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

TOMOYO Linux Cross Reference
Linux/fs/ext2/inode.c

Version: ~ [ linux-5.15-rc7 ] ~ [ linux-5.14.14 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.75 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.155 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.213 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.252 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.287 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.289 ] ~ [ 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 // SPDX-License-Identifier: GPL-2.0
  2 /*
  3  *  linux/fs/ext2/inode.c
  4  *
  5  * Copyright (C) 1992, 1993, 1994, 1995
  6  * Remy Card (card@masi.ibp.fr)
  7  * Laboratoire MASI - Institut Blaise Pascal
  8  * Universite Pierre et Marie Curie (Paris VI)
  9  *
 10  *  from
 11  *
 12  *  linux/fs/minix/inode.c
 13  *
 14  *  Copyright (C) 1991, 1992  Linus Torvalds
 15  *
 16  *  Goal-directed block allocation by Stephen Tweedie
 17  *      (sct@dcs.ed.ac.uk), 1993, 1998
 18  *  Big-endian to little-endian byte-swapping/bitmaps by
 19  *        David S. Miller (davem@caip.rutgers.edu), 1995
 20  *  64-bit file support on 64-bit platforms by Jakub Jelinek
 21  *      (jj@sunsite.ms.mff.cuni.cz)
 22  *
 23  *  Assorted race fixes, rewrite of ext2_get_block() by Al Viro, 2000
 24  */
 25 
 26 #include <linux/time.h>
 27 #include <linux/highuid.h>
 28 #include <linux/pagemap.h>
 29 #include <linux/dax.h>
 30 #include <linux/blkdev.h>
 31 #include <linux/quotaops.h>
 32 #include <linux/writeback.h>
 33 #include <linux/buffer_head.h>
 34 #include <linux/mpage.h>
 35 #include <linux/fiemap.h>
 36 #include <linux/iomap.h>
 37 #include <linux/namei.h>
 38 #include <linux/uio.h>
 39 #include "ext2.h"
 40 #include "acl.h"
 41 #include "xattr.h"
 42 
 43 static int __ext2_write_inode(struct inode *inode, int do_sync);
 44 
 45 /*
 46  * Test whether an inode is a fast symlink.
 47  */
 48 static inline int ext2_inode_is_fast_symlink(struct inode *inode)
 49 {
 50         int ea_blocks = EXT2_I(inode)->i_file_acl ?
 51                 (inode->i_sb->s_blocksize >> 9) : 0;
 52 
 53         return (S_ISLNK(inode->i_mode) &&
 54                 inode->i_blocks - ea_blocks == 0);
 55 }
 56 
 57 static void ext2_truncate_blocks(struct inode *inode, loff_t offset);
 58 
 59 static void ext2_write_failed(struct address_space *mapping, loff_t to)
 60 {
 61         struct inode *inode = mapping->host;
 62 
 63         if (to > inode->i_size) {
 64                 truncate_pagecache(inode, inode->i_size);
 65                 ext2_truncate_blocks(inode, inode->i_size);
 66         }
 67 }
 68 
 69 /*
 70  * Called at the last iput() if i_nlink is zero.
 71  */
 72 void ext2_evict_inode(struct inode * inode)
 73 {
 74         struct ext2_block_alloc_info *rsv;
 75         int want_delete = 0;
 76 
 77         if (!inode->i_nlink && !is_bad_inode(inode)) {
 78                 want_delete = 1;
 79                 dquot_initialize(inode);
 80         } else {
 81                 dquot_drop(inode);
 82         }
 83 
 84         truncate_inode_pages_final(&inode->i_data);
 85 
 86         if (want_delete) {
 87                 sb_start_intwrite(inode->i_sb);
 88                 /* set dtime */
 89                 EXT2_I(inode)->i_dtime  = ktime_get_real_seconds();
 90                 mark_inode_dirty(inode);
 91                 __ext2_write_inode(inode, inode_needs_sync(inode));
 92                 /* truncate to 0 */
 93                 inode->i_size = 0;
 94                 if (inode->i_blocks)
 95                         ext2_truncate_blocks(inode, 0);
 96                 ext2_xattr_delete_inode(inode);
 97         }
 98 
 99         invalidate_inode_buffers(inode);
100         clear_inode(inode);
101 
102         ext2_discard_reservation(inode);
103         rsv = EXT2_I(inode)->i_block_alloc_info;
104         EXT2_I(inode)->i_block_alloc_info = NULL;
105         if (unlikely(rsv))
106                 kfree(rsv);
107 
108         if (want_delete) {
109                 ext2_free_inode(inode);
110                 sb_end_intwrite(inode->i_sb);
111         }
112 }
113 
114 typedef struct {
115         __le32  *p;
116         __le32  key;
117         struct buffer_head *bh;
118 } Indirect;
119 
120 static inline void add_chain(Indirect *p, struct buffer_head *bh, __le32 *v)
121 {
122         p->key = *(p->p = v);
123         p->bh = bh;
124 }
125 
126 static inline int verify_chain(Indirect *from, Indirect *to)
127 {
128         while (from <= to && from->key == *from->p)
129                 from++;
130         return (from > to);
131 }
132 
133 /**
134  *      ext2_block_to_path - parse the block number into array of offsets
135  *      @inode: inode in question (we are only interested in its superblock)
136  *      @i_block: block number to be parsed
137  *      @offsets: array to store the offsets in
138  *      @boundary: set this non-zero if the referred-to block is likely to be
139  *             followed (on disk) by an indirect block.
140  *      To store the locations of file's data ext2 uses a data structure common
141  *      for UNIX filesystems - tree of pointers anchored in the inode, with
142  *      data blocks at leaves and indirect blocks in intermediate nodes.
143  *      This function translates the block number into path in that tree -
144  *      return value is the path length and @offsets[n] is the offset of
145  *      pointer to (n+1)th node in the nth one. If @block is out of range
146  *      (negative or too large) warning is printed and zero returned.
147  *
148  *      Note: function doesn't find node addresses, so no IO is needed. All
149  *      we need to know is the capacity of indirect blocks (taken from the
150  *      inode->i_sb).
151  */
152 
153 /*
154  * Portability note: the last comparison (check that we fit into triple
155  * indirect block) is spelled differently, because otherwise on an
156  * architecture with 32-bit longs and 8Kb pages we might get into trouble
157  * if our filesystem had 8Kb blocks. We might use long long, but that would
158  * kill us on x86. Oh, well, at least the sign propagation does not matter -
159  * i_block would have to be negative in the very beginning, so we would not
160  * get there at all.
161  */
162 
163 static int ext2_block_to_path(struct inode *inode,
164                         long i_block, int offsets[4], int *boundary)
165 {
166         int ptrs = EXT2_ADDR_PER_BLOCK(inode->i_sb);
167         int ptrs_bits = EXT2_ADDR_PER_BLOCK_BITS(inode->i_sb);
168         const long direct_blocks = EXT2_NDIR_BLOCKS,
169                 indirect_blocks = ptrs,
170                 double_blocks = (1 << (ptrs_bits * 2));
171         int n = 0;
172         int final = 0;
173 
174         if (i_block < 0) {
175                 ext2_msg(inode->i_sb, KERN_WARNING,
176                         "warning: %s: block < 0", __func__);
177         } else if (i_block < direct_blocks) {
178                 offsets[n++] = i_block;
179                 final = direct_blocks;
180         } else if ( (i_block -= direct_blocks) < indirect_blocks) {
181                 offsets[n++] = EXT2_IND_BLOCK;
182                 offsets[n++] = i_block;
183                 final = ptrs;
184         } else if ((i_block -= indirect_blocks) < double_blocks) {
185                 offsets[n++] = EXT2_DIND_BLOCK;
186                 offsets[n++] = i_block >> ptrs_bits;
187                 offsets[n++] = i_block & (ptrs - 1);
188                 final = ptrs;
189         } else if (((i_block -= double_blocks) >> (ptrs_bits * 2)) < ptrs) {
190                 offsets[n++] = EXT2_TIND_BLOCK;
191                 offsets[n++] = i_block >> (ptrs_bits * 2);
192                 offsets[n++] = (i_block >> ptrs_bits) & (ptrs - 1);
193                 offsets[n++] = i_block & (ptrs - 1);
194                 final = ptrs;
195         } else {
196                 ext2_msg(inode->i_sb, KERN_WARNING,
197                         "warning: %s: block is too big", __func__);
198         }
199         if (boundary)
200                 *boundary = final - 1 - (i_block & (ptrs - 1));
201 
202         return n;
203 }
204 
205 /**
206  *      ext2_get_branch - read the chain of indirect blocks leading to data
207  *      @inode: inode in question
208  *      @depth: depth of the chain (1 - direct pointer, etc.)
209  *      @offsets: offsets of pointers in inode/indirect blocks
210  *      @chain: place to store the result
211  *      @err: here we store the error value
212  *
213  *      Function fills the array of triples <key, p, bh> and returns %NULL
214  *      if everything went OK or the pointer to the last filled triple
215  *      (incomplete one) otherwise. Upon the return chain[i].key contains
216  *      the number of (i+1)-th block in the chain (as it is stored in memory,
217  *      i.e. little-endian 32-bit), chain[i].p contains the address of that
218  *      number (it points into struct inode for i==0 and into the bh->b_data
219  *      for i>0) and chain[i].bh points to the buffer_head of i-th indirect
220  *      block for i>0 and NULL for i==0. In other words, it holds the block
221  *      numbers of the chain, addresses they were taken from (and where we can
222  *      verify that chain did not change) and buffer_heads hosting these
223  *      numbers.
224  *
225  *      Function stops when it stumbles upon zero pointer (absent block)
226  *              (pointer to last triple returned, *@err == 0)
227  *      or when it gets an IO error reading an indirect block
228  *              (ditto, *@err == -EIO)
229  *      or when it notices that chain had been changed while it was reading
230  *              (ditto, *@err == -EAGAIN)
231  *      or when it reads all @depth-1 indirect blocks successfully and finds
232  *      the whole chain, all way to the data (returns %NULL, *err == 0).
233  */
234 static Indirect *ext2_get_branch(struct inode *inode,
235                                  int depth,
236                                  int *offsets,
237                                  Indirect chain[4],
238                                  int *err)
239 {
240         struct super_block *sb = inode->i_sb;
241         Indirect *p = chain;
242         struct buffer_head *bh;
243 
244         *err = 0;
245         /* i_data is not going away, no lock needed */
246         add_chain (chain, NULL, EXT2_I(inode)->i_data + *offsets);
247         if (!p->key)
248                 goto no_block;
249         while (--depth) {
250                 bh = sb_bread(sb, le32_to_cpu(p->key));
251                 if (!bh)
252                         goto failure;
253                 read_lock(&EXT2_I(inode)->i_meta_lock);
254                 if (!verify_chain(chain, p))
255                         goto changed;
256                 add_chain(++p, bh, (__le32*)bh->b_data + *++offsets);
257                 read_unlock(&EXT2_I(inode)->i_meta_lock);
258                 if (!p->key)
259                         goto no_block;
260         }
261         return NULL;
262 
263 changed:
264         read_unlock(&EXT2_I(inode)->i_meta_lock);
265         brelse(bh);
266         *err = -EAGAIN;
267         goto no_block;
268 failure:
269         *err = -EIO;
270 no_block:
271         return p;
272 }
273 
274 /**
275  *      ext2_find_near - find a place for allocation with sufficient locality
276  *      @inode: owner
277  *      @ind: descriptor of indirect block.
278  *
279  *      This function returns the preferred place for block allocation.
280  *      It is used when heuristic for sequential allocation fails.
281  *      Rules are:
282  *        + if there is a block to the left of our position - allocate near it.
283  *        + if pointer will live in indirect block - allocate near that block.
284  *        + if pointer will live in inode - allocate in the same cylinder group.
285  *
286  * In the latter case we colour the starting block by the callers PID to
287  * prevent it from clashing with concurrent allocations for a different inode
288  * in the same block group.   The PID is used here so that functionally related
289  * files will be close-by on-disk.
290  *
291  *      Caller must make sure that @ind is valid and will stay that way.
292  */
293 
294 static ext2_fsblk_t ext2_find_near(struct inode *inode, Indirect *ind)
295 {
296         struct ext2_inode_info *ei = EXT2_I(inode);
297         __le32 *start = ind->bh ? (__le32 *) ind->bh->b_data : ei->i_data;
298         __le32 *p;
299         ext2_fsblk_t bg_start;
300         ext2_fsblk_t colour;
301 
302         /* Try to find previous block */
303         for (p = ind->p - 1; p >= start; p--)
304                 if (*p)
305                         return le32_to_cpu(*p);
306 
307         /* No such thing, so let's try location of indirect block */
308         if (ind->bh)
309                 return ind->bh->b_blocknr;
310 
311         /*
312          * It is going to be referred from inode itself? OK, just put it into
313          * the same cylinder group then.
314          */
315         bg_start = ext2_group_first_block_no(inode->i_sb, ei->i_block_group);
316         colour = (current->pid % 16) *
317                         (EXT2_BLOCKS_PER_GROUP(inode->i_sb) / 16);
318         return bg_start + colour;
319 }
320 
321 /**
322  *      ext2_find_goal - find a preferred place for allocation.
323  *      @inode: owner
324  *      @block:  block we want
325  *      @partial: pointer to the last triple within a chain
326  *
327  *      Returns preferred place for a block (the goal).
328  */
329 
330 static inline ext2_fsblk_t ext2_find_goal(struct inode *inode, long block,
331                                           Indirect *partial)
332 {
333         struct ext2_block_alloc_info *block_i;
334 
335         block_i = EXT2_I(inode)->i_block_alloc_info;
336 
337         /*
338          * try the heuristic for sequential allocation,
339          * failing that at least try to get decent locality.
340          */
341         if (block_i && (block == block_i->last_alloc_logical_block + 1)
342                 && (block_i->last_alloc_physical_block != 0)) {
343                 return block_i->last_alloc_physical_block + 1;
344         }
345 
346         return ext2_find_near(inode, partial);
347 }
348 
349 /**
350  *      ext2_blks_to_allocate: Look up the block map and count the number
351  *      of direct blocks need to be allocated for the given branch.
352  *
353  *      @branch: chain of indirect blocks
354  *      @k: number of blocks need for indirect blocks
355  *      @blks: number of data blocks to be mapped.
356  *      @blocks_to_boundary:  the offset in the indirect block
357  *
358  *      return the total number of blocks to be allocate, including the
359  *      direct and indirect blocks.
360  */
361 static int
362 ext2_blks_to_allocate(Indirect * branch, int k, unsigned long blks,
363                 int blocks_to_boundary)
364 {
365         unsigned long count = 0;
366 
367         /*
368          * Simple case, [t,d]Indirect block(s) has not allocated yet
369          * then it's clear blocks on that path have not allocated
370          */
371         if (k > 0) {
372                 /* right now don't hanel cross boundary allocation */
373                 if (blks < blocks_to_boundary + 1)
374                         count += blks;
375                 else
376                         count += blocks_to_boundary + 1;
377                 return count;
378         }
379 
380         count++;
381         while (count < blks && count <= blocks_to_boundary
382                 && le32_to_cpu(*(branch[0].p + count)) == 0) {
383                 count++;
384         }
385         return count;
386 }
387 
388 /**
389  *      ext2_alloc_blocks: multiple allocate blocks needed for a branch
390  *      @indirect_blks: the number of blocks need to allocate for indirect
391  *                      blocks
392  *
393  *      @new_blocks: on return it will store the new block numbers for
394  *      the indirect blocks(if needed) and the first direct block,
395  *      @blks:  on return it will store the total number of allocated
396  *              direct blocks
397  */
398 static int ext2_alloc_blocks(struct inode *inode,
399                         ext2_fsblk_t goal, int indirect_blks, int blks,
400                         ext2_fsblk_t new_blocks[4], int *err)
401 {
402         int target, i;
403         unsigned long count = 0;
404         int index = 0;
405         ext2_fsblk_t current_block = 0;
406         int ret = 0;
407 
408         /*
409          * Here we try to allocate the requested multiple blocks at once,
410          * on a best-effort basis.
411          * To build a branch, we should allocate blocks for
412          * the indirect blocks(if not allocated yet), and at least
413          * the first direct block of this branch.  That's the
414          * minimum number of blocks need to allocate(required)
415          */
416         target = blks + indirect_blks;
417 
418         while (1) {
419                 count = target;
420                 /* allocating blocks for indirect blocks and direct blocks */
421                 current_block = ext2_new_blocks(inode,goal,&count,err);
422                 if (*err)
423                         goto failed_out;
424 
425                 target -= count;
426                 /* allocate blocks for indirect blocks */
427                 while (index < indirect_blks && count) {
428                         new_blocks[index++] = current_block++;
429                         count--;
430                 }
431 
432                 if (count > 0)
433                         break;
434         }
435 
436         /* save the new block number for the first direct block */
437         new_blocks[index] = current_block;
438 
439         /* total number of blocks allocated for direct blocks */
440         ret = count;
441         *err = 0;
442         return ret;
443 failed_out:
444         for (i = 0; i <index; i++)
445                 ext2_free_blocks(inode, new_blocks[i], 1);
446         if (index)
447                 mark_inode_dirty(inode);
448         return ret;
449 }
450 
451 /**
452  *      ext2_alloc_branch - allocate and set up a chain of blocks.
453  *      @inode: owner
454  *      @indirect_blks: depth of the chain (number of blocks to allocate)
455  *      @blks: number of allocated direct blocks
456  *      @goal: preferred place for allocation
457  *      @offsets: offsets (in the blocks) to store the pointers to next.
458  *      @branch: place to store the chain in.
459  *
460  *      This function allocates @num blocks, zeroes out all but the last one,
461  *      links them into chain and (if we are synchronous) writes them to disk.
462  *      In other words, it prepares a branch that can be spliced onto the
463  *      inode. It stores the information about that chain in the branch[], in
464  *      the same format as ext2_get_branch() would do. We are calling it after
465  *      we had read the existing part of chain and partial points to the last
466  *      triple of that (one with zero ->key). Upon the exit we have the same
467  *      picture as after the successful ext2_get_block(), except that in one
468  *      place chain is disconnected - *branch->p is still zero (we did not
469  *      set the last link), but branch->key contains the number that should
470  *      be placed into *branch->p to fill that gap.
471  *
472  *      If allocation fails we free all blocks we've allocated (and forget
473  *      their buffer_heads) and return the error value the from failed
474  *      ext2_alloc_block() (normally -ENOSPC). Otherwise we set the chain
475  *      as described above and return 0.
476  */
477 
478 static int ext2_alloc_branch(struct inode *inode,
479                         int indirect_blks, int *blks, ext2_fsblk_t goal,
480                         int *offsets, Indirect *branch)
481 {
482         int blocksize = inode->i_sb->s_blocksize;
483         int i, n = 0;
484         int err = 0;
485         struct buffer_head *bh;
486         int num;
487         ext2_fsblk_t new_blocks[4];
488         ext2_fsblk_t current_block;
489 
490         num = ext2_alloc_blocks(inode, goal, indirect_blks,
491                                 *blks, new_blocks, &err);
492         if (err)
493                 return err;
494 
495         branch[0].key = cpu_to_le32(new_blocks[0]);
496         /*
497          * metadata blocks and data blocks are allocated.
498          */
499         for (n = 1; n <= indirect_blks;  n++) {
500                 /*
501                  * Get buffer_head for parent block, zero it out
502                  * and set the pointer to new one, then send
503                  * parent to disk.
504                  */
505                 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
506                 if (unlikely(!bh)) {
507                         err = -ENOMEM;
508                         goto failed;
509                 }
510                 branch[n].bh = bh;
511                 lock_buffer(bh);
512                 memset(bh->b_data, 0, blocksize);
513                 branch[n].p = (__le32 *) bh->b_data + offsets[n];
514                 branch[n].key = cpu_to_le32(new_blocks[n]);
515                 *branch[n].p = branch[n].key;
516                 if ( n == indirect_blks) {
517                         current_block = new_blocks[n];
518                         /*
519                          * End of chain, update the last new metablock of
520                          * the chain to point to the new allocated
521                          * data blocks numbers
522                          */
523                         for (i=1; i < num; i++)
524                                 *(branch[n].p + i) = cpu_to_le32(++current_block);
525                 }
526                 set_buffer_uptodate(bh);
527                 unlock_buffer(bh);
528                 mark_buffer_dirty_inode(bh, inode);
529                 /* We used to sync bh here if IS_SYNC(inode).
530                  * But we now rely upon generic_write_sync()
531                  * and b_inode_buffers.  But not for directories.
532                  */
533                 if (S_ISDIR(inode->i_mode) && IS_DIRSYNC(inode))
534                         sync_dirty_buffer(bh);
535         }
536         *blks = num;
537         return err;
538 
539 failed:
540         for (i = 1; i < n; i++)
541                 bforget(branch[i].bh);
542         for (i = 0; i < indirect_blks; i++)
543                 ext2_free_blocks(inode, new_blocks[i], 1);
544         ext2_free_blocks(inode, new_blocks[i], num);
545         return err;
546 }
547 
548 /**
549  * ext2_splice_branch - splice the allocated branch onto inode.
550  * @inode: owner
551  * @block: (logical) number of block we are adding
552  * @where: location of missing link
553  * @num:   number of indirect blocks we are adding
554  * @blks:  number of direct blocks we are adding
555  *
556  * This function fills the missing link and does all housekeeping needed in
557  * inode (->i_blocks, etc.). In case of success we end up with the full
558  * chain to new block and return 0.
559  */
560 static void ext2_splice_branch(struct inode *inode,
561                         long block, Indirect *where, int num, int blks)
562 {
563         int i;
564         struct ext2_block_alloc_info *block_i;
565         ext2_fsblk_t current_block;
566 
567         block_i = EXT2_I(inode)->i_block_alloc_info;
568 
569         /* XXX LOCKING probably should have i_meta_lock ?*/
570         /* That's it */
571 
572         *where->p = where->key;
573 
574         /*
575          * Update the host buffer_head or inode to point to more just allocated
576          * direct blocks blocks
577          */
578         if (num == 0 && blks > 1) {
579                 current_block = le32_to_cpu(where->key) + 1;
580                 for (i = 1; i < blks; i++)
581                         *(where->p + i ) = cpu_to_le32(current_block++);
582         }
583 
584         /*
585          * update the most recently allocated logical & physical block
586          * in i_block_alloc_info, to assist find the proper goal block for next
587          * allocation
588          */
589         if (block_i) {
590                 block_i->last_alloc_logical_block = block + blks - 1;
591                 block_i->last_alloc_physical_block =
592                                 le32_to_cpu(where[num].key) + blks - 1;
593         }
594 
595         /* We are done with atomic stuff, now do the rest of housekeeping */
596 
597         /* had we spliced it onto indirect block? */
598         if (where->bh)
599                 mark_buffer_dirty_inode(where->bh, inode);
600 
601         inode->i_ctime = current_time(inode);
602         mark_inode_dirty(inode);
603 }
604 
605 /*
606  * Allocation strategy is simple: if we have to allocate something, we will
607  * have to go the whole way to leaf. So let's do it before attaching anything
608  * to tree, set linkage between the newborn blocks, write them if sync is
609  * required, recheck the path, free and repeat if check fails, otherwise
610  * set the last missing link (that will protect us from any truncate-generated
611  * removals - all blocks on the path are immune now) and possibly force the
612  * write on the parent block.
613  * That has a nice additional property: no special recovery from the failed
614  * allocations is needed - we simply release blocks and do not touch anything
615  * reachable from inode.
616  *
617  * `handle' can be NULL if create == 0.
618  *
619  * return > 0, # of blocks mapped or allocated.
620  * return = 0, if plain lookup failed.
621  * return < 0, error case.
622  */
623 static int ext2_get_blocks(struct inode *inode,
624                            sector_t iblock, unsigned long maxblocks,
625                            u32 *bno, bool *new, bool *boundary,
626                            int create)
627 {
628         int err;
629         int offsets[4];
630         Indirect chain[4];
631         Indirect *partial;
632         ext2_fsblk_t goal;
633         int indirect_blks;
634         int blocks_to_boundary = 0;
635         int depth;
636         struct ext2_inode_info *ei = EXT2_I(inode);
637         int count = 0;
638         ext2_fsblk_t first_block = 0;
639 
640         BUG_ON(maxblocks == 0);
641 
642         depth = ext2_block_to_path(inode,iblock,offsets,&blocks_to_boundary);
643 
644         if (depth == 0)
645                 return -EIO;
646 
647         partial = ext2_get_branch(inode, depth, offsets, chain, &err);
648         /* Simplest case - block found, no allocation needed */
649         if (!partial) {
650                 first_block = le32_to_cpu(chain[depth - 1].key);
651                 count++;
652                 /*map more blocks*/
653                 while (count < maxblocks && count <= blocks_to_boundary) {
654                         ext2_fsblk_t blk;
655 
656                         if (!verify_chain(chain, chain + depth - 1)) {
657                                 /*
658                                  * Indirect block might be removed by
659                                  * truncate while we were reading it.
660                                  * Handling of that case: forget what we've
661                                  * got now, go to reread.
662                                  */
663                                 err = -EAGAIN;
664                                 count = 0;
665                                 partial = chain + depth - 1;
666                                 break;
667                         }
668                         blk = le32_to_cpu(*(chain[depth-1].p + count));
669                         if (blk == first_block + count)
670                                 count++;
671                         else
672                                 break;
673                 }
674                 if (err != -EAGAIN)
675                         goto got_it;
676         }
677 
678         /* Next simple case - plain lookup or failed read of indirect block */
679         if (!create || err == -EIO)
680                 goto cleanup;
681 
682         mutex_lock(&ei->truncate_mutex);
683         /*
684          * If the indirect block is missing while we are reading
685          * the chain(ext2_get_branch() returns -EAGAIN err), or
686          * if the chain has been changed after we grab the semaphore,
687          * (either because another process truncated this branch, or
688          * another get_block allocated this branch) re-grab the chain to see if
689          * the request block has been allocated or not.
690          *
691          * Since we already block the truncate/other get_block
692          * at this point, we will have the current copy of the chain when we
693          * splice the branch into the tree.
694          */
695         if (err == -EAGAIN || !verify_chain(chain, partial)) {
696                 while (partial > chain) {
697                         brelse(partial->bh);
698                         partial--;
699                 }
700                 partial = ext2_get_branch(inode, depth, offsets, chain, &err);
701                 if (!partial) {
702                         count++;
703                         mutex_unlock(&ei->truncate_mutex);
704                         goto got_it;
705                 }
706 
707                 if (err) {
708                         mutex_unlock(&ei->truncate_mutex);
709                         goto cleanup;
710                 }
711         }
712 
713         /*
714          * Okay, we need to do block allocation.  Lazily initialize the block
715          * allocation info here if necessary
716         */
717         if (S_ISREG(inode->i_mode) && (!ei->i_block_alloc_info))
718                 ext2_init_block_alloc_info(inode);
719 
720         goal = ext2_find_goal(inode, iblock, partial);
721 
722         /* the number of blocks need to allocate for [d,t]indirect blocks */
723         indirect_blks = (chain + depth) - partial - 1;
724         /*
725          * Next look up the indirect map to count the total number of
726          * direct blocks to allocate for this branch.
727          */
728         count = ext2_blks_to_allocate(partial, indirect_blks,
729                                         maxblocks, blocks_to_boundary);
730         /*
731          * XXX ???? Block out ext2_truncate while we alter the tree
732          */
733         err = ext2_alloc_branch(inode, indirect_blks, &count, goal,
734                                 offsets + (partial - chain), partial);
735 
736         if (err) {
737                 mutex_unlock(&ei->truncate_mutex);
738                 goto cleanup;
739         }
740 
741         if (IS_DAX(inode)) {
742                 /*
743                  * We must unmap blocks before zeroing so that writeback cannot
744                  * overwrite zeros with stale data from block device page cache.
745                  */
746                 clean_bdev_aliases(inode->i_sb->s_bdev,
747                                    le32_to_cpu(chain[depth-1].key),
748                                    count);
749                 /*
750                  * block must be initialised before we put it in the tree
751                  * so that it's not found by another thread before it's
752                  * initialised
753                  */
754                 err = sb_issue_zeroout(inode->i_sb,
755                                 le32_to_cpu(chain[depth-1].key), count,
756                                 GFP_NOFS);
757                 if (err) {
758                         mutex_unlock(&ei->truncate_mutex);
759                         goto cleanup;
760                 }
761         }
762         *new = true;
763 
764         ext2_splice_branch(inode, iblock, partial, indirect_blks, count);
765         mutex_unlock(&ei->truncate_mutex);
766 got_it:
767         if (count > blocks_to_boundary)
768                 *boundary = true;
769         err = count;
770         /* Clean up and exit */
771         partial = chain + depth - 1;    /* the whole chain */
772 cleanup:
773         while (partial > chain) {
774                 brelse(partial->bh);
775                 partial--;
776         }
777         if (err > 0)
778                 *bno = le32_to_cpu(chain[depth-1].key);
779         return err;
780 }
781 
782 int ext2_get_block(struct inode *inode, sector_t iblock,
783                 struct buffer_head *bh_result, int create)
784 {
785         unsigned max_blocks = bh_result->b_size >> inode->i_blkbits;
786         bool new = false, boundary = false;
787         u32 bno;
788         int ret;
789 
790         ret = ext2_get_blocks(inode, iblock, max_blocks, &bno, &new, &boundary,
791                         create);
792         if (ret <= 0)
793                 return ret;
794 
795         map_bh(bh_result, inode->i_sb, bno);
796         bh_result->b_size = (ret << inode->i_blkbits);
797         if (new)
798                 set_buffer_new(bh_result);
799         if (boundary)
800                 set_buffer_boundary(bh_result);
801         return 0;
802 
803 }
804 
805 #ifdef CONFIG_FS_DAX
806 static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
807                 unsigned flags, struct iomap *iomap, struct iomap *srcmap)
808 {
809         unsigned int blkbits = inode->i_blkbits;
810         unsigned long first_block = offset >> blkbits;
811         unsigned long max_blocks = (length + (1 << blkbits) - 1) >> blkbits;
812         struct ext2_sb_info *sbi = EXT2_SB(inode->i_sb);
813         bool new = false, boundary = false;
814         u32 bno;
815         int ret;
816 
817         ret = ext2_get_blocks(inode, first_block, max_blocks,
818                         &bno, &new, &boundary, flags & IOMAP_WRITE);
819         if (ret < 0)
820                 return ret;
821 
822         iomap->flags = 0;
823         iomap->bdev = inode->i_sb->s_bdev;
824         iomap->offset = (u64)first_block << blkbits;
825         iomap->dax_dev = sbi->s_daxdev;
826 
827         if (ret == 0) {
828                 iomap->type = IOMAP_HOLE;
829                 iomap->addr = IOMAP_NULL_ADDR;
830                 iomap->length = 1 << blkbits;
831         } else {
832                 iomap->type = IOMAP_MAPPED;
833                 iomap->addr = (u64)bno << blkbits;
834                 iomap->length = (u64)ret << blkbits;
835                 iomap->flags |= IOMAP_F_MERGED;
836         }
837 
838         if (new)
839                 iomap->flags |= IOMAP_F_NEW;
840         return 0;
841 }
842 
843 static int
844 ext2_iomap_end(struct inode *inode, loff_t offset, loff_t length,
845                 ssize_t written, unsigned flags, struct iomap *iomap)
846 {
847         if (iomap->type == IOMAP_MAPPED &&
848             written < length &&
849             (flags & IOMAP_WRITE))
850                 ext2_write_failed(inode->i_mapping, offset + length);
851         return 0;
852 }
853 
854 const struct iomap_ops ext2_iomap_ops = {
855         .iomap_begin            = ext2_iomap_begin,
856         .iomap_end              = ext2_iomap_end,
857 };
858 #else
859 /* Define empty ops for !CONFIG_FS_DAX case to avoid ugly ifdefs */
860 const struct iomap_ops ext2_iomap_ops;
861 #endif /* CONFIG_FS_DAX */
862 
863 int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
864                 u64 start, u64 len)
865 {
866         return generic_block_fiemap(inode, fieinfo, start, len,
867                                     ext2_get_block);
868 }
869 
870 static int ext2_writepage(struct page *page, struct writeback_control *wbc)
871 {
872         return block_write_full_page(page, ext2_get_block, wbc);
873 }
874 
875 static int ext2_readpage(struct file *file, struct page *page)
876 {
877         return mpage_readpage(page, ext2_get_block);
878 }
879 
880 static int
881 ext2_readpages(struct file *file, struct address_space *mapping,
882                 struct list_head *pages, unsigned nr_pages)
883 {
884         return mpage_readpages(mapping, pages, nr_pages, ext2_get_block);
885 }
886 
887 static int
888 ext2_write_begin(struct file *file, struct address_space *mapping,
889                 loff_t pos, unsigned len, unsigned flags,
890                 struct page **pagep, void **fsdata)
891 {
892         int ret;
893 
894         ret = block_write_begin(mapping, pos, len, flags, pagep,
895                                 ext2_get_block);
896         if (ret < 0)
897                 ext2_write_failed(mapping, pos + len);
898         return ret;
899 }
900 
901 static int ext2_write_end(struct file *file, struct address_space *mapping,
902                         loff_t pos, unsigned len, unsigned copied,
903                         struct page *page, void *fsdata)
904 {
905         int ret;
906 
907         ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata);
908         if (ret < len)
909                 ext2_write_failed(mapping, pos + len);
910         return ret;
911 }
912 
913 static int
914 ext2_nobh_write_begin(struct file *file, struct address_space *mapping,
915                 loff_t pos, unsigned len, unsigned flags,
916                 struct page **pagep, void **fsdata)
917 {
918         int ret;
919 
920         ret = nobh_write_begin(mapping, pos, len, flags, pagep, fsdata,
921                                ext2_get_block);
922         if (ret < 0)
923                 ext2_write_failed(mapping, pos + len);
924         return ret;
925 }
926 
927 static int ext2_nobh_writepage(struct page *page,
928                         struct writeback_control *wbc)
929 {
930         return nobh_writepage(page, ext2_get_block, wbc);
931 }
932 
933 static sector_t ext2_bmap(struct address_space *mapping, sector_t block)
934 {
935         return generic_block_bmap(mapping,block,ext2_get_block);
936 }
937 
938 static ssize_t
939 ext2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
940 {
941         struct file *file = iocb->ki_filp;
942         struct address_space *mapping = file->f_mapping;
943         struct inode *inode = mapping->host;
944         size_t count = iov_iter_count(iter);
945         loff_t offset = iocb->ki_pos;
946         ssize_t ret;
947 
948         ret = blockdev_direct_IO(iocb, inode, iter, ext2_get_block);
949         if (ret < 0 && iov_iter_rw(iter) == WRITE)
950                 ext2_write_failed(mapping, offset + count);
951         return ret;
952 }
953 
954 static int
955 ext2_writepages(struct address_space *mapping, struct writeback_control *wbc)
956 {
957         return mpage_writepages(mapping, wbc, ext2_get_block);
958 }
959 
960 static int
961 ext2_dax_writepages(struct address_space *mapping, struct writeback_control *wbc)
962 {
963         struct ext2_sb_info *sbi = EXT2_SB(mapping->host->i_sb);
964 
965         return dax_writeback_mapping_range(mapping, sbi->s_daxdev, wbc);
966 }
967 
968 const struct address_space_operations ext2_aops = {
969         .readpage               = ext2_readpage,
970         .readpages              = ext2_readpages,
971         .writepage              = ext2_writepage,
972         .write_begin            = ext2_write_begin,
973         .write_end              = ext2_write_end,
974         .bmap                   = ext2_bmap,
975         .direct_IO              = ext2_direct_IO,
976         .writepages             = ext2_writepages,
977         .migratepage            = buffer_migrate_page,
978         .is_partially_uptodate  = block_is_partially_uptodate,
979         .error_remove_page      = generic_error_remove_page,
980 };
981 
982 const struct address_space_operations ext2_nobh_aops = {
983         .readpage               = ext2_readpage,
984         .readpages              = ext2_readpages,
985         .writepage              = ext2_nobh_writepage,
986         .write_begin            = ext2_nobh_write_begin,
987         .write_end              = nobh_write_end,
988         .bmap                   = ext2_bmap,
989         .direct_IO              = ext2_direct_IO,
990         .writepages             = ext2_writepages,
991         .migratepage            = buffer_migrate_page,
992         .error_remove_page      = generic_error_remove_page,
993 };
994 
995 static const struct address_space_operations ext2_dax_aops = {
996         .writepages             = ext2_dax_writepages,
997         .direct_IO              = noop_direct_IO,
998         .set_page_dirty         = noop_set_page_dirty,
999         .invalidatepage         = noop_invalidatepage,
1000 };
1001 
1002 /*
1003  * Probably it should be a library function... search for first non-zero word
1004  * or memcmp with zero_page, whatever is better for particular architecture.
1005  * Linus?
1006  */
1007 static inline int all_zeroes(__le32 *p, __le32 *q)
1008 {
1009         while (p < q)
1010                 if (*p++)
1011                         return 0;
1012         return 1;
1013 }
1014 
1015 /**
1016  *      ext2_find_shared - find the indirect blocks for partial truncation.
1017  *      @inode:   inode in question
1018  *      @depth:   depth of the affected branch
1019  *      @offsets: offsets of pointers in that branch (see ext2_block_to_path)
1020  *      @chain:   place to store the pointers to partial indirect blocks
1021  *      @top:     place to the (detached) top of branch
1022  *
1023  *      This is a helper function used by ext2_truncate().
1024  *
1025  *      When we do truncate() we may have to clean the ends of several indirect
1026  *      blocks but leave the blocks themselves alive. Block is partially
1027  *      truncated if some data below the new i_size is referred from it (and
1028  *      it is on the path to the first completely truncated data block, indeed).
1029  *      We have to free the top of that path along with everything to the right
1030  *      of the path. Since no allocation past the truncation point is possible
1031  *      until ext2_truncate() finishes, we may safely do the latter, but top
1032  *      of branch may require special attention - pageout below the truncation
1033  *      point might try to populate it.
1034  *
1035  *      We atomically detach the top of branch from the tree, store the block
1036  *      number of its root in *@top, pointers to buffer_heads of partially
1037  *      truncated blocks - in @chain[].bh and pointers to their last elements
1038  *      that should not be removed - in @chain[].p. Return value is the pointer
1039  *      to last filled element of @chain.
1040  *
1041  *      The work left to caller to do the actual freeing of subtrees:
1042  *              a) free the subtree starting from *@top
1043  *              b) free the subtrees whose roots are stored in
1044  *                      (@chain[i].p+1 .. end of @chain[i].bh->b_data)
1045  *              c) free the subtrees growing from the inode past the @chain[0].p
1046  *                      (no partially truncated stuff there).
1047  */
1048 
1049 static Indirect *ext2_find_shared(struct inode *inode,
1050                                 int depth,
1051                                 int offsets[4],
1052                                 Indirect chain[4],
1053                                 __le32 *top)
1054 {
1055         Indirect *partial, *p;
1056         int k, err;
1057 
1058         *top = 0;
1059         for (k = depth; k > 1 && !offsets[k-1]; k--)
1060                 ;
1061         partial = ext2_get_branch(inode, k, offsets, chain, &err);
1062         if (!partial)
1063                 partial = chain + k-1;
1064         /*
1065          * If the branch acquired continuation since we've looked at it -
1066          * fine, it should all survive and (new) top doesn't belong to us.
1067          */
1068         write_lock(&EXT2_I(inode)->i_meta_lock);
1069         if (!partial->key && *partial->p) {
1070                 write_unlock(&EXT2_I(inode)->i_meta_lock);
1071                 goto no_top;
1072         }
1073         for (p=partial; p>chain && all_zeroes((__le32*)p->bh->b_data,p->p); p--)
1074                 ;
1075         /*
1076          * OK, we've found the last block that must survive. The rest of our
1077          * branch should be detached before unlocking. However, if that rest
1078          * of branch is all ours and does not grow immediately from the inode
1079          * it's easier to cheat and just decrement partial->p.
1080          */
1081         if (p == chain + k - 1 && p > chain) {
1082                 p->p--;
1083         } else {
1084                 *top = *p->p;
1085                 *p->p = 0;
1086         }
1087         write_unlock(&EXT2_I(inode)->i_meta_lock);
1088 
1089         while(partial > p)
1090         {
1091                 brelse(partial->bh);
1092                 partial--;
1093         }
1094 no_top:
1095         return partial;
1096 }
1097 
1098 /**
1099  *      ext2_free_data - free a list of data blocks
1100  *      @inode: inode we are dealing with
1101  *      @p:     array of block numbers
1102  *      @q:     points immediately past the end of array
1103  *
1104  *      We are freeing all blocks referred from that array (numbers are
1105  *      stored as little-endian 32-bit) and updating @inode->i_blocks
1106  *      appropriately.
1107  */
1108 static inline void ext2_free_data(struct inode *inode, __le32 *p, __le32 *q)
1109 {
1110         unsigned long block_to_free = 0, count = 0;
1111         unsigned long nr;
1112 
1113         for ( ; p < q ; p++) {
1114                 nr = le32_to_cpu(*p);
1115                 if (nr) {
1116                         *p = 0;
1117                         /* accumulate blocks to free if they're contiguous */
1118                         if (count == 0)
1119                                 goto free_this;
1120                         else if (block_to_free == nr - count)
1121                                 count++;
1122                         else {
1123                                 ext2_free_blocks (inode, block_to_free, count);
1124                                 mark_inode_dirty(inode);
1125                         free_this:
1126                                 block_to_free = nr;
1127                                 count = 1;
1128                         }
1129                 }
1130         }
1131         if (count > 0) {
1132                 ext2_free_blocks (inode, block_to_free, count);
1133                 mark_inode_dirty(inode);
1134         }
1135 }
1136 
1137 /**
1138  *      ext2_free_branches - free an array of branches
1139  *      @inode: inode we are dealing with
1140  *      @p:     array of block numbers
1141  *      @q:     pointer immediately past the end of array
1142  *      @depth: depth of the branches to free
1143  *
1144  *      We are freeing all blocks referred from these branches (numbers are
1145  *      stored as little-endian 32-bit) and updating @inode->i_blocks
1146  *      appropriately.
1147  */
1148 static void ext2_free_branches(struct inode *inode, __le32 *p, __le32 *q, int depth)
1149 {
1150         struct buffer_head * bh;
1151         unsigned long nr;
1152 
1153         if (depth--) {
1154                 int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1155                 for ( ; p < q ; p++) {
1156                         nr = le32_to_cpu(*p);
1157                         if (!nr)
1158                                 continue;
1159                         *p = 0;
1160                         bh = sb_bread(inode->i_sb, nr);
1161                         /*
1162                          * A read failure? Report error and clear slot
1163                          * (should be rare).
1164                          */ 
1165                         if (!bh) {
1166                                 ext2_error(inode->i_sb, "ext2_free_branches",
1167                                         "Read failure, inode=%ld, block=%ld",
1168                                         inode->i_ino, nr);
1169                                 continue;
1170                         }
1171                         ext2_free_branches(inode,
1172                                            (__le32*)bh->b_data,
1173                                            (__le32*)bh->b_data + addr_per_block,
1174                                            depth);
1175                         bforget(bh);
1176                         ext2_free_blocks(inode, nr, 1);
1177                         mark_inode_dirty(inode);
1178                 }
1179         } else
1180                 ext2_free_data(inode, p, q);
1181 }
1182 
1183 /* dax_sem must be held when calling this function */
1184 static void __ext2_truncate_blocks(struct inode *inode, loff_t offset)
1185 {
1186         __le32 *i_data = EXT2_I(inode)->i_data;
1187         struct ext2_inode_info *ei = EXT2_I(inode);
1188         int addr_per_block = EXT2_ADDR_PER_BLOCK(inode->i_sb);
1189         int offsets[4];
1190         Indirect chain[4];
1191         Indirect *partial;
1192         __le32 nr = 0;
1193         int n;
1194         long iblock;
1195         unsigned blocksize;
1196         blocksize = inode->i_sb->s_blocksize;
1197         iblock = (offset + blocksize-1) >> EXT2_BLOCK_SIZE_BITS(inode->i_sb);
1198 
1199 #ifdef CONFIG_FS_DAX
1200         WARN_ON(!rwsem_is_locked(&ei->dax_sem));
1201 #endif
1202 
1203         n = ext2_block_to_path(inode, iblock, offsets, NULL);
1204         if (n == 0)
1205                 return;
1206 
1207         /*
1208          * From here we block out all ext2_get_block() callers who want to
1209          * modify the block allocation tree.
1210          */
1211         mutex_lock(&ei->truncate_mutex);
1212 
1213         if (n == 1) {
1214                 ext2_free_data(inode, i_data+offsets[0],
1215                                         i_data + EXT2_NDIR_BLOCKS);
1216                 goto do_indirects;
1217         }
1218 
1219         partial = ext2_find_shared(inode, n, offsets, chain, &nr);
1220         /* Kill the top of shared branch (already detached) */
1221         if (nr) {
1222                 if (partial == chain)
1223                         mark_inode_dirty(inode);
1224                 else
1225                         mark_buffer_dirty_inode(partial->bh, inode);
1226                 ext2_free_branches(inode, &nr, &nr+1, (chain+n-1) - partial);
1227         }
1228         /* Clear the ends of indirect blocks on the shared branch */
1229         while (partial > chain) {
1230                 ext2_free_branches(inode,
1231                                    partial->p + 1,
1232                                    (__le32*)partial->bh->b_data+addr_per_block,
1233                                    (chain+n-1) - partial);
1234                 mark_buffer_dirty_inode(partial->bh, inode);
1235                 brelse (partial->bh);
1236                 partial--;
1237         }
1238 do_indirects:
1239         /* Kill the remaining (whole) subtrees */
1240         switch (offsets[0]) {
1241                 default:
1242                         nr = i_data[EXT2_IND_BLOCK];
1243                         if (nr) {
1244                                 i_data[EXT2_IND_BLOCK] = 0;
1245                                 mark_inode_dirty(inode);
1246                                 ext2_free_branches(inode, &nr, &nr+1, 1);
1247                         }
1248                         /* fall through */
1249                 case EXT2_IND_BLOCK:
1250                         nr = i_data[EXT2_DIND_BLOCK];
1251                         if (nr) {
1252                                 i_data[EXT2_DIND_BLOCK] = 0;
1253                                 mark_inode_dirty(inode);
1254                                 ext2_free_branches(inode, &nr, &nr+1, 2);
1255                         }
1256                         /* fall through */
1257                 case EXT2_DIND_BLOCK:
1258                         nr = i_data[EXT2_TIND_BLOCK];
1259                         if (nr) {
1260                                 i_data[EXT2_TIND_BLOCK] = 0;
1261                                 mark_inode_dirty(inode);
1262                                 ext2_free_branches(inode, &nr, &nr+1, 3);
1263                         }
1264                 case EXT2_TIND_BLOCK:
1265                         ;
1266         }
1267 
1268         ext2_discard_reservation(inode);
1269 
1270         mutex_unlock(&ei->truncate_mutex);
1271 }
1272 
1273 static void ext2_truncate_blocks(struct inode *inode, loff_t offset)
1274 {
1275         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1276             S_ISLNK(inode->i_mode)))
1277                 return;
1278         if (ext2_inode_is_fast_symlink(inode))
1279                 return;
1280 
1281         dax_sem_down_write(EXT2_I(inode));
1282         __ext2_truncate_blocks(inode, offset);
1283         dax_sem_up_write(EXT2_I(inode));
1284 }
1285 
1286 static int ext2_setsize(struct inode *inode, loff_t newsize)
1287 {
1288         int error;
1289 
1290         if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1291             S_ISLNK(inode->i_mode)))
1292                 return -EINVAL;
1293         if (ext2_inode_is_fast_symlink(inode))
1294                 return -EINVAL;
1295         if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1296                 return -EPERM;
1297 
1298         inode_dio_wait(inode);
1299 
1300         if (IS_DAX(inode)) {
1301                 error = iomap_zero_range(inode, newsize,
1302                                          PAGE_ALIGN(newsize) - newsize, NULL,
1303                                          &ext2_iomap_ops);
1304         } else if (test_opt(inode->i_sb, NOBH))
1305                 error = nobh_truncate_page(inode->i_mapping,
1306                                 newsize, ext2_get_block);
1307         else
1308                 error = block_truncate_page(inode->i_mapping,
1309                                 newsize, ext2_get_block);
1310         if (error)
1311                 return error;
1312 
1313         dax_sem_down_write(EXT2_I(inode));
1314         truncate_setsize(inode, newsize);
1315         __ext2_truncate_blocks(inode, newsize);
1316         dax_sem_up_write(EXT2_I(inode));
1317 
1318         inode->i_mtime = inode->i_ctime = current_time(inode);
1319         if (inode_needs_sync(inode)) {
1320                 sync_mapping_buffers(inode->i_mapping);
1321                 sync_inode_metadata(inode, 1);
1322         } else {
1323                 mark_inode_dirty(inode);
1324         }
1325 
1326         return 0;
1327 }
1328 
1329 static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
1330                                         struct buffer_head **p)
1331 {
1332         struct buffer_head * bh;
1333         unsigned long block_group;
1334         unsigned long block;
1335         unsigned long offset;
1336         struct ext2_group_desc * gdp;
1337 
1338         *p = NULL;
1339         if ((ino != EXT2_ROOT_INO && ino < EXT2_FIRST_INO(sb)) ||
1340             ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
1341                 goto Einval;
1342 
1343         block_group = (ino - 1) / EXT2_INODES_PER_GROUP(sb);
1344         gdp = ext2_get_group_desc(sb, block_group, NULL);
1345         if (!gdp)
1346                 goto Egdp;
1347         /*
1348          * Figure out the offset within the block group inode table
1349          */
1350         offset = ((ino - 1) % EXT2_INODES_PER_GROUP(sb)) * EXT2_INODE_SIZE(sb);
1351         block = le32_to_cpu(gdp->bg_inode_table) +
1352                 (offset >> EXT2_BLOCK_SIZE_BITS(sb));
1353         if (!(bh = sb_bread(sb, block)))
1354                 goto Eio;
1355 
1356         *p = bh;
1357         offset &= (EXT2_BLOCK_SIZE(sb) - 1);
1358         return (struct ext2_inode *) (bh->b_data + offset);
1359 
1360 Einval:
1361         ext2_error(sb, "ext2_get_inode", "bad inode number: %lu",
1362                    (unsigned long) ino);
1363         return ERR_PTR(-EINVAL);
1364 Eio:
1365         ext2_error(sb, "ext2_get_inode",
1366                    "unable to read inode block - inode=%lu, block=%lu",
1367                    (unsigned long) ino, block);
1368 Egdp:
1369         return ERR_PTR(-EIO);
1370 }
1371 
1372 void ext2_set_inode_flags(struct inode *inode)
1373 {
1374         unsigned int flags = EXT2_I(inode)->i_flags;
1375 
1376         inode->i_flags &= ~(S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME |
1377                                 S_DIRSYNC | S_DAX);
1378         if (flags & EXT2_SYNC_FL)
1379                 inode->i_flags |= S_SYNC;
1380         if (flags & EXT2_APPEND_FL)
1381                 inode->i_flags |= S_APPEND;
1382         if (flags & EXT2_IMMUTABLE_FL)
1383                 inode->i_flags |= S_IMMUTABLE;
1384         if (flags & EXT2_NOATIME_FL)
1385                 inode->i_flags |= S_NOATIME;
1386         if (flags & EXT2_DIRSYNC_FL)
1387                 inode->i_flags |= S_DIRSYNC;
1388         if (test_opt(inode->i_sb, DAX) && S_ISREG(inode->i_mode))
1389                 inode->i_flags |= S_DAX;
1390 }
1391 
1392 void ext2_set_file_ops(struct inode *inode)
1393 {
1394         inode->i_op = &ext2_file_inode_operations;
1395         inode->i_fop = &ext2_file_operations;
1396         if (IS_DAX(inode))
1397                 inode->i_mapping->a_ops = &ext2_dax_aops;
1398         else if (test_opt(inode->i_sb, NOBH))
1399                 inode->i_mapping->a_ops = &ext2_nobh_aops;
1400         else
1401                 inode->i_mapping->a_ops = &ext2_aops;
1402 }
1403 
1404 struct inode *ext2_iget (struct super_block *sb, unsigned long ino)
1405 {
1406         struct ext2_inode_info *ei;
1407         struct buffer_head * bh = NULL;
1408         struct ext2_inode *raw_inode;
1409         struct inode *inode;
1410         long ret = -EIO;
1411         int n;
1412         uid_t i_uid;
1413         gid_t i_gid;
1414 
1415         inode = iget_locked(sb, ino);
1416         if (!inode)
1417                 return ERR_PTR(-ENOMEM);
1418         if (!(inode->i_state & I_NEW))
1419                 return inode;
1420 
1421         ei = EXT2_I(inode);
1422         ei->i_block_alloc_info = NULL;
1423 
1424         raw_inode = ext2_get_inode(inode->i_sb, ino, &bh);
1425         if (IS_ERR(raw_inode)) {
1426                 ret = PTR_ERR(raw_inode);
1427                 goto bad_inode;
1428         }
1429 
1430         inode->i_mode = le16_to_cpu(raw_inode->i_mode);
1431         i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low);
1432         i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low);
1433         if (!(test_opt (inode->i_sb, NO_UID32))) {
1434                 i_uid |= le16_to_cpu(raw_inode->i_uid_high) << 16;
1435                 i_gid |= le16_to_cpu(raw_inode->i_gid_high) << 16;
1436         }
1437         i_uid_write(inode, i_uid);
1438         i_gid_write(inode, i_gid);
1439         set_nlink(inode, le16_to_cpu(raw_inode->i_links_count));
1440         inode->i_size = le32_to_cpu(raw_inode->i_size);
1441         inode->i_atime.tv_sec = (signed)le32_to_cpu(raw_inode->i_atime);
1442         inode->i_ctime.tv_sec = (signed)le32_to_cpu(raw_inode->i_ctime);
1443         inode->i_mtime.tv_sec = (signed)le32_to_cpu(raw_inode->i_mtime);
1444         inode->i_atime.tv_nsec = inode->i_mtime.tv_nsec = inode->i_ctime.tv_nsec = 0;
1445         ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
1446         /* We now have enough fields to check if the inode was active or not.
1447          * This is needed because nfsd might try to access dead inodes
1448          * the test is that same one that e2fsck uses
1449          * NeilBrown 1999oct15
1450          */
1451         if (inode->i_nlink == 0 && (inode->i_mode == 0 || ei->i_dtime)) {
1452                 /* this inode is deleted */
1453                 ret = -ESTALE;
1454                 goto bad_inode;
1455         }
1456         inode->i_blocks = le32_to_cpu(raw_inode->i_blocks);
1457         ei->i_flags = le32_to_cpu(raw_inode->i_flags);
1458         ext2_set_inode_flags(inode);
1459         ei->i_faddr = le32_to_cpu(raw_inode->i_faddr);
1460         ei->i_frag_no = raw_inode->i_frag;
1461         ei->i_frag_size = raw_inode->i_fsize;
1462         ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl);
1463         ei->i_dir_acl = 0;
1464 
1465         if (ei->i_file_acl &&
1466             !ext2_data_block_valid(EXT2_SB(sb), ei->i_file_acl, 1)) {
1467                 ext2_error(sb, "ext2_iget", "bad extended attribute block %u",
1468                            ei->i_file_acl);
1469                 ret = -EFSCORRUPTED;
1470                 goto bad_inode;
1471         }
1472 
1473         if (S_ISREG(inode->i_mode))
1474                 inode->i_size |= ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32;
1475         else
1476                 ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl);
1477         if (i_size_read(inode) < 0) {
1478                 ret = -EFSCORRUPTED;
1479                 goto bad_inode;
1480         }
1481         ei->i_dtime = 0;
1482         inode->i_generation = le32_to_cpu(raw_inode->i_generation);
1483         ei->i_state = 0;
1484         ei->i_block_group = (ino - 1) / EXT2_INODES_PER_GROUP(inode->i_sb);
1485         ei->i_dir_start_lookup = 0;
1486 
1487         /*
1488          * NOTE! The in-memory inode i_data array is in little-endian order
1489          * even on big-endian machines: we do NOT byteswap the block numbers!
1490          */
1491         for (n = 0; n < EXT2_N_BLOCKS; n++)
1492                 ei->i_data[n] = raw_inode->i_block[n];
1493 
1494         if (S_ISREG(inode->i_mode)) {
1495                 ext2_set_file_ops(inode);
1496         } else if (S_ISDIR(inode->i_mode)) {
1497                 inode->i_op = &ext2_dir_inode_operations;
1498                 inode->i_fop = &ext2_dir_operations;
1499                 if (test_opt(inode->i_sb, NOBH))
1500                         inode->i_mapping->a_ops = &ext2_nobh_aops;
1501                 else
1502                         inode->i_mapping->a_ops = &ext2_aops;
1503         } else if (S_ISLNK(inode->i_mode)) {
1504                 if (ext2_inode_is_fast_symlink(inode)) {
1505                         inode->i_link = (char *)ei->i_data;
1506                         inode->i_op = &ext2_fast_symlink_inode_operations;
1507                         nd_terminate_link(ei->i_data, inode->i_size,
1508                                 sizeof(ei->i_data) - 1);
1509                 } else {
1510                         inode->i_op = &ext2_symlink_inode_operations;
1511                         inode_nohighmem(inode);
1512                         if (test_opt(inode->i_sb, NOBH))
1513                                 inode->i_mapping->a_ops = &ext2_nobh_aops;
1514                         else
1515                                 inode->i_mapping->a_ops = &ext2_aops;
1516                 }
1517         } else {
1518                 inode->i_op = &ext2_special_inode_operations;
1519                 if (raw_inode->i_block[0])
1520                         init_special_inode(inode, inode->i_mode,
1521                            old_decode_dev(le32_to_cpu(raw_inode->i_block[0])));
1522                 else 
1523                         init_special_inode(inode, inode->i_mode,
1524                            new_decode_dev(le32_to_cpu(raw_inode->i_block[1])));
1525         }
1526         brelse (bh);
1527         unlock_new_inode(inode);
1528         return inode;
1529         
1530 bad_inode:
1531         brelse(bh);
1532         iget_failed(inode);
1533         return ERR_PTR(ret);
1534 }
1535 
1536 static int __ext2_write_inode(struct inode *inode, int do_sync)
1537 {
1538         struct ext2_inode_info *ei = EXT2_I(inode);
1539         struct super_block *sb = inode->i_sb;
1540         ino_t ino = inode->i_ino;
1541         uid_t uid = i_uid_read(inode);
1542         gid_t gid = i_gid_read(inode);
1543         struct buffer_head * bh;
1544         struct ext2_inode * raw_inode = ext2_get_inode(sb, ino, &bh);
1545         int n;
1546         int err = 0;
1547 
1548         if (IS_ERR(raw_inode))
1549                 return -EIO;
1550 
1551         /* For fields not not tracking in the in-memory inode,
1552          * initialise them to zero for new inodes. */
1553         if (ei->i_state & EXT2_STATE_NEW)
1554                 memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
1555 
1556         raw_inode->i_mode = cpu_to_le16(inode->i_mode);
1557         if (!(test_opt(sb, NO_UID32))) {
1558                 raw_inode->i_uid_low = cpu_to_le16(low_16_bits(uid));
1559                 raw_inode->i_gid_low = cpu_to_le16(low_16_bits(gid));
1560 /*
1561  * Fix up interoperability with old kernels. Otherwise, old inodes get
1562  * re-used with the upper 16 bits of the uid/gid intact
1563  */
1564                 if (!ei->i_dtime) {
1565                         raw_inode->i_uid_high = cpu_to_le16(high_16_bits(uid));
1566                         raw_inode->i_gid_high = cpu_to_le16(high_16_bits(gid));
1567                 } else {
1568                         raw_inode->i_uid_high = 0;
1569                         raw_inode->i_gid_high = 0;
1570                 }
1571         } else {
1572                 raw_inode->i_uid_low = cpu_to_le16(fs_high2lowuid(uid));
1573                 raw_inode->i_gid_low = cpu_to_le16(fs_high2lowgid(gid));
1574                 raw_inode->i_uid_high = 0;
1575                 raw_inode->i_gid_high = 0;
1576         }
1577         raw_inode->i_links_count = cpu_to_le16(inode->i_nlink);
1578         raw_inode->i_size = cpu_to_le32(inode->i_size);
1579         raw_inode->i_atime = cpu_to_le32(inode->i_atime.tv_sec);
1580         raw_inode->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec);
1581         raw_inode->i_mtime = cpu_to_le32(inode->i_mtime.tv_sec);
1582 
1583         raw_inode->i_blocks = cpu_to_le32(inode->i_blocks);
1584         raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
1585         raw_inode->i_flags = cpu_to_le32(ei->i_flags);
1586         raw_inode->i_faddr = cpu_to_le32(ei->i_faddr);
1587         raw_inode->i_frag = ei->i_frag_no;
1588         raw_inode->i_fsize = ei->i_frag_size;
1589         raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl);
1590         if (!S_ISREG(inode->i_mode))
1591                 raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl);
1592         else {
1593                 raw_inode->i_size_high = cpu_to_le32(inode->i_size >> 32);
1594                 if (inode->i_size > 0x7fffffffULL) {
1595                         if (!EXT2_HAS_RO_COMPAT_FEATURE(sb,
1596                                         EXT2_FEATURE_RO_COMPAT_LARGE_FILE) ||
1597                             EXT2_SB(sb)->s_es->s_rev_level ==
1598                                         cpu_to_le32(EXT2_GOOD_OLD_REV)) {
1599                                /* If this is the first large file
1600                                 * created, add a flag to the superblock.
1601                                 */
1602                                 spin_lock(&EXT2_SB(sb)->s_lock);
1603                                 ext2_update_dynamic_rev(sb);
1604                                 EXT2_SET_RO_COMPAT_FEATURE(sb,
1605                                         EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
1606                                 spin_unlock(&EXT2_SB(sb)->s_lock);
1607                                 ext2_sync_super(sb, EXT2_SB(sb)->s_es, 1);
1608                         }
1609                 }
1610         }
1611         
1612         raw_inode->i_generation = cpu_to_le32(inode->i_generation);
1613         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
1614                 if (old_valid_dev(inode->i_rdev)) {
1615                         raw_inode->i_block[0] =
1616                                 cpu_to_le32(old_encode_dev(inode->i_rdev));
1617                         raw_inode->i_block[1] = 0;
1618                 } else {
1619                         raw_inode->i_block[0] = 0;
1620                         raw_inode->i_block[1] =
1621                                 cpu_to_le32(new_encode_dev(inode->i_rdev));
1622                         raw_inode->i_block[2] = 0;
1623                 }
1624         } else for (n = 0; n < EXT2_N_BLOCKS; n++)
1625                 raw_inode->i_block[n] = ei->i_data[n];
1626         mark_buffer_dirty(bh);
1627         if (do_sync) {
1628                 sync_dirty_buffer(bh);
1629                 if (buffer_req(bh) && !buffer_uptodate(bh)) {
1630                         printk ("IO error syncing ext2 inode [%s:%08lx]\n",
1631                                 sb->s_id, (unsigned long) ino);
1632                         err = -EIO;
1633                 }
1634         }
1635         ei->i_state &= ~EXT2_STATE_NEW;
1636         brelse (bh);
1637         return err;
1638 }
1639 
1640 int ext2_write_inode(struct inode *inode, struct writeback_control *wbc)
1641 {
1642         return __ext2_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
1643 }
1644 
1645 int ext2_getattr(const struct path *path, struct kstat *stat,
1646                 u32 request_mask, unsigned int query_flags)
1647 {
1648         struct inode *inode = d_inode(path->dentry);
1649         struct ext2_inode_info *ei = EXT2_I(inode);
1650         unsigned int flags;
1651 
1652         flags = ei->i_flags & EXT2_FL_USER_VISIBLE;
1653         if (flags & EXT2_APPEND_FL)
1654                 stat->attributes |= STATX_ATTR_APPEND;
1655         if (flags & EXT2_COMPR_FL)
1656                 stat->attributes |= STATX_ATTR_COMPRESSED;
1657         if (flags & EXT2_IMMUTABLE_FL)
1658                 stat->attributes |= STATX_ATTR_IMMUTABLE;
1659         if (flags & EXT2_NODUMP_FL)
1660                 stat->attributes |= STATX_ATTR_NODUMP;
1661         stat->attributes_mask |= (STATX_ATTR_APPEND |
1662                         STATX_ATTR_COMPRESSED |
1663                         STATX_ATTR_ENCRYPTED |
1664                         STATX_ATTR_IMMUTABLE |
1665                         STATX_ATTR_NODUMP);
1666 
1667         generic_fillattr(inode, stat);
1668         return 0;
1669 }
1670 
1671 int ext2_setattr(struct dentry *dentry, struct iattr *iattr)
1672 {
1673         struct inode *inode = d_inode(dentry);
1674         int error;
1675 
1676         error = setattr_prepare(dentry, iattr);
1677         if (error)
1678                 return error;
1679 
1680         if (is_quota_modification(inode, iattr)) {
1681                 error = dquot_initialize(inode);
1682                 if (error)
1683                         return error;
1684         }
1685         if ((iattr->ia_valid & ATTR_UID && !uid_eq(iattr->ia_uid, inode->i_uid)) ||
1686             (iattr->ia_valid & ATTR_GID && !gid_eq(iattr->ia_gid, inode->i_gid))) {
1687                 error = dquot_transfer(inode, iattr);
1688                 if (error)
1689                         return error;
1690         }
1691         if (iattr->ia_valid & ATTR_SIZE && iattr->ia_size != inode->i_size) {
1692                 error = ext2_setsize(inode, iattr->ia_size);
1693                 if (error)
1694                         return error;
1695         }
1696         setattr_copy(inode, iattr);
1697         if (iattr->ia_valid & ATTR_MODE)
1698                 error = posix_acl_chmod(inode, inode->i_mode);
1699         mark_inode_dirty(inode);
1700 
1701         return error;
1702 }
1703 

~ [ 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