAT_3_5;
1429 key.on_disk_key.k_dir_id = dirino;
1430 key.on_disk_key.k_objectid = inode->i_ino;
1431 key.on_disk_key.k_offset = 0;
1432 key.on_disk_key.k_type = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001434 /* look for the object's stat data */
1435 retval = search_item(inode->i_sb, &key, &path_to_sd);
1436 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001437 reiserfs_error(inode->i_sb, "vs-13070",
1438 "i/o failure occurred trying to find "
1439 "stat data of %K", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001440 reiserfs_make_bad_inode(inode);
1441 return;
1442 }
1443 if (retval != ITEM_FOUND) {
1444 /* a stale NFS handle can trigger this without it being an error */
1445 pathrelse(&path_to_sd);
1446 reiserfs_make_bad_inode(inode);
1447 inode->i_nlink = 0;
1448 return;
1449 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001451 init_inode(inode, &path_to_sd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001453 /* It is possible that knfsd is trying to access inode of a file
1454 that is being removed from the disk by some other thread. As we
1455 update sd on unlink all that is required is to check for nlink
1456 here. This bug was first found by Sizif when debugging
1457 SquidNG/Butterfly, forgotten, and found again after Philippe
Jeff Mahoney0222e652009-03-30 14:02:44 -04001458 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001460 More logical fix would require changes in fs/inode.c:iput() to
1461 remove inode from hash-table _after_ fs cleaned disk stuff up and
1462 in iget() to return NULL if I_FREEING inode is found in
1463 hash-table. */
1464 /* Currently there is one place where it's ok to meet inode with
1465 nlink==0: processing of open-unlinked and half-truncated files
1466 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1467 if ((inode->i_nlink == 0) &&
1468 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001469 reiserfs_warning(inode->i_sb, "vs-13075",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001470 "dead inode read from disk %K. "
1471 "This is likely to be race with knfsd. Ignore",
1472 &key);
1473 reiserfs_make_bad_inode(inode);
1474 }
1475
1476 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
1478}
1479
1480/**
1481 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1482 *
1483 * @inode: inode from hash table to check
1484 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1485 *
1486 * This function is called by iget5_locked() to distinguish reiserfs inodes
1487 * having the same inode numbers. Such inodes can only exist due to some
1488 * error condition. One of them should be bad. Inodes with identical
1489 * inode numbers (objectids) are distinguished by parent directory ids.
1490 *
1491 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001492int reiserfs_find_actor(struct inode *inode, void *opaque)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001494 struct reiserfs_iget_args *args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001496 args = opaque;
1497 /* args is already in CPU order */
1498 return (inode->i_ino == args->objectid) &&
1499 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500}
1501
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001502struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001504 struct inode *inode;
1505 struct reiserfs_iget_args args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001507 args.objectid = key->on_disk_key.k_objectid;
1508 args.dirid = key->on_disk_key.k_dir_id;
Frederic Weisbecker175359f2010-02-11 13:13:10 +01001509 reiserfs_write_unlock(s);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001510 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1511 reiserfs_find_actor, reiserfs_init_locked_inode,
1512 (void *)(&args));
Frederic Weisbecker175359f2010-02-11 13:13:10 +01001513 reiserfs_write_lock(s);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001514 if (!inode)
1515 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001517 if (inode->i_state & I_NEW) {
1518 reiserfs_read_locked_inode(inode, &args);
1519 unlock_new_inode(inode);
1520 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001522 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1523 /* either due to i/o error or a stale NFS handle */
1524 iput(inode);
1525 inode = NULL;
1526 }
1527 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528}
1529
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001530static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1531 u32 objectid, u32 dir_id, u32 generation)
1532
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001534 struct cpu_key key;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001535 struct inode *inode;
1536
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001537 key.on_disk_key.k_objectid = objectid;
1538 key.on_disk_key.k_dir_id = dir_id;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001539 reiserfs_write_lock(sb);
1540 inode = reiserfs_iget(sb, &key);
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001541 if (inode && !IS_ERR(inode) && generation != 0 &&
1542 generation != inode->i_generation) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001543 iput(inode);
1544 inode = NULL;
1545 }
1546 reiserfs_write_unlock(sb);
Christoph Hellwig44003722008-08-11 15:49:04 +02001547
1548 return d_obtain_alias(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549}
1550
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001551struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1552 int fh_len, int fh_type)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001553{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001554 /* fhtype happens to reflect the number of u32s encoded.
1555 * due to a bug in earlier code, fhtype might indicate there
1556 * are more u32s then actually fitted.
1557 * so if fhtype seems to be more than len, reduce fhtype.
1558 * Valid types are:
1559 * 2 - objectid + dir_id - legacy support
1560 * 3 - objectid + dir_id + generation
1561 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1562 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1563 * 6 - as above plus generation of directory
1564 * 6 does not fit in NFSv2 handles
1565 */
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001566 if (fh_type > fh_len) {
1567 if (fh_type != 6 || fh_len != 5)
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001568 reiserfs_warning(sb, "reiserfs-13077",
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001569 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1570 fh_type, fh_len);
1571 fh_type = 5;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001572 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001574 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1575 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1576}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
Christoph Hellwigbe55caf2007-10-21 16:42:13 -07001578struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1579 int fh_len, int fh_type)
1580{
1581 if (fh_type < 4)
1582 return NULL;
1583
1584 return reiserfs_get_dentry(sb,
1585 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1586 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1587 (fh_type == 6) ? fid->raw[5] : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588}
1589
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001590int reiserfs_encode_fh(struct dentry *dentry, __u32 * data, int *lenp,
1591 int need_parent)
1592{
1593 struct inode *inode = dentry->d_inode;
1594 int maxlen = *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595
Aneesh Kumar K.V5fe0c232011-01-29 18:43:25 +05301596 if (need_parent && (maxlen < 5)) {
1597 *lenp = 5;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001598 return 255;
Aneesh Kumar K.V5fe0c232011-01-29 18:43:25 +05301599 } else if (maxlen < 3) {
1600 *lenp = 3;
1601 return 255;
1602 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001604 data[0] = inode->i_ino;
1605 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1606 data[2] = inode->i_generation;
1607 *lenp = 3;
1608 /* no room for directory info? return what we've stored so far */
1609 if (maxlen < 5 || !need_parent)
1610 return 3;
1611
1612 spin_lock(&dentry->d_lock);
1613 inode = dentry->d_parent->d_inode;
1614 data[3] = inode->i_ino;
1615 data[4] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1616 *lenp = 5;
1617 if (maxlen >= 6) {
1618 data[5] = inode->i_generation;
1619 *lenp = 6;
1620 }
1621 spin_unlock(&dentry->d_lock);
1622 return *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623}
1624
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625/* looks for stat data, then copies fields to it, marks the buffer
1626 containing stat data as dirty */
1627/* reiserfs inodes are never really dirty, since the dirty inode call
1628** always logs them. This call allows the VFS inode marking routines
1629** to properly mark inodes for datasync and such, but only actually
1630** does something when called for a synchronous update.
1631*/
Christoph Hellwiga9185b42010-03-05 09:21:37 +01001632int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001633{
1634 struct reiserfs_transaction_handle th;
1635 int jbegin_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001637 if (inode->i_sb->s_flags & MS_RDONLY)
1638 return -EROFS;
1639 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001640 ** these cases are just when the system needs ram, not when the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001641 ** inode needs to reach disk for safety, and they can safely be
1642 ** ignored because the altered inode has already been logged.
1643 */
Christoph Hellwiga9185b42010-03-05 09:21:37 +01001644 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001645 reiserfs_write_lock(inode->i_sb);
1646 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1647 reiserfs_update_sd(&th, inode);
1648 journal_end_sync(&th, inode->i_sb, jbegin_count);
1649 }
1650 reiserfs_write_unlock(inode->i_sb);
1651 }
1652 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653}
1654
1655/* stat data of new object is inserted already, this inserts the item
1656 containing "." and ".." entries */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001657static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1658 struct inode *inode,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001659 struct item_head *ih, struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001660 struct inode *dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001662 struct super_block *sb = th->t_super;
1663 char empty_dir[EMPTY_DIR_SIZE];
1664 char *body = empty_dir;
1665 struct cpu_key key;
1666 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001668 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001670 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1671 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1672 TYPE_DIRENTRY, 3 /*key length */ );
1673
1674 /* compose item head for new item. Directories consist of items of
1675 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1676 is done by reiserfs_new_inode */
1677 if (old_format_only(sb)) {
1678 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1679 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1680
1681 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1682 ih->ih_key.k_objectid,
1683 INODE_PKEY(dir)->k_dir_id,
1684 INODE_PKEY(dir)->k_objectid);
1685 } else {
1686 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1687 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1688
1689 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1690 ih->ih_key.k_objectid,
1691 INODE_PKEY(dir)->k_dir_id,
1692 INODE_PKEY(dir)->k_objectid);
1693 }
1694
1695 /* look for place in the tree for new item */
1696 retval = search_item(sb, &key, path);
1697 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001698 reiserfs_error(sb, "vs-13080",
1699 "i/o failure occurred creating new directory");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001700 return -EIO;
1701 }
1702 if (retval == ITEM_FOUND) {
1703 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001704 reiserfs_warning(sb, "vs-13070",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001705 "object with this key exists (%k)",
1706 &(ih->ih_key));
1707 return -EEXIST;
1708 }
1709
1710 /* insert item, that is empty directory item */
1711 return reiserfs_insert_item(th, path, &key, ih, inode, body);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712}
1713
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714/* stat data of object has been inserted, this inserts the item
1715 containing the body of symlink */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001716static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1717 struct item_head *ih,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001718 struct treepath *path, const char *symname,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001719 int item_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001721 struct super_block *sb = th->t_super;
1722 struct cpu_key key;
1723 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001725 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001727 _make_cpu_key(&key, KEY_FORMAT_3_5,
1728 le32_to_cpu(ih->ih_key.k_dir_id),
1729 le32_to_cpu(ih->ih_key.k_objectid),
1730 1, TYPE_DIRECT, 3 /*key length */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001732 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1733 0 /*free_space */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001735 /* look for place in the tree for new item */
1736 retval = search_item(sb, &key, path);
1737 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001738 reiserfs_error(sb, "vs-13080",
1739 "i/o failure occurred creating new symlink");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001740 return -EIO;
1741 }
1742 if (retval == ITEM_FOUND) {
1743 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001744 reiserfs_warning(sb, "vs-13080",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001745 "object with this key exists (%k)",
1746 &(ih->ih_key));
1747 return -EEXIST;
1748 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001750 /* insert item, that is body of symlink */
1751 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752}
1753
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754/* inserts the stat data into the tree, and then calls
1755 reiserfs_new_directory (to insert ".", ".." item if new object is
1756 directory) or reiserfs_new_symlink (to insert symlink body if new
Jeff Mahoney0222e652009-03-30 14:02:44 -04001757 object is symlink) or nothing (if new object is regular file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
1759 NOTE! uid and gid must already be set in the inode. If we return
1760 non-zero due to an error, we have to drop the quota previously allocated
1761 for the fresh inode. This can only be done outside a transaction, so
1762 if we return non-zero, we also end the transaction. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001763int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1764 struct inode *dir, int mode, const char *symname,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001765 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001766 strlen (symname) for symlinks) */
1767 loff_t i_size, struct dentry *dentry,
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001768 struct inode *inode,
1769 struct reiserfs_security_handle *security)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001771 struct super_block *sb;
Al Viroc1eaa262008-12-30 02:03:58 -05001772 struct reiserfs_iget_args args;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001773 INITIALIZE_PATH(path_to_key);
1774 struct cpu_key key;
1775 struct item_head ih;
1776 struct stat_data sd;
1777 int retval;
1778 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001780 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781
Christoph Hellwig871a2932010-03-03 09:05:07 -05001782 dquot_initialize(inode);
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001783 err = dquot_alloc_inode(inode);
1784 if (err)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001785 goto out_end_trans;
Eric Sesterhenn585b7742006-10-04 02:15:30 -07001786 if (!dir->i_nlink) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001787 err = -EPERM;
1788 goto out_bad_inode;
1789 }
1790
1791 sb = dir->i_sb;
1792
1793 /* item head of new item */
1794 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1795 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1796 if (!ih.ih_key.k_objectid) {
1797 err = -ENOMEM;
1798 goto out_bad_inode;
1799 }
Al Viroc1eaa262008-12-30 02:03:58 -05001800 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
Al Viro2f1169e2009-01-02 08:16:51 -05001801 if (old_format_only(sb))
1802 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1803 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1804 else
1805 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1806 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
Al Viroc1eaa262008-12-30 02:03:58 -05001807 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1808 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1809 if (insert_inode_locked4(inode, args.objectid,
1810 reiserfs_find_actor, &args) < 0) {
1811 err = -EINVAL;
1812 goto out_bad_inode;
1813 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001814 if (old_format_only(sb))
Jeff Mahoney0222e652009-03-30 14:02:44 -04001815 /* not a perfect generation count, as object ids can be reused, but
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001816 ** this is as good as reiserfs can do right now.
1817 ** note that the private part of inode isn't filled in yet, we have
1818 ** to use the directory.
1819 */
1820 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1821 else
1822#if defined( USE_INODE_GENERATION_COUNTER )
1823 inode->i_generation =
1824 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1825#else
1826 inode->i_generation = ++event;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001829 /* fill stat data */
1830 inode->i_nlink = (S_ISDIR(mode) ? 2 : 1);
1831
1832 /* uid and gid must already be set by the caller for quota init */
1833
1834 /* symlink cannot be immutable or append only, right? */
1835 if (S_ISLNK(inode->i_mode))
1836 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1837
1838 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1839 inode->i_size = i_size;
1840 inode->i_blocks = 0;
1841 inode->i_bytes = 0;
1842 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1843 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1844
1845 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1846 REISERFS_I(inode)->i_flags = 0;
1847 REISERFS_I(inode)->i_prealloc_block = 0;
1848 REISERFS_I(inode)->i_prealloc_count = 0;
1849 REISERFS_I(inode)->i_trans_id = 0;
1850 REISERFS_I(inode)->i_jl = NULL;
1851 REISERFS_I(inode)->i_attrs =
1852 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1853 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001854 reiserfs_init_xattr_rwsem(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001855
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001856 /* key to search for correct place for new stat data */
1857 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1858 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1859 TYPE_STAT_DATA, 3 /*key length */ );
1860
1861 /* find proper place for inserting of stat data */
1862 retval = search_item(sb, &key, &path_to_key);
1863 if (retval == IO_ERROR) {
1864 err = -EIO;
1865 goto out_bad_inode;
1866 }
1867 if (retval == ITEM_FOUND) {
1868 pathrelse(&path_to_key);
1869 err = -EEXIST;
1870 goto out_bad_inode;
1871 }
1872 if (old_format_only(sb)) {
1873 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1874 pathrelse(&path_to_key);
1875 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1876 err = -EINVAL;
1877 goto out_bad_inode;
1878 }
1879 inode2sd_v1(&sd, inode, inode->i_size);
1880 } else {
1881 inode2sd(&sd, inode, inode->i_size);
1882 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001883 // store in in-core inode the key of stat data and version all
1884 // object items will have (directory items will have old offset
1885 // format, other new objects will consist of new items)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001886 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1887 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1888 else
1889 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1890 if (old_format_only(sb))
1891 set_inode_sd_version(inode, STAT_DATA_V1);
1892 else
1893 set_inode_sd_version(inode, STAT_DATA_V2);
1894
1895 /* insert the stat data into the tree */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001897 if (REISERFS_I(dir)->new_packing_locality)
1898 th->displace_new_blocks = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001900 retval =
1901 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1902 (char *)(&sd));
1903 if (retval) {
1904 err = retval;
1905 reiserfs_check_path(&path_to_key);
1906 goto out_bad_inode;
1907 }
1908#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1909 if (!th->displace_new_blocks)
1910 REISERFS_I(dir)->new_packing_locality = 0;
1911#endif
1912 if (S_ISDIR(mode)) {
1913 /* insert item with "." and ".." */
1914 retval =
1915 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1916 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001918 if (S_ISLNK(mode)) {
1919 /* insert body of symlink */
1920 if (!old_format_only(sb))
1921 i_size = ROUND_UP(i_size);
1922 retval =
1923 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1924 i_size);
1925 }
1926 if (retval) {
1927 err = retval;
1928 reiserfs_check_path(&path_to_key);
1929 journal_end(th, th->t_super, th->t_blocks_allocated);
1930 goto out_inserted_sd;
1931 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001933 if (reiserfs_posixacl(inode->i_sb)) {
Jeff Mahoney0ab26212009-03-30 14:02:39 -04001934 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001935 if (retval) {
1936 err = retval;
1937 reiserfs_check_path(&path_to_key);
1938 journal_end(th, th->t_super, th->t_blocks_allocated);
1939 goto out_inserted_sd;
1940 }
1941 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001942 reiserfs_warning(inode->i_sb, "jdm-13090",
1943 "ACLs aren't enabled in the fs, "
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001944 "but vfs thinks they are!");
Jeff Mahoney6dfede692009-03-30 14:02:32 -04001945 } else if (IS_PRIVATE(dir))
1946 inode->i_flags |= S_PRIVATE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001948 if (security->name) {
1949 retval = reiserfs_security_write(th, inode, security);
1950 if (retval) {
1951 err = retval;
1952 reiserfs_check_path(&path_to_key);
1953 retval = journal_end(th, th->t_super,
1954 th->t_blocks_allocated);
1955 if (retval)
1956 err = retval;
1957 goto out_inserted_sd;
1958 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 }
1960
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001961 reiserfs_update_sd(th, inode);
1962 reiserfs_check_path(&path_to_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001964 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965
1966/* it looks like you can easily compress these two goto targets into
1967 * one. Keeping it like this doesn't actually hurt anything, and they
1968 * are place holders for what the quota code actually needs.
1969 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001970 out_bad_inode:
1971 /* Invalidate the object, nothing was inserted yet */
1972 INODE_PKEY(inode)->k_objectid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001974 /* Quota change must be inside a transaction for journaling */
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001975 dquot_free_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001977 out_end_trans:
1978 journal_end(th, th->t_super, th->t_blocks_allocated);
1979 /* Drop can be outside and it needs more credits so it's better to have it outside */
Christoph Hellwig9f754752010-03-03 09:05:05 -05001980 dquot_drop(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001981 inode->i_flags |= S_NOQUOTA;
1982 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001984 out_inserted_sd:
1985 inode->i_nlink = 0;
1986 th->t_trans_id = 0; /* so the caller can't use this handle later */
Al Viroc1eaa262008-12-30 02:03:58 -05001987 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
Jeff Mahoneyd9845612009-03-30 14:02:35 -04001988 iput(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001989 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990}
1991
1992/*
1993** finds the tail page in the page cache,
1994** reads the last block in.
1995**
1996** On success, page_result is set to a locked, pinned page, and bh_result
1997** is set to an up to date buffer for the last block in the file. returns 0.
1998**
1999** tail conversion is not done, so bh_result might not be valid for writing
2000** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2001** trying to write the block.
2002**
2003** on failure, nonzero is returned, page_result and bh_result are untouched.
2004*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04002005static int grab_tail_page(struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002006 struct page **page_result,
2007 struct buffer_head **bh_result)
2008{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002010 /* we want the page with the last byte in the file,
2011 ** not the page that will hold the next byte for appending
2012 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002013 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002014 unsigned long pos = 0;
2015 unsigned long start = 0;
Jeff Mahoney995c7622009-03-30 14:02:47 -04002016 unsigned long blocksize = inode->i_sb->s_blocksize;
2017 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002018 struct buffer_head *bh;
2019 struct buffer_head *head;
2020 struct page *page;
2021 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002023 /* we know that we are only called with inode->i_size > 0.
2024 ** we also know that a file tail can never be as big as a block
2025 ** If i_size % blocksize == 0, our file is currently block aligned
2026 ** and it won't need converting or zeroing after a truncate.
2027 */
2028 if ((offset & (blocksize - 1)) == 0) {
2029 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 }
Jeff Mahoney995c7622009-03-30 14:02:47 -04002031 page = grab_cache_page(inode->i_mapping, index);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002032 error = -ENOMEM;
2033 if (!page) {
2034 goto out;
2035 }
2036 /* start within the page of the last block in the file */
2037 start = (offset / blocksize) * blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
Christoph Hellwigebdec242010-10-06 10:47:23 +02002039 error = __block_write_begin(page, start, offset - start,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002040 reiserfs_get_block_create_0);
2041 if (error)
2042 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002044 head = page_buffers(page);
2045 bh = head;
2046 do {
2047 if (pos >= start) {
2048 break;
2049 }
2050 bh = bh->b_this_page;
2051 pos += blocksize;
2052 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002054 if (!buffer_uptodate(bh)) {
2055 /* note, this should never happen, prepare_write should
2056 ** be taking care of this for us. If the buffer isn't up to date,
2057 ** I've screwed up the code to find the buffer, or the code to
2058 ** call prepare_write
2059 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002060 reiserfs_error(inode->i_sb, "clm-6000",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002061 "error reading block %lu", bh->b_blocknr);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002062 error = -EIO;
2063 goto unlock;
2064 }
2065 *bh_result = bh;
2066 *page_result = page;
2067
2068 out:
2069 return error;
2070
2071 unlock:
2072 unlock_page(page);
2073 page_cache_release(page);
2074 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075}
2076
2077/*
2078** vfs version of truncate file. Must NOT be called with
2079** a transaction already started.
2080**
2081** some code taken from block_truncate_page
2082*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04002083int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002084{
2085 struct reiserfs_transaction_handle th;
2086 /* we want the offset for the first byte after the end of the file */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002087 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2088 unsigned blocksize = inode->i_sb->s_blocksize;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002089 unsigned length;
2090 struct page *page = NULL;
2091 int error;
2092 struct buffer_head *bh = NULL;
Jeff Mahoney24996042005-12-14 14:38:05 -05002093 int err2;
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002094 int lock_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002096 lock_depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
Jeff Mahoney995c7622009-03-30 14:02:47 -04002098 if (inode->i_size > 0) {
2099 error = grab_tail_page(inode, &page, &bh);
2100 if (error) {
Jeff Mahoney0222e652009-03-30 14:02:44 -04002101 // -ENOENT means we truncated past the end of the file,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002102 // and get_block_create_0 could not find a block to read in,
2103 // which is ok.
2104 if (error != -ENOENT)
Jeff Mahoney995c7622009-03-30 14:02:47 -04002105 reiserfs_error(inode->i_sb, "clm-6001",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002106 "grab_tail_page failed %d",
2107 error);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002108 page = NULL;
2109 bh = NULL;
2110 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112
Jeff Mahoney0222e652009-03-30 14:02:44 -04002113 /* so, if page != NULL, we have a buffer head for the offset at
2114 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2115 ** then we have an unformatted node. Otherwise, we have a direct item,
2116 ** and no zeroing is required on disk. We zero after the truncate,
2117 ** because the truncate might pack the item anyway
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002118 ** (it will unmap bh if it packs).
2119 */
2120 /* it is enough to reserve space in transaction for 2 balancings:
2121 one for "save" link adding and another for the first
2122 cut_from_item. 1 is for update_sd */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002123 error = journal_begin(&th, inode->i_sb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002124 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2125 if (error)
2126 goto out;
Jeff Mahoney995c7622009-03-30 14:02:47 -04002127 reiserfs_update_inode_transaction(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002128 if (update_timestamps)
2129 /* we are doing real truncate: if the system crashes before the last
2130 transaction of truncating gets committed - on reboot the file
2131 either appears truncated properly or not truncated at all */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002132 add_save_link(&th, inode, 1);
2133 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002134 error =
Jeff Mahoney995c7622009-03-30 14:02:47 -04002135 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002136 if (error)
2137 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138
Jeff Mahoney24996042005-12-14 14:38:05 -05002139 /* check reiserfs_do_truncate after ending the transaction */
2140 if (err2) {
2141 error = err2;
2142 goto out;
2143 }
2144
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002145 if (update_timestamps) {
Jeff Mahoney995c7622009-03-30 14:02:47 -04002146 error = remove_save_link(inode, 1 /* truncate */);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002147 if (error)
2148 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002151 if (page) {
2152 length = offset & (blocksize - 1);
2153 /* if we are not on a block boundary */
2154 if (length) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002155 length = blocksize - length;
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002156 zero_user(page, offset, length);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002157 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2158 mark_buffer_dirty(bh);
2159 }
2160 }
2161 unlock_page(page);
2162 page_cache_release(page);
2163 }
2164
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002165 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2166
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002167 return 0;
2168 out:
2169 if (page) {
2170 unlock_page(page);
2171 page_cache_release(page);
2172 }
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002173
2174 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2175
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002176 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177}
2178
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002179static int map_block_for_writepage(struct inode *inode,
2180 struct buffer_head *bh_result,
2181 unsigned long block)
2182{
2183 struct reiserfs_transaction_handle th;
2184 int fs_gen;
2185 struct item_head tmp_ih;
2186 struct item_head *ih;
2187 struct buffer_head *bh;
2188 __le32 *item;
2189 struct cpu_key key;
2190 INITIALIZE_PATH(path);
2191 int pos_in_item;
2192 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
Oleg Drokin7729ac52005-11-28 13:43:53 -08002193 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002194 int retval;
2195 int use_get_block = 0;
2196 int bytes_copied = 0;
2197 int copy_size;
2198 int trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002200 /* catch places below that try to log something without starting a trans */
2201 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002203 if (!buffer_uptodate(bh_result)) {
2204 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 }
2206
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002207 kmap(bh_result->b_page);
2208 start_over:
2209 reiserfs_write_lock(inode->i_sb);
2210 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002212 research:
2213 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2214 if (retval != POSITION_FOUND) {
2215 use_get_block = 1;
2216 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 }
2218
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002219 bh = get_last_bh(&path);
2220 ih = get_ih(&path);
2221 item = get_item(&path);
2222 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002224 /* we've found an unformatted node */
2225 if (indirect_item_found(retval, ih)) {
2226 if (bytes_copied > 0) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002227 reiserfs_warning(inode->i_sb, "clm-6002",
2228 "bytes_copied %d", bytes_copied);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002229 }
2230 if (!get_block_num(item, pos_in_item)) {
2231 /* crap, we are writing to a hole */
2232 use_get_block = 1;
2233 goto out;
2234 }
2235 set_block_dev_mapped(bh_result,
2236 get_block_num(item, pos_in_item), inode);
2237 } else if (is_direct_le_ih(ih)) {
2238 char *p;
2239 p = page_address(bh_result->b_page);
2240 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2241 copy_size = ih_item_len(ih) - pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002243 fs_gen = get_generation(inode->i_sb);
2244 copy_item_head(&tmp_ih, ih);
2245
2246 if (!trans_running) {
2247 /* vs-3050 is gone, no need to drop the path */
2248 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2249 if (retval)
2250 goto out;
2251 reiserfs_update_inode_transaction(inode);
2252 trans_running = 1;
2253 if (fs_changed(fs_gen, inode->i_sb)
2254 && item_moved(&tmp_ih, &path)) {
2255 reiserfs_restore_prepared_buffer(inode->i_sb,
2256 bh);
2257 goto research;
2258 }
2259 }
2260
2261 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2262
2263 if (fs_changed(fs_gen, inode->i_sb)
2264 && item_moved(&tmp_ih, &path)) {
2265 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2266 goto research;
2267 }
2268
2269 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2270 copy_size);
2271
2272 journal_mark_dirty(&th, inode->i_sb, bh);
2273 bytes_copied += copy_size;
2274 set_block_dev_mapped(bh_result, 0, inode);
2275
2276 /* are there still bytes left? */
2277 if (bytes_copied < bh_result->b_size &&
2278 (byte_offset + bytes_copied) < inode->i_size) {
2279 set_cpu_key_k_offset(&key,
2280 cpu_key_k_offset(&key) +
2281 copy_size);
2282 goto research;
2283 }
2284 } else {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002285 reiserfs_warning(inode->i_sb, "clm-6003",
2286 "bad item inode %lu", inode->i_ino);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002287 retval = -EIO;
2288 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002290 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002292 out:
2293 pathrelse(&path);
2294 if (trans_running) {
2295 int err = journal_end(&th, inode->i_sb, jbegin_count);
2296 if (err)
2297 retval = err;
2298 trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002300 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002302 /* this is where we fill in holes in the file. */
2303 if (use_get_block) {
2304 retval = reiserfs_get_block(inode, block, bh_result,
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002305 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002306 | GET_BLOCK_NO_DANGLE);
2307 if (!retval) {
2308 if (!buffer_mapped(bh_result)
2309 || bh_result->b_blocknr == 0) {
2310 /* get_block failed to find a mapped unformatted node. */
2311 use_get_block = 0;
2312 goto start_over;
2313 }
2314 }
2315 }
2316 kunmap(bh_result->b_page);
2317
2318 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2319 /* we've copied data from the page into the direct item, so the
2320 * buffer in the page is now clean, mark it to reflect that.
2321 */
2322 lock_buffer(bh_result);
2323 clear_buffer_dirty(bh_result);
2324 unlock_buffer(bh_result);
2325 }
2326 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327}
2328
Jeff Mahoney0222e652009-03-30 14:02:44 -04002329/*
2330 * mason@suse.com: updated in 2.5.54 to follow the same general io
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 * start/recovery path as __block_write_full_page, along with special
2332 * code to handle reiserfs tails.
2333 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002334static int reiserfs_write_full_page(struct page *page,
2335 struct writeback_control *wbc)
2336{
2337 struct inode *inode = page->mapping->host;
2338 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2339 int error = 0;
2340 unsigned long block;
Chris Masonb4c76fa2006-08-05 12:15:10 -07002341 sector_t last_block;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002342 struct buffer_head *head, *bh;
2343 int partial = 0;
2344 int nr = 0;
2345 int checked = PageChecked(page);
2346 struct reiserfs_transaction_handle th;
2347 struct super_block *s = inode->i_sb;
2348 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2349 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350
Chris Masone0e851c2006-02-01 03:06:49 -08002351 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2352 if (checked && (current->flags & PF_MEMALLOC)) {
2353 redirty_page_for_writepage(wbc, page);
2354 unlock_page(page);
2355 return 0;
2356 }
2357
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002358 /* The page dirty bit is cleared before writepage is called, which
2359 * means we have to tell create_empty_buffers to make dirty buffers
2360 * The page really should be up to date at this point, so tossing
2361 * in the BH_Uptodate is just a sanity check.
2362 */
2363 if (!page_has_buffers(page)) {
2364 create_empty_buffers(page, s->s_blocksize,
2365 (1 << BH_Dirty) | (1 << BH_Uptodate));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002367 head = page_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002369 /* last page in the file, zero out any contents past the
2370 ** last byte in the file
2371 */
2372 if (page->index >= end_index) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002373 unsigned last_offset;
2374
2375 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2376 /* no file contents in this page */
2377 if (page->index >= end_index + 1 || !last_offset) {
2378 unlock_page(page);
2379 return 0;
2380 }
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002381 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002383 bh = head;
2384 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
Chris Masonb4c76fa2006-08-05 12:15:10 -07002385 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002386 /* first map all the buffers, logging any direct items we find */
2387 do {
Chris Masonb4c76fa2006-08-05 12:15:10 -07002388 if (block > last_block) {
2389 /*
2390 * This can happen when the block size is less than
2391 * the page size. The corresponding bytes in the page
2392 * were zero filled above
2393 */
2394 clear_buffer_dirty(bh);
2395 set_buffer_uptodate(bh);
2396 } else if ((checked || buffer_dirty(bh)) &&
2397 (!buffer_mapped(bh) || (buffer_mapped(bh)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002398 && bh->b_blocknr ==
2399 0))) {
2400 /* not mapped yet, or it points to a direct item, search
2401 * the btree for the mapping info, and log any direct
2402 * items found
2403 */
2404 if ((error = map_block_for_writepage(inode, bh, block))) {
2405 goto fail;
2406 }
2407 }
2408 bh = bh->b_this_page;
2409 block++;
2410 } while (bh != head);
2411
2412 /*
2413 * we start the transaction after map_block_for_writepage,
2414 * because it can create holes in the file (an unbounded operation).
2415 * starting it here, we can make a reliable estimate for how many
2416 * blocks we're going to log
2417 */
2418 if (checked) {
2419 ClearPageChecked(page);
2420 reiserfs_write_lock(s);
2421 error = journal_begin(&th, s, bh_per_page + 1);
2422 if (error) {
2423 reiserfs_write_unlock(s);
2424 goto fail;
2425 }
2426 reiserfs_update_inode_transaction(inode);
2427 }
2428 /* now go through and lock any dirty buffers on the page */
2429 do {
2430 get_bh(bh);
2431 if (!buffer_mapped(bh))
2432 continue;
2433 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2434 continue;
2435
2436 if (checked) {
2437 reiserfs_prepare_for_journal(s, bh, 1);
2438 journal_mark_dirty(&th, s, bh);
2439 continue;
2440 }
2441 /* from this point on, we know the buffer is mapped to a
2442 * real block and not a direct item
2443 */
Wu Fengguang1b430be2010-10-26 14:21:26 -07002444 if (wbc->sync_mode != WB_SYNC_NONE) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002445 lock_buffer(bh);
2446 } else {
Nick Pigginca5de402008-08-02 12:02:13 +02002447 if (!trylock_buffer(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002448 redirty_page_for_writepage(wbc, page);
2449 continue;
2450 }
2451 }
2452 if (test_clear_buffer_dirty(bh)) {
2453 mark_buffer_async_write(bh);
2454 } else {
2455 unlock_buffer(bh);
2456 }
2457 } while ((bh = bh->b_this_page) != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458
2459 if (checked) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002460 error = journal_end(&th, s, bh_per_page + 1);
2461 reiserfs_write_unlock(s);
2462 if (error)
2463 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002465 BUG_ON(PageWriteback(page));
2466 set_page_writeback(page);
2467 unlock_page(page);
2468
2469 /*
Jeff Mahoney0222e652009-03-30 14:02:44 -04002470 * since any buffer might be the only dirty buffer on the page,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002471 * the first submit_bh can bring the page out of writeback.
2472 * be careful with the buffers.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474 do {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002475 struct buffer_head *next = bh->b_this_page;
2476 if (buffer_async_write(bh)) {
2477 submit_bh(WRITE, bh);
2478 nr++;
2479 }
2480 put_bh(bh);
2481 bh = next;
2482 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002484 error = 0;
2485 done:
2486 if (nr == 0) {
2487 /*
2488 * if this page only had a direct item, it is very possible for
Jeff Mahoney0222e652009-03-30 14:02:44 -04002489 * no io to be required without there being an error. Or,
2490 * someone else could have locked them and sent them down the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002491 * pipe without locking the page
2492 */
2493 bh = head;
2494 do {
2495 if (!buffer_uptodate(bh)) {
2496 partial = 1;
2497 break;
2498 }
2499 bh = bh->b_this_page;
2500 } while (bh != head);
2501 if (!partial)
2502 SetPageUptodate(page);
2503 end_page_writeback(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002504 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002505 return error;
2506
2507 fail:
2508 /* catches various errors, we need to make sure any valid dirty blocks
Jeff Mahoney0222e652009-03-30 14:02:44 -04002509 * get to the media. The page is currently locked and not marked for
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002510 * writeback
2511 */
2512 ClearPageUptodate(page);
2513 bh = head;
2514 do {
2515 get_bh(bh);
2516 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2517 lock_buffer(bh);
2518 mark_buffer_async_write(bh);
2519 } else {
2520 /*
2521 * clear any dirty bits that might have come from getting
2522 * attached to a dirty page
2523 */
2524 clear_buffer_dirty(bh);
2525 }
2526 bh = bh->b_this_page;
2527 } while (bh != head);
2528 SetPageError(page);
2529 BUG_ON(PageWriteback(page));
2530 set_page_writeback(page);
2531 unlock_page(page);
2532 do {
2533 struct buffer_head *next = bh->b_this_page;
2534 if (buffer_async_write(bh)) {
2535 clear_buffer_dirty(bh);
2536 submit_bh(WRITE, bh);
2537 nr++;
2538 }
2539 put_bh(bh);
2540 bh = next;
2541 } while (bh != head);
2542 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543}
2544
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002545static int reiserfs_readpage(struct file *f, struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002547 return block_read_full_page(page, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548}
2549
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002550static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002552 struct inode *inode = page->mapping->host;
2553 reiserfs_wait_on_write_block(inode->i_sb);
2554 return reiserfs_write_full_page(page, wbc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555}
2556
Jan Karaec8e2f72009-12-17 15:27:06 -08002557static void reiserfs_truncate_failed_write(struct inode *inode)
2558{
2559 truncate_inode_pages(inode->i_mapping, inode->i_size);
2560 reiserfs_truncate_file(inode, 0);
2561}
2562
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002563static int reiserfs_write_begin(struct file *file,
2564 struct address_space *mapping,
2565 loff_t pos, unsigned len, unsigned flags,
2566 struct page **pagep, void **fsdata)
2567{
2568 struct inode *inode;
2569 struct page *page;
2570 pgoff_t index;
2571 int ret;
2572 int old_ref = 0;
2573
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002574 inode = mapping->host;
2575 *fsdata = 0;
2576 if (flags & AOP_FLAG_CONT_EXPAND &&
2577 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2578 pos ++;
2579 *fsdata = (void *)(unsigned long)flags;
2580 }
2581
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002582 index = pos >> PAGE_CACHE_SHIFT;
Nick Piggin54566b22009-01-04 12:00:53 -08002583 page = grab_cache_page_write_begin(mapping, index, flags);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002584 if (!page)
2585 return -ENOMEM;
2586 *pagep = page;
2587
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002588 reiserfs_wait_on_write_block(inode->i_sb);
2589 fix_tail_page_for_writing(page);
2590 if (reiserfs_transaction_running(inode->i_sb)) {
2591 struct reiserfs_transaction_handle *th;
2592 th = (struct reiserfs_transaction_handle *)current->
2593 journal_info;
2594 BUG_ON(!th->t_refcount);
2595 BUG_ON(!th->t_trans_id);
2596 old_ref = th->t_refcount;
2597 th->t_refcount++;
2598 }
Christoph Hellwig6e1db882010-06-04 11:29:57 +02002599 ret = __block_write_begin(page, pos, len, reiserfs_get_block);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002600 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2601 struct reiserfs_transaction_handle *th = current->journal_info;
2602 /* this gets a little ugly. If reiserfs_get_block returned an
2603 * error and left a transacstion running, we've got to close it,
2604 * and we've got to free handle if it was a persistent transaction.
2605 *
2606 * But, if we had nested into an existing transaction, we need
2607 * to just drop the ref count on the handle.
2608 *
2609 * If old_ref == 0, the transaction is from reiserfs_get_block,
2610 * and it was a persistent trans. Otherwise, it was nested above.
2611 */
2612 if (th->t_refcount > old_ref) {
2613 if (old_ref)
2614 th->t_refcount--;
2615 else {
2616 int err;
2617 reiserfs_write_lock(inode->i_sb);
2618 err = reiserfs_end_persistent_transaction(th);
2619 reiserfs_write_unlock(inode->i_sb);
2620 if (err)
2621 ret = err;
2622 }
2623 }
2624 }
2625 if (ret) {
2626 unlock_page(page);
2627 page_cache_release(page);
Jan Karaec8e2f72009-12-17 15:27:06 -08002628 /* Truncate allocated blocks */
2629 reiserfs_truncate_failed_write(inode);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002630 }
2631 return ret;
2632}
2633
Christoph Hellwigebdec242010-10-06 10:47:23 +02002634int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002636 struct inode *inode = page->mapping->host;
2637 int ret;
2638 int old_ref = 0;
2639
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002640 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002641 reiserfs_wait_on_write_block(inode->i_sb);
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002642 reiserfs_write_lock(inode->i_sb);
2643
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002644 fix_tail_page_for_writing(page);
2645 if (reiserfs_transaction_running(inode->i_sb)) {
2646 struct reiserfs_transaction_handle *th;
2647 th = (struct reiserfs_transaction_handle *)current->
2648 journal_info;
2649 BUG_ON(!th->t_refcount);
2650 BUG_ON(!th->t_trans_id);
2651 old_ref = th->t_refcount;
2652 th->t_refcount++;
2653 }
2654
Christoph Hellwigebdec242010-10-06 10:47:23 +02002655 ret = __block_write_begin(page, from, len, reiserfs_get_block);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002656 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2657 struct reiserfs_transaction_handle *th = current->journal_info;
2658 /* this gets a little ugly. If reiserfs_get_block returned an
2659 * error and left a transacstion running, we've got to close it,
2660 * and we've got to free handle if it was a persistent transaction.
2661 *
2662 * But, if we had nested into an existing transaction, we need
2663 * to just drop the ref count on the handle.
2664 *
2665 * If old_ref == 0, the transaction is from reiserfs_get_block,
2666 * and it was a persistent trans. Otherwise, it was nested above.
2667 */
2668 if (th->t_refcount > old_ref) {
2669 if (old_ref)
2670 th->t_refcount--;
2671 else {
2672 int err;
2673 reiserfs_write_lock(inode->i_sb);
2674 err = reiserfs_end_persistent_transaction(th);
2675 reiserfs_write_unlock(inode->i_sb);
2676 if (err)
2677 ret = err;
2678 }
2679 }
2680 }
2681 return ret;
2682
2683}
2684
2685static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2686{
2687 return generic_block_bmap(as, block, reiserfs_bmap);
2688}
2689
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002690static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2691 loff_t pos, unsigned len, unsigned copied,
2692 struct page *page, void *fsdata)
2693{
2694 struct inode *inode = page->mapping->host;
2695 int ret = 0;
2696 int update_sd = 0;
2697 struct reiserfs_transaction_handle *th;
2698 unsigned start;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002699 int lock_depth = 0;
2700 bool locked = false;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002701
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002702 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2703 pos ++;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002704
2705 reiserfs_wait_on_write_block(inode->i_sb);
2706 if (reiserfs_transaction_running(inode->i_sb))
2707 th = current->journal_info;
2708 else
2709 th = NULL;
2710
2711 start = pos & (PAGE_CACHE_SIZE - 1);
2712 if (unlikely(copied < len)) {
2713 if (!PageUptodate(page))
2714 copied = 0;
2715
2716 page_zero_new_buffers(page, start + copied, start + len);
2717 }
2718 flush_dcache_page(page);
2719
2720 reiserfs_commit_page(inode, page, start, start + copied);
2721
2722 /* generic_commit_write does this for us, but does not update the
2723 ** transaction tracking stuff when the size changes. So, we have
2724 ** to do the i_size updates here.
2725 */
Jan Karaec8e2f72009-12-17 15:27:06 -08002726 if (pos + copied > inode->i_size) {
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002727 struct reiserfs_transaction_handle myth;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002728 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2729 locked = true;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002730 /* If the file have grown beyond the border where it
2731 can have a tail, unmark it as needing a tail
2732 packing */
2733 if ((have_large_tails(inode->i_sb)
2734 && inode->i_size > i_block_size(inode) * 4)
2735 || (have_small_tails(inode->i_sb)
2736 && inode->i_size > i_block_size(inode)))
2737 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2738
2739 ret = journal_begin(&myth, inode->i_sb, 1);
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002740 if (ret)
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002741 goto journal_error;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002742
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002743 reiserfs_update_inode_transaction(inode);
Jan Karaec8e2f72009-12-17 15:27:06 -08002744 inode->i_size = pos + copied;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002745 /*
2746 * this will just nest into our transaction. It's important
2747 * to use mark_inode_dirty so the inode gets pushed around on the
2748 * dirty lists, and so that O_SYNC works as expected
2749 */
2750 mark_inode_dirty(inode);
2751 reiserfs_update_sd(&myth, inode);
2752 update_sd = 1;
2753 ret = journal_end(&myth, inode->i_sb, 1);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002754 if (ret)
2755 goto journal_error;
2756 }
2757 if (th) {
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002758 if (!locked) {
2759 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2760 locked = true;
2761 }
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002762 if (!update_sd)
2763 mark_inode_dirty(inode);
2764 ret = reiserfs_end_persistent_transaction(th);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002765 if (ret)
2766 goto out;
2767 }
2768
2769 out:
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002770 if (locked)
2771 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002772 unlock_page(page);
2773 page_cache_release(page);
Jan Karaec8e2f72009-12-17 15:27:06 -08002774
2775 if (pos + len > inode->i_size)
2776 reiserfs_truncate_failed_write(inode);
2777
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002778 return ret == 0 ? copied : ret;
2779
2780 journal_error:
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002781 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2782 locked = false;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002783 if (th) {
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002784 if (!update_sd)
2785 reiserfs_update_sd(th, inode);
2786 ret = reiserfs_end_persistent_transaction(th);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002787 }
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002788 goto out;
2789}
2790
2791int reiserfs_commit_write(struct file *f, struct page *page,
2792 unsigned from, unsigned to)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002793{
2794 struct inode *inode = page->mapping->host;
2795 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2796 int ret = 0;
2797 int update_sd = 0;
2798 struct reiserfs_transaction_handle *th = NULL;
2799
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002800 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002801 reiserfs_wait_on_write_block(inode->i_sb);
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002802 reiserfs_write_lock(inode->i_sb);
2803
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002804 if (reiserfs_transaction_running(inode->i_sb)) {
2805 th = current->journal_info;
2806 }
2807 reiserfs_commit_page(inode, page, from, to);
2808
2809 /* generic_commit_write does this for us, but does not update the
2810 ** transaction tracking stuff when the size changes. So, we have
2811 ** to do the i_size updates here.
2812 */
2813 if (pos > inode->i_size) {
2814 struct reiserfs_transaction_handle myth;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002815 /* If the file have grown beyond the border where it
2816 can have a tail, unmark it as needing a tail
2817 packing */
2818 if ((have_large_tails(inode->i_sb)
2819 && inode->i_size > i_block_size(inode) * 4)
2820 || (have_small_tails(inode->i_sb)
2821 && inode->i_size > i_block_size(inode)))
2822 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2823
2824 ret = journal_begin(&myth, inode->i_sb, 1);
Frederic Weisbecker7e942772009-08-25 03:38:12 +02002825 if (ret)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002826 goto journal_error;
Frederic Weisbecker7e942772009-08-25 03:38:12 +02002827
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002828 reiserfs_update_inode_transaction(inode);
2829 inode->i_size = pos;
Chris Mason9f037832005-09-13 01:25:17 -07002830 /*
2831 * this will just nest into our transaction. It's important
2832 * to use mark_inode_dirty so the inode gets pushed around on the
2833 * dirty lists, and so that O_SYNC works as expected
2834 */
2835 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002836 reiserfs_update_sd(&myth, inode);
2837 update_sd = 1;
2838 ret = journal_end(&myth, inode->i_sb, 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002839 if (ret)
2840 goto journal_error;
2841 }
2842 if (th) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002843 if (!update_sd)
Chris Mason9f037832005-09-13 01:25:17 -07002844 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002845 ret = reiserfs_end_persistent_transaction(th);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002846 if (ret)
2847 goto out;
2848 }
2849
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002850 out:
2851 return ret;
2852
2853 journal_error:
2854 if (th) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002855 if (!update_sd)
2856 reiserfs_update_sd(th, inode);
2857 ret = reiserfs_end_persistent_transaction(th);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002858 }
2859
2860 return ret;
2861}
2862
2863void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2864{
2865 if (reiserfs_attrs(inode->i_sb)) {
2866 if (sd_attrs & REISERFS_SYNC_FL)
2867 inode->i_flags |= S_SYNC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002869 inode->i_flags &= ~S_SYNC;
2870 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2871 inode->i_flags |= S_IMMUTABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002872 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002873 inode->i_flags &= ~S_IMMUTABLE;
2874 if (sd_attrs & REISERFS_APPEND_FL)
2875 inode->i_flags |= S_APPEND;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002877 inode->i_flags &= ~S_APPEND;
2878 if (sd_attrs & REISERFS_NOATIME_FL)
2879 inode->i_flags |= S_NOATIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002881 inode->i_flags &= ~S_NOATIME;
2882 if (sd_attrs & REISERFS_NOTAIL_FL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002883 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2884 else
2885 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2886 }
2887}
2888
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002889void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002890{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002891 if (reiserfs_attrs(inode->i_sb)) {
2892 if (inode->i_flags & S_IMMUTABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002893 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2894 else
2895 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002896 if (inode->i_flags & S_SYNC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002897 *sd_attrs |= REISERFS_SYNC_FL;
2898 else
2899 *sd_attrs &= ~REISERFS_SYNC_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002900 if (inode->i_flags & S_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002901 *sd_attrs |= REISERFS_NOATIME_FL;
2902 else
2903 *sd_attrs &= ~REISERFS_NOATIME_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002904 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905 *sd_attrs |= REISERFS_NOTAIL_FL;
2906 else
2907 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2908 }
2909}
2910
2911/* decide if this buffer needs to stay around for data logging or ordered
2912** write purposes
2913*/
2914static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2915{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002916 int ret = 1;
2917 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002918
Chris Masond62b1b82006-02-01 03:06:47 -08002919 lock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002920 spin_lock(&j->j_dirty_buffers_lock);
2921 if (!buffer_mapped(bh)) {
2922 goto free_jh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002924 /* the page is locked, and the only places that log a data buffer
2925 * also lock the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002926 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002927 if (reiserfs_file_data_log(inode)) {
2928 /*
2929 * very conservative, leave the buffer pinned if
2930 * anyone might need it.
2931 */
2932 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2933 ret = 0;
2934 }
Chris Masond62b1b82006-02-01 03:06:47 -08002935 } else if (buffer_dirty(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002936 struct reiserfs_journal_list *jl;
2937 struct reiserfs_jh *jh = bh->b_private;
2938
2939 /* why is this safe?
2940 * reiserfs_setattr updates i_size in the on disk
2941 * stat data before allowing vmtruncate to be called.
2942 *
2943 * If buffer was put onto the ordered list for this
2944 * transaction, we know for sure either this transaction
2945 * or an older one already has updated i_size on disk,
2946 * and this ordered data won't be referenced in the file
2947 * if we crash.
2948 *
2949 * if the buffer was put onto the ordered list for an older
2950 * transaction, we need to leave it around
2951 */
2952 if (jh && (jl = jh->jl)
2953 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2954 ret = 0;
2955 }
2956 free_jh:
2957 if (ret && bh->b_private) {
2958 reiserfs_free_jh(bh);
2959 }
2960 spin_unlock(&j->j_dirty_buffers_lock);
Chris Masond62b1b82006-02-01 03:06:47 -08002961 unlock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002962 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002963}
2964
2965/* clm -- taken from fs/buffer.c:block_invalidate_page */
NeilBrown2ff28e22006-03-26 01:37:18 -08002966static void reiserfs_invalidatepage(struct page *page, unsigned long offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002967{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002968 struct buffer_head *head, *bh, *next;
2969 struct inode *inode = page->mapping->host;
2970 unsigned int curr_off = 0;
2971 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002973 BUG_ON(!PageLocked(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002974
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002975 if (offset == 0)
2976 ClearPageChecked(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002977
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002978 if (!page_has_buffers(page))
2979 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002981 head = page_buffers(page);
2982 bh = head;
2983 do {
2984 unsigned int next_off = curr_off + bh->b_size;
2985 next = bh->b_this_page;
2986
2987 /*
2988 * is this block fully invalidated?
2989 */
2990 if (offset <= curr_off) {
2991 if (invalidatepage_can_drop(inode, bh))
2992 reiserfs_unmap_buffer(bh);
2993 else
2994 ret = 0;
2995 }
2996 curr_off = next_off;
2997 bh = next;
2998 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999
3000 /*
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003001 * We release buffers only if the entire page is being invalidated.
3002 * The get_block cached value has been unconditionally invalidated,
3003 * so real IO is not possible anymore.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003004 */
NeilBrown2ff28e22006-03-26 01:37:18 -08003005 if (!offset && ret) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003006 ret = try_to_release_page(page, 0);
NeilBrown2ff28e22006-03-26 01:37:18 -08003007 /* maybe should BUG_ON(!ret); - neilb */
3008 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003009 out:
NeilBrown2ff28e22006-03-26 01:37:18 -08003010 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003011}
3012
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003013static int reiserfs_set_page_dirty(struct page *page)
3014{
3015 struct inode *inode = page->mapping->host;
3016 if (reiserfs_file_data_log(inode)) {
3017 SetPageChecked(page);
3018 return __set_page_dirty_nobuffers(page);
3019 }
3020 return __set_page_dirty_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003021}
3022
3023/*
3024 * Returns 1 if the page's buffers were dropped. The page is locked.
3025 *
3026 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3027 * in the buffers at page_buffers(page).
3028 *
3029 * even in -o notail mode, we can't be sure an old mount without -o notail
3030 * didn't create files with tails.
3031 */
Al Viro27496a82005-10-21 03:20:48 -04003032static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003034 struct inode *inode = page->mapping->host;
3035 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3036 struct buffer_head *head;
3037 struct buffer_head *bh;
3038 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003040 WARN_ON(PageChecked(page));
3041 spin_lock(&j->j_dirty_buffers_lock);
3042 head = page_buffers(page);
3043 bh = head;
3044 do {
3045 if (bh->b_private) {
3046 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3047 reiserfs_free_jh(bh);
3048 } else {
3049 ret = 0;
3050 break;
3051 }
3052 }
3053 bh = bh->b_this_page;
3054 } while (bh != head);
3055 if (ret)
3056 ret = try_to_free_buffers(page);
3057 spin_unlock(&j->j_dirty_buffers_lock);
3058 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003059}
3060
3061/* We thank Mingming Cao for helping us understand in great detail what
3062 to do in this section of the code. */
3063static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003064 const struct iovec *iov, loff_t offset,
3065 unsigned long nr_segs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003066{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003067 struct file *file = iocb->ki_filp;
3068 struct inode *inode = file->f_mapping->host;
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003069 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003071 ret = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003072 offset, nr_segs,
3073 reiserfs_get_blocks_direct_io, NULL);
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003074
3075 /*
3076 * In case of error extending write may have instantiated a few
3077 * blocks outside i_size. Trim these off again.
3078 */
3079 if (unlikely((rw & WRITE) && ret < 0)) {
3080 loff_t isize = i_size_read(inode);
3081 loff_t end = offset + iov_length(iov, nr_segs);
3082
3083 if (end > isize)
3084 vmtruncate(inode, isize);
3085 }
3086
3087 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003088}
3089
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003090int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3091{
3092 struct inode *inode = dentry->d_inode;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003093 unsigned int ia_valid;
Frederic Weisbecker5fe1533f2010-01-04 22:04:01 +01003094 int depth;
3095 int error;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003096
Christoph Hellwigdb78b872010-06-04 11:30:03 +02003097 error = inode_change_ok(inode, attr);
3098 if (error)
3099 return error;
3100
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003101 /* must be turned off for recursive notify_change calls */
3102 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3103
Frederic Weisbecker5fe1533f2010-01-04 22:04:01 +01003104 depth = reiserfs_write_lock_once(inode->i_sb);
Dmitry Monakhov12755622010-04-08 22:04:20 +04003105 if (is_quota_modification(inode, attr))
Christoph Hellwig871a2932010-03-03 09:05:07 -05003106 dquot_initialize(inode);
Christoph Hellwig907f4552010-03-03 09:05:06 -05003107
Dmitry Monakhov12755622010-04-08 22:04:20 +04003108 if (attr->ia_valid & ATTR_SIZE) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003109 /* version 2 items will be caught by the s_maxbytes check
3110 ** done for us in vmtruncate
3111 */
3112 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3113 attr->ia_size > MAX_NON_LFS) {
3114 error = -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003115 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003116 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003117 /* fill in hole pointers in the expanding truncate case. */
3118 if (attr->ia_size > inode->i_size) {
Vladimir Savelievf7557e82007-10-16 01:25:14 -07003119 error = generic_cont_expand_simple(inode, attr->ia_size);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003120 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3121 int err;
3122 struct reiserfs_transaction_handle th;
3123 /* we're changing at most 2 bitmaps, inode + super */
3124 err = journal_begin(&th, inode->i_sb, 4);
3125 if (!err) {
3126 reiserfs_discard_prealloc(&th, inode);
3127 err = journal_end(&th, inode->i_sb, 4);
3128 }
3129 if (err)
3130 error = err;
3131 }
3132 if (error)
3133 goto out;
Vladimir Savelievdd535a52006-07-01 04:36:32 -07003134 /*
3135 * file size is changed, ctime and mtime are
3136 * to be updated
3137 */
3138 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003139 }
3140 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003141
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003142 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
3143 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
3144 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3145 /* stat data of format v3.5 has 16 bit uid and gid */
3146 error = -EINVAL;
3147 goto out;
3148 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003149
Christoph Hellwig10257742010-06-04 11:30:02 +02003150 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
3151 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
3152 struct reiserfs_transaction_handle th;
3153 int jbegin_count =
3154 2 *
3155 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3156 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3157 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003158
Christoph Hellwig10257742010-06-04 11:30:02 +02003159 error = reiserfs_chown_xattrs(inode, attr);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003160
Christoph Hellwig10257742010-06-04 11:30:02 +02003161 if (error)
3162 return error;
3163
3164 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3165 error = journal_begin(&th, inode->i_sb, jbegin_count);
3166 if (error)
3167 goto out;
3168 error = dquot_transfer(inode, attr);
3169 if (error) {
3170 journal_end(&th, inode->i_sb, jbegin_count);
3171 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003172 }
Christoph Hellwig10257742010-06-04 11:30:02 +02003173
3174 /* Update corresponding info in inode so that everything is in
3175 * one transaction */
3176 if (attr->ia_valid & ATTR_UID)
3177 inode->i_uid = attr->ia_uid;
3178 if (attr->ia_valid & ATTR_GID)
3179 inode->i_gid = attr->ia_gid;
3180 mark_inode_dirty(inode);
3181 error = journal_end(&th, inode->i_sb, jbegin_count);
3182 if (error)
3183 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003184 }
3185
Christoph Hellwig10257742010-06-04 11:30:02 +02003186 /*
3187 * Relax the lock here, as it might truncate the
3188 * inode pages and wait for inode pages locks.
3189 * To release such page lock, the owner needs the
3190 * reiserfs lock
3191 */
3192 reiserfs_write_unlock_once(inode->i_sb, depth);
3193 if ((attr->ia_valid & ATTR_SIZE) &&
3194 attr->ia_size != i_size_read(inode))
3195 error = vmtruncate(inode, attr->ia_size);
3196
3197 if (!error) {
3198 setattr_copy(inode, attr);
3199 mark_inode_dirty(inode);
3200 }
3201 depth = reiserfs_write_lock_once(inode->i_sb);
3202
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003203 if (!error && reiserfs_posixacl(inode->i_sb)) {
3204 if (attr->ia_valid & ATTR_MODE)
3205 error = reiserfs_acl_chmod(inode);
3206 }
3207
3208 out:
Frederic Weisbecker5fe1533f2010-01-04 22:04:01 +01003209 reiserfs_write_unlock_once(inode->i_sb, depth);
3210
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003211 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003212}
3213
Christoph Hellwigf5e54d62006-06-28 04:26:44 -07003214const struct address_space_operations reiserfs_address_space_operations = {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003215 .writepage = reiserfs_writepage,
3216 .readpage = reiserfs_readpage,
3217 .readpages = reiserfs_readpages,
3218 .releasepage = reiserfs_releasepage,
3219 .invalidatepage = reiserfs_invalidatepage,
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07003220 .write_begin = reiserfs_write_begin,
3221 .write_end = reiserfs_write_end,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003222 .bmap = reiserfs_aop_bmap,
3223 .direct_IO = reiserfs_direct_IO,
3224 .set_page_dirty = reiserfs_set_page_dirty,
3225};