~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/reiserfs/inode.c

Version: ~ [ linux-5.13-rc1 ] ~ [ linux-5.12.2 ] ~ [ linux-5.11.19 ] ~ [ linux-5.10.35 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.117 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.190 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.232 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.268 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.268 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
  3  */
  4 
  5 #include <linux/time.h>
  6 #include <linux/fs.h>
  7 #include "reiserfs.h"
  8 #include "acl.h"
  9 #include "xattr.h"
 10 #include <linux/exportfs.h>
 11 #include <linux/pagemap.h>
 12 #include <linux/highmem.h>
 13 #include <linux/slab.h>
 14 #include <linux/uaccess.h>
 15 #include <asm/unaligned.h>
 16 #include <linux/buffer_head.h>
 17 #include <linux/mpage.h>
 18 #include <linux/writeback.h>
 19 #include <linux/quotaops.h>
 20 #include <linux/swap.h>
 21 #include <linux/uio.h>
 22 
 23 int reiserfs_commit_write(struct file *f, struct page *page,
 24                           unsigned from, unsigned to);
 25 
 26 void reiserfs_evict_inode(struct inode *inode)
 27 {
 28         /*
 29          * We need blocks for transaction + (user+group) quota
 30          * update (possibly delete)
 31          */
 32         int jbegin_count =
 33             JOURNAL_PER_BALANCE_CNT * 2 +
 34             2 * REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb);
 35         struct reiserfs_transaction_handle th;
 36         int err;
 37 
 38         if (!inode->i_nlink && !is_bad_inode(inode))
 39                 dquot_initialize(inode);
 40 
 41         truncate_inode_pages_final(&inode->i_data);
 42         if (inode->i_nlink)
 43                 goto no_delete;
 44 
 45         /*
 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) {
 51 
 52                 reiserfs_delete_xattrs(inode);
 53 
 54                 reiserfs_write_lock(inode->i_sb);
 55 
 56                 if (journal_begin(&th, inode->i_sb, jbegin_count))
 57                         goto out;
 58                 reiserfs_update_inode_transaction(inode);
 59 
 60                 reiserfs_discard_prealloc(&th, inode);
 61 
 62                 err = reiserfs_delete_object(&th, inode);
 63 
 64                 /*
 65                  * Do quota update inside a transaction for journaled quotas.
 66                  * We must do that after delete_object so that quota updates
 67                  * go into the same transaction as stat data deletion
 68                  */
 69                 if (!err) {
 70                         int depth = reiserfs_write_unlock_nested(inode->i_sb);
 71                         dquot_free_inode(inode);
 72                         reiserfs_write_lock_nested(inode->i_sb, depth);
 73                 }
 74 
 75                 if (journal_end(&th))
 76                         goto out;
 77 
 78                 /*
 79                  * check return value from reiserfs_delete_object after
 80                  * ending the transaction
 81                  */
 82                 if (err)
 83                     goto out;
 84 
 85                 /*
 86                  * all items of file are deleted, so we can remove
 87                  * "save" link
 88                  * we can't do anything about an error here
 89                  */
 90                 remove_save_link(inode, 0 /* not truncate */);
 91 out:
 92                 reiserfs_write_unlock(inode->i_sb);
 93         } else {
 94                 /* no object items are in the tree */
 95                 ;
 96         }
 97 
 98         /* note this must go after the journal_end to prevent deadlock */
 99         clear_inode(inode);
100 
101         dquot_drop(inode);
102         inode->i_blocks = 0;
103         return;
104 
105 no_delete:
106         clear_inode(inode);
107         dquot_drop(inode);
108 }
109 
110 static void _make_cpu_key(struct cpu_key *key, int version, __u32 dirid,
111                           __u32 objectid, loff_t offset, int type, int length)
112 {
113         key->version = version;
114 
115         key->on_disk_key.k_dir_id = dirid;
116         key->on_disk_key.k_objectid = objectid;
117         set_cpu_key_k_offset(key, offset);
118         set_cpu_key_k_type(key, type);
119         key->key_length = length;
120 }
121 
122 /*
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  */
126 void make_cpu_key(struct cpu_key *key, struct inode *inode, loff_t offset,
127                   int type, int length)
128 {
129         _make_cpu_key(key, get_inode_item_key_version(inode),
130                       le32_to_cpu(INODE_PKEY(inode)->k_dir_id),
131                       le32_to_cpu(INODE_PKEY(inode)->k_objectid), offset, type,
132                       length);
133 }
134 
135 /* when key is 0, do not set version and short key */
136 inline void make_le_item_head(struct item_head *ih, const struct cpu_key *key,
137                               int version,
138                               loff_t offset, int type, int length,
139                               int entry_count /*or ih_free_space */ )
140 {
141         if (key) {
142                 ih->ih_key.k_dir_id = cpu_to_le32(key->on_disk_key.k_dir_id);
143                 ih->ih_key.k_objectid =
144                     cpu_to_le32(key->on_disk_key.k_objectid);
145         }
146         put_ih_version(ih, version);
147         set_le_ih_k_offset(ih, offset);
148         set_le_ih_k_type(ih, type);
149         put_ih_item_len(ih, length);
150         /*    set_ih_free_space (ih, 0); */
151         /*
152          * for directory items it is entry count, for directs and stat
153          * datas - 0xffff, for indirects - 0
154          */
155         put_ih_entry_count(ih, entry_count);
156 }
157 
158 /*
159  * FIXME: we might cache recently accessed indirect item
160  * Ugh.  Not too eager for that....
161  * 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....
163  */
164 
165 /*
166  * cutting the code is fine, since it really isn't in use yet and is easy
167  * to add back in.  But, Vladimir has a really good idea here.  Think
168  * about what happens for reading a file.  For each page,
169  * The VFS layer calls reiserfs_readpage, who searches the tree to find
170  * an indirect item.  This indirect item has X number of pointers, where
171  * X is a big number if we've done the block allocation right.  But,
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  */
192 static inline void fix_tail_page_for_writing(struct page *page)
193 {
194         struct buffer_head *head, *next, *bh;
195 
196         if (page && page_has_buffers(page)) {
197                 head = page_buffers(page);
198                 bh = head;
199                 do {
200                         next = bh->b_this_page;
201                         if (buffer_mapped(bh) && bh->b_blocknr == 0) {
202                                 reiserfs_unmap_buffer(bh);
203                         }
204                         bh = next;
205                 } while (bh != head);
206         }
207 }
208 
209 /*
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  */
213 static inline int allocation_needed(int retval, b_blocknr_t allocated,
214                                     struct item_head *ih,
215                                     __le32 * item, int pos_in_item)
216 {
217         if (allocated)
218                 return 0;
219         if (retval == POSITION_FOUND && is_indirect_le_ih(ih) &&
220             get_block_num(item, pos_in_item))
221                 return 0;
222         return 1;
223 }
224 
225 static inline int indirect_item_found(int retval, struct item_head *ih)
226 {
227         return (retval == POSITION_FOUND) && is_indirect_le_ih(ih);
228 }
229 
230 static inline void set_block_dev_mapped(struct buffer_head *bh,
231                                         b_blocknr_t block, struct inode *inode)
232 {
233         map_bh(bh, inode->i_sb, block);
234 }
235 
236 /*
237  * files which were created in the earlier version can not be longer,
238  * than 2 gb
239  */
240 static int file_capable(struct inode *inode, sector_t block)
241 {
242         /* it is new file. */
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)))
246                 return 1;
247 
248         return 0;
249 }
250 
251 static int restart_transaction(struct reiserfs_transaction_handle *th,
252                                struct inode *inode, struct treepath *path)
253 {
254         struct super_block *s = th->t_super;
255         int err;
256 
257         BUG_ON(!th->t_trans_id);
258         BUG_ON(!th->t_refcount);
259 
260         pathrelse(path);
261 
262         /* we cannot restart while nested */
263         if (th->t_refcount > 1) {
264                 return 0;
265         }
266         reiserfs_update_sd(th, inode);
267         err = journal_end(th);
268         if (!err) {
269                 err = journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6);
270                 if (!err)
271                         reiserfs_update_inode_transaction(inode);
272         }
273         return err;
274 }
275 
276 /*
277  * it is called by get_block when create == 0. Returns block number
278  * for 'block'-th logical block of file. When it hits direct item it
279  * returns 0 (being called from bmap) or read direct item into piece
280  * of page (bh_result)
281  * Please improve the english/clarity in the comment above, as it is
282  * hard to understand.
283  */
284 static int _get_block_create_0(struct inode *inode, sector_t block,
285                                struct buffer_head *bh_result, int args)
286 {
287         INITIALIZE_PATH(path);
288         struct cpu_key key;
289         struct buffer_head *bh;
290         struct item_head *ih, tmp_ih;
291         b_blocknr_t blocknr;
292         char *p = NULL;
293         int chars;
294         int ret;
295         int result;
296         int done = 0;
297         unsigned long offset;
298 
299         /* prepare the key to look for the 'block'-th block of file */
300         make_cpu_key(&key, inode,
301                      (loff_t) block * inode->i_sb->s_blocksize + 1, TYPE_ANY,
302                      3);
303 
304         result = search_for_position_by_key(inode->i_sb, &key, &path);
305         if (result != POSITION_FOUND) {
306                 pathrelse(&path);
307                 if (p)
308                         kunmap(bh_result->b_page);
309                 if (result == IO_ERROR)
310                         return -EIO;
311                 /*
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                  */
316                 if ((args & GET_BLOCK_NO_HOLE)
317                     && !PageUptodate(bh_result->b_page)) {
318                         return -ENOENT;
319                 }
320                 return 0;
321         }
322 
323         bh = get_last_bh(&path);
324         ih = tp_item_head(&path);
325         if (is_indirect_le_ih(ih)) {
326                 __le32 *ind_item = (__le32 *) ih_item_body(bh, ih);
327 
328                 /*
329                  * FIXME: here we could cache indirect item or part of it in
330                  * the inode to avoid search_by_key in case of subsequent
331                  * access to file
332                  */
333                 blocknr = get_block_num(ind_item, path.pos_in_item);
334                 ret = 0;
335                 if (blocknr) {
336                         map_bh(bh_result, inode->i_sb, blocknr);
337                         if (path.pos_in_item ==
338                             ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
339                                 set_buffer_boundary(bh_result);
340                         }
341                 } else
342                         /*
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                          */
348                 if ((args & GET_BLOCK_NO_HOLE)
349                             && !PageUptodate(bh_result->b_page)) {
350                         ret = -ENOENT;
351                 }
352 
353                 pathrelse(&path);
354                 if (p)
355                         kunmap(bh_result->b_page);
356                 return ret;
357         }
358         /* requested data are in direct item(s) */
359         if (!(args & GET_BLOCK_READ_DIRECT)) {
360                 /*
361                  * we are called by bmap. FIXME: we can not map block of file
362                  * when it is stored in direct item(s)
363                  */
364                 pathrelse(&path);
365                 if (p)
366                         kunmap(bh_result->b_page);
367                 return -ENOENT;
368         }
369 
370         /*
371          * if we've got a direct item, and the buffer or page was uptodate,
372          * we don't want to pull data off disk again.  skip to the
373          * end, where we map the buffer and return
374          */
375         if (buffer_uptodate(bh_result)) {
376                 goto finished;
377         } else
378                 /*
379                  * grab_tail_page can trigger calls to reiserfs_get_block on
380                  * up to date pages without any buffers.  If the page is up
381                  * to date, we don't want read old data off disk.  Set the up
382                  * to date bit on the buffer instead and jump to the end
383                  */
384         if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
385                 set_buffer_uptodate(bh_result);
386                 goto finished;
387         }
388         /* read file tail into part of page */
389         offset = (cpu_key_k_offset(&key) - 1) & (PAGE_SIZE - 1);
390         copy_item_head(&tmp_ih, ih);
391 
392         /*
393          * we only want to kmap if we are reading the tail into the page.
394          * this is not the common case, so we don't kmap until we are
395          * sure we need to.  But, this means the item might move if
396          * kmap schedules
397          */
398         if (!p)
399                 p = (char *)kmap(bh_result->b_page);
400 
401         p += offset;
402         memset(p, 0, inode->i_sb->s_blocksize);
403         do {
404                 if (!is_direct_le_ih(ih)) {
405                         BUG();
406                 }
407                 /*
408                  * make sure we don't read more bytes than actually exist in
409                  * the file.  This can happen in odd cases where i_size isn't
410                  * correct, and when direct item padding results in a few
411                  * extra bytes at the end of the direct item
412                  */
413                 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
414                         break;
415                 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
416                         chars =
417                             inode->i_size - (le_ih_k_offset(ih) - 1) -
418                             path.pos_in_item;
419                         done = 1;
420                 } else {
421                         chars = ih_item_len(ih) - path.pos_in_item;
422                 }
423                 memcpy(p, ih_item_body(bh, ih) + path.pos_in_item, chars);
424 
425                 if (done)
426                         break;
427 
428                 p += chars;
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                  */
436                 if (PATH_LAST_POSITION(&path) != (B_NR_ITEMS(bh) - 1))
437                         break;
438 
439                 /* update key to look for the next piece */
440                 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + chars);
441                 result = search_for_position_by_key(inode->i_sb, &key, &path);
442                 if (result != POSITION_FOUND)
443                         /* i/o error most likely */
444                         break;
445                 bh = get_last_bh(&path);
446                 ih = tp_item_head(&path);
447         } while (1);
448 
449         flush_dcache_page(bh_result->b_page);
450         kunmap(bh_result->b_page);
451 
452 finished:
453         pathrelse(&path);
454 
455         if (result == IO_ERROR)
456                 return -EIO;
457 
458         /*
459          * this buffer has valid data, but isn't valid for io.  mapping it to
460          * block #0 tells the rest of reiserfs it just has a tail in it
461          */
462         map_bh(bh_result, inode->i_sb, 0);
463         set_buffer_uptodate(bh_result);
464         return 0;
465 }
466 
467 /*
468  * this is called to create file map. So, _get_block_create_0 will not
469  * read direct item
470  */
471 static int reiserfs_bmap(struct inode *inode, sector_t block,
472                          struct buffer_head *bh_result, int create)
473 {
474         if (!file_capable(inode, block))
475                 return -EFBIG;
476 
477         reiserfs_write_lock(inode->i_sb);
478         /* do not read the direct item */
479         _get_block_create_0(inode, block, bh_result, 0);
480         reiserfs_write_unlock(inode->i_sb);
481         return 0;
482 }
483 
484 /*
485  * special version of get_block that is only used by grab_tail_page right
486  * now.  It is sent to __block_write_begin, and when you try to get a
487  * block past the end of the file (or a block from a hole) it returns
488  * -ENOENT instead of a valid buffer.  __block_write_begin expects to
489  * be able to do i/o on the buffers returned, unless an error value
490  * is also returned.
491  *
492  * So, this allows __block_write_begin to be used for reading a single block
493  * in a page.  Where it does not produce a valid page for holes, or past the
494  * end of the file.  This turns out to be exactly what we need for reading
495  * tails for conversion.
496  *
497  * The point of the wrapper is forcing a certain value for create, even
498  * though the VFS layer is calling this function with create==1.  If you
499  * don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
500  * don't use this function.
501 */
502 static int reiserfs_get_block_create_0(struct inode *inode, sector_t block,
503                                        struct buffer_head *bh_result,
504                                        int create)
505 {
506         return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE);
507 }
508 
509 /*
510  * This is special helper for reiserfs_get_block in case we are executing
511  * direct_IO request.
512  */
513 static int reiserfs_get_blocks_direct_io(struct inode *inode,
514                                          sector_t iblock,
515                                          struct buffer_head *bh_result,
516                                          int create)
517 {
518         int ret;
519 
520         bh_result->b_page = NULL;
521 
522         /*
523          * We set the b_size before reiserfs_get_block call since it is
524          * referenced in convert_tail_for_hole() that may be called from
525          * reiserfs_get_block()
526          */
527         bh_result->b_size = (1 << inode->i_blkbits);
528 
529         ret = reiserfs_get_block(inode, iblock, bh_result,
530                                  create | GET_BLOCK_NO_DANGLE);
531         if (ret)
532                 goto out;
533 
534         /* don't allow direct io onto tail pages */
535         if (buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
536                 /*
537                  * make sure future calls to the direct io funcs for this
538                  * offset in the file fail by unmapping the buffer
539                  */
540                 clear_buffer_mapped(bh_result);
541                 ret = -EINVAL;
542         }
543 
544         /*
545          * Possible unpacked tail. Flush the data before pages have
546          * disappeared
547          */
548         if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
549                 int err;
550 
551                 reiserfs_write_lock(inode->i_sb);
552 
553                 err = reiserfs_commit_for_inode(inode);
554                 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
555 
556                 reiserfs_write_unlock(inode->i_sb);
557 
558                 if (err < 0)
559                         ret = err;
560         }
561 out:
562         return ret;
563 }
564 
565 /*
566  * helper function for when reiserfs_get_block is called for a hole
567  * but the file tail is still in a direct item
568  * bh_result is the buffer head for the hole
569  * tail_offset is the offset of the start of the tail in the file
570  *
571  * This calls prepare_write, which will start a new transaction
572  * you should not be in a transaction, or have any paths held when you
573  * call this.
574  */
575 static int convert_tail_for_hole(struct inode *inode,
576                                  struct buffer_head *bh_result,
577                                  loff_t tail_offset)
578 {
579         unsigned long index;
580         unsigned long tail_end;
581         unsigned long tail_start;
582         struct page *tail_page;
583         struct page *hole_page = bh_result->b_page;
584         int retval = 0;
585 
586         if ((tail_offset & (bh_result->b_size - 1)) != 1)
587                 return -EIO;
588 
589         /* always try to read until the end of the block */
590         tail_start = tail_offset & (PAGE_SIZE - 1);
591         tail_end = (tail_start | (bh_result->b_size - 1)) + 1;
592 
593         index = tail_offset >> PAGE_SHIFT;
594         /*
595          * hole_page can be zero in case of direct_io, we are sure
596          * that we cannot get here if we write with O_DIRECT into tail page
597          */
598         if (!hole_page || index != hole_page->index) {
599                 tail_page = grab_cache_page(inode->i_mapping, index);
600                 retval = -ENOMEM;
601                 if (!tail_page) {
602                         goto out;
603                 }
604         } else {
605                 tail_page = hole_page;
606         }
607 
608         /*
609          * we don't have to make sure the conversion did not happen while
610          * we were locking the page because anyone that could convert
611          * must first take i_mutex.
612          *
613          * We must fix the tail page for writing because it might have buffers
614          * that are mapped, but have a block number of 0.  This indicates tail
615          * data that has been read directly into the page, and
616          * __block_write_begin won't trigger a get_block in this case.
617          */
618         fix_tail_page_for_writing(tail_page);
619         retval = __reiserfs_write_begin(tail_page, tail_start,
620                                       tail_end - tail_start);
621         if (retval)
622                 goto unlock;
623 
624         /* tail conversion might change the data in the page */
625         flush_dcache_page(tail_page);
626 
627         retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end);
628 
629 unlock:
630         if (tail_page != hole_page) {
631                 unlock_page(tail_page);
632                 put_page(tail_page);
633         }
634 out:
635         return retval;
636 }
637 
638 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
639                                   sector_t block,
640                                   struct inode *inode,
641                                   b_blocknr_t * allocated_block_nr,
642                                   struct treepath *path, int flags)
643 {
644         BUG_ON(!th->t_trans_id);
645 
646 #ifdef REISERFS_PREALLOCATE
647         if (!(flags & GET_BLOCK_NO_IMUX)) {
648                 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr,
649                                                   path, block);
650         }
651 #endif
652         return reiserfs_new_unf_blocknrs(th, inode, allocated_block_nr, path,
653                                          block);
654 }
655 
656 int reiserfs_get_block(struct inode *inode, sector_t block,
657                        struct buffer_head *bh_result, int create)
658 {
659         int repeat, retval = 0;
660         /* b_blocknr_t is (unsigned) 32 bit int*/
661         b_blocknr_t allocated_block_nr = 0;
662         INITIALIZE_PATH(path);
663         int pos_in_item;
664         struct cpu_key key;
665         struct buffer_head *bh, *unbh = NULL;
666         struct item_head *ih, tmp_ih;
667         __le32 *item;
668         int done;
669         int fs_gen;
670         struct reiserfs_transaction_handle *th = NULL;
671         /*
672          * space reserved in transaction batch:
673          * . 3 balancings in direct->indirect conversion
674          * . 1 block involved into reiserfs_update_sd()
675          * XXX in practically impossible worst case direct2indirect()
676          * can incur (much) more than 3 balancings.
677          * quota update for user, group
678          */
679         int jbegin_count =
680             JOURNAL_PER_BALANCE_CNT * 3 + 1 +
681             2 * REISERFS_QUOTA_TRANS_BLOCKS(inode->i_sb);
682         int version;
683         int dangle = 1;
684         loff_t new_offset =
685             (((loff_t) block) << inode->i_sb->s_blocksize_bits) + 1;
686 
687         reiserfs_write_lock(inode->i_sb);
688         version = get_inode_item_key_version(inode);
689 
690         if (!file_capable(inode, block)) {
691                 reiserfs_write_unlock(inode->i_sb);
692                 return -EFBIG;
693         }
694 
695         /*
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
698          */
699         if (!(create & GET_BLOCK_CREATE)) {
700                 int ret;
701                 /* find number of block-th logical block of the file */
702                 ret = _get_block_create_0(inode, block, bh_result,
703                                           create | GET_BLOCK_READ_DIRECT);
704                 reiserfs_write_unlock(inode->i_sb);
705                 return ret;
706         }
707 
708         /*
709          * if we're already in a transaction, make sure to close
710          * any new transactions we start in this func
711          */
712         if ((create & GET_BLOCK_NO_DANGLE) ||
713             reiserfs_transaction_running(inode->i_sb))
714                 dangle = 0;
715 
716         /*
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
720          */
721         if ((have_large_tails(inode->i_sb)
722              && inode->i_size < i_block_size(inode) * 4)
723             || (have_small_tails(inode->i_sb)
724                 && inode->i_size < i_block_size(inode)))
725                 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask;
726 
727         /* set the key of the first byte in the 'block'-th block of file */
728         make_cpu_key(&key, inode, new_offset, TYPE_ANY, 3 /*key length */ );
729         if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
730 start_trans:
731                 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
732                 if (!th) {
733                         retval = -ENOMEM;
734                         goto failure;
735                 }
736                 reiserfs_update_inode_transaction(inode);
737         }
738 research:
739 
740         retval = search_for_position_by_key(inode->i_sb, &key, &path);
741         if (retval == IO_ERROR) {
742                 retval = -EIO;
743                 goto failure;
744         }
745 
746         bh = get_last_bh(&path);
747         ih = tp_item_head(&path);
748         item = tp_item_body(&path);
749         pos_in_item = path.pos_in_item;
750 
751         fs_gen = get_generation(inode->i_sb);
752         copy_item_head(&tmp_ih, ih);
753 
754         if (allocation_needed
755             (retval, allocated_block_nr, ih, item, pos_in_item)) {
756                 /* we have to allocate block for the unformatted node */
757                 if (!th) {
758                         pathrelse(&path);
759                         goto start_trans;
760                 }
761 
762                 repeat =
763                     _allocate_block(th, block, inode, &allocated_block_nr,
764                                     &path, create);
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                  */
771                 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
772                         SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
773                         retval = restart_transaction(th, inode, &path);
774                         if (retval)
775                                 goto failure;
776                         repeat =
777                             _allocate_block(th, block, inode,
778                                             &allocated_block_nr, NULL, create);
779 
780                         if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
781                                 goto research;
782                         }
783                         if (repeat == QUOTA_EXCEEDED)
784                                 retval = -EDQUOT;
785                         else
786                                 retval = -ENOSPC;
787                         goto failure;
788                 }
789 
790                 if (fs_changed(fs_gen, inode->i_sb)
791                     && item_moved(&tmp_ih, &path)) {
792                         goto research;
793                 }
794         }
795 
796         if (indirect_item_found(retval, ih)) {
797                 b_blocknr_t unfm_ptr;
798                 /*
799                  * 'block'-th block is in the file already (there is
800                  * corresponding cell in some indirect item). But it may be
801                  * zero unformatted node pointer (hole)
802                  */
803                 unfm_ptr = get_block_num(item, pos_in_item);
804                 if (unfm_ptr == 0) {
805                         /* use allocated block to plug the hole */
806                         reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
807                         if (fs_changed(fs_gen, inode->i_sb)
808                             && item_moved(&tmp_ih, &path)) {
809                                 reiserfs_restore_prepared_buffer(inode->i_sb,
810                                                                  bh);
811                                 goto research;
812                         }
813                         set_buffer_new(bh_result);
814                         if (buffer_dirty(bh_result)
815                             && reiserfs_data_ordered(inode->i_sb))
816                                 reiserfs_add_ordered_list(inode, bh_result);
817                         put_block_num(item, pos_in_item, allocated_block_nr);
818                         unfm_ptr = allocated_block_nr;
819                         journal_mark_dirty(th, bh);
820                         reiserfs_update_sd(th, inode);
821                 }
822                 set_block_dev_mapped(bh_result, unfm_ptr, inode);
823                 pathrelse(&path);
824                 retval = 0;
825                 if (!dangle && th)
826                         retval = reiserfs_end_persistent_transaction(th);
827 
828                 reiserfs_write_unlock(inode->i_sb);
829 
830                 /*
831                  * the item was found, so new blocks were not added to the file
832                  * there is no need to make sure the inode is updated with this
833                  * transaction
834                  */
835                 return retval;
836         }
837 
838         if (!th) {
839                 pathrelse(&path);
840                 goto start_trans;
841         }
842 
843         /*
844          * desired position is not found or is in the direct item. We have
845          * to append file with holes up to 'block'-th block converting
846          * direct items to indirect one if necessary
847          */
848         done = 0;
849         do {
850                 if (is_statdata_le_ih(ih)) {
851                         __le32 unp = 0;
852                         struct cpu_key tmp_key;
853 
854                         /* indirect item has to be inserted */
855                         make_le_item_head(&tmp_ih, &key, version, 1,
856                                           TYPE_INDIRECT, UNFM_P_SIZE,
857                                           0 /* free_space */ );
858 
859                         /*
860                          * we are going to add 'block'-th block to the file.
861                          * Use allocated block for that
862                          */
863                         if (cpu_key_k_offset(&key) == 1) {
864                                 unp = cpu_to_le32(allocated_block_nr);
865                                 set_block_dev_mapped(bh_result,
866                                                      allocated_block_nr, inode);
867                                 set_buffer_new(bh_result);
868                                 done = 1;
869                         }
870                         tmp_key = key;  /* ;) */
871                         set_cpu_key_k_offset(&tmp_key, 1);
872                         PATH_LAST_POSITION(&path)++;
873 
874                         retval =
875                             reiserfs_insert_item(th, &path, &tmp_key, &tmp_ih,
876                                                  inode, (char *)&unp);
877                         if (retval) {
878                                 reiserfs_free_block(th, inode,
879                                                     allocated_block_nr, 1);
880                                 /*
881                                  * retval == -ENOSPC, -EDQUOT or -EIO
882                                  * or -EEXIST
883                                  */
884                                 goto failure;
885                         }
886                 } else if (is_direct_le_ih(ih)) {
887                         /* direct item has to be converted */
888                         loff_t tail_offset;
889 
890                         tail_offset =
891                             ((le_ih_k_offset(ih) -
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                          */
899                         if (tail_offset == cpu_key_k_offset(&key)) {
900                                 set_block_dev_mapped(bh_result,
901                                                      allocated_block_nr, inode);
902                                 unbh = bh_result;
903                                 done = 1;
904                         } else {
905                                 /*
906                                  * we have to pad file tail stored in direct
907                                  * item(s) up to block size and convert it
908                                  * to unformatted node. FIXME: this should
909                                  * also get into page cache
910                                  */
911 
912                                 pathrelse(&path);
913                                 /*
914                                  * ugly, but we can only end the transaction if
915                                  * we aren't nested
916                                  */
917                                 BUG_ON(!th->t_refcount);
918                                 if (th->t_refcount == 1) {
919                                         retval =
920                                             reiserfs_end_persistent_transaction
921                                             (th);
922                                         th = NULL;
923                                         if (retval)
924                                                 goto failure;
925                                 }
926 
927                                 retval =
928                                     convert_tail_for_hole(inode, bh_result,
929                                                           tail_offset);
930                                 if (retval) {
931                                         if (retval != -ENOSPC)
932                                                 reiserfs_error(inode->i_sb,
933                                                         "clm-6004",
934                                                         "convert tail failed "
935                                                         "inode %lu, error %d",
936                                                         inode->i_ino,
937                                                         retval);
938                                         if (allocated_block_nr) {
939                                                 /*
940                                                  * the bitmap, the super,
941                                                  * and the stat data == 3
942                                                  */
943                                                 if (!th)
944                                                         th = reiserfs_persistent_transaction(inode->i_sb, 3);
945                                                 if (th)
946                                                         reiserfs_free_block(th,
947                                                                             inode,
948                                                                             allocated_block_nr,
949                                                                             1);
950                                         }
951                                         goto failure;
952                                 }
953                                 goto research;
954                         }
955                         retval =
956                             direct2indirect(th, inode, &path, unbh,
957                                             tail_offset);
958                         if (retval) {
959                                 reiserfs_unmap_buffer(unbh);
960                                 reiserfs_free_block(th, inode,
961                                                     allocated_block_nr, 1);
962                                 goto failure;
963                         }
964                         /*
965                          * it is important the set_buffer_uptodate is done
966                          * after the direct2indirect.  The buffer might
967                          * contain valid data newer than the data on disk
968                          * (read by readpage, changed, and then sent here by
969                          * writepage).  direct2indirect needs to know if unbh
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
973                          */
974                         set_buffer_uptodate(unbh);
975 
976                         /*
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
980                          */
981                         if (unbh->b_page) {
982                                 /*
983                                  * we've converted the tail, so we must
984                                  * flush unbh before the transaction commits
985                                  */
986                                 reiserfs_add_tail_list(inode, unbh);
987 
988                                 /*
989                                  * mark it dirty now to prevent commit_write
990                                  * from adding this buffer to the inode's
991                                  * dirty buffer list
992                                  */
993                                 /*
994                                  * AKPM: changed __mark_buffer_dirty to
995                                  * mark_buffer_dirty().  It's still atomic,
996                                  * but it sets the page dirty too, which makes
997                                  * it eligible for writeback at any time by the
998                                  * VM (which was also the case with
999                                  * __mark_buffer_dirty())
1000                                  */
1001                                 mark_buffer_dirty(unbh);
1002                         }
1003                 } else {
1004                         /*
1005                          * append indirect item with holes if needed, when
1006                          * appending pointer to 'block'-th block use block,
1007                          * which is already allocated
1008                          */
1009                         struct cpu_key tmp_key;
1010                         /*
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;
1015                         unp_t *un;
1016                         __u64 max_to_insert =
1017                             MAX_ITEM_LEN(inode->i_sb->s_blocksize) /
1018                             UNFM_P_SIZE;
1019                         __u64 blocks_needed;
1020 
1021                         RFALSE(pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
1022                                "vs-804: invalid position for append");
1023                         /*
1024                          * indirect item has to be appended,
1025                          * set up key of that position
1026                          * (key type is unimportant)
1027                          */
1028                         make_cpu_key(&tmp_key, inode,
1029                                      le_key_k_offset(version,
1030                                                      &ih->ih_key) +
1031                                      op_bytes_number(ih,
1032                                                      inode->i_sb->s_blocksize),
1033                                      TYPE_INDIRECT, 3);
1034 
1035                         RFALSE(cpu_key_k_offset(&tmp_key) > cpu_key_k_offset(&key),
1036                                "green-805: invalid offset");
1037                         blocks_needed =
1038                             1 +
1039                             ((cpu_key_k_offset(&key) -
1040                               cpu_key_k_offset(&tmp_key)) >> inode->i_sb->
1041                              s_blocksize_bits);
1042 
1043                         if (blocks_needed == 1) {
1044                                 un = &unf_single;
1045                         } else {
1046                                 un = kzalloc(min(blocks_needed, max_to_insert) * UNFM_P_SIZE, GFP_NOFS);
1047                                 if (!un) {
1048                                         un = &unf_single;
1049                                         blocks_needed = 1;
1050                                         max_to_insert = 0;
1051                                 }
1052                         }
1053                         if (blocks_needed <= max_to_insert) {
1054                                 /*
1055                                  * we are going to add target block to
1056                                  * the file. Use allocated block for that
1057                                  */
1058                                 un[blocks_needed - 1] =
1059                                     cpu_to_le32(allocated_block_nr);
1060                                 set_block_dev_mapped(bh_result,
1061                                                      allocated_block_nr, inode);
1062                                 set_buffer_new(bh_result);
1063                                 done = 1;
1064                         } else {
1065                                 /* paste hole to the indirect item */
1066                                 /*
1067                                  * If kmalloc failed, max_to_insert becomes
1068                                  * zero and it means we only have space for
1069                                  * one block
1070                                  */
1071                                 blocks_needed =
1072                                     max_to_insert ? max_to_insert : 1;
1073                         }
1074                         retval =
1075                             reiserfs_paste_into_item(th, &path, &tmp_key, inode,
1076                                                      (char *)un,
1077                                                      UNFM_P_SIZE *
1078                                                      blocks_needed);
1079 
1080                         if (blocks_needed != 1)
1081                                 kfree(un);
1082 
1083                         if (retval) {
1084                                 reiserfs_free_block(th, inode,
1085                                                     allocated_block_nr, 1);
1086                                 goto failure;
1087                         }
1088                         if (!done) {
1089                                 /*
1090                                  * We need to mark new file size in case
1091                                  * this function will be interrupted/aborted
1092                                  * later on. And we may do this only for
1093                                  * holes.
1094                                  */
1095                                 inode->i_size +=
1096                                     inode->i_sb->s_blocksize * blocks_needed;
1097                         }
1098                 }
1099 
1100                 if (done == 1)
1101                         break;
1102 
1103                 /*
1104                  * this loop could log more blocks than we had originally
1105                  * asked for.  So, we have to allow the transaction to end
1106                  * if it is too big or too full.  Update the inode so things
1107                  * are consistent if we crash before the function returns
1108                  * release the path so that anybody waiting on the path before
1109                  * ending their transaction will be able to continue.
1110                  */
1111                 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
1112                         retval = restart_transaction(th, inode, &path);
1113                         if (retval)
1114                                 goto failure;
1115                 }
1116                 /*
1117                  * inserting indirect pointers for a hole can take a
1118                  * long time.  reschedule if needed and also release the write
1119                  * lock for others.
1120                  */
1121                 reiserfs_cond_resched(inode->i_sb);
1122 
1123                 retval = search_for_position_by_key(inode->i_sb, &key, &path);
1124                 if (retval == IO_ERROR) {
1125                         retval = -EIO;
1126                         goto failure;
1127                 }
1128                 if (retval == POSITION_FOUND) {
1129                         reiserfs_warning(inode->i_sb, "vs-825",
1130                                          "%K should not be found", &key);
1131                         retval = -EEXIST;
1132                         if (allocated_block_nr)
1133                                 reiserfs_free_block(th, inode,
1134                                                     allocated_block_nr, 1);
1135                         pathrelse(&path);
1136                         goto failure;
1137                 }
1138                 bh = get_last_bh(&path);
1139                 ih = tp_item_head(&path);
1140                 item = tp_item_body(&path);
1141                 pos_in_item = path.pos_in_item;
1142         } while (1);
1143 
1144         retval = 0;
1145 
1146 failure:
1147         if (th && (!dangle || (retval && !th->t_trans_id))) {
1148                 int err;
1149                 if (th->t_trans_id)
1150                         reiserfs_update_sd(th, inode);
1151                 err = reiserfs_end_persistent_transaction(th);
1152                 if (err)
1153                         retval = err;
1154         }
1155 
1156         reiserfs_write_unlock(inode->i_sb);
1157         reiserfs_check_path(&path);
1158         return retval;
1159 }
1160 
1161 static int
1162 reiserfs_readpages(struct file *file, struct address_space *mapping,
1163                    struct list_head *pages, unsigned nr_pages)
1164 {
1165         return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
1166 }
1167 
1168 /*
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
1172  */
1173 static int real_space_diff(struct inode *inode, int sd_size)
1174 {
1175         int bytes;
1176         loff_t blocksize = inode->i_sb->s_blocksize;
1177 
1178         if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
1179                 return sd_size;
1180 
1181         /*
1182          * End of file is also in full block with indirect reference, so round
1183          * up to the next block.
1184          *
1185          * there is just no way to know if the tail is actually packed
1186          * on the file, so we have to assume it isn't.  When we pack the
1187          * tail, we add 4 bytes to pretend there really is an unformatted
1188          * node pointer
1189          */
1190         bytes =
1191             ((inode->i_size +
1192               (blocksize - 1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE +
1193             sd_size;
1194         return bytes;
1195 }
1196 
1197 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
1198                                         int sd_size)
1199 {
1200         if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1201                 return inode->i_size +
1202                     (loff_t) (real_space_diff(inode, sd_size));
1203         }
1204         return ((loff_t) real_space_diff(inode, sd_size)) +
1205             (((loff_t) blocks) << 9);
1206 }
1207 
1208 /* Compute number of blocks used by file in ReiserFS counting */
1209 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
1210 {
1211         loff_t bytes = inode_get_bytes(inode);
1212         loff_t real_space = real_space_diff(inode, sd_size);
1213 
1214         /* keeps fsck and non-quota versions of reiserfs happy */
1215         if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
1216                 bytes += (loff_t) 511;
1217         }
1218 
1219         /*
1220          * files from before the quota patch might i_blocks such that
1221          * bytes < real_space.  Deal with that here to prevent it from
1222          * going negative.
1223          */
1224         if (bytes < real_space)
1225                 return 0;
1226         return (bytes - real_space) >> 9;
1227 }
1228 
1229 /*
1230  * BAD: new directories have stat data of new type and all other items
1231  * of old type. Version stored in the inode says about body items, so
1232  * in update_stat_data we can not rely on inode, but have to check
1233  * item version directly
1234  */
1235 
1236 /* called by read_locked_inode */
1237 static void init_inode(struct inode *inode, struct treepath *path)
1238 {
1239         struct buffer_head *bh;
1240         struct item_head *ih;
1241         __u32 rdev;
1242 
1243         bh = PATH_PLAST_BUFFER(path);
1244         ih = tp_item_head(path);
1245 
1246         copy_key(INODE_PKEY(inode), &ih->ih_key);
1247 
1248         INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
1249         REISERFS_I(inode)->i_flags = 0;
1250         REISERFS_I(inode)->i_prealloc_block = 0;
1251         REISERFS_I(inode)->i_prealloc_count = 0;
1252         REISERFS_I(inode)->i_trans_id = 0;
1253         REISERFS_I(inode)->i_jl = NULL;
1254         reiserfs_init_xattr_rwsem(inode);
1255 
1256         if (stat_data_v1(ih)) {
1257                 struct stat_data_v1 *sd =
1258                     (struct stat_data_v1 *)ih_item_body(bh, ih);
1259                 unsigned long blocks;
1260 
1261                 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1262                 set_inode_sd_version(inode, STAT_DATA_V1);
1263                 inode->i_mode = sd_v1_mode(sd);
1264                 set_nlink(inode, sd_v1_nlink(sd));
1265                 i_uid_write(inode, sd_v1_uid(sd));
1266                 i_gid_write(inode, sd_v1_gid(sd));
1267                 inode->i_size = sd_v1_size(sd);
1268                 inode->i_atime.tv_sec = sd_v1_atime(sd);
1269                 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1270                 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1271                 inode->i_atime.tv_nsec = 0;
1272                 inode->i_ctime.tv_nsec = 0;
1273                 inode->i_mtime.tv_nsec = 0;
1274 
1275                 inode->i_blocks = sd_v1_blocks(sd);
1276                 inode->i_generation = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1277                 blocks = (inode->i_size + 511) >> 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                  */
1288                 if (inode->i_blocks > blocks) {
1289                         inode->i_blocks = blocks;
1290                 }
1291 
1292                 rdev = sd_v1_rdev(sd);
1293                 REISERFS_I(inode)->i_first_direct_byte =
1294                     sd_v1_first_direct_byte(sd);
1295 
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.
1299                  */
1300                 if (inode->i_blocks & 1) {
1301                         inode->i_blocks++;
1302                 }
1303                 inode_set_bytes(inode,
1304                                 to_real_used_space(inode, inode->i_blocks,
1305                                                    SD_V1_SIZE));
1306                 /*
1307                  * nopack is initially zero for v1 objects. For v2 objects,
1308                  * nopack is initialised from sd_attrs
1309                  */
1310                 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1311         } else {
1312                 /*
1313                  * new stat data found, but object may have old items
1314                  * (directories and symlinks)
1315                  */
1316                 struct stat_data *sd = (struct stat_data *)ih_item_body(bh, ih);
1317 
1318                 inode->i_mode = sd_v2_mode(sd);
1319                 set_nlink(inode, sd_v2_nlink(sd));
1320                 i_uid_write(inode, sd_v2_uid(sd));
1321                 inode->i_size = sd_v2_size(sd);
1322                 i_gid_write(inode, sd_v2_gid(sd));
1323                 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1324                 inode->i_atime.tv_sec = sd_v2_atime(sd);
1325                 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1326                 inode->i_ctime.tv_nsec = 0;
1327                 inode->i_mtime.tv_nsec = 0;
1328                 inode->i_atime.tv_nsec = 0;
1329                 inode->i_blocks = sd_v2_blocks(sd);
1330                 rdev = sd_v2_rdev(sd);
1331                 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1332                         inode->i_generation =
1333                             le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1334                 else
1335                         inode->i_generation = sd_v2_generation(sd);
1336 
1337                 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1338                         set_inode_item_key_version(inode, KEY_FORMAT_3_5);
1339                 else
1340                         set_inode_item_key_version(inode, KEY_FORMAT_3_6);
1341                 REISERFS_I(inode)->i_first_direct_byte = 0;
1342                 set_inode_sd_version(inode, STAT_DATA_V2);
1343                 inode_set_bytes(inode,
1344                                 to_real_used_space(inode, inode->i_blocks,
1345                                                    SD_V2_SIZE));
1346                 /*
1347                  * read persistent inode attributes from sd and initialise
1348                  * generic inode flags from them
1349                  */
1350                 REISERFS_I(inode)->i_attrs = sd_v2_attrs(sd);
1351                 sd_attrs_to_i_attrs(sd_v2_attrs(sd), inode);
1352         }
1353 
1354         pathrelse(path);
1355         if (S_ISREG(inode->i_mode)) {
1356                 inode->i_op = &reiserfs_file_inode_operations;
1357                 inode->i_fop = &reiserfs_file_operations;
1358                 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1359         } else if (S_ISDIR(inode->i_mode)) {
1360                 inode->i_op = &reiserfs_dir_inode_operations;
1361                 inode->i_fop = &reiserfs_dir_operations;
1362         } else if (S_ISLNK(inode->i_mode)) {
1363                 inode->i_op = &reiserfs_symlink_inode_operations;
1364                 inode_nohighmem(inode);
1365                 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1366         } else {
1367                 inode->i_blocks = 0;
1368                 inode->i_op = &reiserfs_special_inode_operations;
1369                 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1370         }
1371 }
1372 
1373 /* update new stat data with inode fields */
1374 static void inode2sd(void *sd, struct inode *inode, loff_t size)
1375 {
1376         struct stat_data *sd_v2 = (struct stat_data *)sd;
1377         __u16 flags;
1378 
1379         set_sd_v2_mode(sd_v2, inode->i_mode);
1380         set_sd_v2_nlink(sd_v2, inode->i_nlink);
1381         set_sd_v2_uid(sd_v2, i_uid_read(inode));
1382         set_sd_v2_size(sd_v2, size);
1383         set_sd_v2_gid(sd_v2, i_gid_read(inode));
1384         set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec);
1385         set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec);
1386         set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec);
1387         set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1388         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1389                 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1390         else
1391                 set_sd_v2_generation(sd_v2, inode->i_generation);
1392         flags = REISERFS_I(inode)->i_attrs;
1393         i_attrs_to_sd_attrs(inode, &flags);
1394         set_sd_v2_attrs(sd_v2, flags);
1395 }
1396 
1397 /* used to copy inode's fields to old stat data */
1398 static void inode2sd_v1(void *sd, struct inode *inode, loff_t size)
1399 {
1400         struct stat_data_v1 *sd_v1 = (struct stat_data_v1 *)sd;
1401 
1402         set_sd_v1_mode(sd_v1, inode->i_mode);
1403         set_sd_v1_uid(sd_v1, i_uid_read(inode));
1404         set_sd_v1_gid(sd_v1, i_gid_read(inode));
1405         set_sd_v1_nlink(sd_v1, inode->i_nlink);
1406         set_sd_v1_size(sd_v1, size);
1407         set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec);
1408         set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec);
1409         set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec);
1410 
1411         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1412                 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1413         else
1414                 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1415 
1416         /* Sigh. i_first_direct_byte is back */
1417         set_sd_v1_first_direct_byte(sd_v1,
1418                                     REISERFS_I(inode)->i_first_direct_byte);
1419 }
1420 
1421 /*
1422  * NOTE, you must prepare the buffer head before sending it here,
1423  * and then log it after the call
1424  */
1425 static void update_stat_data(struct treepath *path, struct inode *inode,
1426                              loff_t size)
1427 {
1428         struct buffer_head *bh;
1429         struct item_head *ih;
1430 
1431         bh = PATH_PLAST_BUFFER(path);
1432         ih = tp_item_head(path);
1433 
1434         if (!is_statdata_le_ih(ih))
1435                 reiserfs_panic(inode->i_sb, "vs-13065", "key %k, found item %h",
1436                                INODE_PKEY(inode), ih);
1437 
1438         /* path points to old stat data */
1439         if (stat_data_v1(ih)) {
1440                 inode2sd_v1(ih_item_body(bh, ih), inode, size);
1441         } else {
1442                 inode2sd(ih_item_body(bh, ih), inode, size);
1443         }
1444 
1445         return;
1446 }
1447 
1448 void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,
1449                              struct inode *inode, loff_t size)
1450 {
1451         struct cpu_key key;
1452         INITIALIZE_PATH(path);
1453         struct buffer_head *bh;
1454         int fs_gen;
1455         struct item_head *ih, tmp_ih;
1456         int retval;
1457 
1458         BUG_ON(!th->t_trans_id);
1459 
1460         /* key type is unimportant */
1461         make_cpu_key(&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);
1462 
1463         for (;;) {
1464                 int pos;
1465                 /* look for the object's stat data */
1466                 retval = search_item(inode->i_sb, &key, &path);
1467                 if (retval == IO_ERROR) {
1468                         reiserfs_error(inode->i_sb, "vs-13050",
1469                                        "i/o failure occurred trying to "
1470                                        "update %K stat data", &key);
1471                         return;
1472                 }
1473                 if (retval == ITEM_NOT_FOUND) {
1474                         pos = PATH_LAST_POSITION(&path);
1475                         pathrelse(&path);
1476                         if (inode->i_nlink == 0) {
1477                                 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found"); */
1478                                 return;
1479                         }
1480                         reiserfs_warning(inode->i_sb, "vs-13060",
1481                                          "stat data of object %k (nlink == %d) "
1482                                          "not found (pos %d)",
1483                                          INODE_PKEY(inode), inode->i_nlink,
1484                                          pos);
1485                         reiserfs_check_path(&path);
1486                         return;
1487                 }
1488 
1489                 /*
1490                  * sigh, prepare_for_journal might schedule.  When it
1491                  * schedules the FS might change.  We have to detect that,
1492                  * and loop back to the search if the stat data item has moved
1493                  */
1494                 bh = get_last_bh(&path);
1495                 ih = tp_item_head(&path);
1496                 copy_item_head(&tmp_ih, ih);
1497                 fs_gen = get_generation(inode->i_sb);
1498                 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
1499 
1500                 /* Stat_data item has been moved after scheduling. */
1501                 if (fs_changed(fs_gen, inode->i_sb)
1502                     && item_moved(&tmp_ih, &path)) {
1503                         reiserfs_restore_prepared_buffer(inode->i_sb, bh);
1504                         continue;
1505                 }
1506                 break;
1507         }
1508         update_stat_data(&path, inode, size);
1509         journal_mark_dirty(th, bh);
1510         pathrelse(&path);
1511         return;
1512 }
1513 
1514 /*
1515  * reiserfs_read_locked_inode is called to read the inode off disk, and it
1516  * does a make_bad_inode when things go wrong.  But, we need to make sure
1517  * and clear the key in the private portion of the inode, otherwise a
1518  * corresponding iput might try to delete whatever object the inode last
1519  * represented.
1520  */
1521 static void reiserfs_make_bad_inode(struct inode *inode)
1522 {
1523         memset(INODE_PKEY(inode), 0, KEY_SIZE);
1524         make_bad_inode(inode);
1525 }
1526 
1527 /*
1528  * initially this function was derived from minix or ext2's analog and
1529  * evolved as the prototype did
1530  */
1531 int reiserfs_init_locked_inode(struct inode *inode, void *p)
1532 {
1533         struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p;
1534         inode->i_ino = args->objectid;
1535         INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1536         return 0;
1537 }
1538 
1539 /*
1540  * looks for stat data in the tree, and fills up the fields of in-core
1541  * inode stat data fields
1542  */
1543 void reiserfs_read_locked_inode(struct inode *inode,
1544                                 struct reiserfs_iget_args *args)
1545 {
1546         INITIALIZE_PATH(path_to_sd);
1547         struct cpu_key key;
1548         unsigned long dirino;
1549         int retval;
1550 
1551         dirino = args->dirid;
1552 
1553         /*
1554          * set version 1, version 2 could be used too, because stat data
1555          * key is the same in both versions
1556          */
1557         key.version = KEY_FORMAT_3_5;
1558         key.on_disk_key.k_dir_id = dirino;
1559         key.on_disk_key.k_objectid = inode->i_ino;
1560         key.on_disk_key.k_offset = 0;
1561         key.on_disk_key.k_type = 0;
1562 
1563         /* look for the object's stat data */
1564         retval = search_item(inode->i_sb, &key, &path_to_sd);
1565         if (retval == IO_ERROR) {
1566                 reiserfs_error(inode->i_sb, "vs-13070",
1567                                "i/o failure occurred trying to find "
1568                                "stat data of %K", &key);
1569                 reiserfs_make_bad_inode(inode);
1570                 return;
1571         }
1572 
1573         /* a stale NFS handle can trigger this without it being an error */
1574         if (retval != ITEM_FOUND) {
1575                 pathrelse(&path_to_sd);
1576                 reiserfs_make_bad_inode(inode);
1577                 clear_nlink(inode);
1578                 return;
1579         }
1580 
1581         init_inode(inode, &path_to_sd);
1582 
1583         /*
1584          * It is possible that knfsd is trying to access inode of a file
1585          * that is being removed from the disk by some other thread. As we
1586          * update sd on unlink all that is required is to check for nlink
1587          * here. This bug was first found by Sizif when debugging
1588          * SquidNG/Butterfly, forgotten, and found again after Philippe
1589          * Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1590 
1591          * More logical fix would require changes in fs/inode.c:iput() to
1592          * remove inode from hash-table _after_ fs cleaned disk stuff up and
1593          * in iget() to return NULL if I_FREEING inode is found in
1594          * hash-table.
1595          */
1596 
1597         /*
1598          * Currently there is one place where it's ok to meet inode with
1599          * nlink==0: processing of open-unlinked and half-truncated files
1600          * during mount (fs/reiserfs/super.c:finish_unfinished()).
1601          */
1602         if ((inode->i_nlink == 0) &&
1603             !REISERFS_SB(inode->i_sb)->s_is_unlinked_ok) {
1604                 reiserfs_warning(inode->i_sb, "vs-13075",
1605                                  "dead inode read from disk %K. "
1606                                  "This is likely to be race with knfsd. Ignore",
1607                                  &key);
1608                 reiserfs_make_bad_inode(inode);
1609         }
1610 
1611         /* init inode should be relsing */
1612         reiserfs_check_path(&path_to_sd);
1613 
1614         /*
1615          * Stat data v1 doesn't support ACLs.
1616          */
1617         if (get_inode_sd_version(inode) == STAT_DATA_V1)
1618                 cache_no_acl(inode);
1619 }
1620 
1621 /*
1622  * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1623  *
1624  * @inode:    inode from hash table to check
1625  * @opaque:   "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1626  *
1627  * This function is called by iget5_locked() to distinguish reiserfs inodes
1628  * having the same inode numbers. Such inodes can only exist due to some
1629  * error condition. One of them should be bad. Inodes with identical
1630  * inode numbers (objectids) are distinguished by parent directory ids.
1631  *
1632  */
1633 int reiserfs_find_actor(struct inode *inode, void *opaque)
1634 {
1635         struct reiserfs_iget_args *args;
1636 
1637         args = opaque;
1638         /* args is already in CPU order */
1639         return (inode->i_ino == args->objectid) &&
1640             (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1641 }
1642 
1643 struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key)
1644 {
1645         struct inode *inode;
1646         struct reiserfs_iget_args args;
1647         int depth;
1648 
1649         args.objectid = key->on_disk_key.k_objectid;
1650         args.dirid = key->on_disk_key.k_dir_id;
1651         depth = reiserfs_write_unlock_nested(s);
1652         inode = iget5_locked(s, key->on_disk_key.k_objectid,
1653                              reiserfs_find_actor, reiserfs_init_locked_inode,
1654                              (void *)(&args));
1655         reiserfs_write_lock_nested(s, depth);
1656         if (!inode)
1657                 return ERR_PTR(-ENOMEM);
1658 
1659         if (inode->i_state & I_NEW) {
1660                 reiserfs_read_locked_inode(inode, &args);
1661                 unlock_new_inode(inode);
1662         }
1663 
1664         if (comp_short_keys(INODE_PKEY(inode), key) || is_bad_inode(inode)) {
1665                 /* either due to i/o error or a stale NFS handle */
1666                 iput(inode);
1667                 inode = NULL;
1668         }
1669         return inode;
1670 }
1671 
1672 static struct dentry *reiserfs_get_dentry(struct super_block *sb,
1673         u32 objectid, u32 dir_id, u32 generation)
1674 
1675 {
1676         struct cpu_key key;
1677         struct inode *inode;
1678 
1679         key.on_disk_key.k_objectid = objectid;
1680         key.on_disk_key.k_dir_id = dir_id;
1681         reiserfs_write_lock(sb);
1682         inode = reiserfs_iget(sb, &key);
1683         if (inode && !IS_ERR(inode) && generation != 0 &&
1684             generation != inode->i_generation) {
1685                 iput(inode);
1686                 inode = NULL;
1687         }
1688         reiserfs_write_unlock(sb);
1689 
1690         return d_obtain_alias(inode);
1691 }
1692 
1693 struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
1694                 int fh_len, int fh_type)
1695 {
1696         /*
1697          * fhtype happens to reflect the number of u32s encoded.
1698          * due to a bug in earlier code, fhtype might indicate there
1699          * are more u32s then actually fitted.
1700          * so if fhtype seems to be more than len, reduce fhtype.
1701          * Valid types are:
1702          *   2 - objectid + dir_id - legacy support
1703          *   3 - objectid + dir_id + generation
1704          *   4 - objectid + dir_id + objectid and dirid of parent - legacy
1705          *   5 - objectid + dir_id + generation + objectid and dirid of parent
1706          *   6 - as above plus generation of directory
1707          * 6 does not fit in NFSv2 handles
1708          */
1709         if (fh_type > fh_len) {
1710                 if (fh_type != 6 || fh_len != 5)
1711                         reiserfs_warning(sb, "reiserfs-13077",
1712                                 "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1713                                 fh_type, fh_len);
1714                 fh_type = fh_len;
1715         }
1716         if (fh_len < 2)
1717                 return NULL;
1718 
1719         return reiserfs_get_dentry(sb, fid->raw[0], fid->raw[1],
1720                 (fh_type == 3 || fh_type >= 5) ? fid->raw[2] : 0);
1721 }
1722 
1723 struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,
1724                 int fh_len, int fh_type)
1725 {
1726         if (fh_type > fh_len)
1727                 fh_type = fh_len;
1728         if (fh_type < 4)
1729                 return NULL;
1730 
1731         return reiserfs_get_dentry(sb,
1732                 (fh_type >= 5) ? fid->raw[3] : fid->raw[2],
1733                 (fh_type >= 5) ? fid->raw[4] : fid->raw[3],
1734                 (fh_type == 6) ? fid->raw[5] : 0);
1735 }
1736 
1737 int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,
1738                        struct inode *parent)
1739 {
1740         int maxlen = *lenp;
1741 
1742         if (parent && (maxlen < 5)) {
1743                 *lenp = 5;
1744                 return FILEID_INVALID;
1745         } else if (maxlen < 3) {
1746                 *lenp = 3;
1747                 return FILEID_INVALID;
1748         }
1749 
1750         data[0] = inode->i_ino;
1751         data[1] = le32_to_cpu(INODE_PKEY(inode)->k_dir_id);
1752         data[2] = inode->i_generation;
1753         *lenp = 3;
1754         if (parent) {
1755                 data[3] = parent->i_ino;
1756                 data[4] = le32_to_cpu(INODE_PKEY(parent)->k_dir_id);
1757                 *lenp = 5;
1758                 if (maxlen >= 6) {
1759                         data[5] = parent->i_generation;
1760                         *lenp = 6;
1761                 }
1762         }
1763         return *lenp;
1764 }
1765 
1766 /*
1767  * looks for stat data, then copies fields to it, marks the buffer
1768  * containing stat data as dirty
1769  */
1770 /*
1771  * reiserfs inodes are never really dirty, since the dirty inode call
1772  * always logs them.  This call allows the VFS inode marking routines
1773  * to properly mark inodes for datasync and such, but only actually
1774  * does something when called for a synchronous update.
1775  */
1776 int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc)
1777 {
1778         struct reiserfs_transaction_handle th;
1779         int jbegin_count = 1;
1780 
1781         if (inode->i_sb->s_flags & MS_RDONLY)
1782                 return -EROFS;
1783         /*
1784          * memory pressure can sometimes initiate write_inode calls with
1785          * sync == 1,
1786          * these cases are just when the system needs ram, not when the
1787          * inode needs to reach disk for safety, and they can safely be
1788          * ignored because the altered inode has already been logged.
1789          */
1790         if (wbc->sync_mode == WB_SYNC_ALL && !(current->flags & PF_MEMALLOC)) {
1791                 reiserfs_write_lock(inode->i_sb);
1792                 if (!journal_begin(&th, inode->i_sb, jbegin_count)) {
1793                         reiserfs_update_sd(&th, inode);
1794                         journal_end_sync(&th);
1795                 }
1796                 reiserfs_write_unlock(inode->i_sb);
1797         }
1798         return 0;
1799 }
1800 
1801 /*
1802  * stat data of new object is inserted already, this inserts the item
1803  * containing "." and ".." entries
1804  */
1805 static int reiserfs_new_directory(struct reiserfs_transaction_handle *th,
1806                                   struct inode *inode,
1807                                   struct item_head *ih, struct treepath *path,
1808                                   struct inode *dir)
1809 {
1810         struct super_block *sb = th->t_super;
1811         char empty_dir[EMPTY_DIR_SIZE];
1812         char *body = empty_dir;
1813         struct cpu_key key;
1814         int retval;
1815 
1816         BUG_ON(!th->t_trans_id);
1817 
1818         _make_cpu_key(&key, KEY_FORMAT_3_5, le32_to_cpu(ih->ih_key.k_dir_id),
1819                       le32_to_cpu(ih->ih_key.k_objectid), DOT_OFFSET,
1820                       TYPE_DIRENTRY, 3 /*key length */ );
1821 
1822         /*
1823          * compose item head for new item. Directories consist of items of
1824          * old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1825          * is done by reiserfs_new_inode
1826          */
1827         if (old_format_only(sb)) {
1828                 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1829                                   TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1830 
1831                 make_empty_dir_item_v1(body, ih->ih_key.k_dir_id,
1832                                        ih->ih_key.k_objectid,
1833                                        INODE_PKEY(dir)->k_dir_id,
1834                                        INODE_PKEY(dir)->k_objectid);
1835         } else {
1836                 make_le_item_head(ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET,
1837                                   TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1838 
1839                 make_empty_dir_item(body, ih->ih_key.k_dir_id,
1840                                     ih->ih_key.k_objectid,
1841                                     INODE_PKEY(dir)->k_dir_id,
1842                                     INODE_PKEY(dir)->k_objectid);
1843         }
1844 
1845         /* look for place in the tree for new item */
1846         retval = search_item(sb, &key, path);
1847         if (retval == IO_ERROR) {
1848                 reiserfs_error(sb, "vs-13080",
1849                                "i/o failure occurred creating new directory");
1850                 return -EIO;
1851         }
1852         if (retval == ITEM_FOUND) {
1853                 pathrelse(path);
1854                 reiserfs_warning(sb, "vs-13070",
1855                                  "object with this key exists (%k)",
1856                                  &(ih->ih_key));
1857                 return -EEXIST;
1858         }
1859 
1860         /* insert item, that is empty directory item */
1861         return reiserfs_insert_item(th, path, &key, ih, inode, body);
1862 }
1863 
1864 /*
1865  * stat data of object has been inserted, this inserts the item
1866  * containing the body of symlink
1867  */
1868 static int reiserfs_new_symlink(struct reiserfs_transaction_handle *th,
1869                                 struct inode *inode,
1870                                 struct item_head *ih,
1871                                 struct treepath *path, const char *symname,
1872                                 int item_len)
1873 {
1874         struct super_block *sb = th->t_super;
1875         struct cpu_key key;
1876         int retval;
1877 
1878         BUG_ON(!th->t_trans_id);
1879 
1880         _make_cpu_key(&key, KEY_FORMAT_3_5,
1881                       le32_to_cpu(ih->ih_key.k_dir_id),
1882                       le32_to_cpu(ih->ih_key.k_objectid),
1883                       1, TYPE_DIRECT, 3 /*key length */ );
1884 
1885         make_le_item_head(ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len,
1886                           0 /*free_space */ );
1887 
1888         /* look for place in the tree for new item */
1889         retval = search_item(sb, &key, path);
1890         if (retval == IO_ERROR) {
1891                 reiserfs_error(sb, "vs-13080",
1892                                "i/o failure occurred creating new symlink");
1893                 return -EIO;
1894         }
1895         if (retval == ITEM_FOUND) {
1896                 pathrelse(path);
1897                 reiserfs_warning(sb, "vs-13080",
1898                                  "object with this key exists (%k)",
1899                                  &(ih->ih_key));
1900                 return -EEXIST;
1901         }
1902 
1903         /* insert item, that is body of symlink */
1904         return reiserfs_insert_item(th, path, &key, ih, inode, symname);
1905 }
1906 
1907 /*
1908  * inserts the stat data into the tree, and then calls
1909  * reiserfs_new_directory (to insert ".", ".." item if new object is
1910  * directory) or reiserfs_new_symlink (to insert symlink body if new
1911  * object is symlink) or nothing (if new object is regular file)
1912 
1913  * NOTE! uid and gid must already be set in the inode.  If we return
1914  * non-zero due to an error, we have to drop the quota previously allocated
1915  * for the fresh inode.  This can only be done outside a transaction, so
1916  * if we return non-zero, we also end the transaction.
1917  *
1918  * @th: active transaction handle
1919  * @dir: parent directory for new inode
1920  * @mode: mode of new inode
1921  * @symname: symlink contents if inode is symlink
1922  * @isize: 0 for regular file, EMPTY_DIR_SIZE for dirs, strlen(symname) for
1923  *         symlinks
1924  * @inode: inode to be filled
1925  * @security: optional security context to associate with this inode
1926  */
1927 int reiserfs_new_inode(struct reiserfs_transaction_handle *th,
1928                        struct inode *dir, umode_t mode, const char *symname,
1929                        /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1930                           strlen (symname) for symlinks) */
1931                        loff_t i_size, struct dentry *dentry,
1932                        struct inode *inode,
1933                        struct reiserfs_security_handle *security)
1934 {
1935         struct super_block *sb = dir->i_sb;
1936         struct reiserfs_iget_args args;
1937         INITIALIZE_PATH(path_to_key);
1938         struct cpu_key key;
1939         struct item_head ih;
1940         struct stat_data sd;
1941         int retval;
1942         int err;
1943         int depth;
1944 
1945         BUG_ON(!th->t_trans_id);
1946 
1947         depth = reiserfs_write_unlock_nested(sb);
1948         err = dquot_alloc_inode(inode);
1949         reiserfs_write_lock_nested(sb, depth);
1950         if (err)
1951                 goto out_end_trans;
1952         if (!dir->i_nlink) {
1953                 err = -EPERM;
1954                 goto out_bad_inode;
1955         }
1956 
1957         /* item head of new item */
1958         ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1959         ih.ih_key.k_objectid = cpu_to_le32(reiserfs_get_unused_objectid(th));
1960         if (!ih.ih_key.k_objectid) {
1961                 err = -ENOMEM;
1962                 goto out_bad_inode;
1963         }
1964         args.objectid = inode->i_ino = le32_to_cpu(ih.ih_key.k_objectid);
1965         if (old_format_only(sb))
1966                 make_le_item_head(&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET,
1967                                   TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1968         else
1969                 make_le_item_head(&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET,
1970                                   TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1971         memcpy(INODE_PKEY(inode), &ih.ih_key, KEY_SIZE);
1972         args.dirid = le32_to_cpu(ih.ih_key.k_dir_id);
1973 
1974         depth = reiserfs_write_unlock_nested(inode->i_sb);
1975         err = insert_inode_locked4(inode, args.objectid,
1976                              reiserfs_find_actor, &args);
1977         reiserfs_write_lock_nested(inode->i_sb, depth);
1978         if (err) {
1979                 err = -EINVAL;
1980                 goto out_bad_inode;
1981         }
1982 
1983         if (old_format_only(sb))
1984                 /*
1985                  * not a perfect generation count, as object ids can be reused,
1986                  * but this is as good as reiserfs can do right now.
1987                  * note that the private part of inode isn't filled in yet,
1988                  * we have to use the directory.
1989                  */
1990                 inode->i_generation = le32_to_cpu(INODE_PKEY(dir)->k_objectid);
1991         else
1992 #if defined( USE_INODE_GENERATION_COUNTER )
1993                 inode->i_generation =
1994                     le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1995 #else
1996                 inode->i_generation = ++event;
1997 #endif
1998 
1999         /* fill stat data */
2000         set_nlink(inode, (S_ISDIR(mode) ? 2 : 1));
2001 
2002         /* uid and gid must already be set by the caller for quota init */
2003 
2004         /* symlink cannot be immutable or append only, right? */
2005         if (S_ISLNK(inode->i_mode))
2006                 inode->i_flags &= ~(S_IMMUTABLE | S_APPEND);
2007 
2008         inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
2009         inode->i_size = i_size;
2010         inode->i_blocks = 0;
2011         inode->i_bytes = 0;
2012         REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
2013             U32_MAX /*NO_BYTES_IN_DIRECT_ITEM */ ;
2014 
2015         INIT_LIST_HEAD(&REISERFS_I(inode)->i_prealloc_list);
2016         REISERFS_I(inode)->i_flags = 0;
2017         REISERFS_I(inode)->i_prealloc_block = 0;
2018         REISERFS_I(inode)->i_prealloc_count = 0;
2019         REISERFS_I(inode)->i_trans_id = 0;
2020         REISERFS_I(inode)->i_jl = NULL;
2021         REISERFS_I(inode)->i_attrs =
2022             REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
2023         sd_attrs_to_i_attrs(REISERFS_I(inode)->i_attrs, inode);
2024         reiserfs_init_xattr_rwsem(inode);
2025 
2026         /* key to search for correct place for new stat data */
2027         _make_cpu_key(&key, KEY_FORMAT_3_6, le32_to_cpu(ih.ih_key.k_dir_id),
2028                       le32_to_cpu(ih.ih_key.k_objectid), SD_OFFSET,
2029                       TYPE_STAT_DATA, 3 /*key length */ );
2030 
2031         /* find proper place for inserting of stat data */
2032         retval = search_item(sb, &key, &path_to_key);
2033         if (retval == IO_ERROR) {
2034                 err = -EIO;
2035                 goto out_bad_inode;
2036         }
2037         if (retval == ITEM_FOUND) {
2038                 pathrelse(&path_to_key);
2039                 err = -EEXIST;
2040                 goto out_bad_inode;
2041         }
2042         if (old_format_only(sb)) {
2043                 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
2044                 if (i_uid_read(inode) & ~0xffff || i_gid_read(inode) & ~0xffff) {
2045                         pathrelse(&path_to_key);
2046                         err = -EINVAL;
2047                         goto out_bad_inode;
2048                 }
2049                 inode2sd_v1(&sd, inode, inode->i_size);
2050         } else {
2051                 inode2sd(&sd, inode, inode->i_size);
2052         }
2053         /*
2054          * store in in-core inode the key of stat data and version all
2055          * object items will have (directory items will have old offset
2056          * format, other new objects will consist of new items)
2057          */
2058         if (old_format_only(sb) || S_ISDIR(mode) || S_ISLNK(mode))
2059                 set_inode_item_key_version(inode, KEY_FORMAT_3_5);
2060         else
2061                 set_inode_item_key_version(inode, KEY_FORMAT_3_6);
2062         if (old_format_only(sb))
2063                 set_inode_sd_version(inode, STAT_DATA_V1);
2064         else
2065                 set_inode_sd_version(inode, STAT_DATA_V2);
2066 
2067         /* insert the stat data into the tree */
2068 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2069         if (REISERFS_I(dir)->new_packing_locality)
2070                 th->displace_new_blocks = 1;
2071 #endif
2072         retval =
2073             reiserfs_insert_item(th, &path_to_key, &key, &ih, inode,
2074                                  (char *)(&sd));
2075         if (retval) {
2076                 err = retval;
2077                 reiserfs_check_path(&path_to_key);
2078                 goto out_bad_inode;
2079         }
2080 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2081         if (!th->displace_new_blocks)
2082                 REISERFS_I(dir)->new_packing_locality = 0;
2083 #endif
2084         if (S_ISDIR(mode)) {
2085                 /* insert item with "." and ".." */
2086                 retval =
2087                     reiserfs_new_directory(th, inode, &ih, &path_to_key, dir);
2088         }
2089 
2090         if (S_ISLNK(mode)) {
2091                 /* insert body of symlink */
2092                 if (!old_format_only(sb))
2093                         i_size = ROUND_UP(i_size);
2094                 retval =
2095                     reiserfs_new_symlink(th, inode, &ih, &path_to_key, symname,
2096                                          i_size);
2097         }
2098         if (retval) {
2099                 err = retval;
2100                 reiserfs_check_path(&path_to_key);
2101                 journal_end(th);
2102                 goto out_inserted_sd;
2103         }
2104 
2105         if (reiserfs_posixacl(inode->i_sb)) {
2106                 reiserfs_write_unlock(inode->i_sb);
2107                 retval = reiserfs_inherit_default_acl(th, dir, dentry, inode);
2108                 reiserfs_write_lock(inode->i_sb);
2109                 if (retval) {
2110                         err = retval;
2111                         reiserfs_check_path(&path_to_key);
2112                         journal_end(th);
2113                         goto out_inserted_sd;
2114                 }
2115         } else if (inode->i_sb->s_flags & MS_POSIXACL) {
2116                 reiserfs_warning(inode->i_sb, "jdm-13090",
2117                                  "ACLs aren't enabled in the fs, "
2118                                  "but vfs thinks they are!");
2119         } else if (IS_PRIVATE(dir))
2120                 inode->i_flags |= S_PRIVATE;
2121 
2122         if (security->name) {
2123                 reiserfs_write_unlock(inode->i_sb);
2124                 retval = reiserfs_security_write(th, inode, security);
2125                 reiserfs_write_lock(inode->i_sb);
2126                 if (retval) {
2127                         err = retval;
2128                         reiserfs_check_path(&path_to_key);
2129                         retval = journal_end(th);
2130                         if (retval)
2131                                 err = retval;
2132                         goto out_inserted_sd;
2133                 }
2134         }
2135 
2136         reiserfs_update_sd(th, inode);
2137         reiserfs_check_path(&path_to_key);
2138 
2139         return 0;
2140 
2141 out_bad_inode:
2142         /* Invalidate the object, nothing was inserted yet */
2143         INODE_PKEY(inode)->k_objectid = 0;
2144 
2145         /* Quota change must be inside a transaction for journaling */
2146         depth = reiserfs_write_unlock_nested(inode->i_sb);
2147         dquot_free_inode(inode);
2148         reiserfs_write_lock_nested(inode->i_sb, depth);
2149 
2150 out_end_trans:
2151         journal_end(th);
2152         /*
2153          * Drop can be outside and it needs more credits so it's better
2154          * to have it outside
2155          */
2156         depth = reiserfs_write_unlock_nested(inode->i_sb);
2157         dquot_drop(inode);
2158         reiserfs_write_lock_nested(inode->i_sb, depth);
2159         inode->i_flags |= S_NOQUOTA;
2160         make_bad_inode(inode);
2161 
2162 out_inserted_sd:
2163         clear_nlink(inode);
2164         th->t_trans_id = 0;     /* so the caller can't use this handle later */
2165         unlock_new_inode(inode); /* OK to do even if we hadn't locked it */
2166         iput(inode);
2167         return err;
2168 }
2169 
2170 /*
2171  * finds the tail page in the page cache,
2172  * reads the last block in.
2173  *
2174  * On success, page_result is set to a locked, pinned page, and bh_result
2175  * is set to an up to date buffer for the last block in the file.  returns 0.
2176  *
2177  * tail conversion is not done, so bh_result might not be valid for writing
2178  * check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
2179  * trying to write the block.
2180  *
2181  * on failure, nonzero is returned, page_result and bh_result are untouched.
2182  */
2183 static int grab_tail_page(struct inode *inode,
2184                           struct page **page_result,
2185                           struct buffer_head **bh_result)
2186 {
2187 
2188         /*
2189          * we want the page with the last byte in the file,
2190          * not the page that will hold the next byte for appending
2191          */
2192         unsigned long index = (inode->i_size - 1) >> PAGE_SHIFT;
2193         unsigned long pos = 0;
2194         unsigned long start = 0;
2195         unsigned long blocksize = inode->i_sb->s_blocksize;
2196         unsigned long offset = (inode->i_size) & (PAGE_SIZE - 1);
2197         struct buffer_head *bh;
2198         struct buffer_head *head;
2199         struct page *page;
2200         int error;
2201 
2202         /*
2203          * we know that we are only called with inode->i_size > 0.
2204          * we also know that a file tail can never be as big as a block
2205          * If i_size % blocksize == 0, our file is currently block aligned
2206          * and it won't need converting or zeroing after a truncate.
2207          */
2208         if ((offset & (blocksize - 1)) == 0) {
2209                 return -ENOENT;
2210         }
2211         page = grab_cache_page(inode->i_mapping, index);
2212         error = -ENOMEM;
2213         if (!page) {
2214                 goto out;
2215         }
2216         /* start within the page of the last block in the file */
2217         start = (offset / blocksize) * blocksize;
2218 
2219         error = __block_write_begin(page, start, offset - start,
2220                                     reiserfs_get_block_create_0);
2221         if (error)
2222                 goto unlock;
2223 
2224         head = page_buffers(page);
2225         bh = head;
2226         do {
2227                 if (pos >= start) {
2228                         break;
2229                 }
2230                 bh = bh->b_this_page;
2231                 pos += blocksize;
2232         } while (bh != head);
2233 
2234         if (!buffer_uptodate(bh)) {
2235                 /*
2236                  * note, this should never happen, prepare_write should be
2237                  * taking care of this for us.  If the buffer isn't up to
2238                  * date, I've screwed up the code to find the buffer, or the
2239                  * code to call prepare_write
2240                  */
2241                 reiserfs_error(inode->i_sb, "clm-6000",
2242                                "error reading block %lu", bh->b_blocknr);
2243                 error = -EIO;
2244                 goto unlock;
2245         }
2246         *bh_result = bh;
2247         *page_result = page;
2248 
2249 out:
2250         return error;
2251 
2252 unlock:
2253         unlock_page(page);
2254         put_page(page);
2255         return error;
2256 }
2257 
2258 /*
2259  * vfs version of truncate file.  Must NOT be called with
2260  * a transaction already started.
2261  *
2262  * some code taken from block_truncate_page
2263  */
2264 int reiserfs_truncate_file(struct inode *inode, int update_timestamps)
2265 {
2266         struct reiserfs_transaction_handle th;
2267         /* we want the offset for the first byte after the end of the file */
2268         unsigned long offset = inode->i_size & (PAGE_SIZE - 1);
2269         unsigned blocksize = inode->i_sb->s_blocksize;
2270         unsigned length;
2271         struct page *page = NULL;
2272         int error;
2273         struct buffer_head *bh = NULL;
2274         int err2;
2275 
2276         reiserfs_write_lock(inode->i_sb);
2277 
2278         if (inode->i_size > 0) {
2279                 error = grab_tail_page(inode, &page, &bh);
2280                 if (error) {
2281                         /*
2282                          * -ENOENT means we truncated past the end of the
2283                          * file, and get_block_create_0 could not find a
2284                          * block to read in, which is ok.
2285                          */
2286                         if (error != -ENOENT)
2287                                 reiserfs_error(inode->i_sb, "clm-6001",
2288                                                "grab_tail_page failed %d",
2289                                                error);
2290                         page = NULL;
2291                         bh = NULL;
2292                 }
2293         }
2294 
2295         /*
2296          * so, if page != NULL, we have a buffer head for the offset at
2297          * the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
2298          * then we have an unformatted node.  Otherwise, we have a direct item,
2299          * and no zeroing is required on disk.  We zero after the truncate,
2300          * because the truncate might pack the item anyway
2301          * (it will unmap bh if it packs).
2302          *
2303          * it is enough to reserve space in transaction for 2 balancings:
2304          * one for "save" link adding and another for the first
2305          * cut_from_item. 1 is for update_sd
2306          */
2307         error = journal_begin(&th, inode->i_sb,
2308                               JOURNAL_PER_BALANCE_CNT * 2 + 1);
2309         if (error)
2310                 goto out;
2311         reiserfs_update_inode_transaction(inode);
2312         if (update_timestamps)
2313                 /*
2314                  * we are doing real truncate: if the system crashes
2315                  * before the last transaction of truncating gets committed
2316                  * - on reboot the file either appears truncated properly
2317                  * or not truncated at all
2318                  */
2319                 add_save_link(&th, inode, 1);
2320         err2 = reiserfs_do_truncate(&th, inode, page, update_timestamps);
2321         error = journal_end(&th);
2322         if (error)
2323                 goto out;
2324 
2325         /* check reiserfs_do_truncate after ending the transaction */
2326         if (err2) {
2327                 error = err2;
2328                 goto out;
2329         }
2330         
2331         if (update_timestamps) {
2332                 error = remove_save_link(inode, 1 /* truncate */);
2333                 if (error)
2334                         goto out;
2335         }
2336 
2337         if (page) {
2338                 length = offset & (blocksize - 1);
2339                 /* if we are not on a block boundary */
2340                 if (length) {
2341                         length = blocksize - length;
2342                         zero_user(page, offset, length);
2343                         if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2344                                 mark_buffer_dirty(bh);
2345                         }
2346                 }
2347                 unlock_page(page);
2348                 put_page(page);
2349         }
2350 
2351         reiserfs_write_unlock(inode->i_sb);
2352 
2353         return 0;
2354 out:
2355         if (page) {
2356                 unlock_page(page);
2357                 put_page(page);
2358         }
2359 
2360         reiserfs_write_unlock(inode->i_sb);
2361 
2362         return error;
2363 }
2364 
2365 static int map_block_for_writepage(struct inode *inode,
2366                                    struct buffer_head *bh_result,
2367                                    unsigned long block)
2368 {
2369         struct reiserfs_transaction_handle th;
2370         int fs_gen;
2371         struct item_head tmp_ih;
2372         struct item_head *ih;
2373         struct buffer_head *bh;
2374         __le32 *item;
2375         struct cpu_key key;
2376         INITIALIZE_PATH(path);
2377         int pos_in_item;
2378         int jbegin_count = JOURNAL_PER_BALANCE_CNT;
2379         loff_t byte_offset = ((loff_t)block << inode->i_sb->s_blocksize_bits)+1;
2380         int retval;
2381         int use_get_block = 0;
2382         int bytes_copied = 0;
2383         int copy_size;
2384         int trans_running = 0;
2385 
2386         /*
2387          * catch places below that try to log something without
2388          * starting a trans
2389          */
2390         th.t_trans_id = 0;
2391 
2392         if (!buffer_uptodate(bh_result)) {
2393                 return -EIO;
2394         }
2395 
2396         kmap(bh_result->b_page);
2397 start_over:
2398         reiserfs_write_lock(inode->i_sb);
2399         make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3);
2400 
2401 research:
2402         retval = search_for_position_by_key(inode->i_sb, &key, &path);
2403         if (retval != POSITION_FOUND) {
2404                 use_get_block = 1;
2405                 goto out;
2406         }
2407 
2408         bh = get_last_bh(&path);
2409         ih = tp_item_head(&path);
2410         item = tp_item_body(&path);
2411         pos_in_item = path.pos_in_item;
2412 
2413         /* we've found an unformatted node */
2414         if (indirect_item_found(retval, ih)) {
2415                 if (bytes_copied > 0) {
2416                         reiserfs_warning(inode->i_sb, "clm-6002",
2417                                          "bytes_copied %d", bytes_copied);
2418                 }
2419                 if (!get_block_num(item, pos_in_item)) {
2420                         /* crap, we are writing to a hole */
2421                         use_get_block = 1;
2422                         goto out;
2423                 }
2424                 set_block_dev_mapped(bh_result,
2425                                      get_block_num(item, pos_in_item), inode);
2426         } else if (is_direct_le_ih(ih)) {
2427                 char *p;
2428                 p = page_address(bh_result->b_page);
2429                 p += (byte_offset - 1) & (PAGE_SIZE - 1);
2430                 copy_size = ih_item_len(ih) - pos_in_item;
2431 
2432                 fs_gen = get_generation(inode->i_sb);
2433                 copy_item_head(&tmp_ih, ih);
2434 
2435                 if (!trans_running) {
2436                         /* vs-3050 is gone, no need to drop the path */
2437                         retval = journal_begin(&th, inode->i_sb, jbegin_count);
2438                         if (retval)
2439                                 goto out;
2440                         reiserfs_update_inode_transaction(inode);
2441                         trans_running = 1;
2442                         if (fs_changed(fs_gen, inode->i_sb)
2443                             && item_moved(&tmp_ih, &path)) {
2444                                 reiserfs_restore_prepared_buffer(inode->i_sb,
2445                                                                  bh);
2446                                 goto research;
2447                         }
2448                 }
2449 
2450                 reiserfs_prepare_for_journal(inode->i_sb, bh, 1);
2451 
2452                 if (fs_changed(fs_gen, inode->i_sb)
2453                     && item_moved(&tmp_ih, &path)) {
2454                         reiserfs_restore_prepared_buffer(inode->i_sb, bh);
2455                         goto research;
2456                 }
2457 
2458                 memcpy(ih_item_body(bh, ih) + pos_in_item, p + bytes_copied,
2459                        copy_size);
2460 
2461                 journal_mark_dirty(&th, bh);
2462                 bytes_copied += copy_size;
2463                 set_block_dev_mapped(bh_result, 0, inode);
2464 
2465                 /* are there still bytes left? */
2466                 if (bytes_copied < bh_result->b_size &&
2467                     (byte_offset + bytes_copied) < inode->i_size) {
2468                         set_cpu_key_k_offset(&key,
2469                                              cpu_key_k_offset(&key) +
2470                                              copy_size);
2471                         goto research;
2472                 }
2473         } else {
2474                 reiserfs_warning(inode->i_sb, "clm-6003",
2475                                  "bad item inode %lu", inode->i_ino);
2476                 retval = -EIO;
2477                 goto out;
2478         }
2479         retval = 0;
2480 
2481 out:
2482         pathrelse(&path);
2483         if (trans_running) {
2484                 int err = journal_end(&th);
2485                 if (err)
2486                         retval = err;
2487                 trans_running = 0;
2488         }
2489         reiserfs_write_unlock(inode->i_sb);
2490 
2491         /* this is where we fill in holes in the file. */
2492         if (use_get_block) {
2493                 retval = reiserfs_get_block(inode, block, bh_result,
2494                                             GET_BLOCK_CREATE | GET_BLOCK_NO_IMUX
2495                                             | GET_BLOCK_NO_DANGLE);
2496                 if (!retval) {
2497                         if (!buffer_mapped(bh_result)
2498                             || bh_result->b_blocknr == 0) {
2499                                 /* get_block failed to find a mapped unformatted node. */
2500                                 use_get_block = 0;
2501                                 goto start_over;
2502                         }
2503                 }
2504         }
2505         kunmap(bh_result->b_page);
2506 
2507         if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2508                 /*
2509                  * we've copied data from the page into the direct item, so the
2510                  * buffer in the page is now clean, mark it to reflect that.
2511                  */
2512                 lock_buffer(bh_result);
2513                 clear_buffer_dirty(bh_result);
2514                 unlock_buffer(bh_result);
2515         }
2516         return retval;
2517 }
2518 
2519 /*
2520  * mason@suse.com: updated in 2.5.54 to follow the same general io
2521  * start/recovery path as __block_write_full_page, along with special
2522  * code to handle reiserfs tails.
2523  */
2524 static int reiserfs_write_full_page(struct page *page,
2525                                     struct writeback_control *wbc)
2526 {
2527         struct inode *inode = page->mapping->host;
2528         unsigned long end_index = inode->i_size >> PAGE_SHIFT;
2529         int error = 0;
2530         unsigned long block;
2531         sector_t last_block;
2532         struct buffer_head *head, *bh;
2533         int partial = 0;
2534         int nr = 0;
2535         int checked = PageChecked(page);
2536         struct reiserfs_transaction_handle th;
2537         struct super_block *s = inode->i_sb;
2538         int bh_per_page = PAGE_SIZE / s->s_blocksize;
2539         th.t_trans_id = 0;
2540 
2541         /* no logging allowed when nonblocking or from PF_MEMALLOC */
2542         if (checked && (current->flags & PF_MEMALLOC)) {
2543                 redirty_page_for_writepage(wbc, page);
2544                 unlock_page(page);
2545                 return 0;
2546         }
2547 
2548         /*
2549          * The page dirty bit is cleared before writepage is called, which
2550          * means we have to tell create_empty_buffers to make dirty buffers
2551          * The page really should be up to date at this point, so tossing
2552          * in the BH_Uptodate is just a sanity check.
2553          */
2554         if (!page_has_buffers(page)) {
2555                 create_empty_buffers(page, s->s_blocksize,
2556                                      (1 << BH_Dirty) | (1 << BH_Uptodate));
2557         }
2558         head = page_buffers(page);
2559 
2560         /*
2561          * last page in the file, zero out any contents past the
2562          * last byte in the file
2563          */
2564         if (page->index >= end_index) {
2565                 unsigned last_offset;
2566 
2567                 last_offset = inode->i_size & (PAGE_SIZE - 1);
2568                 /* no file contents in this page */
2569                 if (page->index >= end_index + 1 || !last_offset) {
2570                         unlock_page(page);
2571                         return 0;
2572                 }
2573                 zero_user_segment(page, last_offset, PAGE_SIZE);
2574         }
2575         bh = head;
2576         block = page->index << (PAGE_SHIFT - s->s_blocksize_bits);
2577         last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
2578         /* first map all the buffers, logging any direct items we find */
2579         do {
2580                 if (block > last_block) {
2581                         /*
2582                          * This can happen when the block size is less than
2583                          * the page size.  The corresponding bytes in the page
2584                          * were zero filled above
2585                          */
2586                         clear_buffer_dirty(bh);
2587                         set_buffer_uptodate(bh);
2588                 } else if ((checked || buffer_dirty(bh)) &&
2589                            (!buffer_mapped(bh) || (buffer_mapped(bh)
2590                                                        && bh->b_blocknr ==
2591                                                        0))) {
2592                         /*
2593                          * not mapped yet, or it points to a direct item, search
2594                          * the btree for the mapping info, and log any direct
2595                          * items found
2596                          */
2597                         if ((error = map_block_for_writepage(inode, bh, block))) {
2598                                 goto fail;
2599                         }
2600                 }
2601                 bh = bh->b_this_page;
2602                 block++;
2603         } while (bh != head);
2604 
2605         /*
2606          * we start the transaction after map_block_for_writepage,
2607          * because it can create holes in the file (an unbounded operation).
2608          * starting it here, we can make a reliable estimate for how many
2609          * blocks we're going to log
2610          */
2611         if (checked) {
2612                 ClearPageChecked(page);
2613                 reiserfs_write_lock(s);
2614                 error = journal_begin(&th, s, bh_per_page + 1);
2615                 if (error) {
2616                         reiserfs_write_unlock(s);
2617                         goto fail;
2618                 }
2619                 reiserfs_update_inode_transaction(inode);
2620         }
2621         /* now go through and lock any dirty buffers on the page */
2622         do {
2623                 get_bh(bh);
2624                 if (!buffer_mapped(bh))
2625                         continue;
2626                 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2627                         continue;
2628 
2629                 if (checked) {
2630                         reiserfs_prepare_for_journal(s, bh, 1);
2631                         journal_mark_dirty(&th, bh);
2632                         continue;
2633                 }
2634                 /*
2635                  * from this point on, we know the buffer is mapped to a
2636                  * real block and not a direct item
2637                  */
2638                 if (wbc->sync_mode != WB_SYNC_NONE) {
2639                         lock_buffer(bh);
2640                 } else {
2641                         if (!trylock_buffer(bh)) {
2642                                 redirty_page_for_writepage(wbc, page);
2643                                 continue;
2644                         }
2645                 }
2646                 if (test_clear_buffer_dirty(bh)) {
2647                         mark_buffer_async_write(bh);
2648                 } else {
2649                         unlock_buffer(bh);
2650                 }
2651         } while ((bh = bh->b_this_page) != head);
2652 
2653         if (checked) {
2654                 error = journal_end(&th);
2655                 reiserfs_write_unlock(s);
2656                 if (error)
2657                         goto fail;
2658         }
2659         BUG_ON(PageWriteback(page));
2660         set_page_writeback(page);
2661         unlock_page(page);
2662 
2663         /*
2664          * since any buffer might be the only dirty buffer on the page,
2665          * the first submit_bh can bring the page out of writeback.
2666          * be careful with the buffers.
2667          */
2668         do {
2669                 struct buffer_head *next = bh->b_this_page;
2670                 if (buffer_async_write(bh)) {
2671                         submit_bh(WRITE, bh);
2672                         nr++;
2673                 }
2674                 put_bh(bh);
2675                 bh = next;
2676         } while (bh != head);
2677 
2678         error = 0;
2679 done:
2680         if (nr == 0) {
2681                 /*
2682                  * if this page only had a direct item, it is very possible for
2683                  * no io to be required without there being an error.  Or,
2684                  * someone else could have locked them and sent them down the
2685                  * pipe without locking the page
2686                  */
2687                 bh = head;
2688                 do {
2689                         if (!buffer_uptodate(bh)) {
2690                                 partial = 1;
2691                                 break;
2692                         }
2693                         bh = bh->b_this_page;
2694                 } while (bh != head);
2695                 if (!partial)
2696                         SetPageUptodate(page);
2697                 end_page_writeback(page);
2698         }
2699         return error;
2700 
2701 fail:
2702         /*
2703          * catches various errors, we need to make sure any valid dirty blocks
2704          * get to the media.  The page is currently locked and not marked for
2705          * writeback
2706          */
2707         ClearPageUptodate(page);
2708         bh = head;
2709         do {
2710                 get_bh(bh);
2711                 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2712                         lock_buffer(bh);
2713                         mark_buffer_async_write(bh);
2714                 } else {
2715                         /*
2716                          * clear any dirty bits that might have come from
2717                          * getting attached to a dirty page
2718                          */
2719                         clear_buffer_dirty(bh);
2720                 }
2721                 bh = bh->b_this_page;
2722         } while (bh != head);
2723         SetPageError(page);
2724         BUG_ON(PageWriteback(page));
2725         set_page_writeback(page);
2726         unlock_page(page);
2727         do {
2728                 struct buffer_head *next = bh->b_this_page;
2729                 if (buffer_async_write(bh)) {
2730                         clear_buffer_dirty(bh);
2731                         submit_bh(WRITE, bh);
2732                         nr++;
2733                 }
2734                 put_bh(bh);
2735                 bh = next;
2736         } while (bh != head);
2737         goto done;
2738 }
2739 
2740 static int reiserfs_readpage(struct file *f, struct page *page)
2741 {
2742         return block_read_full_page(page, reiserfs_get_block);
2743 }
2744 
2745 static int reiserfs_writepage(struct page *page, struct writeback_control *wbc)
2746 {
2747         struct inode *inode = page->mapping->host;
2748         reiserfs_wait_on_write_block(inode->i_sb);
2749         return reiserfs_write_full_page(page, wbc);
2750 }
2751 
2752 static void reiserfs_truncate_failed_write(struct inode *inode)
2753 {
2754         truncate_inode_pages(inode->i_mapping, inode->i_size);
2755         reiserfs_truncate_file(inode, 0);
2756 }
2757 
2758 static int reiserfs_write_begin(struct file *file,
2759                                 struct address_space *mapping,
2760                                 loff_t pos, unsigned len, unsigned flags,
2761                                 struct page **pagep, void **fsdata)
2762 {
2763         struct inode *inode;
2764         struct page *page;
2765         pgoff_t index;
2766         int ret;
2767         int old_ref = 0;
2768 
2769         inode = mapping->host;
2770         *fsdata = NULL;
2771         if (flags & AOP_FLAG_CONT_EXPAND &&
2772             (pos & (inode->i_sb->s_blocksize - 1)) == 0) {
2773                 pos ++;
2774                 *fsdata = (void *)(unsigned long)flags;
2775         }
2776 
2777         index = pos >> PAGE_SHIFT;
2778         page = grab_cache_page_write_begin(mapping, index, flags);
2779         if (!page)
2780                 return -ENOMEM;
2781         *pagep = page;
2782 
2783         reiserfs_wait_on_write_block(inode->i_sb);
2784         fix_tail_page_for_writing(page);
2785         if (reiserfs_transaction_running(inode->i_sb)) {
2786                 struct reiserfs_transaction_handle *th;
2787                 th = (struct reiserfs_transaction_handle *)current->
2788                     journal_info;
2789                 BUG_ON(!th->t_refcount);
2790                 BUG_ON(!th->t_trans_id);
2791                 old_ref = th->t_refcount;
2792                 th->t_refcount++;
2793         }
2794         ret = __block_write_begin(page, pos, len, reiserfs_get_block);
2795         if (ret && reiserfs_transaction_running(inode->i_sb)) {
2796                 struct reiserfs_transaction_handle *th = current->journal_info;
2797                 /*
2798                  * this gets a little ugly.  If reiserfs_get_block returned an
2799                  * error and left a transacstion running, we've got to close
2800                  * it, and we've got to free handle if it was a persistent
2801                  * transaction.
2802                  *
2803                  * But, if we had nested into an existing transaction, we need
2804                  * to just drop the ref count on the handle.
2805                  *
2806                  * If old_ref == 0, the transaction is from reiserfs_get_block,
2807                  * and it was a persistent trans.  Otherwise, it was nested
2808                  * above.
2809                  */
2810                 if (th->t_refcount > old_ref) {
2811                         if (old_ref)
2812                                 th->t_refcount--;
2813                         else {
2814                                 int err;
2815                                 reiserfs_write_lock(inode->i_sb);
2816                                 err = reiserfs_end_persistent_transaction(th);
2817                                 reiserfs_write_unlock(inode->i_sb);
2818                                 if (err)
2819                                         ret = err;
2820                         }
2821                 }
2822         }
2823         if (ret) {
2824                 unlock_page(page);
2825                 put_page(page);
2826                 /* Truncate allocated blocks */
2827                 reiserfs_truncate_failed_write(inode);
2828         }
2829         return ret;
2830 }
2831 
2832 int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len)
2833 {
2834         struct inode *inode = page->mapping->host;
2835         int ret;
2836         int old_ref = 0;
2837         int depth;
2838 
2839         depth = reiserfs_write_unlock_nested(inode->i_sb);
2840         reiserfs_wait_on_write_block(inode->i_sb);
2841         reiserfs_write_lock_nested(inode->i_sb, depth);
2842 
2843         fix_tail_page_for_writing(page);
2844         if (reiserfs_transaction_running(inode->i_sb)) {
2845                 struct reiserfs_transaction_handle *th;
2846                 th = (struct reiserfs_transaction_handle *)current->
2847                     journal_info;
2848                 BUG_ON(!th->t_refcount);
2849                 BUG_ON(!th->t_trans_id);
2850                 old_ref = th->t_refcount;
2851                 th->t_refcount++;
2852         }
2853 
2854         ret = __block_write_begin(page, from, len, reiserfs_get_block);
2855         if (ret && reiserfs_transaction_running(inode->i_sb)) {
2856                 struct reiserfs_transaction_handle *th = current->journal_info;
2857                 /*
2858                  * this gets a little ugly.  If reiserfs_get_block returned an
2859                  * error and left a transacstion running, we've got to close
2860                  * it, and we've got to free handle if it was a persistent
2861                  * transaction.
2862                  *
2863                  * But, if we had nested into an existing transaction, we need
2864                  * to just drop the ref count on the handle.
2865                  *
2866                  * If old_ref == 0, the transaction is from reiserfs_get_block,
2867                  * and it was a persistent trans.  Otherwise, it was nested
2868                  * above.
2869                  */
2870                 if (th->t_refcount > old_ref) {
2871                         if (old_ref)
2872                                 th->t_refcount--;
2873                         else {
2874                                 int err;
2875                                 reiserfs_write_lock(inode->i_sb);
2876                                 err = reiserfs_end_persistent_transaction(th);
2877                                 reiserfs_write_unlock(inode->i_sb);
2878                                 if (err)
2879                                         ret = err;
2880                         }
2881                 }
2882         }
2883         return ret;
2884 
2885 }
2886 
2887 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block)
2888 {
2889         return generic_block_bmap(as, block, reiserfs_bmap);
2890 }
2891 
2892 static int reiserfs_write_end(struct file *file, struct address_space *mapping,
2893                               loff_t pos, unsigned len, unsigned copied,
2894                               struct page *page, void *fsdata)
2895 {
2896         struct inode *inode = page->mapping->host;
2897         int ret = 0;
2898         int update_sd = 0;
2899         struct reiserfs_transaction_handle *th;
2900         unsigned start;
2901         bool locked = false;
2902 
2903         if ((unsigned long)fsdata & AOP_FLAG_CONT_EXPAND)
2904                 pos ++;
2905 
2906         reiserfs_wait_on_write_block(inode->i_sb);
2907         if (reiserfs_transaction_running(inode->i_sb))
2908                 th = current->journal_info;
2909         else
2910                 th = NULL;
2911 
2912         start = pos & (PAGE_SIZE - 1);
2913         if (unlikely(copied < len)) {
2914                 if (!PageUptodate(page))
2915                         copied = 0;
2916 
2917                 page_zero_new_buffers(page, start + copied, start + len);
2918         }
2919         flush_dcache_page(page);
2920 
2921         reiserfs_commit_page(inode, page, start, start + copied);
2922 
2923         /*
2924          * generic_commit_write does this for us, but does not update the
2925          * transaction tracking stuff when the size changes.  So, we have
2926          * to do the i_size updates here.
2927          */
2928         if (pos + copied > inode->i_size) {
2929                 struct reiserfs_transaction_handle myth;
2930                 reiserfs_write_lock(inode->i_sb);
2931                 locked = true;
2932                 /*
2933                  * If the file have grown beyond the border where it
2934                  * can have a tail, unmark it as needing a tail
2935                  * packing
2936                  */
2937                 if ((have_large_tails(inode->i_sb)
2938                      && inode->i_size > i_block_size(inode) * 4)
2939                     || (have_small_tails(inode->i_sb)
2940                         && inode->i_size > i_block_size(inode)))
2941                         REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
2942 
2943                 ret = journal_begin(&myth, inode->i_sb, 1);
2944                 if (ret)
2945                         goto journal_error;
2946 
2947                 reiserfs_update_inode_transaction(inode);
2948                 inode->i_size = pos + copied;
2949                 /*
2950                  * this will just nest into our transaction.  It's important
2951                  * to use mark_inode_dirty so the inode gets pushed around on
2952                  * the dirty lists, and so that O_SYNC works as expected
2953                  */
2954                 mark_inode_dirty(inode);
2955                 reiserfs_update_sd(&myth, inode);
2956                 update_sd = 1;
2957                 ret = journal_end(&myth);
2958                 if (ret)
2959                         goto journal_error;
2960         }
2961         if (th) {
2962                 if (!locked) {
2963                         reiserfs_write_lock(inode->i_sb);
2964                         locked = true;
2965                 }
2966                 if (!update_sd)
2967                         mark_inode_dirty(inode);
2968                 ret = reiserfs_end_persistent_transaction(th);
2969                 if (ret)
2970                         goto out;
2971         }
2972 
2973 out:
2974         if (locked)
2975                 reiserfs_write_unlock(inode->i_sb);
2976         unlock_page(page);
2977         put_page(page);
2978 
2979         if (pos + len > inode->i_size)
2980                 reiserfs_truncate_failed_write(inode);
2981 
2982         return ret == 0 ? copied : ret;
2983 
2984 journal_error:
2985         reiserfs_write_unlock(inode->i_sb);
2986         locked = false;
2987         if (th) {
2988                 if (!update_sd)
2989                         reiserfs_update_sd(th, inode);
2990                 ret = reiserfs_end_persistent_transaction(th);
2991         }
2992         goto out;
2993 }
2994 
2995 int reiserfs_commit_write(struct file *f, struct page *page,
2996                           unsigned from, unsigned to)
2997 {
2998         struct inode *inode = page->mapping->host;
2999         loff_t pos = ((loff_t) page->index << PAGE_SHIFT) + to;
3000         int ret = 0;
3001         int update_sd = 0;
3002         struct reiserfs_transaction_handle *th = NULL;
3003         int depth;
3004 
3005         depth = reiserfs_write_unlock_nested(inode->i_sb);
3006         reiserfs_wait_on_write_block(inode->i_sb);
3007         reiserfs_write_lock_nested(inode->i_sb, depth);
3008 
3009         if (reiserfs_transaction_running(inode->i_sb)) {
3010                 th = current->journal_info;
3011         }
3012         reiserfs_commit_page(inode, page, from, to);
3013 
3014         /*
3015          * generic_commit_write does this for us, but does not update the
3016          * transaction tracking stuff when the size changes.  So, we have
3017          * to do the i_size updates here.
3018          */
3019         if (pos > inode->i_size) {
3020                 struct reiserfs_transaction_handle myth;
3021                 /*
3022                  * If the file have grown beyond the border where it
3023                  * can have a tail, unmark it as needing a tail
3024                  * packing
3025                  */
3026                 if ((have_large_tails(inode->i_sb)
3027                      && inode->i_size > i_block_size(inode) * 4)
3028                     || (have_small_tails(inode->i_sb)
3029                         && inode->i_size > i_block_size(inode)))
3030                         REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
3031 
3032                 ret = journal_begin(&myth, inode->i_sb, 1);
3033                 if (ret)
3034                         goto journal_error;
3035 
3036                 reiserfs_update_inode_transaction(inode);
3037                 inode->i_size = pos;
3038                 /*
3039                  * this will just nest into our transaction.  It's important
3040                  * to use mark_inode_dirty so the inode gets pushed around
3041                  * on the dirty lists, and so that O_SYNC works as expected
3042                  */
3043                 mark_inode_dirty(inode);
3044                 reiserfs_update_sd(&myth, inode);
3045                 update_sd = 1;
3046                 ret = journal_end(&myth);
3047                 if (ret)
3048                         goto journal_error;
3049         }
3050         if (th) {
3051                 if (!update_sd)
3052                         mark_inode_dirty(inode);
3053                 ret = reiserfs_end_persistent_transaction(th);
3054                 if (ret)
3055                         goto out;
3056         }
3057 
3058 out:
3059         return ret;
3060 
3061 journal_error:
3062         if (th) {
3063                 if (!update_sd)
3064                         reiserfs_update_sd(th, inode);
3065                 ret = reiserfs_end_persistent_transaction(th);
3066         }
3067 
3068         return ret;
3069 }
3070 
3071 void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode)
3072 {
3073         if (reiserfs_attrs(inode->i_sb)) {
3074                 if (sd_attrs & REISERFS_SYNC_FL)
3075                         inode->i_flags |= S_SYNC;
3076                 else
3077                         inode->i_flags &= ~S_SYNC;
3078                 if (sd_attrs & REISERFS_IMMUTABLE_FL)
3079                         inode->i_flags |= S_IMMUTABLE;
3080                 else
3081                         inode->i_flags &= ~S_IMMUTABLE;
3082                 if (sd_attrs & REISERFS_APPEND_FL)
3083                         inode->i_flags |= S_APPEND;
3084                 else
3085                         inode->i_flags &= ~S_APPEND;
3086                 if (sd_attrs & REISERFS_NOATIME_FL)
3087                         inode->i_flags |= S_NOATIME;
3088                 else
3089                         inode->i_flags &= ~S_NOATIME;
3090                 if (sd_attrs & REISERFS_NOTAIL_FL)
3091                         REISERFS_I(inode)->i_flags |= i_nopack_mask;
3092                 else
3093                         REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
3094         }
3095 }
3096 
3097 void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs)
3098 {
3099         if (reiserfs_attrs(inode->i_sb)) {
3100                 if (inode->i_flags & S_IMMUTABLE)
3101                         *sd_attrs |= REISERFS_IMMUTABLE_FL;
3102                 else
3103                         *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
3104                 if (inode->i_flags & S_SYNC)
3105                         *sd_attrs |= REISERFS_SYNC_FL;
3106                 else
3107                         *sd_attrs &= ~REISERFS_SYNC_FL;
3108                 if (inode->i_flags & S_NOATIME)
3109                         *sd_attrs |= REISERFS_NOATIME_FL;
3110                 else
3111                         *sd_attrs &= ~REISERFS_NOATIME_FL;
3112                 if (REISERFS_I(inode)->i_flags & i_nopack_mask)
3113                         *sd_attrs |= REISERFS_NOTAIL_FL;
3114                 else
3115                         *sd_attrs &= ~REISERFS_NOTAIL_FL;
3116         }
3117 }
3118 
3119 /*
3120  * decide if this buffer needs to stay around for data logging or ordered
3121  * write purposes
3122  */
3123 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
3124 {
3125         int ret = 1;
3126         struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3127 
3128         lock_buffer(bh);
3129         spin_lock(&j->j_dirty_buffers_lock);
3130         if (!buffer_mapped(bh)) {
3131                 goto free_jh;
3132         }
3133         /*
3134          * the page is locked, and the only places that log a data buffer
3135          * also lock the page.
3136          */
3137         if (reiserfs_file_data_log(inode)) {
3138                 /*
3139                  * very conservative, leave the buffer pinned if
3140                  * anyone might need it.
3141                  */
3142                 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
3143                         ret = 0;
3144                 }
3145         } else  if (buffer_dirty(bh)) {
3146                 struct reiserfs_journal_list *jl;
3147                 struct reiserfs_jh *jh = bh->b_private;
3148 
3149                 /*
3150                  * why is this safe?
3151                  * reiserfs_setattr updates i_size in the on disk
3152                  * stat data before allowing vmtruncate to be called.
3153                  *
3154                  * If buffer was put onto the ordered list for this
3155                  * transaction, we know for sure either this transaction
3156                  * or an older one already has updated i_size on disk,
3157                  * and this ordered data won't be referenced in the file
3158                  * if we crash.
3159                  *
3160                  * if the buffer was put onto the ordered list for an older
3161                  * transaction, we need to leave it around
3162                  */
3163                 if (jh && (jl = jh->jl)
3164                     && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
3165                         ret = 0;
3166         }
3167 free_jh:
3168         if (ret && bh->b_private) {
3169                 reiserfs_free_jh(bh);
3170         }
3171         spin_unlock(&j->j_dirty_buffers_lock);
3172         unlock_buffer(bh);
3173         return ret;
3174 }
3175 
3176 /* clm -- taken from fs/buffer.c:block_invalidate_page */
3177 static void reiserfs_invalidatepage(struct page *page, unsigned int offset,
3178                                     unsigned int length)
3179 {
3180         struct buffer_head *head, *bh, *next;
3181         struct inode *inode = page->mapping->host;
3182         unsigned int curr_off = 0;
3183         unsigned int stop = offset + length;
3184         int partial_page = (offset || length < PAGE_SIZE);
3185         int ret = 1;
3186 
3187         BUG_ON(!PageLocked(page));
3188 
3189         if (!partial_page)
3190                 ClearPageChecked(page);
3191 
3192         if (!page_has_buffers(page))
3193                 goto out;
3194 
3195         head = page_buffers(page);
3196         bh = head;
3197         do {
3198                 unsigned int next_off = curr_off + bh->b_size;
3199                 next = bh->b_this_page;
3200 
3201                 if (next_off > stop)
3202                         goto out;
3203 
3204                 /*
3205                  * is this block fully invalidated?
3206                  */
3207                 if (offset <= curr_off) {
3208                         if (invalidatepage_can_drop(inode, bh))
3209                                 reiserfs_unmap_buffer(bh);
3210                         else
3211                                 ret = 0;
3212                 }
3213                 curr_off = next_off;
3214                 bh = next;
3215         } while (bh != head);
3216 
3217         /*
3218          * We release buffers only if the entire page is being invalidated.
3219          * The get_block cached value has been unconditionally invalidated,
3220          * so real IO is not possible anymore.
3221          */
3222         if (!partial_page && ret) {
3223                 ret = try_to_release_page(page, 0);
3224                 /* maybe should BUG_ON(!ret); - neilb */
3225         }
3226 out:
3227         return;
3228 }
3229 
3230 static int reiserfs_set_page_dirty(struct page *page)
3231 {
3232         struct inode *inode = page->mapping->host;
3233         if (reiserfs_file_data_log(inode)) {
3234                 SetPageChecked(page);
3235                 return __set_page_dirty_nobuffers(page);
3236         }
3237         return __set_page_dirty_buffers(page);
3238 }
3239 
3240 /*
3241  * Returns 1 if the page's buffers were dropped.  The page is locked.
3242  *
3243  * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
3244  * in the buffers at page_buffers(page).
3245  *
3246  * even in -o notail mode, we can't be sure an old mount without -o notail
3247  * didn't create files with tails.
3248  */
3249 static int reiserfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
3250 {
3251         struct inode *inode = page->mapping->host;
3252         struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb);
3253         struct buffer_head *head;
3254         struct buffer_head *bh;
3255         int ret = 1;
3256 
3257         WARN_ON(PageChecked(page));
3258         spin_lock(&j->j_dirty_buffers_lock);
3259         head = page_buffers(page);
3260         bh = head;
3261         do {
3262                 if (bh->b_private) {
3263                         if (!buffer_dirty(bh) && !buffer_locked(bh)) {
3264                                 reiserfs_free_jh(bh);
3265                         } else {
3266                                 ret = 0;
3267                                 break;
3268                         }
3269                 }
3270                 bh = bh->b_this_page;
3271         } while (bh != head);
3272         if (ret)
3273                 ret = try_to_free_buffers(page);
3274         spin_unlock(&j->j_dirty_buffers_lock);
3275         return ret;
3276 }
3277 
3278 /*
3279  * We thank Mingming Cao for helping us understand in great detail what
3280  * to do in this section of the code.
3281  */
3282 static ssize_t reiserfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
3283                                   loff_t offset)
3284 {
3285         struct file *file = iocb->ki_filp;
3286         struct inode *inode = file->f_mapping->host;
3287         size_t count = iov_iter_count(iter);
3288         ssize_t ret;
3289 
3290         ret = blockdev_direct_IO(iocb, inode, iter, offset,
3291                                  reiserfs_get_blocks_direct_io);
3292 
3293         /*
3294          * In case of error extending write may have instantiated a few
3295          * blocks outside i_size. Trim these off again.
3296          */
3297         if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) {
3298                 loff_t isize = i_size_read(inode);
3299                 loff_t end = offset + count;
3300 
3301                 if ((end > isize) && inode_newsize_ok(inode, isize) == 0) {
3302                         truncate_setsize(inode, isize);
3303                         reiserfs_vfs_truncate_file(inode);
3304                 }
3305         }
3306 
3307         return ret;
3308 }
3309 
3310 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)
3311 {
3312         struct inode *inode = d_inode(dentry);
3313         unsigned int ia_valid;
3314         int error;
3315 
3316         error = inode_change_ok(inode, attr);
3317         if (error)
3318                 return error;
3319 
3320         /* must be turned off for recursive notify_change calls */
3321         ia_valid = attr->ia_valid &= ~(ATTR_KILL_SUID|ATTR_KILL_SGID);
3322 
3323         if (is_quota_modification(inode, attr)) {
3324                 error = dquot_initialize(inode);
3325                 if (error)
3326                         return error;
3327         }
3328         reiserfs_write_lock(inode->i_sb);
3329         if (attr->ia_valid & ATTR_SIZE) {
3330                 /*
3331                  * version 2 items will be caught by the s_maxbytes check
3332                  * done for us in vmtruncate
3333                  */
3334                 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
3335                     attr->ia_size > MAX_NON_LFS) {
3336                         reiserfs_write_unlock(inode->i_sb);
3337                         error = -EFBIG;
3338                         goto out;
3339                 }
3340 
3341                 inode_dio_wait(inode);
3342 
3343                 /* fill in hole pointers in the expanding truncate case. */
3344                 if (attr->ia_size > inode->i_size) {
3345                         error = generic_cont_expand_simple(inode, attr->ia_size);
3346                         if (REISERFS_I(inode)->i_prealloc_count > 0) {
3347                                 int err;
3348                                 struct reiserfs_transaction_handle th;
3349                                 /* we're changing at most 2 bitmaps, inode + super */
3350                                 err = journal_begin(&th, inode->i_sb, 4);
3351                                 if (!err) {
3352                                         reiserfs_discard_prealloc(&th, inode);
3353                                         err = journal_end(&th);
3354                                 }
3355                                 if (err)
3356                                         error = err;
3357                         }
3358                         if (error) {
3359                                 reiserfs_write_unlock(inode->i_sb);
3360                                 goto out;
3361                         }
3362                         /*
3363                          * file size is changed, ctime and mtime are
3364                          * to be updated
3365                          */
3366                         attr->ia_valid |= (ATTR_MTIME | ATTR_CTIME);
3367                 }
3368         }
3369         reiserfs_write_unlock(inode->i_sb);
3370 
3371         if ((((attr->ia_valid & ATTR_UID) && (from_kuid(&init_user_ns, attr->ia_uid) & ~0xffff)) ||
3372              ((attr->ia_valid & ATTR_GID) && (from_kgid(&init_user_ns, attr->ia_gid) & ~0xffff))) &&
3373             (get_inode_sd_version(inode) == STAT_DATA_V1)) {
3374                 /* stat data of format v3.5 has 16 bit uid and gid */
3375                 error = -EINVAL;
3376                 goto out;
3377         }
3378 
3379         if ((ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
3380             (ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
3381                 struct reiserfs_transaction_handle th;
3382                 int jbegin_count =
3383                     2 *
3384                     (REISERFS_QUOTA_INIT_BLOCKS(inode->i_sb) +
3385                      REISERFS_QUOTA_DEL_BLOCKS(inode->i_sb)) +
3386                     2;
3387 
3388                 error = reiserfs_chown_xattrs(inode, attr);
3389 
3390                 if (error)
3391                         return error;
3392 
3393                 /*
3394                  * (user+group)*(old+new) structure - we count quota
3395                  * info and , inode write (sb, inode)
3396                  */
3397                 reiserfs_write_lock(inode->i_sb);
3398                 error = journal_begin(&th, inode->i_sb, jbegin_count);
3399                 reiserfs_write_unlock(inode->i_sb);
3400                 if (error)
3401                         goto out;
3402                 error = dquot_transfer(inode, attr);
3403                 reiserfs_write_lock(inode->i_sb);
3404                 if (error) {
3405                         journal_end(&th);
3406                         reiserfs_write_unlock(inode->i_sb);
3407                         goto out;
3408                 }
3409 
3410                 /*
3411                  * Update corresponding info in inode so that everything
3412                  * is in one transaction
3413                  */
3414                 if (attr->ia_valid & ATTR_UID)
3415                         inode->i_uid = attr->ia_uid;
3416                 if (attr->ia_valid & ATTR_GID)
3417                         inode->i_gid = attr->ia_gid;
3418                 mark_inode_dirty(inode);
3419                 error = journal_end(&th);
3420                 reiserfs_write_unlock(inode->i_sb);
3421                 if (error)
3422                         goto out;
3423         }
3424 
3425         if ((attr->ia_valid & ATTR_SIZE) &&
3426             attr->ia_size != i_size_read(inode)) {
3427                 error = inode_newsize_ok(inode, attr->ia_size);
3428                 if (!error) {
3429                         /*
3430                          * Could race against reiserfs_file_release
3431                          * if called from NFS, so take tailpack mutex.
3432                          */
3433                         mutex_lock(&REISERFS_I(inode)->tailpack);
3434                         truncate_setsize(inode, attr->ia_size);
3435                         reiserfs_truncate_file(inode, 1);
3436                         mutex_unlock(&REISERFS_I(inode)->tailpack);
3437                 }
3438         }
3439 
3440         if (!error) {
3441                 setattr_copy(inode, attr);
3442                 mark_inode_dirty(inode);
3443         }
3444 
3445         if (!error && reiserfs_posixacl(inode->i_sb)) {
3446                 if (attr->ia_valid & ATTR_MODE)
3447                         error = reiserfs_acl_chmod(inode);
3448         }
3449 
3450 out:
3451         return error;
3452 }
3453 
3454 const struct address_space_operations reiserfs_address_space_operations = {
3455         .writepage = reiserfs_writepage,
3456         .readpage = reiserfs_readpage,
3457         .readpages = reiserfs_readpages,
3458         .releasepage = reiserfs_releasepage,
3459         .invalidatepage = reiserfs_invalidatepage,
3460         .write_begin = reiserfs_write_begin,
3461         .write_end = reiserfs_write_end,
3462         .bmap = reiserfs_aop_bmap,
3463         .direct_IO = reiserfs_direct_IO,
3464         .set_page_dirty = reiserfs_set_page_dirty,
3465 };
3466 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp