aboutsummaryrefslogtreecommitdiffstats
path: root/fs/reiserfs/inode.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/reiserfs/inode.c')
-rw-r--r--fs/reiserfs/inode.c1206
1 files changed, 707 insertions, 499 deletions
diff --git a/fs/reiserfs/inode.c b/fs/reiserfs/inode.c
index bc8b8009897d..e3ca04894919 100644
--- a/fs/reiserfs/inode.c
+++ b/fs/reiserfs/inode.c
@@ -25,7 +25,10 @@ int reiserfs_commit_write(struct file *f, struct page *page,
25 25
26void reiserfs_evict_inode(struct inode *inode) 26void reiserfs_evict_inode(struct inode *inode)
27{ 27{
28 /* We need blocks for transaction + (user+group) quota update (possibly delete) */ 28 /*
29 * We need blocks for transaction + (user+group) quota
30 * update (possibly delete)
31 */
29 int jbegin_count = 32 int jbegin_count =
30 JOURNAL_PER_BALANCE_CNT * 2 + 33 JOURNAL_PER_BALANCE_CNT * 2 +
31 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb); 34 2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
@@ -39,8 +42,12 @@ void reiserfs_evict_inode(struct inode *inode)
39 if (inode->i_nlink) 42 if (inode->i_nlink)
40 goto no_delete; 43 goto no_delete;
41 44
42 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */ 45 /*
43 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */ 46 * The = 0 happens when we abort creating a new inode
47 * for some reason like lack of space..
48 * also handles bad_inode case
49 */
50 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) {
44 51
45 reiserfs_delete_xattrs(inode); 52 reiserfs_delete_xattrs(inode);
46 53
@@ -54,34 +61,43 @@ void reiserfs_evict_inode(struct inode *inode)
54 61
55 err = reiserfs_delete_object(&th, inode); 62 err = reiserfs_delete_object(&th, inode);
56 63
57 /* Do quota update inside a transaction for journaled quotas. We must do that 64 /*
58 * after delete_object so that quota updates go into the same transaction as 65 * Do quota update inside a transaction for journaled quotas.
59 * stat data deletion */ 66 * We must do that after delete_object so that quota updates
67 * go into the same transaction as stat data deletion
68 */
60 if (!err) { 69 if (!err) {
61 int depth = reiserfs_write_unlock_nested(inode->i_sb); 70 int depth = reiserfs_write_unlock_nested(inode->i_sb);
62 dquot_free_inode(inode); 71 dquot_free_inode(inode);
63 reiserfs_write_lock_nested(inode->i_sb, depth); 72 reiserfs_write_lock_nested(inode->i_sb, depth);
64 } 73 }
65 74
66 if (journal_end(&th, inode->i_sb, jbegin_count)) 75 if (journal_end(&th))
67 goto out; 76 goto out;
68 77
69 /* check return value from reiserfs_delete_object after 78 /*
79 * check return value from reiserfs_delete_object after
70 * ending the transaction 80 * ending the transaction
71 */ 81 */
72 if (err) 82 if (err)
73 goto out; 83 goto out;
74 84
75 /* all items of file are deleted, so we can remove "save" link */ 85 /*
76 remove_save_link(inode, 0 /* not truncate */ ); /* we can't do anything 86 * all items of file are deleted, so we can remove
77 * about an error here */ 87 * "save" link
88 * we can't do anything about an error here
89 */
90 remove_save_link(inode, 0 /* not truncate */);
78out: 91out:
79 reiserfs_write_unlock(inode->i_sb); 92 reiserfs_write_unlock(inode->i_sb);
80 } else { 93 } else {
81 /* no object items are in the tree */ 94 /* no object items are in the tree */
82 ; 95 ;
83 } 96 }
84 clear_inode(inode); /* note this must go after the journal_end to prevent deadlock */ 97
98 /* note this must go after the journal_end to prevent deadlock */
99 clear_inode(inode);
100
85 dquot_drop(inode); 101 dquot_drop(inode);
86 inode->i_blocks = 0; 102 inode->i_blocks = 0;
87 return; 103 return;
@@ -103,8 +119,10 @@ static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
103 key->key_length = length; 119 key->key_length = length;
104} 120}
105 121
106/* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set 122/*
107 offset and type of key */ 123 * take base of inode_key (it comes from inode always) (dirid, objectid)
124 * and version from an inode, set offset and type of key
125 */
108void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset, 126void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
109 int type, int length) 127 int type, int length)
110{ 128{
@@ -114,9 +132,7 @@ void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
114 length); 132 length);
115} 133}
116 134
117// 135/* when key is 0, do not set version and short key */
118// when key is 0, do not set version and short key
119//
120inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key, 136inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
121 int version, 137 int version,
122 loff_t offset, int type, int length, 138 loff_t offset, int type, int length,
@@ -132,43 +148,47 @@ inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
132 set_le_ih_k_type(ih, type); 148 set_le_ih_k_type(ih, type);
133 put_ih_item_len(ih, length); 149 put_ih_item_len(ih, length);
134 /* set_ih_free_space (ih, 0); */ 150 /* set_ih_free_space (ih, 0); */
135 // for directory items it is entry count, for directs and stat 151 /*
136 // datas - 0xffff, for indirects - 0 152 * for directory items it is entry count, for directs and stat
153 * datas - 0xffff, for indirects - 0
154 */
137 put_ih_entry_count(ih, entry_count); 155 put_ih_entry_count(ih, entry_count);
138} 156}
139 157
140// 158/*
141// FIXME: we might cache recently accessed indirect item 159 * FIXME: we might cache recently accessed indirect item
142 160 * Ugh. Not too eager for that....
143// Ugh. Not too eager for that.... 161 * I cut the code until such time as I see a convincing argument (benchmark).
144// I cut the code until such time as I see a convincing argument (benchmark). 162 * I don't want a bloated inode struct..., and I don't like code complexity....
145// I don't want a bloated inode struct..., and I don't like code complexity.... 163 */
146
147/* cutting the code is fine, since it really isn't in use yet and is easy
148** to add back in. But, Vladimir has a really good idea here. Think
149** about what happens for reading a file. For each page,
150** The VFS layer calls reiserfs_readpage, who searches the tree to find
151** an indirect item. This indirect item has X number of pointers, where
152** X is a big number if we've done the block allocation right. But,
153** we only use one or two of these pointers during each call to readpage,
154** needlessly researching again later on.
155**
156** The size of the cache could be dynamic based on the size of the file.
157**
158** I'd also like to see us cache the location the stat data item, since
159** we are needlessly researching for that frequently.
160**
161** --chris
162*/
163 164
164/* If this page has a file tail in it, and 165/*
165** it was read in by get_block_create_0, the page data is valid, 166 * cutting the code is fine, since it really isn't in use yet and is easy
166** but tail is still sitting in a direct item, and we can't write to 167 * to add back in. But, Vladimir has a really good idea here. Think
167** it. So, look through this page, and check all the mapped buffers 168 * about what happens for reading a file. For each page,
168** to make sure they have valid block numbers. Any that don't need 169 * The VFS layer calls reiserfs_readpage, who searches the tree to find
169** to be unmapped, so that __block_write_begin will correctly call 170 * an indirect item. This indirect item has X number of pointers, where
170** reiserfs_get_block to convert the tail into an unformatted node 171 * X is a big number if we've done the block allocation right. But,
171*/ 172 * we only use one or two of these pointers during each call to readpage,
173 * needlessly researching again later on.
174 *
175 * The size of the cache could be dynamic based on the size of the file.
176 *
177 * I'd also like to see us cache the location the stat data item, since
178 * we are needlessly researching for that frequently.
179 *
180 * --chris
181 */
182
183/*
184 * If this page has a file tail in it, and
185 * it was read in by get_block_create_0, the page data is valid,
186 * but tail is still sitting in a direct item, and we can't write to
187 * it. So, look through this page, and check all the mapped buffers
188 * to make sure they have valid block numbers. Any that don't need
189 * to be unmapped, so that __block_write_begin will correctly call
190 * reiserfs_get_block to convert the tail into an unformatted node
191 */
172static inline void fix_tail_page_for_writing(struct page *page) 192static inline void fix_tail_page_for_writing(struct page *page)
173{ 193{
174 struct buffer_head *head, *next, *bh; 194 struct buffer_head *head, *next, *bh;
@@ -186,8 +206,10 @@ static inline void fix_tail_page_for_writing(struct page *page)
186 } 206 }
187} 207}
188 208
189/* reiserfs_get_block does not need to allocate a block only if it has been 209/*
190 done already or non-hole position has been found in the indirect item */ 210 * reiserfs_get_block does not need to allocate a block only if it has been
211 * done already or non-hole position has been found in the indirect item
212 */
191static inline int allocation_needed(int retval, b_blocknr_t allocated, 213static inline int allocation_needed(int retval, b_blocknr_t allocated,
192 struct item_head *ih, 214 struct item_head *ih,
193 __le32 * item, int pos_in_item) 215 __le32 * item, int pos_in_item)
@@ -211,14 +233,16 @@ static inline void set_block_dev_mapped(struct buffer_head *bh,
211 map_bh(bh, inode->i_sb, block); 233 map_bh(bh, inode->i_sb, block);
212} 234}
213 235
214// 236/*
215// files which were created in the earlier version can not be longer, 237 * files which were created in the earlier version can not be longer,
216// than 2 gb 238 * than 2 gb
217// 239 */
218static int file_capable(struct inode *inode, sector_t block) 240static int file_capable(struct inode *inode, sector_t block)
219{ 241{
220 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 || // it is new file. 242 /* it is new file. */
221 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb 243 if (get_inode_item_key_version(inode) != KEY_FORMAT_3_5 ||
244 /* old file, but 'block' is inside of 2gb */
245 block < (1 << (31 - inode->i_sb->s_blocksize_bits)))
222 return 1; 246 return 1;
223 247
224 return 0; 248 return 0;
@@ -228,7 +252,6 @@ static int restart_transaction(struct reiserfs_transaction_handle *th,
228 struct inode *inode, struct treepath *path) 252 struct inode *inode, struct treepath *path)
229{ 253{
230 struct super_block *s = th->t_super; 254 struct super_block *s = th->t_super;
231 int len = th->t_blocks_allocated;
232 int err; 255 int err;
233 256
234 BUG_ON(!th->t_trans_id); 257 BUG_ON(!th->t_trans_id);
@@ -241,7 +264,7 @@ static int restart_transaction(struct reiserfs_transaction_handle *th,
241 return 0; 264 return 0;
242 } 265 }
243 reiserfs_update_sd(th, inode); 266 reiserfs_update_sd(th, inode);
244 err = journal_end(th, s, len); 267 err = journal_end(th);
245 if (!err) { 268 if (!err) {
246 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6); 269 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
247 if (!err) 270 if (!err)
@@ -250,14 +273,14 @@ static int restart_transaction(struct reiserfs_transaction_handle *th,
250 return err; 273 return err;
251} 274}
252 275
253// it is called by get_block when create == 0. Returns block number 276/*
254// for 'block'-th logical block of file. When it hits direct item it 277 * it is called by get_block when create == 0. Returns block number
255// returns 0 (being called from bmap) or read direct item into piece 278 * for 'block'-th logical block of file. When it hits direct item it
256// of page (bh_result) 279 * returns 0 (being called from bmap) or read direct item into piece
257 280 * of page (bh_result)
258// Please improve the english/clarity in the comment above, as it is 281 * Please improve the english/clarity in the comment above, as it is
259// hard to understand. 282 * hard to understand.
260 283 */
261static int _get_block_create_0(struct inode *inode, sector_t block, 284static int _get_block_create_0(struct inode *inode, sector_t block,
262 struct buffer_head *bh_result, int args) 285 struct buffer_head *bh_result, int args)
263{ 286{
@@ -273,7 +296,7 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
273 int done = 0; 296 int done = 0;
274 unsigned long offset; 297 unsigned long offset;
275 298
276 // prepare the key to look for the 'block'-th block of file 299 /* prepare the key to look for the 'block'-th block of file */
277 make_cpu_key(&key, inode, 300 make_cpu_key(&key, inode,
278 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY, 301 (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
279 3); 302 3);
@@ -285,23 +308,28 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
285 kunmap(bh_result->b_page); 308 kunmap(bh_result->b_page);
286 if (result == IO_ERROR) 309 if (result == IO_ERROR)
287 return -EIO; 310 return -EIO;
288 // We do not return -ENOENT if there is a hole but page is uptodate, because it means 311 /*
289 // That there is some MMAPED data associated with it that is yet to be written to disk. 312 * We do not return -ENOENT if there is a hole but page is
313 * uptodate, because it means that there is some MMAPED data
314 * associated with it that is yet to be written to disk.
315 */
290 if ((args & GET_BLOCK_NO_HOLE) 316 if ((args & GET_BLOCK_NO_HOLE)
291 && !PageUptodate(bh_result->b_page)) { 317 && !PageUptodate(bh_result->b_page)) {
292 return -ENOENT; 318 return -ENOENT;
293 } 319 }
294 return 0; 320 return 0;
295 } 321 }
296 // 322
297 bh = get_last_bh(&path); 323 bh = get_last_bh(&path);
298 ih = get_ih(&path); 324 ih = tp_item_head(&path);
299 if (is_indirect_le_ih(ih)) { 325 if (is_indirect_le_ih(ih)) {
300 __le32 *ind_item = (__le32 *) B_I_PITEM(bh, ih); 326 __le32 *ind_item = (__le32 *) ih_item_body(bh, ih);
301 327
302 /* FIXME: here we could cache indirect item or part of it in 328 /*
303 the inode to avoid search_by_key in case of subsequent 329 * FIXME: here we could cache indirect item or part of it in
304 access to file */ 330 * the inode to avoid search_by_key in case of subsequent
331 * access to file
332 */
305 blocknr = get_block_num(ind_item, path.pos_in_item); 333 blocknr = get_block_num(ind_item, path.pos_in_item);
306 ret = 0; 334 ret = 0;
307 if (blocknr) { 335 if (blocknr) {
@@ -311,8 +339,12 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
311 set_buffer_boundary(bh_result); 339 set_buffer_boundary(bh_result);
312 } 340 }
313 } else 341 } else
314 // We do not return -ENOENT if there is a hole but page is uptodate, because it means 342 /*
315 // That there is some MMAPED data associated with it that is yet to be written to disk. 343 * We do not return -ENOENT if there is a hole but
344 * page is uptodate, because it means that there is
345 * some MMAPED data associated with it that is
346 * yet to be written to disk.
347 */
316 if ((args & GET_BLOCK_NO_HOLE) 348 if ((args & GET_BLOCK_NO_HOLE)
317 && !PageUptodate(bh_result->b_page)) { 349 && !PageUptodate(bh_result->b_page)) {
318 ret = -ENOENT; 350 ret = -ENOENT;
@@ -323,41 +355,45 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
323 kunmap(bh_result->b_page); 355 kunmap(bh_result->b_page);
324 return ret; 356 return ret;
325 } 357 }
326 // requested data are in direct item(s) 358 /* requested data are in direct item(s) */
327 if (!(args & GET_BLOCK_READ_DIRECT)) { 359 if (!(args & GET_BLOCK_READ_DIRECT)) {
328 // we are called by bmap. FIXME: we can not map block of file 360 /*
329 // when it is stored in direct item(s) 361 * we are called by bmap. FIXME: we can not map block of file
362 * when it is stored in direct item(s)
363 */
330 pathrelse(&path); 364 pathrelse(&path);
331 if (p) 365 if (p)
332 kunmap(bh_result->b_page); 366 kunmap(bh_result->b_page);
333 return -ENOENT; 367 return -ENOENT;
334 } 368 }
335 369
336 /* if we've got a direct item, and the buffer or page was uptodate, 370 /*
337 ** we don't want to pull data off disk again. skip to the 371 * if we've got a direct item, and the buffer or page was uptodate,
338 ** end, where we map the buffer and return 372 * we don't want to pull data off disk again. skip to the
373 * end, where we map the buffer and return
339 */ 374 */
340 if (buffer_uptodate(bh_result)) { 375 if (buffer_uptodate(bh_result)) {
341 goto finished; 376 goto finished;
342 } else 377 } else
343 /* 378 /*
344 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date 379 * grab_tail_page can trigger calls to reiserfs_get_block on
345 ** pages without any buffers. If the page is up to date, we don't want 380 * up to date pages without any buffers. If the page is up
346 ** read old data off disk. Set the up to date bit on the buffer instead 381 * to date, we don't want read old data off disk. Set the up
347 ** and jump to the end 382 * to date bit on the buffer instead and jump to the end
348 */ 383 */
349 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) { 384 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
350 set_buffer_uptodate(bh_result); 385 set_buffer_uptodate(bh_result);
351 goto finished; 386 goto finished;
352 } 387 }
353 // read file tail into part of page 388 /* read file tail into part of page */
354 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1); 389 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1);
355 copy_item_head(&tmp_ih, ih); 390 copy_item_head(&tmp_ih, ih);
356 391
357 /* we only want to kmap if we are reading the tail into the page. 392 /*
358 ** this is not the common case, so we don't kmap until we are 393 * we only want to kmap if we are reading the tail into the page.
359 ** sure we need to. But, this means the item might move if 394 * this is not the common case, so we don't kmap until we are
360 ** kmap schedules 395 * sure we need to. But, this means the item might move if
396 * kmap schedules
361 */ 397 */
362 if (!p) 398 if (!p)
363 p = (char *)kmap(bh_result->b_page); 399 p = (char *)kmap(bh_result->b_page);
@@ -368,10 +404,11 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
368 if (!is_direct_le_ih(ih)) { 404 if (!is_direct_le_ih(ih)) {
369 BUG(); 405 BUG();
370 } 406 }
371 /* make sure we don't read more bytes than actually exist in 407 /*
372 ** the file. This can happen in odd cases where i_size isn't 408 * make sure we don't read more bytes than actually exist in
373 ** correct, and when direct item padding results in a few 409 * the file. This can happen in odd cases where i_size isn't
374 ** extra bytes at the end of the direct item 410 * correct, and when direct item padding results in a few
411 * extra bytes at the end of the direct item
375 */ 412 */
376 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size) 413 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
377 break; 414 break;
@@ -383,40 +420,43 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
383 } else { 420 } else {
384 chars = ih_item_len(ih) - path.pos_in_item; 421 chars = ih_item_len(ih) - path.pos_in_item;
385 } 422 }
386 memcpy(p, B_I_PITEM(bh, ih) + path.pos_in_item, chars); 423 memcpy(p, ih_item_body(bh, ih) + path.pos_in_item, chars);
387 424
388 if (done) 425 if (done)
389 break; 426 break;
390 427
391 p += chars; 428 p += chars;
392 429
430 /*
431 * we done, if read direct item is not the last item of
432 * node FIXME: we could try to check right delimiting key
433 * to see whether direct item continues in the right
434 * neighbor or rely on i_size
435 */
393 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1)) 436 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
394 // we done, if read direct item is not the last item of
395 // node FIXME: we could try to check right delimiting key
396 // to see whether direct item continues in the right
397 // neighbor or rely on i_size
398 break; 437 break;
399 438
400 // update key to look for the next piece 439 /* update key to look for the next piece */
401 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars); 440 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
402 result = search_for_position_by_key(inode->i_sb, &key, &path); 441 result = search_for_position_by_key(inode->i_sb, &key, &path);
403 if (result != POSITION_FOUND) 442 if (result != POSITION_FOUND)
404 // i/o error most likely 443 /* i/o error most likely */
405 break; 444 break;
406 bh = get_last_bh(&path); 445 bh = get_last_bh(&path);
407 ih = get_ih(&path); 446 ih = tp_item_head(&path);
408 } while (1); 447 } while (1);
409 448
410 flush_dcache_page(bh_result->b_page); 449 flush_dcache_page(bh_result->b_page);
411 kunmap(bh_result->b_page); 450 kunmap(bh_result->b_page);
412 451
413 finished: 452finished:
414 pathrelse(&path); 453 pathrelse(&path);
415 454
416 if (result == IO_ERROR) 455 if (result == IO_ERROR)
417 return -EIO; 456 return -EIO;
418 457
419 /* this buffer has valid data, but isn't valid for io. mapping it to 458 /*
459 * this buffer has valid data, but isn't valid for io. mapping it to
420 * block #0 tells the rest of reiserfs it just has a tail in it 460 * block #0 tells the rest of reiserfs it just has a tail in it
421 */ 461 */
422 map_bh(bh_result, inode->i_sb, 0); 462 map_bh(bh_result, inode->i_sb, 0);
@@ -424,8 +464,10 @@ static int _get_block_create_0(struct inode *inode, sector_t block,
424 return 0; 464 return 0;
425} 465}
426 466
427// this is called to create file map. So, _get_block_create_0 will not 467/*
428// read direct item 468 * this is called to create file map. So, _get_block_create_0 will not
469 * read direct item
470 */
429static int reiserfs_bmap(struct inode *inode, sector_t block, 471static int reiserfs_bmap(struct inode *inode, sector_t block,
430 struct buffer_head *bh_result, int create) 472 struct buffer_head *bh_result, int create)
431{ 473{
@@ -439,22 +481,23 @@ static int reiserfs_bmap(struct inode *inode, sector_t block,
439 return 0; 481 return 0;
440} 482}
441 483
442/* special version of get_block that is only used by grab_tail_page right 484/*
443** now. It is sent to __block_write_begin, and when you try to get a 485 * special version of get_block that is only used by grab_tail_page right
444** block past the end of the file (or a block from a hole) it returns 486 * now. It is sent to __block_write_begin, and when you try to get a
445** -ENOENT instead of a valid buffer. __block_write_begin expects to 487 * block past the end of the file (or a block from a hole) it returns
446** be able to do i/o on the buffers returned, unless an error value 488 * -ENOENT instead of a valid buffer. __block_write_begin expects to
447** is also returned. 489 * be able to do i/o on the buffers returned, unless an error value
448** 490 * is also returned.
449** So, this allows __block_write_begin to be used for reading a single block 491 *
450** in a page. Where it does not produce a valid page for holes, or past the 492 * So, this allows __block_write_begin to be used for reading a single block
451** end of the file. This turns out to be exactly what we need for reading 493 * in a page. Where it does not produce a valid page for holes, or past the
452** tails for conversion. 494 * end of the file. This turns out to be exactly what we need for reading
453** 495 * tails for conversion.
454** The point of the wrapper is forcing a certain value for create, even 496 *
455** though the VFS layer is calling this function with create==1. If you 497 * The point of the wrapper is forcing a certain value for create, even
456** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block, 498 * though the VFS layer is calling this function with create==1. If you
457** don't use this function. 499 * don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
500 * don't use this function.
458*/ 501*/
459static int reiserfs_get_block_create_0(struct inode *inode, sector_t block, 502static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
460 struct buffer_head *bh_result, 503 struct buffer_head *bh_result,
@@ -463,8 +506,10 @@ static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
463 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE); 506 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
464} 507}
465 508
466/* This is special helper for reiserfs_get_block in case we are executing 509/*
467 direct_IO request. */ 510 * This is special helper for reiserfs_get_block in case we are executing
511 * direct_IO request.
512 */
468static int reiserfs_get_blocks_direct_io(struct inode *inode, 513static int reiserfs_get_blocks_direct_io(struct inode *inode,
469 sector_t iblock, 514 sector_t iblock,
470 struct buffer_head *bh_result, 515 struct buffer_head *bh_result,
@@ -474,9 +519,11 @@ static int reiserfs_get_blocks_direct_io(struct inode *inode,
474 519
475 bh_result->b_page = NULL; 520 bh_result->b_page = NULL;
476 521
477 /* We set the b_size before reiserfs_get_block call since it is 522 /*
478 referenced in convert_tail_for_hole() that may be called from 523 * We set the b_size before reiserfs_get_block call since it is
479 reiserfs_get_block() */ 524 * referenced in convert_tail_for_hole() that may be called from
525 * reiserfs_get_block()
526 */
480 bh_result->b_size = (1 << inode->i_blkbits); 527 bh_result->b_size = (1 << inode->i_blkbits);
481 528
482 ret = reiserfs_get_block(inode, iblock, bh_result, 529 ret = reiserfs_get_block(inode, iblock, bh_result,
@@ -486,14 +533,18 @@ static int reiserfs_get_blocks_direct_io(struct inode *inode,
486 533
487 /* don't allow direct io onto tail pages */ 534 /* don't allow direct io onto tail pages */
488 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) { 535 if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
489 /* make sure future calls to the direct io funcs for this offset 536 /*
490 ** in the file fail by unmapping the buffer 537 * make sure future calls to the direct io funcs for this
538 * offset in the file fail by unmapping the buffer
491 */ 539 */
492 clear_buffer_mapped(bh_result); 540 clear_buffer_mapped(bh_result);
493 ret = -EINVAL; 541 ret = -EINVAL;
494 } 542 }
495 /* Possible unpacked tail. Flush the data before pages have 543
496 disappeared */ 544 /*
545 * Possible unpacked tail. Flush the data before pages have
546 * disappeared
547 */
497 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) { 548 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
498 int err; 549 int err;
499 550
@@ -507,20 +558,20 @@ static int reiserfs_get_blocks_direct_io(struct inode *inode,
507 if (err < 0) 558 if (err < 0)
508 ret = err; 559 ret = err;
509 } 560 }
510 out: 561out:
511 return ret; 562 return ret;
512} 563}
513 564
514/* 565/*
515** helper function for when reiserfs_get_block is called for a hole 566 * helper function for when reiserfs_get_block is called for a hole
516** but the file tail is still in a direct item 567 * but the file tail is still in a direct item
517** bh_result is the buffer head for the hole 568 * bh_result is the buffer head for the hole
518** tail_offset is the offset of the start of the tail in the file 569 * tail_offset is the offset of the start of the tail in the file
519** 570 *
520** This calls prepare_write, which will start a new transaction 571 * This calls prepare_write, which will start a new transaction
521** you should not be in a transaction, or have any paths held when you 572 * you should not be in a transaction, or have any paths held when you
522** call this. 573 * call this.
523*/ 574 */
524static int convert_tail_for_hole(struct inode *inode, 575static int convert_tail_for_hole(struct inode *inode,
525 struct buffer_head *bh_result, 576 struct buffer_head *bh_result,
526 loff_t tail_offset) 577 loff_t tail_offset)
@@ -540,9 +591,10 @@ static int convert_tail_for_hole(struct inode *inode,
540 tail_end = (tail_start | (bh_result->b_size - 1)) + 1; 591 tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
541 592
542 index = tail_offset >> PAGE_CACHE_SHIFT; 593 index = tail_offset >> PAGE_CACHE_SHIFT;
543 /* hole_page can be zero in case of direct_io, we are sure 594 /*
544 that we cannot get here if we write with O_DIRECT into 595 * hole_page can be zero in case of direct_io, we are sure
545 tail page */ 596 * that we cannot get here if we write with O_DIRECT into tail page
597 */
546 if (!hole_page || index != hole_page->index) { 598 if (!hole_page || index != hole_page->index) {
547 tail_page = grab_cache_page(inode->i_mapping, index); 599 tail_page = grab_cache_page(inode->i_mapping, index);
548 retval = -ENOMEM; 600 retval = -ENOMEM;
@@ -553,14 +605,15 @@ static int convert_tail_for_hole(struct inode *inode,
553 tail_page = hole_page; 605 tail_page = hole_page;
554 } 606 }
555 607
556 /* we don't have to make sure the conversion did not happen while 608 /*
557 ** we were locking the page because anyone that could convert 609 * we don't have to make sure the conversion did not happen while
558 ** must first take i_mutex. 610 * we were locking the page because anyone that could convert
559 ** 611 * must first take i_mutex.
560 ** We must fix the tail page for writing because it might have buffers 612 *
561 ** that are mapped, but have a block number of 0. This indicates tail 613 * We must fix the tail page for writing because it might have buffers
562 ** data that has been read directly into the page, and 614 * that are mapped, but have a block number of 0. This indicates tail
563 ** __block_write_begin won't trigger a get_block in this case. 615 * data that has been read directly into the page, and
616 * __block_write_begin won't trigger a get_block in this case.
564 */ 617 */
565 fix_tail_page_for_writing(tail_page); 618 fix_tail_page_for_writing(tail_page);
566 retval = __reiserfs_write_begin(tail_page, tail_start, 619 retval = __reiserfs_write_begin(tail_page, tail_start,
@@ -573,12 +626,12 @@ static int convert_tail_for_hole(struct inode *inode,
573 626
574 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end); 627 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
575 628
576 unlock: 629unlock:
577 if (tail_page != hole_page) { 630 if (tail_page != hole_page) {
578 unlock_page(tail_page); 631 unlock_page(tail_page);
579 page_cache_release(tail_page); 632 page_cache_release(tail_page);
580 } 633 }
581 out: 634out:
582 return retval; 635 return retval;
583} 636}
584 637
@@ -604,7 +657,8 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
604 struct buffer_head *bh_result, int create) 657 struct buffer_head *bh_result, int create)
605{ 658{
606 int repeat, retval = 0; 659 int repeat, retval = 0;
607 b_blocknr_t allocated_block_nr = 0; // b_blocknr_t is (unsigned) 32 bit int 660 /* b_blocknr_t is (unsigned) 32 bit int*/
661 b_blocknr_t allocated_block_nr = 0;
608 INITIALIZE_PATH(path); 662 INITIALIZE_PATH(path);
609 int pos_in_item; 663 int pos_in_item;
610 struct cpu_key key; 664 struct cpu_key key;
@@ -614,12 +668,14 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
614 int done; 668 int done;
615 int fs_gen; 669 int fs_gen;
616 struct reiserfs_transaction_handle *th = NULL; 670 struct reiserfs_transaction_handle *th = NULL;
617 /* space reserved in transaction batch: 671 /*
618 . 3 balancings in direct->indirect conversion 672 * space reserved in transaction batch:
619 . 1 block involved into reiserfs_update_sd() 673 * . 3 balancings in direct->indirect conversion
620 XXX in practically impossible worst case direct2indirect() 674 * . 1 block involved into reiserfs_update_sd()
621 can incur (much) more than 3 balancings. 675 * XXX in practically impossible worst case direct2indirect()
622 quota update for user, group */ 676 * can incur (much) more than 3 balancings.
677 * quota update for user, group
678 */
623 int jbegin_count = 679 int jbegin_count =
624 JOURNAL_PER_BALANCE_CNT * 3 + 1 + 680 JOURNAL_PER_BALANCE_CNT * 3 + 1 +
625 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb); 681 2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
@@ -636,8 +692,9 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
636 return -EFBIG; 692 return -EFBIG;
637 } 693 }
638 694
639 /* if !create, we aren't changing the FS, so we don't need to 695 /*
640 ** log anything, so we don't need to start a transaction 696 * if !create, we aren't changing the FS, so we don't need to
697 * log anything, so we don't need to start a transaction
641 */ 698 */
642 if (!(create & GET_BLOCK_CREATE)) { 699 if (!(create & GET_BLOCK_CREATE)) {
643 int ret; 700 int ret;
@@ -647,6 +704,7 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
647 reiserfs_write_unlock(inode->i_sb); 704 reiserfs_write_unlock(inode->i_sb);
648 return ret; 705 return ret;
649 } 706 }
707
650 /* 708 /*
651 * if we're already in a transaction, make sure to close 709 * if we're already in a transaction, make sure to close
652 * any new transactions we start in this func 710 * any new transactions we start in this func
@@ -655,8 +713,10 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
655 reiserfs_transaction_running(inode->i_sb)) 713 reiserfs_transaction_running(inode->i_sb))
656 dangle = 0; 714 dangle = 0;
657 715
658 /* If file is of such a size, that it might have a tail and tails are enabled 716 /*
659 ** we should mark it as possibly needing tail packing on close 717 * If file is of such a size, that it might have a tail and
718 * tails are enabled we should mark it as possibly needing
719 * tail packing on close
660 */ 720 */
661 if ((have_large_tails(inode->i_sb) 721 if ((have_large_tails(inode->i_sb)
662 && inode->i_size < i_block_size(inode) * 4) 722 && inode->i_size < i_block_size(inode) * 4)
@@ -667,7 +727,7 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
667 /* set the key of the first byte in the 'block'-th block of file */ 727 /* set the key of the first byte in the 'block'-th block of file */
668 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ ); 728 make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
669 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) { 729 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
670 start_trans: 730start_trans:
671 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count); 731 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
672 if (!th) { 732 if (!th) {
673 retval = -ENOMEM; 733 retval = -ENOMEM;
@@ -675,7 +735,7 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
675 } 735 }
676 reiserfs_update_inode_transaction(inode); 736 reiserfs_update_inode_transaction(inode);
677 } 737 }
678 research: 738research:
679 739
680 retval = search_for_position_by_key(inode->i_sb, &key, &path); 740 retval = search_for_position_by_key(inode->i_sb, &key, &path);
681 if (retval == IO_ERROR) { 741 if (retval == IO_ERROR) {
@@ -684,8 +744,8 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
684 } 744 }
685 745
686 bh = get_last_bh(&path); 746 bh = get_last_bh(&path);
687 ih = get_ih(&path); 747 ih = tp_item_head(&path);
688 item = get_item(&path); 748 item = tp_item_body(&path);
689 pos_in_item = path.pos_in_item; 749 pos_in_item = path.pos_in_item;
690 750
691 fs_gen = get_generation(inode->i_sb); 751 fs_gen = get_generation(inode->i_sb);
@@ -703,11 +763,12 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
703 _allocate_block(th, block, inode, &allocated_block_nr, 763 _allocate_block(th, block, inode, &allocated_block_nr,
704 &path, create); 764 &path, create);
705 765
766 /*
767 * restart the transaction to give the journal a chance to free
768 * some blocks. releases the path, so we have to go back to
769 * research if we succeed on the second try
770 */
706 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) { 771 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
707 /* restart the transaction to give the journal a chance to free
708 ** some blocks. releases the path, so we have to go back to
709 ** research if we succeed on the second try
710 */
711 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1; 772 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
712 retval = restart_transaction(th, inode, &path); 773 retval = restart_transaction(th, inode, &path);
713 if (retval) 774 if (retval)
@@ -734,9 +795,11 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
734 795
735 if (indirect_item_found(retval, ih)) { 796 if (indirect_item_found(retval, ih)) {
736 b_blocknr_t unfm_ptr; 797 b_blocknr_t unfm_ptr;
737 /* 'block'-th block is in the file already (there is 798 /*
738 corresponding cell in some indirect item). But it may be 799 * 'block'-th block is in the file already (there is
739 zero unformatted node pointer (hole) */ 800 * corresponding cell in some indirect item). But it may be
801 * zero unformatted node pointer (hole)
802 */
740 unfm_ptr = get_block_num(item, pos_in_item); 803 unfm_ptr = get_block_num(item, pos_in_item);
741 if (unfm_ptr == 0) { 804 if (unfm_ptr == 0) {
742 /* use allocated block to plug the hole */ 805 /* use allocated block to plug the hole */
@@ -753,7 +816,7 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
753 reiserfs_add_ordered_list(inode, bh_result); 816 reiserfs_add_ordered_list(inode, bh_result);
754 put_block_num(item, pos_in_item, allocated_block_nr); 817 put_block_num(item, pos_in_item, allocated_block_nr);
755 unfm_ptr = allocated_block_nr; 818 unfm_ptr = allocated_block_nr;
756 journal_mark_dirty(th, inode->i_sb, bh); 819 journal_mark_dirty(th, bh);
757 reiserfs_update_sd(th, inode); 820 reiserfs_update_sd(th, inode);
758 } 821 }
759 set_block_dev_mapped(bh_result, unfm_ptr, inode); 822 set_block_dev_mapped(bh_result, unfm_ptr, inode);
@@ -764,9 +827,10 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
764 827
765 reiserfs_write_unlock(inode->i_sb); 828 reiserfs_write_unlock(inode->i_sb);
766 829
767 /* the item was found, so new blocks were not added to the file 830 /*
768 ** there is no need to make sure the inode is updated with this 831 * the item was found, so new blocks were not added to the file
769 ** transaction 832 * there is no need to make sure the inode is updated with this
833 * transaction
770 */ 834 */
771 return retval; 835 return retval;
772 } 836 }
@@ -776,9 +840,11 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
776 goto start_trans; 840 goto start_trans;
777 } 841 }
778 842
779 /* desired position is not found or is in the direct item. We have 843 /*
780 to append file with holes up to 'block'-th block converting 844 * desired position is not found or is in the direct item. We have
781 direct items to indirect one if necessary */ 845 * to append file with holes up to 'block'-th block converting
846 * direct items to indirect one if necessary
847 */
782 done = 0; 848 done = 0;
783 do { 849 do {
784 if (is_statdata_le_ih(ih)) { 850 if (is_statdata_le_ih(ih)) {
@@ -790,16 +856,18 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
790 TYPE_INDIRECT, UNFM_P_SIZE, 856 TYPE_INDIRECT, UNFM_P_SIZE,
791 0 /* free_space */ ); 857 0 /* free_space */ );
792 858
859 /*
860 * we are going to add 'block'-th block to the file.
861 * Use allocated block for that
862 */
793 if (cpu_key_k_offset(&key) == 1) { 863 if (cpu_key_k_offset(&key) == 1) {
794 /* we are going to add 'block'-th block to the file. Use
795 allocated block for that */
796 unp = cpu_to_le32(allocated_block_nr); 864 unp = cpu_to_le32(allocated_block_nr);
797 set_block_dev_mapped(bh_result, 865 set_block_dev_mapped(bh_result,
798 allocated_block_nr, inode); 866 allocated_block_nr, inode);
799 set_buffer_new(bh_result); 867 set_buffer_new(bh_result);
800 done = 1; 868 done = 1;
801 } 869 }
802 tmp_key = key; // ;) 870 tmp_key = key; /* ;) */
803 set_cpu_key_k_offset(&tmp_key, 1); 871 set_cpu_key_k_offset(&tmp_key, 1);
804 PATH_LAST_POSITION(&path)++; 872 PATH_LAST_POSITION(&path)++;
805 873
@@ -809,9 +877,12 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
809 if (retval) { 877 if (retval) {
810 reiserfs_free_block(th, inode, 878 reiserfs_free_block(th, inode,
811 allocated_block_nr, 1); 879 allocated_block_nr, 1);
812 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST 880 /*
881 * retval == -ENOSPC, -EDQUOT or -EIO
882 * or -EEXIST
883 */
884 goto failure;
813 } 885 }
814 //mark_tail_converted (inode);
815 } else if (is_direct_le_ih(ih)) { 886 } else if (is_direct_le_ih(ih)) {
816 /* direct item has to be converted */ 887 /* direct item has to be converted */
817 loff_t tail_offset; 888 loff_t tail_offset;
@@ -819,18 +890,24 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
819 tail_offset = 890 tail_offset =
820 ((le_ih_k_offset(ih) - 891 ((le_ih_k_offset(ih) -
821 1) & ~(inode->i_sb->s_blocksize - 1)) + 1; 892 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
893
894 /*
895 * direct item we just found fits into block we have
896 * to map. Convert it into unformatted node: use
897 * bh_result for the conversion
898 */
822 if (tail_offset == cpu_key_k_offset(&key)) { 899 if (tail_offset == cpu_key_k_offset(&key)) {
823 /* direct item we just found fits into block we have
824 to map. Convert it into unformatted node: use
825 bh_result for the conversion */
826 set_block_dev_mapped(bh_result, 900 set_block_dev_mapped(bh_result,
827 allocated_block_nr, inode); 901 allocated_block_nr, inode);
828 unbh = bh_result; 902 unbh = bh_result;
829 done = 1; 903 done = 1;
830 } else { 904 } else {
831 /* we have to padd file tail stored in direct item(s) 905 /*
832 up to block size and convert it to unformatted 906 * we have to pad file tail stored in direct
833 node. FIXME: this should also get into page cache */ 907 * item(s) up to block size and convert it
908 * to unformatted node. FIXME: this should
909 * also get into page cache
910 */
834 911
835 pathrelse(&path); 912 pathrelse(&path);
836 /* 913 /*
@@ -859,7 +936,10 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
859 inode->i_ino, 936 inode->i_ino,
860 retval); 937 retval);
861 if (allocated_block_nr) { 938 if (allocated_block_nr) {
862 /* the bitmap, the super, and the stat data == 3 */ 939 /*
940 * the bitmap, the super,
941 * and the stat data == 3
942 */
863 if (!th) 943 if (!th)
864 th = reiserfs_persistent_transaction(inode->i_sb, 3); 944 th = reiserfs_persistent_transaction(inode->i_sb, 3);
865 if (th) 945 if (th)
@@ -881,43 +961,57 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
881 allocated_block_nr, 1); 961 allocated_block_nr, 1);
882 goto failure; 962 goto failure;
883 } 963 }
884 /* it is important the set_buffer_uptodate is done after 964 /*
885 ** the direct2indirect. The buffer might contain valid 965 * it is important the set_buffer_uptodate is done
886 ** data newer than the data on disk (read by readpage, changed, 966 * after the direct2indirect. The buffer might
887 ** and then sent here by writepage). direct2indirect needs 967 * contain valid data newer than the data on disk
888 ** to know if unbh was already up to date, so it can decide 968 * (read by readpage, changed, and then sent here by
889 ** if the data in unbh needs to be replaced with data from 969 * writepage). direct2indirect needs to know if unbh
890 ** the disk 970 * was already up to date, so it can decide if the
971 * data in unbh needs to be replaced with data from
972 * the disk
891 */ 973 */
892 set_buffer_uptodate(unbh); 974 set_buffer_uptodate(unbh);
893 975
894 /* unbh->b_page == NULL in case of DIRECT_IO request, this means 976 /*
895 buffer will disappear shortly, so it should not be added to 977 * unbh->b_page == NULL in case of DIRECT_IO request,
978 * this means buffer will disappear shortly, so it
979 * should not be added to
896 */ 980 */
897 if (unbh->b_page) { 981 if (unbh->b_page) {
898 /* we've converted the tail, so we must 982 /*
899 ** flush unbh before the transaction commits 983 * we've converted the tail, so we must
984 * flush unbh before the transaction commits
900 */ 985 */
901 reiserfs_add_tail_list(inode, unbh); 986 reiserfs_add_tail_list(inode, unbh);
902 987
903 /* mark it dirty now to prevent commit_write from adding 988 /*
904 ** this buffer to the inode's dirty buffer list 989 * mark it dirty now to prevent commit_write
990 * from adding this buffer to the inode's
991 * dirty buffer list
905 */ 992 */
906 /* 993 /*
907 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty(). 994 * AKPM: changed __mark_buffer_dirty to
908 * It's still atomic, but it sets the page dirty too, 995 * mark_buffer_dirty(). It's still atomic,
909 * which makes it eligible for writeback at any time by the 996 * but it sets the page dirty too, which makes
910 * VM (which was also the case with __mark_buffer_dirty()) 997 * it eligible for writeback at any time by the
998 * VM (which was also the case with
999 * __mark_buffer_dirty())
911 */ 1000 */
912 mark_buffer_dirty(unbh); 1001 mark_buffer_dirty(unbh);
913 } 1002 }
914 } else { 1003 } else {
915 /* append indirect item with holes if needed, when appending 1004 /*
916 pointer to 'block'-th block use block, which is already 1005 * append indirect item with holes if needed, when
917 allocated */ 1006 * appending pointer to 'block'-th block use block,
1007 * which is already allocated
1008 */
918 struct cpu_key tmp_key; 1009 struct cpu_key tmp_key;
919 unp_t unf_single = 0; // We use this in case we need to allocate only 1010 /*
920 // one block which is a fastpath 1011 * We use this in case we need to allocate
1012 * only one block which is a fastpath
1013 */
1014 unp_t unf_single = 0;
921 unp_t *un; 1015 unp_t *un;
922 __u64 max_to_insert = 1016 __u64 max_to_insert =
923 MAX_ITEM_LEN(inode->i_sb->s_blocksize) / 1017 MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
@@ -926,14 +1020,17 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
926 1020
927 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE, 1021 RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
928 "vs-804: invalid position for append"); 1022 "vs-804: invalid position for append");
929 /* indirect item has to be appended, set up key of that position */ 1023 /*
1024 * indirect item has to be appended,
1025 * set up key of that position
1026 * (key type is unimportant)
1027 */
930 make_cpu_key(&tmp_key, inode, 1028 make_cpu_key(&tmp_key, inode,
931 le_key_k_offset(version, 1029 le_key_k_offset(version,
932 &(ih->ih_key)) + 1030 &ih->ih_key) +
933 op_bytes_number(ih, 1031 op_bytes_number(ih,
934 inode->i_sb->s_blocksize), 1032 inode->i_sb->s_blocksize),
935 //pos_in_item * inode->i_sb->s_blocksize, 1033 TYPE_INDIRECT, 3);
936 TYPE_INDIRECT, 3); // key type is unimportant
937 1034
938 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key), 1035 RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
939 "green-805: invalid offset"); 1036 "green-805: invalid offset");
@@ -954,8 +1051,10 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
954 } 1051 }
955 } 1052 }
956 if (blocks_needed <= max_to_insert) { 1053 if (blocks_needed <= max_to_insert) {
957 /* we are going to add target block to the file. Use allocated 1054 /*
958 block for that */ 1055 * we are going to add target block to
1056 * the file. Use allocated block for that
1057 */
959 un[blocks_needed - 1] = 1058 un[blocks_needed - 1] =
960 cpu_to_le32(allocated_block_nr); 1059 cpu_to_le32(allocated_block_nr);
961 set_block_dev_mapped(bh_result, 1060 set_block_dev_mapped(bh_result,
@@ -964,8 +1063,11 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
964 done = 1; 1063 done = 1;
965 } else { 1064 } else {
966 /* paste hole to the indirect item */ 1065 /* paste hole to the indirect item */
967 /* If kmalloc failed, max_to_insert becomes zero and it means we 1066 /*
968 only have space for one block */ 1067 * If kmalloc failed, max_to_insert becomes
1068 * zero and it means we only have space for
1069 * one block
1070 */
969 blocks_needed = 1071 blocks_needed =
970 max_to_insert ? max_to_insert : 1; 1072 max_to_insert ? max_to_insert : 1;
971 } 1073 }
@@ -984,9 +1086,12 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
984 goto failure; 1086 goto failure;
985 } 1087 }
986 if (!done) { 1088 if (!done) {
987 /* We need to mark new file size in case this function will be 1089 /*
988 interrupted/aborted later on. And we may do this only for 1090 * We need to mark new file size in case
989 holes. */ 1091 * this function will be interrupted/aborted
1092 * later on. And we may do this only for
1093 * holes.
1094 */
990 inode->i_size += 1095 inode->i_size +=
991 inode->i_sb->s_blocksize * blocks_needed; 1096 inode->i_sb->s_blocksize * blocks_needed;
992 } 1097 }
@@ -995,13 +1100,13 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
995 if (done == 1) 1100 if (done == 1)
996 break; 1101 break;
997 1102
998 /* this loop could log more blocks than we had originally asked 1103 /*
999 ** for. So, we have to allow the transaction to end if it is 1104 * this loop could log more blocks than we had originally
1000 ** too big or too full. Update the inode so things are 1105 * asked for. So, we have to allow the transaction to end
1001 ** consistent if we crash before the function returns 1106 * if it is too big or too full. Update the inode so things
1002 ** 1107 * are consistent if we crash before the function returns
1003 ** release the path so that anybody waiting on the path before 1108 * release the path so that anybody waiting on the path before
1004 ** ending their transaction will be able to continue. 1109 * ending their transaction will be able to continue.
1005 */ 1110 */
1006 if (journal_transaction_should_end(th, th->t_blocks_allocated)) { 1111 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1007 retval = restart_transaction(th, inode, &path); 1112 retval = restart_transaction(th, inode, &path);
@@ -1031,14 +1136,14 @@ int reiserfs_get_block(struct inode *inode, sector_t block,
1031 goto failure; 1136 goto failure;
1032 } 1137 }
1033 bh = get_last_bh(&path); 1138 bh = get_last_bh(&path);
1034 ih = get_ih(&path); 1139 ih = tp_item_head(&path);
1035 item = get_item(&path); 1140 item = tp_item_body(&path);
1036 pos_in_item = path.pos_in_item; 1141 pos_in_item = path.pos_in_item;
1037 } while (1); 1142 } while (1);
1038 1143
1039 retval = 0; 1144 retval = 0;
1040 1145
1041 failure: 1146failure:
1042 if (th && (!dangle || (retval && !th->t_trans_id))) { 1147 if (th && (!dangle || (retval && !th->t_trans_id))) {
1043 int err; 1148 int err;
1044 if (th->t_trans_id) 1149 if (th->t_trans_id)
@@ -1060,8 +1165,10 @@ reiserfs_readpages(struct file *file, struct address_space *mapping,
1060 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block); 1165 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1061} 1166}
1062 1167
1063/* Compute real number of used bytes by file 1168/*
1064 * Following three functions can go away when we'll have enough space in stat item 1169 * Compute real number of used bytes by file
1170 * Following three functions can go away when we'll have enough space in
1171 * stat item
1065 */ 1172 */
1066static int real_space_diff(struct inode *inode, int sd_size) 1173static int real_space_diff(struct inode *inode, int sd_size)
1067{ 1174{
@@ -1071,13 +1178,14 @@ static int real_space_diff(struct inode *inode, int sd_size)
1071 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) 1178 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1072 return sd_size; 1179 return sd_size;
1073 1180
1074 /* End of file is also in full block with indirect reference, so round 1181 /*
1075 ** up to the next block. 1182 * End of file is also in full block with indirect reference, so round
1076 ** 1183 * up to the next block.
1077 ** there is just no way to know if the tail is actually packed 1184 *
1078 ** on the file, so we have to assume it isn't. When we pack the 1185 * there is just no way to know if the tail is actually packed
1079 ** tail, we add 4 bytes to pretend there really is an unformatted 1186 * on the file, so we have to assume it isn't. When we pack the
1080 ** node pointer 1187 * tail, we add 4 bytes to pretend there really is an unformatted
1188 * node pointer
1081 */ 1189 */
1082 bytes = 1190 bytes =
1083 ((inode->i_size + 1191 ((inode->i_size +
@@ -1108,36 +1216,36 @@ static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1108 bytes += (loff_t) 511; 1216 bytes += (loff_t) 511;
1109 } 1217 }
1110 1218
1111 /* files from before the quota patch might i_blocks such that 1219 /*
1112 ** bytes < real_space. Deal with that here to prevent it from 1220 * files from before the quota patch might i_blocks such that
1113 ** going negative. 1221 * bytes < real_space. Deal with that here to prevent it from
1222 * going negative.
1114 */ 1223 */
1115 if (bytes < real_space) 1224 if (bytes < real_space)
1116 return 0; 1225 return 0;
1117 return (bytes - real_space) >> 9; 1226 return (bytes - real_space) >> 9;
1118} 1227}
1119 1228
1120// 1229/*
1121// BAD: new directories have stat data of new type and all other items 1230 * BAD: new directories have stat data of new type and all other items
1122// of old type. Version stored in the inode says about body items, so 1231 * of old type. Version stored in the inode says about body items, so
1123// in update_stat_data we can not rely on inode, but have to check 1232 * in update_stat_data we can not rely on inode, but have to check
1124// item version directly 1233 * item version directly
1125// 1234 */
1126 1235
1127// called by read_locked_inode 1236/* called by read_locked_inode */
1128static void init_inode(struct inode *inode, struct treepath *path) 1237static void init_inode(struct inode *inode, struct treepath *path)
1129{ 1238{
1130 struct buffer_head *bh; 1239 struct buffer_head *bh;
1131 struct item_head *ih; 1240 struct item_head *ih;
1132 __u32 rdev; 1241 __u32 rdev;
1133 //int version = ITEM_VERSION_1;
1134 1242
1135 bh = PATH_PLAST_BUFFER(path); 1243 bh = PATH_PLAST_BUFFER(path);
1136 ih = PATH_PITEM_HEAD(path); 1244 ih = tp_item_head(path);
1137 1245
1138 copy_key(INODE_PKEY(inode), &(ih->ih_key)); 1246 copy_key(INODE_PKEY(inode), &ih->ih_key);
1139 1247
1140 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list)); 1248 INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
1141 REISERFS_I(inode)->i_flags = 0; 1249 REISERFS_I(inode)->i_flags = 0;
1142 REISERFS_I(inode)->i_prealloc_block = 0; 1250 REISERFS_I(inode)->i_prealloc_block = 0;
1143 REISERFS_I(inode)->i_prealloc_count = 0; 1251 REISERFS_I(inode)->i_prealloc_count = 0;
@@ -1147,7 +1255,7 @@ static void init_inode(struct inode *inode, struct treepath *path)
1147 1255
1148 if (stat_data_v1(ih)) { 1256 if (stat_data_v1(ih)) {
1149 struct stat_data_v1 *sd = 1257 struct stat_data_v1 *sd =
1150 (struct stat_data_v1 *)B_I_PITEM(bh, ih); 1258 (struct stat_data_v1 *)ih_item_body(bh, ih);
1151 unsigned long blocks; 1259 unsigned long blocks;
1152 1260
1153 set_inode_item_key_version(inode, KEY_FORMAT_3_5); 1261 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
@@ -1168,20 +1276,26 @@ static void init_inode(struct inode *inode, struct treepath *path)
1168 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id); 1276 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1169 blocks = (inode->i_size + 511) >> 9; 1277 blocks = (inode->i_size + 511) >> 9;
1170 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9); 1278 blocks = _ROUND_UP(blocks, inode->i_sb->s_blocksize >> 9);
1279
1280 /*
1281 * there was a bug in <=3.5.23 when i_blocks could take
1282 * negative values. Starting from 3.5.17 this value could
1283 * even be stored in stat data. For such files we set
1284 * i_blocks based on file size. Just 2 notes: this can be
1285 * wrong for sparse files. On-disk value will be only
1286 * updated if file's inode will ever change
1287 */
1171 if (inode->i_blocks > blocks) { 1288 if (inode->i_blocks > blocks) {
1172 // there was a bug in <=3.5.23 when i_blocks could take negative
1173 // values. Starting from 3.5.17 this value could even be stored in
1174 // stat data. For such files we set i_blocks based on file
1175 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1176 // only updated if file's inode will ever change
1177 inode->i_blocks = blocks; 1289 inode->i_blocks = blocks;
1178 } 1290 }
1179 1291
1180 rdev = sd_v1_rdev(sd); 1292 rdev = sd_v1_rdev(sd);
1181 REISERFS_I(inode)->i_first_direct_byte = 1293 REISERFS_I(inode)->i_first_direct_byte =
1182 sd_v1_first_direct_byte(sd); 1294 sd_v1_first_direct_byte(sd);
1183 /* an early bug in the quota code can give us an odd number for the 1295
1184 ** block count. This is incorrect, fix it here. 1296 /*
1297 * an early bug in the quota code can give us an odd
1298 * number for the block count. This is incorrect, fix it here.
1185 */ 1299 */
1186 if (inode->i_blocks & 1) { 1300 if (inode->i_blocks & 1) {
1187 inode->i_blocks++; 1301 inode->i_blocks++;
@@ -1189,13 +1303,17 @@ static void init_inode(struct inode *inode, struct treepath *path)
1189 inode_set_bytes(inode, 1303 inode_set_bytes(inode,
1190 to_real_used_space(inode, inode->i_blocks, 1304 to_real_used_space(inode, inode->i_blocks,
1191 SD_V1_SIZE)); 1305 SD_V1_SIZE));
1192 /* nopack is initially zero for v1 objects. For v2 objects, 1306 /*
1193 nopack is initialised from sd_attrs */ 1307 * nopack is initially zero for v1 objects. For v2 objects,
1308 * nopack is initialised from sd_attrs
1309 */
1194 REISERFS_I(inode)->i_flags &= ~i_nopack_mask; 1310 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1195 } else { 1311 } else {
1196 // new stat data found, but object may have old items 1312 /*
1197 // (directories and symlinks) 1313 * new stat data found, but object may have old items
1198 struct stat_data *sd = (struct stat_data *)B_I_PITEM(bh, ih); 1314 * (directories and symlinks)
1315 */
1316 struct stat_data *sd = (struct stat_data *)ih_item_body(bh, ih);
1199 1317
1200 inode->i_mode = sd_v2_mode(sd); 1318 inode->i_mode = sd_v2_mode(sd);
1201 set_nlink(inode, sd_v2_nlink(sd)); 1319 set_nlink(inode, sd_v2_nlink(sd));
@@ -1225,8 +1343,10 @@ static void init_inode(struct inode *inode, struct treepath *path)
1225 inode_set_bytes(inode, 1343 inode_set_bytes(inode,
1226 to_real_used_space(inode, inode->i_blocks, 1344 to_real_used_space(inode, inode->i_blocks,
1227 SD_V2_SIZE)); 1345 SD_V2_SIZE));
1228 /* read persistent inode attributes from sd and initialise 1346 /*
1229 generic inode flags from them */ 1347 * read persistent inode attributes from sd and initialise
1348 * generic inode flags from them
1349 */
1230 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd); 1350 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1231 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode); 1351 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1232 } 1352 }
@@ -1249,7 +1369,7 @@ static void init_inode(struct inode *inode, struct treepath *path)
1249 } 1369 }
1250} 1370}
1251 1371
1252// update new stat data with inode fields 1372/* update new stat data with inode fields */
1253static void inode2sd(void *sd, struct inode *inode, loff_t size) 1373static void inode2sd(void *sd, struct inode *inode, loff_t size)
1254{ 1374{
1255 struct stat_data *sd_v2 = (struct stat_data *)sd; 1375 struct stat_data *sd_v2 = (struct stat_data *)sd;
@@ -1273,7 +1393,7 @@ static void inode2sd(void *sd, struct inode *inode, loff_t size)
1273 set_sd_v2_attrs(sd_v2, flags); 1393 set_sd_v2_attrs(sd_v2, flags);
1274} 1394}
1275 1395
1276// used to copy inode's fields to old stat data 1396/* used to copy inode's fields to old stat data */
1277static void inode2sd_v1(void *sd, struct inode *inode, loff_t size) 1397static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1278{ 1398{
1279 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd; 1399 struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
@@ -1292,14 +1412,15 @@ static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1292 else 1412 else
1293 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE)); 1413 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1294 1414
1295 // Sigh. i_first_direct_byte is back 1415 /* Sigh. i_first_direct_byte is back */
1296 set_sd_v1_first_direct_byte(sd_v1, 1416 set_sd_v1_first_direct_byte(sd_v1,
1297 REISERFS_I(inode)->i_first_direct_byte); 1417 REISERFS_I(inode)->i_first_direct_byte);
1298} 1418}
1299 1419
1300/* NOTE, you must prepare the buffer head before sending it here, 1420/*
1301** and then log it after the call 1421 * NOTE, you must prepare the buffer head before sending it here,
1302*/ 1422 * and then log it after the call
1423 */
1303static void update_stat_data(struct treepath *path, struct inode *inode, 1424static void update_stat_data(struct treepath *path, struct inode *inode,
1304 loff_t size) 1425 loff_t size)
1305{ 1426{
@@ -1307,17 +1428,17 @@ static void update_stat_data(struct treepath *path, struct inode *inode,
1307 struct item_head *ih; 1428 struct item_head *ih;
1308 1429
1309 bh = PATH_PLAST_BUFFER(path); 1430 bh = PATH_PLAST_BUFFER(path);
1310 ih = PATH_PITEM_HEAD(path); 1431 ih = tp_item_head(path);
1311 1432
1312 if (!is_statdata_le_ih(ih)) 1433 if (!is_statdata_le_ih(ih))
1313 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h", 1434 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
1314 INODE_PKEY(inode), ih); 1435 INODE_PKEY(inode), ih);
1315 1436
1437 /* path points to old stat data */
1316 if (stat_data_v1(ih)) { 1438 if (stat_data_v1(ih)) {
1317 // path points to old stat data 1439 inode2sd_v1(ih_item_body(bh, ih), inode, size);
1318 inode2sd_v1(B_I_PITEM(bh, ih), inode, size);
1319 } else { 1440 } else {
1320 inode2sd(B_I_PITEM(bh, ih), inode, size); 1441 inode2sd(ih_item_body(bh, ih), inode, size);
1321 } 1442 }
1322 1443
1323 return; 1444 return;
@@ -1335,7 +1456,8 @@ void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1335 1456
1336 BUG_ON(!th->t_trans_id); 1457 BUG_ON(!th->t_trans_id);
1337 1458
1338 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3); //key type is unimportant 1459 /* key type is unimportant */
1460 make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);
1339 1461
1340 for (;;) { 1462 for (;;) {
1341 int pos; 1463 int pos;
@@ -1363,45 +1485,48 @@ void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1363 return; 1485 return;
1364 } 1486 }
1365 1487
1366 /* sigh, prepare_for_journal might schedule. When it schedules the 1488 /*
1367 ** FS might change. We have to detect that, and loop back to the 1489 * sigh, prepare_for_journal might schedule. When it
1368 ** search if the stat data item has moved 1490 * schedules the FS might change. We have to detect that,
1491 * and loop back to the search if the stat data item has moved
1369 */ 1492 */
1370 bh = get_last_bh(&path); 1493 bh = get_last_bh(&path);
1371 ih = get_ih(&path); 1494 ih = tp_item_head(&path);
1372 copy_item_head(&tmp_ih, ih); 1495 copy_item_head(&tmp_ih, ih);
1373 fs_gen = get_generation(inode->i_sb); 1496 fs_gen = get_generation(inode->i_sb);
1374 reiserfs_prepare_for_journal(inode->i_sb, bh, 1); 1497 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1498
1499 /* Stat_data item has been moved after scheduling. */
1375 if (fs_changed(fs_gen, inode->i_sb) 1500 if (fs_changed(fs_gen, inode->i_sb)
1376 && item_moved(&tmp_ih, &path)) { 1501 && item_moved(&tmp_ih, &path)) {
1377 reiserfs_restore_prepared_buffer(inode->i_sb, bh); 1502 reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1378 continue; /* Stat_data item has been moved after scheduling. */ 1503 continue;
1379 } 1504 }
1380 break; 1505 break;
1381 } 1506 }
1382 update_stat_data(&path, inode, size); 1507 update_stat_data(&path, inode, size);
1383 journal_mark_dirty(th, th->t_super, bh); 1508 journal_mark_dirty(th, bh);
1384 pathrelse(&path); 1509 pathrelse(&path);
1385 return; 1510 return;
1386} 1511}
1387 1512
1388/* reiserfs_read_locked_inode is called to read the inode off disk, and it 1513/*
1389** does a make_bad_inode when things go wrong. But, we need to make sure 1514 * reiserfs_read_locked_inode is called to read the inode off disk, and it
1390** and clear the key in the private portion of the inode, otherwise a 1515 * does a make_bad_inode when things go wrong. But, we need to make sure
1391** corresponding iput might try to delete whatever object the inode last 1516 * and clear the key in the private portion of the inode, otherwise a
1392** represented. 1517 * corresponding iput might try to delete whatever object the inode last
1393*/ 1518 * represented.
1519 */
1394static void reiserfs_make_bad_inode(struct inode *inode) 1520static void reiserfs_make_bad_inode(struct inode *inode)
1395{ 1521{
1396 memset(INODE_PKEY(inode), 0, KEY_SIZE); 1522 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1397 make_bad_inode(inode); 1523 make_bad_inode(inode);
1398} 1524}
1399 1525
1400// 1526/*
1401// initially this function was derived from minix or ext2's analog and 1527 * initially this function was derived from minix or ext2's analog and
1402// evolved as the prototype did 1528 * evolved as the prototype did
1403// 1529 */
1404
1405int reiserfs_init_locked_inode(struct inode *inode, void *p) 1530int reiserfs_init_locked_inode(struct inode *inode, void *p)
1406{ 1531{
1407 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p; 1532 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
@@ -1410,8 +1535,10 @@ int reiserfs_init_locked_inode(struct inode *inode, void *p)
1410 return 0; 1535 return 0;
1411} 1536}
1412 1537
1413/* looks for stat data in the tree, and fills up the fields of in-core 1538/*
1414 inode stat data fields */ 1539 * looks for stat data in the tree, and fills up the fields of in-core
1540 * inode stat data fields
1541 */
1415void reiserfs_read_locked_inode(struct inode *inode, 1542void reiserfs_read_locked_inode(struct inode *inode,
1416 struct reiserfs_iget_args *args) 1543 struct reiserfs_iget_args *args)
1417{ 1544{
@@ -1422,8 +1549,10 @@ void reiserfs_read_locked_inode(struct inode *inode,
1422 1549
1423 dirino = args->dirid; 1550 dirino = args->dirid;
1424 1551
1425 /* set version 1, version 2 could be used too, because stat data 1552 /*
1426 key is the same in both versions */ 1553 * set version 1, version 2 could be used too, because stat data
1554 * key is the same in both versions
1555 */
1427 key.version = KEY_FORMAT_3_5; 1556 key.version = KEY_FORMAT_3_5;
1428 key.on_disk_key.k_dir_id = dirino; 1557 key.on_disk_key.k_dir_id = dirino;
1429 key.on_disk_key.k_objectid = inode->i_ino; 1558 key.on_disk_key.k_objectid = inode->i_ino;
@@ -1439,8 +1568,9 @@ void reiserfs_read_locked_inode(struct inode *inode,
1439 reiserfs_make_bad_inode(inode); 1568 reiserfs_make_bad_inode(inode);
1440 return; 1569 return;
1441 } 1570 }
1571
1572 /* a stale NFS handle can trigger this without it being an error */
1442 if (retval != ITEM_FOUND) { 1573 if (retval != ITEM_FOUND) {
1443 /* a stale NFS handle can trigger this without it being an error */
1444 pathrelse(&path_to_sd); 1574 pathrelse(&path_to_sd);
1445 reiserfs_make_bad_inode(inode); 1575 reiserfs_make_bad_inode(inode);
1446 clear_nlink(inode); 1576 clear_nlink(inode);
@@ -1449,20 +1579,25 @@ void reiserfs_read_locked_inode(struct inode *inode,
1449 1579
1450 init_inode(inode, &path_to_sd); 1580 init_inode(inode, &path_to_sd);
1451 1581
1452 /* It is possible that knfsd is trying to access inode of a file 1582 /*
1453 that is being removed from the disk by some other thread. As we 1583 * It is possible that knfsd is trying to access inode of a file
1454 update sd on unlink all that is required is to check for nlink 1584 * that is being removed from the disk by some other thread. As we
1455 here. This bug was first found by Sizif when debugging 1585 * update sd on unlink all that is required is to check for nlink
1456 SquidNG/Butterfly, forgotten, and found again after Philippe 1586 * here. This bug was first found by Sizif when debugging
1457 Gramoulle <philippe.gramoulle@mmania.com> reproduced it. 1587 * SquidNG/Butterfly, forgotten, and found again after Philippe
1458 1588 * Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1459 More logical fix would require changes in fs/inode.c:iput() to 1589
1460 remove inode from hash-table _after_ fs cleaned disk stuff up and 1590 * More logical fix would require changes in fs/inode.c:iput() to
1461 in iget() to return NULL if I_FREEING inode is found in 1591 * remove inode from hash-table _after_ fs cleaned disk stuff up and
1462 hash-table. */ 1592 * in iget() to return NULL if I_FREEING inode is found in
1463 /* Currently there is one place where it's ok to meet inode with 1593 * hash-table.
1464 nlink==0: processing of open-unlinked and half-truncated files 1594 */
1465 during mount (fs/reiserfs/super.c:finish_unfinished()). */ 1595
1596 /*
1597 * Currently there is one place where it's ok to meet inode with
1598 * nlink==0: processing of open-unlinked and half-truncated files
1599 * during mount (fs/reiserfs/super.c:finish_unfinished()).
1600 */
1466 if ((inode->i_nlink == 0) && 1601 if ((inode->i_nlink == 0) &&
1467 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) { 1602 !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1468 reiserfs_warning(inode->i_sb, "vs-13075", 1603 reiserfs_warning(inode->i_sb, "vs-13075",
@@ -1472,7 +1607,8 @@ void reiserfs_read_locked_inode(struct inode *inode,
1472 reiserfs_make_bad_inode(inode); 1607 reiserfs_make_bad_inode(inode);
1473 } 1608 }
1474 1609
1475 reiserfs_check_path(&path_to_sd); /* init inode should be relsing */ 1610 /* init inode should be relsing */
1611 reiserfs_check_path(&path_to_sd);
1476 1612
1477 /* 1613 /*
1478 * Stat data v1 doesn't support ACLs. 1614 * Stat data v1 doesn't support ACLs.
@@ -1481,7 +1617,7 @@ void reiserfs_read_locked_inode(struct inode *inode,
1481 cache_no_acl(inode); 1617 cache_no_acl(inode);
1482} 1618}
1483 1619
1484/** 1620/*
1485 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked(). 1621 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1486 * 1622 *
1487 * @inode: inode from hash table to check 1623 * @inode: inode from hash table to check
@@ -1556,7 +1692,8 @@ static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1556struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid, 1692struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1557 int fh_len, int fh_type) 1693 int fh_len, int fh_type)
1558{ 1694{
1559 /* fhtype happens to reflect the number of u32s encoded. 1695 /*
1696 * fhtype happens to reflect the number of u32s encoded.
1560 * due to a bug in earlier code, fhtype might indicate there 1697 * due to a bug in earlier code, fhtype might indicate there
1561 * are more u32s then actually fitted. 1698 * are more u32s then actually fitted.
1562 * so if fhtype seems to be more than len, reduce fhtype. 1699 * so if fhtype seems to be more than len, reduce fhtype.
@@ -1625,13 +1762,16 @@ int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
1625 return *lenp; 1762 return *lenp;
1626} 1763}
1627 1764
1628/* looks for stat data, then copies fields to it, marks the buffer 1765/*
1629 containing stat data as dirty */ 1766 * looks for stat data, then copies fields to it, marks the buffer
1630/* reiserfs inodes are never really dirty, since the dirty inode call 1767 * containing stat data as dirty
1631** always logs them. This call allows the VFS inode marking routines 1768 */
1632** to properly mark inodes for datasync and such, but only actually 1769/*
1633** does something when called for a synchronous update. 1770 * reiserfs inodes are never really dirty, since the dirty inode call
1634*/ 1771 * always logs them. This call allows the VFS inode marking routines
1772 * to properly mark inodes for datasync and such, but only actually
1773 * does something when called for a synchronous update.
1774 */
1635int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc) 1775int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1636{ 1776{
1637 struct reiserfs_transaction_handle th; 1777 struct reiserfs_transaction_handle th;
@@ -1639,24 +1779,28 @@ int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1639 1779
1640 if (inode->i_sb->s_flags & MS_RDONLY) 1780 if (inode->i_sb->s_flags & MS_RDONLY)
1641 return -EROFS; 1781 return -EROFS;
1642 /* memory pressure can sometimes initiate write_inode calls with sync == 1, 1782 /*
1643 ** these cases are just when the system needs ram, not when the 1783 * memory pressure can sometimes initiate write_inode calls with
1644 ** inode needs to reach disk for safety, and they can safely be 1784 * sync == 1,
1645 ** ignored because the altered inode has already been logged. 1785 * these cases are just when the system needs ram, not when the
1786 * inode needs to reach disk for safety, and they can safely be
1787 * ignored because the altered inode has already been logged.
1646 */ 1788 */
1647 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) { 1789 if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
1648 reiserfs_write_lock(inode->i_sb); 1790 reiserfs_write_lock(inode->i_sb);
1649 if (!journal_begin(&th, inode->i_sb, jbegin_count)) { 1791 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1650 reiserfs_update_sd(&th, inode); 1792 reiserfs_update_sd(&th, inode);
1651 journal_end_sync(&th, inode->i_sb, jbegin_count); 1793 journal_end_sync(&th);
1652 } 1794 }
1653 reiserfs_write_unlock(inode->i_sb); 1795 reiserfs_write_unlock(inode->i_sb);
1654 } 1796 }
1655 return 0; 1797 return 0;
1656} 1798}
1657 1799
1658/* stat data of new object is inserted already, this inserts the item 1800/*
1659 containing "." and ".." entries */ 1801 * stat data of new object is inserted already, this inserts the item
1802 * containing "." and ".." entries
1803 */
1660static int reiserfs_new_directory(struct reiserfs_transaction_handle *th, 1804static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1661 struct inode *inode, 1805 struct inode *inode,
1662 struct item_head *ih, struct treepath *path, 1806 struct item_head *ih, struct treepath *path,
@@ -1674,9 +1818,11 @@ static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1674 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET, 1818 le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1675 TYPE_DIRENTRY, 3 /*key length */ ); 1819 TYPE_DIRENTRY, 3 /*key length */ );
1676 1820
1677 /* compose item head for new item. Directories consist of items of 1821 /*
1678 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it 1822 * compose item head for new item. Directories consist of items of
1679 is done by reiserfs_new_inode */ 1823 * old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1824 * is done by reiserfs_new_inode
1825 */
1680 if (old_format_only(sb)) { 1826 if (old_format_only(sb)) {
1681 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, 1827 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1682 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2); 1828 TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
@@ -1714,9 +1860,12 @@ static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1714 return reiserfs_insert_item(th, path, &key, ih, inode, body); 1860 return reiserfs_insert_item(th, path, &key, ih, inode, body);
1715} 1861}
1716 1862
1717/* stat data of object has been inserted, this inserts the item 1863/*
1718 containing the body of symlink */ 1864 * stat data of object has been inserted, this inserts the item
1719static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct inode *inode, /* Inode of symlink */ 1865 * containing the body of symlink
1866 */
1867static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th,
1868 struct inode *inode,
1720 struct item_head *ih, 1869 struct item_head *ih,
1721 struct treepath *path, const char *symname, 1870 struct treepath *path, const char *symname,
1722 int item_len) 1871 int item_len)
@@ -1754,15 +1903,26 @@ static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th, struct i
1754 return reiserfs_insert_item(th, path, &key, ih, inode, symname); 1903 return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1755} 1904}
1756 1905
1757/* inserts the stat data into the tree, and then calls 1906/*
1758 reiserfs_new_directory (to insert ".", ".." item if new object is 1907 * inserts the stat data into the tree, and then calls
1759 directory) or reiserfs_new_symlink (to insert symlink body if new 1908 * reiserfs_new_directory (to insert ".", ".." item if new object is
1760 object is symlink) or nothing (if new object is regular file) 1909 * directory) or reiserfs_new_symlink (to insert symlink body if new
1761 1910 * object is symlink) or nothing (if new object is regular file)
1762 NOTE! uid and gid must already be set in the inode. If we return 1911
1763 non-zero due to an error, we have to drop the quota previously allocated 1912 * NOTE! uid and gid must already be set in the inode. If we return
1764 for the fresh inode. This can only be done outside a transaction, so 1913 * non-zero due to an error, we have to drop the quota previously allocated
1765 if we return non-zero, we also end the transaction. */ 1914 * for the fresh inode. This can only be done outside a transaction, so
1915 * if we return non-zero, we also end the transaction.
1916 *
1917 * @th: active transaction handle
1918 * @dir: parent directory for new inode
1919 * @mode: mode of new inode
1920 * @symname: symlink contents if inode is symlink
1921 * @isize: 0 for regular file, EMPTY_DIR_SIZE for dirs, strlen(symname) for
1922 * symlinks
1923 * @inode: inode to be filled
1924 * @security: optional security context to associate with this inode
1925 */
1766int reiserfs_new_inode(struct reiserfs_transaction_handle *th, 1926int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1767 struct inode *dir, umode_t mode, const char *symname, 1927 struct inode *dir, umode_t mode, const char *symname,
1768 /* 0 for regular, EMTRY_DIR_SIZE for dirs, 1928 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
@@ -1807,7 +1967,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1807 else 1967 else
1808 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET, 1968 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1809 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT); 1969 TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1810 memcpy(INODE_PKEY(inode), &(ih.ih_key), KEY_SIZE); 1970 memcpy(INODE_PKEY(inode), &ih.ih_key, KEY_SIZE);
1811 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id); 1971 args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1812 1972
1813 depth = reiserfs_write_unlock_nested(inode->i_sb); 1973 depth = reiserfs_write_unlock_nested(inode->i_sb);
@@ -1820,10 +1980,11 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1820 } 1980 }
1821 1981
1822 if (old_format_only(sb)) 1982 if (old_format_only(sb))
1823 /* not a perfect generation count, as object ids can be reused, but 1983 /*
1824 ** this is as good as reiserfs can do right now. 1984 * not a perfect generation count, as object ids can be reused,
1825 ** note that the private part of inode isn't filled in yet, we have 1985 * but this is as good as reiserfs can do right now.
1826 ** to use the directory. 1986 * note that the private part of inode isn't filled in yet,
1987 * we have to use the directory.
1827 */ 1988 */
1828 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid); 1989 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1829 else 1990 else
@@ -1850,7 +2011,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1850 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 : 2011 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1851 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ; 2012 U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
1852 2013
1853 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list)); 2014 INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
1854 REISERFS_I(inode)->i_flags = 0; 2015 REISERFS_I(inode)->i_flags = 0;
1855 REISERFS_I(inode)->i_prealloc_block = 0; 2016 REISERFS_I(inode)->i_prealloc_block = 0;
1856 REISERFS_I(inode)->i_prealloc_count = 0; 2017 REISERFS_I(inode)->i_prealloc_count = 0;
@@ -1878,9 +2039,9 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1878 goto out_bad_inode; 2039 goto out_bad_inode;
1879 } 2040 }
1880 if (old_format_only(sb)) { 2041 if (old_format_only(sb)) {
2042 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1881 if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) { 2043 if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) {
1882 pathrelse(&path_to_key); 2044 pathrelse(&path_to_key);
1883 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1884 err = -EINVAL; 2045 err = -EINVAL;
1885 goto out_bad_inode; 2046 goto out_bad_inode;
1886 } 2047 }
@@ -1888,9 +2049,11 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1888 } else { 2049 } else {
1889 inode2sd(&sd, inode, inode->i_size); 2050 inode2sd(&sd, inode, inode->i_size);
1890 } 2051 }
1891 // store in in-core inode the key of stat data and version all 2052 /*
1892 // object items will have (directory items will have old offset 2053 * store in in-core inode the key of stat data and version all
1893 // format, other new objects will consist of new items) 2054 * object items will have (directory items will have old offset
2055 * format, other new objects will consist of new items)
2056 */
1894 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode)) 2057 if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
1895 set_inode_item_key_version(inode, KEY_FORMAT_3_5); 2058 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1896 else 2059 else
@@ -1934,7 +2097,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1934 if (retval) { 2097 if (retval) {
1935 err = retval; 2098 err = retval;
1936 reiserfs_check_path(&path_to_key); 2099 reiserfs_check_path(&path_to_key);
1937 journal_end(th, th->t_super, th->t_blocks_allocated); 2100 journal_end(th);
1938 goto out_inserted_sd; 2101 goto out_inserted_sd;
1939 } 2102 }
1940 2103
@@ -1945,7 +2108,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1945 if (retval) { 2108 if (retval) {
1946 err = retval; 2109 err = retval;
1947 reiserfs_check_path(&path_to_key); 2110 reiserfs_check_path(&path_to_key);
1948 journal_end(th, th->t_super, th->t_blocks_allocated); 2111 journal_end(th);
1949 goto out_inserted_sd; 2112 goto out_inserted_sd;
1950 } 2113 }
1951 } else if (inode->i_sb->s_flags & MS_POSIXACL) { 2114 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
@@ -1962,8 +2125,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1962 if (retval) { 2125 if (retval) {
1963 err = retval; 2126 err = retval;
1964 reiserfs_check_path(&path_to_key); 2127 reiserfs_check_path(&path_to_key);
1965 retval = journal_end(th, th->t_super, 2128 retval = journal_end(th);
1966 th->t_blocks_allocated);
1967 if (retval) 2129 if (retval)
1968 err = retval; 2130 err = retval;
1969 goto out_inserted_sd; 2131 goto out_inserted_sd;
@@ -1975,11 +2137,7 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1975 2137
1976 return 0; 2138 return 0;
1977 2139
1978/* it looks like you can easily compress these two goto targets into 2140out_bad_inode:
1979 * one. Keeping it like this doesn't actually hurt anything, and they
1980 * are place holders for what the quota code actually needs.
1981 */
1982 out_bad_inode:
1983 /* Invalidate the object, nothing was inserted yet */ 2141 /* Invalidate the object, nothing was inserted yet */
1984 INODE_PKEY(inode)->k_objectid = 0; 2142 INODE_PKEY(inode)->k_objectid = 0;
1985 2143
@@ -1988,16 +2146,19 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1988 dquot_free_inode(inode); 2146 dquot_free_inode(inode);
1989 reiserfs_write_lock_nested(inode->i_sb, depth); 2147 reiserfs_write_lock_nested(inode->i_sb, depth);
1990 2148
1991 out_end_trans: 2149out_end_trans:
1992 journal_end(th, th->t_super, th->t_blocks_allocated); 2150 journal_end(th);
1993 /* Drop can be outside and it needs more credits so it's better to have it outside */ 2151 /*
2152 * Drop can be outside and it needs more credits so it's better
2153 * to have it outside
2154 */
1994 depth = reiserfs_write_unlock_nested(inode->i_sb); 2155 depth = reiserfs_write_unlock_nested(inode->i_sb);
1995 dquot_drop(inode); 2156 dquot_drop(inode);
1996 reiserfs_write_lock_nested(inode->i_sb, depth); 2157 reiserfs_write_lock_nested(inode->i_sb, depth);
1997 inode->i_flags |= S_NOQUOTA; 2158 inode->i_flags |= S_NOQUOTA;
1998 make_bad_inode(inode); 2159 make_bad_inode(inode);
1999 2160
2000 out_inserted_sd: 2161out_inserted_sd:
2001 clear_nlink(inode); 2162 clear_nlink(inode);
2002 th->t_trans_id = 0; /* so the caller can't use this handle later */ 2163 th->t_trans_id = 0; /* so the caller can't use this handle later */
2003 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */ 2164 unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
@@ -2006,25 +2167,26 @@ int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
2006} 2167}
2007 2168
2008/* 2169/*
2009** finds the tail page in the page cache, 2170 * finds the tail page in the page cache,
2010** reads the last block in. 2171 * reads the last block in.
2011** 2172 *
2012** On success, page_result is set to a locked, pinned page, and bh_result 2173 * On success, page_result is set to a locked, pinned page, and bh_result
2013** is set to an up to date buffer for the last block in the file. returns 0. 2174 * is set to an up to date buffer for the last block in the file. returns 0.
2014** 2175 *
2015** tail conversion is not done, so bh_result might not be valid for writing 2176 * tail conversion is not done, so bh_result might not be valid for writing
2016** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before 2177 * check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2017** trying to write the block. 2178 * trying to write the block.
2018** 2179 *
2019** on failure, nonzero is returned, page_result and bh_result are untouched. 2180 * on failure, nonzero is returned, page_result and bh_result are untouched.
2020*/ 2181 */
2021static int grab_tail_page(struct inode *inode, 2182static int grab_tail_page(struct inode *inode,
2022 struct page **page_result, 2183 struct page **page_result,
2023 struct buffer_head **bh_result) 2184 struct buffer_head **bh_result)
2024{ 2185{
2025 2186
2026 /* we want the page with the last byte in the file, 2187 /*
2027 ** not the page that will hold the next byte for appending 2188 * we want the page with the last byte in the file,
2189 * not the page that will hold the next byte for appending
2028 */ 2190 */
2029 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; 2191 unsigned long index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
2030 unsigned long pos = 0; 2192 unsigned long pos = 0;
@@ -2036,10 +2198,11 @@ static int grab_tail_page(struct inode *inode,
2036 struct page *page; 2198 struct page *page;
2037 int error; 2199 int error;
2038 2200
2039 /* we know that we are only called with inode->i_size > 0. 2201 /*
2040 ** we also know that a file tail can never be as big as a block 2202 * we know that we are only called with inode->i_size > 0.
2041 ** If i_size % blocksize == 0, our file is currently block aligned 2203 * we also know that a file tail can never be as big as a block
2042 ** and it won't need converting or zeroing after a truncate. 2204 * If i_size % blocksize == 0, our file is currently block aligned
2205 * and it won't need converting or zeroing after a truncate.
2043 */ 2206 */
2044 if ((offset & (blocksize - 1)) == 0) { 2207 if ((offset & (blocksize - 1)) == 0) {
2045 return -ENOENT; 2208 return -ENOENT;
@@ -2068,10 +2231,11 @@ static int grab_tail_page(struct inode *inode,
2068 } while (bh != head); 2231 } while (bh != head);
2069 2232
2070 if (!buffer_uptodate(bh)) { 2233 if (!buffer_uptodate(bh)) {
2071 /* note, this should never happen, prepare_write should 2234 /*
2072 ** be taking care of this for us. If the buffer isn't up to date, 2235 * note, this should never happen, prepare_write should be
2073 ** I've screwed up the code to find the buffer, or the code to 2236 * taking care of this for us. If the buffer isn't up to
2074 ** call prepare_write 2237 * date, I've screwed up the code to find the buffer, or the
2238 * code to call prepare_write
2075 */ 2239 */
2076 reiserfs_error(inode->i_sb, "clm-6000", 2240 reiserfs_error(inode->i_sb, "clm-6000",
2077 "error reading block %lu", bh->b_blocknr); 2241 "error reading block %lu", bh->b_blocknr);
@@ -2081,21 +2245,21 @@ static int grab_tail_page(struct inode *inode,
2081 *bh_result = bh; 2245 *bh_result = bh;
2082 *page_result = page; 2246 *page_result = page;
2083 2247
2084 out: 2248out:
2085 return error; 2249 return error;
2086 2250
2087 unlock: 2251unlock:
2088 unlock_page(page); 2252 unlock_page(page);
2089 page_cache_release(page); 2253 page_cache_release(page);
2090 return error; 2254 return error;
2091} 2255}
2092 2256
2093/* 2257/*
2094** vfs version of truncate file. Must NOT be called with 2258 * vfs version of truncate file. Must NOT be called with
2095** a transaction already started. 2259 * a transaction already started.
2096** 2260 *
2097** some code taken from block_truncate_page 2261 * some code taken from block_truncate_page
2098*/ 2262 */
2099int reiserfs_truncate_file(struct inode *inode, int update_timestamps) 2263int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2100{ 2264{
2101 struct reiserfs_transaction_handle th; 2265 struct reiserfs_transaction_handle th;
@@ -2113,9 +2277,11 @@ int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2113 if (inode->i_size > 0) { 2277 if (inode->i_size > 0) {
2114 error = grab_tail_page(inode, &page, &bh); 2278 error = grab_tail_page(inode, &page, &bh);
2115 if (error) { 2279 if (error) {
2116 // -ENOENT means we truncated past the end of the file, 2280 /*
2117 // and get_block_create_0 could not find a block to read in, 2281 * -ENOENT means we truncated past the end of the
2118 // which is ok. 2282 * file, and get_block_create_0 could not find a
2283 * block to read in, which is ok.
2284 */
2119 if (error != -ENOENT) 2285 if (error != -ENOENT)
2120 reiserfs_error(inode->i_sb, "clm-6001", 2286 reiserfs_error(inode->i_sb, "clm-6001",
2121 "grab_tail_page failed %d", 2287 "grab_tail_page failed %d",
@@ -2125,29 +2291,33 @@ int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2125 } 2291 }
2126 } 2292 }
2127 2293
2128 /* so, if page != NULL, we have a buffer head for the offset at 2294 /*
2129 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0, 2295 * so, if page != NULL, we have a buffer head for the offset at
2130 ** then we have an unformatted node. Otherwise, we have a direct item, 2296 * the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2131 ** and no zeroing is required on disk. We zero after the truncate, 2297 * then we have an unformatted node. Otherwise, we have a direct item,
2132 ** because the truncate might pack the item anyway 2298 * and no zeroing is required on disk. We zero after the truncate,
2133 ** (it will unmap bh if it packs). 2299 * because the truncate might pack the item anyway
2300 * (it will unmap bh if it packs).
2301 *
2302 * it is enough to reserve space in transaction for 2 balancings:
2303 * one for "save" link adding and another for the first
2304 * cut_from_item. 1 is for update_sd
2134 */ 2305 */
2135 /* it is enough to reserve space in transaction for 2 balancings:
2136 one for "save" link adding and another for the first
2137 cut_from_item. 1 is for update_sd */
2138 error = journal_begin(&th, inode->i_sb, 2306 error = journal_begin(&th, inode->i_sb,
2139 JOURNAL_PER_BALANCE_CNT * 2 + 1); 2307 JOURNAL_PER_BALANCE_CNT * 2 + 1);
2140 if (error) 2308 if (error)
2141 goto out; 2309 goto out;
2142 reiserfs_update_inode_transaction(inode); 2310 reiserfs_update_inode_transaction(inode);
2143 if (update_timestamps) 2311 if (update_timestamps)
2144 /* we are doing real truncate: if the system crashes before the last 2312 /*
2145 transaction of truncating gets committed - on reboot the file 2313 * we are doing real truncate: if the system crashes
2146 either appears truncated properly or not truncated at all */ 2314 * before the last transaction of truncating gets committed
2315 * - on reboot the file either appears truncated properly
2316 * or not truncated at all
2317 */
2147 add_save_link(&th, inode, 1); 2318 add_save_link(&th, inode, 1);
2148 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps); 2319 err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
2149 error = 2320 error = journal_end(&th);
2150 journal_end(&th, inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1);
2151 if (error) 2321 if (error)
2152 goto out; 2322 goto out;
2153 2323
@@ -2180,7 +2350,7 @@ int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2180 reiserfs_write_unlock(inode->i_sb); 2350 reiserfs_write_unlock(inode->i_sb);
2181 2351
2182 return 0; 2352 return 0;
2183 out: 2353out:
2184 if (page) { 2354 if (page) {
2185 unlock_page(page); 2355 unlock_page(page);
2186 page_cache_release(page); 2356 page_cache_release(page);
@@ -2212,7 +2382,10 @@ static int map_block_for_writepage(struct inode *inode,
2212 int copy_size; 2382 int copy_size;
2213 int trans_running = 0; 2383 int trans_running = 0;
2214 2384
2215 /* catch places below that try to log something without starting a trans */ 2385 /*
2386 * catch places below that try to log something without
2387 * starting a trans
2388 */
2216 th.t_trans_id = 0; 2389 th.t_trans_id = 0;
2217 2390
2218 if (!buffer_uptodate(bh_result)) { 2391 if (!buffer_uptodate(bh_result)) {
@@ -2220,11 +2393,11 @@ static int map_block_for_writepage(struct inode *inode,
2220 } 2393 }
2221 2394
2222 kmap(bh_result->b_page); 2395 kmap(bh_result->b_page);
2223 start_over: 2396start_over:
2224 reiserfs_write_lock(inode->i_sb); 2397 reiserfs_write_lock(inode->i_sb);
2225 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3); 2398 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2226 2399
2227 research: 2400research:
2228 retval = search_for_position_by_key(inode->i_sb, &key, &path); 2401 retval = search_for_position_by_key(inode->i_sb, &key, &path);
2229 if (retval != POSITION_FOUND) { 2402 if (retval != POSITION_FOUND) {
2230 use_get_block = 1; 2403 use_get_block = 1;
@@ -2232,8 +2405,8 @@ static int map_block_for_writepage(struct inode *inode,
2232 } 2405 }
2233 2406
2234 bh = get_last_bh(&path); 2407 bh = get_last_bh(&path);
2235 ih = get_ih(&path); 2408 ih = tp_item_head(&path);
2236 item = get_item(&path); 2409 item = tp_item_body(&path);
2237 pos_in_item = path.pos_in_item; 2410 pos_in_item = path.pos_in_item;
2238 2411
2239 /* we've found an unformatted node */ 2412 /* we've found an unformatted node */
@@ -2281,10 +2454,10 @@ static int map_block_for_writepage(struct inode *inode,
2281 goto research; 2454 goto research;
2282 } 2455 }
2283 2456
2284 memcpy(B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied, 2457 memcpy(ih_item_body(bh, ih) + pos_in_item, p + bytes_copied,
2285 copy_size); 2458 copy_size);
2286 2459
2287 journal_mark_dirty(&th, inode->i_sb, bh); 2460 journal_mark_dirty(&th, bh);
2288 bytes_copied += copy_size; 2461 bytes_copied += copy_size;
2289 set_block_dev_mapped(bh_result, 0, inode); 2462 set_block_dev_mapped(bh_result, 0, inode);
2290 2463
@@ -2304,10 +2477,10 @@ static int map_block_for_writepage(struct inode *inode,
2304 } 2477 }
2305 retval = 0; 2478 retval = 0;
2306 2479
2307 out: 2480out:
2308 pathrelse(&path); 2481 pathrelse(&path);
2309 if (trans_running) { 2482 if (trans_running) {
2310 int err = journal_end(&th, inode->i_sb, jbegin_count); 2483 int err = journal_end(&th);
2311 if (err) 2484 if (err)
2312 retval = err; 2485 retval = err;
2313 trans_running = 0; 2486 trans_running = 0;
@@ -2331,7 +2504,8 @@ static int map_block_for_writepage(struct inode *inode,
2331 kunmap(bh_result->b_page); 2504 kunmap(bh_result->b_page);
2332 2505
2333 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) { 2506 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2334 /* we've copied data from the page into the direct item, so the 2507 /*
2508 * we've copied data from the page into the direct item, so the
2335 * buffer in the page is now clean, mark it to reflect that. 2509 * buffer in the page is now clean, mark it to reflect that.
2336 */ 2510 */
2337 lock_buffer(bh_result); 2511 lock_buffer(bh_result);
@@ -2370,7 +2544,8 @@ static int reiserfs_write_full_page(struct page *page,
2370 return 0; 2544 return 0;
2371 } 2545 }
2372 2546
2373 /* The page dirty bit is cleared before writepage is called, which 2547 /*
2548 * The page dirty bit is cleared before writepage is called, which
2374 * means we have to tell create_empty_buffers to make dirty buffers 2549 * means we have to tell create_empty_buffers to make dirty buffers
2375 * The page really should be up to date at this point, so tossing 2550 * The page really should be up to date at this point, so tossing
2376 * in the BH_Uptodate is just a sanity check. 2551 * in the BH_Uptodate is just a sanity check.
@@ -2381,8 +2556,9 @@ static int reiserfs_write_full_page(struct page *page,
2381 } 2556 }
2382 head = page_buffers(page); 2557 head = page_buffers(page);
2383 2558
2384 /* last page in the file, zero out any contents past the 2559 /*
2385 ** last byte in the file 2560 * last page in the file, zero out any contents past the
2561 * last byte in the file
2386 */ 2562 */
2387 if (page->index >= end_index) { 2563 if (page->index >= end_index) {
2388 unsigned last_offset; 2564 unsigned last_offset;
@@ -2412,7 +2588,8 @@ static int reiserfs_write_full_page(struct page *page,
2412 (!buffer_mapped(bh) || (buffer_mapped(bh) 2588 (!buffer_mapped(bh) || (buffer_mapped(bh)
2413 && bh->b_blocknr == 2589 && bh->b_blocknr ==
2414 0))) { 2590 0))) {
2415 /* not mapped yet, or it points to a direct item, search 2591 /*
2592 * not mapped yet, or it points to a direct item, search
2416 * the btree for the mapping info, and log any direct 2593 * the btree for the mapping info, and log any direct
2417 * items found 2594 * items found
2418 */ 2595 */
@@ -2450,10 +2627,11 @@ static int reiserfs_write_full_page(struct page *page,
2450 2627
2451 if (checked) { 2628 if (checked) {
2452 reiserfs_prepare_for_journal(s, bh, 1); 2629 reiserfs_prepare_for_journal(s, bh, 1);
2453 journal_mark_dirty(&th, s, bh); 2630 journal_mark_dirty(&th, bh);
2454 continue; 2631 continue;
2455 } 2632 }
2456 /* from this point on, we know the buffer is mapped to a 2633 /*
2634 * from this point on, we know the buffer is mapped to a
2457 * real block and not a direct item 2635 * real block and not a direct item
2458 */ 2636 */
2459 if (wbc->sync_mode != WB_SYNC_NONE) { 2637 if (wbc->sync_mode != WB_SYNC_NONE) {
@@ -2472,7 +2650,7 @@ static int reiserfs_write_full_page(struct page *page,
2472 } while ((bh = bh->b_this_page) != head); 2650 } while ((bh = bh->b_this_page) != head);
2473 2651
2474 if (checked) { 2652 if (checked) {
2475 error = journal_end(&th, s, bh_per_page + 1); 2653 error = journal_end(&th);
2476 reiserfs_write_unlock(s); 2654 reiserfs_write_unlock(s);
2477 if (error) 2655 if (error)
2478 goto fail; 2656 goto fail;
@@ -2497,7 +2675,7 @@ static int reiserfs_write_full_page(struct page *page,
2497 } while (bh != head); 2675 } while (bh != head);
2498 2676
2499 error = 0; 2677 error = 0;
2500 done: 2678done:
2501 if (nr == 0) { 2679 if (nr == 0) {
2502 /* 2680 /*
2503 * if this page only had a direct item, it is very possible for 2681 * if this page only had a direct item, it is very possible for
@@ -2519,8 +2697,9 @@ static int reiserfs_write_full_page(struct page *page,
2519 } 2697 }
2520 return error; 2698 return error;
2521 2699
2522 fail: 2700fail:
2523 /* catches various errors, we need to make sure any valid dirty blocks 2701 /*
2702 * catches various errors, we need to make sure any valid dirty blocks
2524 * get to the media. The page is currently locked and not marked for 2703 * get to the media. The page is currently locked and not marked for
2525 * writeback 2704 * writeback
2526 */ 2705 */
@@ -2533,8 +2712,8 @@ static int reiserfs_write_full_page(struct page *page,
2533 mark_buffer_async_write(bh); 2712 mark_buffer_async_write(bh);
2534 } else { 2713 } else {
2535 /* 2714 /*
2536 * clear any dirty bits that might have come from getting 2715 * clear any dirty bits that might have come from
2537 * attached to a dirty page 2716 * getting attached to a dirty page
2538 */ 2717 */
2539 clear_buffer_dirty(bh); 2718 clear_buffer_dirty(bh);
2540 } 2719 }
@@ -2614,15 +2793,18 @@ static int reiserfs_write_begin(struct file *file,
2614 ret = __block_write_begin(page, pos, len, reiserfs_get_block); 2793 ret = __block_write_begin(page, pos, len, reiserfs_get_block);
2615 if (ret && reiserfs_transaction_running(inode->i_sb)) { 2794 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2616 struct reiserfs_transaction_handle *th = current->journal_info; 2795 struct reiserfs_transaction_handle *th = current->journal_info;
2617 /* this gets a little ugly. If reiserfs_get_block returned an 2796 /*
2618 * error and left a transacstion running, we've got to close it, 2797 * this gets a little ugly. If reiserfs_get_block returned an
2619 * and we've got to free handle if it was a persistent transaction. 2798 * error and left a transacstion running, we've got to close
2799 * it, and we've got to free handle if it was a persistent
2800 * transaction.
2620 * 2801 *
2621 * But, if we had nested into an existing transaction, we need 2802 * But, if we had nested into an existing transaction, we need
2622 * to just drop the ref count on the handle. 2803 * to just drop the ref count on the handle.
2623 * 2804 *
2624 * If old_ref == 0, the transaction is from reiserfs_get_block, 2805 * If old_ref == 0, the transaction is from reiserfs_get_block,
2625 * and it was a persistent trans. Otherwise, it was nested above. 2806 * and it was a persistent trans. Otherwise, it was nested
2807 * above.
2626 */ 2808 */
2627 if (th->t_refcount > old_ref) { 2809 if (th->t_refcount > old_ref) {
2628 if (old_ref) 2810 if (old_ref)
@@ -2671,15 +2853,18 @@ int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
2671 ret = __block_write_begin(page, from, len, reiserfs_get_block); 2853 ret = __block_write_begin(page, from, len, reiserfs_get_block);
2672 if (ret && reiserfs_transaction_running(inode->i_sb)) { 2854 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2673 struct reiserfs_transaction_handle *th = current->journal_info; 2855 struct reiserfs_transaction_handle *th = current->journal_info;
2674 /* this gets a little ugly. If reiserfs_get_block returned an 2856 /*
2675 * error and left a transacstion running, we've got to close it, 2857 * this gets a little ugly. If reiserfs_get_block returned an
2676 * and we've got to free handle if it was a persistent transaction. 2858 * error and left a transacstion running, we've got to close
2859 * it, and we've got to free handle if it was a persistent
2860 * transaction.
2677 * 2861 *
2678 * But, if we had nested into an existing transaction, we need 2862 * But, if we had nested into an existing transaction, we need
2679 * to just drop the ref count on the handle. 2863 * to just drop the ref count on the handle.
2680 * 2864 *
2681 * If old_ref == 0, the transaction is from reiserfs_get_block, 2865 * If old_ref == 0, the transaction is from reiserfs_get_block,
2682 * and it was a persistent trans. Otherwise, it was nested above. 2866 * and it was a persistent trans. Otherwise, it was nested
2867 * above.
2683 */ 2868 */
2684 if (th->t_refcount > old_ref) { 2869 if (th->t_refcount > old_ref) {
2685 if (old_ref) 2870 if (old_ref)
@@ -2734,17 +2919,20 @@ static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2734 2919
2735 reiserfs_commit_page(inode, page, start, start + copied); 2920 reiserfs_commit_page(inode, page, start, start + copied);
2736 2921
2737 /* generic_commit_write does this for us, but does not update the 2922 /*
2738 ** transaction tracking stuff when the size changes. So, we have 2923 * generic_commit_write does this for us, but does not update the
2739 ** to do the i_size updates here. 2924 * transaction tracking stuff when the size changes. So, we have
2925 * to do the i_size updates here.
2740 */ 2926 */
2741 if (pos + copied > inode->i_size) { 2927 if (pos + copied > inode->i_size) {
2742 struct reiserfs_transaction_handle myth; 2928 struct reiserfs_transaction_handle myth;
2743 reiserfs_write_lock(inode->i_sb); 2929 reiserfs_write_lock(inode->i_sb);
2744 locked = true; 2930 locked = true;
2745 /* If the file have grown beyond the border where it 2931 /*
2746 can have a tail, unmark it as needing a tail 2932 * If the file have grown beyond the border where it
2747 packing */ 2933 * can have a tail, unmark it as needing a tail
2934 * packing
2935 */
2748 if ((have_large_tails(inode->i_sb) 2936 if ((have_large_tails(inode->i_sb)
2749 && inode->i_size > i_block_size(inode) * 4) 2937 && inode->i_size > i_block_size(inode) * 4)
2750 || (have_small_tails(inode->i_sb) 2938 || (have_small_tails(inode->i_sb)
@@ -2759,13 +2947,13 @@ static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2759 inode->i_size = pos + copied; 2947 inode->i_size = pos + copied;
2760 /* 2948 /*
2761 * this will just nest into our transaction. It's important 2949 * this will just nest into our transaction. It's important
2762 * to use mark_inode_dirty so the inode gets pushed around on the 2950 * to use mark_inode_dirty so the inode gets pushed around on
2763 * dirty lists, and so that O_SYNC works as expected 2951 * the dirty lists, and so that O_SYNC works as expected
2764 */ 2952 */
2765 mark_inode_dirty(inode); 2953 mark_inode_dirty(inode);
2766 reiserfs_update_sd(&myth, inode); 2954 reiserfs_update_sd(&myth, inode);
2767 update_sd = 1; 2955 update_sd = 1;
2768 ret = journal_end(&myth, inode->i_sb, 1); 2956 ret = journal_end(&myth);
2769 if (ret) 2957 if (ret)
2770 goto journal_error; 2958 goto journal_error;
2771 } 2959 }
@@ -2781,7 +2969,7 @@ static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2781 goto out; 2969 goto out;
2782 } 2970 }
2783 2971
2784 out: 2972out:
2785 if (locked) 2973 if (locked)
2786 reiserfs_write_unlock(inode->i_sb); 2974 reiserfs_write_unlock(inode->i_sb);
2787 unlock_page(page); 2975 unlock_page(page);
@@ -2792,7 +2980,7 @@ static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2792 2980
2793 return ret == 0 ? copied : ret; 2981 return ret == 0 ? copied : ret;
2794 2982
2795 journal_error: 2983journal_error:
2796 reiserfs_write_unlock(inode->i_sb); 2984 reiserfs_write_unlock(inode->i_sb);
2797 locked = false; 2985 locked = false;
2798 if (th) { 2986 if (th) {
@@ -2822,15 +3010,18 @@ int reiserfs_commit_write(struct file *f, struct page *page,
2822 } 3010 }
2823 reiserfs_commit_page(inode, page, from, to); 3011 reiserfs_commit_page(inode, page, from, to);
2824 3012
2825 /* generic_commit_write does this for us, but does not update the 3013 /*
2826 ** transaction tracking stuff when the size changes. So, we have 3014 * generic_commit_write does this for us, but does not update the
2827 ** to do the i_size updates here. 3015 * transaction tracking stuff when the size changes. So, we have
3016 * to do the i_size updates here.
2828 */ 3017 */
2829 if (pos > inode->i_size) { 3018 if (pos > inode->i_size) {
2830 struct reiserfs_transaction_handle myth; 3019 struct reiserfs_transaction_handle myth;
2831 /* If the file have grown beyond the border where it 3020 /*
2832 can have a tail, unmark it as needing a tail 3021 * If the file have grown beyond the border where it
2833 packing */ 3022 * can have a tail, unmark it as needing a tail
3023 * packing
3024 */
2834 if ((have_large_tails(inode->i_sb) 3025 if ((have_large_tails(inode->i_sb)
2835 && inode->i_size > i_block_size(inode) * 4) 3026 && inode->i_size > i_block_size(inode) * 4)
2836 || (have_small_tails(inode->i_sb) 3027 || (have_small_tails(inode->i_sb)
@@ -2845,13 +3036,13 @@ int reiserfs_commit_write(struct file *f, struct page *page,
2845 inode->i_size = pos; 3036 inode->i_size = pos;
2846 /* 3037 /*
2847 * this will just nest into our transaction. It's important 3038 * this will just nest into our transaction. It's important
2848 * to use mark_inode_dirty so the inode gets pushed around on the 3039 * to use mark_inode_dirty so the inode gets pushed around
2849 * dirty lists, and so that O_SYNC works as expected 3040 * on the dirty lists, and so that O_SYNC works as expected
2850 */ 3041 */
2851 mark_inode_dirty(inode); 3042 mark_inode_dirty(inode);
2852 reiserfs_update_sd(&myth, inode); 3043 reiserfs_update_sd(&myth, inode);
2853 update_sd = 1; 3044 update_sd = 1;
2854 ret = journal_end(&myth, inode->i_sb, 1); 3045 ret = journal_end(&myth);
2855 if (ret) 3046 if (ret)
2856 goto journal_error; 3047 goto journal_error;
2857 } 3048 }
@@ -2863,10 +3054,10 @@ int reiserfs_commit_write(struct file *f, struct page *page,
2863 goto out; 3054 goto out;
2864 } 3055 }
2865 3056
2866 out: 3057out:
2867 return ret; 3058 return ret;
2868 3059
2869 journal_error: 3060journal_error:
2870 if (th) { 3061 if (th) {
2871 if (!update_sd) 3062 if (!update_sd)
2872 reiserfs_update_sd(th, inode); 3063 reiserfs_update_sd(th, inode);
@@ -2924,9 +3115,10 @@ void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
2924 } 3115 }
2925} 3116}
2926 3117
2927/* decide if this buffer needs to stay around for data logging or ordered 3118/*
2928** write purposes 3119 * decide if this buffer needs to stay around for data logging or ordered
2929*/ 3120 * write purposes
3121 */
2930static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh) 3122static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2931{ 3123{
2932 int ret = 1; 3124 int ret = 1;
@@ -2937,7 +3129,8 @@ static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2937 if (!buffer_mapped(bh)) { 3129 if (!buffer_mapped(bh)) {
2938 goto free_jh; 3130 goto free_jh;
2939 } 3131 }
2940 /* the page is locked, and the only places that log a data buffer 3132 /*
3133 * the page is locked, and the only places that log a data buffer
2941 * also lock the page. 3134 * also lock the page.
2942 */ 3135 */
2943 if (reiserfs_file_data_log(inode)) { 3136 if (reiserfs_file_data_log(inode)) {
@@ -2952,7 +3145,8 @@ static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2952 struct reiserfs_journal_list *jl; 3145 struct reiserfs_journal_list *jl;
2953 struct reiserfs_jh *jh = bh->b_private; 3146 struct reiserfs_jh *jh = bh->b_private;
2954 3147
2955 /* why is this safe? 3148 /*
3149 * why is this safe?
2956 * reiserfs_setattr updates i_size in the on disk 3150 * reiserfs_setattr updates i_size in the on disk
2957 * stat data before allowing vmtruncate to be called. 3151 * stat data before allowing vmtruncate to be called.
2958 * 3152 *
@@ -2969,7 +3163,7 @@ static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2969 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl) 3163 && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2970 ret = 0; 3164 ret = 0;
2971 } 3165 }
2972 free_jh: 3166free_jh:
2973 if (ret && bh->b_private) { 3167 if (ret && bh->b_private) {
2974 reiserfs_free_jh(bh); 3168 reiserfs_free_jh(bh);
2975 } 3169 }
@@ -3028,7 +3222,7 @@ static void reiserfs_invalidatepage(struct page *page, unsigned int offset,
3028 ret = try_to_release_page(page, 0); 3222 ret = try_to_release_page(page, 0);
3029 /* maybe should BUG_ON(!ret); - neilb */ 3223 /* maybe should BUG_ON(!ret); - neilb */
3030 } 3224 }
3031 out: 3225out:
3032 return; 3226 return;
3033} 3227}
3034 3228
@@ -3080,8 +3274,10 @@ static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
3080 return ret; 3274 return ret;
3081} 3275}
3082 3276
3083/* We thank Mingming Cao for helping us understand in great detail what 3277/*
3084 to do in this section of the code. */ 3278 * We thank Mingming Cao for helping us understand in great detail what
3279 * to do in this section of the code.
3280 */
3085static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb, 3281static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
3086 const struct iovec *iov, loff_t offset, 3282 const struct iovec *iov, loff_t offset,
3087 unsigned long nr_segs) 3283 unsigned long nr_segs)
@@ -3127,8 +3323,9 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3127 dquot_initialize(inode); 3323 dquot_initialize(inode);
3128 reiserfs_write_lock(inode->i_sb); 3324 reiserfs_write_lock(inode->i_sb);
3129 if (attr->ia_valid & ATTR_SIZE) { 3325 if (attr->ia_valid & ATTR_SIZE) {
3130 /* version 2 items will be caught by the s_maxbytes check 3326 /*
3131 ** done for us in vmtruncate 3327 * version 2 items will be caught by the s_maxbytes check
3328 * done for us in vmtruncate
3132 */ 3329 */
3133 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 && 3330 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3134 attr->ia_size > MAX_NON_LFS) { 3331 attr->ia_size > MAX_NON_LFS) {
@@ -3149,7 +3346,7 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3149 err = journal_begin(&th, inode->i_sb, 4); 3346 err = journal_begin(&th, inode->i_sb, 4);
3150 if (!err) { 3347 if (!err) {
3151 reiserfs_discard_prealloc(&th, inode); 3348 reiserfs_discard_prealloc(&th, inode);
3152 err = journal_end(&th, inode->i_sb, 4); 3349 err = journal_end(&th);
3153 } 3350 }
3154 if (err) 3351 if (err)
3155 error = err; 3352 error = err;
@@ -3189,7 +3386,10 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3189 if (error) 3386 if (error)
3190 return error; 3387 return error;
3191 3388
3192 /* (user+group)*(old+new) structure - we count quota info and , inode write (sb, inode) */ 3389 /*
3390 * (user+group)*(old+new) structure - we count quota
3391 * info and , inode write (sb, inode)
3392 */
3193 reiserfs_write_lock(inode->i_sb); 3393 reiserfs_write_lock(inode->i_sb);
3194 error = journal_begin(&th, inode->i_sb, jbegin_count); 3394 error = journal_begin(&th, inode->i_sb, jbegin_count);
3195 reiserfs_write_unlock(inode->i_sb); 3395 reiserfs_write_unlock(inode->i_sb);
@@ -3198,19 +3398,21 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3198 error = dquot_transfer(inode, attr); 3398 error = dquot_transfer(inode, attr);
3199 reiserfs_write_lock(inode->i_sb); 3399 reiserfs_write_lock(inode->i_sb);
3200 if (error) { 3400 if (error) {
3201 journal_end(&th, inode->i_sb, jbegin_count); 3401 journal_end(&th);
3202 reiserfs_write_unlock(inode->i_sb); 3402 reiserfs_write_unlock(inode->i_sb);
3203 goto out; 3403 goto out;
3204 } 3404 }
3205 3405
3206 /* Update corresponding info in inode so that everything is in 3406 /*
3207 * one transaction */ 3407 * Update corresponding info in inode so that everything
3408 * is in one transaction
3409 */
3208 if (attr->ia_valid & ATTR_UID) 3410 if (attr->ia_valid & ATTR_UID)
3209 inode->i_uid = attr->ia_uid; 3411 inode->i_uid = attr->ia_uid;
3210 if (attr->ia_valid & ATTR_GID) 3412 if (attr->ia_valid & ATTR_GID)
3211 inode->i_gid = attr->ia_gid; 3413 inode->i_gid = attr->ia_gid;
3212 mark_inode_dirty(inode); 3414 mark_inode_dirty(inode);
3213 error = journal_end(&th, inode->i_sb, jbegin_count); 3415 error = journal_end(&th);
3214 reiserfs_write_unlock(inode->i_sb); 3416 reiserfs_write_unlock(inode->i_sb);
3215 if (error) 3417 if (error)
3216 goto out; 3418 goto out;
@@ -3220,8 +3422,14 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3220 attr->ia_size != i_size_read(inode)) { 3422 attr->ia_size != i_size_read(inode)) {
3221 error = inode_newsize_ok(inode, attr->ia_size); 3423 error = inode_newsize_ok(inode, attr->ia_size);
3222 if (!error) { 3424 if (!error) {
3425 /*
3426 * Could race against reiserfs_file_release
3427 * if called from NFS, so take tailpack mutex.
3428 */
3429 mutex_lock(&REISERFS_I(inode)->tailpack);
3223 truncate_setsize(inode, attr->ia_size); 3430 truncate_setsize(inode, attr->ia_size);
3224 reiserfs_vfs_truncate_file(inode); 3431 reiserfs_truncate_file(inode, 1);
3432 mutex_unlock(&REISERFS_I(inode)->tailpack);
3225 } 3433 }
3226 } 3434 }
3227 3435