blob: 0048cc16a6a8c5e6d54ebd1d9dd16f29f1d310ea [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>
Al Virof466c6f2012-03-17 01:16:43 -04007#include "reiserfs.h"
Al Viroa3063ab2012-03-17 01:03:10 -04008#include "acl.h"
Al Viroc45ac882012-03-17 00:59:06 -04009#include "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>
Kent Overstreeta27bb332013-05-07 16:19:08 -070021#include <linux/aio.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -070023int reiserfs_commit_write(struct file *f, struct page *page,
24 unsigned from, unsigned to);
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Al Viro845a2cc2010-06-07 11:37:37 -040026void reiserfs_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -070027{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070028 /* We need blocks for transaction + (user+group) quota update (possibly delete) */
29 int jbegin_count =
30 JOURNAL_PER_BALANCE_CNT * 2 +
31 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
32 struct reiserfs_transaction_handle th;
Frederic Weisbeckercb1c2e52009-12-13 23:32:06 +010033 int depth;
Jeff Mahoney24996042005-12-14 14:38:05 -050034 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Al Viro845a2cc2010-06-07 11:37:37 -040036 if (!inode->i_nlink && !is_bad_inode(inode))
Christoph Hellwig871a2932010-03-03 09:05:07 -050037 dquot_initialize(inode);
Christoph Hellwig907f4552010-03-03 09:05:06 -050038
Mark Fashehfef26652005-09-09 13:01:31 -070039 truncate_inode_pages(&inode->i_data, 0);
Al Viro845a2cc2010-06-07 11:37:37 -040040 if (inode->i_nlink)
41 goto no_delete;
Mark Fashehfef26652005-09-09 13:01:31 -070042
Frederic Weisbeckercb1c2e52009-12-13 23:32:06 +010043 depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070045 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
46 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070047 reiserfs_delete_xattrs(inode);
48
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070049 if (journal_begin(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070050 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070051 reiserfs_update_inode_transaction(inode);
52
Jeff Mahoneyeb35c212008-07-08 14:37:06 -040053 reiserfs_discard_prealloc(&th, inode);
54
Jeff Mahoney24996042005-12-14 14:38:05 -050055 err = reiserfs_delete_object(&th, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070056
57 /* Do quota update inside a transaction for journaled quotas. We must do that
58 * after delete_object so that quota updates go into the same transaction as
59 * stat data deletion */
Jeff Mahoney24996042005-12-14 14:38:05 -050060 if (!err)
Christoph Hellwig63936dd2010-03-03 09:05:01 -050061 dquot_free_inode(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070062
Alexander Zarochentsevb0b33de2006-08-05 12:14:01 -070063 if (journal_end(&th, inode->i_sb, jbegin_count))
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070064 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070065
Jeff Mahoney24996042005-12-14 14:38:05 -050066 /* check return value from reiserfs_delete_object after
67 * ending the transaction
68 */
69 if (err)
70 goto out;
71
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070072 /* all items of file are deleted, so we can remove "save" link */
73 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything
74 * about an error here */
75 } else {
76 /* no object items are in the tree */
77 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070079 out:
Jeff Mahoney48d17882012-08-02 21:36:04 -040080 reiserfs_write_unlock_once(inode->i_sb, depth);
Jan Karadbd57682012-05-03 14:48:02 +020081 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */
Al Viro845a2cc2010-06-07 11:37:37 -040082 dquot_drop(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070083 inode->i_blocks = 0;
Sergey Senozhatskyf4ae2fa2010-08-11 14:07:01 +030084 return;
Al Viro845a2cc2010-06-07 11:37:37 -040085
86no_delete:
Jan Karadbd57682012-05-03 14:48:02 +020087 clear_inode(inode);
Al Viro845a2cc2010-06-07 11:37:37 -040088 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089}
90
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070091static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
92 __u32 objectid, loff_t offset, int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070093{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070094 key->version = version;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070096 key->on_disk_key.k_dir_id = dirid;
97 key->on_disk_key.k_objectid = objectid;
98 set_cpu_key_k_offset(key, offset);
99 set_cpu_key_k_type(key, type);
100 key->key_length = length;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101}
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103/* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
104 offset and type of key */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700105void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
106 int type, int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700108 _make_cpu_key(key, get_inode_item_key_version(inode),
109 le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
110 le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
111 length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112}
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114//
115// when key is 0, do not set version and short key
116//
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700117inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
118 int version,
119 loff_t offset, int type, int length,
120 int entry_count /*or ih_free_space */ )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700122 if (key) {
123 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
124 ih->ih_key.k_objectid =
125 cpu_to_le32(key->on_disk_key.k_objectid);
126 }
127 put_ih_version(ih, version);
128 set_le_ih_k_offset(ih, offset);
129 set_le_ih_k_type(ih, type);
130 put_ih_item_len(ih, length);
131 /* set_ih_free_space (ih, 0); */
132 // for directory items it is entry count, for directs and stat
133 // datas - 0xffff, for indirects - 0
134 put_ih_entry_count(ih, entry_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135}
136
137//
138// FIXME: we might cache recently accessed indirect item
139
140// Ugh. Not too eager for that....
141// I cut the code until such time as I see a convincing argument (benchmark).
142// I don't want a bloated inode struct..., and I don't like code complexity....
143
144/* cutting the code is fine, since it really isn't in use yet and is easy
145** to add back in. But, Vladimir has a really good idea here. Think
146** about what happens for reading a file. For each page,
147** The VFS layer calls reiserfs_readpage, who searches the tree to find
148** an indirect item. This indirect item has X number of pointers, where
149** X is a big number if we've done the block allocation right. But,
150** we only use one or two of these pointers during each call to readpage,
151** needlessly researching again later on.
152**
153** The size of the cache could be dynamic based on the size of the file.
154**
155** I'd also like to see us cache the location the stat data item, since
156** we are needlessly researching for that frequently.
157**
158** --chris
159*/
160
161/* If this page has a file tail in it, and
162** it was read in by get_block_create_0, the page data is valid,
163** but tail is still sitting in a direct item, and we can't write to
164** it. So, look through this page, and check all the mapped buffers
165** to make sure they have valid block numbers. Any that don't need
Christoph Hellwigebdec242010-10-06 10:47:23 +0200166** to be unmapped, so that __block_write_begin will correctly call
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167** reiserfs_get_block to convert the tail into an unformatted node
168*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700169static inline void fix_tail_page_for_writing(struct page *page)
170{
171 struct buffer_head *head, *next, *bh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700173 if (page && page_has_buffers(page)) {
174 head = page_buffers(page);
175 bh = head;
176 do {
177 next = bh->b_this_page;
178 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
179 reiserfs_unmap_buffer(bh);
180 }
181 bh = next;
182 } while (bh != head);
183 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184}
185
186/* reiserfs_get_block does not need to allocate a block only if it has been
187 done already or non-hole position has been found in the indirect item */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700188static inline int allocation_needed(int retval, b_blocknr_t allocated,
189 struct item_head *ih,
190 __le32 * item, int pos_in_item)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700192 if (allocated)
193 return 0;
194 if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
195 get_block_num(item, pos_in_item))
196 return 0;
197 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198}
199
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700200static inline int indirect_item_found(int retval, struct item_head *ih)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700202 return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203}
204
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700205static inline void set_block_dev_mapped(struct buffer_head *bh,
206 b_blocknr_t block, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
208 map_bh(bh, inode->i_sb, block);
209}
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211//
212// files which were created in the earlier version can not be longer,
213// than 2 gb
214//
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700215static int file_capable(struct inode *inode, sector_t block)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700217 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file.
218 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
219 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700221 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222}
223
Adrian Bunkdeba0f42007-10-16 23:26:03 -0700224static int restart_transaction(struct reiserfs_transaction_handle *th,
225 struct inode *inode, struct treepath *path)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700226{
227 struct super_block *s = th->t_super;
228 int len = th->t_blocks_allocated;
229 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700231 BUG_ON(!th->t_trans_id);
232 BUG_ON(!th->t_refcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Suzuki K P87b41262006-12-06 20:36:10 -0800234 pathrelse(path);
235
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700236 /* we cannot restart while nested */
237 if (th->t_refcount > 1) {
238 return 0;
239 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700240 reiserfs_update_sd(th, inode);
241 err = journal_end(th, s, len);
242 if (!err) {
243 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
244 if (!err)
245 reiserfs_update_inode_transaction(inode);
246 }
247 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248}
249
250// it is called by get_block when create == 0. Returns block number
251// for 'block'-th logical block of file. When it hits direct item it
252// returns 0 (being called from bmap) or read direct item into piece
253// of page (bh_result)
254
255// Please improve the english/clarity in the comment above, as it is
256// hard to understand.
257
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700258static int _get_block_create_0(struct inode *inode, sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700259 struct buffer_head *bh_result, int args)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700261 INITIALIZE_PATH(path);
262 struct cpu_key key;
263 struct buffer_head *bh;
264 struct item_head *ih, tmp_ih;
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700265 b_blocknr_t blocknr;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700266 char *p = NULL;
267 int chars;
268 int ret;
269 int result;
270 int done = 0;
271 unsigned long offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700273 // prepare the key to look for the 'block'-th block of file
274 make_cpu_key(&key, inode,
275 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
276 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700278 result = search_for_position_by_key(inode->i_sb, &key, &path);
279 if (result != POSITION_FOUND) {
280 pathrelse(&path);
281 if (p)
282 kunmap(bh_result->b_page);
283 if (result == IO_ERROR)
284 return -EIO;
285 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
286 // That there is some MMAPED data associated with it that is yet to be written to disk.
287 if ((args & GET_BLOCK_NO_HOLE)
288 && !PageUptodate(bh_result->b_page)) {
289 return -ENOENT;
290 }
291 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700293 //
294 bh = get_last_bh(&path);
295 ih = get_ih(&path);
296 if (is_indirect_le_ih(ih)) {
297 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700299 /* FIXME: here we could cache indirect item or part of it in
300 the inode to avoid search_by_key in case of subsequent
301 access to file */
302 blocknr = get_block_num(ind_item, path.pos_in_item);
303 ret = 0;
304 if (blocknr) {
305 map_bh(bh_result, inode->i_sb, blocknr);
306 if (path.pos_in_item ==
307 ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
308 set_buffer_boundary(bh_result);
309 }
310 } else
311 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
312 // That there is some MMAPED data associated with it that is yet to be written to disk.
313 if ((args & GET_BLOCK_NO_HOLE)
314 && !PageUptodate(bh_result->b_page)) {
315 ret = -ENOENT;
316 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700318 pathrelse(&path);
319 if (p)
320 kunmap(bh_result->b_page);
321 return ret;
322 }
323 // requested data are in direct item(s)
324 if (!(args & GET_BLOCK_READ_DIRECT)) {
325 // we are called by bmap. FIXME: we can not map block of file
326 // when it is stored in direct item(s)
327 pathrelse(&path);
328 if (p)
329 kunmap(bh_result->b_page);
330 return -ENOENT;
331 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700333 /* if we've got a direct item, and the buffer or page was uptodate,
334 ** we don't want to pull data off disk again. skip to the
335 ** end, where we map the buffer and return
336 */
337 if (buffer_uptodate(bh_result)) {
338 goto finished;
339 } else
340 /*
341 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
342 ** pages without any buffers. If the page is up to date, we don't want
343 ** read old data off disk. Set the up to date bit on the buffer instead
344 ** and jump to the end
345 */
346 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 set_buffer_uptodate(bh_result);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700348 goto finished;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700350 // read file tail into part of page
351 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700352 copy_item_head(&tmp_ih, ih);
353
354 /* we only want to kmap if we are reading the tail into the page.
355 ** this is not the common case, so we don't kmap until we are
356 ** sure we need to. But, this means the item might move if
357 ** kmap schedules
358 */
Frederic Weisbecker27b3a5c2009-10-14 23:34:31 +0200359 if (!p)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700360 p = (char *)kmap(bh_result->b_page);
Frederic Weisbecker27b3a5c2009-10-14 23:34:31 +0200361
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700362 p += offset;
363 memset(p, 0, inode->i_sb->s_blocksize);
364 do {
365 if (!is_direct_le_ih(ih)) {
366 BUG();
367 }
368 /* make sure we don't read more bytes than actually exist in
369 ** the file. This can happen in odd cases where i_size isn't
Jeff Mahoney0222e652009-03-30 14:02:44 -0400370 ** correct, and when direct item padding results in a few
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700371 ** extra bytes at the end of the direct item
372 */
373 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
374 break;
375 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
376 chars =
377 inode->i_size - (le_ih_k_offset(ih) - 1) -
378 path.pos_in_item;
379 done = 1;
380 } else {
381 chars = ih_item_len(ih) - path.pos_in_item;
382 }
383 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700385 if (done)
386 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700388 p += chars;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700390 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
391 // we done, if read direct item is not the last item of
392 // node FIXME: we could try to check right delimiting key
393 // to see whether direct item continues in the right
394 // neighbor or rely on i_size
395 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700397 // update key to look for the next piece
398 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
399 result = search_for_position_by_key(inode->i_sb, &key, &path);
400 if (result != POSITION_FOUND)
401 // i/o error most likely
402 break;
403 bh = get_last_bh(&path);
404 ih = get_ih(&path);
405 } while (1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700407 flush_dcache_page(bh_result->b_page);
408 kunmap(bh_result->b_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700410 finished:
411 pathrelse(&path);
Qu Fuping6283d582005-06-25 14:55:44 -0700412
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700413 if (result == IO_ERROR)
414 return -EIO;
Qu Fuping6283d582005-06-25 14:55:44 -0700415
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700416 /* this buffer has valid data, but isn't valid for io. mapping it to
417 * block #0 tells the rest of reiserfs it just has a tail in it
418 */
419 map_bh(bh_result, inode->i_sb, 0);
420 set_buffer_uptodate(bh_result);
421 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422}
423
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424// this is called to create file map. So, _get_block_create_0 will not
425// read direct item
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700426static int reiserfs_bmap(struct inode *inode, sector_t block,
427 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700429 if (!file_capable(inode, block))
430 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700432 reiserfs_write_lock(inode->i_sb);
433 /* do not read the direct item */
434 _get_block_create_0(inode, block, bh_result, 0);
435 reiserfs_write_unlock(inode->i_sb);
436 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437}
438
439/* special version of get_block that is only used by grab_tail_page right
Christoph Hellwigebdec242010-10-06 10:47:23 +0200440** now. It is sent to __block_write_begin, and when you try to get a
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441** block past the end of the file (or a block from a hole) it returns
Christoph Hellwigebdec242010-10-06 10:47:23 +0200442** -ENOENT instead of a valid buffer. __block_write_begin expects to
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443** be able to do i/o on the buffers returned, unless an error value
444** is also returned.
Jeff Mahoney0222e652009-03-30 14:02:44 -0400445**
Christoph Hellwigebdec242010-10-06 10:47:23 +0200446** So, this allows __block_write_begin to be used for reading a single block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447** in a page. Where it does not produce a valid page for holes, or past the
448** end of the file. This turns out to be exactly what we need for reading
449** tails for conversion.
450**
451** The point of the wrapper is forcing a certain value for create, even
Jeff Mahoney0222e652009-03-30 14:02:44 -0400452** though the VFS layer is calling this function with create==1. If you
453** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454** don't use this function.
455*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700456static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
457 struct buffer_head *bh_result,
458 int create)
459{
460 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461}
462
463/* This is special helper for reiserfs_get_block in case we are executing
464 direct_IO request. */
465static int reiserfs_get_blocks_direct_io(struct inode *inode,
466 sector_t iblock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 struct buffer_head *bh_result,
468 int create)
469{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700470 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700472 bh_result->b_page = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700474 /* We set the b_size before reiserfs_get_block call since it is
475 referenced in convert_tail_for_hole() that may be called from
476 reiserfs_get_block() */
477 bh_result->b_size = (1 << inode->i_blkbits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700479 ret = reiserfs_get_block(inode, iblock, bh_result,
480 create | GET_BLOCK_NO_DANGLE);
481 if (ret)
482 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700484 /* don't allow direct io onto tail pages */
485 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
486 /* make sure future calls to the direct io funcs for this offset
487 ** in the file fail by unmapping the buffer
488 */
489 clear_buffer_mapped(bh_result);
490 ret = -EINVAL;
491 }
492 /* Possible unpacked tail. Flush the data before pages have
493 disappeared */
494 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
495 int err;
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +0200496
497 reiserfs_write_lock(inode->i_sb);
498
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700499 err = reiserfs_commit_for_inode(inode);
500 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +0200501
502 reiserfs_write_unlock(inode->i_sb);
503
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700504 if (err < 0)
505 ret = err;
506 }
507 out:
508 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509}
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511/*
512** helper function for when reiserfs_get_block is called for a hole
513** but the file tail is still in a direct item
514** bh_result is the buffer head for the hole
515** tail_offset is the offset of the start of the tail in the file
516**
517** This calls prepare_write, which will start a new transaction
518** you should not be in a transaction, or have any paths held when you
519** call this.
520*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700521static int convert_tail_for_hole(struct inode *inode,
522 struct buffer_head *bh_result,
523 loff_t tail_offset)
524{
525 unsigned long index;
526 unsigned long tail_end;
527 unsigned long tail_start;
528 struct page *tail_page;
529 struct page *hole_page = bh_result->b_page;
530 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700532 if ((tail_offset & (bh_result->b_size - 1)) != 1)
533 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700535 /* always try to read until the end of the block */
536 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1);
537 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700539 index = tail_offset >> PAGE_CACHE_SHIFT;
540 /* hole_page can be zero in case of direct_io, we are sure
541 that we cannot get here if we write with O_DIRECT into
542 tail page */
543 if (!hole_page || index != hole_page->index) {
544 tail_page = grab_cache_page(inode->i_mapping, index);
545 retval = -ENOMEM;
546 if (!tail_page) {
547 goto out;
548 }
549 } else {
550 tail_page = hole_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700553 /* we don't have to make sure the conversion did not happen while
554 ** we were locking the page because anyone that could convert
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800555 ** must first take i_mutex.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700556 **
557 ** We must fix the tail page for writing because it might have buffers
558 ** that are mapped, but have a block number of 0. This indicates tail
Christoph Hellwigebdec242010-10-06 10:47:23 +0200559 ** data that has been read directly into the page, and
560 ** __block_write_begin won't trigger a get_block in this case.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700561 */
562 fix_tail_page_for_writing(tail_page);
Christoph Hellwigebdec242010-10-06 10:47:23 +0200563 retval = __reiserfs_write_begin(tail_page, tail_start,
564 tail_end - tail_start);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700565 if (retval)
566 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700568 /* tail conversion might change the data in the page */
569 flush_dcache_page(tail_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700571 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700573 unlock:
574 if (tail_page != hole_page) {
575 unlock_page(tail_page);
576 page_cache_release(tail_page);
577 }
578 out:
579 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580}
581
582static inline int _allocate_block(struct reiserfs_transaction_handle *th,
Jeff Mahoney3ee16672007-10-18 23:39:25 -0700583 sector_t block,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700584 struct inode *inode,
585 b_blocknr_t * allocated_block_nr,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -0800586 struct treepath *path, int flags)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700587{
588 BUG_ON(!th->t_trans_id);
589
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590#ifdef REISERFS_PREALLOCATE
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800591 if (!(flags & GET_BLOCK_NO_IMUX)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700592 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
593 path, block);
594 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700596 return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
597 block);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598}
599
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700600int reiserfs_get_block(struct inode *inode, sector_t block,
601 struct buffer_head *bh_result, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700603 int repeat, retval = 0;
604 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int
605 INITIALIZE_PATH(path);
606 int pos_in_item;
607 struct cpu_key key;
608 struct buffer_head *bh, *unbh = NULL;
609 struct item_head *ih, tmp_ih;
610 __le32 *item;
611 int done;
612 int fs_gen;
Frederic Weisbecker269313092009-05-07 23:48:44 +0200613 int lock_depth;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700614 struct reiserfs_transaction_handle *th = NULL;
Jeff Mahoney0222e652009-03-30 14:02:44 -0400615 /* space reserved in transaction batch:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700616 . 3 balancings in direct->indirect conversion
617 . 1 block involved into reiserfs_update_sd()
618 XXX in practically impossible worst case direct2indirect()
619 can incur (much) more than 3 balancings.
620 quota update for user, group */
621 int jbegin_count =
622 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
623 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
624 int version;
625 int dangle = 1;
626 loff_t new_offset =
627 (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
Frederic Weisbecker269313092009-05-07 23:48:44 +0200629 lock_depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700630 version = get_inode_item_key_version(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700632 if (!file_capable(inode, block)) {
Frederic Weisbecker269313092009-05-07 23:48:44 +0200633 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700634 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 }
636
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700637 /* if !create, we aren't changing the FS, so we don't need to
638 ** log anything, so we don't need to start a transaction
639 */
640 if (!(create & GET_BLOCK_CREATE)) {
641 int ret;
642 /* find number of block-th logical block of the file */
643 ret = _get_block_create_0(inode, block, bh_result,
644 create | GET_BLOCK_READ_DIRECT);
Frederic Weisbecker269313092009-05-07 23:48:44 +0200645 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700646 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700648 /*
649 * if we're already in a transaction, make sure to close
650 * any new transactions we start in this func
651 */
652 if ((create & GET_BLOCK_NO_DANGLE) ||
653 reiserfs_transaction_running(inode->i_sb))
654 dangle = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700656 /* If file is of such a size, that it might have a tail and tails are enabled
657 ** we should mark it as possibly needing tail packing on close
658 */
659 if ((have_large_tails(inode->i_sb)
660 && inode->i_size < i_block_size(inode) * 4)
661 || (have_small_tails(inode->i_sb)
662 && inode->i_size < i_block_size(inode)))
663 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
664
665 /* set the key of the first byte in the 'block'-th block of file */
666 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
667 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
668 start_trans:
669 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
670 if (!th) {
671 retval = -ENOMEM;
672 goto failure;
673 }
674 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700676 research:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700678 retval = search_for_position_by_key(inode->i_sb, &key, &path);
679 if (retval == IO_ERROR) {
680 retval = -EIO;
681 goto failure;
682 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700684 bh = get_last_bh(&path);
685 ih = get_ih(&path);
686 item = get_item(&path);
687 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700689 fs_gen = get_generation(inode->i_sb);
690 copy_item_head(&tmp_ih, ih);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700692 if (allocation_needed
693 (retval, allocated_block_nr, ih, item, pos_in_item)) {
694 /* we have to allocate block for the unformatted node */
695 if (!th) {
696 pathrelse(&path);
697 goto start_trans;
698 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700700 repeat =
701 _allocate_block(th, block, inode, &allocated_block_nr,
702 &path, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700704 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
705 /* restart the transaction to give the journal a chance to free
706 ** some blocks. releases the path, so we have to go back to
707 ** research if we succeed on the second try
708 */
709 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
710 retval = restart_transaction(th, inode, &path);
711 if (retval)
712 goto failure;
713 repeat =
714 _allocate_block(th, block, inode,
715 &allocated_block_nr, NULL, create);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700717 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
718 goto research;
719 }
720 if (repeat == QUOTA_EXCEEDED)
721 retval = -EDQUOT;
722 else
723 retval = -ENOSPC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 goto failure;
725 }
726
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700727 if (fs_changed(fs_gen, inode->i_sb)
728 && item_moved(&tmp_ih, &path)) {
729 goto research;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700731 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700733 if (indirect_item_found(retval, ih)) {
734 b_blocknr_t unfm_ptr;
735 /* 'block'-th block is in the file already (there is
736 corresponding cell in some indirect item). But it may be
737 zero unformatted node pointer (hole) */
738 unfm_ptr = get_block_num(item, pos_in_item);
739 if (unfm_ptr == 0) {
740 /* use allocated block to plug the hole */
741 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
742 if (fs_changed(fs_gen, inode->i_sb)
743 && item_moved(&tmp_ih, &path)) {
744 reiserfs_restore_prepared_buffer(inode->i_sb,
745 bh);
746 goto research;
747 }
748 set_buffer_new(bh_result);
749 if (buffer_dirty(bh_result)
750 && reiserfs_data_ordered(inode->i_sb))
751 reiserfs_add_ordered_list(inode, bh_result);
752 put_block_num(item, pos_in_item, allocated_block_nr);
753 unfm_ptr = allocated_block_nr;
754 journal_mark_dirty(th, inode->i_sb, bh);
755 reiserfs_update_sd(th, inode);
756 }
757 set_block_dev_mapped(bh_result, unfm_ptr, inode);
758 pathrelse(&path);
759 retval = 0;
760 if (!dangle && th)
761 retval = reiserfs_end_persistent_transaction(th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Frederic Weisbecker269313092009-05-07 23:48:44 +0200763 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700764
765 /* the item was found, so new blocks were not added to the file
Jeff Mahoney0222e652009-03-30 14:02:44 -0400766 ** there is no need to make sure the inode is updated with this
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700767 ** transaction
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700769 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 }
771
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700772 if (!th) {
773 pathrelse(&path);
774 goto start_trans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700777 /* desired position is not found or is in the direct item. We have
778 to append file with holes up to 'block'-th block converting
779 direct items to indirect one if necessary */
780 done = 0;
781 do {
782 if (is_statdata_le_ih(ih)) {
783 __le32 unp = 0;
784 struct cpu_key tmp_key;
785
786 /* indirect item has to be inserted */
787 make_le_item_head(&tmp_ih, &key, version, 1,
788 TYPE_INDIRECT, UNFM_P_SIZE,
789 0 /* free_space */ );
790
791 if (cpu_key_k_offset(&key) == 1) {
792 /* we are going to add 'block'-th block to the file. Use
793 allocated block for that */
794 unp = cpu_to_le32(allocated_block_nr);
795 set_block_dev_mapped(bh_result,
796 allocated_block_nr, inode);
797 set_buffer_new(bh_result);
798 done = 1;
799 }
800 tmp_key = key; // ;)
801 set_cpu_key_k_offset(&tmp_key, 1);
802 PATH_LAST_POSITION(&path)++;
803
804 retval =
805 reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
806 inode, (char *)&unp);
807 if (retval) {
808 reiserfs_free_block(th, inode,
809 allocated_block_nr, 1);
810 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
811 }
812 //mark_tail_converted (inode);
813 } else if (is_direct_le_ih(ih)) {
814 /* direct item has to be converted */
815 loff_t tail_offset;
816
817 tail_offset =
818 ((le_ih_k_offset(ih) -
819 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
820 if (tail_offset == cpu_key_k_offset(&key)) {
821 /* direct item we just found fits into block we have
822 to map. Convert it into unformatted node: use
823 bh_result for the conversion */
824 set_block_dev_mapped(bh_result,
825 allocated_block_nr, inode);
826 unbh = bh_result;
827 done = 1;
828 } else {
829 /* we have to padd file tail stored in direct item(s)
830 up to block size and convert it to unformatted
831 node. FIXME: this should also get into page cache */
832
833 pathrelse(&path);
834 /*
835 * ugly, but we can only end the transaction if
836 * we aren't nested
837 */
838 BUG_ON(!th->t_refcount);
839 if (th->t_refcount == 1) {
840 retval =
841 reiserfs_end_persistent_transaction
842 (th);
843 th = NULL;
844 if (retval)
845 goto failure;
846 }
847
848 retval =
849 convert_tail_for_hole(inode, bh_result,
850 tail_offset);
851 if (retval) {
852 if (retval != -ENOSPC)
Jeff Mahoney0030b642009-03-30 14:02:28 -0400853 reiserfs_error(inode->i_sb,
854 "clm-6004",
855 "convert tail failed "
856 "inode %lu, error %d",
857 inode->i_ino,
858 retval);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700859 if (allocated_block_nr) {
860 /* the bitmap, the super, and the stat data == 3 */
861 if (!th)
862 th = reiserfs_persistent_transaction(inode->i_sb, 3);
863 if (th)
864 reiserfs_free_block(th,
865 inode,
866 allocated_block_nr,
867 1);
868 }
869 goto failure;
870 }
871 goto research;
872 }
873 retval =
874 direct2indirect(th, inode, &path, unbh,
875 tail_offset);
876 if (retval) {
877 reiserfs_unmap_buffer(unbh);
878 reiserfs_free_block(th, inode,
879 allocated_block_nr, 1);
880 goto failure;
881 }
882 /* it is important the set_buffer_uptodate is done after
883 ** the direct2indirect. The buffer might contain valid
884 ** data newer than the data on disk (read by readpage, changed,
885 ** and then sent here by writepage). direct2indirect needs
886 ** to know if unbh was already up to date, so it can decide
887 ** if the data in unbh needs to be replaced with data from
888 ** the disk
889 */
890 set_buffer_uptodate(unbh);
891
892 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
893 buffer will disappear shortly, so it should not be added to
894 */
895 if (unbh->b_page) {
896 /* we've converted the tail, so we must
897 ** flush unbh before the transaction commits
898 */
899 reiserfs_add_tail_list(inode, unbh);
900
901 /* mark it dirty now to prevent commit_write from adding
902 ** this buffer to the inode's dirty buffer list
903 */
904 /*
905 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
906 * It's still atomic, but it sets the page dirty too,
907 * which makes it eligible for writeback at any time by the
908 * VM (which was also the case with __mark_buffer_dirty())
909 */
910 mark_buffer_dirty(unbh);
911 }
912 } else {
913 /* append indirect item with holes if needed, when appending
914 pointer to 'block'-th block use block, which is already
915 allocated */
916 struct cpu_key tmp_key;
917 unp_t unf_single = 0; // We use this in case we need to allocate only
918 // one block which is a fastpath
919 unp_t *un;
920 __u64 max_to_insert =
921 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
922 UNFM_P_SIZE;
923 __u64 blocks_needed;
924
925 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
926 "vs-804: invalid position for append");
927 /* indirect item has to be appended, set up key of that position */
928 make_cpu_key(&tmp_key, inode,
929 le_key_k_offset(version,
930 &(ih->ih_key)) +
931 op_bytes_number(ih,
932 inode->i_sb->s_blocksize),
933 //pos_in_item * inode->i_sb->s_blocksize,
934 TYPE_INDIRECT, 3); // key type is unimportant
935
Vladimir V. Savelievc499ec22006-03-02 02:54:39 -0800936 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
937 "green-805: invalid offset");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700938 blocks_needed =
939 1 +
940 ((cpu_key_k_offset(&key) -
941 cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
942 s_blocksize_bits);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700943
944 if (blocks_needed == 1) {
945 un = &unf_single;
946 } else {
Frederic Weisbecker1d2c6cf2009-11-20 04:11:30 +0100947 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_NOFS);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700948 if (!un) {
949 un = &unf_single;
950 blocks_needed = 1;
951 max_to_insert = 0;
Yan Burman01afb212006-12-06 20:39:01 -0800952 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700953 }
954 if (blocks_needed <= max_to_insert) {
955 /* we are going to add target block to the file. Use allocated
956 block for that */
957 un[blocks_needed - 1] =
958 cpu_to_le32(allocated_block_nr);
959 set_block_dev_mapped(bh_result,
960 allocated_block_nr, inode);
961 set_buffer_new(bh_result);
962 done = 1;
963 } else {
964 /* paste hole to the indirect item */
965 /* If kmalloc failed, max_to_insert becomes zero and it means we
966 only have space for one block */
967 blocks_needed =
968 max_to_insert ? max_to_insert : 1;
969 }
970 retval =
971 reiserfs_paste_into_item(th, &path, &tmp_key, inode,
972 (char *)un,
973 UNFM_P_SIZE *
974 blocks_needed);
975
976 if (blocks_needed != 1)
977 kfree(un);
978
979 if (retval) {
980 reiserfs_free_block(th, inode,
981 allocated_block_nr, 1);
982 goto failure;
983 }
984 if (!done) {
985 /* We need to mark new file size in case this function will be
986 interrupted/aborted later on. And we may do this only for
987 holes. */
988 inode->i_size +=
989 inode->i_sb->s_blocksize * blocks_needed;
990 }
991 }
992
993 if (done == 1)
994 break;
995
996 /* this loop could log more blocks than we had originally asked
997 ** for. So, we have to allow the transaction to end if it is
Jeff Mahoney0222e652009-03-30 14:02:44 -0400998 ** too big or too full. Update the inode so things are
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700999 ** consistent if we crash before the function returns
1000 **
1001 ** release the path so that anybody waiting on the path before
1002 ** ending their transaction will be able to continue.
1003 */
1004 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1005 retval = restart_transaction(th, inode, &path);
1006 if (retval)
1007 goto failure;
1008 }
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02001009 /*
1010 * inserting indirect pointers for a hole can take a
1011 * long time. reschedule if needed and also release the write
1012 * lock for others.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001013 */
Frederic Weisbecker269313092009-05-07 23:48:44 +02001014 if (need_resched()) {
1015 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
1016 schedule();
1017 lock_depth = reiserfs_write_lock_once(inode->i_sb);
1018 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001019
1020 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1021 if (retval == IO_ERROR) {
1022 retval = -EIO;
1023 goto failure;
1024 }
1025 if (retval == POSITION_FOUND) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001026 reiserfs_warning(inode->i_sb, "vs-825",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001027 "%K should not be found", &key);
1028 retval = -EEXIST;
1029 if (allocated_block_nr)
1030 reiserfs_free_block(th, inode,
1031 allocated_block_nr, 1);
1032 pathrelse(&path);
1033 goto failure;
1034 }
1035 bh = get_last_bh(&path);
1036 ih = get_ih(&path);
1037 item = get_item(&path);
1038 pos_in_item = path.pos_in_item;
1039 } while (1);
1040
1041 retval = 0;
1042
1043 failure:
1044 if (th && (!dangle || (retval && !th->t_trans_id))) {
1045 int err;
1046 if (th->t_trans_id)
1047 reiserfs_update_sd(th, inode);
1048 err = reiserfs_end_persistent_transaction(th);
1049 if (err)
1050 retval = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
Frederic Weisbecker269313092009-05-07 23:48:44 +02001053 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001054 reiserfs_check_path(&path);
1055 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056}
1057
1058static int
1059reiserfs_readpages(struct file *file, struct address_space *mapping,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001060 struct list_head *pages, unsigned nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001062 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063}
1064
1065/* Compute real number of used bytes by file
1066 * Following three functions can go away when we'll have enough space in stat item
1067 */
1068static int real_space_diff(struct inode *inode, int sd_size)
1069{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001070 int bytes;
1071 loff_t blocksize = inode->i_sb->s_blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001073 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1074 return sd_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001076 /* End of file is also in full block with indirect reference, so round
1077 ** up to the next block.
1078 **
1079 ** there is just no way to know if the tail is actually packed
1080 ** on the file, so we have to assume it isn't. When we pack the
1081 ** tail, we add 4 bytes to pretend there really is an unformatted
1082 ** node pointer
1083 */
1084 bytes =
1085 ((inode->i_size +
1086 (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1087 sd_size;
1088 return bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089}
1090
1091static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001092 int sd_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001094 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1095 return inode->i_size +
1096 (loff_t) (real_space_diff(inode, sd_size));
1097 }
1098 return ((loff_t) real_space_diff(inode, sd_size)) +
1099 (((loff_t) blocks) << 9);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100}
1101
1102/* Compute number of blocks used by file in ReiserFS counting */
1103static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1104{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001105 loff_t bytes = inode_get_bytes(inode);
1106 loff_t real_space = real_space_diff(inode, sd_size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001108 /* keeps fsck and non-quota versions of reiserfs happy */
1109 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1110 bytes += (loff_t) 511;
1111 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001113 /* files from before the quota patch might i_blocks such that
1114 ** bytes < real_space. Deal with that here to prevent it from
1115 ** going negative.
1116 */
1117 if (bytes < real_space)
1118 return 0;
1119 return (bytes - real_space) >> 9;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120}
1121
1122//
1123// BAD: new directories have stat data of new type and all other items
1124// of old type. Version stored in the inode says about body items, so
1125// in update_stat_data we can not rely on inode, but have to check
1126// item version directly
1127//
1128
1129// called by read_locked_inode
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001130static void init_inode(struct inode *inode, struct treepath *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001132 struct buffer_head *bh;
1133 struct item_head *ih;
1134 __u32 rdev;
1135 //int version = ITEM_VERSION_1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001137 bh = PATH_PLAST_BUFFER(path);
1138 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001140 copy_key(INODE_PKEY(inode), &(ih->ih_key));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001142 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1143 REISERFS_I(inode)->i_flags = 0;
1144 REISERFS_I(inode)->i_prealloc_block = 0;
1145 REISERFS_I(inode)->i_prealloc_count = 0;
1146 REISERFS_I(inode)->i_trans_id = 0;
1147 REISERFS_I(inode)->i_jl = NULL;
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001148 reiserfs_init_xattr_rwsem(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001150 if (stat_data_v1(ih)) {
1151 struct stat_data_v1 *sd =
1152 (struct stat_data_v1 *)B_I_PITEM(bh, ih);
1153 unsigned long blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001155 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1156 set_inode_sd_version(inode, STAT_DATA_V1);
1157 inode->i_mode = sd_v1_mode(sd);
Miklos Szeredibfe86842011-10-28 14:13:29 +02001158 set_nlink(inode, sd_v1_nlink(sd));
Eric W. Biedermandf814652012-02-07 16:30:06 -08001159 i_uid_write(inode, sd_v1_uid(sd));
1160 i_gid_write(inode, sd_v1_gid(sd));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001161 inode->i_size = sd_v1_size(sd);
1162 inode->i_atime.tv_sec = sd_v1_atime(sd);
1163 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1164 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1165 inode->i_atime.tv_nsec = 0;
1166 inode->i_ctime.tv_nsec = 0;
1167 inode->i_mtime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001169 inode->i_blocks = sd_v1_blocks(sd);
1170 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1171 blocks = (inode->i_size + 511) >> 9;
1172 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1173 if (inode->i_blocks > blocks) {
1174 // there was a bug in <=3.5.23 when i_blocks could take negative
1175 // values. Starting from 3.5.17 this value could even be stored in
1176 // stat data. For such files we set i_blocks based on file
1177 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1178 // only updated if file's inode will ever change
1179 inode->i_blocks = blocks;
1180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001182 rdev = sd_v1_rdev(sd);
1183 REISERFS_I(inode)->i_first_direct_byte =
1184 sd_v1_first_direct_byte(sd);
1185 /* an early bug in the quota code can give us an odd number for the
1186 ** block count. This is incorrect, fix it here.
1187 */
1188 if (inode->i_blocks & 1) {
1189 inode->i_blocks++;
1190 }
1191 inode_set_bytes(inode,
1192 to_real_used_space(inode, inode->i_blocks,
1193 SD_V1_SIZE));
1194 /* nopack is initially zero for v1 objects. For v2 objects,
1195 nopack is initialised from sd_attrs */
1196 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1197 } else {
1198 // new stat data found, but object may have old items
1199 // (directories and symlinks)
1200 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih);
1201
1202 inode->i_mode = sd_v2_mode(sd);
Miklos Szeredibfe86842011-10-28 14:13:29 +02001203 set_nlink(inode, sd_v2_nlink(sd));
Eric W. Biedermandf814652012-02-07 16:30:06 -08001204 i_uid_write(inode, sd_v2_uid(sd));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001205 inode->i_size = sd_v2_size(sd);
Eric W. Biedermandf814652012-02-07 16:30:06 -08001206 i_gid_write(inode, sd_v2_gid(sd));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001207 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1208 inode->i_atime.tv_sec = sd_v2_atime(sd);
1209 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1210 inode->i_ctime.tv_nsec = 0;
1211 inode->i_mtime.tv_nsec = 0;
1212 inode->i_atime.tv_nsec = 0;
1213 inode->i_blocks = sd_v2_blocks(sd);
1214 rdev = sd_v2_rdev(sd);
1215 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1216 inode->i_generation =
1217 le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1218 else
1219 inode->i_generation = sd_v2_generation(sd);
1220
1221 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1222 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1223 else
1224 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1225 REISERFS_I(inode)->i_first_direct_byte = 0;
1226 set_inode_sd_version(inode, STAT_DATA_V2);
1227 inode_set_bytes(inode,
1228 to_real_used_space(inode, inode->i_blocks,
1229 SD_V2_SIZE));
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02001230 /* read persistent inode attributes from sd and initialise
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001231 generic inode flags from them */
1232 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1233 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 }
1235
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001236 pathrelse(path);
1237 if (S_ISREG(inode->i_mode)) {
1238 inode->i_op = &reiserfs_file_inode_operations;
1239 inode->i_fop = &reiserfs_file_operations;
1240 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1241 } else if (S_ISDIR(inode->i_mode)) {
1242 inode->i_op = &reiserfs_dir_inode_operations;
1243 inode->i_fop = &reiserfs_dir_operations;
1244 } else if (S_ISLNK(inode->i_mode)) {
1245 inode->i_op = &reiserfs_symlink_inode_operations;
1246 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1247 } else {
1248 inode->i_blocks = 0;
1249 inode->i_op = &reiserfs_special_inode_operations;
1250 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252}
1253
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254// update new stat data with inode fields
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001255static void inode2sd(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001257 struct stat_data *sd_v2 = (struct stat_data *)sd;
1258 __u16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001260 set_sd_v2_mode(sd_v2, inode->i_mode);
1261 set_sd_v2_nlink(sd_v2, inode->i_nlink);
Eric W. Biedermandf814652012-02-07 16:30:06 -08001262 set_sd_v2_uid(sd_v2, i_uid_read(inode));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001263 set_sd_v2_size(sd_v2, size);
Eric W. Biedermandf814652012-02-07 16:30:06 -08001264 set_sd_v2_gid(sd_v2, i_gid_read(inode));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001265 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1266 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1267 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1268 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1269 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1270 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1271 else
1272 set_sd_v2_generation(sd_v2, inode->i_generation);
1273 flags = REISERFS_I(inode)->i_attrs;
1274 i_attrs_to_sd_attrs(inode, &flags);
1275 set_sd_v2_attrs(sd_v2, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276}
1277
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278// used to copy inode's fields to old stat data
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001279static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001281 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001283 set_sd_v1_mode(sd_v1, inode->i_mode);
Eric W. Biedermandf814652012-02-07 16:30:06 -08001284 set_sd_v1_uid(sd_v1, i_uid_read(inode));
1285 set_sd_v1_gid(sd_v1, i_gid_read(inode));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001286 set_sd_v1_nlink(sd_v1, inode->i_nlink);
1287 set_sd_v1_size(sd_v1, size);
1288 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1289 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1290 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001292 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1293 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1294 else
1295 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001297 // Sigh. i_first_direct_byte is back
1298 set_sd_v1_first_direct_byte(sd_v1,
1299 REISERFS_I(inode)->i_first_direct_byte);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300}
1301
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302/* NOTE, you must prepare the buffer head before sending it here,
1303** and then log it after the call
1304*/
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001305static void update_stat_data(struct treepath *path, struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001306 loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001308 struct buffer_head *bh;
1309 struct item_head *ih;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001311 bh = PATH_PLAST_BUFFER(path);
1312 ih = PATH_PITEM_HEAD(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001314 if (!is_statdata_le_ih(ih))
Jeff Mahoneyc3a9c212009-03-30 14:02:25 -04001315 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001316 INODE_PKEY(inode), ih);
1317
1318 if (stat_data_v1(ih)) {
1319 // path points to old stat data
1320 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1321 } else {
1322 inode2sd(B_I_PITEM(bh, ih), inode, size);
1323 }
1324
1325 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326}
1327
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001328void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1329 struct inode *inode, loff_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001331 struct cpu_key key;
1332 INITIALIZE_PATH(path);
1333 struct buffer_head *bh;
1334 int fs_gen;
1335 struct item_head *ih, tmp_ih;
1336 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001338 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001340 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant
1341
1342 for (;;) {
1343 int pos;
1344 /* look for the object's stat data */
1345 retval = search_item(inode->i_sb, &key, &path);
1346 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001347 reiserfs_error(inode->i_sb, "vs-13050",
1348 "i/o failure occurred trying to "
1349 "update %K stat data", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001350 return;
1351 }
1352 if (retval == ITEM_NOT_FOUND) {
1353 pos = PATH_LAST_POSITION(&path);
1354 pathrelse(&path);
1355 if (inode->i_nlink == 0) {
1356 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1357 return;
1358 }
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001359 reiserfs_warning(inode->i_sb, "vs-13060",
1360 "stat data of object %k (nlink == %d) "
1361 "not found (pos %d)",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001362 INODE_PKEY(inode), inode->i_nlink,
1363 pos);
1364 reiserfs_check_path(&path);
1365 return;
1366 }
1367
1368 /* sigh, prepare_for_journal might schedule. When it schedules the
1369 ** FS might change. We have to detect that, and loop back to the
1370 ** search if the stat data item has moved
1371 */
1372 bh = get_last_bh(&path);
1373 ih = get_ih(&path);
1374 copy_item_head(&tmp_ih, ih);
1375 fs_gen = get_generation(inode->i_sb);
1376 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1377 if (fs_changed(fs_gen, inode->i_sb)
1378 && item_moved(&tmp_ih, &path)) {
1379 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1380 continue; /* Stat_data item has been moved after scheduling. */
1381 }
1382 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001384 update_stat_data(&path, inode, size);
1385 journal_mark_dirty(th, th->t_super, bh);
1386 pathrelse(&path);
1387 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388}
1389
1390/* reiserfs_read_locked_inode is called to read the inode off disk, and it
1391** does a make_bad_inode when things go wrong. But, we need to make sure
1392** and clear the key in the private portion of the inode, otherwise a
1393** corresponding iput might try to delete whatever object the inode last
1394** represented.
1395*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001396static void reiserfs_make_bad_inode(struct inode *inode)
1397{
1398 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1399 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400}
1401
1402//
1403// initially this function was derived from minix or ext2's analog and
1404// evolved as the prototype did
1405//
1406
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001407int reiserfs_init_locked_inode(struct inode *inode, void *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001409 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1410 inode->i_ino = args->objectid;
1411 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1412 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413}
1414
1415/* looks for stat data in the tree, and fills up the fields of in-core
1416 inode stat data fields */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001417void reiserfs_read_locked_inode(struct inode *inode,
1418 struct reiserfs_iget_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001420 INITIALIZE_PATH(path_to_sd);
1421 struct cpu_key key;
1422 unsigned long dirino;
1423 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001425 dirino = args->dirid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001427 /* set version 1, version 2 could be used too, because stat data
1428 key is the same in both versions */
1429 key.version = KEY_FORMAT_3_5;
1430 key.on_disk_key.k_dir_id = dirino;
1431 key.on_disk_key.k_objectid = inode->i_ino;
1432 key.on_disk_key.k_offset = 0;
1433 key.on_disk_key.k_type = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001435 /* look for the object's stat data */
1436 retval = search_item(inode->i_sb, &key, &path_to_sd);
1437 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001438 reiserfs_error(inode->i_sb, "vs-13070",
1439 "i/o failure occurred trying to find "
1440 "stat data of %K", &key);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001441 reiserfs_make_bad_inode(inode);
1442 return;
1443 }
1444 if (retval != ITEM_FOUND) {
1445 /* a stale NFS handle can trigger this without it being an error */
1446 pathrelse(&path_to_sd);
1447 reiserfs_make_bad_inode(inode);
Miklos Szeredi6d6b77f2011-10-28 14:13:28 +02001448 clear_nlink(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001449 return;
1450 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001452 init_inode(inode, &path_to_sd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001454 /* It is possible that knfsd is trying to access inode of a file
1455 that is being removed from the disk by some other thread. As we
1456 update sd on unlink all that is required is to check for nlink
1457 here. This bug was first found by Sizif when debugging
1458 SquidNG/Butterfly, forgotten, and found again after Philippe
Jeff Mahoney0222e652009-03-30 14:02:44 -04001459 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001461 More logical fix would require changes in fs/inode.c:iput() to
1462 remove inode from hash-table _after_ fs cleaned disk stuff up and
1463 in iget() to return NULL if I_FREEING inode is found in
1464 hash-table. */
1465 /* Currently there is one place where it's ok to meet inode with
1466 nlink==0: processing of open-unlinked and half-truncated files
1467 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1468 if ((inode->i_nlink == 0) &&
1469 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001470 reiserfs_warning(inode->i_sb, "vs-13075",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001471 "dead inode read from disk %K. "
1472 "This is likely to be race with knfsd. Ignore",
1473 &key);
1474 reiserfs_make_bad_inode(inode);
1475 }
1476
1477 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
Christoph Hellwig4482a082011-07-23 17:37:03 +02001479 /*
1480 * Stat data v1 doesn't support ACLs.
1481 */
1482 if (get_inode_sd_version(inode) == STAT_DATA_V1)
1483 cache_no_acl(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484}
1485
1486/**
1487 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1488 *
1489 * @inode: inode from hash table to check
1490 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1491 *
1492 * This function is called by iget5_locked() to distinguish reiserfs inodes
1493 * having the same inode numbers. Such inodes can only exist due to some
1494 * error condition. One of them should be bad. Inodes with identical
1495 * inode numbers (objectids) are distinguished by parent directory ids.
1496 *
1497 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001498int reiserfs_find_actor(struct inode *inode, void *opaque)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001500 struct reiserfs_iget_args *args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001502 args = opaque;
1503 /* args is already in CPU order */
1504 return (inode->i_ino == args->objectid) &&
1505 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506}
1507
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001508struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001510 struct inode *inode;
1511 struct reiserfs_iget_args args;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001513 args.objectid = key->on_disk_key.k_objectid;
1514 args.dirid = key->on_disk_key.k_dir_id;
Frederic Weisbecker175359f2010-02-11 13:13:10 +01001515 reiserfs_write_unlock(s);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001516 inode = iget5_locked(s, key->on_disk_key.k_objectid,
1517 reiserfs_find_actor, reiserfs_init_locked_inode,
1518 (void *)(&args));
Frederic Weisbecker175359f2010-02-11 13:13:10 +01001519 reiserfs_write_lock(s);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001520 if (!inode)
1521 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001523 if (inode->i_state & I_NEW) {
1524 reiserfs_read_locked_inode(inode, &args);
1525 unlock_new_inode(inode);
1526 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001528 if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1529 /* either due to i/o error or a stale NFS handle */
1530 iput(inode);
1531 inode = NULL;
1532 }
1533 return inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534}
1535
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001536static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1537 u32 objectid, u32 dir_id, u32 generation)
1538
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001540 struct cpu_key key;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001541 struct inode *inode;
1542
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001543 key.on_disk_key.k_objectid = objectid;
1544 key.on_disk_key.k_dir_id = dir_id;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001545 reiserfs_write_lock(sb);
1546 inode = reiserfs_iget(sb, &key);
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001547 if (inode && !IS_ERR(inode) && generation != 0 &&
1548 generation != inode->i_generation) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001549 iput(inode);
1550 inode = NULL;
1551 }
1552 reiserfs_write_unlock(sb);
Christoph Hellwig44003722008-08-11 15:49:04 +02001553
1554 return d_obtain_alias(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555}
1556
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001557struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1558 int fh_len, int fh_type)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001559{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001560 /* fhtype happens to reflect the number of u32s encoded.
1561 * due to a bug in earlier code, fhtype might indicate there
1562 * are more u32s then actually fitted.
1563 * so if fhtype seems to be more than len, reduce fhtype.
1564 * Valid types are:
1565 * 2 - objectid + dir_id - legacy support
1566 * 3 - objectid + dir_id + generation
1567 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1568 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1569 * 6 - as above plus generation of directory
1570 * 6 does not fit in NFSv2 handles
1571 */
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001572 if (fh_type > fh_len) {
1573 if (fh_type != 6 || fh_len != 5)
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001574 reiserfs_warning(sb, "reiserfs-13077",
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001575 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1576 fh_type, fh_len);
Hugh Dickins35c2a7f2012-10-07 20:32:51 -07001577 fh_type = fh_len;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001578 }
Hugh Dickins35c2a7f2012-10-07 20:32:51 -07001579 if (fh_len < 2)
1580 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001582 return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1583 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1584}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001586struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1587 int fh_len, int fh_type)
1588{
Hugh Dickins35c2a7f2012-10-07 20:32:51 -07001589 if (fh_type > fh_len)
1590 fh_type = fh_len;
Christoph Hellwigbe55caf12007-10-21 16:42:13 -07001591 if (fh_type < 4)
1592 return NULL;
1593
1594 return reiserfs_get_dentry(sb,
1595 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1596 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1597 (fh_type == 6) ? fid->raw[5] : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598}
1599
Al Virob0b03822012-04-02 14:34:06 -04001600int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
1601 struct inode *parent)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001602{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001603 int maxlen = *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
Al Virob0b03822012-04-02 14:34:06 -04001605 if (parent && (maxlen < 5)) {
Aneesh Kumar K.V5fe0c232011-01-29 18:43:25 +05301606 *lenp = 5;
Namjae Jeon94e07a752013-02-17 15:48:11 +09001607 return FILEID_INVALID;
Aneesh Kumar K.V5fe0c232011-01-29 18:43:25 +05301608 } else if (maxlen < 3) {
1609 *lenp = 3;
Namjae Jeon94e07a752013-02-17 15:48:11 +09001610 return FILEID_INVALID;
Aneesh Kumar K.V5fe0c232011-01-29 18:43:25 +05301611 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001613 data[0] = inode->i_ino;
1614 data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1615 data[2] = inode->i_generation;
1616 *lenp = 3;
Al Virob0b03822012-04-02 14:34:06 -04001617 if (parent) {
1618 data[3] = parent->i_ino;
1619 data[4] = le32_to_cpu(INODE_PKEY(parent)->k_dir_id);
1620 *lenp = 5;
1621 if (maxlen >= 6) {
1622 data[5] = parent->i_generation;
1623 *lenp = 6;
1624 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001625 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001626 return *lenp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627}
1628
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629/* looks for stat data, then copies fields to it, marks the buffer
1630 containing stat data as dirty */
1631/* reiserfs inodes are never really dirty, since the dirty inode call
1632** always logs them. This call allows the VFS inode marking routines
1633** to properly mark inodes for datasync and such, but only actually
1634** does something when called for a synchronous update.
1635*/
Christoph Hellwiga9185b42010-03-05 09:21:37 +01001636int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001637{
1638 struct reiserfs_transaction_handle th;
1639 int jbegin_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001641 if (inode->i_sb->s_flags & MS_RDONLY)
1642 return -EROFS;
1643 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001644 ** these cases are just when the system needs ram, not when the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001645 ** inode needs to reach disk for safety, and they can safely be
1646 ** ignored because the altered inode has already been logged.
1647 */
Christoph Hellwiga9185b42010-03-05 09:21:37 +01001648 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001649 reiserfs_write_lock(inode->i_sb);
1650 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1651 reiserfs_update_sd(&th, inode);
1652 journal_end_sync(&th, inode->i_sb, jbegin_count);
1653 }
1654 reiserfs_write_unlock(inode->i_sb);
1655 }
1656 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657}
1658
1659/* stat data of new object is inserted already, this inserts the item
1660 containing "." and ".." entries */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001661static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1662 struct inode *inode,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001663 struct item_head *ih, struct treepath *path,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001664 struct inode *dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001666 struct super_block *sb = th->t_super;
1667 char empty_dir[EMPTY_DIR_SIZE];
1668 char *body = empty_dir;
1669 struct cpu_key key;
1670 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001672 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001674 _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1675 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1676 TYPE_DIRENTRY, 3 /*key length */ );
1677
1678 /* compose item head for new item. Directories consist of items of
1679 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1680 is done by reiserfs_new_inode */
1681 if (old_format_only(sb)) {
1682 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1683 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1684
1685 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1686 ih->ih_key.k_objectid,
1687 INODE_PKEY(dir)->k_dir_id,
1688 INODE_PKEY(dir)->k_objectid);
1689 } else {
1690 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1691 TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1692
1693 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1694 ih->ih_key.k_objectid,
1695 INODE_PKEY(dir)->k_dir_id,
1696 INODE_PKEY(dir)->k_objectid);
1697 }
1698
1699 /* look for place in the tree for new item */
1700 retval = search_item(sb, &key, path);
1701 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001702 reiserfs_error(sb, "vs-13080",
1703 "i/o failure occurred creating new directory");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001704 return -EIO;
1705 }
1706 if (retval == ITEM_FOUND) {
1707 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001708 reiserfs_warning(sb, "vs-13070",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001709 "object with this key exists (%k)",
1710 &(ih->ih_key));
1711 return -EEXIST;
1712 }
1713
1714 /* insert item, that is empty directory item */
1715 return reiserfs_insert_item(th, path, &key, ih, inode, body);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716}
1717
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718/* stat data of object has been inserted, this inserts the item
1719 containing the body of symlink */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001720static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */
1721 struct item_head *ih,
Josef "Jeff" Sipekfec6d052006-12-08 02:36:32 -08001722 struct treepath *path, const char *symname,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001723 int item_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001725 struct super_block *sb = th->t_super;
1726 struct cpu_key key;
1727 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001729 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001731 _make_cpu_key(&key, KEY_FORMAT_3_5,
1732 le32_to_cpu(ih->ih_key.k_dir_id),
1733 le32_to_cpu(ih->ih_key.k_objectid),
1734 1, TYPE_DIRECT, 3 /*key length */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001736 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1737 0 /*free_space */ );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001739 /* look for place in the tree for new item */
1740 retval = search_item(sb, &key, path);
1741 if (retval == IO_ERROR) {
Jeff Mahoney0030b642009-03-30 14:02:28 -04001742 reiserfs_error(sb, "vs-13080",
1743 "i/o failure occurred creating new symlink");
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001744 return -EIO;
1745 }
1746 if (retval == ITEM_FOUND) {
1747 pathrelse(path);
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001748 reiserfs_warning(sb, "vs-13080",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001749 "object with this key exists (%k)",
1750 &(ih->ih_key));
1751 return -EEXIST;
1752 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001754 /* insert item, that is body of symlink */
1755 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756}
1757
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758/* inserts the stat data into the tree, and then calls
1759 reiserfs_new_directory (to insert ".", ".." item if new object is
1760 directory) or reiserfs_new_symlink (to insert symlink body if new
Jeff Mahoney0222e652009-03-30 14:02:44 -04001761 object is symlink) or nothing (if new object is regular file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
1763 NOTE! uid and gid must already be set in the inode. If we return
1764 non-zero due to an error, we have to drop the quota previously allocated
1765 for the fresh inode. This can only be done outside a transaction, so
1766 if we return non-zero, we also end the transaction. */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001767int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
Al Viro8e071892011-07-26 02:50:53 -04001768 struct inode *dir, umode_t mode, const char *symname,
Jeff Mahoney0222e652009-03-30 14:02:44 -04001769 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001770 strlen (symname) for symlinks) */
1771 loff_t i_size, struct dentry *dentry,
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001772 struct inode *inode,
1773 struct reiserfs_security_handle *security)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001775 struct super_block *sb;
Al Viroc1eaa262008-12-30 02:03:58 -05001776 struct reiserfs_iget_args args;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001777 INITIALIZE_PATH(path_to_key);
1778 struct cpu_key key;
1779 struct item_head ih;
1780 struct stat_data sd;
1781 int retval;
1782 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001784 BUG_ON(!th->t_trans_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785
Jan Kara7af11682012-11-13 17:05:14 +01001786 reiserfs_write_unlock(inode->i_sb);
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001787 err = dquot_alloc_inode(inode);
Jan Kara7af11682012-11-13 17:05:14 +01001788 reiserfs_write_lock(inode->i_sb);
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001789 if (err)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001790 goto out_end_trans;
Eric Sesterhenn585b7742006-10-04 02:15:30 -07001791 if (!dir->i_nlink) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001792 err = -EPERM;
1793 goto out_bad_inode;
1794 }
1795
1796 sb = dir->i_sb;
1797
1798 /* item head of new item */
1799 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1800 ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1801 if (!ih.ih_key.k_objectid) {
1802 err = -ENOMEM;
1803 goto out_bad_inode;
1804 }
Al Viroc1eaa262008-12-30 02:03:58 -05001805 args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
Al Viro2f1169e2009-01-02 08:16:51 -05001806 if (old_format_only(sb))
1807 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1808 TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1809 else
1810 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1811 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
Al Viroc1eaa262008-12-30 02:03:58 -05001812 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE);
1813 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
Jeff Mahoneya1457c02013-05-31 15:51:17 -04001814
1815 reiserfs_write_unlock(inode->i_sb);
1816 err = insert_inode_locked4(inode, args.objectid,
1817 reiserfs_find_actor, &args);
1818 reiserfs_write_lock(inode->i_sb);
1819 if (err) {
Al Viroc1eaa262008-12-30 02:03:58 -05001820 err = -EINVAL;
1821 goto out_bad_inode;
1822 }
Jeff Mahoneya1457c02013-05-31 15:51:17 -04001823
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001824 if (old_format_only(sb))
Jeff Mahoney0222e652009-03-30 14:02:44 -04001825 /* not a perfect generation count, as object ids can be reused, but
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001826 ** this is as good as reiserfs can do right now.
1827 ** note that the private part of inode isn't filled in yet, we have
1828 ** to use the directory.
1829 */
1830 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1831 else
1832#if defined( USE_INODE_GENERATION_COUNTER )
1833 inode->i_generation =
1834 le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1835#else
1836 inode->i_generation = ++event;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001839 /* fill stat data */
Miklos Szeredibfe86842011-10-28 14:13:29 +02001840 set_nlink(inode, (S_ISDIR(mode) ? 2 : 1));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001841
1842 /* uid and gid must already be set by the caller for quota init */
1843
1844 /* symlink cannot be immutable or append only, right? */
1845 if (S_ISLNK(inode->i_mode))
1846 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
1847
1848 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
1849 inode->i_size = i_size;
1850 inode->i_blocks = 0;
1851 inode->i_bytes = 0;
1852 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1853 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1854
1855 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list));
1856 REISERFS_I(inode)->i_flags = 0;
1857 REISERFS_I(inode)->i_prealloc_block = 0;
1858 REISERFS_I(inode)->i_prealloc_count = 0;
1859 REISERFS_I(inode)->i_trans_id = 0;
1860 REISERFS_I(inode)->i_jl = NULL;
1861 REISERFS_I(inode)->i_attrs =
1862 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1863 sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
Alexey Dobriyan068fbb32006-09-29 01:59:58 -07001864 reiserfs_init_xattr_rwsem(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001865
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001866 /* key to search for correct place for new stat data */
1867 _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
1868 le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
1869 TYPE_STAT_DATA, 3 /*key length */ );
1870
1871 /* find proper place for inserting of stat data */
1872 retval = search_item(sb, &key, &path_to_key);
1873 if (retval == IO_ERROR) {
1874 err = -EIO;
1875 goto out_bad_inode;
1876 }
1877 if (retval == ITEM_FOUND) {
1878 pathrelse(&path_to_key);
1879 err = -EEXIST;
1880 goto out_bad_inode;
1881 }
1882 if (old_format_only(sb)) {
Eric W. Biedermandf814652012-02-07 16:30:06 -08001883 if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001884 pathrelse(&path_to_key);
1885 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1886 err = -EINVAL;
1887 goto out_bad_inode;
1888 }
1889 inode2sd_v1(&sd, inode, inode->i_size);
1890 } else {
1891 inode2sd(&sd, inode, inode->i_size);
1892 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001893 // store in in-core inode the key of stat data and version all
1894 // object items will have (directory items will have old offset
1895 // format, other new objects will consist of new items)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001896 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1897 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1898 else
1899 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1900 if (old_format_only(sb))
1901 set_inode_sd_version(inode, STAT_DATA_V1);
1902 else
1903 set_inode_sd_version(inode, STAT_DATA_V2);
1904
1905 /* insert the stat data into the tree */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906#ifdef DISPLACE_NEW_PACKING_LOCALITIES
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001907 if (REISERFS_I(dir)->new_packing_locality)
1908 th->displace_new_blocks = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001910 retval =
1911 reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
1912 (char *)(&sd));
1913 if (retval) {
1914 err = retval;
1915 reiserfs_check_path(&path_to_key);
1916 goto out_bad_inode;
1917 }
1918#ifdef DISPLACE_NEW_PACKING_LOCALITIES
1919 if (!th->displace_new_blocks)
1920 REISERFS_I(dir)->new_packing_locality = 0;
1921#endif
1922 if (S_ISDIR(mode)) {
1923 /* insert item with "." and ".." */
1924 retval =
1925 reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
1926 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001928 if (S_ISLNK(mode)) {
1929 /* insert body of symlink */
1930 if (!old_format_only(sb))
1931 i_size = ROUND_UP(i_size);
1932 retval =
1933 reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
1934 i_size);
1935 }
1936 if (retval) {
1937 err = retval;
1938 reiserfs_check_path(&path_to_key);
1939 journal_end(th, th->t_super, th->t_blocks_allocated);
1940 goto out_inserted_sd;
1941 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001943 if (reiserfs_posixacl(inode->i_sb)) {
Jeff Mahoney0ab26212009-03-30 14:02:39 -04001944 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001945 if (retval) {
1946 err = retval;
1947 reiserfs_check_path(&path_to_key);
1948 journal_end(th, th->t_super, th->t_blocks_allocated);
1949 goto out_inserted_sd;
1950 }
1951 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04001952 reiserfs_warning(inode->i_sb, "jdm-13090",
1953 "ACLs aren't enabled in the fs, "
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001954 "but vfs thinks they are!");
Jeff Mahoney6dfede692009-03-30 14:02:32 -04001955 } else if (IS_PRIVATE(dir))
1956 inode->i_flags |= S_PRIVATE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Jeff Mahoney57fe60d2009-03-30 14:02:41 -04001958 if (security->name) {
1959 retval = reiserfs_security_write(th, inode, security);
1960 if (retval) {
1961 err = retval;
1962 reiserfs_check_path(&path_to_key);
1963 retval = journal_end(th, th->t_super,
1964 th->t_blocks_allocated);
1965 if (retval)
1966 err = retval;
1967 goto out_inserted_sd;
1968 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969 }
1970
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001971 reiserfs_update_sd(th, inode);
1972 reiserfs_check_path(&path_to_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001974 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975
1976/* it looks like you can easily compress these two goto targets into
1977 * one. Keeping it like this doesn't actually hurt anything, and they
1978 * are place holders for what the quota code actually needs.
1979 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001980 out_bad_inode:
1981 /* Invalidate the object, nothing was inserted yet */
1982 INODE_PKEY(inode)->k_objectid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001984 /* Quota change must be inside a transaction for journaling */
Christoph Hellwig63936dd2010-03-03 09:05:01 -05001985 dquot_free_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001987 out_end_trans:
1988 journal_end(th, th->t_super, th->t_blocks_allocated);
Jan Kara7af11682012-11-13 17:05:14 +01001989 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001990 /* Drop can be outside and it needs more credits so it's better to have it outside */
Christoph Hellwig9f754752010-03-03 09:05:05 -05001991 dquot_drop(inode);
Jan Kara7af11682012-11-13 17:05:14 +01001992 reiserfs_write_lock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001993 inode->i_flags |= S_NOQUOTA;
1994 make_bad_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001996 out_inserted_sd:
Miklos Szeredi6d6b77f2011-10-28 14:13:28 +02001997 clear_nlink(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07001998 th->t_trans_id = 0; /* so the caller can't use this handle later */
Al Viroc1eaa262008-12-30 02:03:58 -05001999 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
Jeff Mahoneyd9845612009-03-30 14:02:35 -04002000 iput(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002001 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002}
2003
2004/*
2005** finds the tail page in the page cache,
2006** reads the last block in.
2007**
2008** On success, page_result is set to a locked, pinned page, and bh_result
2009** is set to an up to date buffer for the last block in the file. returns 0.
2010**
2011** tail conversion is not done, so bh_result might not be valid for writing
2012** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2013** trying to write the block.
2014**
2015** on failure, nonzero is returned, page_result and bh_result are untouched.
2016*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04002017static int grab_tail_page(struct inode *inode,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002018 struct page **page_result,
2019 struct buffer_head **bh_result)
2020{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002022 /* we want the page with the last byte in the file,
2023 ** not the page that will hold the next byte for appending
2024 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002025 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002026 unsigned long pos = 0;
2027 unsigned long start = 0;
Jeff Mahoney995c7622009-03-30 14:02:47 -04002028 unsigned long blocksize = inode->i_sb->s_blocksize;
2029 unsigned long offset = (inode->i_size) & (PAGE_CACHE_SIZE - 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002030 struct buffer_head *bh;
2031 struct buffer_head *head;
2032 struct page *page;
2033 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002035 /* we know that we are only called with inode->i_size > 0.
2036 ** we also know that a file tail can never be as big as a block
2037 ** If i_size % blocksize == 0, our file is currently block aligned
2038 ** and it won't need converting or zeroing after a truncate.
2039 */
2040 if ((offset & (blocksize - 1)) == 0) {
2041 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 }
Jeff Mahoney995c7622009-03-30 14:02:47 -04002043 page = grab_cache_page(inode->i_mapping, index);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002044 error = -ENOMEM;
2045 if (!page) {
2046 goto out;
2047 }
2048 /* start within the page of the last block in the file */
2049 start = (offset / blocksize) * blocksize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050
Christoph Hellwigebdec242010-10-06 10:47:23 +02002051 error = __block_write_begin(page, start, offset - start,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002052 reiserfs_get_block_create_0);
2053 if (error)
2054 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002056 head = page_buffers(page);
2057 bh = head;
2058 do {
2059 if (pos >= start) {
2060 break;
2061 }
2062 bh = bh->b_this_page;
2063 pos += blocksize;
2064 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002066 if (!buffer_uptodate(bh)) {
2067 /* note, this should never happen, prepare_write should
2068 ** be taking care of this for us. If the buffer isn't up to date,
2069 ** I've screwed up the code to find the buffer, or the code to
2070 ** call prepare_write
2071 */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002072 reiserfs_error(inode->i_sb, "clm-6000",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002073 "error reading block %lu", bh->b_blocknr);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002074 error = -EIO;
2075 goto unlock;
2076 }
2077 *bh_result = bh;
2078 *page_result = page;
2079
2080 out:
2081 return error;
2082
2083 unlock:
2084 unlock_page(page);
2085 page_cache_release(page);
2086 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087}
2088
2089/*
2090** vfs version of truncate file. Must NOT be called with
2091** a transaction already started.
2092**
2093** some code taken from block_truncate_page
2094*/
Jeff Mahoney995c7622009-03-30 14:02:47 -04002095int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002096{
2097 struct reiserfs_transaction_handle th;
2098 /* we want the offset for the first byte after the end of the file */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002099 unsigned long offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2100 unsigned blocksize = inode->i_sb->s_blocksize;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002101 unsigned length;
2102 struct page *page = NULL;
2103 int error;
2104 struct buffer_head *bh = NULL;
Jeff Mahoney24996042005-12-14 14:38:05 -05002105 int err2;
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002106 int lock_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002108 lock_depth = reiserfs_write_lock_once(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109
Jeff Mahoney995c7622009-03-30 14:02:47 -04002110 if (inode->i_size > 0) {
2111 error = grab_tail_page(inode, &page, &bh);
2112 if (error) {
Jeff Mahoney0222e652009-03-30 14:02:44 -04002113 // -ENOENT means we truncated past the end of the file,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002114 // and get_block_create_0 could not find a block to read in,
2115 // which is ok.
2116 if (error != -ENOENT)
Jeff Mahoney995c7622009-03-30 14:02:47 -04002117 reiserfs_error(inode->i_sb, "clm-6001",
Jeff Mahoney0030b642009-03-30 14:02:28 -04002118 "grab_tail_page failed %d",
2119 error);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002120 page = NULL;
2121 bh = NULL;
2122 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124
Jeff Mahoney0222e652009-03-30 14:02:44 -04002125 /* so, if page != NULL, we have a buffer head for the offset at
2126 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2127 ** then we have an unformatted node. Otherwise, we have a direct item,
2128 ** and no zeroing is required on disk. We zero after the truncate,
2129 ** because the truncate might pack the item anyway
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002130 ** (it will unmap bh if it packs).
2131 */
2132 /* it is enough to reserve space in transaction for 2 balancings:
2133 one for "save" link adding and another for the first
2134 cut_from_item. 1 is for update_sd */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002135 error = journal_begin(&th, inode->i_sb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002136 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2137 if (error)
2138 goto out;
Jeff Mahoney995c7622009-03-30 14:02:47 -04002139 reiserfs_update_inode_transaction(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002140 if (update_timestamps)
2141 /* we are doing real truncate: if the system crashes before the last
2142 transaction of truncating gets committed - on reboot the file
2143 either appears truncated properly or not truncated at all */
Jeff Mahoney995c7622009-03-30 14:02:47 -04002144 add_save_link(&th, inode, 1);
2145 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002146 error =
Jeff Mahoney995c7622009-03-30 14:02:47 -04002147 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002148 if (error)
2149 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150
Jeff Mahoney24996042005-12-14 14:38:05 -05002151 /* check reiserfs_do_truncate after ending the transaction */
2152 if (err2) {
2153 error = err2;
2154 goto out;
2155 }
2156
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002157 if (update_timestamps) {
Jeff Mahoney995c7622009-03-30 14:02:47 -04002158 error = remove_save_link(inode, 1 /* truncate */);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002159 if (error)
2160 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002163 if (page) {
2164 length = offset & (blocksize - 1);
2165 /* if we are not on a block boundary */
2166 if (length) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002167 length = blocksize - length;
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002168 zero_user(page, offset, length);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002169 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2170 mark_buffer_dirty(bh);
2171 }
2172 }
2173 unlock_page(page);
2174 page_cache_release(page);
2175 }
2176
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002177 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2178
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002179 return 0;
2180 out:
2181 if (page) {
2182 unlock_page(page);
2183 page_cache_release(page);
2184 }
Frederic Weisbecker22c963a2009-04-14 05:34:24 +02002185
2186 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2187
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002188 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189}
2190
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002191static int map_block_for_writepage(struct inode *inode,
2192 struct buffer_head *bh_result,
2193 unsigned long block)
2194{
2195 struct reiserfs_transaction_handle th;
2196 int fs_gen;
2197 struct item_head tmp_ih;
2198 struct item_head *ih;
2199 struct buffer_head *bh;
2200 __le32 *item;
2201 struct cpu_key key;
2202 INITIALIZE_PATH(path);
2203 int pos_in_item;
2204 int jbegin_count = JOURNAL_PER_BALANCE_CNT;
Oleg Drokin7729ac52005-11-28 13:43:53 -08002205 loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002206 int retval;
2207 int use_get_block = 0;
2208 int bytes_copied = 0;
2209 int copy_size;
2210 int trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002212 /* catch places below that try to log something without starting a trans */
2213 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002215 if (!buffer_uptodate(bh_result)) {
2216 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 }
2218
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002219 kmap(bh_result->b_page);
2220 start_over:
2221 reiserfs_write_lock(inode->i_sb);
2222 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002224 research:
2225 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2226 if (retval != POSITION_FOUND) {
2227 use_get_block = 1;
2228 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 }
2230
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002231 bh = get_last_bh(&path);
2232 ih = get_ih(&path);
2233 item = get_item(&path);
2234 pos_in_item = path.pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002236 /* we've found an unformatted node */
2237 if (indirect_item_found(retval, ih)) {
2238 if (bytes_copied > 0) {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002239 reiserfs_warning(inode->i_sb, "clm-6002",
2240 "bytes_copied %d", bytes_copied);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002241 }
2242 if (!get_block_num(item, pos_in_item)) {
2243 /* crap, we are writing to a hole */
2244 use_get_block = 1;
2245 goto out;
2246 }
2247 set_block_dev_mapped(bh_result,
2248 get_block_num(item, pos_in_item), inode);
2249 } else if (is_direct_le_ih(ih)) {
2250 char *p;
2251 p = page_address(bh_result->b_page);
2252 p += (byte_offset - 1) & (PAGE_CACHE_SIZE - 1);
2253 copy_size = ih_item_len(ih) - pos_in_item;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002255 fs_gen = get_generation(inode->i_sb);
2256 copy_item_head(&tmp_ih, ih);
2257
2258 if (!trans_running) {
2259 /* vs-3050 is gone, no need to drop the path */
2260 retval = journal_begin(&th, inode->i_sb, jbegin_count);
2261 if (retval)
2262 goto out;
2263 reiserfs_update_inode_transaction(inode);
2264 trans_running = 1;
2265 if (fs_changed(fs_gen, inode->i_sb)
2266 && item_moved(&tmp_ih, &path)) {
2267 reiserfs_restore_prepared_buffer(inode->i_sb,
2268 bh);
2269 goto research;
2270 }
2271 }
2272
2273 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2274
2275 if (fs_changed(fs_gen, inode->i_sb)
2276 && item_moved(&tmp_ih, &path)) {
2277 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2278 goto research;
2279 }
2280
2281 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied,
2282 copy_size);
2283
2284 journal_mark_dirty(&th, inode->i_sb, bh);
2285 bytes_copied += copy_size;
2286 set_block_dev_mapped(bh_result, 0, inode);
2287
2288 /* are there still bytes left? */
2289 if (bytes_copied < bh_result->b_size &&
2290 (byte_offset + bytes_copied) < inode->i_size) {
2291 set_cpu_key_k_offset(&key,
2292 cpu_key_k_offset(&key) +
2293 copy_size);
2294 goto research;
2295 }
2296 } else {
Jeff Mahoney45b03d52009-03-30 14:02:21 -04002297 reiserfs_warning(inode->i_sb, "clm-6003",
2298 "bad item inode %lu", inode->i_ino);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002299 retval = -EIO;
2300 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002302 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002304 out:
2305 pathrelse(&path);
2306 if (trans_running) {
2307 int err = journal_end(&th, inode->i_sb, jbegin_count);
2308 if (err)
2309 retval = err;
2310 trans_running = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002312 reiserfs_write_unlock(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002314 /* this is where we fill in holes in the file. */
2315 if (use_get_block) {
2316 retval = reiserfs_get_block(inode, block, bh_result,
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002317 GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002318 | GET_BLOCK_NO_DANGLE);
2319 if (!retval) {
2320 if (!buffer_mapped(bh_result)
2321 || bh_result->b_blocknr == 0) {
2322 /* get_block failed to find a mapped unformatted node. */
2323 use_get_block = 0;
2324 goto start_over;
2325 }
2326 }
2327 }
2328 kunmap(bh_result->b_page);
2329
2330 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2331 /* we've copied data from the page into the direct item, so the
2332 * buffer in the page is now clean, mark it to reflect that.
2333 */
2334 lock_buffer(bh_result);
2335 clear_buffer_dirty(bh_result);
2336 unlock_buffer(bh_result);
2337 }
2338 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339}
2340
Jeff Mahoney0222e652009-03-30 14:02:44 -04002341/*
2342 * mason@suse.com: updated in 2.5.54 to follow the same general io
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 * start/recovery path as __block_write_full_page, along with special
2344 * code to handle reiserfs tails.
2345 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002346static int reiserfs_write_full_page(struct page *page,
2347 struct writeback_control *wbc)
2348{
2349 struct inode *inode = page->mapping->host;
2350 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT;
2351 int error = 0;
2352 unsigned long block;
Chris Masonb4c76fa2006-08-05 12:15:10 -07002353 sector_t last_block;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002354 struct buffer_head *head, *bh;
2355 int partial = 0;
2356 int nr = 0;
2357 int checked = PageChecked(page);
2358 struct reiserfs_transaction_handle th;
2359 struct super_block *s = inode->i_sb;
2360 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2361 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362
Chris Masone0e851c2006-02-01 03:06:49 -08002363 /* no logging allowed when nonblocking or from PF_MEMALLOC */
2364 if (checked && (current->flags & PF_MEMALLOC)) {
2365 redirty_page_for_writepage(wbc, page);
2366 unlock_page(page);
2367 return 0;
2368 }
2369
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002370 /* The page dirty bit is cleared before writepage is called, which
2371 * means we have to tell create_empty_buffers to make dirty buffers
2372 * The page really should be up to date at this point, so tossing
2373 * in the BH_Uptodate is just a sanity check.
2374 */
2375 if (!page_has_buffers(page)) {
2376 create_empty_buffers(page, s->s_blocksize,
2377 (1 << BH_Dirty) | (1 << BH_Uptodate));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002379 head = page_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002381 /* last page in the file, zero out any contents past the
2382 ** last byte in the file
2383 */
2384 if (page->index >= end_index) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002385 unsigned last_offset;
2386
2387 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1);
2388 /* no file contents in this page */
2389 if (page->index >= end_index + 1 || !last_offset) {
2390 unlock_page(page);
2391 return 0;
2392 }
Christoph Lametereebd2aa2008-02-04 22:28:29 -08002393 zero_user_segment(page, last_offset, PAGE_CACHE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002395 bh = head;
2396 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits);
Chris Masonb4c76fa2006-08-05 12:15:10 -07002397 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002398 /* first map all the buffers, logging any direct items we find */
2399 do {
Chris Masonb4c76fa2006-08-05 12:15:10 -07002400 if (block > last_block) {
2401 /*
2402 * This can happen when the block size is less than
2403 * the page size. The corresponding bytes in the page
2404 * were zero filled above
2405 */
2406 clear_buffer_dirty(bh);
2407 set_buffer_uptodate(bh);
2408 } else if ((checked || buffer_dirty(bh)) &&
2409 (!buffer_mapped(bh) || (buffer_mapped(bh)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002410 && bh->b_blocknr ==
2411 0))) {
2412 /* not mapped yet, or it points to a direct item, search
2413 * the btree for the mapping info, and log any direct
2414 * items found
2415 */
2416 if ((error = map_block_for_writepage(inode, bh, block))) {
2417 goto fail;
2418 }
2419 }
2420 bh = bh->b_this_page;
2421 block++;
2422 } while (bh != head);
2423
2424 /*
2425 * we start the transaction after map_block_for_writepage,
2426 * because it can create holes in the file (an unbounded operation).
2427 * starting it here, we can make a reliable estimate for how many
2428 * blocks we're going to log
2429 */
2430 if (checked) {
2431 ClearPageChecked(page);
2432 reiserfs_write_lock(s);
2433 error = journal_begin(&th, s, bh_per_page + 1);
2434 if (error) {
2435 reiserfs_write_unlock(s);
2436 goto fail;
2437 }
2438 reiserfs_update_inode_transaction(inode);
2439 }
2440 /* now go through and lock any dirty buffers on the page */
2441 do {
2442 get_bh(bh);
2443 if (!buffer_mapped(bh))
2444 continue;
2445 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2446 continue;
2447
2448 if (checked) {
2449 reiserfs_prepare_for_journal(s, bh, 1);
2450 journal_mark_dirty(&th, s, bh);
2451 continue;
2452 }
2453 /* from this point on, we know the buffer is mapped to a
2454 * real block and not a direct item
2455 */
Wu Fengguang1b430be2010-10-26 14:21:26 -07002456 if (wbc->sync_mode != WB_SYNC_NONE) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002457 lock_buffer(bh);
2458 } else {
Nick Pigginca5de402008-08-02 12:02:13 +02002459 if (!trylock_buffer(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002460 redirty_page_for_writepage(wbc, page);
2461 continue;
2462 }
2463 }
2464 if (test_clear_buffer_dirty(bh)) {
2465 mark_buffer_async_write(bh);
2466 } else {
2467 unlock_buffer(bh);
2468 }
2469 } while ((bh = bh->b_this_page) != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470
2471 if (checked) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002472 error = journal_end(&th, s, bh_per_page + 1);
2473 reiserfs_write_unlock(s);
2474 if (error)
2475 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002476 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002477 BUG_ON(PageWriteback(page));
2478 set_page_writeback(page);
2479 unlock_page(page);
2480
2481 /*
Jeff Mahoney0222e652009-03-30 14:02:44 -04002482 * since any buffer might be the only dirty buffer on the page,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002483 * the first submit_bh can bring the page out of writeback.
2484 * be careful with the buffers.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486 do {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002487 struct buffer_head *next = bh->b_this_page;
2488 if (buffer_async_write(bh)) {
2489 submit_bh(WRITE, bh);
2490 nr++;
2491 }
2492 put_bh(bh);
2493 bh = next;
2494 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002496 error = 0;
2497 done:
2498 if (nr == 0) {
2499 /*
2500 * if this page only had a direct item, it is very possible for
Jeff Mahoney0222e652009-03-30 14:02:44 -04002501 * no io to be required without there being an error. Or,
2502 * someone else could have locked them and sent them down the
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002503 * pipe without locking the page
2504 */
2505 bh = head;
2506 do {
2507 if (!buffer_uptodate(bh)) {
2508 partial = 1;
2509 break;
2510 }
2511 bh = bh->b_this_page;
2512 } while (bh != head);
2513 if (!partial)
2514 SetPageUptodate(page);
2515 end_page_writeback(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002517 return error;
2518
2519 fail:
2520 /* catches various errors, we need to make sure any valid dirty blocks
Jeff Mahoney0222e652009-03-30 14:02:44 -04002521 * get to the media. The page is currently locked and not marked for
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002522 * writeback
2523 */
2524 ClearPageUptodate(page);
2525 bh = head;
2526 do {
2527 get_bh(bh);
2528 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2529 lock_buffer(bh);
2530 mark_buffer_async_write(bh);
2531 } else {
2532 /*
2533 * clear any dirty bits that might have come from getting
2534 * attached to a dirty page
2535 */
2536 clear_buffer_dirty(bh);
2537 }
2538 bh = bh->b_this_page;
2539 } while (bh != head);
2540 SetPageError(page);
2541 BUG_ON(PageWriteback(page));
2542 set_page_writeback(page);
2543 unlock_page(page);
2544 do {
2545 struct buffer_head *next = bh->b_this_page;
2546 if (buffer_async_write(bh)) {
2547 clear_buffer_dirty(bh);
2548 submit_bh(WRITE, bh);
2549 nr++;
2550 }
2551 put_bh(bh);
2552 bh = next;
2553 } while (bh != head);
2554 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555}
2556
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002557static int reiserfs_readpage(struct file *f, struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002559 return block_read_full_page(page, reiserfs_get_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002560}
2561
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002562static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002563{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002564 struct inode *inode = page->mapping->host;
2565 reiserfs_wait_on_write_block(inode->i_sb);
2566 return reiserfs_write_full_page(page, wbc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567}
2568
Jan Karaec8e2f72009-12-17 15:27:06 -08002569static void reiserfs_truncate_failed_write(struct inode *inode)
2570{
2571 truncate_inode_pages(inode->i_mapping, inode->i_size);
2572 reiserfs_truncate_file(inode, 0);
2573}
2574
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002575static int reiserfs_write_begin(struct file *file,
2576 struct address_space *mapping,
2577 loff_t pos, unsigned len, unsigned flags,
2578 struct page **pagep, void **fsdata)
2579{
2580 struct inode *inode;
2581 struct page *page;
2582 pgoff_t index;
2583 int ret;
2584 int old_ref = 0;
2585
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002586 inode = mapping->host;
2587 *fsdata = 0;
2588 if (flags & AOP_FLAG_CONT_EXPAND &&
2589 (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2590 pos ++;
2591 *fsdata = (void *)(unsigned long)flags;
2592 }
2593
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002594 index = pos >> PAGE_CACHE_SHIFT;
Nick Piggin54566b22009-01-04 12:00:53 -08002595 page = grab_cache_page_write_begin(mapping, index, flags);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002596 if (!page)
2597 return -ENOMEM;
2598 *pagep = page;
2599
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002600 reiserfs_wait_on_write_block(inode->i_sb);
2601 fix_tail_page_for_writing(page);
2602 if (reiserfs_transaction_running(inode->i_sb)) {
2603 struct reiserfs_transaction_handle *th;
2604 th = (struct reiserfs_transaction_handle *)current->
2605 journal_info;
2606 BUG_ON(!th->t_refcount);
2607 BUG_ON(!th->t_trans_id);
2608 old_ref = th->t_refcount;
2609 th->t_refcount++;
2610 }
Christoph Hellwig6e1db882010-06-04 11:29:57 +02002611 ret = __block_write_begin(page, pos, len, reiserfs_get_block);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002612 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2613 struct reiserfs_transaction_handle *th = current->journal_info;
2614 /* this gets a little ugly. If reiserfs_get_block returned an
2615 * error and left a transacstion running, we've got to close it,
2616 * and we've got to free handle if it was a persistent transaction.
2617 *
2618 * But, if we had nested into an existing transaction, we need
2619 * to just drop the ref count on the handle.
2620 *
2621 * If old_ref == 0, the transaction is from reiserfs_get_block,
2622 * and it was a persistent trans. Otherwise, it was nested above.
2623 */
2624 if (th->t_refcount > old_ref) {
2625 if (old_ref)
2626 th->t_refcount--;
2627 else {
2628 int err;
2629 reiserfs_write_lock(inode->i_sb);
2630 err = reiserfs_end_persistent_transaction(th);
2631 reiserfs_write_unlock(inode->i_sb);
2632 if (err)
2633 ret = err;
2634 }
2635 }
2636 }
2637 if (ret) {
2638 unlock_page(page);
2639 page_cache_release(page);
Jan Karaec8e2f72009-12-17 15:27:06 -08002640 /* Truncate allocated blocks */
2641 reiserfs_truncate_failed_write(inode);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002642 }
2643 return ret;
2644}
2645
Christoph Hellwigebdec242010-10-06 10:47:23 +02002646int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002647{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002648 struct inode *inode = page->mapping->host;
2649 int ret;
2650 int old_ref = 0;
2651
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002652 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002653 reiserfs_wait_on_write_block(inode->i_sb);
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002654 reiserfs_write_lock(inode->i_sb);
2655
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002656 fix_tail_page_for_writing(page);
2657 if (reiserfs_transaction_running(inode->i_sb)) {
2658 struct reiserfs_transaction_handle *th;
2659 th = (struct reiserfs_transaction_handle *)current->
2660 journal_info;
2661 BUG_ON(!th->t_refcount);
2662 BUG_ON(!th->t_trans_id);
2663 old_ref = th->t_refcount;
2664 th->t_refcount++;
2665 }
2666
Christoph Hellwigebdec242010-10-06 10:47:23 +02002667 ret = __block_write_begin(page, from, len, reiserfs_get_block);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002668 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2669 struct reiserfs_transaction_handle *th = current->journal_info;
2670 /* this gets a little ugly. If reiserfs_get_block returned an
2671 * error and left a transacstion running, we've got to close it,
2672 * and we've got to free handle if it was a persistent transaction.
2673 *
2674 * But, if we had nested into an existing transaction, we need
2675 * to just drop the ref count on the handle.
2676 *
2677 * If old_ref == 0, the transaction is from reiserfs_get_block,
2678 * and it was a persistent trans. Otherwise, it was nested above.
2679 */
2680 if (th->t_refcount > old_ref) {
2681 if (old_ref)
2682 th->t_refcount--;
2683 else {
2684 int err;
2685 reiserfs_write_lock(inode->i_sb);
2686 err = reiserfs_end_persistent_transaction(th);
2687 reiserfs_write_unlock(inode->i_sb);
2688 if (err)
2689 ret = err;
2690 }
2691 }
2692 }
2693 return ret;
2694
2695}
2696
2697static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2698{
2699 return generic_block_bmap(as, block, reiserfs_bmap);
2700}
2701
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002702static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2703 loff_t pos, unsigned len, unsigned copied,
2704 struct page *page, void *fsdata)
2705{
2706 struct inode *inode = page->mapping->host;
2707 int ret = 0;
2708 int update_sd = 0;
2709 struct reiserfs_transaction_handle *th;
2710 unsigned start;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002711 int lock_depth = 0;
2712 bool locked = false;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002713
Vladimir Savelievf7557e82007-10-16 01:25:14 -07002714 if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2715 pos ++;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002716
2717 reiserfs_wait_on_write_block(inode->i_sb);
2718 if (reiserfs_transaction_running(inode->i_sb))
2719 th = current->journal_info;
2720 else
2721 th = NULL;
2722
2723 start = pos & (PAGE_CACHE_SIZE - 1);
2724 if (unlikely(copied < len)) {
2725 if (!PageUptodate(page))
2726 copied = 0;
2727
2728 page_zero_new_buffers(page, start + copied, start + len);
2729 }
2730 flush_dcache_page(page);
2731
2732 reiserfs_commit_page(inode, page, start, start + copied);
2733
2734 /* generic_commit_write does this for us, but does not update the
2735 ** transaction tracking stuff when the size changes. So, we have
2736 ** to do the i_size updates here.
2737 */
Jan Karaec8e2f72009-12-17 15:27:06 -08002738 if (pos + copied > inode->i_size) {
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002739 struct reiserfs_transaction_handle myth;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002740 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2741 locked = true;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002742 /* If the file have grown beyond the border where it
2743 can have a tail, unmark it as needing a tail
2744 packing */
2745 if ((have_large_tails(inode->i_sb)
2746 && inode->i_size > i_block_size(inode) * 4)
2747 || (have_small_tails(inode->i_sb)
2748 && inode->i_size > i_block_size(inode)))
2749 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2750
2751 ret = journal_begin(&myth, inode->i_sb, 1);
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002752 if (ret)
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002753 goto journal_error;
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002754
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002755 reiserfs_update_inode_transaction(inode);
Jan Karaec8e2f72009-12-17 15:27:06 -08002756 inode->i_size = pos + copied;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002757 /*
2758 * this will just nest into our transaction. It's important
2759 * to use mark_inode_dirty so the inode gets pushed around on the
2760 * dirty lists, and so that O_SYNC works as expected
2761 */
2762 mark_inode_dirty(inode);
2763 reiserfs_update_sd(&myth, inode);
2764 update_sd = 1;
2765 ret = journal_end(&myth, inode->i_sb, 1);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002766 if (ret)
2767 goto journal_error;
2768 }
2769 if (th) {
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002770 if (!locked) {
2771 lock_depth = reiserfs_write_lock_once(inode->i_sb);
2772 locked = true;
2773 }
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002774 if (!update_sd)
2775 mark_inode_dirty(inode);
2776 ret = reiserfs_end_persistent_transaction(th);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002777 if (ret)
2778 goto out;
2779 }
2780
2781 out:
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002782 if (locked)
2783 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002784 unlock_page(page);
2785 page_cache_release(page);
Jan Karaec8e2f72009-12-17 15:27:06 -08002786
2787 if (pos + len > inode->i_size)
2788 reiserfs_truncate_failed_write(inode);
2789
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002790 return ret == 0 ? copied : ret;
2791
2792 journal_error:
Frederic Weisbeckerd6f5b0a2009-05-08 14:53:52 +02002793 reiserfs_write_unlock_once(inode->i_sb, lock_depth);
2794 locked = false;
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002795 if (th) {
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002796 if (!update_sd)
2797 reiserfs_update_sd(th, inode);
2798 ret = reiserfs_end_persistent_transaction(th);
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002799 }
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07002800 goto out;
2801}
2802
2803int reiserfs_commit_write(struct file *f, struct page *page,
2804 unsigned from, unsigned to)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002805{
2806 struct inode *inode = page->mapping->host;
2807 loff_t pos = ((loff_t) page->index << PAGE_CACHE_SHIFT) + to;
2808 int ret = 0;
2809 int update_sd = 0;
2810 struct reiserfs_transaction_handle *th = NULL;
2811
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002812 reiserfs_write_unlock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002813 reiserfs_wait_on_write_block(inode->i_sb);
Frederic Weisbecker8ebc4232009-04-07 04:19:49 +02002814 reiserfs_write_lock(inode->i_sb);
2815
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002816 if (reiserfs_transaction_running(inode->i_sb)) {
2817 th = current->journal_info;
2818 }
2819 reiserfs_commit_page(inode, page, from, to);
2820
2821 /* generic_commit_write does this for us, but does not update the
2822 ** transaction tracking stuff when the size changes. So, we have
2823 ** to do the i_size updates here.
2824 */
2825 if (pos > inode->i_size) {
2826 struct reiserfs_transaction_handle myth;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002827 /* If the file have grown beyond the border where it
2828 can have a tail, unmark it as needing a tail
2829 packing */
2830 if ((have_large_tails(inode->i_sb)
2831 && inode->i_size > i_block_size(inode) * 4)
2832 || (have_small_tails(inode->i_sb)
2833 && inode->i_size > i_block_size(inode)))
2834 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2835
2836 ret = journal_begin(&myth, inode->i_sb, 1);
Frederic Weisbecker7e942772009-08-25 03:38:12 +02002837 if (ret)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002838 goto journal_error;
Frederic Weisbecker7e942772009-08-25 03:38:12 +02002839
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002840 reiserfs_update_inode_transaction(inode);
2841 inode->i_size = pos;
Chris Mason9f037832005-09-13 01:25:17 -07002842 /*
2843 * this will just nest into our transaction. It's important
2844 * to use mark_inode_dirty so the inode gets pushed around on the
2845 * dirty lists, and so that O_SYNC works as expected
2846 */
2847 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002848 reiserfs_update_sd(&myth, inode);
2849 update_sd = 1;
2850 ret = journal_end(&myth, inode->i_sb, 1);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002851 if (ret)
2852 goto journal_error;
2853 }
2854 if (th) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002855 if (!update_sd)
Chris Mason9f037832005-09-13 01:25:17 -07002856 mark_inode_dirty(inode);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002857 ret = reiserfs_end_persistent_transaction(th);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002858 if (ret)
2859 goto out;
2860 }
2861
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002862 out:
2863 return ret;
2864
2865 journal_error:
2866 if (th) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002867 if (!update_sd)
2868 reiserfs_update_sd(th, inode);
2869 ret = reiserfs_end_persistent_transaction(th);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002870 }
2871
2872 return ret;
2873}
2874
2875void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
2876{
2877 if (reiserfs_attrs(inode->i_sb)) {
2878 if (sd_attrs & REISERFS_SYNC_FL)
2879 inode->i_flags |= S_SYNC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002881 inode->i_flags &= ~S_SYNC;
2882 if (sd_attrs & REISERFS_IMMUTABLE_FL)
2883 inode->i_flags |= S_IMMUTABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002885 inode->i_flags &= ~S_IMMUTABLE;
2886 if (sd_attrs & REISERFS_APPEND_FL)
2887 inode->i_flags |= S_APPEND;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002888 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002889 inode->i_flags &= ~S_APPEND;
2890 if (sd_attrs & REISERFS_NOATIME_FL)
2891 inode->i_flags |= S_NOATIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892 else
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002893 inode->i_flags &= ~S_NOATIME;
2894 if (sd_attrs & REISERFS_NOTAIL_FL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002895 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2896 else
2897 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2898 }
2899}
2900
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002901void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002903 if (reiserfs_attrs(inode->i_sb)) {
2904 if (inode->i_flags & S_IMMUTABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2906 else
2907 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002908 if (inode->i_flags & S_SYNC)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002909 *sd_attrs |= REISERFS_SYNC_FL;
2910 else
2911 *sd_attrs &= ~REISERFS_SYNC_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002912 if (inode->i_flags & S_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913 *sd_attrs |= REISERFS_NOATIME_FL;
2914 else
2915 *sd_attrs &= ~REISERFS_NOATIME_FL;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002916 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002917 *sd_attrs |= REISERFS_NOTAIL_FL;
2918 else
2919 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2920 }
2921}
2922
2923/* decide if this buffer needs to stay around for data logging or ordered
2924** write purposes
2925*/
2926static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2927{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002928 int ret = 1;
2929 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002930
Chris Masond62b1b82006-02-01 03:06:47 -08002931 lock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002932 spin_lock(&j->j_dirty_buffers_lock);
2933 if (!buffer_mapped(bh)) {
2934 goto free_jh;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002935 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002936 /* the page is locked, and the only places that log a data buffer
2937 * also lock the page.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002938 */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002939 if (reiserfs_file_data_log(inode)) {
2940 /*
2941 * very conservative, leave the buffer pinned if
2942 * anyone might need it.
2943 */
2944 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2945 ret = 0;
2946 }
Chris Masond62b1b82006-02-01 03:06:47 -08002947 } else if (buffer_dirty(bh)) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002948 struct reiserfs_journal_list *jl;
2949 struct reiserfs_jh *jh = bh->b_private;
2950
2951 /* why is this safe?
2952 * reiserfs_setattr updates i_size in the on disk
2953 * stat data before allowing vmtruncate to be called.
2954 *
2955 * If buffer was put onto the ordered list for this
2956 * transaction, we know for sure either this transaction
2957 * or an older one already has updated i_size on disk,
2958 * and this ordered data won't be referenced in the file
2959 * if we crash.
2960 *
2961 * if the buffer was put onto the ordered list for an older
2962 * transaction, we need to leave it around
2963 */
2964 if (jh && (jl = jh->jl)
2965 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2966 ret = 0;
2967 }
2968 free_jh:
2969 if (ret && bh->b_private) {
2970 reiserfs_free_jh(bh);
2971 }
2972 spin_unlock(&j->j_dirty_buffers_lock);
Chris Masond62b1b82006-02-01 03:06:47 -08002973 unlock_buffer(bh);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002974 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975}
2976
2977/* clm -- taken from fs/buffer.c:block_invalidate_page */
Lukas Czernerd47992f2013-05-21 23:17:23 -04002978static void reiserfs_invalidatepage(struct page *page, unsigned int offset,
2979 unsigned int length)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002981 struct buffer_head *head, *bh, *next;
2982 struct inode *inode = page->mapping->host;
2983 unsigned int curr_off = 0;
Lukas Czernerbad54832013-05-21 23:58:51 -04002984 unsigned int stop = offset + length;
2985 int partial_page = (offset || length < PAGE_CACHE_SIZE);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002986 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002988 BUG_ON(!PageLocked(page));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002989
Lukas Czernerbad54832013-05-21 23:58:51 -04002990 if (!partial_page)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002991 ClearPageChecked(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002992
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002993 if (!page_has_buffers(page))
2994 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002995
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07002996 head = page_buffers(page);
2997 bh = head;
2998 do {
2999 unsigned int next_off = curr_off + bh->b_size;
3000 next = bh->b_this_page;
3001
Lukas Czernerbad54832013-05-21 23:58:51 -04003002 if (next_off > stop)
3003 goto out;
3004
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003005 /*
3006 * is this block fully invalidated?
3007 */
3008 if (offset <= curr_off) {
3009 if (invalidatepage_can_drop(inode, bh))
3010 reiserfs_unmap_buffer(bh);
3011 else
3012 ret = 0;
3013 }
3014 curr_off = next_off;
3015 bh = next;
3016 } while (bh != head);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003017
3018 /*
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003019 * We release buffers only if the entire page is being invalidated.
3020 * The get_block cached value has been unconditionally invalidated,
3021 * so real IO is not possible anymore.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003022 */
Lukas Czernerbad54832013-05-21 23:58:51 -04003023 if (!partial_page && ret) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003024 ret = try_to_release_page(page, 0);
NeilBrown2ff28e22006-03-26 01:37:18 -08003025 /* maybe should BUG_ON(!ret); - neilb */
3026 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003027 out:
NeilBrown2ff28e22006-03-26 01:37:18 -08003028 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003029}
3030
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003031static int reiserfs_set_page_dirty(struct page *page)
3032{
3033 struct inode *inode = page->mapping->host;
3034 if (reiserfs_file_data_log(inode)) {
3035 SetPageChecked(page);
3036 return __set_page_dirty_nobuffers(page);
3037 }
3038 return __set_page_dirty_buffers(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039}
3040
3041/*
3042 * Returns 1 if the page's buffers were dropped. The page is locked.
3043 *
3044 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3045 * in the buffers at page_buffers(page).
3046 *
3047 * even in -o notail mode, we can't be sure an old mount without -o notail
3048 * didn't create files with tails.
3049 */
Al Viro27496a82005-10-21 03:20:48 -04003050static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003052 struct inode *inode = page->mapping->host;
3053 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3054 struct buffer_head *head;
3055 struct buffer_head *bh;
3056 int ret = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003058 WARN_ON(PageChecked(page));
3059 spin_lock(&j->j_dirty_buffers_lock);
3060 head = page_buffers(page);
3061 bh = head;
3062 do {
3063 if (bh->b_private) {
3064 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3065 reiserfs_free_jh(bh);
3066 } else {
3067 ret = 0;
3068 break;
3069 }
3070 }
3071 bh = bh->b_this_page;
3072 } while (bh != head);
3073 if (ret)
3074 ret = try_to_free_buffers(page);
3075 spin_unlock(&j->j_dirty_buffers_lock);
3076 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077}
3078
3079/* We thank Mingming Cao for helping us understand in great detail what
3080 to do in this section of the code. */
3081static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003082 const struct iovec *iov, loff_t offset,
3083 unsigned long nr_segs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003084{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003085 struct file *file = iocb->ki_filp;
3086 struct inode *inode = file->f_mapping->host;
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003087 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003088
Christoph Hellwigaacfc19c2011-06-24 14:29:47 -04003089 ret = blockdev_direct_IO(rw, iocb, inode, iov, offset, nr_segs,
3090 reiserfs_get_blocks_direct_io);
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003091
3092 /*
3093 * In case of error extending write may have instantiated a few
3094 * blocks outside i_size. Trim these off again.
3095 */
3096 if (unlikely((rw & WRITE) && ret < 0)) {
3097 loff_t isize = i_size_read(inode);
3098 loff_t end = offset + iov_length(iov, nr_segs);
3099
Marco Stornellicfac4b42012-12-15 11:47:31 +01003100 if ((end > isize) && inode_newsize_ok(inode, isize) == 0) {
3101 truncate_setsize(inode, isize);
3102 reiserfs_vfs_truncate_file(inode);
3103 }
Christoph Hellwigeafdc7d2010-06-04 11:29:53 +02003104 }
3105
3106 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003107}
3108
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003109int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3110{
3111 struct inode *inode = dentry->d_inode;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003112 unsigned int ia_valid;
Frederic Weisbecker5fe1533f2010-01-04 22:04:01 +01003113 int depth;
3114 int error;
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003115
Christoph Hellwigdb78b872010-06-04 11:30:03 +02003116 error = inode_change_ok(inode, attr);
3117 if (error)
3118 return error;
3119
Jeff Laytoncdd6fe62007-10-18 03:05:19 -07003120 /* must be turned off for recursive notify_change calls */
3121 ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3122
Dmitry Monakhov12755622010-04-08 22:04:20 +04003123 if (is_quota_modification(inode, attr))
Christoph Hellwig871a2932010-03-03 09:05:07 -05003124 dquot_initialize(inode);
Jan Kara7af11682012-11-13 17:05:14 +01003125 depth = reiserfs_write_lock_once(inode->i_sb);
Dmitry Monakhov12755622010-04-08 22:04:20 +04003126 if (attr->ia_valid & ATTR_SIZE) {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003127 /* version 2 items will be caught by the s_maxbytes check
3128 ** done for us in vmtruncate
3129 */
3130 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3131 attr->ia_size > MAX_NON_LFS) {
3132 error = -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003133 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003134 }
Christoph Hellwig562c72a2011-06-24 14:29:45 -04003135
3136 inode_dio_wait(inode);
3137
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003138 /* fill in hole pointers in the expanding truncate case. */
3139 if (attr->ia_size > inode->i_size) {
Vladimir Savelievf7557e82007-10-16 01:25:14 -07003140 error = generic_cont_expand_simple(inode, attr->ia_size);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003141 if (REISERFS_I(inode)->i_prealloc_count > 0) {
3142 int err;
3143 struct reiserfs_transaction_handle th;
3144 /* we're changing at most 2 bitmaps, inode + super */
3145 err = journal_begin(&th, inode->i_sb, 4);
3146 if (!err) {
3147 reiserfs_discard_prealloc(&th, inode);
3148 err = journal_end(&th, inode->i_sb, 4);
3149 }
3150 if (err)
3151 error = err;
3152 }
3153 if (error)
3154 goto out;
Vladimir Savelievdd535a52006-07-01 04:36:32 -07003155 /*
3156 * file size is changed, ctime and mtime are
3157 * to be updated
3158 */
3159 attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003160 }
3161 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003162
Eric W. Biedermandf814652012-02-07 16:30:06 -08003163 if ((((attr->ia_valid & ATTR_UID) && (from_kuid(&init_user_ns, attr->ia_uid) & ~0xffff)) ||
3164 ((attr->ia_valid & ATTR_GID) && (from_kgid(&init_user_ns, attr->ia_gid) & ~0xffff))) &&
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003165 (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3166 /* stat data of format v3.5 has 16 bit uid and gid */
3167 error = -EINVAL;
3168 goto out;
3169 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003170
Eric W. Biedermandf814652012-02-07 16:30:06 -08003171 if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
3172 (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
Christoph Hellwig10257742010-06-04 11:30:02 +02003173 struct reiserfs_transaction_handle th;
3174 int jbegin_count =
3175 2 *
3176 (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3177 REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3178 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003179
Christoph Hellwig10257742010-06-04 11:30:02 +02003180 error = reiserfs_chown_xattrs(inode, attr);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003181
Christoph Hellwig10257742010-06-04 11:30:02 +02003182 if (error)
3183 return error;
3184
3185 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */
3186 error = journal_begin(&th, inode->i_sb, jbegin_count);
3187 if (error)
3188 goto out;
Jan Kara7af11682012-11-13 17:05:14 +01003189 reiserfs_write_unlock_once(inode->i_sb, depth);
Christoph Hellwig10257742010-06-04 11:30:02 +02003190 error = dquot_transfer(inode, attr);
Jan Kara7af11682012-11-13 17:05:14 +01003191 depth = reiserfs_write_lock_once(inode->i_sb);
Christoph Hellwig10257742010-06-04 11:30:02 +02003192 if (error) {
3193 journal_end(&th, inode->i_sb, jbegin_count);
3194 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003195 }
Christoph Hellwig10257742010-06-04 11:30:02 +02003196
3197 /* Update corresponding info in inode so that everything is in
3198 * one transaction */
3199 if (attr->ia_valid & ATTR_UID)
3200 inode->i_uid = attr->ia_uid;
3201 if (attr->ia_valid & ATTR_GID)
3202 inode->i_gid = attr->ia_gid;
3203 mark_inode_dirty(inode);
3204 error = journal_end(&th, inode->i_sb, jbegin_count);
3205 if (error)
3206 goto out;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003207 }
3208
Christoph Hellwig10257742010-06-04 11:30:02 +02003209 /*
3210 * Relax the lock here, as it might truncate the
3211 * inode pages and wait for inode pages locks.
3212 * To release such page lock, the owner needs the
3213 * reiserfs lock
3214 */
3215 reiserfs_write_unlock_once(inode->i_sb, depth);
3216 if ((attr->ia_valid & ATTR_SIZE) &&
Marco Stornellicfac4b42012-12-15 11:47:31 +01003217 attr->ia_size != i_size_read(inode)) {
3218 error = inode_newsize_ok(inode, attr->ia_size);
3219 if (!error) {
3220 truncate_setsize(inode, attr->ia_size);
3221 reiserfs_vfs_truncate_file(inode);
3222 }
3223 }
Christoph Hellwig10257742010-06-04 11:30:02 +02003224
3225 if (!error) {
3226 setattr_copy(inode, attr);
3227 mark_inode_dirty(inode);
3228 }
3229 depth = reiserfs_write_lock_once(inode->i_sb);
3230
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003231 if (!error && reiserfs_posixacl(inode->i_sb)) {
3232 if (attr->ia_valid & ATTR_MODE)
3233 error = reiserfs_acl_chmod(inode);
3234 }
3235
3236 out:
Frederic Weisbecker5fe1533f2010-01-04 22:04:01 +01003237 reiserfs_write_unlock_once(inode->i_sb, depth);
3238
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003239 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003240}
3241
Christoph Hellwigf5e54d62006-06-28 04:26:44 -07003242const struct address_space_operations reiserfs_address_space_operations = {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003243 .writepage = reiserfs_writepage,
3244 .readpage = reiserfs_readpage,
3245 .readpages = reiserfs_readpages,
3246 .releasepage = reiserfs_releasepage,
3247 .invalidatepage = reiserfs_invalidatepage,
Vladimir Savelievba9d8ce2007-10-16 01:25:14 -07003248 .write_begin = reiserfs_write_begin,
3249 .write_end = reiserfs_write_end,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -07003250 .bmap = reiserfs_aop_bmap,
3251 .direct_IO = reiserfs_direct_IO,
3252 .set_page_dirty = reiserfs_set_page_dirty,
3253};