xfs: fix various typos
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / xfs / xfs_bmap.c
CommitLineData
1da177e4 1/*
3e57ecf6 2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
7b718769 3 * All Rights Reserved.
1da177e4 4 *
7b718769
NS
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
1da177e4
LT
7 * published by the Free Software Foundation.
8 *
7b718769
NS
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
1da177e4 13 *
7b718769
NS
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1da177e4 17 */
1da177e4 18#include "xfs.h"
a844f451 19#include "xfs_fs.h"
1da177e4 20#include "xfs_types.h"
a844f451 21#include "xfs_bit.h"
1da177e4 22#include "xfs_log.h"
a844f451 23#include "xfs_inum.h"
1da177e4
LT
24#include "xfs_trans.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
1da177e4 27#include "xfs_dir2.h"
a844f451 28#include "xfs_da_btree.h"
1da177e4 29#include "xfs_bmap_btree.h"
a844f451 30#include "xfs_alloc_btree.h"
1da177e4 31#include "xfs_ialloc_btree.h"
1da177e4 32#include "xfs_dir2_sf.h"
a844f451 33#include "xfs_attr_sf.h"
1da177e4 34#include "xfs_dinode.h"
1da177e4 35#include "xfs_inode.h"
a844f451
NS
36#include "xfs_btree.h"
37#include "xfs_dmapi.h"
38#include "xfs_mount.h"
39#include "xfs_ialloc.h"
1da177e4 40#include "xfs_itable.h"
f6c2d1fa
NS
41#include "xfs_dir2_data.h"
42#include "xfs_dir2_leaf.h"
43#include "xfs_dir2_block.h"
a844f451 44#include "xfs_inode_item.h"
1da177e4
LT
45#include "xfs_extfree_item.h"
46#include "xfs_alloc.h"
47#include "xfs_bmap.h"
48#include "xfs_rtalloc.h"
49#include "xfs_error.h"
d8cc890d 50#include "xfs_attr_leaf.h"
1da177e4
LT
51#include "xfs_rw.h"
52#include "xfs_quota.h"
53#include "xfs_trans_space.h"
54#include "xfs_buf_item.h"
2a82b8be 55#include "xfs_filestream.h"
739bfb2a 56#include "xfs_vnodeops.h"
1da177e4
LT
57
58
59#ifdef DEBUG
60STATIC void
61xfs_bmap_check_leaf_extents(xfs_btree_cur_t *cur, xfs_inode_t *ip, int whichfork);
62#endif
63
64kmem_zone_t *xfs_bmap_free_item_zone;
65
66/*
67 * Prototypes for internal bmap routines.
68 */
69
70
71/*
72 * Called from xfs_bmap_add_attrfork to handle extents format files.
73 */
74STATIC int /* error */
75xfs_bmap_add_attrfork_extents(
76 xfs_trans_t *tp, /* transaction pointer */
77 xfs_inode_t *ip, /* incore inode pointer */
78 xfs_fsblock_t *firstblock, /* first block allocated */
79 xfs_bmap_free_t *flist, /* blocks to free at commit */
80 int *flags); /* inode logging flags */
81
82/*
83 * Called from xfs_bmap_add_attrfork to handle local format files.
84 */
85STATIC int /* error */
86xfs_bmap_add_attrfork_local(
87 xfs_trans_t *tp, /* transaction pointer */
88 xfs_inode_t *ip, /* incore inode pointer */
89 xfs_fsblock_t *firstblock, /* first block allocated */
90 xfs_bmap_free_t *flist, /* blocks to free at commit */
91 int *flags); /* inode logging flags */
92
93/*
4eea22f0 94 * Called by xfs_bmapi to update file extent records and the btree
1da177e4
LT
95 * after allocating space (or doing a delayed allocation).
96 */
97STATIC int /* error */
98xfs_bmap_add_extent(
99 xfs_inode_t *ip, /* incore inode pointer */
100 xfs_extnum_t idx, /* extent number to update/insert */
101 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 102 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4
LT
103 xfs_fsblock_t *first, /* pointer to firstblock variable */
104 xfs_bmap_free_t *flist, /* list of extents to be freed */
105 int *logflagsp, /* inode logging flags */
3e57ecf6 106 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
107 int whichfork, /* data or attr fork */
108 int rsvd); /* OK to allocate reserved blocks */
109
110/*
111 * Called by xfs_bmap_add_extent to handle cases converting a delayed
112 * allocation to a real allocation.
113 */
114STATIC int /* error */
115xfs_bmap_add_extent_delay_real(
116 xfs_inode_t *ip, /* incore inode pointer */
117 xfs_extnum_t idx, /* extent number to update/insert */
118 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 119 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4
LT
120 xfs_filblks_t *dnew, /* new delayed-alloc indirect blocks */
121 xfs_fsblock_t *first, /* pointer to firstblock variable */
122 xfs_bmap_free_t *flist, /* list of extents to be freed */
123 int *logflagsp, /* inode logging flags */
3e57ecf6 124 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
125 int rsvd); /* OK to allocate reserved blocks */
126
127/*
128 * Called by xfs_bmap_add_extent to handle cases converting a hole
129 * to a delayed allocation.
130 */
131STATIC int /* error */
132xfs_bmap_add_extent_hole_delay(
133 xfs_inode_t *ip, /* incore inode pointer */
134 xfs_extnum_t idx, /* extent number to update/insert */
4eea22f0 135 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4 136 int *logflagsp,/* inode logging flags */
3e57ecf6 137 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
138 int rsvd); /* OK to allocate reserved blocks */
139
140/*
141 * Called by xfs_bmap_add_extent to handle cases converting a hole
142 * to a real allocation.
143 */
144STATIC int /* error */
145xfs_bmap_add_extent_hole_real(
146 xfs_inode_t *ip, /* incore inode pointer */
147 xfs_extnum_t idx, /* extent number to update/insert */
148 xfs_btree_cur_t *cur, /* if null, not a btree */
4eea22f0 149 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4 150 int *logflagsp, /* inode logging flags */
3e57ecf6 151 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
152 int whichfork); /* data or attr fork */
153
154/*
155 * Called by xfs_bmap_add_extent to handle cases converting an unwritten
156 * allocation to a real allocation or vice versa.
157 */
158STATIC int /* error */
159xfs_bmap_add_extent_unwritten_real(
160 xfs_inode_t *ip, /* incore inode pointer */
161 xfs_extnum_t idx, /* extent number to update/insert */
162 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 163 xfs_bmbt_irec_t *new, /* new data to add to file extents */
3e57ecf6
OW
164 int *logflagsp, /* inode logging flags */
165 xfs_extdelta_t *delta); /* Change made to incore extents */
1da177e4
LT
166
167/*
168 * xfs_bmap_alloc is called by xfs_bmapi to allocate an extent for a file.
169 * It figures out where to ask the underlying allocator to put the new extent.
170 */
171STATIC int /* error */
172xfs_bmap_alloc(
173 xfs_bmalloca_t *ap); /* bmap alloc argument struct */
174
175/*
176 * Transform a btree format file with only one leaf node, where the
177 * extents list will fit in the inode, into an extents format file.
4eea22f0 178 * Since the file extents are already in-core, all we have to do is
1da177e4
LT
179 * give up the space for the btree root and pitch the leaf block.
180 */
181STATIC int /* error */
182xfs_bmap_btree_to_extents(
183 xfs_trans_t *tp, /* transaction pointer */
184 xfs_inode_t *ip, /* incore inode pointer */
185 xfs_btree_cur_t *cur, /* btree cursor */
186 int *logflagsp, /* inode logging flags */
187 int whichfork); /* data or attr fork */
188
1da177e4 189/*
4eea22f0 190 * Called by xfs_bmapi to update file extent records and the btree
1da177e4
LT
191 * after removing space (or undoing a delayed allocation).
192 */
193STATIC int /* error */
194xfs_bmap_del_extent(
195 xfs_inode_t *ip, /* incore inode pointer */
196 xfs_trans_t *tp, /* current trans pointer */
197 xfs_extnum_t idx, /* extent number to update/insert */
198 xfs_bmap_free_t *flist, /* list of extents to be freed */
199 xfs_btree_cur_t *cur, /* if null, not a btree */
4eea22f0 200 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4 201 int *logflagsp,/* inode logging flags */
3e57ecf6 202 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
203 int whichfork, /* data or attr fork */
204 int rsvd); /* OK to allocate reserved blocks */
205
206/*
207 * Remove the entry "free" from the free item list. Prev points to the
208 * previous entry, unless "free" is the head of the list.
209 */
210STATIC void
211xfs_bmap_del_free(
212 xfs_bmap_free_t *flist, /* free item list header */
213 xfs_bmap_free_item_t *prev, /* previous item on list, if any */
214 xfs_bmap_free_item_t *free); /* list item to be freed */
215
1da177e4
LT
216/*
217 * Convert an extents-format file into a btree-format file.
218 * The new file will have a root block (in the inode) and a single child block.
219 */
220STATIC int /* error */
221xfs_bmap_extents_to_btree(
222 xfs_trans_t *tp, /* transaction pointer */
223 xfs_inode_t *ip, /* incore inode pointer */
224 xfs_fsblock_t *firstblock, /* first-block-allocated */
225 xfs_bmap_free_t *flist, /* blocks freed in xaction */
226 xfs_btree_cur_t **curp, /* cursor returned to caller */
227 int wasdel, /* converting a delayed alloc */
228 int *logflagsp, /* inode logging flags */
229 int whichfork); /* data or attr fork */
230
1da177e4
LT
231/*
232 * Convert a local file to an extents file.
233 * This code is sort of bogus, since the file data needs to get
234 * logged so it won't be lost. The bmap-level manipulations are ok, though.
235 */
236STATIC int /* error */
237xfs_bmap_local_to_extents(
238 xfs_trans_t *tp, /* transaction pointer */
239 xfs_inode_t *ip, /* incore inode pointer */
240 xfs_fsblock_t *firstblock, /* first block allocated in xaction */
241 xfs_extlen_t total, /* total blocks needed by transaction */
242 int *logflagsp, /* inode logging flags */
243 int whichfork); /* data or attr fork */
244
245/*
246 * Search the extents list for the inode, for the extent containing bno.
247 * If bno lies in a hole, point to the next entry. If bno lies past eof,
248 * *eofp will be set, and *prevp will contain the last entry (null if none).
249 * Else, *lastxp will be set to the index of the found
250 * entry; *gotp will contain the entry.
251 */
a6f64d4a 252STATIC xfs_bmbt_rec_host_t * /* pointer to found extent entry */
1da177e4
LT
253xfs_bmap_search_extents(
254 xfs_inode_t *ip, /* incore inode pointer */
255 xfs_fileoff_t bno, /* block number searched for */
256 int whichfork, /* data or attr fork */
257 int *eofp, /* out: end of file found */
258 xfs_extnum_t *lastxp, /* out: last extent index */
259 xfs_bmbt_irec_t *gotp, /* out: extent entry found */
260 xfs_bmbt_irec_t *prevp); /* out: previous extent entry found */
261
ba0f32d4
CH
262/*
263 * Check the last inode extent to determine whether this allocation will result
264 * in blocks being allocated at the end of the file. When we allocate new data
265 * blocks at the end of the file which do not start at the previous data block,
266 * we will try to align the new blocks at stripe unit boundaries.
267 */
268STATIC int /* error */
269xfs_bmap_isaeof(
270 xfs_inode_t *ip, /* incore inode pointer */
271 xfs_fileoff_t off, /* file offset in fsblocks */
272 int whichfork, /* data or attribute fork */
273 char *aeof); /* return value */
274
1da177e4 275#ifdef XFS_BMAP_TRACE
1da177e4 276/*
4eea22f0 277 * Add bmap trace entry prior to a call to xfs_iext_remove.
1da177e4
LT
278 */
279STATIC void
280xfs_bmap_trace_delete(
3a59c94c 281 const char *fname, /* function name */
1da177e4
LT
282 char *desc, /* operation description */
283 xfs_inode_t *ip, /* incore inode pointer */
284 xfs_extnum_t idx, /* index of entry(entries) deleted */
285 xfs_extnum_t cnt, /* count of entries deleted, 1 or 2 */
286 int whichfork); /* data or attr fork */
287
288/*
4eea22f0 289 * Add bmap trace entry prior to a call to xfs_iext_insert, or
1da177e4
LT
290 * reading in the extents list from the disk (in the btree).
291 */
292STATIC void
293xfs_bmap_trace_insert(
3a59c94c 294 const char *fname, /* function name */
1da177e4
LT
295 char *desc, /* operation description */
296 xfs_inode_t *ip, /* incore inode pointer */
297 xfs_extnum_t idx, /* index of entry(entries) inserted */
298 xfs_extnum_t cnt, /* count of entries inserted, 1 or 2 */
299 xfs_bmbt_irec_t *r1, /* inserted record 1 */
300 xfs_bmbt_irec_t *r2, /* inserted record 2 or null */
301 int whichfork); /* data or attr fork */
302
303/*
4eea22f0 304 * Add bmap trace entry after updating an extent record in place.
1da177e4
LT
305 */
306STATIC void
307xfs_bmap_trace_post_update(
3a59c94c 308 const char *fname, /* function name */
1da177e4
LT
309 char *desc, /* operation description */
310 xfs_inode_t *ip, /* incore inode pointer */
311 xfs_extnum_t idx, /* index of entry updated */
312 int whichfork); /* data or attr fork */
313
314/*
4eea22f0 315 * Add bmap trace entry prior to updating an extent record in place.
1da177e4
LT
316 */
317STATIC void
318xfs_bmap_trace_pre_update(
3a59c94c 319 const char *fname, /* function name */
1da177e4
LT
320 char *desc, /* operation description */
321 xfs_inode_t *ip, /* incore inode pointer */
322 xfs_extnum_t idx, /* index of entry to be updated */
323 int whichfork); /* data or attr fork */
324
3a59c94c 325#define XFS_BMAP_TRACE_DELETE(d,ip,i,c,w) \
34a622b2 326 xfs_bmap_trace_delete(__func__,d,ip,i,c,w)
3a59c94c 327#define XFS_BMAP_TRACE_INSERT(d,ip,i,c,r1,r2,w) \
34a622b2 328 xfs_bmap_trace_insert(__func__,d,ip,i,c,r1,r2,w)
3a59c94c 329#define XFS_BMAP_TRACE_POST_UPDATE(d,ip,i,w) \
34a622b2 330 xfs_bmap_trace_post_update(__func__,d,ip,i,w)
3a59c94c 331#define XFS_BMAP_TRACE_PRE_UPDATE(d,ip,i,w) \
34a622b2 332 xfs_bmap_trace_pre_update(__func__,d,ip,i,w)
1da177e4 333#else
3a59c94c
ES
334#define XFS_BMAP_TRACE_DELETE(d,ip,i,c,w)
335#define XFS_BMAP_TRACE_INSERT(d,ip,i,c,r1,r2,w)
336#define XFS_BMAP_TRACE_POST_UPDATE(d,ip,i,w)
337#define XFS_BMAP_TRACE_PRE_UPDATE(d,ip,i,w)
1da177e4
LT
338#endif /* XFS_BMAP_TRACE */
339
340/*
341 * Compute the worst-case number of indirect blocks that will be used
342 * for ip's delayed extent of length "len".
343 */
344STATIC xfs_filblks_t
345xfs_bmap_worst_indlen(
346 xfs_inode_t *ip, /* incore inode pointer */
347 xfs_filblks_t len); /* delayed extent length */
348
349#ifdef DEBUG
350/*
351 * Perform various validation checks on the values being returned
352 * from xfs_bmapi().
353 */
354STATIC void
355xfs_bmap_validate_ret(
356 xfs_fileoff_t bno,
357 xfs_filblks_t len,
358 int flags,
359 xfs_bmbt_irec_t *mval,
360 int nmap,
361 int ret_nmap);
362#else
363#define xfs_bmap_validate_ret(bno,len,flags,mval,onmap,nmap)
364#endif /* DEBUG */
365
366#if defined(XFS_RW_TRACE)
367STATIC void
368xfs_bunmap_trace(
369 xfs_inode_t *ip,
370 xfs_fileoff_t bno,
371 xfs_filblks_t len,
372 int flags,
373 inst_t *ra);
374#else
375#define xfs_bunmap_trace(ip, bno, len, flags, ra)
376#endif /* XFS_RW_TRACE */
377
378STATIC int
379xfs_bmap_count_tree(
380 xfs_mount_t *mp,
381 xfs_trans_t *tp,
4eea22f0 382 xfs_ifork_t *ifp,
1da177e4
LT
383 xfs_fsblock_t blockno,
384 int levelin,
385 int *count);
386
c94312de 387STATIC void
1da177e4 388xfs_bmap_count_leaves(
4eea22f0
MK
389 xfs_ifork_t *ifp,
390 xfs_extnum_t idx,
1da177e4
LT
391 int numrecs,
392 int *count);
393
c94312de 394STATIC void
91e11088 395xfs_bmap_disk_count_leaves(
136341b4 396 struct xfs_mount *mp,
7cc95a82 397 struct xfs_btree_block *block,
91e11088
YL
398 int numrecs,
399 int *count);
400
1da177e4
LT
401/*
402 * Bmap internal routines.
403 */
404
fe033cc8
CH
405STATIC int /* error */
406xfs_bmbt_lookup_eq(
407 struct xfs_btree_cur *cur,
408 xfs_fileoff_t off,
409 xfs_fsblock_t bno,
410 xfs_filblks_t len,
411 int *stat) /* success/failure */
412{
413 cur->bc_rec.b.br_startoff = off;
414 cur->bc_rec.b.br_startblock = bno;
415 cur->bc_rec.b.br_blockcount = len;
416 return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
417}
418
419STATIC int /* error */
420xfs_bmbt_lookup_ge(
421 struct xfs_btree_cur *cur,
422 xfs_fileoff_t off,
423 xfs_fsblock_t bno,
424 xfs_filblks_t len,
425 int *stat) /* success/failure */
426{
427 cur->bc_rec.b.br_startoff = off;
428 cur->bc_rec.b.br_startblock = bno;
429 cur->bc_rec.b.br_blockcount = len;
430 return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
431}
432
278d0ca1
CH
433/*
434* Update the record referred to by cur to the value given
435 * by [off, bno, len, state].
436 * This either works (return 0) or gets an EFSCORRUPTED error.
437 */
438STATIC int
439xfs_bmbt_update(
440 struct xfs_btree_cur *cur,
441 xfs_fileoff_t off,
442 xfs_fsblock_t bno,
443 xfs_filblks_t len,
444 xfs_exntst_t state)
445{
446 union xfs_btree_rec rec;
447
448 xfs_bmbt_disk_set_allf(&rec.bmbt, off, bno, len, state);
449 return xfs_btree_update(cur, &rec);
450}
fe033cc8 451
1da177e4
LT
452/*
453 * Called from xfs_bmap_add_attrfork to handle btree format files.
454 */
455STATIC int /* error */
456xfs_bmap_add_attrfork_btree(
457 xfs_trans_t *tp, /* transaction pointer */
458 xfs_inode_t *ip, /* incore inode pointer */
459 xfs_fsblock_t *firstblock, /* first block allocated */
460 xfs_bmap_free_t *flist, /* blocks to free at commit */
461 int *flags) /* inode logging flags */
462{
463 xfs_btree_cur_t *cur; /* btree cursor */
464 int error; /* error return value */
465 xfs_mount_t *mp; /* file system mount struct */
466 int stat; /* newroot status */
467
468 mp = ip->i_mount;
469 if (ip->i_df.if_broot_bytes <= XFS_IFORK_DSIZE(ip))
470 *flags |= XFS_ILOG_DBROOT;
471 else {
561f7d17 472 cur = xfs_bmbt_init_cursor(mp, tp, ip, XFS_DATA_FORK);
1da177e4
LT
473 cur->bc_private.b.flist = flist;
474 cur->bc_private.b.firstblock = *firstblock;
475 if ((error = xfs_bmbt_lookup_ge(cur, 0, 0, 0, &stat)))
476 goto error0;
6bd8fc8a
LM
477 /* must be at least one entry */
478 XFS_WANT_CORRUPTED_GOTO(stat == 1, error0);
ea77b0a6 479 if ((error = xfs_btree_new_iroot(cur, flags, &stat)))
1da177e4
LT
480 goto error0;
481 if (stat == 0) {
482 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
483 return XFS_ERROR(ENOSPC);
484 }
485 *firstblock = cur->bc_private.b.firstblock;
486 cur->bc_private.b.allocated = 0;
487 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
488 }
489 return 0;
490error0:
491 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
492 return error;
493}
494
495/*
496 * Called from xfs_bmap_add_attrfork to handle extents format files.
497 */
498STATIC int /* error */
499xfs_bmap_add_attrfork_extents(
500 xfs_trans_t *tp, /* transaction pointer */
501 xfs_inode_t *ip, /* incore inode pointer */
502 xfs_fsblock_t *firstblock, /* first block allocated */
503 xfs_bmap_free_t *flist, /* blocks to free at commit */
504 int *flags) /* inode logging flags */
505{
506 xfs_btree_cur_t *cur; /* bmap btree cursor */
507 int error; /* error return value */
508
509 if (ip->i_d.di_nextents * sizeof(xfs_bmbt_rec_t) <= XFS_IFORK_DSIZE(ip))
510 return 0;
511 cur = NULL;
512 error = xfs_bmap_extents_to_btree(tp, ip, firstblock, flist, &cur, 0,
513 flags, XFS_DATA_FORK);
514 if (cur) {
515 cur->bc_private.b.allocated = 0;
516 xfs_btree_del_cursor(cur,
517 error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
518 }
519 return error;
520}
521
522/*
523 * Called from xfs_bmap_add_attrfork to handle local format files.
524 */
525STATIC int /* error */
526xfs_bmap_add_attrfork_local(
527 xfs_trans_t *tp, /* transaction pointer */
528 xfs_inode_t *ip, /* incore inode pointer */
529 xfs_fsblock_t *firstblock, /* first block allocated */
530 xfs_bmap_free_t *flist, /* blocks to free at commit */
531 int *flags) /* inode logging flags */
532{
533 xfs_da_args_t dargs; /* args for dir/attr code */
534 int error; /* error return value */
535 xfs_mount_t *mp; /* mount structure pointer */
536
537 if (ip->i_df.if_bytes <= XFS_IFORK_DSIZE(ip))
538 return 0;
539 if ((ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
540 mp = ip->i_mount;
541 memset(&dargs, 0, sizeof(dargs));
542 dargs.dp = ip;
543 dargs.firstblock = firstblock;
544 dargs.flist = flist;
545 dargs.total = mp->m_dirblkfsbs;
546 dargs.whichfork = XFS_DATA_FORK;
547 dargs.trans = tp;
f6c2d1fa 548 error = xfs_dir2_sf_to_block(&dargs);
1da177e4
LT
549 } else
550 error = xfs_bmap_local_to_extents(tp, ip, firstblock, 1, flags,
551 XFS_DATA_FORK);
552 return error;
553}
554
555/*
4eea22f0 556 * Called by xfs_bmapi to update file extent records and the btree
1da177e4
LT
557 * after allocating space (or doing a delayed allocation).
558 */
559STATIC int /* error */
560xfs_bmap_add_extent(
561 xfs_inode_t *ip, /* incore inode pointer */
562 xfs_extnum_t idx, /* extent number to update/insert */
563 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 564 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4
LT
565 xfs_fsblock_t *first, /* pointer to firstblock variable */
566 xfs_bmap_free_t *flist, /* list of extents to be freed */
567 int *logflagsp, /* inode logging flags */
3e57ecf6 568 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
569 int whichfork, /* data or attr fork */
570 int rsvd) /* OK to use reserved data blocks */
571{
572 xfs_btree_cur_t *cur; /* btree cursor or null */
573 xfs_filblks_t da_new; /* new count del alloc blocks used */
574 xfs_filblks_t da_old; /* old count del alloc blocks used */
575 int error; /* error return value */
1da177e4
LT
576 xfs_ifork_t *ifp; /* inode fork ptr */
577 int logflags; /* returned value */
578 xfs_extnum_t nextents; /* number of extents in file now */
579
580 XFS_STATS_INC(xs_add_exlist);
581 cur = *curp;
582 ifp = XFS_IFORK_PTR(ip, whichfork);
583 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
584 ASSERT(idx <= nextents);
585 da_old = da_new = 0;
586 error = 0;
587 /*
588 * This is the first extent added to a new/empty file.
589 * Special case this one, so other routines get to assume there are
590 * already extents in the list.
591 */
592 if (nextents == 0) {
3a59c94c
ES
593 XFS_BMAP_TRACE_INSERT("insert empty", ip, 0, 1, new, NULL,
594 whichfork);
4eea22f0 595 xfs_iext_insert(ifp, 0, 1, new);
1da177e4
LT
596 ASSERT(cur == NULL);
597 ifp->if_lastex = 0;
9d87c319 598 if (!isnullstartblock(new->br_startblock)) {
1da177e4 599 XFS_IFORK_NEXT_SET(ip, whichfork, 1);
9d87c319 600 logflags = XFS_ILOG_CORE | xfs_ilog_fext(whichfork);
1da177e4
LT
601 } else
602 logflags = 0;
3e57ecf6
OW
603 /* DELTA: single new extent */
604 if (delta) {
605 if (delta->xed_startoff > new->br_startoff)
606 delta->xed_startoff = new->br_startoff;
607 if (delta->xed_blockcount <
608 new->br_startoff + new->br_blockcount)
609 delta->xed_blockcount = new->br_startoff +
610 new->br_blockcount;
611 }
1da177e4
LT
612 }
613 /*
614 * Any kind of new delayed allocation goes here.
615 */
9d87c319 616 else if (isnullstartblock(new->br_startblock)) {
1da177e4
LT
617 if (cur)
618 ASSERT((cur->bc_private.b.flags &
619 XFS_BTCUR_BPRV_WASDEL) == 0);
e9ed9d22 620 if ((error = xfs_bmap_add_extent_hole_delay(ip, idx, new,
3e57ecf6 621 &logflags, delta, rsvd)))
1da177e4
LT
622 goto done;
623 }
624 /*
625 * Real allocation off the end of the file.
626 */
627 else if (idx == nextents) {
628 if (cur)
629 ASSERT((cur->bc_private.b.flags &
630 XFS_BTCUR_BPRV_WASDEL) == 0);
631 if ((error = xfs_bmap_add_extent_hole_real(ip, idx, cur, new,
3e57ecf6 632 &logflags, delta, whichfork)))
1da177e4
LT
633 goto done;
634 } else {
635 xfs_bmbt_irec_t prev; /* old extent at offset idx */
636
637 /*
638 * Get the record referred to by idx.
639 */
4eea22f0 640 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx), &prev);
1da177e4
LT
641 /*
642 * If it's a real allocation record, and the new allocation ends
643 * after the start of the referred to record, then we're filling
644 * in a delayed or unwritten allocation with a real one, or
645 * converting real back to unwritten.
646 */
9d87c319 647 if (!isnullstartblock(new->br_startblock) &&
1da177e4
LT
648 new->br_startoff + new->br_blockcount > prev.br_startoff) {
649 if (prev.br_state != XFS_EXT_UNWRITTEN &&
9d87c319
ES
650 isnullstartblock(prev.br_startblock)) {
651 da_old = startblockval(prev.br_startblock);
1da177e4
LT
652 if (cur)
653 ASSERT(cur->bc_private.b.flags &
654 XFS_BTCUR_BPRV_WASDEL);
655 if ((error = xfs_bmap_add_extent_delay_real(ip,
656 idx, &cur, new, &da_new, first, flist,
3e57ecf6 657 &logflags, delta, rsvd)))
1da177e4
LT
658 goto done;
659 } else if (new->br_state == XFS_EXT_NORM) {
660 ASSERT(new->br_state == XFS_EXT_NORM);
661 if ((error = xfs_bmap_add_extent_unwritten_real(
3e57ecf6 662 ip, idx, &cur, new, &logflags, delta)))
1da177e4
LT
663 goto done;
664 } else {
665 ASSERT(new->br_state == XFS_EXT_UNWRITTEN);
666 if ((error = xfs_bmap_add_extent_unwritten_real(
3e57ecf6 667 ip, idx, &cur, new, &logflags, delta)))
1da177e4
LT
668 goto done;
669 }
670 ASSERT(*curp == cur || *curp == NULL);
671 }
672 /*
673 * Otherwise we're filling in a hole with an allocation.
674 */
675 else {
676 if (cur)
677 ASSERT((cur->bc_private.b.flags &
678 XFS_BTCUR_BPRV_WASDEL) == 0);
679 if ((error = xfs_bmap_add_extent_hole_real(ip, idx, cur,
3e57ecf6 680 new, &logflags, delta, whichfork)))
1da177e4
LT
681 goto done;
682 }
683 }
684
685 ASSERT(*curp == cur || *curp == NULL);
686 /*
687 * Convert to a btree if necessary.
688 */
689 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS &&
690 XFS_IFORK_NEXTENTS(ip, whichfork) > ifp->if_ext_max) {
691 int tmp_logflags; /* partial log flag return val */
692
693 ASSERT(cur == NULL);
694 error = xfs_bmap_extents_to_btree(ip->i_transp, ip, first,
695 flist, &cur, da_old > 0, &tmp_logflags, whichfork);
696 logflags |= tmp_logflags;
697 if (error)
698 goto done;
699 }
700 /*
701 * Adjust for changes in reserved delayed indirect blocks.
702 * Nothing to do for disk quotas here.
703 */
704 if (da_old || da_new) {
705 xfs_filblks_t nblks;
706
707 nblks = da_new;
708 if (cur)
709 nblks += cur->bc_private.b.allocated;
710 ASSERT(nblks <= da_old);
711 if (nblks < da_old)
712 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS,
20f4ebf2 713 (int64_t)(da_old - nblks), rsvd);
1da177e4
LT
714 }
715 /*
716 * Clear out the allocated field, done with it now in any case.
717 */
718 if (cur) {
719 cur->bc_private.b.allocated = 0;
720 *curp = cur;
721 }
722done:
723#ifdef DEBUG
724 if (!error)
725 xfs_bmap_check_leaf_extents(*curp, ip, whichfork);
726#endif
727 *logflagsp = logflags;
728 return error;
729}
730
731/*
732 * Called by xfs_bmap_add_extent to handle cases converting a delayed
733 * allocation to a real allocation.
734 */
735STATIC int /* error */
736xfs_bmap_add_extent_delay_real(
737 xfs_inode_t *ip, /* incore inode pointer */
738 xfs_extnum_t idx, /* extent number to update/insert */
739 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 740 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4
LT
741 xfs_filblks_t *dnew, /* new delayed-alloc indirect blocks */
742 xfs_fsblock_t *first, /* pointer to firstblock variable */
743 xfs_bmap_free_t *flist, /* list of extents to be freed */
744 int *logflagsp, /* inode logging flags */
3e57ecf6 745 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
746 int rsvd) /* OK to use reserved data block allocation */
747{
1da177e4
LT
748 xfs_btree_cur_t *cur; /* btree cursor */
749 int diff; /* temp value */
a6f64d4a 750 xfs_bmbt_rec_host_t *ep; /* extent entry for idx */
1da177e4 751 int error; /* error return value */
1da177e4 752 int i; /* temp state */
4eea22f0 753 xfs_ifork_t *ifp; /* inode fork pointer */
1da177e4
LT
754 xfs_fileoff_t new_endoff; /* end offset of new entry */
755 xfs_bmbt_irec_t r[3]; /* neighbor extent entries */
756 /* left is 0, right is 1, prev is 2 */
757 int rval=0; /* return value (logging flags) */
758 int state = 0;/* state bits, accessed thru macros */
3e57ecf6
OW
759 xfs_filblks_t temp=0; /* value for dnew calculations */
760 xfs_filblks_t temp2=0;/* value for dnew calculations */
1da177e4
LT
761 int tmp_rval; /* partial logging flags */
762 enum { /* bit number definitions for state */
763 LEFT_CONTIG, RIGHT_CONTIG,
764 LEFT_FILLING, RIGHT_FILLING,
765 LEFT_DELAY, RIGHT_DELAY,
766 LEFT_VALID, RIGHT_VALID
767 };
768
769#define LEFT r[0]
770#define RIGHT r[1]
771#define PREV r[2]
772#define MASK(b) (1 << (b))
773#define MASK2(a,b) (MASK(a) | MASK(b))
774#define MASK3(a,b,c) (MASK2(a,b) | MASK(c))
775#define MASK4(a,b,c,d) (MASK3(a,b,c) | MASK(d))
776#define STATE_SET(b,v) ((v) ? (state |= MASK(b)) : (state &= ~MASK(b)))
777#define STATE_TEST(b) (state & MASK(b))
778#define STATE_SET_TEST(b,v) ((v) ? ((state |= MASK(b)), 1) : \
779 ((state &= ~MASK(b)), 0))
780#define SWITCH_STATE \
781 (state & MASK4(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG))
782
783 /*
784 * Set up a bunch of variables to make the tests simpler.
785 */
786 cur = *curp;
4eea22f0
MK
787 ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
788 ep = xfs_iext_get_ext(ifp, idx);
1da177e4
LT
789 xfs_bmbt_get_all(ep, &PREV);
790 new_endoff = new->br_startoff + new->br_blockcount;
791 ASSERT(PREV.br_startoff <= new->br_startoff);
792 ASSERT(PREV.br_startoff + PREV.br_blockcount >= new_endoff);
793 /*
794 * Set flags determining what part of the previous delayed allocation
795 * extent is being replaced by a real allocation.
796 */
797 STATE_SET(LEFT_FILLING, PREV.br_startoff == new->br_startoff);
798 STATE_SET(RIGHT_FILLING,
799 PREV.br_startoff + PREV.br_blockcount == new_endoff);
800 /*
801 * Check and set flags if this segment has a left neighbor.
802 * Don't set contiguous if the combined extent would be too large.
803 */
804 if (STATE_SET_TEST(LEFT_VALID, idx > 0)) {
4eea22f0 805 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx - 1), &LEFT);
9d87c319 806 STATE_SET(LEFT_DELAY, isnullstartblock(LEFT.br_startblock));
1da177e4
LT
807 }
808 STATE_SET(LEFT_CONTIG,
809 STATE_TEST(LEFT_VALID) && !STATE_TEST(LEFT_DELAY) &&
810 LEFT.br_startoff + LEFT.br_blockcount == new->br_startoff &&
811 LEFT.br_startblock + LEFT.br_blockcount == new->br_startblock &&
812 LEFT.br_state == new->br_state &&
813 LEFT.br_blockcount + new->br_blockcount <= MAXEXTLEN);
814 /*
815 * Check and set flags if this segment has a right neighbor.
816 * Don't set contiguous if the combined extent would be too large.
817 * Also check for all-three-contiguous being too large.
818 */
819 if (STATE_SET_TEST(RIGHT_VALID,
820 idx <
821 ip->i_df.if_bytes / (uint)sizeof(xfs_bmbt_rec_t) - 1)) {
4eea22f0 822 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx + 1), &RIGHT);
9d87c319 823 STATE_SET(RIGHT_DELAY, isnullstartblock(RIGHT.br_startblock));
1da177e4
LT
824 }
825 STATE_SET(RIGHT_CONTIG,
826 STATE_TEST(RIGHT_VALID) && !STATE_TEST(RIGHT_DELAY) &&
827 new_endoff == RIGHT.br_startoff &&
828 new->br_startblock + new->br_blockcount ==
829 RIGHT.br_startblock &&
830 new->br_state == RIGHT.br_state &&
831 new->br_blockcount + RIGHT.br_blockcount <= MAXEXTLEN &&
832 ((state & MASK3(LEFT_CONTIG, LEFT_FILLING, RIGHT_FILLING)) !=
833 MASK3(LEFT_CONTIG, LEFT_FILLING, RIGHT_FILLING) ||
834 LEFT.br_blockcount + new->br_blockcount + RIGHT.br_blockcount
835 <= MAXEXTLEN));
836 error = 0;
837 /*
838 * Switch out based on the FILLING and CONTIG state bits.
839 */
840 switch (SWITCH_STATE) {
841
842 case MASK4(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
843 /*
844 * Filling in all of a previously delayed allocation extent.
845 * The left and right neighbors are both contiguous with new.
846 */
3a59c94c 847 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|LC|RC", ip, idx - 1,
1da177e4 848 XFS_DATA_FORK);
4eea22f0 849 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4
LT
850 LEFT.br_blockcount + PREV.br_blockcount +
851 RIGHT.br_blockcount);
3a59c94c 852 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|LC|RC", ip, idx - 1,
1da177e4 853 XFS_DATA_FORK);
3a59c94c 854 XFS_BMAP_TRACE_DELETE("LF|RF|LC|RC", ip, idx, 2, XFS_DATA_FORK);
4eea22f0 855 xfs_iext_remove(ifp, idx, 2);
1da177e4
LT
856 ip->i_df.if_lastex = idx - 1;
857 ip->i_d.di_nextents--;
858 if (cur == NULL)
859 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
860 else {
861 rval = XFS_ILOG_CORE;
862 if ((error = xfs_bmbt_lookup_eq(cur, RIGHT.br_startoff,
863 RIGHT.br_startblock,
864 RIGHT.br_blockcount, &i)))
865 goto done;
6bd8fc8a 866 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 867 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 868 goto done;
6bd8fc8a 869 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 870 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4 871 goto done;
6bd8fc8a 872 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
873 if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
874 LEFT.br_startblock,
875 LEFT.br_blockcount +
876 PREV.br_blockcount +
877 RIGHT.br_blockcount, LEFT.br_state)))
878 goto done;
879 }
880 *dnew = 0;
3e57ecf6
OW
881 /* DELTA: Three in-core extents are replaced by one. */
882 temp = LEFT.br_startoff;
883 temp2 = LEFT.br_blockcount +
884 PREV.br_blockcount +
885 RIGHT.br_blockcount;
1da177e4
LT
886 break;
887
888 case MASK3(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG):
889 /*
890 * Filling in all of a previously delayed allocation extent.
891 * The left neighbor is contiguous, the right is not.
892 */
3a59c94c 893 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|LC", ip, idx - 1,
1da177e4 894 XFS_DATA_FORK);
4eea22f0 895 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4 896 LEFT.br_blockcount + PREV.br_blockcount);
3a59c94c 897 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|LC", ip, idx - 1,
1da177e4
LT
898 XFS_DATA_FORK);
899 ip->i_df.if_lastex = idx - 1;
3a59c94c 900 XFS_BMAP_TRACE_DELETE("LF|RF|LC", ip, idx, 1, XFS_DATA_FORK);
4eea22f0 901 xfs_iext_remove(ifp, idx, 1);
1da177e4
LT
902 if (cur == NULL)
903 rval = XFS_ILOG_DEXT;
904 else {
905 rval = 0;
906 if ((error = xfs_bmbt_lookup_eq(cur, LEFT.br_startoff,
907 LEFT.br_startblock, LEFT.br_blockcount,
908 &i)))
909 goto done;
6bd8fc8a 910 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
911 if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
912 LEFT.br_startblock,
913 LEFT.br_blockcount +
914 PREV.br_blockcount, LEFT.br_state)))
915 goto done;
916 }
917 *dnew = 0;
3e57ecf6
OW
918 /* DELTA: Two in-core extents are replaced by one. */
919 temp = LEFT.br_startoff;
920 temp2 = LEFT.br_blockcount +
921 PREV.br_blockcount;
1da177e4
LT
922 break;
923
924 case MASK3(LEFT_FILLING, RIGHT_FILLING, RIGHT_CONTIG):
925 /*
926 * Filling in all of a previously delayed allocation extent.
927 * The right neighbor is contiguous, the left is not.
928 */
3a59c94c 929 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|RC", ip, idx, XFS_DATA_FORK);
1da177e4
LT
930 xfs_bmbt_set_startblock(ep, new->br_startblock);
931 xfs_bmbt_set_blockcount(ep,
932 PREV.br_blockcount + RIGHT.br_blockcount);
3a59c94c 933 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|RC", ip, idx, XFS_DATA_FORK);
1da177e4 934 ip->i_df.if_lastex = idx;
3a59c94c 935 XFS_BMAP_TRACE_DELETE("LF|RF|RC", ip, idx + 1, 1, XFS_DATA_FORK);
4eea22f0 936 xfs_iext_remove(ifp, idx + 1, 1);
1da177e4
LT
937 if (cur == NULL)
938 rval = XFS_ILOG_DEXT;
939 else {
940 rval = 0;
941 if ((error = xfs_bmbt_lookup_eq(cur, RIGHT.br_startoff,
942 RIGHT.br_startblock,
943 RIGHT.br_blockcount, &i)))
944 goto done;
6bd8fc8a 945 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
946 if ((error = xfs_bmbt_update(cur, PREV.br_startoff,
947 new->br_startblock,
948 PREV.br_blockcount +
949 RIGHT.br_blockcount, PREV.br_state)))
950 goto done;
951 }
952 *dnew = 0;
3e57ecf6
OW
953 /* DELTA: Two in-core extents are replaced by one. */
954 temp = PREV.br_startoff;
955 temp2 = PREV.br_blockcount +
956 RIGHT.br_blockcount;
1da177e4
LT
957 break;
958
959 case MASK2(LEFT_FILLING, RIGHT_FILLING):
960 /*
961 * Filling in all of a previously delayed allocation extent.
962 * Neither the left nor right neighbors are contiguous with
963 * the new one.
964 */
3a59c94c 965 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF", ip, idx, XFS_DATA_FORK);
1da177e4 966 xfs_bmbt_set_startblock(ep, new->br_startblock);
3a59c94c 967 XFS_BMAP_TRACE_POST_UPDATE("LF|RF", ip, idx, XFS_DATA_FORK);
1da177e4
LT
968 ip->i_df.if_lastex = idx;
969 ip->i_d.di_nextents++;
970 if (cur == NULL)
971 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
972 else {
973 rval = XFS_ILOG_CORE;
974 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
975 new->br_startblock, new->br_blockcount,
976 &i)))
977 goto done;
6bd8fc8a 978 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 979 cur->bc_rec.b.br_state = XFS_EXT_NORM;
4b22a571 980 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 981 goto done;
6bd8fc8a 982 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
983 }
984 *dnew = 0;
3e57ecf6
OW
985 /* DELTA: The in-core extent described by new changed type. */
986 temp = new->br_startoff;
987 temp2 = new->br_blockcount;
1da177e4
LT
988 break;
989
990 case MASK2(LEFT_FILLING, LEFT_CONTIG):
991 /*
992 * Filling in the first part of a previous delayed allocation.
993 * The left neighbor is contiguous.
994 */
3a59c94c 995 XFS_BMAP_TRACE_PRE_UPDATE("LF|LC", ip, idx - 1, XFS_DATA_FORK);
4eea22f0 996 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4
LT
997 LEFT.br_blockcount + new->br_blockcount);
998 xfs_bmbt_set_startoff(ep,
999 PREV.br_startoff + new->br_blockcount);
3a59c94c 1000 XFS_BMAP_TRACE_POST_UPDATE("LF|LC", ip, idx - 1, XFS_DATA_FORK);
1da177e4 1001 temp = PREV.br_blockcount - new->br_blockcount;
3a59c94c 1002 XFS_BMAP_TRACE_PRE_UPDATE("LF|LC", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1003 xfs_bmbt_set_blockcount(ep, temp);
1004 ip->i_df.if_lastex = idx - 1;
1005 if (cur == NULL)
1006 rval = XFS_ILOG_DEXT;
1007 else {
1008 rval = 0;
1009 if ((error = xfs_bmbt_lookup_eq(cur, LEFT.br_startoff,
1010 LEFT.br_startblock, LEFT.br_blockcount,
1011 &i)))
1012 goto done;
6bd8fc8a 1013 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1014 if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
1015 LEFT.br_startblock,
1016 LEFT.br_blockcount +
1017 new->br_blockcount,
1018 LEFT.br_state)))
1019 goto done;
1020 }
1021 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
9d87c319
ES
1022 startblockval(PREV.br_startblock));
1023 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 1024 XFS_BMAP_TRACE_POST_UPDATE("LF|LC", ip, idx, XFS_DATA_FORK);
1da177e4 1025 *dnew = temp;
3e57ecf6
OW
1026 /* DELTA: The boundary between two in-core extents moved. */
1027 temp = LEFT.br_startoff;
1028 temp2 = LEFT.br_blockcount +
1029 PREV.br_blockcount;
1da177e4
LT
1030 break;
1031
1032 case MASK(LEFT_FILLING):
1033 /*
1034 * Filling in the first part of a previous delayed allocation.
1035 * The left neighbor is not contiguous.
1036 */
3a59c94c 1037 XFS_BMAP_TRACE_PRE_UPDATE("LF", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1038 xfs_bmbt_set_startoff(ep, new_endoff);
1039 temp = PREV.br_blockcount - new->br_blockcount;
1040 xfs_bmbt_set_blockcount(ep, temp);
3a59c94c 1041 XFS_BMAP_TRACE_INSERT("LF", ip, idx, 1, new, NULL,
1da177e4 1042 XFS_DATA_FORK);
4eea22f0 1043 xfs_iext_insert(ifp, idx, 1, new);
1da177e4
LT
1044 ip->i_df.if_lastex = idx;
1045 ip->i_d.di_nextents++;
1046 if (cur == NULL)
1047 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1048 else {
1049 rval = XFS_ILOG_CORE;
1050 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1051 new->br_startblock, new->br_blockcount,
1052 &i)))
1053 goto done;
6bd8fc8a 1054 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 1055 cur->bc_rec.b.br_state = XFS_EXT_NORM;
4b22a571 1056 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1057 goto done;
6bd8fc8a 1058 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1059 }
1060 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
1061 ip->i_d.di_nextents > ip->i_df.if_ext_max) {
1062 error = xfs_bmap_extents_to_btree(ip->i_transp, ip,
1063 first, flist, &cur, 1, &tmp_rval,
1064 XFS_DATA_FORK);
1065 rval |= tmp_rval;
1066 if (error)
1067 goto done;
1068 }
1069 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
9d87c319 1070 startblockval(PREV.br_startblock) -
1da177e4 1071 (cur ? cur->bc_private.b.allocated : 0));
4eea22f0 1072 ep = xfs_iext_get_ext(ifp, idx + 1);
9d87c319 1073 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 1074 XFS_BMAP_TRACE_POST_UPDATE("LF", ip, idx + 1, XFS_DATA_FORK);
1da177e4 1075 *dnew = temp;
3e57ecf6
OW
1076 /* DELTA: One in-core extent is split in two. */
1077 temp = PREV.br_startoff;
1078 temp2 = PREV.br_blockcount;
1da177e4
LT
1079 break;
1080
1081 case MASK2(RIGHT_FILLING, RIGHT_CONTIG):
1082 /*
1083 * Filling in the last part of a previous delayed allocation.
1084 * The right neighbor is contiguous with the new allocation.
1085 */
1086 temp = PREV.br_blockcount - new->br_blockcount;
3a59c94c
ES
1087 XFS_BMAP_TRACE_PRE_UPDATE("RF|RC", ip, idx, XFS_DATA_FORK);
1088 XFS_BMAP_TRACE_PRE_UPDATE("RF|RC", ip, idx + 1, XFS_DATA_FORK);
1da177e4 1089 xfs_bmbt_set_blockcount(ep, temp);
4eea22f0
MK
1090 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp, idx + 1),
1091 new->br_startoff, new->br_startblock,
1da177e4
LT
1092 new->br_blockcount + RIGHT.br_blockcount,
1093 RIGHT.br_state);
3a59c94c 1094 XFS_BMAP_TRACE_POST_UPDATE("RF|RC", ip, idx + 1, XFS_DATA_FORK);
1da177e4
LT
1095 ip->i_df.if_lastex = idx + 1;
1096 if (cur == NULL)
1097 rval = XFS_ILOG_DEXT;
1098 else {
1099 rval = 0;
1100 if ((error = xfs_bmbt_lookup_eq(cur, RIGHT.br_startoff,
1101 RIGHT.br_startblock,
1102 RIGHT.br_blockcount, &i)))
1103 goto done;
6bd8fc8a 1104 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1105 if ((error = xfs_bmbt_update(cur, new->br_startoff,
1106 new->br_startblock,
1107 new->br_blockcount +
1108 RIGHT.br_blockcount,
1109 RIGHT.br_state)))
1110 goto done;
1111 }
1112 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
9d87c319
ES
1113 startblockval(PREV.br_startblock));
1114 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 1115 XFS_BMAP_TRACE_POST_UPDATE("RF|RC", ip, idx, XFS_DATA_FORK);
1da177e4 1116 *dnew = temp;
3e57ecf6
OW
1117 /* DELTA: The boundary between two in-core extents moved. */
1118 temp = PREV.br_startoff;
1119 temp2 = PREV.br_blockcount +
1120 RIGHT.br_blockcount;
1da177e4
LT
1121 break;
1122
1123 case MASK(RIGHT_FILLING):
1124 /*
1125 * Filling in the last part of a previous delayed allocation.
1126 * The right neighbor is not contiguous.
1127 */
1128 temp = PREV.br_blockcount - new->br_blockcount;
3a59c94c 1129 XFS_BMAP_TRACE_PRE_UPDATE("RF", ip, idx, XFS_DATA_FORK);
1da177e4 1130 xfs_bmbt_set_blockcount(ep, temp);
3a59c94c
ES
1131 XFS_BMAP_TRACE_INSERT("RF", ip, idx + 1, 1, new, NULL,
1132 XFS_DATA_FORK);
4eea22f0 1133 xfs_iext_insert(ifp, idx + 1, 1, new);
1da177e4
LT
1134 ip->i_df.if_lastex = idx + 1;
1135 ip->i_d.di_nextents++;
1136 if (cur == NULL)
1137 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1138 else {
1139 rval = XFS_ILOG_CORE;
1140 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1141 new->br_startblock, new->br_blockcount,
1142 &i)))
1143 goto done;
6bd8fc8a 1144 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 1145 cur->bc_rec.b.br_state = XFS_EXT_NORM;
4b22a571 1146 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1147 goto done;
6bd8fc8a 1148 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1149 }
1150 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
1151 ip->i_d.di_nextents > ip->i_df.if_ext_max) {
1152 error = xfs_bmap_extents_to_btree(ip->i_transp, ip,
1153 first, flist, &cur, 1, &tmp_rval,
1154 XFS_DATA_FORK);
1155 rval |= tmp_rval;
1156 if (error)
1157 goto done;
1158 }
1159 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
9d87c319 1160 startblockval(PREV.br_startblock) -
1da177e4 1161 (cur ? cur->bc_private.b.allocated : 0));
4eea22f0 1162 ep = xfs_iext_get_ext(ifp, idx);
9d87c319 1163 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 1164 XFS_BMAP_TRACE_POST_UPDATE("RF", ip, idx, XFS_DATA_FORK);
1da177e4 1165 *dnew = temp;
3e57ecf6
OW
1166 /* DELTA: One in-core extent is split in two. */
1167 temp = PREV.br_startoff;
1168 temp2 = PREV.br_blockcount;
1da177e4
LT
1169 break;
1170
1171 case 0:
1172 /*
1173 * Filling in the middle part of a previous delayed allocation.
1174 * Contiguity is impossible here.
1175 * This case is avoided almost all the time.
1176 */
1177 temp = new->br_startoff - PREV.br_startoff;
3a59c94c 1178 XFS_BMAP_TRACE_PRE_UPDATE("0", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1179 xfs_bmbt_set_blockcount(ep, temp);
1180 r[0] = *new;
d2133717
LM
1181 r[1].br_state = PREV.br_state;
1182 r[1].br_startblock = 0;
1da177e4
LT
1183 r[1].br_startoff = new_endoff;
1184 temp2 = PREV.br_startoff + PREV.br_blockcount - new_endoff;
1185 r[1].br_blockcount = temp2;
3a59c94c 1186 XFS_BMAP_TRACE_INSERT("0", ip, idx + 1, 2, &r[0], &r[1],
1da177e4 1187 XFS_DATA_FORK);
4eea22f0 1188 xfs_iext_insert(ifp, idx + 1, 2, &r[0]);
1da177e4
LT
1189 ip->i_df.if_lastex = idx + 1;
1190 ip->i_d.di_nextents++;
1191 if (cur == NULL)
1192 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1193 else {
1194 rval = XFS_ILOG_CORE;
1195 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1196 new->br_startblock, new->br_blockcount,
1197 &i)))
1198 goto done;
6bd8fc8a 1199 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 1200 cur->bc_rec.b.br_state = XFS_EXT_NORM;
4b22a571 1201 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1202 goto done;
6bd8fc8a 1203 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1204 }
1205 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
1206 ip->i_d.di_nextents > ip->i_df.if_ext_max) {
1207 error = xfs_bmap_extents_to_btree(ip->i_transp, ip,
1208 first, flist, &cur, 1, &tmp_rval,
1209 XFS_DATA_FORK);
1210 rval |= tmp_rval;
1211 if (error)
1212 goto done;
1213 }
1214 temp = xfs_bmap_worst_indlen(ip, temp);
1215 temp2 = xfs_bmap_worst_indlen(ip, temp2);
9d87c319 1216 diff = (int)(temp + temp2 - startblockval(PREV.br_startblock) -
1da177e4
LT
1217 (cur ? cur->bc_private.b.allocated : 0));
1218 if (diff > 0 &&
20f4ebf2 1219 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd)) {
1da177e4
LT
1220 /*
1221 * Ick gross gag me with a spoon.
1222 */
1223 ASSERT(0); /* want to see if this ever happens! */
1224 while (diff > 0) {
1225 if (temp) {
1226 temp--;
1227 diff--;
1228 if (!diff ||
1229 !xfs_mod_incore_sb(ip->i_mount,
20f4ebf2 1230 XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd))
1da177e4
LT
1231 break;
1232 }
1233 if (temp2) {
1234 temp2--;
1235 diff--;
1236 if (!diff ||
1237 !xfs_mod_incore_sb(ip->i_mount,
20f4ebf2 1238 XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd))
1da177e4
LT
1239 break;
1240 }
1241 }
1242 }
4eea22f0 1243 ep = xfs_iext_get_ext(ifp, idx);
9d87c319 1244 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c
ES
1245 XFS_BMAP_TRACE_POST_UPDATE("0", ip, idx, XFS_DATA_FORK);
1246 XFS_BMAP_TRACE_PRE_UPDATE("0", ip, idx + 2, XFS_DATA_FORK);
4eea22f0 1247 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp, idx + 2),
9d87c319 1248 nullstartblock((int)temp2));
3a59c94c 1249 XFS_BMAP_TRACE_POST_UPDATE("0", ip, idx + 2, XFS_DATA_FORK);
1da177e4 1250 *dnew = temp + temp2;
3e57ecf6
OW
1251 /* DELTA: One in-core extent is split in three. */
1252 temp = PREV.br_startoff;
1253 temp2 = PREV.br_blockcount;
1da177e4
LT
1254 break;
1255
1256 case MASK3(LEFT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
1257 case MASK3(RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
1258 case MASK2(LEFT_FILLING, RIGHT_CONTIG):
1259 case MASK2(RIGHT_FILLING, LEFT_CONTIG):
1260 case MASK2(LEFT_CONTIG, RIGHT_CONTIG):
1261 case MASK(LEFT_CONTIG):
1262 case MASK(RIGHT_CONTIG):
1263 /*
1264 * These cases are all impossible.
1265 */
1266 ASSERT(0);
1267 }
1268 *curp = cur;
3e57ecf6
OW
1269 if (delta) {
1270 temp2 += temp;
1271 if (delta->xed_startoff > temp)
1272 delta->xed_startoff = temp;
1273 if (delta->xed_blockcount < temp2)
1274 delta->xed_blockcount = temp2;
1275 }
1da177e4
LT
1276done:
1277 *logflagsp = rval;
1278 return error;
1279#undef LEFT
1280#undef RIGHT
1281#undef PREV
1282#undef MASK
1283#undef MASK2
1284#undef MASK3
1285#undef MASK4
1286#undef STATE_SET
1287#undef STATE_TEST
1288#undef STATE_SET_TEST
1289#undef SWITCH_STATE
1290}
1291
1292/*
1293 * Called by xfs_bmap_add_extent to handle cases converting an unwritten
1294 * allocation to a real allocation or vice versa.
1295 */
1296STATIC int /* error */
1297xfs_bmap_add_extent_unwritten_real(
1298 xfs_inode_t *ip, /* incore inode pointer */
1299 xfs_extnum_t idx, /* extent number to update/insert */
1300 xfs_btree_cur_t **curp, /* if *curp is null, not a btree */
4eea22f0 1301 xfs_bmbt_irec_t *new, /* new data to add to file extents */
3e57ecf6
OW
1302 int *logflagsp, /* inode logging flags */
1303 xfs_extdelta_t *delta) /* Change made to incore extents */
1da177e4 1304{
1da177e4 1305 xfs_btree_cur_t *cur; /* btree cursor */
a6f64d4a 1306 xfs_bmbt_rec_host_t *ep; /* extent entry for idx */
1da177e4 1307 int error; /* error return value */
1da177e4 1308 int i; /* temp state */
4eea22f0 1309 xfs_ifork_t *ifp; /* inode fork pointer */
1da177e4
LT
1310 xfs_fileoff_t new_endoff; /* end offset of new entry */
1311 xfs_exntst_t newext; /* new extent state */
1312 xfs_exntst_t oldext; /* old extent state */
1313 xfs_bmbt_irec_t r[3]; /* neighbor extent entries */
1314 /* left is 0, right is 1, prev is 2 */
1315 int rval=0; /* return value (logging flags) */
1316 int state = 0;/* state bits, accessed thru macros */
3e57ecf6
OW
1317 xfs_filblks_t temp=0;
1318 xfs_filblks_t temp2=0;
1da177e4
LT
1319 enum { /* bit number definitions for state */
1320 LEFT_CONTIG, RIGHT_CONTIG,
1321 LEFT_FILLING, RIGHT_FILLING,
1322 LEFT_DELAY, RIGHT_DELAY,
1323 LEFT_VALID, RIGHT_VALID
1324 };
1325
1326#define LEFT r[0]
1327#define RIGHT r[1]
1328#define PREV r[2]
1329#define MASK(b) (1 << (b))
1330#define MASK2(a,b) (MASK(a) | MASK(b))
1331#define MASK3(a,b,c) (MASK2(a,b) | MASK(c))
1332#define MASK4(a,b,c,d) (MASK3(a,b,c) | MASK(d))
1333#define STATE_SET(b,v) ((v) ? (state |= MASK(b)) : (state &= ~MASK(b)))
1334#define STATE_TEST(b) (state & MASK(b))
1335#define STATE_SET_TEST(b,v) ((v) ? ((state |= MASK(b)), 1) : \
1336 ((state &= ~MASK(b)), 0))
1337#define SWITCH_STATE \
1338 (state & MASK4(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG))
1339
1340 /*
1341 * Set up a bunch of variables to make the tests simpler.
1342 */
1343 error = 0;
1344 cur = *curp;
4eea22f0
MK
1345 ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
1346 ep = xfs_iext_get_ext(ifp, idx);
1da177e4
LT
1347 xfs_bmbt_get_all(ep, &PREV);
1348 newext = new->br_state;
1349 oldext = (newext == XFS_EXT_UNWRITTEN) ?
1350 XFS_EXT_NORM : XFS_EXT_UNWRITTEN;
1351 ASSERT(PREV.br_state == oldext);
1352 new_endoff = new->br_startoff + new->br_blockcount;
1353 ASSERT(PREV.br_startoff <= new->br_startoff);
1354 ASSERT(PREV.br_startoff + PREV.br_blockcount >= new_endoff);
1355 /*
1356 * Set flags determining what part of the previous oldext allocation
1357 * extent is being replaced by a newext allocation.
1358 */
1359 STATE_SET(LEFT_FILLING, PREV.br_startoff == new->br_startoff);
1360 STATE_SET(RIGHT_FILLING,
1361 PREV.br_startoff + PREV.br_blockcount == new_endoff);
1362 /*
1363 * Check and set flags if this segment has a left neighbor.
1364 * Don't set contiguous if the combined extent would be too large.
1365 */
1366 if (STATE_SET_TEST(LEFT_VALID, idx > 0)) {
4eea22f0 1367 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx - 1), &LEFT);
9d87c319 1368 STATE_SET(LEFT_DELAY, isnullstartblock(LEFT.br_startblock));
1da177e4
LT
1369 }
1370 STATE_SET(LEFT_CONTIG,
1371 STATE_TEST(LEFT_VALID) && !STATE_TEST(LEFT_DELAY) &&
1372 LEFT.br_startoff + LEFT.br_blockcount == new->br_startoff &&
1373 LEFT.br_startblock + LEFT.br_blockcount == new->br_startblock &&
1374 LEFT.br_state == newext &&
1375 LEFT.br_blockcount + new->br_blockcount <= MAXEXTLEN);
1376 /*
1377 * Check and set flags if this segment has a right neighbor.
1378 * Don't set contiguous if the combined extent would be too large.
1379 * Also check for all-three-contiguous being too large.
1380 */
1381 if (STATE_SET_TEST(RIGHT_VALID,
1382 idx <
1383 ip->i_df.if_bytes / (uint)sizeof(xfs_bmbt_rec_t) - 1)) {
4eea22f0 1384 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx + 1), &RIGHT);
9d87c319 1385 STATE_SET(RIGHT_DELAY, isnullstartblock(RIGHT.br_startblock));
1da177e4
LT
1386 }
1387 STATE_SET(RIGHT_CONTIG,
1388 STATE_TEST(RIGHT_VALID) && !STATE_TEST(RIGHT_DELAY) &&
1389 new_endoff == RIGHT.br_startoff &&
1390 new->br_startblock + new->br_blockcount ==
1391 RIGHT.br_startblock &&
1392 newext == RIGHT.br_state &&
1393 new->br_blockcount + RIGHT.br_blockcount <= MAXEXTLEN &&
1394 ((state & MASK3(LEFT_CONTIG, LEFT_FILLING, RIGHT_FILLING)) !=
1395 MASK3(LEFT_CONTIG, LEFT_FILLING, RIGHT_FILLING) ||
1396 LEFT.br_blockcount + new->br_blockcount + RIGHT.br_blockcount
1397 <= MAXEXTLEN));
1398 /*
1399 * Switch out based on the FILLING and CONTIG state bits.
1400 */
1401 switch (SWITCH_STATE) {
1402
1403 case MASK4(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
1404 /*
1405 * Setting all of a previous oldext extent to newext.
1406 * The left and right neighbors are both contiguous with new.
1407 */
3a59c94c 1408 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|LC|RC", ip, idx - 1,
1da177e4 1409 XFS_DATA_FORK);
4eea22f0 1410 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4
LT
1411 LEFT.br_blockcount + PREV.br_blockcount +
1412 RIGHT.br_blockcount);
3a59c94c 1413 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|LC|RC", ip, idx - 1,
1da177e4 1414 XFS_DATA_FORK);
3a59c94c 1415 XFS_BMAP_TRACE_DELETE("LF|RF|LC|RC", ip, idx, 2, XFS_DATA_FORK);
4eea22f0 1416 xfs_iext_remove(ifp, idx, 2);
1da177e4
LT
1417 ip->i_df.if_lastex = idx - 1;
1418 ip->i_d.di_nextents -= 2;
1419 if (cur == NULL)
1420 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1421 else {
1422 rval = XFS_ILOG_CORE;
1423 if ((error = xfs_bmbt_lookup_eq(cur, RIGHT.br_startoff,
1424 RIGHT.br_startblock,
1425 RIGHT.br_blockcount, &i)))
1426 goto done;
6bd8fc8a 1427 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 1428 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 1429 goto done;
6bd8fc8a 1430 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 1431 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4 1432 goto done;
6bd8fc8a 1433 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 1434 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 1435 goto done;
6bd8fc8a 1436 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 1437 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4 1438 goto done;
6bd8fc8a 1439 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1440 if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
1441 LEFT.br_startblock,
1442 LEFT.br_blockcount + PREV.br_blockcount +
1443 RIGHT.br_blockcount, LEFT.br_state)))
1444 goto done;
1445 }
3e57ecf6
OW
1446 /* DELTA: Three in-core extents are replaced by one. */
1447 temp = LEFT.br_startoff;
1448 temp2 = LEFT.br_blockcount +
1449 PREV.br_blockcount +
1450 RIGHT.br_blockcount;
1da177e4
LT
1451 break;
1452
1453 case MASK3(LEFT_FILLING, RIGHT_FILLING, LEFT_CONTIG):
1454 /*
1455 * Setting all of a previous oldext extent to newext.
1456 * The left neighbor is contiguous, the right is not.
1457 */
3a59c94c 1458 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|LC", ip, idx - 1,
1da177e4 1459 XFS_DATA_FORK);
4eea22f0 1460 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4 1461 LEFT.br_blockcount + PREV.br_blockcount);
3a59c94c 1462 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|LC", ip, idx - 1,
1da177e4
LT
1463 XFS_DATA_FORK);
1464 ip->i_df.if_lastex = idx - 1;
3a59c94c 1465 XFS_BMAP_TRACE_DELETE("LF|RF|LC", ip, idx, 1, XFS_DATA_FORK);
4eea22f0 1466 xfs_iext_remove(ifp, idx, 1);
1da177e4
LT
1467 ip->i_d.di_nextents--;
1468 if (cur == NULL)
1469 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1470 else {
1471 rval = XFS_ILOG_CORE;
1472 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1473 PREV.br_startblock, PREV.br_blockcount,
1474 &i)))
1475 goto done;
6bd8fc8a 1476 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 1477 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 1478 goto done;
6bd8fc8a 1479 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 1480 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4 1481 goto done;
6bd8fc8a 1482 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1483 if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
1484 LEFT.br_startblock,
1485 LEFT.br_blockcount + PREV.br_blockcount,
1486 LEFT.br_state)))
1487 goto done;
1488 }
3e57ecf6
OW
1489 /* DELTA: Two in-core extents are replaced by one. */
1490 temp = LEFT.br_startoff;
1491 temp2 = LEFT.br_blockcount +
1492 PREV.br_blockcount;
1da177e4
LT
1493 break;
1494
1495 case MASK3(LEFT_FILLING, RIGHT_FILLING, RIGHT_CONTIG):
1496 /*
1497 * Setting all of a previous oldext extent to newext.
1498 * The right neighbor is contiguous, the left is not.
1499 */
3a59c94c 1500 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF|RC", ip, idx,
1da177e4
LT
1501 XFS_DATA_FORK);
1502 xfs_bmbt_set_blockcount(ep,
1503 PREV.br_blockcount + RIGHT.br_blockcount);
1504 xfs_bmbt_set_state(ep, newext);
3a59c94c 1505 XFS_BMAP_TRACE_POST_UPDATE("LF|RF|RC", ip, idx,
1da177e4
LT
1506 XFS_DATA_FORK);
1507 ip->i_df.if_lastex = idx;
3a59c94c 1508 XFS_BMAP_TRACE_DELETE("LF|RF|RC", ip, idx + 1, 1, XFS_DATA_FORK);
4eea22f0 1509 xfs_iext_remove(ifp, idx + 1, 1);
1da177e4
LT
1510 ip->i_d.di_nextents--;
1511 if (cur == NULL)
1512 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1513 else {
1514 rval = XFS_ILOG_CORE;
1515 if ((error = xfs_bmbt_lookup_eq(cur, RIGHT.br_startoff,
1516 RIGHT.br_startblock,
1517 RIGHT.br_blockcount, &i)))
1518 goto done;
6bd8fc8a 1519 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 1520 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 1521 goto done;
6bd8fc8a 1522 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 1523 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4 1524 goto done;
6bd8fc8a 1525 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1526 if ((error = xfs_bmbt_update(cur, new->br_startoff,
1527 new->br_startblock,
1528 new->br_blockcount + RIGHT.br_blockcount,
1529 newext)))
1530 goto done;
1531 }
3e57ecf6
OW
1532 /* DELTA: Two in-core extents are replaced by one. */
1533 temp = PREV.br_startoff;
1534 temp2 = PREV.br_blockcount +
1535 RIGHT.br_blockcount;
1da177e4
LT
1536 break;
1537
1538 case MASK2(LEFT_FILLING, RIGHT_FILLING):
1539 /*
1540 * Setting all of a previous oldext extent to newext.
1541 * Neither the left nor right neighbors are contiguous with
1542 * the new one.
1543 */
3a59c94c 1544 XFS_BMAP_TRACE_PRE_UPDATE("LF|RF", ip, idx,
1da177e4
LT
1545 XFS_DATA_FORK);
1546 xfs_bmbt_set_state(ep, newext);
3a59c94c 1547 XFS_BMAP_TRACE_POST_UPDATE("LF|RF", ip, idx,
1da177e4
LT
1548 XFS_DATA_FORK);
1549 ip->i_df.if_lastex = idx;
1550 if (cur == NULL)
1551 rval = XFS_ILOG_DEXT;
1552 else {
1553 rval = 0;
1554 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1555 new->br_startblock, new->br_blockcount,
1556 &i)))
1557 goto done;
6bd8fc8a 1558 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1559 if ((error = xfs_bmbt_update(cur, new->br_startoff,
1560 new->br_startblock, new->br_blockcount,
1561 newext)))
1562 goto done;
1563 }
3e57ecf6
OW
1564 /* DELTA: The in-core extent described by new changed type. */
1565 temp = new->br_startoff;
1566 temp2 = new->br_blockcount;
1da177e4
LT
1567 break;
1568
1569 case MASK2(LEFT_FILLING, LEFT_CONTIG):
1570 /*
1571 * Setting the first part of a previous oldext extent to newext.
1572 * The left neighbor is contiguous.
1573 */
3a59c94c 1574 XFS_BMAP_TRACE_PRE_UPDATE("LF|LC", ip, idx - 1,
1da177e4 1575 XFS_DATA_FORK);
4eea22f0 1576 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4
LT
1577 LEFT.br_blockcount + new->br_blockcount);
1578 xfs_bmbt_set_startoff(ep,
1579 PREV.br_startoff + new->br_blockcount);
3a59c94c 1580 XFS_BMAP_TRACE_POST_UPDATE("LF|LC", ip, idx - 1,
1da177e4 1581 XFS_DATA_FORK);
3a59c94c 1582 XFS_BMAP_TRACE_PRE_UPDATE("LF|LC", ip, idx,
1da177e4
LT
1583 XFS_DATA_FORK);
1584 xfs_bmbt_set_startblock(ep,
1585 new->br_startblock + new->br_blockcount);
1586 xfs_bmbt_set_blockcount(ep,
1587 PREV.br_blockcount - new->br_blockcount);
3a59c94c 1588 XFS_BMAP_TRACE_POST_UPDATE("LF|LC", ip, idx,
1da177e4
LT
1589 XFS_DATA_FORK);
1590 ip->i_df.if_lastex = idx - 1;
1591 if (cur == NULL)
1592 rval = XFS_ILOG_DEXT;
1593 else {
1594 rval = 0;
1595 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1596 PREV.br_startblock, PREV.br_blockcount,
1597 &i)))
1598 goto done;
6bd8fc8a 1599 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1600 if ((error = xfs_bmbt_update(cur,
1601 PREV.br_startoff + new->br_blockcount,
1602 PREV.br_startblock + new->br_blockcount,
1603 PREV.br_blockcount - new->br_blockcount,
1604 oldext)))
1605 goto done;
8df4da4a 1606 if ((error = xfs_btree_decrement(cur, 0, &i)))
1da177e4
LT
1607 goto done;
1608 if (xfs_bmbt_update(cur, LEFT.br_startoff,
1609 LEFT.br_startblock,
1610 LEFT.br_blockcount + new->br_blockcount,
1611 LEFT.br_state))
1612 goto done;
1613 }
3e57ecf6
OW
1614 /* DELTA: The boundary between two in-core extents moved. */
1615 temp = LEFT.br_startoff;
1616 temp2 = LEFT.br_blockcount +
1617 PREV.br_blockcount;
1da177e4
LT
1618 break;
1619
1620 case MASK(LEFT_FILLING):
1621 /*
1622 * Setting the first part of a previous oldext extent to newext.
1623 * The left neighbor is not contiguous.
1624 */
3a59c94c 1625 XFS_BMAP_TRACE_PRE_UPDATE("LF", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1626 ASSERT(ep && xfs_bmbt_get_state(ep) == oldext);
1627 xfs_bmbt_set_startoff(ep, new_endoff);
1628 xfs_bmbt_set_blockcount(ep,
1629 PREV.br_blockcount - new->br_blockcount);
1630 xfs_bmbt_set_startblock(ep,
1631 new->br_startblock + new->br_blockcount);
3a59c94c
ES
1632 XFS_BMAP_TRACE_POST_UPDATE("LF", ip, idx, XFS_DATA_FORK);
1633 XFS_BMAP_TRACE_INSERT("LF", ip, idx, 1, new, NULL,
1da177e4 1634 XFS_DATA_FORK);
4eea22f0 1635 xfs_iext_insert(ifp, idx, 1, new);
1da177e4
LT
1636 ip->i_df.if_lastex = idx;
1637 ip->i_d.di_nextents++;
1638 if (cur == NULL)
1639 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1640 else {
1641 rval = XFS_ILOG_CORE;
1642 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1643 PREV.br_startblock, PREV.br_blockcount,
1644 &i)))
1645 goto done;
6bd8fc8a 1646 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1647 if ((error = xfs_bmbt_update(cur,
1648 PREV.br_startoff + new->br_blockcount,
1649 PREV.br_startblock + new->br_blockcount,
1650 PREV.br_blockcount - new->br_blockcount,
1651 oldext)))
1652 goto done;
1653 cur->bc_rec.b = *new;
4b22a571 1654 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1655 goto done;
6bd8fc8a 1656 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4 1657 }
3e57ecf6
OW
1658 /* DELTA: One in-core extent is split in two. */
1659 temp = PREV.br_startoff;
1660 temp2 = PREV.br_blockcount;
1da177e4
LT
1661 break;
1662
1663 case MASK2(RIGHT_FILLING, RIGHT_CONTIG):
1664 /*
1665 * Setting the last part of a previous oldext extent to newext.
1666 * The right neighbor is contiguous with the new allocation.
1667 */
3a59c94c 1668 XFS_BMAP_TRACE_PRE_UPDATE("RF|RC", ip, idx,
1da177e4 1669 XFS_DATA_FORK);
3a59c94c 1670 XFS_BMAP_TRACE_PRE_UPDATE("RF|RC", ip, idx + 1,
1da177e4
LT
1671 XFS_DATA_FORK);
1672 xfs_bmbt_set_blockcount(ep,
1673 PREV.br_blockcount - new->br_blockcount);
3a59c94c 1674 XFS_BMAP_TRACE_POST_UPDATE("RF|RC", ip, idx,
1da177e4 1675 XFS_DATA_FORK);
4eea22f0
MK
1676 xfs_bmbt_set_allf(xfs_iext_get_ext(ifp, idx + 1),
1677 new->br_startoff, new->br_startblock,
1da177e4 1678 new->br_blockcount + RIGHT.br_blockcount, newext);
3a59c94c 1679 XFS_BMAP_TRACE_POST_UPDATE("RF|RC", ip, idx + 1,
1da177e4
LT
1680 XFS_DATA_FORK);
1681 ip->i_df.if_lastex = idx + 1;
1682 if (cur == NULL)
1683 rval = XFS_ILOG_DEXT;
1684 else {
1685 rval = 0;
1686 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1687 PREV.br_startblock,
1688 PREV.br_blockcount, &i)))
1689 goto done;
6bd8fc8a 1690 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1691 if ((error = xfs_bmbt_update(cur, PREV.br_startoff,
1692 PREV.br_startblock,
1693 PREV.br_blockcount - new->br_blockcount,
1694 oldext)))
1695 goto done;
637aa50f 1696 if ((error = xfs_btree_increment(cur, 0, &i)))
1da177e4
LT
1697 goto done;
1698 if ((error = xfs_bmbt_update(cur, new->br_startoff,
1699 new->br_startblock,
1700 new->br_blockcount + RIGHT.br_blockcount,
1701 newext)))
1702 goto done;
1703 }
3e57ecf6
OW
1704 /* DELTA: The boundary between two in-core extents moved. */
1705 temp = PREV.br_startoff;
1706 temp2 = PREV.br_blockcount +
1707 RIGHT.br_blockcount;
1da177e4
LT
1708 break;
1709
1710 case MASK(RIGHT_FILLING):
1711 /*
1712 * Setting the last part of a previous oldext extent to newext.
1713 * The right neighbor is not contiguous.
1714 */
3a59c94c 1715 XFS_BMAP_TRACE_PRE_UPDATE("RF", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1716 xfs_bmbt_set_blockcount(ep,
1717 PREV.br_blockcount - new->br_blockcount);
3a59c94c
ES
1718 XFS_BMAP_TRACE_POST_UPDATE("RF", ip, idx, XFS_DATA_FORK);
1719 XFS_BMAP_TRACE_INSERT("RF", ip, idx + 1, 1, new, NULL,
1720 XFS_DATA_FORK);
4eea22f0 1721 xfs_iext_insert(ifp, idx + 1, 1, new);
1da177e4
LT
1722 ip->i_df.if_lastex = idx + 1;
1723 ip->i_d.di_nextents++;
1724 if (cur == NULL)
1725 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1726 else {
1727 rval = XFS_ILOG_CORE;
1728 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1729 PREV.br_startblock, PREV.br_blockcount,
1730 &i)))
1731 goto done;
6bd8fc8a 1732 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1733 if ((error = xfs_bmbt_update(cur, PREV.br_startoff,
1734 PREV.br_startblock,
1735 PREV.br_blockcount - new->br_blockcount,
1736 oldext)))
1737 goto done;
1738 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1739 new->br_startblock, new->br_blockcount,
1740 &i)))
1741 goto done;
6bd8fc8a 1742 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 1743 cur->bc_rec.b.br_state = XFS_EXT_NORM;
4b22a571 1744 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1745 goto done;
6bd8fc8a 1746 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4 1747 }
3e57ecf6
OW
1748 /* DELTA: One in-core extent is split in two. */
1749 temp = PREV.br_startoff;
1750 temp2 = PREV.br_blockcount;
1da177e4
LT
1751 break;
1752
1753 case 0:
1754 /*
1755 * Setting the middle part of a previous oldext extent to
1756 * newext. Contiguity is impossible here.
1757 * One extent becomes three extents.
1758 */
3a59c94c 1759 XFS_BMAP_TRACE_PRE_UPDATE("0", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1760 xfs_bmbt_set_blockcount(ep,
1761 new->br_startoff - PREV.br_startoff);
3a59c94c 1762 XFS_BMAP_TRACE_POST_UPDATE("0", ip, idx, XFS_DATA_FORK);
1da177e4
LT
1763 r[0] = *new;
1764 r[1].br_startoff = new_endoff;
1765 r[1].br_blockcount =
1766 PREV.br_startoff + PREV.br_blockcount - new_endoff;
1767 r[1].br_startblock = new->br_startblock + new->br_blockcount;
1768 r[1].br_state = oldext;
3a59c94c 1769 XFS_BMAP_TRACE_INSERT("0", ip, idx + 1, 2, &r[0], &r[1],
1da177e4 1770 XFS_DATA_FORK);
4eea22f0 1771 xfs_iext_insert(ifp, idx + 1, 2, &r[0]);
1da177e4
LT
1772 ip->i_df.if_lastex = idx + 1;
1773 ip->i_d.di_nextents += 2;
1774 if (cur == NULL)
1775 rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
1776 else {
1777 rval = XFS_ILOG_CORE;
1778 if ((error = xfs_bmbt_lookup_eq(cur, PREV.br_startoff,
1779 PREV.br_startblock, PREV.br_blockcount,
1780 &i)))
1781 goto done;
6bd8fc8a 1782 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
1783 /* new right extent - oldext */
1784 if ((error = xfs_bmbt_update(cur, r[1].br_startoff,
1785 r[1].br_startblock, r[1].br_blockcount,
1786 r[1].br_state)))
1787 goto done;
1788 /* new left extent - oldext */
1da177e4 1789 cur->bc_rec.b = PREV;
6a617dd2
TS
1790 cur->bc_rec.b.br_blockcount =
1791 new->br_startoff - PREV.br_startoff;
4b22a571 1792 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1793 goto done;
6bd8fc8a 1794 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
ddea2d52
LM
1795 /*
1796 * Reset the cursor to the position of the new extent
1797 * we are about to insert as we can't trust it after
1798 * the previous insert.
1799 */
1800 if ((error = xfs_bmbt_lookup_eq(cur, new->br_startoff,
1801 new->br_startblock, new->br_blockcount,
1802 &i)))
1da177e4 1803 goto done;
ddea2d52 1804 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
1da177e4 1805 /* new middle extent - newext */
ddea2d52 1806 cur->bc_rec.b.br_state = new->br_state;
4b22a571 1807 if ((error = xfs_btree_insert(cur, &i)))
1da177e4 1808 goto done;
6bd8fc8a 1809 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4 1810 }
3e57ecf6
OW
1811 /* DELTA: One in-core extent is split in three. */
1812 temp = PREV.br_startoff;
1813 temp2 = PREV.br_blockcount;
1da177e4
LT
1814 break;
1815
1816 case MASK3(LEFT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
1817 case MASK3(RIGHT_FILLING, LEFT_CONTIG, RIGHT_CONTIG):
1818 case MASK2(LEFT_FILLING, RIGHT_CONTIG):
1819 case MASK2(RIGHT_FILLING, LEFT_CONTIG):
1820 case MASK2(LEFT_CONTIG, RIGHT_CONTIG):
1821 case MASK(LEFT_CONTIG):
1822 case MASK(RIGHT_CONTIG):
1823 /*
1824 * These cases are all impossible.
1825 */
1826 ASSERT(0);
1827 }
1828 *curp = cur;
3e57ecf6
OW
1829 if (delta) {
1830 temp2 += temp;
1831 if (delta->xed_startoff > temp)
1832 delta->xed_startoff = temp;
1833 if (delta->xed_blockcount < temp2)
1834 delta->xed_blockcount = temp2;
1835 }
1da177e4
LT
1836done:
1837 *logflagsp = rval;
1838 return error;
1839#undef LEFT
1840#undef RIGHT
1841#undef PREV
1842#undef MASK
1843#undef MASK2
1844#undef MASK3
1845#undef MASK4
1846#undef STATE_SET
1847#undef STATE_TEST
1848#undef STATE_SET_TEST
1849#undef SWITCH_STATE
1850}
1851
1852/*
1853 * Called by xfs_bmap_add_extent to handle cases converting a hole
1854 * to a delayed allocation.
1855 */
1856/*ARGSUSED*/
1857STATIC int /* error */
1858xfs_bmap_add_extent_hole_delay(
1859 xfs_inode_t *ip, /* incore inode pointer */
1860 xfs_extnum_t idx, /* extent number to update/insert */
4eea22f0 1861 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4 1862 int *logflagsp, /* inode logging flags */
3e57ecf6 1863 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
1864 int rsvd) /* OK to allocate reserved blocks */
1865{
a6f64d4a 1866 xfs_bmbt_rec_host_t *ep; /* extent record for idx */
4eea22f0 1867 xfs_ifork_t *ifp; /* inode fork pointer */
1da177e4
LT
1868 xfs_bmbt_irec_t left; /* left neighbor extent entry */
1869 xfs_filblks_t newlen=0; /* new indirect size */
1870 xfs_filblks_t oldlen=0; /* old indirect size */
1871 xfs_bmbt_irec_t right; /* right neighbor extent entry */
1872 int state; /* state bits, accessed thru macros */
3e57ecf6
OW
1873 xfs_filblks_t temp=0; /* temp for indirect calculations */
1874 xfs_filblks_t temp2=0;
1da177e4
LT
1875 enum { /* bit number definitions for state */
1876 LEFT_CONTIG, RIGHT_CONTIG,
1877 LEFT_DELAY, RIGHT_DELAY,
1878 LEFT_VALID, RIGHT_VALID
1879 };
1880
1881#define MASK(b) (1 << (b))
1882#define MASK2(a,b) (MASK(a) | MASK(b))
1883#define STATE_SET(b,v) ((v) ? (state |= MASK(b)) : (state &= ~MASK(b)))
1884#define STATE_TEST(b) (state & MASK(b))
1885#define STATE_SET_TEST(b,v) ((v) ? ((state |= MASK(b)), 1) : \
1886 ((state &= ~MASK(b)), 0))
1887#define SWITCH_STATE (state & MASK2(LEFT_CONTIG, RIGHT_CONTIG))
1888
4eea22f0
MK
1889 ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
1890 ep = xfs_iext_get_ext(ifp, idx);
1da177e4 1891 state = 0;
9d87c319 1892 ASSERT(isnullstartblock(new->br_startblock));
1da177e4
LT
1893 /*
1894 * Check and set flags if this segment has a left neighbor
1895 */
1896 if (STATE_SET_TEST(LEFT_VALID, idx > 0)) {
4eea22f0 1897 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx - 1), &left);
9d87c319 1898 STATE_SET(LEFT_DELAY, isnullstartblock(left.br_startblock));
1da177e4
LT
1899 }
1900 /*
1901 * Check and set flags if the current (right) segment exists.
1902 * If it doesn't exist, we're converting the hole at end-of-file.
1903 */
1904 if (STATE_SET_TEST(RIGHT_VALID,
1905 idx <
1906 ip->i_df.if_bytes / (uint)sizeof(xfs_bmbt_rec_t))) {
1907 xfs_bmbt_get_all(ep, &right);
9d87c319 1908 STATE_SET(RIGHT_DELAY, isnullstartblock(right.br_startblock));
1da177e4
LT
1909 }
1910 /*
1911 * Set contiguity flags on the left and right neighbors.
1912 * Don't let extents get too large, even if the pieces are contiguous.
1913 */
1914 STATE_SET(LEFT_CONTIG,
1915 STATE_TEST(LEFT_VALID) && STATE_TEST(LEFT_DELAY) &&
1916 left.br_startoff + left.br_blockcount == new->br_startoff &&
1917 left.br_blockcount + new->br_blockcount <= MAXEXTLEN);
1918 STATE_SET(RIGHT_CONTIG,
1919 STATE_TEST(RIGHT_VALID) && STATE_TEST(RIGHT_DELAY) &&
1920 new->br_startoff + new->br_blockcount == right.br_startoff &&
1921 new->br_blockcount + right.br_blockcount <= MAXEXTLEN &&
1922 (!STATE_TEST(LEFT_CONTIG) ||
1923 (left.br_blockcount + new->br_blockcount +
1924 right.br_blockcount <= MAXEXTLEN)));
1925 /*
1926 * Switch out based on the contiguity flags.
1927 */
1928 switch (SWITCH_STATE) {
1929
1930 case MASK2(LEFT_CONTIG, RIGHT_CONTIG):
1931 /*
1932 * New allocation is contiguous with delayed allocations
1933 * on the left and on the right.
4eea22f0 1934 * Merge all three into a single extent record.
1da177e4
LT
1935 */
1936 temp = left.br_blockcount + new->br_blockcount +
1937 right.br_blockcount;
3a59c94c 1938 XFS_BMAP_TRACE_PRE_UPDATE("LC|RC", ip, idx - 1,
1da177e4 1939 XFS_DATA_FORK);
4eea22f0 1940 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1), temp);
9d87c319
ES
1941 oldlen = startblockval(left.br_startblock) +
1942 startblockval(new->br_startblock) +
1943 startblockval(right.br_startblock);
1da177e4 1944 newlen = xfs_bmap_worst_indlen(ip, temp);
4eea22f0 1945 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp, idx - 1),
9d87c319 1946 nullstartblock((int)newlen));
3a59c94c 1947 XFS_BMAP_TRACE_POST_UPDATE("LC|RC", ip, idx - 1,
1da177e4 1948 XFS_DATA_FORK);
3a59c94c 1949 XFS_BMAP_TRACE_DELETE("LC|RC", ip, idx, 1, XFS_DATA_FORK);
4eea22f0 1950 xfs_iext_remove(ifp, idx, 1);
1da177e4 1951 ip->i_df.if_lastex = idx - 1;
3e57ecf6
OW
1952 /* DELTA: Two in-core extents were replaced by one. */
1953 temp2 = temp;
1954 temp = left.br_startoff;
1da177e4
LT
1955 break;
1956
1957 case MASK(LEFT_CONTIG):
1958 /*
1959 * New allocation is contiguous with a delayed allocation
1960 * on the left.
1961 * Merge the new allocation with the left neighbor.
1962 */
1963 temp = left.br_blockcount + new->br_blockcount;
3a59c94c 1964 XFS_BMAP_TRACE_PRE_UPDATE("LC", ip, idx - 1,
1da177e4 1965 XFS_DATA_FORK);
4eea22f0 1966 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1), temp);
9d87c319
ES
1967 oldlen = startblockval(left.br_startblock) +
1968 startblockval(new->br_startblock);
1da177e4 1969 newlen = xfs_bmap_worst_indlen(ip, temp);
4eea22f0 1970 xfs_bmbt_set_startblock(xfs_iext_get_ext(ifp, idx - 1),
9d87c319 1971 nullstartblock((int)newlen));
3a59c94c 1972 XFS_BMAP_TRACE_POST_UPDATE("LC", ip, idx - 1,
1da177e4
LT
1973 XFS_DATA_FORK);
1974 ip->i_df.if_lastex = idx - 1;
3e57ecf6
OW
1975 /* DELTA: One in-core extent grew into a hole. */
1976 temp2 = temp;
1977 temp = left.br_startoff;
1da177e4
LT
1978 break;
1979
1980 case MASK(RIGHT_CONTIG):
1981 /*
1982 * New allocation is contiguous with a delayed allocation
1983 * on the right.
1984 * Merge the new allocation with the right neighbor.
1985 */
3a59c94c 1986 XFS_BMAP_TRACE_PRE_UPDATE("RC", ip, idx, XFS_DATA_FORK);
1da177e4 1987 temp = new->br_blockcount + right.br_blockcount;
9d87c319
ES
1988 oldlen = startblockval(new->br_startblock) +
1989 startblockval(right.br_startblock);
1da177e4
LT
1990 newlen = xfs_bmap_worst_indlen(ip, temp);
1991 xfs_bmbt_set_allf(ep, new->br_startoff,
9d87c319 1992 nullstartblock((int)newlen), temp, right.br_state);
3a59c94c 1993 XFS_BMAP_TRACE_POST_UPDATE("RC", ip, idx, XFS_DATA_FORK);
1da177e4 1994 ip->i_df.if_lastex = idx;
3e57ecf6
OW
1995 /* DELTA: One in-core extent grew into a hole. */
1996 temp2 = temp;
1997 temp = new->br_startoff;
1da177e4
LT
1998 break;
1999
2000 case 0:
2001 /*
2002 * New allocation is not contiguous with another
2003 * delayed allocation.
2004 * Insert a new entry.
2005 */
2006 oldlen = newlen = 0;
3a59c94c 2007 XFS_BMAP_TRACE_INSERT("0", ip, idx, 1, new, NULL,
1da177e4 2008 XFS_DATA_FORK);
4eea22f0 2009 xfs_iext_insert(ifp, idx, 1, new);
1da177e4 2010 ip->i_df.if_lastex = idx;
3e57ecf6
OW
2011 /* DELTA: A new in-core extent was added in a hole. */
2012 temp2 = new->br_blockcount;
2013 temp = new->br_startoff;
1da177e4
LT
2014 break;
2015 }
2016 if (oldlen != newlen) {
2017 ASSERT(oldlen > newlen);
2018 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS,
20f4ebf2 2019 (int64_t)(oldlen - newlen), rsvd);
1da177e4
LT
2020 /*
2021 * Nothing to do for disk quota accounting here.
2022 */
2023 }
3e57ecf6
OW
2024 if (delta) {
2025 temp2 += temp;
2026 if (delta->xed_startoff > temp)
2027 delta->xed_startoff = temp;
2028 if (delta->xed_blockcount < temp2)
2029 delta->xed_blockcount = temp2;
2030 }
1da177e4
LT
2031 *logflagsp = 0;
2032 return 0;
2033#undef MASK
2034#undef MASK2
2035#undef STATE_SET
2036#undef STATE_TEST
2037#undef STATE_SET_TEST
2038#undef SWITCH_STATE
2039}
2040
2041/*
2042 * Called by xfs_bmap_add_extent to handle cases converting a hole
2043 * to a real allocation.
2044 */
2045STATIC int /* error */
2046xfs_bmap_add_extent_hole_real(
2047 xfs_inode_t *ip, /* incore inode pointer */
2048 xfs_extnum_t idx, /* extent number to update/insert */
2049 xfs_btree_cur_t *cur, /* if null, not a btree */
4eea22f0 2050 xfs_bmbt_irec_t *new, /* new data to add to file extents */
1da177e4 2051 int *logflagsp, /* inode logging flags */
3e57ecf6 2052 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
2053 int whichfork) /* data or attr fork */
2054{
a6f64d4a 2055 xfs_bmbt_rec_host_t *ep; /* pointer to extent entry ins. point */
1da177e4 2056 int error; /* error return value */
1da177e4
LT
2057 int i; /* temp state */
2058 xfs_ifork_t *ifp; /* inode fork pointer */
2059 xfs_bmbt_irec_t left; /* left neighbor extent entry */
2060 xfs_bmbt_irec_t right; /* right neighbor extent entry */
3e57ecf6 2061 int rval=0; /* return value (logging flags) */
1da177e4 2062 int state; /* state bits, accessed thru macros */
3e57ecf6
OW
2063 xfs_filblks_t temp=0;
2064 xfs_filblks_t temp2=0;
1da177e4
LT
2065 enum { /* bit number definitions for state */
2066 LEFT_CONTIG, RIGHT_CONTIG,
2067 LEFT_DELAY, RIGHT_DELAY,
2068 LEFT_VALID, RIGHT_VALID
2069 };
2070
2071#define MASK(b) (1 << (b))
2072#define MASK2(a,b) (MASK(a) | MASK(b))
2073#define STATE_SET(b,v) ((v) ? (state |= MASK(b)) : (state &= ~MASK(b)))
2074#define STATE_TEST(b) (state & MASK(b))
2075#define STATE_SET_TEST(b,v) ((v) ? ((state |= MASK(b)), 1) : \
2076 ((state &= ~MASK(b)), 0))
2077#define SWITCH_STATE (state & MASK2(LEFT_CONTIG, RIGHT_CONTIG))
2078
2079 ifp = XFS_IFORK_PTR(ip, whichfork);
2080 ASSERT(idx <= ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t));
4eea22f0 2081 ep = xfs_iext_get_ext(ifp, idx);
1da177e4
LT
2082 state = 0;
2083 /*
2084 * Check and set flags if this segment has a left neighbor.
2085 */
2086 if (STATE_SET_TEST(LEFT_VALID, idx > 0)) {
4eea22f0 2087 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, idx - 1), &left);
9d87c319 2088 STATE_SET(LEFT_DELAY, isnullstartblock(left.br_startblock));
1da177e4
LT
2089 }
2090 /*
2091 * Check and set flags if this segment has a current value.
2092 * Not true if we're inserting into the "hole" at eof.
2093 */
2094 if (STATE_SET_TEST(RIGHT_VALID,
2095 idx <
2096 ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t))) {
2097 xfs_bmbt_get_all(ep, &right);
9d87c319 2098 STATE_SET(RIGHT_DELAY, isnullstartblock(right.br_startblock));
1da177e4
LT
2099 }
2100 /*
2101 * We're inserting a real allocation between "left" and "right".
2102 * Set the contiguity flags. Don't let extents get too large.
2103 */
2104 STATE_SET(LEFT_CONTIG,
2105 STATE_TEST(LEFT_VALID) && !STATE_TEST(LEFT_DELAY) &&
2106 left.br_startoff + left.br_blockcount == new->br_startoff &&
2107 left.br_startblock + left.br_blockcount == new->br_startblock &&
2108 left.br_state == new->br_state &&
2109 left.br_blockcount + new->br_blockcount <= MAXEXTLEN);
2110 STATE_SET(RIGHT_CONTIG,
2111 STATE_TEST(RIGHT_VALID) && !STATE_TEST(RIGHT_DELAY) &&
2112 new->br_startoff + new->br_blockcount == right.br_startoff &&
2113 new->br_startblock + new->br_blockcount ==
2114 right.br_startblock &&
2115 new->br_state == right.br_state &&
2116 new->br_blockcount + right.br_blockcount <= MAXEXTLEN &&
2117 (!STATE_TEST(LEFT_CONTIG) ||
2118 left.br_blockcount + new->br_blockcount +
2119 right.br_blockcount <= MAXEXTLEN));
2120
3e57ecf6 2121 error = 0;
1da177e4
LT
2122 /*
2123 * Select which case we're in here, and implement it.
2124 */
2125 switch (SWITCH_STATE) {
2126
2127 case MASK2(LEFT_CONTIG, RIGHT_CONTIG):
2128 /*
2129 * New allocation is contiguous with real allocations on the
2130 * left and on the right.
4eea22f0 2131 * Merge all three into a single extent record.
1da177e4 2132 */
3a59c94c 2133 XFS_BMAP_TRACE_PRE_UPDATE("LC|RC", ip, idx - 1,
1da177e4 2134 whichfork);
4eea22f0 2135 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4
LT
2136 left.br_blockcount + new->br_blockcount +
2137 right.br_blockcount);
3a59c94c 2138 XFS_BMAP_TRACE_POST_UPDATE("LC|RC", ip, idx - 1,
1da177e4 2139 whichfork);
3a59c94c 2140 XFS_BMAP_TRACE_DELETE("LC|RC", ip, idx, 1, whichfork);
4eea22f0 2141 xfs_iext_remove(ifp, idx, 1);
1da177e4
LT
2142 ifp->if_lastex = idx - 1;
2143 XFS_IFORK_NEXT_SET(ip, whichfork,
2144 XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
2145 if (cur == NULL) {
9d87c319 2146 rval = XFS_ILOG_CORE | xfs_ilog_fext(whichfork);
3e57ecf6
OW
2147 } else {
2148 rval = XFS_ILOG_CORE;
2149 if ((error = xfs_bmbt_lookup_eq(cur,
2150 right.br_startoff,
2151 right.br_startblock,
2152 right.br_blockcount, &i)))
2153 goto done;
6bd8fc8a 2154 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
91cca5df 2155 if ((error = xfs_btree_delete(cur, &i)))
3e57ecf6 2156 goto done;
6bd8fc8a 2157 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
8df4da4a 2158 if ((error = xfs_btree_decrement(cur, 0, &i)))
3e57ecf6 2159 goto done;
6bd8fc8a 2160 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
3e57ecf6
OW
2161 if ((error = xfs_bmbt_update(cur, left.br_startoff,
2162 left.br_startblock,
2163 left.br_blockcount +
2164 new->br_blockcount +
2165 right.br_blockcount,
2166 left.br_state)))
2167 goto done;
1da177e4 2168 }
3e57ecf6
OW
2169 /* DELTA: Two in-core extents were replaced by one. */
2170 temp = left.br_startoff;
2171 temp2 = left.br_blockcount +
2172 new->br_blockcount +
2173 right.br_blockcount;
2174 break;
1da177e4
LT
2175
2176 case MASK(LEFT_CONTIG):
2177 /*
2178 * New allocation is contiguous with a real allocation
2179 * on the left.
2180 * Merge the new allocation with the left neighbor.
2181 */
3a59c94c 2182 XFS_BMAP_TRACE_PRE_UPDATE("LC", ip, idx - 1, whichfork);
4eea22f0 2183 xfs_bmbt_set_blockcount(xfs_iext_get_ext(ifp, idx - 1),
1da177e4 2184 left.br_blockcount + new->br_blockcount);
3a59c94c 2185 XFS_BMAP_TRACE_POST_UPDATE("LC", ip, idx - 1, whichfork);
1da177e4
LT
2186 ifp->if_lastex = idx - 1;
2187 if (cur == NULL) {
9d87c319 2188 rval = xfs_ilog_fext(whichfork);
3e57ecf6
OW
2189 } else {
2190 rval = 0;
2191 if ((error = xfs_bmbt_lookup_eq(cur,
2192 left.br_startoff,
2193 left.br_startblock,
2194 left.br_blockcount, &i)))
2195 goto done;
6bd8fc8a 2196 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
3e57ecf6
OW
2197 if ((error = xfs_bmbt_update(cur, left.br_startoff,
2198 left.br_startblock,
2199 left.br_blockcount +
2200 new->br_blockcount,
2201 left.br_state)))
2202 goto done;
1da177e4 2203 }
3e57ecf6
OW
2204 /* DELTA: One in-core extent grew. */
2205 temp = left.br_startoff;
2206 temp2 = left.br_blockcount +
2207 new->br_blockcount;
2208 break;
1da177e4
LT
2209
2210 case MASK(RIGHT_CONTIG):
2211 /*
2212 * New allocation is contiguous with a real allocation
2213 * on the right.
2214 * Merge the new allocation with the right neighbor.
2215 */
3a59c94c 2216 XFS_BMAP_TRACE_PRE_UPDATE("RC", ip, idx, whichfork);
1da177e4
LT
2217 xfs_bmbt_set_allf(ep, new->br_startoff, new->br_startblock,
2218 new->br_blockcount + right.br_blockcount,
2219 right.br_state);
3a59c94c 2220 XFS_BMAP_TRACE_POST_UPDATE("RC", ip, idx, whichfork);
1da177e4
LT
2221 ifp->if_lastex = idx;
2222 if (cur == NULL) {
9d87c319 2223 rval = xfs_ilog_fext(whichfork);
3e57ecf6
OW
2224 } else {
2225 rval = 0;
2226 if ((error = xfs_bmbt_lookup_eq(cur,
2227 right.br_startoff,
2228 right.br_startblock,
2229 right.br_blockcount, &i)))
2230 goto done;
6bd8fc8a 2231 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
3e57ecf6
OW
2232 if ((error = xfs_bmbt_update(cur, new->br_startoff,
2233 new->br_startblock,
2234 new->br_blockcount +
2235 right.br_blockcount,
2236 right.br_state)))
2237 goto done;
1da177e4 2238 }
3e57ecf6
OW
2239 /* DELTA: One in-core extent grew. */
2240 temp = new->br_startoff;
2241 temp2 = new->br_blockcount +
2242 right.br_blockcount;
2243 break;
1da177e4
LT
2244
2245 case 0:
2246 /*
2247 * New allocation is not contiguous with another
2248 * real allocation.
2249 * Insert a new entry.
2250 */
3a59c94c 2251 XFS_BMAP_TRACE_INSERT("0", ip, idx, 1, new, NULL, whichfork);
4eea22f0 2252 xfs_iext_insert(ifp, idx, 1, new);
1da177e4
LT
2253 ifp->if_lastex = idx;
2254 XFS_IFORK_NEXT_SET(ip, whichfork,
2255 XFS_IFORK_NEXTENTS(ip, whichfork) + 1);
2256 if (cur == NULL) {
9d87c319 2257 rval = XFS_ILOG_CORE | xfs_ilog_fext(whichfork);
3e57ecf6
OW
2258 } else {
2259 rval = XFS_ILOG_CORE;
2260 if ((error = xfs_bmbt_lookup_eq(cur,
2261 new->br_startoff,
2262 new->br_startblock,
2263 new->br_blockcount, &i)))
2264 goto done;
6bd8fc8a 2265 XFS_WANT_CORRUPTED_GOTO(i == 0, done);
3e57ecf6 2266 cur->bc_rec.b.br_state = new->br_state;
4b22a571 2267 if ((error = xfs_btree_insert(cur, &i)))
3e57ecf6 2268 goto done;
6bd8fc8a 2269 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4 2270 }
3e57ecf6
OW
2271 /* DELTA: A new extent was added in a hole. */
2272 temp = new->br_startoff;
2273 temp2 = new->br_blockcount;
2274 break;
1da177e4 2275 }
3e57ecf6
OW
2276 if (delta) {
2277 temp2 += temp;
2278 if (delta->xed_startoff > temp)
2279 delta->xed_startoff = temp;
2280 if (delta->xed_blockcount < temp2)
2281 delta->xed_blockcount = temp2;
2282 }
2283done:
2284 *logflagsp = rval;
2285 return error;
1da177e4
LT
2286#undef MASK
2287#undef MASK2
2288#undef STATE_SET
2289#undef STATE_TEST
2290#undef STATE_SET_TEST
2291#undef SWITCH_STATE
1da177e4
LT
2292}
2293
dd9f438e
NS
2294/*
2295 * Adjust the size of the new extent based on di_extsize and rt extsize.
2296 */
2297STATIC int
2298xfs_bmap_extsize_align(
2299 xfs_mount_t *mp,
2300 xfs_bmbt_irec_t *gotp, /* next extent pointer */
2301 xfs_bmbt_irec_t *prevp, /* previous extent pointer */
2302 xfs_extlen_t extsz, /* align to this extent size */
2303 int rt, /* is this a realtime inode? */
2304 int eof, /* is extent at end-of-file? */
2305 int delay, /* creating delalloc extent? */
2306 int convert, /* overwriting unwritten extent? */
2307 xfs_fileoff_t *offp, /* in/out: aligned offset */
2308 xfs_extlen_t *lenp) /* in/out: aligned length */
2309{
2310 xfs_fileoff_t orig_off; /* original offset */
2311 xfs_extlen_t orig_alen; /* original length */
2312 xfs_fileoff_t orig_end; /* original off+len */
2313 xfs_fileoff_t nexto; /* next file offset */
2314 xfs_fileoff_t prevo; /* previous file offset */
2315 xfs_fileoff_t align_off; /* temp for offset */
2316 xfs_extlen_t align_alen; /* temp for length */
2317 xfs_extlen_t temp; /* temp for calculations */
2318
2319 if (convert)
2320 return 0;
2321
2322 orig_off = align_off = *offp;
2323 orig_alen = align_alen = *lenp;
2324 orig_end = orig_off + orig_alen;
2325
2326 /*
2327 * If this request overlaps an existing extent, then don't
2328 * attempt to perform any additional alignment.
2329 */
2330 if (!delay && !eof &&
2331 (orig_off >= gotp->br_startoff) &&
2332 (orig_end <= gotp->br_startoff + gotp->br_blockcount)) {
2333 return 0;
2334 }
2335
2336 /*
2337 * If the file offset is unaligned vs. the extent size
2338 * we need to align it. This will be possible unless
2339 * the file was previously written with a kernel that didn't
2340 * perform this alignment, or if a truncate shot us in the
2341 * foot.
2342 */
2343 temp = do_mod(orig_off, extsz);
2344 if (temp) {
2345 align_alen += temp;
2346 align_off -= temp;
2347 }
2348 /*
2349 * Same adjustment for the end of the requested area.
2350 */
2351 if ((temp = (align_alen % extsz))) {
2352 align_alen += extsz - temp;
2353 }
2354 /*
2355 * If the previous block overlaps with this proposed allocation
2356 * then move the start forward without adjusting the length.
2357 */
2358 if (prevp->br_startoff != NULLFILEOFF) {
2359 if (prevp->br_startblock == HOLESTARTBLOCK)
2360 prevo = prevp->br_startoff;
2361 else
2362 prevo = prevp->br_startoff + prevp->br_blockcount;
2363 } else
2364 prevo = 0;
2365 if (align_off != orig_off && align_off < prevo)
2366 align_off = prevo;
2367 /*
2368 * If the next block overlaps with this proposed allocation
2369 * then move the start back without adjusting the length,
2370 * but not before offset 0.
2371 * This may of course make the start overlap previous block,
2372 * and if we hit the offset 0 limit then the next block
2373 * can still overlap too.
2374 */
2375 if (!eof && gotp->br_startoff != NULLFILEOFF) {
2376 if ((delay && gotp->br_startblock == HOLESTARTBLOCK) ||
2377 (!delay && gotp->br_startblock == DELAYSTARTBLOCK))
2378 nexto = gotp->br_startoff + gotp->br_blockcount;
2379 else
2380 nexto = gotp->br_startoff;
2381 } else
2382 nexto = NULLFILEOFF;
2383 if (!eof &&
2384 align_off + align_alen != orig_end &&
2385 align_off + align_alen > nexto)
2386 align_off = nexto > align_alen ? nexto - align_alen : 0;
2387 /*
2388 * If we're now overlapping the next or previous extent that
2389 * means we can't fit an extsz piece in this hole. Just move
2390 * the start forward to the first valid spot and set
2391 * the length so we hit the end.
2392 */
2393 if (align_off != orig_off && align_off < prevo)
2394 align_off = prevo;
2395 if (align_off + align_alen != orig_end &&
2396 align_off + align_alen > nexto &&
2397 nexto != NULLFILEOFF) {
2398 ASSERT(nexto > prevo);
2399 align_alen = nexto - align_off;
2400 }
2401
2402 /*
2403 * If realtime, and the result isn't a multiple of the realtime
2404 * extent size we need to remove blocks until it is.
2405 */
2406 if (rt && (temp = (align_alen % mp->m_sb.sb_rextsize))) {
2407 /*
2408 * We're not covering the original request, or
2409 * we won't be able to once we fix the length.
2410 */
2411 if (orig_off < align_off ||
2412 orig_end > align_off + align_alen ||
2413 align_alen - temp < orig_alen)
2414 return XFS_ERROR(EINVAL);
2415 /*
2416 * Try to fix it by moving the start up.
2417 */
2418 if (align_off + temp <= orig_off) {
2419 align_alen -= temp;
2420 align_off += temp;
2421 }
2422 /*
2423 * Try to fix it by moving the end in.
2424 */
2425 else if (align_off + align_alen - temp >= orig_end)
2426 align_alen -= temp;
2427 /*
2428 * Set the start to the minimum then trim the length.
2429 */
2430 else {
2431 align_alen -= orig_off - align_off;
2432 align_off = orig_off;
2433 align_alen -= align_alen % mp->m_sb.sb_rextsize;
2434 }
2435 /*
2436 * Result doesn't cover the request, fail it.
2437 */
2438 if (orig_off < align_off || orig_end > align_off + align_alen)
2439 return XFS_ERROR(EINVAL);
2440 } else {
2441 ASSERT(orig_off >= align_off);
2442 ASSERT(orig_end <= align_off + align_alen);
2443 }
2444
2445#ifdef DEBUG
2446 if (!eof && gotp->br_startoff != NULLFILEOFF)
2447 ASSERT(align_off + align_alen <= gotp->br_startoff);
2448 if (prevp->br_startoff != NULLFILEOFF)
2449 ASSERT(align_off >= prevp->br_startoff + prevp->br_blockcount);
2450#endif
2451
2452 *lenp = align_alen;
2453 *offp = align_off;
2454 return 0;
2455}
2456
1da177e4
LT
2457#define XFS_ALLOC_GAP_UNITS 4
2458
c2b1cba6 2459STATIC void
a365bdd5 2460xfs_bmap_adjacent(
1da177e4
LT
2461 xfs_bmalloca_t *ap) /* bmap alloc argument struct */
2462{
2463 xfs_fsblock_t adjust; /* adjustment to block numbers */
1da177e4
LT
2464 xfs_agnumber_t fb_agno; /* ag number of ap->firstblock */
2465 xfs_mount_t *mp; /* mount point structure */
2466 int nullfb; /* true if ap->firstblock isn't set */
2467 int rt; /* true if inode is realtime */
1da177e4
LT
2468
2469#define ISVALID(x,y) \
2470 (rt ? \
2471 (x) < mp->m_sb.sb_rblocks : \
2472 XFS_FSB_TO_AGNO(mp, x) == XFS_FSB_TO_AGNO(mp, y) && \
2473 XFS_FSB_TO_AGNO(mp, x) < mp->m_sb.sb_agcount && \
2474 XFS_FSB_TO_AGBNO(mp, x) < mp->m_sb.sb_agblocks)
2475
1da177e4
LT
2476 mp = ap->ip->i_mount;
2477 nullfb = ap->firstblock == NULLFSBLOCK;
2478 rt = XFS_IS_REALTIME_INODE(ap->ip) && ap->userdata;
2479 fb_agno = nullfb ? NULLAGNUMBER : XFS_FSB_TO_AGNO(mp, ap->firstblock);
1da177e4
LT
2480 /*
2481 * If allocating at eof, and there's a previous real block,
9da096fd 2482 * try to use its last block as our starting point.
1da177e4
LT
2483 */
2484 if (ap->eof && ap->prevp->br_startoff != NULLFILEOFF &&
9d87c319 2485 !isnullstartblock(ap->prevp->br_startblock) &&
1da177e4
LT
2486 ISVALID(ap->prevp->br_startblock + ap->prevp->br_blockcount,
2487 ap->prevp->br_startblock)) {
2488 ap->rval = ap->prevp->br_startblock + ap->prevp->br_blockcount;
2489 /*
2490 * Adjust for the gap between prevp and us.
2491 */
2492 adjust = ap->off -
2493 (ap->prevp->br_startoff + ap->prevp->br_blockcount);
2494 if (adjust &&
2495 ISVALID(ap->rval + adjust, ap->prevp->br_startblock))
2496 ap->rval += adjust;
2497 }
2498 /*
2499 * If not at eof, then compare the two neighbor blocks.
2500 * Figure out whether either one gives us a good starting point,
2501 * and pick the better one.
2502 */
2503 else if (!ap->eof) {
2504 xfs_fsblock_t gotbno; /* right side block number */
2505 xfs_fsblock_t gotdiff=0; /* right side difference */
2506 xfs_fsblock_t prevbno; /* left side block number */
2507 xfs_fsblock_t prevdiff=0; /* left side difference */
2508
2509 /*
2510 * If there's a previous (left) block, select a requested
2511 * start block based on it.
2512 */
2513 if (ap->prevp->br_startoff != NULLFILEOFF &&
9d87c319 2514 !isnullstartblock(ap->prevp->br_startblock) &&
1da177e4
LT
2515 (prevbno = ap->prevp->br_startblock +
2516 ap->prevp->br_blockcount) &&
2517 ISVALID(prevbno, ap->prevp->br_startblock)) {
2518 /*
2519 * Calculate gap to end of previous block.
2520 */
2521 adjust = prevdiff = ap->off -
2522 (ap->prevp->br_startoff +
2523 ap->prevp->br_blockcount);
2524 /*
2525 * Figure the startblock based on the previous block's
2526 * end and the gap size.
2527 * Heuristic!
2528 * If the gap is large relative to the piece we're
2529 * allocating, or using it gives us an invalid block
2530 * number, then just use the end of the previous block.
2531 */
2532 if (prevdiff <= XFS_ALLOC_GAP_UNITS * ap->alen &&
2533 ISVALID(prevbno + prevdiff,
2534 ap->prevp->br_startblock))
2535 prevbno += adjust;
2536 else
2537 prevdiff += adjust;
2538 /*
2539 * If the firstblock forbids it, can't use it,
2540 * must use default.
2541 */
2542 if (!rt && !nullfb &&
2543 XFS_FSB_TO_AGNO(mp, prevbno) != fb_agno)
2544 prevbno = NULLFSBLOCK;
2545 }
2546 /*
2547 * No previous block or can't follow it, just default.
2548 */
2549 else
2550 prevbno = NULLFSBLOCK;
2551 /*
2552 * If there's a following (right) block, select a requested
2553 * start block based on it.
2554 */
9d87c319 2555 if (!isnullstartblock(ap->gotp->br_startblock)) {
1da177e4
LT
2556 /*
2557 * Calculate gap to start of next block.
2558 */
2559 adjust = gotdiff = ap->gotp->br_startoff - ap->off;
2560 /*
2561 * Figure the startblock based on the next block's
2562 * start and the gap size.
2563 */
2564 gotbno = ap->gotp->br_startblock;
2565 /*
2566 * Heuristic!
2567 * If the gap is large relative to the piece we're
2568 * allocating, or using it gives us an invalid block
2569 * number, then just use the start of the next block
2570 * offset by our length.
2571 */
2572 if (gotdiff <= XFS_ALLOC_GAP_UNITS * ap->alen &&
2573 ISVALID(gotbno - gotdiff, gotbno))
2574 gotbno -= adjust;
2575 else if (ISVALID(gotbno - ap->alen, gotbno)) {
2576 gotbno -= ap->alen;
2577 gotdiff += adjust - ap->alen;
2578 } else
2579 gotdiff += adjust;
2580 /*
2581 * If the firstblock forbids it, can't use it,
2582 * must use default.
2583 */
2584 if (!rt && !nullfb &&
2585 XFS_FSB_TO_AGNO(mp, gotbno) != fb_agno)
2586 gotbno = NULLFSBLOCK;
2587 }
2588 /*
2589 * No next block, just default.
2590 */
2591 else
2592 gotbno = NULLFSBLOCK;
2593 /*
2594 * If both valid, pick the better one, else the only good
2595 * one, else ap->rval is already set (to 0 or the inode block).
2596 */
2597 if (prevbno != NULLFSBLOCK && gotbno != NULLFSBLOCK)
2598 ap->rval = prevdiff <= gotdiff ? prevbno : gotbno;
2599 else if (prevbno != NULLFSBLOCK)
2600 ap->rval = prevbno;
2601 else if (gotbno != NULLFSBLOCK)
2602 ap->rval = gotbno;
2603 }
a365bdd5 2604#undef ISVALID
a365bdd5
NS
2605}
2606
2607STATIC int
2608xfs_bmap_rtalloc(
2609 xfs_bmalloca_t *ap) /* bmap alloc argument struct */
2610{
2611 xfs_alloctype_t atype = 0; /* type for allocation routines */
2612 int error; /* error return value */
2613 xfs_mount_t *mp; /* mount point structure */
2614 xfs_extlen_t prod = 0; /* product factor for allocators */
2615 xfs_extlen_t ralen = 0; /* realtime allocation length */
2616 xfs_extlen_t align; /* minimum allocation alignment */
a365bdd5
NS
2617 xfs_rtblock_t rtb;
2618
2619 mp = ap->ip->i_mount;
957d0ebe 2620 align = xfs_get_extsz_hint(ap->ip);
a365bdd5
NS
2621 prod = align / mp->m_sb.sb_rextsize;
2622 error = xfs_bmap_extsize_align(mp, ap->gotp, ap->prevp,
2623 align, 1, ap->eof, 0,
2624 ap->conv, &ap->off, &ap->alen);
2625 if (error)
2626 return error;
2627 ASSERT(ap->alen);
2628 ASSERT(ap->alen % mp->m_sb.sb_rextsize == 0);
2629
2630 /*
2631 * If the offset & length are not perfectly aligned
2632 * then kill prod, it will just get us in trouble.
2633 */
2634 if (do_mod(ap->off, align) || ap->alen % align)
2635 prod = 1;
2636 /*
2637 * Set ralen to be the actual requested length in rtextents.
2638 */
2639 ralen = ap->alen / mp->m_sb.sb_rextsize;
2640 /*
2641 * If the old value was close enough to MAXEXTLEN that
2642 * we rounded up to it, cut it back so it's valid again.
2643 * Note that if it's a really large request (bigger than
2644 * MAXEXTLEN), we don't hear about that number, and can't
2645 * adjust the starting point to match it.
2646 */
2647 if (ralen * mp->m_sb.sb_rextsize >= MAXEXTLEN)
2648 ralen = MAXEXTLEN / mp->m_sb.sb_rextsize;
2649 /*
2650 * If it's an allocation to an empty file at offset 0,
2651 * pick an extent that will space things out in the rt area.
2652 */
2653 if (ap->eof && ap->off == 0) {
0892ccd6
AM
2654 xfs_rtblock_t uninitialized_var(rtx); /* realtime extent no */
2655
a365bdd5
NS
2656 error = xfs_rtpick_extent(mp, ap->tp, ralen, &rtx);
2657 if (error)
2658 return error;
2659 ap->rval = rtx * mp->m_sb.sb_rextsize;
2660 } else {
2661 ap->rval = 0;
2662 }
2663
2664 xfs_bmap_adjacent(ap);
2665
2666 /*
2667 * Realtime allocation, done through xfs_rtallocate_extent.
2668 */
2669 atype = ap->rval == 0 ? XFS_ALLOCTYPE_ANY_AG : XFS_ALLOCTYPE_NEAR_BNO;
2670 do_div(ap->rval, mp->m_sb.sb_rextsize);
2671 rtb = ap->rval;
2672 ap->alen = ralen;
2673 if ((error = xfs_rtallocate_extent(ap->tp, ap->rval, 1, ap->alen,
2674 &ralen, atype, ap->wasdel, prod, &rtb)))
2675 return error;
2676 if (rtb == NULLFSBLOCK && prod > 1 &&
2677 (error = xfs_rtallocate_extent(ap->tp, ap->rval, 1,
2678 ap->alen, &ralen, atype,
2679 ap->wasdel, 1, &rtb)))
2680 return error;
2681 ap->rval = rtb;
2682 if (ap->rval != NULLFSBLOCK) {
2683 ap->rval *= mp->m_sb.sb_rextsize;
2684 ralen *= mp->m_sb.sb_rextsize;
2685 ap->alen = ralen;
2686 ap->ip->i_d.di_nblocks += ralen;
2687 xfs_trans_log_inode(ap->tp, ap->ip, XFS_ILOG_CORE);
2688 if (ap->wasdel)
2689 ap->ip->i_delayed_blks -= ralen;
2690 /*
2691 * Adjust the disk quota also. This was reserved
2692 * earlier.
2693 */
2694 XFS_TRANS_MOD_DQUOT_BYINO(mp, ap->tp, ap->ip,
2695 ap->wasdel ? XFS_TRANS_DQ_DELRTBCOUNT :
2696 XFS_TRANS_DQ_RTBCOUNT, (long) ralen);
2697 } else {
2698 ap->alen = 0;
2699 }
2700 return 0;
2701}
2702
2703STATIC int
2704xfs_bmap_btalloc(
2705 xfs_bmalloca_t *ap) /* bmap alloc argument struct */
2706{
2707 xfs_mount_t *mp; /* mount point structure */
2708 xfs_alloctype_t atype = 0; /* type for allocation routines */
2709 xfs_extlen_t align; /* minimum allocation alignment */
2710 xfs_agnumber_t ag;
2711 xfs_agnumber_t fb_agno; /* ag number of ap->firstblock */
2712 xfs_agnumber_t startag;
2713 xfs_alloc_arg_t args;
2714 xfs_extlen_t blen;
a365bdd5
NS
2715 xfs_extlen_t nextminlen = 0;
2716 xfs_perag_t *pag;
2717 int nullfb; /* true if ap->firstblock isn't set */
2718 int isaligned;
2719 int notinit;
2720 int tryagain;
2721 int error;
2722
2723 mp = ap->ip->i_mount;
957d0ebe 2724 align = ap->userdata ? xfs_get_extsz_hint(ap->ip) : 0;
a365bdd5
NS
2725 if (unlikely(align)) {
2726 error = xfs_bmap_extsize_align(mp, ap->gotp, ap->prevp,
2727 align, 0, ap->eof, 0, ap->conv,
2728 &ap->off, &ap->alen);
2729 ASSERT(!error);
2730 ASSERT(ap->alen);
2731 }
2732 nullfb = ap->firstblock == NULLFSBLOCK;
2733 fb_agno = nullfb ? NULLAGNUMBER : XFS_FSB_TO_AGNO(mp, ap->firstblock);
2a82b8be
DC
2734 if (nullfb) {
2735 if (ap->userdata && xfs_inode_is_filestream(ap->ip)) {
2736 ag = xfs_filestream_lookup_ag(ap->ip);
2737 ag = (ag != NULLAGNUMBER) ? ag : 0;
2738 ap->rval = XFS_AGB_TO_FSB(mp, ag, 0);
2739 } else {
2740 ap->rval = XFS_INO_TO_FSB(mp, ap->ip->i_ino);
2741 }
2742 } else
a365bdd5
NS
2743 ap->rval = ap->firstblock;
2744
2745 xfs_bmap_adjacent(ap);
2746
1da177e4
LT
2747 /*
2748 * If allowed, use ap->rval; otherwise must use firstblock since
2749 * it's in the right allocation group.
2750 */
a365bdd5 2751 if (nullfb || XFS_FSB_TO_AGNO(mp, ap->rval) == fb_agno)
1da177e4
LT
2752 ;
2753 else
2754 ap->rval = ap->firstblock;
2755 /*
a365bdd5 2756 * Normal allocation, done through xfs_alloc_vextent.
1da177e4 2757 */
a365bdd5
NS
2758 tryagain = isaligned = 0;
2759 args.tp = ap->tp;
2760 args.mp = mp;
2761 args.fsbno = ap->rval;
2762 args.maxlen = MIN(ap->alen, mp->m_sb.sb_agblocks);
d210a28c 2763 args.firstblock = ap->firstblock;
a365bdd5
NS
2764 blen = 0;
2765 if (nullfb) {
2a82b8be
DC
2766 if (ap->userdata && xfs_inode_is_filestream(ap->ip))
2767 args.type = XFS_ALLOCTYPE_NEAR_BNO;
2768 else
2769 args.type = XFS_ALLOCTYPE_START_BNO;
a365bdd5 2770 args.total = ap->total;
2a82b8be 2771
a365bdd5 2772 /*
2a82b8be
DC
2773 * Search for an allocation group with a single extent
2774 * large enough for the request.
2775 *
2776 * If one isn't found, then adjust the minimum allocation
2777 * size to the largest space found.
a365bdd5
NS
2778 */
2779 startag = ag = XFS_FSB_TO_AGNO(mp, args.fsbno);
2a82b8be
DC
2780 if (startag == NULLAGNUMBER)
2781 startag = ag = 0;
a365bdd5
NS
2782 notinit = 0;
2783 down_read(&mp->m_peraglock);
2784 while (blen < ap->alen) {
2785 pag = &mp->m_perag[ag];
2786 if (!pag->pagf_init &&
2787 (error = xfs_alloc_pagf_init(mp, args.tp,
2788 ag, XFS_ALLOC_FLAG_TRYLOCK))) {
2789 up_read(&mp->m_peraglock);
2790 return error;
2791 }
1da177e4 2792 /*
a365bdd5 2793 * See xfs_alloc_fix_freelist...
1da177e4 2794 */
a365bdd5 2795 if (pag->pagf_init) {
6cc87645
DC
2796 xfs_extlen_t longest;
2797 longest = xfs_alloc_longest_free_extent(mp, pag);
a365bdd5
NS
2798 if (blen < longest)
2799 blen = longest;
2800 } else
2801 notinit = 1;
2a82b8be
DC
2802
2803 if (xfs_inode_is_filestream(ap->ip)) {
2804 if (blen >= ap->alen)
2805 break;
2806
2807 if (ap->userdata) {
2808 /*
2809 * If startag is an invalid AG, we've
2810 * come here once before and
2811 * xfs_filestream_new_ag picked the
2812 * best currently available.
2813 *
2814 * Don't continue looping, since we
2815 * could loop forever.
2816 */
2817 if (startag == NULLAGNUMBER)
2818 break;
2819
2820 error = xfs_filestream_new_ag(ap, &ag);
2821 if (error) {
2822 up_read(&mp->m_peraglock);
2823 return error;
2824 }
2825
2826 /* loop again to set 'blen'*/
2827 startag = NULLAGNUMBER;
2828 continue;
2829 }
2830 }
a365bdd5
NS
2831 if (++ag == mp->m_sb.sb_agcount)
2832 ag = 0;
2833 if (ag == startag)
2834 break;
2835 }
2836 up_read(&mp->m_peraglock);
2837 /*
2838 * Since the above loop did a BUF_TRYLOCK, it is
2839 * possible that there is space for this request.
2840 */
2841 if (notinit || blen < ap->minlen)
2842 args.minlen = ap->minlen;
2843 /*
2844 * If the best seen length is less than the request
2845 * length, use the best as the minimum.
2846 */
2847 else if (blen < ap->alen)
2848 args.minlen = blen;
2849 /*
2850 * Otherwise we've seen an extent as big as alen,
2851 * use that as the minimum.
2852 */
2853 else
2854 args.minlen = ap->alen;
2a82b8be
DC
2855
2856 /*
2857 * set the failure fallback case to look in the selected
2858 * AG as the stream may have moved.
2859 */
2860 if (xfs_inode_is_filestream(ap->ip))
2861 ap->rval = args.fsbno = XFS_AGB_TO_FSB(mp, ag, 0);
a365bdd5 2862 } else if (ap->low) {
2a82b8be
DC
2863 if (xfs_inode_is_filestream(ap->ip))
2864 args.type = XFS_ALLOCTYPE_FIRST_AG;
2865 else
2866 args.type = XFS_ALLOCTYPE_START_BNO;
a365bdd5
NS
2867 args.total = args.minlen = ap->minlen;
2868 } else {
2869 args.type = XFS_ALLOCTYPE_NEAR_BNO;
2870 args.total = ap->total;
2871 args.minlen = ap->minlen;
2872 }
957d0ebe
DC
2873 /* apply extent size hints if obtained earlier */
2874 if (unlikely(align)) {
2875 args.prod = align;
a365bdd5
NS
2876 if ((args.mod = (xfs_extlen_t)do_mod(ap->off, args.prod)))
2877 args.mod = (xfs_extlen_t)(args.prod - args.mod);
e6a4b37f 2878 } else if (mp->m_sb.sb_blocksize >= PAGE_CACHE_SIZE) {
a365bdd5
NS
2879 args.prod = 1;
2880 args.mod = 0;
2881 } else {
e6a4b37f 2882 args.prod = PAGE_CACHE_SIZE >> mp->m_sb.sb_blocklog;
a365bdd5
NS
2883 if ((args.mod = (xfs_extlen_t)(do_mod(ap->off, args.prod))))
2884 args.mod = (xfs_extlen_t)(args.prod - args.mod);
1da177e4
LT
2885 }
2886 /*
a365bdd5
NS
2887 * If we are not low on available data blocks, and the
2888 * underlying logical volume manager is a stripe, and
2889 * the file offset is zero then try to allocate data
2890 * blocks on stripe unit boundary.
2891 * NOTE: ap->aeof is only set if the allocation length
2892 * is >= the stripe unit and the allocation offset is
2893 * at the end of file.
1da177e4 2894 */
a365bdd5
NS
2895 if (!ap->low && ap->aeof) {
2896 if (!ap->off) {
2897 args.alignment = mp->m_dalign;
2898 atype = args.type;
2899 isaligned = 1;
1da177e4 2900 /*
a365bdd5 2901 * Adjust for alignment
1da177e4 2902 */
a365bdd5
NS
2903 if (blen > args.alignment && blen <= ap->alen)
2904 args.minlen = blen - args.alignment;
2905 args.minalignslop = 0;
2906 } else {
1da177e4 2907 /*
a365bdd5
NS
2908 * First try an exact bno allocation.
2909 * If it fails then do a near or start bno
2910 * allocation with alignment turned on.
1da177e4 2911 */
a365bdd5
NS
2912 atype = args.type;
2913 tryagain = 1;
2914 args.type = XFS_ALLOCTYPE_THIS_BNO;
2915 args.alignment = 1;
1da177e4 2916 /*
a365bdd5
NS
2917 * Compute the minlen+alignment for the
2918 * next case. Set slop so that the value
2919 * of minlen+alignment+slop doesn't go up
2920 * between the calls.
1da177e4 2921 */
a365bdd5
NS
2922 if (blen > mp->m_dalign && blen <= ap->alen)
2923 nextminlen = blen - mp->m_dalign;
1da177e4 2924 else
a365bdd5
NS
2925 nextminlen = args.minlen;
2926 if (nextminlen + mp->m_dalign > args.minlen + 1)
2927 args.minalignslop =
2928 nextminlen + mp->m_dalign -
2929 args.minlen - 1;
2930 else
2931 args.minalignslop = 0;
1da177e4 2932 }
a365bdd5
NS
2933 } else {
2934 args.alignment = 1;
2935 args.minalignslop = 0;
2936 }
2937 args.minleft = ap->minleft;
2938 args.wasdel = ap->wasdel;
2939 args.isfl = 0;
2940 args.userdata = ap->userdata;
2941 if ((error = xfs_alloc_vextent(&args)))
2942 return error;
2943 if (tryagain && args.fsbno == NULLFSBLOCK) {
1da177e4 2944 /*
a365bdd5
NS
2945 * Exact allocation failed. Now try with alignment
2946 * turned on.
1da177e4 2947 */
a365bdd5
NS
2948 args.type = atype;
2949 args.fsbno = ap->rval;
2950 args.alignment = mp->m_dalign;
2951 args.minlen = nextminlen;
2952 args.minalignslop = 0;
2953 isaligned = 1;
1da177e4
LT
2954 if ((error = xfs_alloc_vextent(&args)))
2955 return error;
a365bdd5
NS
2956 }
2957 if (isaligned && args.fsbno == NULLFSBLOCK) {
2958 /*
2959 * allocation failed, so turn off alignment and
2960 * try again.
2961 */
2962 args.type = atype;
2963 args.fsbno = ap->rval;
2964 args.alignment = 0;
2965 if ((error = xfs_alloc_vextent(&args)))
2966 return error;
2967 }
2968 if (args.fsbno == NULLFSBLOCK && nullfb &&
2969 args.minlen > ap->minlen) {
2970 args.minlen = ap->minlen;
2971 args.type = XFS_ALLOCTYPE_START_BNO;
2972 args.fsbno = ap->rval;
2973 if ((error = xfs_alloc_vextent(&args)))
2974 return error;
2975 }
2976 if (args.fsbno == NULLFSBLOCK && nullfb) {
2977 args.fsbno = 0;
2978 args.type = XFS_ALLOCTYPE_FIRST_AG;
2979 args.total = ap->minlen;
2980 args.minleft = 0;
2981 if ((error = xfs_alloc_vextent(&args)))
2982 return error;
2983 ap->low = 1;
2984 }
2985 if (args.fsbno != NULLFSBLOCK) {
2986 ap->firstblock = ap->rval = args.fsbno;
2987 ASSERT(nullfb || fb_agno == args.agno ||
2988 (ap->low && fb_agno < args.agno));
2989 ap->alen = args.len;
2990 ap->ip->i_d.di_nblocks += args.len;
2991 xfs_trans_log_inode(ap->tp, ap->ip, XFS_ILOG_CORE);
2992 if (ap->wasdel)
2993 ap->ip->i_delayed_blks -= args.len;
2994 /*
2995 * Adjust the disk quota also. This was reserved
2996 * earlier.
2997 */
2998 XFS_TRANS_MOD_DQUOT_BYINO(mp, ap->tp, ap->ip,
2999 ap->wasdel ? XFS_TRANS_DQ_DELBCOUNT :
3000 XFS_TRANS_DQ_BCOUNT,
3001 (long) args.len);
3002 } else {
3003 ap->rval = NULLFSBLOCK;
3004 ap->alen = 0;
1da177e4
LT
3005 }
3006 return 0;
a365bdd5
NS
3007}
3008
3009/*
3010 * xfs_bmap_alloc is called by xfs_bmapi to allocate an extent for a file.
3011 * It figures out where to ask the underlying allocator to put the new extent.
3012 */
3013STATIC int
3014xfs_bmap_alloc(
3015 xfs_bmalloca_t *ap) /* bmap alloc argument struct */
3016{
71ddabb9 3017 if (XFS_IS_REALTIME_INODE(ap->ip) && ap->userdata)
a365bdd5
NS
3018 return xfs_bmap_rtalloc(ap);
3019 return xfs_bmap_btalloc(ap);
1da177e4
LT
3020}
3021
3022/*
3023 * Transform a btree format file with only one leaf node, where the
3024 * extents list will fit in the inode, into an extents format file.
4eea22f0 3025 * Since the file extents are already in-core, all we have to do is
1da177e4
LT
3026 * give up the space for the btree root and pitch the leaf block.
3027 */
3028STATIC int /* error */
3029xfs_bmap_btree_to_extents(
3030 xfs_trans_t *tp, /* transaction pointer */
3031 xfs_inode_t *ip, /* incore inode pointer */
3032 xfs_btree_cur_t *cur, /* btree cursor */
3033 int *logflagsp, /* inode logging flags */
3034 int whichfork) /* data or attr fork */
3035{
3036 /* REFERENCED */
7cc95a82 3037 struct xfs_btree_block *cblock;/* child btree block */
1da177e4
LT
3038 xfs_fsblock_t cbno; /* child block number */
3039 xfs_buf_t *cbp; /* child block's buffer */
3040 int error; /* error return value */
3041 xfs_ifork_t *ifp; /* inode fork data */
3042 xfs_mount_t *mp; /* mount point structure */
576039cf 3043 __be64 *pp; /* ptr to block address */
7cc95a82 3044 struct xfs_btree_block *rblock;/* root btree block */
1da177e4 3045
60197e8d 3046 mp = ip->i_mount;
1da177e4
LT
3047 ifp = XFS_IFORK_PTR(ip, whichfork);
3048 ASSERT(ifp->if_flags & XFS_IFEXTENTS);
3049 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
3050 rblock = ifp->if_broot;
16259e7d
CH
3051 ASSERT(be16_to_cpu(rblock->bb_level) == 1);
3052 ASSERT(be16_to_cpu(rblock->bb_numrecs) == 1);
60197e8d
CH
3053 ASSERT(xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0) == 1);
3054 pp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, ifp->if_broot_bytes);
576039cf 3055 cbno = be64_to_cpu(*pp);
1da177e4
LT
3056 *logflagsp = 0;
3057#ifdef DEBUG
576039cf 3058 if ((error = xfs_btree_check_lptr(cur, cbno, 1)))
1da177e4
LT
3059 return error;
3060#endif
1da177e4
LT
3061 if ((error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp,
3062 XFS_BMAP_BTREE_REF)))
3063 return error;
7cc95a82
CH
3064 cblock = XFS_BUF_TO_BLOCK(cbp);
3065 if ((error = xfs_btree_check_block(cur, cblock, 0, cbp)))
1da177e4
LT
3066 return error;
3067 xfs_bmap_add_free(cbno, 1, cur->bc_private.b.flist, mp);
3068 ip->i_d.di_nblocks--;
3069 XFS_TRANS_MOD_DQUOT_BYINO(mp, tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
3070 xfs_trans_binval(tp, cbp);
3071 if (cur->bc_bufs[0] == cbp)
3072 cur->bc_bufs[0] = NULL;
3073 xfs_iroot_realloc(ip, -1, whichfork);
3074 ASSERT(ifp->if_broot == NULL);
3075 ASSERT((ifp->if_flags & XFS_IFBROOT) == 0);
3076 XFS_IFORK_FMT_SET(ip, whichfork, XFS_DINODE_FMT_EXTENTS);
9d87c319 3077 *logflagsp = XFS_ILOG_CORE | xfs_ilog_fext(whichfork);
1da177e4
LT
3078 return 0;
3079}
3080
3081/*
4eea22f0 3082 * Called by xfs_bmapi to update file extent records and the btree
1da177e4
LT
3083 * after removing space (or undoing a delayed allocation).
3084 */
3085STATIC int /* error */
3086xfs_bmap_del_extent(
3087 xfs_inode_t *ip, /* incore inode pointer */
3088 xfs_trans_t *tp, /* current transaction pointer */
3089 xfs_extnum_t idx, /* extent number to update/delete */
3090 xfs_bmap_free_t *flist, /* list of extents to be freed */
3091 xfs_btree_cur_t *cur, /* if null, not a btree */
4eea22f0 3092 xfs_bmbt_irec_t *del, /* data to remove from extents */
1da177e4 3093 int *logflagsp, /* inode logging flags */
3e57ecf6 3094 xfs_extdelta_t *delta, /* Change made to incore extents */
1da177e4
LT
3095 int whichfork, /* data or attr fork */
3096 int rsvd) /* OK to allocate reserved blocks */
3097{
3098 xfs_filblks_t da_new; /* new delay-alloc indirect blocks */
3099 xfs_filblks_t da_old; /* old delay-alloc indirect blocks */
3100 xfs_fsblock_t del_endblock=0; /* first block past del */
3101 xfs_fileoff_t del_endoff; /* first offset past del */
3102 int delay; /* current block is delayed allocated */
3103 int do_fx; /* free extent at end of routine */
a6f64d4a 3104 xfs_bmbt_rec_host_t *ep; /* current extent entry pointer */
1da177e4
LT
3105 int error; /* error return value */
3106 int flags; /* inode logging flags */
1da177e4
LT
3107 xfs_bmbt_irec_t got; /* current extent entry */
3108 xfs_fileoff_t got_endoff; /* first offset past got */
3109 int i; /* temp state */
3110 xfs_ifork_t *ifp; /* inode fork pointer */
3111 xfs_mount_t *mp; /* mount structure */
3112 xfs_filblks_t nblks; /* quota/sb block count */
3113 xfs_bmbt_irec_t new; /* new record to be inserted */
3114 /* REFERENCED */
1da177e4
LT
3115 uint qfield; /* quota field to update */
3116 xfs_filblks_t temp; /* for indirect length calculations */
3117 xfs_filblks_t temp2; /* for indirect length calculations */
3118
3119 XFS_STATS_INC(xs_del_exlist);
3120 mp = ip->i_mount;
3121 ifp = XFS_IFORK_PTR(ip, whichfork);
4eea22f0
MK
3122 ASSERT((idx >= 0) && (idx < ifp->if_bytes /
3123 (uint)sizeof(xfs_bmbt_rec_t)));
1da177e4 3124 ASSERT(del->br_blockcount > 0);
4eea22f0 3125 ep = xfs_iext_get_ext(ifp, idx);
1da177e4
LT
3126 xfs_bmbt_get_all(ep, &got);
3127 ASSERT(got.br_startoff <= del->br_startoff);
3128 del_endoff = del->br_startoff + del->br_blockcount;
3129 got_endoff = got.br_startoff + got.br_blockcount;
3130 ASSERT(got_endoff >= del_endoff);
9d87c319
ES
3131 delay = isnullstartblock(got.br_startblock);
3132 ASSERT(isnullstartblock(del->br_startblock) == delay);
1da177e4
LT
3133 flags = 0;
3134 qfield = 0;
3135 error = 0;
3136 /*
3137 * If deleting a real allocation, must free up the disk space.
3138 */
3139 if (!delay) {
3140 flags = XFS_ILOG_CORE;
3141 /*
3142 * Realtime allocation. Free it and record di_nblocks update.
3143 */
71ddabb9 3144 if (whichfork == XFS_DATA_FORK && XFS_IS_REALTIME_INODE(ip)) {
1da177e4
LT
3145 xfs_fsblock_t bno;
3146 xfs_filblks_t len;
3147
3148 ASSERT(do_mod(del->br_blockcount,
3149 mp->m_sb.sb_rextsize) == 0);
3150 ASSERT(do_mod(del->br_startblock,
3151 mp->m_sb.sb_rextsize) == 0);
3152 bno = del->br_startblock;
3153 len = del->br_blockcount;
3154 do_div(bno, mp->m_sb.sb_rextsize);
3155 do_div(len, mp->m_sb.sb_rextsize);
3156 if ((error = xfs_rtfree_extent(ip->i_transp, bno,
3157 (xfs_extlen_t)len)))
3158 goto done;
3159 do_fx = 0;
3160 nblks = len * mp->m_sb.sb_rextsize;
3161 qfield = XFS_TRANS_DQ_RTBCOUNT;
3162 }
3163 /*
3164 * Ordinary allocation.
3165 */
3166 else {
3167 do_fx = 1;
3168 nblks = del->br_blockcount;
3169 qfield = XFS_TRANS_DQ_BCOUNT;
3170 }
3171 /*
3172 * Set up del_endblock and cur for later.
3173 */
3174 del_endblock = del->br_startblock + del->br_blockcount;
3175 if (cur) {
3176 if ((error = xfs_bmbt_lookup_eq(cur, got.br_startoff,
3177 got.br_startblock, got.br_blockcount,
3178 &i)))
3179 goto done;
6bd8fc8a 3180 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
3181 }
3182 da_old = da_new = 0;
3183 } else {
9d87c319 3184 da_old = startblockval(got.br_startblock);
1da177e4
LT
3185 da_new = 0;
3186 nblks = 0;
3187 do_fx = 0;
3188 }
3189 /*
3190 * Set flag value to use in switch statement.
3191 * Left-contig is 2, right-contig is 1.
3192 */
3193 switch (((got.br_startoff == del->br_startoff) << 1) |
3194 (got_endoff == del_endoff)) {
3195 case 3:
3196 /*
3197 * Matches the whole extent. Delete the entry.
3198 */
3a59c94c 3199 XFS_BMAP_TRACE_DELETE("3", ip, idx, 1, whichfork);
4eea22f0 3200 xfs_iext_remove(ifp, idx, 1);
1da177e4
LT
3201 ifp->if_lastex = idx;
3202 if (delay)
3203 break;
3204 XFS_IFORK_NEXT_SET(ip, whichfork,
3205 XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
3206 flags |= XFS_ILOG_CORE;
3207 if (!cur) {
9d87c319 3208 flags |= xfs_ilog_fext(whichfork);
1da177e4
LT
3209 break;
3210 }
91cca5df 3211 if ((error = xfs_btree_delete(cur, &i)))
1da177e4 3212 goto done;
6bd8fc8a 3213 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
3214 break;
3215
3216 case 2:
3217 /*
3218 * Deleting the first part of the extent.
3219 */
3a59c94c 3220 XFS_BMAP_TRACE_PRE_UPDATE("2", ip, idx, whichfork);
1da177e4
LT
3221 xfs_bmbt_set_startoff(ep, del_endoff);
3222 temp = got.br_blockcount - del->br_blockcount;
3223 xfs_bmbt_set_blockcount(ep, temp);
3224 ifp->if_lastex = idx;
3225 if (delay) {
3226 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
3227 da_old);
9d87c319 3228 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 3229 XFS_BMAP_TRACE_POST_UPDATE("2", ip, idx,
1da177e4
LT
3230 whichfork);
3231 da_new = temp;
3232 break;
3233 }
3234 xfs_bmbt_set_startblock(ep, del_endblock);
3a59c94c 3235 XFS_BMAP_TRACE_POST_UPDATE("2", ip, idx, whichfork);
1da177e4 3236 if (!cur) {
9d87c319 3237 flags |= xfs_ilog_fext(whichfork);
1da177e4
LT
3238 break;
3239 }
3240 if ((error = xfs_bmbt_update(cur, del_endoff, del_endblock,
3241 got.br_blockcount - del->br_blockcount,
3242 got.br_state)))
3243 goto done;
3244 break;
3245
3246 case 1:
3247 /*
3248 * Deleting the last part of the extent.
3249 */
3250 temp = got.br_blockcount - del->br_blockcount;
3a59c94c 3251 XFS_BMAP_TRACE_PRE_UPDATE("1", ip, idx, whichfork);
1da177e4
LT
3252 xfs_bmbt_set_blockcount(ep, temp);
3253 ifp->if_lastex = idx;
3254 if (delay) {
3255 temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
3256 da_old);
9d87c319 3257 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
3a59c94c 3258 XFS_BMAP_TRACE_POST_UPDATE("1", ip, idx,
1da177e4
LT
3259 whichfork);
3260 da_new = temp;
3261 break;
3262 }
3a59c94c 3263 XFS_BMAP_TRACE_POST_UPDATE("1", ip, idx, whichfork);
1da177e4 3264 if (!cur) {
9d87c319 3265 flags |= xfs_ilog_fext(whichfork);
1da177e4
LT
3266 break;
3267 }
3268 if ((error = xfs_bmbt_update(cur, got.br_startoff,
3269 got.br_startblock,
3270 got.br_blockcount - del->br_blockcount,
3271 got.br_state)))
3272 goto done;
3273 break;
3274
3275 case 0:
3276 /*
3277 * Deleting the middle of the extent.
3278 */
3279 temp = del->br_startoff - got.br_startoff;
3a59c94c 3280 XFS_BMAP_TRACE_PRE_UPDATE("0", ip, idx, whichfork);
1da177e4
LT
3281 xfs_bmbt_set_blockcount(ep, temp);
3282 new.br_startoff = del_endoff;
3283 temp2 = got_endoff - del_endoff;
3284 new.br_blockcount = temp2;
3285 new.br_state = got.br_state;
3286 if (!delay) {
3287 new.br_startblock = del_endblock;
3288 flags |= XFS_ILOG_CORE;
3289 if (cur) {
3290 if ((error = xfs_bmbt_update(cur,
3291 got.br_startoff,
3292 got.br_startblock, temp,
3293 got.br_state)))
3294 goto done;
637aa50f 3295 if ((error = xfs_btree_increment(cur, 0, &i)))
1da177e4
LT
3296 goto done;
3297 cur->bc_rec.b = new;
4b22a571 3298 error = xfs_btree_insert(cur, &i);
1da177e4
LT
3299 if (error && error != ENOSPC)
3300 goto done;
3301 /*
3302 * If get no-space back from btree insert,
3303 * it tried a split, and we have a zero
3304 * block reservation.
3305 * Fix up our state and return the error.
3306 */
3307 if (error == ENOSPC) {
3308 /*
3309 * Reset the cursor, don't trust
3310 * it after any insert operation.
3311 */
3312 if ((error = xfs_bmbt_lookup_eq(cur,
3313 got.br_startoff,
3314 got.br_startblock,
3315 temp, &i)))
3316 goto done;
6bd8fc8a 3317 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4
LT
3318 /*
3319 * Update the btree record back
3320 * to the original value.
3321 */
3322 if ((error = xfs_bmbt_update(cur,
3323 got.br_startoff,
3324 got.br_startblock,
3325 got.br_blockcount,
3326 got.br_state)))
3327 goto done;
3328 /*
3329 * Reset the extent record back
3330 * to the original value.
3331 */
3332 xfs_bmbt_set_blockcount(ep,
3333 got.br_blockcount);
3334 flags = 0;
3335 error = XFS_ERROR(ENOSPC);
3336 goto done;
3337 }
6bd8fc8a 3338 XFS_WANT_CORRUPTED_GOTO(i == 1, done);
1da177e4 3339 } else
9d87c319 3340 flags |= xfs_ilog_fext(whichfork);
1da177e4
LT
3341 XFS_IFORK_NEXT_SET(ip, whichfork,
3342 XFS_IFORK_NEXTENTS(ip, whichfork) + 1);
3343 } else {
3344 ASSERT(whichfork == XFS_DATA_FORK);
3345 temp = xfs_bmap_worst_indlen(ip, temp);
9d87c319 3346 xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
1da177e4 3347 temp2 = xfs_bmap_worst_indlen(ip, temp2);
9d87c319 3348 new.br_startblock = nullstartblock((int)temp2);
1da177e4
LT
3349 da_new = temp + temp2;
3350 while (da_new > da_old) {
3351 if (temp) {
3352 temp--;
3353 da_new--;
3354 xfs_bmbt_set_startblock(ep,
9d87c319 3355 nullstartblock((int)temp));
1da177e4
LT
3356 }
3357 if (da_new == da_old)
3358 break;
3359 if (temp2) {
3360 temp2--;
3361 da_new--;
3362 new.br_startblock =
9d87c319 3363 nullstartblock((int)temp2);
1da177e4
LT
3364 }
3365 }
3366 }
3a59c94c
ES
3367 XFS_BMAP_TRACE_POST_UPDATE("0", ip, idx, whichfork);
3368 XFS_BMAP_TRACE_INSERT("0", ip, idx + 1, 1, &new, NULL,
1da177e4 3369 whichfork);
4eea22f0 3370 xfs_iext_insert(ifp, idx + 1, 1, &new);
1da177e4
LT
3371 ifp->if_lastex = idx + 1;
3372 break;
3373 }
3374 /*
3375 * If we need to, add to list of extents to delete.
3376 */
3377 if (do_fx)
3378 xfs_bmap_add_free(del->br_startblock, del->br_blockcount, flist,
3379 mp);
3380 /*
3381 * Adjust inode # blocks in the file.
3382 */
3383 if (nblks)
3384 ip->i_d.di_nblocks -= nblks;
3385 /*
3386 * Adjust quota data.
3387 */
3388 if (qfield)
3389 XFS_TRANS_MOD_DQUOT_BYINO(mp, tp, ip, qfield, (long)-nblks);
3390
3391 /*
3392 * Account for change in delayed indirect blocks.
3393 * Nothing to do for disk quota accounting here.
3394 */
3395 ASSERT(da_old >= da_new);
3396 if (da_old > da_new)
20f4ebf2 3397 xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, (int64_t)(da_old - da_new),
1da177e4 3398 rsvd);
3e57ecf6
OW
3399 if (delta) {
3400 /* DELTA: report the original extent. */
3401 if (delta->xed_startoff > got.br_startoff)
3402 delta->xed_startoff = got.br_startoff;
3403 if (delta->xed_blockcount < got.br_startoff+got.br_blockcount)
3404 delta->xed_blockcount = got.br_startoff +
3405 got.br_blockcount;
3406 }
1da177e4
LT
3407done:
3408 *logflagsp = flags;
3409 return error;
3410}
3411
3412/*
3413 * Remove the entry "free" from the free item list. Prev points to the
3414 * previous entry, unless "free" is the head of the list.
3415 */
3416STATIC void
3417xfs_bmap_del_free(
3418 xfs_bmap_free_t *flist, /* free item list header */
3419 xfs_bmap_free_item_t *prev, /* previous item on list, if any */
3420 xfs_bmap_free_item_t *free) /* list item to be freed */
3421{
3422 if (prev)
3423 prev->xbfi_next = free->xbfi_next;
3424 else
3425 flist->xbf_first = free->xbfi_next;
3426 flist->xbf_count--;
3427 kmem_zone_free(xfs_bmap_free_item_zone, free);
3428}
3429
1da177e4
LT
3430/*
3431 * Convert an extents-format file into a btree-format file.
3432 * The new file will have a root block (in the inode) and a single child block.
3433 */
3434STATIC int /* error */
3435xfs_bmap_extents_to_btree(
3436 xfs_trans_t *tp, /* transaction pointer */
3437 xfs_inode_t *ip, /* incore inode pointer */
3438 xfs_fsblock_t *firstblock, /* first-block-allocated */
3439 xfs_bmap_free_t *flist, /* blocks freed in xaction */
3440 xfs_btree_cur_t **curp, /* cursor returned to caller */
3441 int wasdel, /* converting a delayed alloc */
3442 int *logflagsp, /* inode logging flags */
3443 int whichfork) /* data or attr fork */
3444{
7cc95a82 3445 struct xfs_btree_block *ablock; /* allocated (child) bt block */
1da177e4
LT
3446 xfs_buf_t *abp; /* buffer for ablock */
3447 xfs_alloc_arg_t args; /* allocation arguments */
3448 xfs_bmbt_rec_t *arp; /* child record pointer */
7cc95a82 3449 struct xfs_btree_block *block; /* btree root block */
1da177e4 3450 xfs_btree_cur_t *cur; /* bmap btree cursor */
a6f64d4a 3451 xfs_bmbt_rec_host_t *ep; /* extent record pointer */
1da177e4 3452 int error; /* error return value */
4eea22f0 3453 xfs_extnum_t i, cnt; /* extent record index */
1da177e4
LT
3454 xfs_ifork_t *ifp; /* inode fork pointer */
3455 xfs_bmbt_key_t *kp; /* root block key pointer */
3456 xfs_mount_t *mp; /* mount structure */
4eea22f0 3457 xfs_extnum_t nextents; /* number of file extents */
1da177e4
LT
3458 xfs_bmbt_ptr_t *pp; /* root block address pointer */
3459
3460 ifp = XFS_IFORK_PTR(ip, whichfork);
3461 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS);
3462 ASSERT(ifp->if_ext_max ==
3463 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
3464 /*
3465 * Make space in the inode incore.
3466 */
3467 xfs_iroot_realloc(ip, 1, whichfork);
3468 ifp->if_flags |= XFS_IFBROOT;
7cc95a82 3469
1da177e4
LT
3470 /*
3471 * Fill in the root.
3472 */
3473 block = ifp->if_broot;
16259e7d
CH
3474 block->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
3475 block->bb_level = cpu_to_be16(1);
3476 block->bb_numrecs = cpu_to_be16(1);
7cc95a82
CH
3477 block->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
3478 block->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
3479
1da177e4
LT
3480 /*
3481 * Need a cursor. Can't allocate until bb_level is filled in.
3482 */
3483 mp = ip->i_mount;
561f7d17 3484 cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
1da177e4
LT
3485 cur->bc_private.b.firstblock = *firstblock;
3486 cur->bc_private.b.flist = flist;
3487 cur->bc_private.b.flags = wasdel ? XFS_BTCUR_BPRV_WASDEL : 0;
3488 /*
3489 * Convert to a btree with two levels, one record in root.
3490 */
3491 XFS_IFORK_FMT_SET(ip, whichfork, XFS_DINODE_FMT_BTREE);
3492 args.tp = tp;
3493 args.mp = mp;
d210a28c 3494 args.firstblock = *firstblock;
1da177e4
LT
3495 if (*firstblock == NULLFSBLOCK) {
3496 args.type = XFS_ALLOCTYPE_START_BNO;
3497 args.fsbno = XFS_INO_TO_FSB(mp, ip->i_ino);
3498 } else if (flist->xbf_low) {
3499 args.type = XFS_ALLOCTYPE_START_BNO;
3500 args.fsbno = *firstblock;
3501 } else {
3502 args.type = XFS_ALLOCTYPE_NEAR_BNO;
3503 args.fsbno = *firstblock;
3504 }
3505 args.minlen = args.maxlen = args.prod = 1;
3506 args.total = args.minleft = args.alignment = args.mod = args.isfl =
3507 args.minalignslop = 0;
3508 args.wasdel = wasdel;
3509 *logflagsp = 0;
3510 if ((error = xfs_alloc_vextent(&args))) {
3511 xfs_iroot_realloc(ip, -1, whichfork);
3512 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
3513 return error;
3514 }
3515 /*
3516 * Allocation can't fail, the space was reserved.
3517 */
3518 ASSERT(args.fsbno != NULLFSBLOCK);
3519 ASSERT(*firstblock == NULLFSBLOCK ||
3520 args.agno == XFS_FSB_TO_AGNO(mp, *firstblock) ||
3521 (flist->xbf_low &&
3522 args.agno > XFS_FSB_TO_AGNO(mp, *firstblock)));
3523 *firstblock = cur->bc_private.b.firstblock = args.fsbno;
3524 cur->bc_private.b.allocated++;
3525 ip->i_d.di_nblocks++;
3526 XFS_TRANS_MOD_DQUOT_BYINO(mp, tp, ip, XFS_TRANS_DQ_BCOUNT, 1L);
3527 abp = xfs_btree_get_bufl(mp, tp, args.fsbno, 0);
3528 /*
3529 * Fill in the child block.
3530 */
7cc95a82 3531 ablock = XFS_BUF_TO_BLOCK(abp);
16259e7d 3532 ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
1da177e4 3533 ablock->bb_level = 0;
7cc95a82
CH
3534 ablock->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
3535 ablock->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
136341b4 3536 arp = XFS_BMBT_REC_ADDR(mp, ablock, 1);
1da177e4 3537 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
4eea22f0
MK
3538 for (cnt = i = 0; i < nextents; i++) {
3539 ep = xfs_iext_get_ext(ifp, i);
9d87c319 3540 if (!isnullstartblock(xfs_bmbt_get_startblock(ep))) {
cd8b0a97
CH
3541 arp->l0 = cpu_to_be64(ep->l0);
3542 arp->l1 = cpu_to_be64(ep->l1);
1da177e4
LT
3543 arp++; cnt++;
3544 }
3545 }
16259e7d 3546 ASSERT(cnt == XFS_IFORK_NEXTENTS(ip, whichfork));
7cc95a82
CH
3547 xfs_btree_set_numrecs(ablock, cnt);
3548
1da177e4
LT
3549 /*
3550 * Fill in the root key and pointer.
3551 */
136341b4
CH
3552 kp = XFS_BMBT_KEY_ADDR(mp, block, 1);
3553 arp = XFS_BMBT_REC_ADDR(mp, ablock, 1);
8801bb99 3554 kp->br_startoff = cpu_to_be64(xfs_bmbt_disk_get_startoff(arp));
136341b4
CH
3555 pp = XFS_BMBT_PTR_ADDR(mp, block, 1, xfs_bmbt_get_maxrecs(cur,
3556 be16_to_cpu(block->bb_level)));
576039cf 3557 *pp = cpu_to_be64(args.fsbno);
136341b4 3558
1da177e4
LT
3559 /*
3560 * Do all this logging at the end so that
3561 * the root is at the right level.
3562 */
fd6bcc5b
CH
3563 xfs_btree_log_block(cur, abp, XFS_BB_ALL_BITS);
3564 xfs_btree_log_recs(cur, abp, 1, be16_to_cpu(ablock->bb_numrecs));
1da177e4
LT
3565 ASSERT(*curp == NULL);
3566 *curp = cur;
9d87c319 3567 *logflagsp = XFS_ILOG_CORE | xfs_ilog_fbroot(whichfork);
1da177e4
LT
3568 return 0;
3569}
3570
d8cc890d
NS
3571/*
3572 * Helper routine to reset inode di_forkoff field when switching
3573 * attribute fork from local to extent format - we reset it where
3574 * possible to make space available for inline data fork extents.
3575 */
3576STATIC void
3577xfs_bmap_forkoff_reset(
3578 xfs_mount_t *mp,
3579 xfs_inode_t *ip,
3580 int whichfork)
3581{
3582 if (whichfork == XFS_ATTR_FORK &&
3583 (ip->i_d.di_format != XFS_DINODE_FMT_DEV) &&
3584 (ip->i_d.di_format != XFS_DINODE_FMT_UUID) &&
e5889e90 3585 (ip->i_d.di_format != XFS_DINODE_FMT_BTREE) &&
d8cc890d
NS
3586 ((mp->m_attroffset >> 3) > ip->i_d.di_forkoff)) {
3587 ip->i_d.di_forkoff = mp->m_attroffset >> 3;
3588 ip->i_df.if_ext_max = XFS_IFORK_DSIZE(ip) /
3589 (uint)sizeof(xfs_bmbt_rec_t);
3590 ip->i_afp->if_ext_max = XFS_IFORK_ASIZE(ip) /
3591 (uint)sizeof(xfs_bmbt_rec_t);
3592 }
3593}
3594
1da177e4
LT
3595/*
3596 * Convert a local file to an extents file.
3597 * This code is out of bounds for data forks of regular files,
3598 * since the file data needs to get logged so things will stay consistent.
3599 * (The bmap-level manipulations are ok, though).
3600 */
3601STATIC int /* error */
3602xfs_bmap_local_to_extents(
3603 xfs_trans_t *tp, /* transaction pointer */
3604 xfs_inode_t *ip, /* incore inode pointer */
3605 xfs_fsblock_t *firstblock, /* first block allocated in xaction */
3606 xfs_extlen_t total, /* total blocks needed by transaction */
3607 int *logflagsp, /* inode logging flags */
3608 int whichfork) /* data or attr fork */
3609{
3610 int error; /* error return value */
3611 int flags; /* logging flags returned */
1da177e4
LT
3612 xfs_ifork_t *ifp; /* inode fork pointer */
3613
3614 /*
3615 * We don't want to deal with the case of keeping inode data inline yet.
3616 * So sending the data fork of a regular inode is invalid.
3617 */
3618 ASSERT(!((ip->i_d.di_mode & S_IFMT) == S_IFREG &&
3619 whichfork == XFS_DATA_FORK));
3620 ifp = XFS_IFORK_PTR(ip, whichfork);
3621 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL);
3622 flags = 0;
3623 error = 0;
3624 if (ifp->if_bytes) {
3625 xfs_alloc_arg_t args; /* allocation arguments */
4eea22f0 3626 xfs_buf_t *bp; /* buffer for extent block */
a6f64d4a 3627 xfs_bmbt_rec_host_t *ep;/* extent record pointer */
1da177e4
LT
3628
3629 args.tp = tp;
3630 args.mp = ip->i_mount;
d210a28c 3631 args.firstblock = *firstblock;
f020b67f
MK
3632 ASSERT((ifp->if_flags &
3633 (XFS_IFINLINE|XFS_IFEXTENTS|XFS_IFEXTIREC)) == XFS_IFINLINE);
1da177e4
LT
3634 /*
3635 * Allocate a block. We know we need only one, since the
3636 * file currently fits in an inode.
3637 */
3638 if (*firstblock == NULLFSBLOCK) {
3639 args.fsbno = XFS_INO_TO_FSB(args.mp, ip->i_ino);
3640 args.type = XFS_ALLOCTYPE_START_BNO;
3641 } else {
3642 args.fsbno = *firstblock;
3643 args.type = XFS_ALLOCTYPE_NEAR_BNO;
3644 }
3645 args.total = total;
3646 args.mod = args.minleft = args.alignment = args.wasdel =
3647 args.isfl = args.minalignslop = 0;
3648 args.minlen = args.maxlen = args.prod = 1;
3649 if ((error = xfs_alloc_vextent(&args)))
3650 goto done;
3651 /*
3652 * Can't fail, the space was reserved.
3653 */
3654 ASSERT(args.fsbno != NULLFSBLOCK);
3655 ASSERT(args.len == 1);
3656 *firstblock = args.fsbno;
3657 bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
3658 memcpy((char *)XFS_BUF_PTR(bp), ifp->if_u1.if_data,
3659 ifp->if_bytes);
3660 xfs_trans_log_buf(tp, bp, 0, ifp->if_bytes - 1);
d8cc890d 3661 xfs_bmap_forkoff_reset(args.mp, ip, whichfork);
1da177e4 3662 xfs_idata_realloc(ip, -ifp->if_bytes, whichfork);
4eea22f0
MK
3663 xfs_iext_add(ifp, 0, 1);
3664 ep = xfs_iext_get_ext(ifp, 0);
1da177e4 3665 xfs_bmbt_set_allf(ep, 0, args.fsbno, 1, XFS_EXT_NORM);
3a59c94c 3666 XFS_BMAP_TRACE_POST_UPDATE("new", ip, 0, whichfork);
1da177e4
LT
3667 XFS_IFORK_NEXT_SET(ip, whichfork, 1);
3668 ip->i_d.di_nblocks = 1;
3669 XFS_TRANS_MOD_DQUOT_BYINO(args.mp, tp, ip,
3670 XFS_TRANS_DQ_BCOUNT, 1L);
9d87c319 3671 flags |= xfs_ilog_fext(whichfork);
d8cc890d 3672 } else {
1da177e4 3673 ASSERT(XFS_IFORK_NEXTENTS(ip, whichfork) == 0);
d8cc890d
NS
3674 xfs_bmap_forkoff_reset(ip->i_mount, ip, whichfork);
3675 }
1da177e4
LT
3676 ifp->if_flags &= ~XFS_IFINLINE;
3677 ifp->if_flags |= XFS_IFEXTENTS;
3678 XFS_IFORK_FMT_SET(ip, whichfork, XFS_DINODE_FMT_EXTENTS);
3679 flags |= XFS_ILOG_CORE;
3680done:
3681 *logflagsp = flags;
3682 return error;
3683}
3684
0293ce3a 3685/*
8867bc9b
MK
3686 * Search the extent records for the entry containing block bno.
3687 * If bno lies in a hole, point to the next entry. If bno lies
3688 * past eof, *eofp will be set, and *prevp will contain the last
3689 * entry (null if none). Else, *lastxp will be set to the index
3690 * of the found entry; *gotp will contain the entry.
0293ce3a 3691 */
a6f64d4a 3692xfs_bmbt_rec_host_t * /* pointer to found extent entry */
0293ce3a
MK
3693xfs_bmap_search_multi_extents(
3694 xfs_ifork_t *ifp, /* inode fork pointer */
3695 xfs_fileoff_t bno, /* block number searched for */
3696 int *eofp, /* out: end of file found */
3697 xfs_extnum_t *lastxp, /* out: last extent index */
3698 xfs_bmbt_irec_t *gotp, /* out: extent entry found */
3699 xfs_bmbt_irec_t *prevp) /* out: previous extent entry found */
3700{
a6f64d4a 3701 xfs_bmbt_rec_host_t *ep; /* extent record pointer */
0293ce3a 3702 xfs_extnum_t lastx; /* last extent index */
0293ce3a
MK
3703
3704 /*
8867bc9b
MK
3705 * Initialize the extent entry structure to catch access to
3706 * uninitialized br_startblock field.
0293ce3a 3707 */
8867bc9b
MK
3708 gotp->br_startoff = 0xffa5a5a5a5a5a5a5LL;
3709 gotp->br_blockcount = 0xa55a5a5a5a5a5a5aLL;
3710 gotp->br_state = XFS_EXT_INVALID;
3711#if XFS_BIG_BLKNOS
3712 gotp->br_startblock = 0xffffa5a5a5a5a5a5LL;
3713#else
3714 gotp->br_startblock = 0xffffa5a5;
3715#endif
3716 prevp->br_startoff = NULLFILEOFF;
3717
3718 ep = xfs_iext_bno_to_ext(ifp, bno, &lastx);
3719 if (lastx > 0) {
3720 xfs_bmbt_get_all(xfs_iext_get_ext(ifp, lastx - 1), prevp);
0293ce3a 3721 }
8867bc9b
MK
3722 if (lastx < (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t))) {
3723 xfs_bmbt_get_all(ep, gotp);
3724 *eofp = 0;
3725 } else {
3726 if (lastx > 0) {
3727 *gotp = *prevp;
3728 }
3729 *eofp = 1;
3730 ep = NULL;
0293ce3a 3731 }
8867bc9b 3732 *lastxp = lastx;
0293ce3a
MK
3733 return ep;
3734}
3735
1da177e4
LT
3736/*
3737 * Search the extents list for the inode, for the extent containing bno.
3738 * If bno lies in a hole, point to the next entry. If bno lies past eof,
3739 * *eofp will be set, and *prevp will contain the last entry (null if none).
3740 * Else, *lastxp will be set to the index of the found
3741 * entry; *gotp will contain the entry.
3742 */
a6f64d4a 3743STATIC xfs_bmbt_rec_host_t * /* pointer to found extent entry */
1da177e4
LT
3744xfs_bmap_search_extents(
3745 xfs_inode_t *ip, /* incore inode pointer */
3746 xfs_fileoff_t bno, /* block number searched for */
572d95f4 3747 int fork, /* data or attr fork */
1da177e4
LT
3748 int *eofp, /* out: end of file found */
3749 xfs_extnum_t *lastxp, /* out: last extent index */
3750 xfs_bmbt_irec_t *gotp, /* out: extent entry found */
3751 xfs_bmbt_irec_t *prevp) /* out: previous extent entry found */
3752{
3753 xfs_ifork_t *ifp; /* inode fork pointer */
a6f64d4a 3754 xfs_bmbt_rec_host_t *ep; /* extent record pointer */
1da177e4
LT
3755
3756 XFS_STATS_INC(xs_look_exlist);
572d95f4 3757 ifp = XFS_IFORK_PTR(ip, fork);
1da177e4 3758
0293ce3a
MK
3759 ep = xfs_bmap_search_multi_extents(ifp, bno, eofp, lastxp, gotp, prevp);
3760
572d95f4
NS
3761 if (unlikely(!(gotp->br_startblock) && (*lastxp != NULLEXTNUM) &&
3762 !(XFS_IS_REALTIME_INODE(ip) && fork == XFS_DATA_FORK))) {
3763 xfs_cmn_err(XFS_PTAG_FSBLOCK_ZERO, CE_ALERT, ip->i_mount,
3764 "Access to block zero in inode %llu "
3765 "start_block: %llx start_off: %llx "
3766 "blkcnt: %llx extent-state: %x lastx: %x\n",
3767 (unsigned long long)ip->i_ino,
3ddb8fa9
NS
3768 (unsigned long long)gotp->br_startblock,
3769 (unsigned long long)gotp->br_startoff,
3770 (unsigned long long)gotp->br_blockcount,
572d95f4
NS
3771 gotp->br_state, *lastxp);
3772 *lastxp = NULLEXTNUM;
3773 *eofp = 1;
3774 return NULL;
3775 }
3776 return ep;
1da177e4
LT
3777}
3778
3779
3780#ifdef XFS_BMAP_TRACE
3781ktrace_t *xfs_bmap_trace_buf;
3782
3783/*
3784 * Add a bmap trace buffer entry. Base routine for the others.
3785 */
3786STATIC void
3787xfs_bmap_trace_addentry(
3788 int opcode, /* operation */
3a59c94c 3789 const char *fname, /* function name */
1da177e4
LT
3790 char *desc, /* operation description */
3791 xfs_inode_t *ip, /* incore inode pointer */
3792 xfs_extnum_t idx, /* index of entry(ies) */
3793 xfs_extnum_t cnt, /* count of entries, 1 or 2 */
a6f64d4a
CH
3794 xfs_bmbt_rec_host_t *r1, /* first record */
3795 xfs_bmbt_rec_host_t *r2, /* second record or null */
1da177e4
LT
3796 int whichfork) /* data or attr fork */
3797{
a6f64d4a 3798 xfs_bmbt_rec_host_t tr2;
1da177e4
LT
3799
3800 ASSERT(cnt == 1 || cnt == 2);
3801 ASSERT(r1 != NULL);
3802 if (cnt == 1) {
3803 ASSERT(r2 == NULL);
3804 r2 = &tr2;
3805 memset(&tr2, 0, sizeof(tr2));
3806 } else
3807 ASSERT(r2 != NULL);
3808 ktrace_enter(xfs_bmap_trace_buf,
3809 (void *)(__psint_t)(opcode | (whichfork << 16)),
3810 (void *)fname, (void *)desc, (void *)ip,
3811 (void *)(__psint_t)idx,
3812 (void *)(__psint_t)cnt,
3813 (void *)(__psunsigned_t)(ip->i_ino >> 32),
3814 (void *)(__psunsigned_t)(unsigned)ip->i_ino,
3815 (void *)(__psunsigned_t)(r1->l0 >> 32),
3816 (void *)(__psunsigned_t)(unsigned)(r1->l0),
3817 (void *)(__psunsigned_t)(r1->l1 >> 32),
3818 (void *)(__psunsigned_t)(unsigned)(r1->l1),
3819 (void *)(__psunsigned_t)(r2->l0 >> 32),
3820 (void *)(__psunsigned_t)(unsigned)(r2->l0),
3821 (void *)(__psunsigned_t)(r2->l1 >> 32),
3822 (void *)(__psunsigned_t)(unsigned)(r2->l1)
3823 );
3824 ASSERT(ip->i_xtrace);
3825 ktrace_enter(ip->i_xtrace,
3826 (void *)(__psint_t)(opcode | (whichfork << 16)),
3827 (void *)fname, (void *)desc, (void *)ip,
3828 (void *)(__psint_t)idx,
3829 (void *)(__psint_t)cnt,
3830 (void *)(__psunsigned_t)(ip->i_ino >> 32),
3831 (void *)(__psunsigned_t)(unsigned)ip->i_ino,
3832 (void *)(__psunsigned_t)(r1->l0 >> 32),
3833 (void *)(__psunsigned_t)(unsigned)(r1->l0),
3834 (void *)(__psunsigned_t)(r1->l1 >> 32),
3835 (void *)(__psunsigned_t)(unsigned)(r1->l1),
3836 (void *)(__psunsigned_t)(r2->l0 >> 32),
3837 (void *)(__psunsigned_t)(unsigned)(r2->l0),
3838 (void *)(__psunsigned_t)(r2->l1 >> 32),
3839 (void *)(__psunsigned_t)(unsigned)(r2->l1)
3840 );
3841}
3842
3843/*
4eea22f0 3844 * Add bmap trace entry prior to a call to xfs_iext_remove.
1da177e4
LT
3845 */
3846STATIC void
3847xfs_bmap_trace_delete(
3a59c94c 3848 const char *fname, /* function name */
1da177e4
LT
3849 char *desc, /* operation description */
3850 xfs_inode_t *ip, /* incore inode pointer */
3851 xfs_extnum_t idx, /* index of entry(entries) deleted */
3852 xfs_extnum_t cnt, /* count of entries deleted, 1 or 2 */
3853 int whichfork) /* data or attr fork */
3854{
3855 xfs_ifork_t *ifp; /* inode fork pointer */
3856
3857 ifp = XFS_IFORK_PTR(ip, whichfork);
3858 xfs_bmap_trace_addentry(XFS_BMAP_KTRACE_DELETE, fname, desc, ip, idx,
4eea22f0
MK
3859 cnt, xfs_iext_get_ext(ifp, idx),
3860 cnt == 2 ? xfs_iext_get_ext(ifp, idx + 1) : NULL,
1da177e4
LT
3861 whichfork);
3862}
3863
3864/*
4eea22f0 3865 * Add bmap trace entry prior to a call to xfs_iext_insert, or
1da177e4
LT
3866 * reading in the extents list from the disk (in the btree).
3867 */
3868STATIC void
3869xfs_bmap_trace_insert(
3a59c94c 3870 const char *fname, /* function name */
1da177e4
LT
3871 char *desc, /* operation description */
3872 xfs_inode_t *ip, /* incore inode pointer */
3873 xfs_extnum_t idx, /* index of entry(entries) inserted */
3874 xfs_extnum_t cnt, /* count of entries inserted, 1 or 2 */
3875 xfs_bmbt_irec_t *r1, /* inserted record 1 */
3876 xfs_bmbt_irec_t *r2, /* inserted record 2 or null */
3877 int whichfork) /* data or attr fork */
3878{
a6f64d4a
CH
3879 xfs_bmbt_rec_host_t tr1; /* compressed record 1 */
3880 xfs_bmbt_rec_host_t tr2; /* compressed record 2 if needed */
1da177e4
LT
3881
3882 xfs_bmbt_set_all(&tr1, r1);
3883 if (cnt == 2) {
3884 ASSERT(r2 != NULL);
3885 xfs_bmbt_set_all(&tr2, r2);
3886 } else {
3887 ASSERT(cnt == 1);
3888 ASSERT(r2 == NULL);
3889 }
3890 xfs_bmap_trace_addentry(XFS_BMAP_KTRACE_INSERT, fname, desc, ip, idx,
3891 cnt, &tr1, cnt == 2 ? &tr2 : NULL, whichfork);
3892}
3893
3894/*
4eea22f0 3895 * Add bmap trace entry after updating an extent record in place.
1da177e4
LT
3896 */
3897STATIC void
3898xfs_bmap_trace_post_update(
3a59c94c 3899 const char *fname, /* function name */
1da177e4
LT
3900 char *desc, /* operation description */
3901 xfs_inode_t *ip, /* incore inode pointer */
3902 xfs_extnum_t idx, /* index of entry updated */
3903 int whichfork) /* data or attr fork */
3904{
3905 xfs_ifork_t *ifp; /* inode fork pointer */
3906
3907 ifp = XFS_IFORK_PTR(ip, whichfork);
3908 xfs_bmap_trace_addentry(XFS_BMAP_KTRACE_POST_UP, fname, desc, ip, idx,
4eea22f0 3909 1, xfs_iext_get_ext(ifp, idx), NULL, whichfork);
1da177e4
LT
3910}
3911
3912/*
4eea22f0 3913 * Add bmap trace entry prior to updating an extent record in place.
1da177e4
LT
3914 */
3915STATIC void
3916xfs_bmap_trace_pre_update(
3a59c94c 3917 const char *fname, /* function name */
1da177e4
LT
3918 char *desc, /* operation description */
3919 xfs_inode_t *ip, /* incore inode pointer */
3920 xfs_extnum_t idx, /* index of entry to be updated */
3921 int whichfork) /* data or attr fork */
3922{
3923 xfs_ifork_t *ifp; /* inode fork pointer */
3924
3925 ifp = XFS_IFORK_PTR(ip, whichfork);
3926 xfs_bmap_trace_addentry(XFS_BMAP_KTRACE_PRE_UP, fname, desc, ip, idx, 1,
4eea22f0 3927 xfs_iext_get_ext(ifp, idx), NULL, whichfork);
1da177e4
LT
3928}
3929#endif /* XFS_BMAP_TRACE */
3930
3931/*
3932 * Compute the worst-case number of indirect blocks that will be used
3933 * for ip's delayed extent of length "len".
3934 */
3935STATIC xfs_filblks_t
3936xfs_bmap_worst_indlen(
3937 xfs_inode_t *ip, /* incore inode pointer */
3938 xfs_filblks_t len) /* delayed extent length */
3939{
3940 int level; /* btree level number */
3941 int maxrecs; /* maximum record count at this level */
3942 xfs_mount_t *mp; /* mount structure */
3943 xfs_filblks_t rval; /* return value */
3944
3945 mp = ip->i_mount;
3946 maxrecs = mp->m_bmap_dmxr[0];
3947 for (level = 0, rval = 0;
3948 level < XFS_BM_MAXLEVELS(mp, XFS_DATA_FORK);
3949 level++) {
3950 len += maxrecs - 1;
3951 do_div(len, maxrecs);
3952 rval += len;
3953 if (len == 1)
3954 return rval + XFS_BM_MAXLEVELS(mp, XFS_DATA_FORK) -
3955 level - 1;
3956 if (level == 0)
3957 maxrecs = mp->m_bmap_dmxr[1];
3958 }
3959 return rval;
3960}
3961
3962#if defined(XFS_RW_TRACE)
3963STATIC void
3964xfs_bunmap_trace(
3965 xfs_inode_t *ip,
3966 xfs_fileoff_t bno,
3967 xfs_filblks_t len,
3968 int flags,
3969 inst_t *ra)
3970{
3971 if (ip->i_rwtrace == NULL)
3972 return;
3973 ktrace_enter(ip->i_rwtrace,
3e57ecf6 3974 (void *)(__psint_t)XFS_BUNMAP,
1da177e4
LT
3975 (void *)ip,
3976 (void *)(__psint_t)((ip->i_d.di_size >> 32) & 0xffffffff),
3977 (void *)(__psint_t)(ip->i_d.di_size & 0xffffffff),
3978 (void *)(__psint_t)(((xfs_dfiloff_t)bno >> 32) & 0xffffffff),
3979 (void *)(__psint_t)((xfs_dfiloff_t)bno & 0xffffffff),
3980 (void *)(__psint_t)len,
3981 (void *)(__psint_t)flags,
3982 (void *)(unsigned long)current_cpu(),
3983 (void *)ra,
3984 (void *)0,
3985 (void *)0,
3986 (void *)0,
3987 (void *)0,
3988 (void *)0,
3989 (void *)0);
3990}
3991#endif
3992
3993/*
3994 * Convert inode from non-attributed to attributed.
3995 * Must not be in a transaction, ip must not be locked.
3996 */
3997int /* error code */
3998xfs_bmap_add_attrfork(
3999 xfs_inode_t *ip, /* incore inode pointer */
d8cc890d
NS
4000 int size, /* space new attribute needs */
4001 int rsvd) /* xact may use reserved blks */
1da177e4 4002{
1da177e4 4003 xfs_fsblock_t firstblock; /* 1st block/ag allocated */
4eea22f0 4004 xfs_bmap_free_t flist; /* freed extent records */
1da177e4 4005 xfs_mount_t *mp; /* mount structure */
1da177e4 4006 xfs_trans_t *tp; /* transaction pointer */
d8cc890d
NS
4007 int blks; /* space reservation */
4008 int version = 1; /* superblock attr version */
4009 int committed; /* xaction was committed */
4010 int logflags; /* logging flags */
4011 int error; /* error return value */
1da177e4 4012
d8cc890d 4013 ASSERT(XFS_IFORK_Q(ip) == 0);
1da177e4
LT
4014 ASSERT(ip->i_df.if_ext_max ==
4015 XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t));
d8cc890d 4016
1da177e4
LT
4017 mp = ip->i_mount;
4018 ASSERT(!XFS_NOT_DQATTACHED(mp, ip));
4019 tp = xfs_trans_alloc(mp, XFS_TRANS_ADDAFORK);
4020 blks = XFS_ADDAFORK_SPACE_RES(mp);
4021 if (rsvd)
4022 tp->t_flags |= XFS_TRANS_RESERVE;
4023 if ((error = xfs_trans_reserve(tp, blks, XFS_ADDAFORK_LOG_RES(mp), 0,
4024 XFS_TRANS_PERM_LOG_RES, XFS_ADDAFORK_LOG_COUNT)))
4025 goto error0;
4026 xfs_ilock(ip, XFS_ILOCK_EXCL);
4027 error = XFS_TRANS_RESERVE_QUOTA_NBLKS(mp, tp, ip, blks, 0, rsvd ?
4028 XFS_QMOPT_RES_REGBLKS | XFS_QMOPT_FORCE_RES :
4029 XFS_QMOPT_RES_REGBLKS);
4030 if (error) {
4031 xfs_iunlock(ip, XFS_ILOCK_EXCL);
4032 xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES);
4033 return error;
4034 }
4035 if (XFS_IFORK_Q(ip))
4036 goto error1;
4037 if (ip->i_d.di_aformat != XFS_DINODE_FMT_EXTENTS) {
4038 /*
4039 * For inodes coming from pre-6.2 filesystems.
4040 */
4041 ASSERT(ip->i_d.di_aformat == 0);
4042 ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS;
4043 }
4044 ASSERT(ip->i_d.di_anextents == 0);
0b1f9177 4045 IHOLD(ip);
1da177e4
LT
4046 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
4047 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
4048 switch (ip->i_d.di_format) {
4049 case XFS_DINODE_FMT_DEV:
4050 ip->i_d.di_forkoff = roundup(sizeof(xfs_dev_t), 8) >> 3;
4051 break;
4052 case XFS_DINODE_FMT_UUID:
4053 ip->i_d.di_forkoff = roundup(sizeof(uuid_t), 8) >> 3;
4054 break;
4055 case XFS_DINODE_FMT_LOCAL:
4056 case XFS_DINODE_FMT_EXTENTS:
4057 case XFS_DINODE_FMT_BTREE:
d8cc890d
NS
4058 ip->i_d.di_forkoff = xfs_attr_shortform_bytesfit(ip, size);
4059 if (!ip->i_d.di_forkoff)
4060 ip->i_d.di_forkoff = mp->m_attroffset >> 3;
13059ff0 4061 else if (mp->m_flags & XFS_MOUNT_ATTR2)
d8cc890d 4062 version = 2;
1da177e4
LT
4063 break;
4064 default:
4065 ASSERT(0);
4066 error = XFS_ERROR(EINVAL);
4067 goto error1;
4068 }
4069 ip->i_df.if_ext_max =
4070 XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
4071 ASSERT(ip->i_afp == NULL);
4072 ip->i_afp = kmem_zone_zalloc(xfs_ifork_zone, KM_SLEEP);
4073 ip->i_afp->if_ext_max =
4074 XFS_IFORK_ASIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
4075 ip->i_afp->if_flags = XFS_IFEXTENTS;
4076 logflags = 0;
9d87c319 4077 xfs_bmap_init(&flist, &firstblock);
1da177e4
LT
4078 switch (ip->i_d.di_format) {
4079 case XFS_DINODE_FMT_LOCAL:
4080 error = xfs_bmap_add_attrfork_local(tp, ip, &firstblock, &flist,
4081 &logflags);
4082 break;
4083 case XFS_DINODE_FMT_EXTENTS:
4084 error = xfs_bmap_add_attrfork_extents(tp, ip, &firstblock,
4085 &flist, &logflags);
4086 break;
4087 case XFS_DINODE_FMT_BTREE:
4088 error = xfs_bmap_add_attrfork_btree(tp, ip, &firstblock, &flist,
4089 &logflags);
4090 break;
4091 default:
4092 error = 0;
4093 break;
4094 }
4095 if (logflags)
4096 xfs_trans_log_inode(tp, ip, logflags);
4097 if (error)
4098 goto error2;
62118709
ES
4099 if (!xfs_sb_version_hasattr(&mp->m_sb) ||
4100 (!xfs_sb_version_hasattr2(&mp->m_sb) && version == 2)) {
da087bad
NS
4101 __int64_t sbfields = 0;
4102
3685c2a1 4103 spin_lock(&mp->m_sb_lock);
62118709
ES
4104 if (!xfs_sb_version_hasattr(&mp->m_sb)) {
4105 xfs_sb_version_addattr(&mp->m_sb);
da087bad 4106 sbfields |= XFS_SB_VERSIONNUM;
d8cc890d 4107 }
62118709
ES
4108 if (!xfs_sb_version_hasattr2(&mp->m_sb) && version == 2) {
4109 xfs_sb_version_addattr2(&mp->m_sb);
da087bad 4110 sbfields |= (XFS_SB_VERSIONNUM | XFS_SB_FEATURES2);
d8cc890d 4111 }
da087bad 4112 if (sbfields) {
3685c2a1 4113 spin_unlock(&mp->m_sb_lock);
da087bad 4114 xfs_mod_sb(tp, sbfields);
1da177e4 4115 } else
3685c2a1 4116 spin_unlock(&mp->m_sb_lock);
1da177e4 4117 }
f7c99b6f 4118 if ((error = xfs_bmap_finish(&tp, &flist, &committed)))
1da177e4 4119 goto error2;
1c72bf90 4120 error = xfs_trans_commit(tp, XFS_TRANS_PERM_LOG_RES);
1da177e4
LT
4121 ASSERT(ip->i_df.if_ext_max ==
4122 XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t));
4123 return error;
4124error2:
4125 xfs_bmap_cancel(&flist);
4126error1:
1da177e4
LT
4127 xfs_iunlock(ip, XFS_ILOCK_EXCL);
4128error0:
4129 xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT);
4130 ASSERT(ip->i_df.if_ext_max ==
4131 XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t));
4132 return error;
4133}
4134
4135/*
4136 * Add the extent to the list of extents to be free at transaction end.
4137 * The list is maintained sorted (by block number).
4138 */
4139/* ARGSUSED */
4140void
4141xfs_bmap_add_free(
4142 xfs_fsblock_t bno, /* fs block number of extent */
4143 xfs_filblks_t len, /* length of extent */
4144 xfs_bmap_free_t *flist, /* list of extents */
4145 xfs_mount_t *mp) /* mount point structure */
4146{
4147 xfs_bmap_free_item_t *cur; /* current (next) element */
4148 xfs_bmap_free_item_t *new; /* new element */
4149 xfs_bmap_free_item_t *prev; /* previous element */
4150#ifdef DEBUG
4151 xfs_agnumber_t agno;
4152 xfs_agblock_t agbno;
4153
4154 ASSERT(bno != NULLFSBLOCK);
4155 ASSERT(len > 0);
4156 ASSERT(len <= MAXEXTLEN);
9d87c319 4157 ASSERT(!isnullstartblock(bno));
1da177e4
LT
4158 agno = XFS_FSB_TO_AGNO(mp, bno);
4159 agbno = XFS_FSB_TO_AGBNO(mp, bno);
4160 ASSERT(agno < mp->m_sb.sb_agcount);
4161 ASSERT(agbno < mp->m_sb.sb_agblocks);
4162 ASSERT(len < mp->m_sb.sb_agblocks);
4163 ASSERT(agbno + len <= mp->m_sb.sb_agblocks);
4164#endif
4165 ASSERT(xfs_bmap_free_item_zone != NULL);
4166 new = kmem_zone_alloc(xfs_bmap_free_item_zone, KM_SLEEP);
4167 new->xbfi_startblock = bno;
4168 new->xbfi_blockcount = (xfs_extlen_t)len;
4169 for (prev = NULL, cur = flist->xbf_first;
4170 cur != NULL;
4171 prev = cur, cur = cur->xbfi_next) {
4172 if (cur->xbfi_startblock >= bno)
4173 break;
4174 }
4175 if (prev)
4176 prev->xbfi_next = new;
4177 else
4178 flist->xbf_first = new;
4179 new->xbfi_next = cur;
4180 flist->xbf_count++;
4181}
4182
4183/*
4184 * Compute and fill in the value of the maximum depth of a bmap btree
4185 * in this filesystem. Done once, during mount.
4186 */
4187void
4188xfs_bmap_compute_maxlevels(
4189 xfs_mount_t *mp, /* file system mount structure */
4190 int whichfork) /* data or attr fork */
4191{
4192 int level; /* btree level */
4193 uint maxblocks; /* max blocks at this level */
4194 uint maxleafents; /* max leaf entries possible */
4195 int maxrootrecs; /* max records in root block */
4196 int minleafrecs; /* min records in leaf block */
4197 int minnoderecs; /* min records in node block */
4198 int sz; /* root block size */
4199
4200 /*
4201 * The maximum number of extents in a file, hence the maximum
4202 * number of leaf entries, is controlled by the type of di_nextents
4203 * (a signed 32-bit number, xfs_extnum_t), or by di_anextents
4204 * (a signed 16-bit number, xfs_aextnum_t).
6d1337b2
TS
4205 *
4206 * Note that we can no longer assume that if we are in ATTR1 that
4207 * the fork offset of all the inodes will be (m_attroffset >> 3)
4208 * because we could have mounted with ATTR2 and then mounted back
4209 * with ATTR1, keeping the di_forkoff's fixed but probably at
4210 * various positions. Therefore, for both ATTR1 and ATTR2
4211 * we have to assume the worst case scenario of a minimum size
4212 * available.
1da177e4 4213 */
d8cc890d
NS
4214 if (whichfork == XFS_DATA_FORK) {
4215 maxleafents = MAXEXTNUM;
6d1337b2 4216 sz = XFS_BMDR_SPACE_CALC(MINDBTPTRS);
d8cc890d
NS
4217 } else {
4218 maxleafents = MAXAEXTNUM;
6d1337b2 4219 sz = XFS_BMDR_SPACE_CALC(MINABTPTRS);
d8cc890d 4220 }
60197e8d 4221 maxrootrecs = xfs_bmdr_maxrecs(mp, sz, 0);
1da177e4
LT
4222 minleafrecs = mp->m_bmap_dmnr[0];
4223 minnoderecs = mp->m_bmap_dmnr[1];
1da177e4
LT
4224 maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs;
4225 for (level = 1; maxblocks > 1; level++) {
4226 if (maxblocks <= maxrootrecs)
4227 maxblocks = 1;
4228 else
4229 maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs;
4230 }
4231 mp->m_bm_maxlevels[whichfork] = level;
4232}
4233
4234/*
4235 * Routine to be called at transaction's end by xfs_bmapi, xfs_bunmapi
4236 * caller. Frees all the extents that need freeing, which must be done
4237 * last due to locking considerations. We never free any extents in
4238 * the first transaction. This is to allow the caller to make the first
4239 * transaction a synchronous one so that the pointers to the data being
4240 * broken in this transaction will be permanent before the data is actually
4241 * freed. This is necessary to prevent blocks from being reallocated
4242 * and written to before the free and reallocation are actually permanent.
4243 * We do not just make the first transaction synchronous here, because
4244 * there are more efficient ways to gain the same protection in some cases
4245 * (see the file truncation code).
4246 *
4247 * Return 1 if the given transaction was committed and a new one
4248 * started, and 0 otherwise in the committed parameter.
4249 */
4250/*ARGSUSED*/
4251int /* error */
4252xfs_bmap_finish(
4253 xfs_trans_t **tp, /* transaction pointer addr */
4254 xfs_bmap_free_t *flist, /* i/o: list extents to free */
1da177e4
LT
4255 int *committed) /* xact committed or not */
4256{
4257 xfs_efd_log_item_t *efd; /* extent free data */
4258 xfs_efi_log_item_t *efi; /* extent free intention */
4259 int error; /* error return value */
4eea22f0 4260 xfs_bmap_free_item_t *free; /* free extent item */
1da177e4
LT
4261 unsigned int logres; /* new log reservation */
4262 unsigned int logcount; /* new log count */
4263 xfs_mount_t *mp; /* filesystem mount structure */
4264 xfs_bmap_free_item_t *next; /* next item on free list */
4265 xfs_trans_t *ntp; /* new transaction pointer */
4266
4267 ASSERT((*tp)->t_flags & XFS_TRANS_PERM_LOG_RES);
4268 if (flist->xbf_count == 0) {
4269 *committed = 0;
4270 return 0;
4271 }
4272 ntp = *tp;
4273 efi = xfs_trans_get_efi(ntp, flist->xbf_count);
4274 for (free = flist->xbf_first; free; free = free->xbfi_next)
4275 xfs_trans_log_efi_extent(ntp, efi, free->xbfi_startblock,
4276 free->xbfi_blockcount);
4277 logres = ntp->t_log_res;
4278 logcount = ntp->t_log_count;
4279 ntp = xfs_trans_dup(*tp);
1c72bf90 4280 error = xfs_trans_commit(*tp, 0);
1da177e4
LT
4281 *tp = ntp;
4282 *committed = 1;
4283 /*
4284 * We have a new transaction, so we should return committed=1,
4285 * even though we're returning an error.
4286 */
cc09c0dc 4287 if (error)
1da177e4 4288 return error;
cc09c0dc
DC
4289
4290 /*
4291 * transaction commit worked ok so we can drop the extra ticket
4292 * reference that we gained in xfs_trans_dup()
4293 */
4294 xfs_log_ticket_put(ntp->t_ticket);
4295
1da177e4
LT
4296 if ((error = xfs_trans_reserve(ntp, 0, logres, 0, XFS_TRANS_PERM_LOG_RES,
4297 logcount)))
4298 return error;
4299 efd = xfs_trans_get_efd(ntp, efi, flist->xbf_count);
4300 for (free = flist->xbf_first; free != NULL; free = next) {
4301 next = free->xbfi_next;
4302 if ((error = xfs_free_extent(ntp, free->xbfi_startblock,
4303 free->xbfi_blockcount))) {
4304 /*
4305 * The bmap free list will be cleaned up at a
4306 * higher level. The EFI will be canceled when
4307 * this transaction is aborted.
4308 * Need to force shutdown here to make sure it
4309 * happens, since this transaction may not be
4310 * dirty yet.
4311 */
4312 mp = ntp->t_mountp;
4313 if (!XFS_FORCED_SHUTDOWN(mp))
4314 xfs_force_shutdown(mp,
4315 (error == EFSCORRUPTED) ?
7d04a335
NS
4316 SHUTDOWN_CORRUPT_INCORE :
4317 SHUTDOWN_META_IO_ERROR);
1da177e4
LT
4318 return error;
4319 }
4320 xfs_trans_log_efd_extent(ntp, efd, free->xbfi_startblock,
4321 free->xbfi_blockcount);
4322 xfs_bmap_del_free(flist, NULL, free);
4323 }
4324 return 0;
4325}
4326
4327/*
4328 * Free up any items left in the list.
4329 */
4330void
4331xfs_bmap_cancel(
4332 xfs_bmap_free_t *flist) /* list of bmap_free_items */
4333{
4334 xfs_bmap_free_item_t *free; /* free list item */
4335 xfs_bmap_free_item_t *next;
4336
4337 if (flist->xbf_count == 0)
4338 return;
4339 ASSERT(flist->xbf_first != NULL);
4340 for (free = flist->xbf_first; free; free = next) {
4341 next = free->xbfi_next;
4342 xfs_bmap_del_free(flist, NULL, free);
4343 }
4344 ASSERT(flist->xbf_count == 0);
4345}
4346
4347/*
4348 * Returns the file-relative block number of the first unused block(s)
4349 * in the file with at least "len" logically contiguous blocks free.
4350 * This is the lowest-address hole if the file has holes, else the first block
4351 * past the end of file.
4352 * Return 0 if the file is currently local (in-inode).
4353 */
4354int /* error */
4355xfs_bmap_first_unused(
4356 xfs_trans_t *tp, /* transaction pointer */
4357 xfs_inode_t *ip, /* incore inode */
4358 xfs_extlen_t len, /* size of hole to find */
4359 xfs_fileoff_t *first_unused, /* unused block */
4360 int whichfork) /* data or attr fork */
4361{
1da177e4 4362 int error; /* error return value */
4eea22f0 4363 int idx; /* extent record index */
1da177e4
LT
4364 xfs_ifork_t *ifp; /* inode fork pointer */
4365 xfs_fileoff_t lastaddr; /* last block number seen */
4366 xfs_fileoff_t lowest; /* lowest useful block */
4367 xfs_fileoff_t max; /* starting useful block */
4368 xfs_fileoff_t off; /* offset for this block */
4369 xfs_extnum_t nextents; /* number of extent entries */
4370
4371 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE ||
4372 XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS ||
4373 XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL);
4374 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL) {
4375 *first_unused = 0;
4376 return 0;
4377 }
4378 ifp = XFS_IFORK_PTR(ip, whichfork);
4379 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
4380 (error = xfs_iread_extents(tp, ip, whichfork)))
4381 return error;
4382 lowest = *first_unused;
4383 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
4eea22f0 4384 for (idx = 0, lastaddr = 0, max = lowest; idx < nextents; idx++) {
a6f64d4a 4385 xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, idx);
1da177e4
LT
4386 off = xfs_bmbt_get_startoff(ep);
4387 /*
4388 * See if the hole before this extent will work.
4389 */
4390 if (off >= lowest + len && off - max >= len) {
4391 *first_unused = max;
4392 return 0;
4393 }
4394 lastaddr = off + xfs_bmbt_get_blockcount(ep);
4395 max = XFS_FILEOFF_MAX(lastaddr, lowest);
4396 }
4397 *first_unused = max;
4398 return 0;
4399}
4400
4401/*
4402 * Returns the file-relative block number of the last block + 1 before
4403 * last_block (input value) in the file.
4eea22f0
MK
4404 * This is not based on i_size, it is based on the extent records.
4405 * Returns 0 for local files, as they do not have extent records.
1da177e4
LT
4406 */
4407int /* error */
4408xfs_bmap_last_before(
4409 xfs_trans_t *tp, /* transaction pointer */
4410 xfs_inode_t *ip, /* incore inode */
4411 xfs_fileoff_t *last_block, /* last block */
4412 int whichfork) /* data or attr fork */
4413{
4414 xfs_fileoff_t bno; /* input file offset */
4415 int eof; /* hit end of file */
a6f64d4a 4416 xfs_bmbt_rec_host_t *ep; /* pointer to last extent */
1da177e4
LT
4417 int error; /* error return value */
4418 xfs_bmbt_irec_t got; /* current extent value */
4419 xfs_ifork_t *ifp; /* inode fork pointer */
4420 xfs_extnum_t lastx; /* last extent used */
4421 xfs_bmbt_irec_t prev; /* previous extent value */
4422
4423 if (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE &&
4424 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
4425 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL)
4426 return XFS_ERROR(EIO);
4427 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL) {
4428 *last_block = 0;
4429 return 0;
4430 }
4431 ifp = XFS_IFORK_PTR(ip, whichfork);
4432 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
4433 (error = xfs_iread_extents(tp, ip, whichfork)))
4434 return error;
4435 bno = *last_block - 1;
4436 ep = xfs_bmap_search_extents(ip, bno, whichfork, &eof, &lastx, &got,
4437 &prev);
4438 if (eof || xfs_bmbt_get_startoff(ep) > bno) {
4439 if (prev.br_startoff == NULLFILEOFF)
4440 *last_block = 0;
4441 else
4442 *last_block = prev.br_startoff + prev.br_blockcount;
4443 }
4444 /*
4445 * Otherwise *last_block is already the right answer.
4446 */
4447 return 0;
4448}
4449
4450/*
4451 * Returns the file-relative block number of the first block past eof in
4eea22f0
MK
4452 * the file. This is not based on i_size, it is based on the extent records.
4453 * Returns 0 for local files, as they do not have extent records.
1da177e4
LT
4454 */
4455int /* error */
4456xfs_bmap_last_offset(
4457 xfs_trans_t *tp, /* transaction pointer */
4458 xfs_inode_t *ip, /* incore inode */
4459 xfs_fileoff_t *last_block, /* last block */
4460 int whichfork) /* data or attr fork */
4461{
a6f64d4a 4462 xfs_bmbt_rec_host_t *ep; /* pointer to last extent */
1da177e4
LT
4463 int error; /* error return value */
4464 xfs_ifork_t *ifp; /* inode fork pointer */
4465 xfs_extnum_t nextents; /* number of extent entries */
4466
4467 if (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE &&
4468 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
4469 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL)
4470 return XFS_ERROR(EIO);
4471 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL) {
4472 *last_block = 0;
4473 return 0;
4474 }
4475 ifp = XFS_IFORK_PTR(ip, whichfork);
4476 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
4477 (error = xfs_iread_extents(tp, ip, whichfork)))
4478 return error;
4479 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
4480 if (!nextents) {
4481 *last_block = 0;
4482 return 0;
4483 }
4eea22f0 4484 ep = xfs_iext_get_ext(ifp, nextents - 1);
1da177e4
LT
4485 *last_block = xfs_bmbt_get_startoff(ep) + xfs_bmbt_get_blockcount(ep);
4486 return 0;
4487}
4488
4489/*
4490 * Returns whether the selected fork of the inode has exactly one
4491 * block or not. For the data fork we check this matches di_size,
4492 * implying the file's range is 0..bsize-1.
4493 */
4494int /* 1=>1 block, 0=>otherwise */
4495xfs_bmap_one_block(
4496 xfs_inode_t *ip, /* incore inode */
4497 int whichfork) /* data or attr fork */
4498{
a6f64d4a 4499 xfs_bmbt_rec_host_t *ep; /* ptr to fork's extent */
1da177e4
LT
4500 xfs_ifork_t *ifp; /* inode fork pointer */
4501 int rval; /* return value */
4502 xfs_bmbt_irec_t s; /* internal version of extent */
4503
4504#ifndef DEBUG
ba87ea69
LM
4505 if (whichfork == XFS_DATA_FORK) {
4506 return ((ip->i_d.di_mode & S_IFMT) == S_IFREG) ?
4507 (ip->i_size == ip->i_mount->m_sb.sb_blocksize) :
4508 (ip->i_d.di_size == ip->i_mount->m_sb.sb_blocksize);
4509 }
1da177e4
LT
4510#endif /* !DEBUG */
4511 if (XFS_IFORK_NEXTENTS(ip, whichfork) != 1)
4512 return 0;
4513 if (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS)
4514 return 0;
4515 ifp = XFS_IFORK_PTR(ip, whichfork);
4516 ASSERT(ifp->if_flags & XFS_IFEXTENTS);
4eea22f0 4517 ep = xfs_iext_get_ext(ifp, 0);
1da177e4
LT
4518 xfs_bmbt_get_all(ep, &s);
4519 rval = s.br_startoff == 0 && s.br_blockcount == 1;
4520 if (rval && whichfork == XFS_DATA_FORK)
ba87ea69 4521 ASSERT(ip->i_size == ip->i_mount->m_sb.sb_blocksize);
1da177e4
LT
4522 return rval;
4523}
4524
4e8938fe
CH
4525STATIC int
4526xfs_bmap_sanity_check(
4527 struct xfs_mount *mp,
4528 struct xfs_buf *bp,
4529 int level)
4530{
4531 struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
4532
4533 if (be32_to_cpu(block->bb_magic) != XFS_BMAP_MAGIC ||
4534 be16_to_cpu(block->bb_level) != level ||
4535 be16_to_cpu(block->bb_numrecs) == 0 ||
4536 be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
4537 return 0;
4538 return 1;
4539}
4540
1da177e4
LT
4541/*
4542 * Read in the extents to if_extents.
4543 * All inode fields are set up by caller, we just traverse the btree
4544 * and copy the records in. If the file system cannot contain unwritten
4545 * extents, the records are checked for no "state" flags.
4546 */
4547int /* error */
4548xfs_bmap_read_extents(
4549 xfs_trans_t *tp, /* transaction pointer */
4550 xfs_inode_t *ip, /* incore inode */
4551 int whichfork) /* data or attr fork */
4552{
7cc95a82 4553 struct xfs_btree_block *block; /* current btree block */
1da177e4
LT
4554 xfs_fsblock_t bno; /* block # of "block" */
4555 xfs_buf_t *bp; /* buffer for "block" */
4556 int error; /* error return value */
4557 xfs_exntfmt_t exntf; /* XFS_EXTFMT_NOSTATE, if checking */
1da177e4
LT
4558 xfs_extnum_t i, j; /* index into the extents list */
4559 xfs_ifork_t *ifp; /* fork structure */
4560 int level; /* btree level, for checking */
4561 xfs_mount_t *mp; /* file system mount structure */
576039cf 4562 __be64 *pp; /* pointer to block address */
1da177e4
LT
4563 /* REFERENCED */
4564 xfs_extnum_t room; /* number of entries there's room for */
1da177e4
LT
4565
4566 bno = NULLFSBLOCK;
4567 mp = ip->i_mount;
4568 ifp = XFS_IFORK_PTR(ip, whichfork);
4569 exntf = (whichfork != XFS_DATA_FORK) ? XFS_EXTFMT_NOSTATE :
4570 XFS_EXTFMT_INODE(ip);
4571 block = ifp->if_broot;
4572 /*
4573 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
4574 */
16259e7d
CH
4575 level = be16_to_cpu(block->bb_level);
4576 ASSERT(level > 0);
60197e8d 4577 pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
576039cf
CH
4578 bno = be64_to_cpu(*pp);
4579 ASSERT(bno != NULLDFSBNO);
4580 ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
4581 ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
1da177e4
LT
4582 /*
4583 * Go down the tree until leaf level is reached, following the first
4584 * pointer (leftmost) at each level.
4585 */
4586 while (level-- > 0) {
4587 if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
4588 XFS_BMAP_BTREE_REF)))
4589 return error;
7cc95a82 4590 block = XFS_BUF_TO_BLOCK(bp);
1da177e4 4591 XFS_WANT_CORRUPTED_GOTO(
4e8938fe 4592 xfs_bmap_sanity_check(mp, bp, level),
1da177e4
LT
4593 error0);
4594 if (level == 0)
4595 break;
136341b4 4596 pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
576039cf
CH
4597 bno = be64_to_cpu(*pp);
4598 XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0);
1da177e4
LT
4599 xfs_trans_brelse(tp, bp);
4600 }
4601 /*
4602 * Here with bp and block set to the leftmost leaf node in the tree.
4603 */
4eea22f0 4604 room = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
1da177e4
LT
4605 i = 0;
4606 /*
4eea22f0 4607 * Loop over all leaf nodes. Copy information to the extent records.
1da177e4
LT
4608 */
4609 for (;;) {
a6f64d4a 4610 xfs_bmbt_rec_t *frp;
1da177e4
LT
4611 xfs_fsblock_t nextbno;
4612 xfs_extnum_t num_recs;
4eea22f0 4613 xfs_extnum_t start;
1da177e4
LT
4614
4615
7cc95a82 4616 num_recs = xfs_btree_get_numrecs(block);
1da177e4
LT
4617 if (unlikely(i + num_recs > room)) {
4618 ASSERT(i + num_recs <= room);
3762ec6b
NS
4619 xfs_fs_repair_cmn_err(CE_WARN, ip->i_mount,
4620 "corrupt dinode %Lu, (btree extents).",
1da177e4
LT
4621 (unsigned long long) ip->i_ino);
4622 XFS_ERROR_REPORT("xfs_bmap_read_extents(1)",
4623 XFS_ERRLEVEL_LOW,
4624 ip->i_mount);
4625 goto error0;
4626 }
4627 XFS_WANT_CORRUPTED_GOTO(
4e8938fe 4628 xfs_bmap_sanity_check(mp, bp, 0),
1da177e4
LT
4629 error0);
4630 /*
4631 * Read-ahead the next leaf block, if any.
4632 */
7cc95a82 4633 nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
1da177e4
LT
4634 if (nextbno != NULLFSBLOCK)
4635 xfs_btree_reada_bufl(mp, nextbno, 1);
4636 /*
4eea22f0 4637 * Copy records into the extent records.
1da177e4 4638 */
136341b4 4639 frp = XFS_BMBT_REC_ADDR(mp, block, 1);
4eea22f0
MK
4640 start = i;
4641 for (j = 0; j < num_recs; j++, i++, frp++) {
a6f64d4a 4642 xfs_bmbt_rec_host_t *trp = xfs_iext_get_ext(ifp, i);
cd8b0a97
CH
4643 trp->l0 = be64_to_cpu(frp->l0);
4644 trp->l1 = be64_to_cpu(frp->l1);
1da177e4
LT
4645 }
4646 if (exntf == XFS_EXTFMT_NOSTATE) {
4647 /*
4648 * Check all attribute bmap btree records and
4649 * any "older" data bmap btree records for a
4650 * set bit in the "extent flag" position.
4651 */
4eea22f0
MK
4652 if (unlikely(xfs_check_nostate_extents(ifp,
4653 start, num_recs))) {
1da177e4
LT
4654 XFS_ERROR_REPORT("xfs_bmap_read_extents(2)",
4655 XFS_ERRLEVEL_LOW,
4656 ip->i_mount);
4657 goto error0;
4658 }
4659 }
1da177e4
LT
4660 xfs_trans_brelse(tp, bp);
4661 bno = nextbno;
4662 /*
4663 * If we've reached the end, stop.
4664 */
4665 if (bno == NULLFSBLOCK)
4666 break;
4667 if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
4668 XFS_BMAP_BTREE_REF)))
4669 return error;
7cc95a82 4670 block = XFS_BUF_TO_BLOCK(bp);
1da177e4 4671 }
4eea22f0 4672 ASSERT(i == (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t)));
1da177e4 4673 ASSERT(i == XFS_IFORK_NEXTENTS(ip, whichfork));
3a59c94c 4674 XFS_BMAP_TRACE_EXLIST(ip, i, whichfork);
1da177e4
LT
4675 return 0;
4676error0:
4677 xfs_trans_brelse(tp, bp);
4678 return XFS_ERROR(EFSCORRUPTED);
4679}
4680
4681#ifdef XFS_BMAP_TRACE
4682/*
4eea22f0 4683 * Add bmap trace insert entries for all the contents of the extent records.
1da177e4
LT
4684 */
4685void
4686xfs_bmap_trace_exlist(
3a59c94c 4687 const char *fname, /* function name */
1da177e4
LT
4688 xfs_inode_t *ip, /* incore inode pointer */
4689 xfs_extnum_t cnt, /* count of entries in the list */
4690 int whichfork) /* data or attr fork */
4691{
a6f64d4a 4692 xfs_bmbt_rec_host_t *ep; /* current extent record */
4eea22f0 4693 xfs_extnum_t idx; /* extent record index */
1da177e4 4694 xfs_ifork_t *ifp; /* inode fork pointer */
4eea22f0 4695 xfs_bmbt_irec_t s; /* file extent record */
1da177e4
LT
4696
4697 ifp = XFS_IFORK_PTR(ip, whichfork);
4eea22f0
MK
4698 ASSERT(cnt == (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t)));
4699 for (idx = 0; idx < cnt; idx++) {
4700 ep = xfs_iext_get_ext(ifp, idx);
1da177e4 4701 xfs_bmbt_get_all(ep, &s);
3a59c94c 4702 XFS_BMAP_TRACE_INSERT("exlist", ip, idx, 1, &s, NULL,
1da177e4
LT
4703 whichfork);
4704 }
4705}
4706#endif
4707
4708#ifdef DEBUG
4709/*
4710 * Validate that the bmbt_irecs being returned from bmapi are valid
4711 * given the callers original parameters. Specifically check the
4712 * ranges of the returned irecs to ensure that they only extent beyond
4713 * the given parameters if the XFS_BMAPI_ENTIRE flag was set.
4714 */
4715STATIC void
4716xfs_bmap_validate_ret(
4717 xfs_fileoff_t bno,
4718 xfs_filblks_t len,
4719 int flags,
4720 xfs_bmbt_irec_t *mval,
4721 int nmap,
4722 int ret_nmap)
4723{
4724 int i; /* index to map values */
4725
4726 ASSERT(ret_nmap <= nmap);
4727
4728 for (i = 0; i < ret_nmap; i++) {
4729 ASSERT(mval[i].br_blockcount > 0);
4730 if (!(flags & XFS_BMAPI_ENTIRE)) {
4731 ASSERT(mval[i].br_startoff >= bno);
4732 ASSERT(mval[i].br_blockcount <= len);
4733 ASSERT(mval[i].br_startoff + mval[i].br_blockcount <=
4734 bno + len);
4735 } else {
4736 ASSERT(mval[i].br_startoff < bno + len);
4737 ASSERT(mval[i].br_startoff + mval[i].br_blockcount >
4738 bno);
4739 }
4740 ASSERT(i == 0 ||
4741 mval[i - 1].br_startoff + mval[i - 1].br_blockcount ==
4742 mval[i].br_startoff);
4743 if ((flags & XFS_BMAPI_WRITE) && !(flags & XFS_BMAPI_DELAY))
4744 ASSERT(mval[i].br_startblock != DELAYSTARTBLOCK &&
4745 mval[i].br_startblock != HOLESTARTBLOCK);
4746 ASSERT(mval[i].br_state == XFS_EXT_NORM ||
4747 mval[i].br_state == XFS_EXT_UNWRITTEN);
4748 }
4749}
4750#endif /* DEBUG */
4751
4752
4753/*
4754 * Map file blocks to filesystem blocks.
4755 * File range is given by the bno/len pair.
4756 * Adds blocks to file if a write ("flags & XFS_BMAPI_WRITE" set)
4757 * into a hole or past eof.
4758 * Only allocates blocks from a single allocation group,
4759 * to avoid locking problems.
4760 * The returned value in "firstblock" from the first call in a transaction
4761 * must be remembered and presented to subsequent calls in "firstblock".
4762 * An upper bound for the number of blocks to be allocated is supplied to
4763 * the first call in "total"; if no allocation group has that many free
4764 * blocks then the call will fail (return NULLFSBLOCK in "firstblock").
4765 */
4766int /* error */
4767xfs_bmapi(
4768 xfs_trans_t *tp, /* transaction pointer */
4769 xfs_inode_t *ip, /* incore inode */
4770 xfs_fileoff_t bno, /* starting file offs. mapped */
4771 xfs_filblks_t len, /* length to map in file */
4772 int flags, /* XFS_BMAPI_... */
4773 xfs_fsblock_t *firstblock, /* first allocated block
4774 controls a.g. for allocs */
4775 xfs_extlen_t total, /* total blocks needed */
4776 xfs_bmbt_irec_t *mval, /* output: map values */
4777 int *nmap, /* i/o: mval size/count */
3e57ecf6
OW
4778 xfs_bmap_free_t *flist, /* i/o: list extents to free */
4779 xfs_extdelta_t *delta) /* o: change made to incore extents */
1da177e4
LT
4780{
4781 xfs_fsblock_t abno; /* allocated block number */
4782 xfs_extlen_t alen; /* allocated extent length */
4783 xfs_fileoff_t aoff; /* allocated file offset */
4784 xfs_bmalloca_t bma; /* args for xfs_bmap_alloc */
1da177e4 4785 xfs_btree_cur_t *cur; /* bmap btree cursor */
1da177e4 4786 xfs_fileoff_t end; /* end of mapped file region */
4eea22f0 4787 int eof; /* we've hit the end of extents */
a6f64d4a 4788 xfs_bmbt_rec_host_t *ep; /* extent record pointer */
1da177e4 4789 int error; /* error return */
4eea22f0 4790 xfs_bmbt_irec_t got; /* current file extent record */
1da177e4
LT
4791 xfs_ifork_t *ifp; /* inode fork pointer */
4792 xfs_extlen_t indlen; /* indirect blocks length */
1da177e4
LT
4793 xfs_extnum_t lastx; /* last useful extent number */
4794 int logflags; /* flags for transaction logging */
4795 xfs_extlen_t minleft; /* min blocks left after allocation */
4796 xfs_extlen_t minlen; /* min allocation size */
4797 xfs_mount_t *mp; /* xfs mount structure */
4798 int n; /* current extent index */
9da096fd 4799 int nallocs; /* number of extents alloc'd */
1da177e4
LT
4800 xfs_extnum_t nextents; /* number of extents in file */
4801 xfs_fileoff_t obno; /* old block number (offset) */
4eea22f0 4802 xfs_bmbt_irec_t prev; /* previous file extent record */
1da177e4 4803 int tmp_logflags; /* temp flags holder */
06d10dd9
NS
4804 int whichfork; /* data or attr fork */
4805 char inhole; /* current location is hole in file */
1da177e4 4806 char wasdelay; /* old extent was delayed */
1da177e4 4807 char wr; /* this is a write request */
06d10dd9 4808 char rt; /* this is a realtime file */
1da177e4
LT
4809#ifdef DEBUG
4810 xfs_fileoff_t orig_bno; /* original block number value */
4811 int orig_flags; /* original flags arg value */
4812 xfs_filblks_t orig_len; /* original value of len arg */
4813 xfs_bmbt_irec_t *orig_mval; /* original value of mval */
4814 int orig_nmap; /* original value of *nmap */
4815
4816 orig_bno = bno;
4817 orig_len = len;
4818 orig_flags = flags;
4819 orig_mval = mval;
4820 orig_nmap = *nmap;
4821#endif
4822 ASSERT(*nmap >= 1);
4823 ASSERT(*nmap <= XFS_BMAP_MAX_NMAP || !(flags & XFS_BMAPI_WRITE));
4824 whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
4825 XFS_ATTR_FORK : XFS_DATA_FORK;
4826 mp = ip->i_mount;
4827 if (unlikely(XFS_TEST_ERROR(
4828 (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
4829 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE &&
4830 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL),
4831 mp, XFS_ERRTAG_BMAPIFORMAT, XFS_RANDOM_BMAPIFORMAT))) {
4832 XFS_ERROR_REPORT("xfs_bmapi", XFS_ERRLEVEL_LOW, mp);
4833 return XFS_ERROR(EFSCORRUPTED);
4834 }
4835 if (XFS_FORCED_SHUTDOWN(mp))
4836 return XFS_ERROR(EIO);
dd9f438e 4837 rt = (whichfork == XFS_DATA_FORK) && XFS_IS_REALTIME_INODE(ip);
1da177e4
LT
4838 ifp = XFS_IFORK_PTR(ip, whichfork);
4839 ASSERT(ifp->if_ext_max ==
4840 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
4841 if ((wr = (flags & XFS_BMAPI_WRITE)) != 0)
4842 XFS_STATS_INC(xs_blk_mapw);
4843 else
4844 XFS_STATS_INC(xs_blk_mapr);
1da177e4 4845 /*
39269e29 4846 * IGSTATE flag is used to combine extents which
1da177e4
LT
4847 * differ only due to the state of the extents.
4848 * This technique is used from xfs_getbmap()
4849 * when the caller does not wish to see the
4850 * separation (which is the default).
4851 *
4852 * This technique is also used when writing a
4853 * buffer which has been partially written,
4854 * (usually by being flushed during a chunkread),
4855 * to ensure one write takes place. This also
4856 * prevents a change in the xfs inode extents at
4857 * this time, intentionally. This change occurs
4858 * on completion of the write operation, in
4859 * xfs_strat_comp(), where the xfs_bmapi() call
4860 * is transactioned, and the extents combined.
4861 */
39269e29
NS
4862 if ((flags & XFS_BMAPI_IGSTATE) && wr) /* if writing unwritten space */
4863 wr = 0; /* no allocations are allowed */
4864 ASSERT(wr || !(flags & XFS_BMAPI_DELAY));
1da177e4
LT
4865 logflags = 0;
4866 nallocs = 0;
4867 cur = NULL;
4868 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL) {
4869 ASSERT(wr && tp);
4870 if ((error = xfs_bmap_local_to_extents(tp, ip,
4871 firstblock, total, &logflags, whichfork)))
4872 goto error0;
4873 }
4874 if (wr && *firstblock == NULLFSBLOCK) {
4875 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE)
16259e7d 4876 minleft = be16_to_cpu(ifp->if_broot->bb_level) + 1;
1da177e4
LT
4877 else
4878 minleft = 1;
4879 } else
4880 minleft = 0;
4881 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
4882 (error = xfs_iread_extents(tp, ip, whichfork)))
4883 goto error0;
4884 ep = xfs_bmap_search_extents(ip, bno, whichfork, &eof, &lastx, &got,
4885 &prev);
4886 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
4887 n = 0;
4888 end = bno + len;
4889 obno = bno;
4890 bma.ip = NULL;
3e57ecf6
OW
4891 if (delta) {
4892 delta->xed_startoff = NULLFILEOFF;
4893 delta->xed_blockcount = 0;
4894 }
1da177e4
LT
4895 while (bno < end && n < *nmap) {
4896 /*
4897 * Reading past eof, act as though there's a hole
4898 * up to end.
4899 */
4900 if (eof && !wr)
4901 got.br_startoff = end;
4902 inhole = eof || got.br_startoff > bno;
39269e29 4903 wasdelay = wr && !inhole && !(flags & XFS_BMAPI_DELAY) &&
9d87c319 4904 isnullstartblock(got.br_startblock);
1da177e4
LT
4905 /*
4906 * First, deal with the hole before the allocated space
4907 * that we found, if any.
4908 */
4909 if (wr && (inhole || wasdelay)) {
4910 /*
4911 * For the wasdelay case, we could also just
4912 * allocate the stuff asked for in this bmap call
4913 * but that wouldn't be as good.
4914 */
39269e29 4915 if (wasdelay && !(flags & XFS_BMAPI_EXACT)) {
1da177e4
LT
4916 alen = (xfs_extlen_t)got.br_blockcount;
4917 aoff = got.br_startoff;
4918 if (lastx != NULLEXTNUM && lastx) {
4eea22f0 4919 ep = xfs_iext_get_ext(ifp, lastx - 1);
1da177e4
LT
4920 xfs_bmbt_get_all(ep, &prev);
4921 }
4922 } else if (wasdelay) {
4923 alen = (xfs_extlen_t)
4924 XFS_FILBLKS_MIN(len,
4925 (got.br_startoff +
4926 got.br_blockcount) - bno);
4927 aoff = bno;
4928 } else {
4929 alen = (xfs_extlen_t)
4930 XFS_FILBLKS_MIN(len, MAXEXTLEN);
4931 if (!eof)
4932 alen = (xfs_extlen_t)
4933 XFS_FILBLKS_MIN(alen,
4934 got.br_startoff - bno);
4935 aoff = bno;
4936 }
39269e29
NS
4937 minlen = (flags & XFS_BMAPI_CONTIG) ? alen : 1;
4938 if (flags & XFS_BMAPI_DELAY) {
dd9f438e 4939 xfs_extlen_t extsz;
06d10dd9
NS
4940
4941 /* Figure out the extent size, adjust alen */
957d0ebe 4942 extsz = xfs_get_extsz_hint(ip);
dd9f438e
NS
4943 if (extsz) {
4944 error = xfs_bmap_extsize_align(mp,
4945 &got, &prev, extsz,
39269e29
NS
4946 rt, eof,
4947 flags&XFS_BMAPI_DELAY,
4948 flags&XFS_BMAPI_CONVERT,
dd9f438e
NS
4949 &aoff, &alen);
4950 ASSERT(!error);
06d10dd9
NS
4951 }
4952
dd9f438e
NS
4953 if (rt)
4954 extsz = alen / mp->m_sb.sb_rextsize;
4955
1da177e4
LT
4956 /*
4957 * Make a transaction-less quota reservation for
4958 * delayed allocation blocks. This number gets
9a2a7de2
NS
4959 * adjusted later. We return if we haven't
4960 * allocated blocks already inside this loop.
1da177e4 4961 */
9a2a7de2 4962 if ((error = XFS_TRANS_RESERVE_QUOTA_NBLKS(
06d10dd9
NS
4963 mp, NULL, ip, (long)alen, 0,
4964 rt ? XFS_QMOPT_RES_RTBLKS :
9a2a7de2 4965 XFS_QMOPT_RES_REGBLKS))) {
1da177e4
LT
4966 if (n == 0) {
4967 *nmap = 0;
4968 ASSERT(cur == NULL);
9a2a7de2 4969 return error;
1da177e4
LT
4970 }
4971 break;
4972 }
4973
4974 /*
4975 * Split changing sb for alen and indlen since
4976 * they could be coming from different places.
4977 */
06d10dd9
NS
4978 indlen = (xfs_extlen_t)
4979 xfs_bmap_worst_indlen(ip, alen);
4980 ASSERT(indlen > 0);
1da177e4 4981
dd9f438e 4982 if (rt) {
06d10dd9
NS
4983 error = xfs_mod_incore_sb(mp,
4984 XFS_SBS_FREXTENTS,
20f4ebf2 4985 -((int64_t)extsz), (flags &
39269e29 4986 XFS_BMAPI_RSVBLOCKS));
dd9f438e 4987 } else {
06d10dd9
NS
4988 error = xfs_mod_incore_sb(mp,
4989 XFS_SBS_FDBLOCKS,
20f4ebf2 4990 -((int64_t)alen), (flags &
39269e29 4991 XFS_BMAPI_RSVBLOCKS));
dd9f438e 4992 }
3bdbfb10 4993 if (!error) {
06d10dd9
NS
4994 error = xfs_mod_incore_sb(mp,
4995 XFS_SBS_FDBLOCKS,
20f4ebf2 4996 -((int64_t)indlen), (flags &
39269e29 4997 XFS_BMAPI_RSVBLOCKS));
3ddb8fa9
NS
4998 if (error && rt)
4999 xfs_mod_incore_sb(mp,
3bdbfb10 5000 XFS_SBS_FREXTENTS,
20f4ebf2 5001 (int64_t)extsz, (flags &
39269e29 5002 XFS_BMAPI_RSVBLOCKS));
3ddb8fa9
NS
5003 else if (error)
5004 xfs_mod_incore_sb(mp,
3bdbfb10 5005 XFS_SBS_FDBLOCKS,
20f4ebf2 5006 (int64_t)alen, (flags &
39269e29 5007 XFS_BMAPI_RSVBLOCKS));
3bdbfb10 5008 }
06d10dd9
NS
5009
5010 if (error) {
3ddb8fa9 5011 if (XFS_IS_QUOTA_ON(mp))
06d10dd9 5012 /* unreserve the blocks now */
dd9f438e 5013 (void)
06d10dd9
NS
5014 XFS_TRANS_UNRESERVE_QUOTA_NBLKS(
5015 mp, NULL, ip,
5016 (long)alen, 0, rt ?
5017 XFS_QMOPT_RES_RTBLKS :
5018 XFS_QMOPT_RES_REGBLKS);
1da177e4
LT
5019 break;
5020 }
06d10dd9 5021
1da177e4 5022 ip->i_delayed_blks += alen;
9d87c319 5023 abno = nullstartblock(indlen);
1da177e4
LT
5024 } else {
5025 /*
5026 * If first time, allocate and fill in
5027 * once-only bma fields.
5028 */
5029 if (bma.ip == NULL) {
5030 bma.tp = tp;
5031 bma.ip = ip;
5032 bma.prevp = &prev;
5033 bma.gotp = &got;
5034 bma.total = total;
5035 bma.userdata = 0;
5036 }
5037 /* Indicate if this is the first user data
5038 * in the file, or just any user data.
5039 */
39269e29 5040 if (!(flags & XFS_BMAPI_METADATA)) {
1da177e4
LT
5041 bma.userdata = (aoff == 0) ?
5042 XFS_ALLOC_INITIAL_USER_DATA :
5043 XFS_ALLOC_USERDATA;
5044 }
5045 /*
5046 * Fill in changeable bma fields.
5047 */
5048 bma.eof = eof;
5049 bma.firstblock = *firstblock;
5050 bma.alen = alen;
5051 bma.off = aoff;
7288026b 5052 bma.conv = !!(flags & XFS_BMAPI_CONVERT);
1da177e4
LT
5053 bma.wasdel = wasdelay;
5054 bma.minlen = minlen;
5055 bma.low = flist->xbf_low;
5056 bma.minleft = minleft;
5057 /*
5058 * Only want to do the alignment at the
5059 * eof if it is userdata and allocation length
5060 * is larger than a stripe unit.
5061 */
5062 if (mp->m_dalign && alen >= mp->m_dalign &&
39269e29
NS
5063 (!(flags & XFS_BMAPI_METADATA)) &&
5064 (whichfork == XFS_DATA_FORK)) {
1da177e4
LT
5065 if ((error = xfs_bmap_isaeof(ip, aoff,
5066 whichfork, &bma.aeof)))
5067 goto error0;
5068 } else
5069 bma.aeof = 0;
5070 /*
5071 * Call allocator.
5072 */
5073 if ((error = xfs_bmap_alloc(&bma)))
5074 goto error0;
5075 /*
5076 * Copy out result fields.
5077 */
5078 abno = bma.rval;
5079 if ((flist->xbf_low = bma.low))
5080 minleft = 0;
5081 alen = bma.alen;
5082 aoff = bma.off;
5083 ASSERT(*firstblock == NULLFSBLOCK ||
5084 XFS_FSB_TO_AGNO(mp, *firstblock) ==
5085 XFS_FSB_TO_AGNO(mp, bma.firstblock) ||
5086 (flist->xbf_low &&
5087 XFS_FSB_TO_AGNO(mp, *firstblock) <
5088 XFS_FSB_TO_AGNO(mp, bma.firstblock)));
5089 *firstblock = bma.firstblock;
5090 if (cur)
5091 cur->bc_private.b.firstblock =
5092 *firstblock;
5093 if (abno == NULLFSBLOCK)
5094 break;
5095 if ((ifp->if_flags & XFS_IFBROOT) && !cur) {
561f7d17 5096 cur = xfs_bmbt_init_cursor(mp, tp,
1da177e4
LT
5097 ip, whichfork);
5098 cur->bc_private.b.firstblock =
5099 *firstblock;
5100 cur->bc_private.b.flist = flist;
5101 }
5102 /*
5103 * Bump the number of extents we've allocated
5104 * in this call.
5105 */
5106 nallocs++;
5107 }
5108 if (cur)
5109 cur->bc_private.b.flags =
5110 wasdelay ? XFS_BTCUR_BPRV_WASDEL : 0;
5111 got.br_startoff = aoff;
5112 got.br_startblock = abno;
5113 got.br_blockcount = alen;
5114 got.br_state = XFS_EXT_NORM; /* assume normal */
5115 /*
5116 * Determine state of extent, and the filesystem.
5117 * A wasdelay extent has been initialized, so
5118 * shouldn't be flagged as unwritten.
5119 */
62118709 5120 if (wr && xfs_sb_version_hasextflgbit(&mp->m_sb)) {
1da177e4
LT
5121 if (!wasdelay && (flags & XFS_BMAPI_PREALLOC))
5122 got.br_state = XFS_EXT_UNWRITTEN;
5123 }
5124 error = xfs_bmap_add_extent(ip, lastx, &cur, &got,
3e57ecf6
OW
5125 firstblock, flist, &tmp_logflags, delta,
5126 whichfork, (flags & XFS_BMAPI_RSVBLOCKS));
1da177e4
LT
5127 logflags |= tmp_logflags;
5128 if (error)
5129 goto error0;
5130 lastx = ifp->if_lastex;
4eea22f0 5131 ep = xfs_iext_get_ext(ifp, lastx);
1da177e4
LT
5132 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
5133 xfs_bmbt_get_all(ep, &got);
5134 ASSERT(got.br_startoff <= aoff);
5135 ASSERT(got.br_startoff + got.br_blockcount >=
5136 aoff + alen);
5137#ifdef DEBUG
39269e29 5138 if (flags & XFS_BMAPI_DELAY) {
9d87c319
ES
5139 ASSERT(isnullstartblock(got.br_startblock));
5140 ASSERT(startblockval(got.br_startblock) > 0);
1da177e4
LT
5141 }
5142 ASSERT(got.br_state == XFS_EXT_NORM ||
5143 got.br_state == XFS_EXT_UNWRITTEN);
5144#endif
5145 /*
5146 * Fall down into the found allocated space case.
5147 */
5148 } else if (inhole) {
5149 /*
5150 * Reading in a hole.
5151 */
5152 mval->br_startoff = bno;
5153 mval->br_startblock = HOLESTARTBLOCK;
5154 mval->br_blockcount =
5155 XFS_FILBLKS_MIN(len, got.br_startoff - bno);
5156 mval->br_state = XFS_EXT_NORM;
5157 bno += mval->br_blockcount;
5158 len -= mval->br_blockcount;
5159 mval++;
5160 n++;
5161 continue;
5162 }
5163 /*
5164 * Then deal with the allocated space we found.
5165 */
5166 ASSERT(ep != NULL);
39269e29
NS
5167 if (!(flags & XFS_BMAPI_ENTIRE) &&
5168 (got.br_startoff + got.br_blockcount > obno)) {
1da177e4
LT
5169 if (obno > bno)
5170 bno = obno;
5171 ASSERT((bno >= obno) || (n == 0));
5172 ASSERT(bno < end);
5173 mval->br_startoff = bno;
9d87c319 5174 if (isnullstartblock(got.br_startblock)) {
39269e29 5175 ASSERT(!wr || (flags & XFS_BMAPI_DELAY));
1da177e4
LT
5176 mval->br_startblock = DELAYSTARTBLOCK;
5177 } else
5178 mval->br_startblock =
5179 got.br_startblock +
5180 (bno - got.br_startoff);
5181 /*
5182 * Return the minimum of what we got and what we
5183 * asked for for the length. We can use the len
5184 * variable here because it is modified below
5185 * and we could have been there before coming
5186 * here if the first part of the allocation
5187 * didn't overlap what was asked for.
5188 */
5189 mval->br_blockcount =
5190 XFS_FILBLKS_MIN(end - bno, got.br_blockcount -
5191 (bno - got.br_startoff));
5192 mval->br_state = got.br_state;
5193 ASSERT(mval->br_blockcount <= len);
5194 } else {
5195 *mval = got;
9d87c319 5196 if (isnullstartblock(mval->br_startblock)) {
39269e29 5197 ASSERT(!wr || (flags & XFS_BMAPI_DELAY));
1da177e4
LT
5198 mval->br_startblock = DELAYSTARTBLOCK;
5199 }
5200 }
5201
5202 /*
5203 * Check if writing previously allocated but
5204 * unwritten extents.
5205 */
5206 if (wr && mval->br_state == XFS_EXT_UNWRITTEN &&
5207 ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_DELAY)) == 0)) {
5208 /*
5209 * Modify (by adding) the state flag, if writing.
5210 */
5211 ASSERT(mval->br_blockcount <= len);
5212 if ((ifp->if_flags & XFS_IFBROOT) && !cur) {
561f7d17
CH
5213 cur = xfs_bmbt_init_cursor(mp,
5214 tp, ip, whichfork);
1da177e4
LT
5215 cur->bc_private.b.firstblock =
5216 *firstblock;
5217 cur->bc_private.b.flist = flist;
5218 }
5219 mval->br_state = XFS_EXT_NORM;
5220 error = xfs_bmap_add_extent(ip, lastx, &cur, mval,
3e57ecf6
OW
5221 firstblock, flist, &tmp_logflags, delta,
5222 whichfork, (flags & XFS_BMAPI_RSVBLOCKS));
1da177e4
LT
5223 logflags |= tmp_logflags;
5224 if (error)
5225 goto error0;
5226 lastx = ifp->if_lastex;
4eea22f0 5227 ep = xfs_iext_get_ext(ifp, lastx);
1da177e4
LT
5228 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
5229 xfs_bmbt_get_all(ep, &got);
5230 /*
5231 * We may have combined previously unwritten
5232 * space with written space, so generate
5233 * another request.
5234 */
5235 if (mval->br_blockcount < len)
5236 continue;
5237 }
5238
39269e29 5239 ASSERT((flags & XFS_BMAPI_ENTIRE) ||
1da177e4 5240 ((mval->br_startoff + mval->br_blockcount) <= end));
39269e29
NS
5241 ASSERT((flags & XFS_BMAPI_ENTIRE) ||
5242 (mval->br_blockcount <= len) ||
1da177e4
LT
5243 (mval->br_startoff < obno));
5244 bno = mval->br_startoff + mval->br_blockcount;
5245 len = end - bno;
5246 if (n > 0 && mval->br_startoff == mval[-1].br_startoff) {
5247 ASSERT(mval->br_startblock == mval[-1].br_startblock);
5248 ASSERT(mval->br_blockcount > mval[-1].br_blockcount);
5249 ASSERT(mval->br_state == mval[-1].br_state);
5250 mval[-1].br_blockcount = mval->br_blockcount;
5251 mval[-1].br_state = mval->br_state;
5252 } else if (n > 0 && mval->br_startblock != DELAYSTARTBLOCK &&
5253 mval[-1].br_startblock != DELAYSTARTBLOCK &&
5254 mval[-1].br_startblock != HOLESTARTBLOCK &&
5255 mval->br_startblock ==
5256 mval[-1].br_startblock + mval[-1].br_blockcount &&
39269e29
NS
5257 ((flags & XFS_BMAPI_IGSTATE) ||
5258 mval[-1].br_state == mval->br_state)) {
1da177e4
LT
5259 ASSERT(mval->br_startoff ==
5260 mval[-1].br_startoff + mval[-1].br_blockcount);
5261 mval[-1].br_blockcount += mval->br_blockcount;
5262 } else if (n > 0 &&
5263 mval->br_startblock == DELAYSTARTBLOCK &&
5264 mval[-1].br_startblock == DELAYSTARTBLOCK &&
5265 mval->br_startoff ==
5266 mval[-1].br_startoff + mval[-1].br_blockcount) {
5267 mval[-1].br_blockcount += mval->br_blockcount;
5268 mval[-1].br_state = mval->br_state;
5269 } else if (!((n == 0) &&
5270 ((mval->br_startoff + mval->br_blockcount) <=
5271 obno))) {
5272 mval++;
5273 n++;
5274 }
5275 /*
5276 * If we're done, stop now. Stop when we've allocated
5277 * XFS_BMAP_MAX_NMAP extents no matter what. Otherwise
5278 * the transaction may get too big.
5279 */
5280 if (bno >= end || n >= *nmap || nallocs >= *nmap)
5281 break;
5282 /*
5283 * Else go on to the next record.
5284 */
4eea22f0 5285 ep = xfs_iext_get_ext(ifp, ++lastx);
4e5ae838
DC
5286 prev = got;
5287 if (lastx >= nextents)
1da177e4 5288 eof = 1;
4e5ae838 5289 else
1da177e4
LT
5290 xfs_bmbt_get_all(ep, &got);
5291 }
5292 ifp->if_lastex = lastx;
5293 *nmap = n;
5294 /*
5295 * Transform from btree to extents, give it cur.
5296 */
5297 if (tp && XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE &&
5298 XFS_IFORK_NEXTENTS(ip, whichfork) <= ifp->if_ext_max) {
5299 ASSERT(wr && cur);
5300 error = xfs_bmap_btree_to_extents(tp, ip, cur,
5301 &tmp_logflags, whichfork);
5302 logflags |= tmp_logflags;
5303 if (error)
5304 goto error0;
5305 }
5306 ASSERT(ifp->if_ext_max ==
5307 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
5308 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE ||
5309 XFS_IFORK_NEXTENTS(ip, whichfork) > ifp->if_ext_max);
5310 error = 0;
3e57ecf6
OW
5311 if (delta && delta->xed_startoff != NULLFILEOFF) {
5312 /* A change was actually made.
5313 * Note that delta->xed_blockount is an offset at this
5314 * point and needs to be converted to a block count.
5315 */
5316 ASSERT(delta->xed_blockcount > delta->xed_startoff);
5317 delta->xed_blockcount -= delta->xed_startoff;
5318 }
1da177e4
LT
5319error0:
5320 /*
5321 * Log everything. Do this after conversion, there's no point in
4eea22f0 5322 * logging the extent records if we've converted to btree format.
1da177e4 5323 */
9d87c319 5324 if ((logflags & xfs_ilog_fext(whichfork)) &&
1da177e4 5325 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS)
9d87c319
ES
5326 logflags &= ~xfs_ilog_fext(whichfork);
5327 else if ((logflags & xfs_ilog_fbroot(whichfork)) &&
1da177e4 5328 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE)
9d87c319 5329 logflags &= ~xfs_ilog_fbroot(whichfork);
1da177e4
LT
5330 /*
5331 * Log whatever the flags say, even if error. Otherwise we might miss
5332 * detecting a case where the data is changed, there's an error,
5333 * and it's not logged so we don't shutdown when we should.
5334 */
5335 if (logflags) {
5336 ASSERT(tp && wr);
5337 xfs_trans_log_inode(tp, ip, logflags);
5338 }
5339 if (cur) {
5340 if (!error) {
5341 ASSERT(*firstblock == NULLFSBLOCK ||
5342 XFS_FSB_TO_AGNO(mp, *firstblock) ==
5343 XFS_FSB_TO_AGNO(mp,
5344 cur->bc_private.b.firstblock) ||
5345 (flist->xbf_low &&
5346 XFS_FSB_TO_AGNO(mp, *firstblock) <
5347 XFS_FSB_TO_AGNO(mp,
5348 cur->bc_private.b.firstblock)));
5349 *firstblock = cur->bc_private.b.firstblock;
5350 }
5351 xfs_btree_del_cursor(cur,
5352 error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
5353 }
5354 if (!error)
5355 xfs_bmap_validate_ret(orig_bno, orig_len, orig_flags, orig_mval,
5356 orig_nmap, *nmap);
5357 return error;
5358}
5359
5360/*
5361 * Map file blocks to filesystem blocks, simple version.
5362 * One block (extent) only, read-only.
5363 * For flags, only the XFS_BMAPI_ATTRFORK flag is examined.
5364 * For the other flag values, the effect is as if XFS_BMAPI_METADATA
5365 * was set and all the others were clear.
5366 */
5367int /* error */
5368xfs_bmapi_single(
5369 xfs_trans_t *tp, /* transaction pointer */
5370 xfs_inode_t *ip, /* incore inode */
5371 int whichfork, /* data or attr fork */
5372 xfs_fsblock_t *fsb, /* output: mapped block */
5373 xfs_fileoff_t bno) /* starting file offs. mapped */
5374{
4eea22f0 5375 int eof; /* we've hit the end of extents */
1da177e4 5376 int error; /* error return */
4eea22f0 5377 xfs_bmbt_irec_t got; /* current file extent record */
1da177e4
LT
5378 xfs_ifork_t *ifp; /* inode fork pointer */
5379 xfs_extnum_t lastx; /* last useful extent number */
4eea22f0 5380 xfs_bmbt_irec_t prev; /* previous file extent record */
1da177e4
LT
5381
5382 ifp = XFS_IFORK_PTR(ip, whichfork);
5383 if (unlikely(
5384 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE &&
5385 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS)) {
5386 XFS_ERROR_REPORT("xfs_bmapi_single", XFS_ERRLEVEL_LOW,
5387 ip->i_mount);
5388 return XFS_ERROR(EFSCORRUPTED);
5389 }
5390 if (XFS_FORCED_SHUTDOWN(ip->i_mount))
5391 return XFS_ERROR(EIO);
5392 XFS_STATS_INC(xs_blk_mapr);
5393 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
5394 (error = xfs_iread_extents(tp, ip, whichfork)))
5395 return error;
5396 (void)xfs_bmap_search_extents(ip, bno, whichfork, &eof, &lastx, &got,
5397 &prev);
5398 /*
5399 * Reading past eof, act as though there's a hole
5400 * up to end.
5401 */
5402 if (eof || got.br_startoff > bno) {
5403 *fsb = NULLFSBLOCK;
5404 return 0;
5405 }
9d87c319 5406 ASSERT(!isnullstartblock(got.br_startblock));
1da177e4
LT
5407 ASSERT(bno < got.br_startoff + got.br_blockcount);
5408 *fsb = got.br_startblock + (bno - got.br_startoff);
5409 ifp->if_lastex = lastx;
5410 return 0;
5411}
5412
5413/*
5414 * Unmap (remove) blocks from a file.
5415 * If nexts is nonzero then the number of extents to remove is limited to
5416 * that value. If not all extents in the block range can be removed then
5417 * *done is set.
5418 */
5419int /* error */
5420xfs_bunmapi(
5421 xfs_trans_t *tp, /* transaction pointer */
5422 struct xfs_inode *ip, /* incore inode */
5423 xfs_fileoff_t bno, /* starting offset to unmap */
5424 xfs_filblks_t len, /* length to unmap in file */
5425 int flags, /* misc flags */
5426 xfs_extnum_t nexts, /* number of extents max */
5427 xfs_fsblock_t *firstblock, /* first allocated block
5428 controls a.g. for allocs */
5429 xfs_bmap_free_t *flist, /* i/o: list extents to free */
3e57ecf6
OW
5430 xfs_extdelta_t *delta, /* o: change made to incore
5431 extents */
1da177e4
LT
5432 int *done) /* set if not done yet */
5433{
5434 xfs_btree_cur_t *cur; /* bmap btree cursor */
5435 xfs_bmbt_irec_t del; /* extent being deleted */
5436 int eof; /* is deleting at eof */
a6f64d4a 5437 xfs_bmbt_rec_host_t *ep; /* extent record pointer */
1da177e4
LT
5438 int error; /* error return value */
5439 xfs_extnum_t extno; /* extent number in list */
4eea22f0 5440 xfs_bmbt_irec_t got; /* current extent record */
1da177e4
LT
5441 xfs_ifork_t *ifp; /* inode fork pointer */
5442 int isrt; /* freeing in rt area */
5443 xfs_extnum_t lastx; /* last extent index used */
5444 int logflags; /* transaction logging flags */
5445 xfs_extlen_t mod; /* rt extent offset */
5446 xfs_mount_t *mp; /* mount structure */
4eea22f0
MK
5447 xfs_extnum_t nextents; /* number of file extents */
5448 xfs_bmbt_irec_t prev; /* previous extent record */
1da177e4
LT
5449 xfs_fileoff_t start; /* first file offset deleted */
5450 int tmp_logflags; /* partial logging flags */
5451 int wasdel; /* was a delayed alloc extent */
5452 int whichfork; /* data or attribute fork */
5453 int rsvd; /* OK to allocate reserved blocks */
5454 xfs_fsblock_t sum;
5455
5456 xfs_bunmap_trace(ip, bno, len, flags, (inst_t *)__return_address);
5457 whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
5458 XFS_ATTR_FORK : XFS_DATA_FORK;
5459 ifp = XFS_IFORK_PTR(ip, whichfork);
5460 if (unlikely(
5461 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
5462 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE)) {
5463 XFS_ERROR_REPORT("xfs_bunmapi", XFS_ERRLEVEL_LOW,
5464 ip->i_mount);
5465 return XFS_ERROR(EFSCORRUPTED);
5466 }
5467 mp = ip->i_mount;
5468 if (XFS_FORCED_SHUTDOWN(mp))
5469 return XFS_ERROR(EIO);
5470 rsvd = (flags & XFS_BMAPI_RSVBLOCKS) != 0;
5471 ASSERT(len > 0);
5472 ASSERT(nexts >= 0);
5473 ASSERT(ifp->if_ext_max ==
5474 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
5475 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
5476 (error = xfs_iread_extents(tp, ip, whichfork)))
5477 return error;
5478 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
5479 if (nextents == 0) {
5480 *done = 1;
5481 return 0;
5482 }
5483 XFS_STATS_INC(xs_blk_unmap);
dd9f438e 5484 isrt = (whichfork == XFS_DATA_FORK) && XFS_IS_REALTIME_INODE(ip);
1da177e4
LT
5485 start = bno;
5486 bno = start + len - 1;
5487 ep = xfs_bmap_search_extents(ip, bno, whichfork, &eof, &lastx, &got,
5488 &prev);
3e57ecf6
OW
5489 if (delta) {
5490 delta->xed_startoff = NULLFILEOFF;
5491 delta->xed_blockcount = 0;
5492 }
1da177e4
LT
5493 /*
5494 * Check to see if the given block number is past the end of the
5495 * file, back up to the last block if so...
5496 */
5497 if (eof) {
4eea22f0 5498 ep = xfs_iext_get_ext(ifp, --lastx);
1da177e4
LT
5499 xfs_bmbt_get_all(ep, &got);
5500 bno = got.br_startoff + got.br_blockcount - 1;
5501 }
5502 logflags = 0;
5503 if (ifp->if_flags & XFS_IFBROOT) {
5504 ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
561f7d17 5505 cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
1da177e4
LT
5506 cur->bc_private.b.firstblock = *firstblock;
5507 cur->bc_private.b.flist = flist;
5508 cur->bc_private.b.flags = 0;
5509 } else
5510 cur = NULL;
5511 extno = 0;
5512 while (bno != (xfs_fileoff_t)-1 && bno >= start && lastx >= 0 &&
5513 (nexts == 0 || extno < nexts)) {
5514 /*
5515 * Is the found extent after a hole in which bno lives?
5516 * Just back up to the previous extent, if so.
5517 */
5518 if (got.br_startoff > bno) {
5519 if (--lastx < 0)
5520 break;
4eea22f0 5521 ep = xfs_iext_get_ext(ifp, lastx);
1da177e4
LT
5522 xfs_bmbt_get_all(ep, &got);
5523 }
5524 /*
5525 * Is the last block of this extent before the range
5526 * we're supposed to delete? If so, we're done.
5527 */
5528 bno = XFS_FILEOFF_MIN(bno,
5529 got.br_startoff + got.br_blockcount - 1);
5530 if (bno < start)
5531 break;
5532 /*
5533 * Then deal with the (possibly delayed) allocated space
5534 * we found.
5535 */
5536 ASSERT(ep != NULL);
5537 del = got;
9d87c319 5538 wasdel = isnullstartblock(del.br_startblock);
1da177e4
LT
5539 if (got.br_startoff < start) {
5540 del.br_startoff = start;
5541 del.br_blockcount -= start - got.br_startoff;
5542 if (!wasdel)
5543 del.br_startblock += start - got.br_startoff;
5544 }
5545 if (del.br_startoff + del.br_blockcount > bno + 1)
5546 del.br_blockcount = bno + 1 - del.br_startoff;
5547 sum = del.br_startblock + del.br_blockcount;
5548 if (isrt &&
5549 (mod = do_mod(sum, mp->m_sb.sb_rextsize))) {
5550 /*
5551 * Realtime extent not lined up at the end.
5552 * The extent could have been split into written
5553 * and unwritten pieces, or we could just be
5554 * unmapping part of it. But we can't really
5555 * get rid of part of a realtime extent.
5556 */
5557 if (del.br_state == XFS_EXT_UNWRITTEN ||
62118709 5558 !xfs_sb_version_hasextflgbit(&mp->m_sb)) {
1da177e4
LT
5559 /*
5560 * This piece is unwritten, or we're not
5561 * using unwritten extents. Skip over it.
5562 */
5563 ASSERT(bno >= mod);
5564 bno -= mod > del.br_blockcount ?
5565 del.br_blockcount : mod;
5566 if (bno < got.br_startoff) {
5567 if (--lastx >= 0)
4eea22f0
MK
5568 xfs_bmbt_get_all(xfs_iext_get_ext(
5569 ifp, lastx), &got);
1da177e4
LT
5570 }
5571 continue;
5572 }
5573 /*
5574 * It's written, turn it unwritten.
5575 * This is better than zeroing it.
5576 */
5577 ASSERT(del.br_state == XFS_EXT_NORM);
5578 ASSERT(xfs_trans_get_block_res(tp) > 0);
5579 /*
5580 * If this spans a realtime extent boundary,
5581 * chop it back to the start of the one we end at.
5582 */
5583 if (del.br_blockcount > mod) {
5584 del.br_startoff += del.br_blockcount - mod;
5585 del.br_startblock += del.br_blockcount - mod;
5586 del.br_blockcount = mod;
5587 }
5588 del.br_state = XFS_EXT_UNWRITTEN;
5589 error = xfs_bmap_add_extent(ip, lastx, &cur, &del,
3e57ecf6
OW
5590 firstblock, flist, &logflags, delta,
5591 XFS_DATA_FORK, 0);
1da177e4
LT
5592 if (error)
5593 goto error0;
5594 goto nodelete;
5595 }
5596 if (isrt && (mod = do_mod(del.br_startblock, mp->m_sb.sb_rextsize))) {
5597 /*
5598 * Realtime extent is lined up at the end but not
5599 * at the front. We'll get rid of full extents if
5600 * we can.
5601 */
5602 mod = mp->m_sb.sb_rextsize - mod;
5603 if (del.br_blockcount > mod) {
5604 del.br_blockcount -= mod;
5605 del.br_startoff += mod;
5606 del.br_startblock += mod;
5607 } else if ((del.br_startoff == start &&
5608 (del.br_state == XFS_EXT_UNWRITTEN ||
5609 xfs_trans_get_block_res(tp) == 0)) ||
62118709 5610 !xfs_sb_version_hasextflgbit(&mp->m_sb)) {
1da177e4
LT
5611 /*
5612 * Can't make it unwritten. There isn't
5613 * a full extent here so just skip it.
5614 */
5615 ASSERT(bno >= del.br_blockcount);
5616 bno -= del.br_blockcount;
5617 if (bno < got.br_startoff) {
5618 if (--lastx >= 0)
5619 xfs_bmbt_get_all(--ep, &got);
5620 }
5621 continue;
5622 } else if (del.br_state == XFS_EXT_UNWRITTEN) {
5623 /*
5624 * This one is already unwritten.
5625 * It must have a written left neighbor.
5626 * Unwrite the killed part of that one and
5627 * try again.
5628 */
5629 ASSERT(lastx > 0);
4eea22f0
MK
5630 xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
5631 lastx - 1), &prev);
1da177e4 5632 ASSERT(prev.br_state == XFS_EXT_NORM);
9d87c319 5633 ASSERT(!isnullstartblock(prev.br_startblock));
1da177e4
LT
5634 ASSERT(del.br_startblock ==
5635 prev.br_startblock + prev.br_blockcount);
5636 if (prev.br_startoff < start) {
5637 mod = start - prev.br_startoff;
5638 prev.br_blockcount -= mod;
5639 prev.br_startblock += mod;
5640 prev.br_startoff = start;
5641 }
5642 prev.br_state = XFS_EXT_UNWRITTEN;
5643 error = xfs_bmap_add_extent(ip, lastx - 1, &cur,
5644 &prev, firstblock, flist, &logflags,
3e57ecf6 5645 delta, XFS_DATA_FORK, 0);
1da177e4
LT
5646 if (error)
5647 goto error0;
5648 goto nodelete;
5649 } else {
5650 ASSERT(del.br_state == XFS_EXT_NORM);
5651 del.br_state = XFS_EXT_UNWRITTEN;
5652 error = xfs_bmap_add_extent(ip, lastx, &cur,
5653 &del, firstblock, flist, &logflags,
3e57ecf6 5654 delta, XFS_DATA_FORK, 0);
1da177e4
LT
5655 if (error)
5656 goto error0;
5657 goto nodelete;
5658 }
5659 }
5660 if (wasdel) {
9d87c319 5661 ASSERT(startblockval(del.br_startblock) > 0);
dd9f438e 5662 /* Update realtime/data freespace, unreserve quota */
06d10dd9
NS
5663 if (isrt) {
5664 xfs_filblks_t rtexts;
5665
5666 rtexts = XFS_FSB_TO_B(mp, del.br_blockcount);
5667 do_div(rtexts, mp->m_sb.sb_rextsize);
5668 xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS,
20f4ebf2 5669 (int64_t)rtexts, rsvd);
dd9f438e
NS
5670 (void)XFS_TRANS_RESERVE_QUOTA_NBLKS(mp,
5671 NULL, ip, -((long)del.br_blockcount), 0,
06d10dd9
NS
5672 XFS_QMOPT_RES_RTBLKS);
5673 } else {
5674 xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS,
20f4ebf2 5675 (int64_t)del.br_blockcount, rsvd);
dd9f438e
NS
5676 (void)XFS_TRANS_RESERVE_QUOTA_NBLKS(mp,
5677 NULL, ip, -((long)del.br_blockcount), 0,
1da177e4 5678 XFS_QMOPT_RES_REGBLKS);
06d10dd9 5679 }
1da177e4
LT
5680 ip->i_delayed_blks -= del.br_blockcount;
5681 if (cur)
5682 cur->bc_private.b.flags |=
5683 XFS_BTCUR_BPRV_WASDEL;
5684 } else if (cur)
5685 cur->bc_private.b.flags &= ~XFS_BTCUR_BPRV_WASDEL;
5686 /*
5687 * If it's the case where the directory code is running
5688 * with no block reservation, and the deleted block is in
5689 * the middle of its extent, and the resulting insert
5690 * of an extent would cause transformation to btree format,
5691 * then reject it. The calling code will then swap
5692 * blocks around instead.
5693 * We have to do this now, rather than waiting for the
5694 * conversion to btree format, since the transaction
5695 * will be dirty.
5696 */
5697 if (!wasdel && xfs_trans_get_block_res(tp) == 0 &&
5698 XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS &&
5699 XFS_IFORK_NEXTENTS(ip, whichfork) >= ifp->if_ext_max &&
5700 del.br_startoff > got.br_startoff &&
5701 del.br_startoff + del.br_blockcount <
5702 got.br_startoff + got.br_blockcount) {
5703 error = XFS_ERROR(ENOSPC);
5704 goto error0;
5705 }
5706 error = xfs_bmap_del_extent(ip, tp, lastx, flist, cur, &del,
3e57ecf6 5707 &tmp_logflags, delta, whichfork, rsvd);
1da177e4
LT
5708 logflags |= tmp_logflags;
5709 if (error)
5710 goto error0;
5711 bno = del.br_startoff - 1;
5712nodelete:
5713 lastx = ifp->if_lastex;
5714 /*
5715 * If not done go on to the next (previous) record.
5716 * Reset ep in case the extents array was re-alloced.
5717 */
4eea22f0 5718 ep = xfs_iext_get_ext(ifp, lastx);
1da177e4
LT
5719 if (bno != (xfs_fileoff_t)-1 && bno >= start) {
5720 if (lastx >= XFS_IFORK_NEXTENTS(ip, whichfork) ||
5721 xfs_bmbt_get_startoff(ep) > bno) {
4eea22f0
MK
5722 if (--lastx >= 0)
5723 ep = xfs_iext_get_ext(ifp, lastx);
1da177e4
LT
5724 }
5725 if (lastx >= 0)
5726 xfs_bmbt_get_all(ep, &got);
5727 extno++;
5728 }
5729 }
5730 ifp->if_lastex = lastx;
5731 *done = bno == (xfs_fileoff_t)-1 || bno < start || lastx < 0;
5732 ASSERT(ifp->if_ext_max ==
5733 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
5734 /*
5735 * Convert to a btree if necessary.
5736 */
5737 if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS &&
5738 XFS_IFORK_NEXTENTS(ip, whichfork) > ifp->if_ext_max) {
5739 ASSERT(cur == NULL);
5740 error = xfs_bmap_extents_to_btree(tp, ip, firstblock, flist,
5741 &cur, 0, &tmp_logflags, whichfork);
5742 logflags |= tmp_logflags;
5743 if (error)
5744 goto error0;
5745 }
5746 /*
5747 * transform from btree to extents, give it cur
5748 */
5749 else if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE &&
5750 XFS_IFORK_NEXTENTS(ip, whichfork) <= ifp->if_ext_max) {
5751 ASSERT(cur != NULL);
5752 error = xfs_bmap_btree_to_extents(tp, ip, cur, &tmp_logflags,
5753 whichfork);
5754 logflags |= tmp_logflags;
5755 if (error)
5756 goto error0;
5757 }
5758 /*
5759 * transform from extents to local?
5760 */
5761 ASSERT(ifp->if_ext_max ==
5762 XFS_IFORK_SIZE(ip, whichfork) / (uint)sizeof(xfs_bmbt_rec_t));
5763 error = 0;
3e57ecf6
OW
5764 if (delta && delta->xed_startoff != NULLFILEOFF) {
5765 /* A change was actually made.
5766 * Note that delta->xed_blockount is an offset at this
5767 * point and needs to be converted to a block count.
5768 */
5769 ASSERT(delta->xed_blockcount > delta->xed_startoff);
5770 delta->xed_blockcount -= delta->xed_startoff;
5771 }
1da177e4
LT
5772error0:
5773 /*
5774 * Log everything. Do this after conversion, there's no point in
4eea22f0 5775 * logging the extent records if we've converted to btree format.
1da177e4 5776 */
9d87c319 5777 if ((logflags & xfs_ilog_fext(whichfork)) &&
1da177e4 5778 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS)
9d87c319
ES
5779 logflags &= ~xfs_ilog_fext(whichfork);
5780 else if ((logflags & xfs_ilog_fbroot(whichfork)) &&
1da177e4 5781 XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE)
9d87c319 5782 logflags &= ~xfs_ilog_fbroot(whichfork);
1da177e4
LT
5783 /*
5784 * Log inode even in the error case, if the transaction
5785 * is dirty we'll need to shut down the filesystem.
5786 */
5787 if (logflags)
5788 xfs_trans_log_inode(tp, ip, logflags);
5789 if (cur) {
5790 if (!error) {
5791 *firstblock = cur->bc_private.b.firstblock;
5792 cur->bc_private.b.allocated = 0;
5793 }
5794 xfs_btree_del_cursor(cur,
5795 error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
5796 }
5797 return error;
5798}
5799
3bacbcd8
VA
5800/*
5801 * returns 1 for success, 0 if we failed to map the extent.
5802 */
5803STATIC int
5804xfs_getbmapx_fix_eof_hole(
5805 xfs_inode_t *ip, /* xfs incore inode pointer */
8a7141a8 5806 struct getbmapx *out, /* output structure */
3bacbcd8 5807 int prealloced, /* this is a file with
8a7141a8 5808 * preallocated data space */
3bacbcd8
VA
5809 __int64_t end, /* last block requested */
5810 xfs_fsblock_t startblock)
5811{
5812 __int64_t fixlen;
5813 xfs_mount_t *mp; /* file system mount point */
5af317c9
ES
5814 xfs_ifork_t *ifp; /* inode fork pointer */
5815 xfs_extnum_t lastx; /* last extent pointer */
5816 xfs_fileoff_t fileblock;
3bacbcd8
VA
5817
5818 if (startblock == HOLESTARTBLOCK) {
5819 mp = ip->i_mount;
5820 out->bmv_block = -1;
5821 fixlen = XFS_FSB_TO_BB(mp, XFS_B_TO_FSB(mp, ip->i_size));
5822 fixlen -= out->bmv_offset;
5823 if (prealloced && out->bmv_offset + out->bmv_length == end) {
5824 /* Came to hole at EOF. Trim it. */
5825 if (fixlen <= 0)
5826 return 0;
5827 out->bmv_length = fixlen;
5828 }
5829 } else {
5af317c9
ES
5830 if (startblock == DELAYSTARTBLOCK)
5831 out->bmv_block = -2;
5832 else
9d87c319 5833 out->bmv_block = xfs_fsb_to_db(ip, startblock);
5af317c9
ES
5834 fileblock = XFS_BB_TO_FSB(ip->i_mount, out->bmv_offset);
5835 ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
5836 if (xfs_iext_bno_to_ext(ifp, fileblock, &lastx) &&
5837 (lastx == (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t))-1))
5838 out->bmv_oflags |= BMV_OF_LAST;
3bacbcd8
VA
5839 }
5840
5841 return 1;
5842}
5843
1da177e4 5844/*
8a7141a8
ES
5845 * Get inode's extents as described in bmv, and format for output.
5846 * Calls formatter to fill the user's buffer until all extents
5847 * are mapped, until the passed-in bmv->bmv_count slots have
5848 * been filled, or until the formatter short-circuits the loop,
5849 * if it is tracking filled-in extents on its own.
1da177e4
LT
5850 */
5851int /* error code */
5852xfs_getbmap(
993386c1 5853 xfs_inode_t *ip,
8a7141a8
ES
5854 struct getbmapx *bmv, /* user bmap structure */
5855 xfs_bmap_format_t formatter, /* format to user */
5856 void *arg) /* formatter arg */
1da177e4
LT
5857{
5858 __int64_t bmvend; /* last block requested */
5859 int error; /* return value */
5860 __int64_t fixlen; /* length for -1 case */
5861 int i; /* extent number */
1da177e4
LT
5862 int lock; /* lock state */
5863 xfs_bmbt_irec_t *map; /* buffer for user's data */
5864 xfs_mount_t *mp; /* file system mount point */
5865 int nex; /* # of user extents can do */
5866 int nexleft; /* # of user extents left */
5867 int subnex; /* # of bmapi's can do */
5868 int nmap; /* number of map entries */
8a7141a8 5869 struct getbmapx out; /* output structure */
1da177e4
LT
5870 int whichfork; /* data or attr fork */
5871 int prealloced; /* this is a file with
5872 * preallocated data space */
8a7141a8 5873 int iflags; /* interface flags */
1da177e4 5874 int bmapi_flags; /* flags for xfs_bmapi */
1da177e4 5875
1da177e4 5876 mp = ip->i_mount;
8a7141a8 5877 iflags = bmv->bmv_iflags;
1da177e4 5878
8a7141a8 5879 whichfork = iflags & BMV_IF_ATTRFORK ? XFS_ATTR_FORK : XFS_DATA_FORK;
1da177e4
LT
5880
5881 /* If the BMV_IF_NO_DMAPI_READ interface bit specified, do not
5882 * generate a DMAPI read event. Otherwise, if the DM_EVENT_READ
5883 * bit is set for the file, generate a read event in order
5884 * that the DMAPI application may do its thing before we return
5885 * the extents. Usually this means restoring user file data to
5886 * regions of the file that look like holes.
5887 *
5888 * The "old behavior" (from XFS_IOC_GETBMAP) is to not specify
5889 * BMV_IF_NO_DMAPI_READ so that read events are generated.
5890 * If this were not true, callers of ioctl( XFS_IOC_GETBMAP )
5891 * could misinterpret holes in a DMAPI file as true holes,
5892 * when in fact they may represent offline user data.
5893 */
8a7141a8 5894 if ((iflags & BMV_IF_NO_DMAPI_READ) == 0 &&
eb9df39d
CH
5895 DM_EVENT_ENABLED(ip, DM_EVENT_READ) &&
5896 whichfork == XFS_DATA_FORK) {
bc4ac74a 5897 error = XFS_SEND_DATA(mp, DM_EVENT_READ, ip, 0, 0, 0, NULL);
1da177e4
LT
5898 if (error)
5899 return XFS_ERROR(error);
5900 }
5901
5902 if (whichfork == XFS_ATTR_FORK) {
5903 if (XFS_IFORK_Q(ip)) {
5904 if (ip->i_d.di_aformat != XFS_DINODE_FMT_EXTENTS &&
5905 ip->i_d.di_aformat != XFS_DINODE_FMT_BTREE &&
5906 ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)
5907 return XFS_ERROR(EINVAL);
5908 } else if (unlikely(
5909 ip->i_d.di_aformat != 0 &&
5910 ip->i_d.di_aformat != XFS_DINODE_FMT_EXTENTS)) {
5911 XFS_ERROR_REPORT("xfs_getbmap", XFS_ERRLEVEL_LOW,
5912 ip->i_mount);
5913 return XFS_ERROR(EFSCORRUPTED);
5914 }
5915 } else if (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS &&
5916 ip->i_d.di_format != XFS_DINODE_FMT_BTREE &&
5917 ip->i_d.di_format != XFS_DINODE_FMT_LOCAL)
5918 return XFS_ERROR(EINVAL);
5919 if (whichfork == XFS_DATA_FORK) {
957d0ebe 5920 if (xfs_get_extsz_hint(ip) ||
dd9f438e 5921 ip->i_d.di_flags & (XFS_DIFLAG_PREALLOC|XFS_DIFLAG_APPEND)){
1da177e4
LT
5922 prealloced = 1;
5923 fixlen = XFS_MAXIOFFSET(mp);
5924 } else {
5925 prealloced = 0;
ba87ea69 5926 fixlen = ip->i_size;
1da177e4
LT
5927 }
5928 } else {
5929 prealloced = 0;
5930 fixlen = 1LL << 32;
5931 }
5932
5933 if (bmv->bmv_length == -1) {
5934 fixlen = XFS_FSB_TO_BB(mp, XFS_B_TO_FSB(mp, fixlen));
5935 bmv->bmv_length = MAX( (__int64_t)(fixlen - bmv->bmv_offset),
5936 (__int64_t)0);
5937 } else if (bmv->bmv_length < 0)
5938 return XFS_ERROR(EINVAL);
5939 if (bmv->bmv_length == 0) {
5940 bmv->bmv_entries = 0;
5941 return 0;
5942 }
5943 nex = bmv->bmv_count - 1;
5944 if (nex <= 0)
5945 return XFS_ERROR(EINVAL);
5946 bmvend = bmv->bmv_offset + bmv->bmv_length;
5947
5948 xfs_ilock(ip, XFS_IOLOCK_SHARED);
5949
5af317c9
ES
5950 if (((iflags & BMV_IF_DELALLOC) == 0) &&
5951 (whichfork == XFS_DATA_FORK) &&
5952 (ip->i_delayed_blks || ip->i_size > ip->i_d.di_size)) {
1da177e4 5953 /* xfs_fsize_t last_byte = xfs_file_last_byte(ip); */
739bfb2a
CH
5954 error = xfs_flush_pages(ip, (xfs_off_t)0,
5955 -1, 0, FI_REMAPF);
e12070a5
NS
5956 if (error) {
5957 xfs_iunlock(ip, XFS_IOLOCK_SHARED);
5958 return error;
5959 }
1da177e4
LT
5960 }
5961
5af317c9
ES
5962 ASSERT(whichfork == XFS_ATTR_FORK || (iflags & BMV_IF_DELALLOC) ||
5963 ip->i_delayed_blks == 0);
1da177e4
LT
5964
5965 lock = xfs_ilock_map_shared(ip);
5966
5967 /*
5968 * Don't let nex be bigger than the number of extents
5969 * we can have assuming alternating holes and real extents.
5970 */
5971 if (nex > XFS_IFORK_NEXTENTS(ip, whichfork) * 2 + 1)
5972 nex = XFS_IFORK_NEXTENTS(ip, whichfork) * 2 + 1;
5973
9d87c319 5974 bmapi_flags = xfs_bmapi_aflag(whichfork) |
5af317c9 5975 ((iflags & BMV_IF_PREALLOC) ? 0 : XFS_BMAPI_IGSTATE);
1da177e4
LT
5976
5977 /*
5978 * Allocate enough space to handle "subnex" maps at a time.
5979 */
5980 subnex = 16;
5981 map = kmem_alloc(subnex * sizeof(*map), KM_SLEEP);
5982
5983 bmv->bmv_entries = 0;
5984
5af317c9
ES
5985 if ((XFS_IFORK_NEXTENTS(ip, whichfork) == 0)) {
5986 if (((iflags & BMV_IF_DELALLOC) == 0) ||
5987 whichfork == XFS_ATTR_FORK) {
5988 error = 0;
5989 goto unlock_and_return;
5990 }
1da177e4
LT
5991 }
5992
5993 nexleft = nex;
5994
5995 do {
5996 nmap = (nexleft > subnex) ? subnex : nexleft;
5997 error = xfs_bmapi(NULL, ip, XFS_BB_TO_FSBT(mp, bmv->bmv_offset),
5998 XFS_BB_TO_FSB(mp, bmv->bmv_length),
3e57ecf6
OW
5999 bmapi_flags, NULL, 0, map, &nmap,
6000 NULL, NULL);
1da177e4
LT
6001 if (error)
6002 goto unlock_and_return;
6003 ASSERT(nmap <= subnex);
6004
6005 for (i = 0; i < nmap && nexleft && bmv->bmv_length; i++) {
5af317c9
ES
6006 out.bmv_oflags = 0;
6007 if (map[i].br_state == XFS_EXT_UNWRITTEN)
6008 out.bmv_oflags |= BMV_OF_PREALLOC;
6009 else if (map[i].br_startblock == DELAYSTARTBLOCK)
6010 out.bmv_oflags |= BMV_OF_DELALLOC;
1da177e4
LT
6011 out.bmv_offset = XFS_FSB_TO_BB(mp, map[i].br_startoff);
6012 out.bmv_length = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
8a7141a8 6013 out.bmv_unused1 = out.bmv_unused2 = 0;
5af317c9
ES
6014 ASSERT(((iflags & BMV_IF_DELALLOC) != 0) ||
6015 (map[i].br_startblock != DELAYSTARTBLOCK));
9af0a70c 6016 if (map[i].br_startblock == HOLESTARTBLOCK &&
3bacbcd8
VA
6017 whichfork == XFS_ATTR_FORK) {
6018 /* came to the end of attribute fork */
5af317c9 6019 out.bmv_oflags |= BMV_OF_LAST;
1da177e4
LT
6020 goto unlock_and_return;
6021 } else {
8a7141a8
ES
6022 int full = 0; /* user array is full */
6023
3bacbcd8
VA
6024 if (!xfs_getbmapx_fix_eof_hole(ip, &out,
6025 prealloced, bmvend,
6026 map[i].br_startblock)) {
6027 goto unlock_and_return;
6028 }
1da177e4 6029
8a7141a8
ES
6030 /* format results & advance arg */
6031 error = formatter(&arg, &out, &full);
6032 if (error || full)
6033 goto unlock_and_return;
6034 nexleft--;
1da177e4
LT
6035 bmv->bmv_offset =
6036 out.bmv_offset + out.bmv_length;
6037 bmv->bmv_length = MAX((__int64_t)0,
6038 (__int64_t)(bmvend - bmv->bmv_offset));
6039 bmv->bmv_entries++;
1da177e4
LT
6040 }
6041 }
6042 } while (nmap && nexleft && bmv->bmv_length);
6043
6044unlock_and_return:
6045 xfs_iunlock_map_shared(ip, lock);
6046 xfs_iunlock(ip, XFS_IOLOCK_SHARED);
6047
f0e2d93c 6048 kmem_free(map);
1da177e4
LT
6049
6050 return error;
6051}
6052
6053/*
6054 * Check the last inode extent to determine whether this allocation will result
6055 * in blocks being allocated at the end of the file. When we allocate new data
6056 * blocks at the end of the file which do not start at the previous data block,
6057 * we will try to align the new blocks at stripe unit boundaries.
6058 */
ba0f32d4 6059STATIC int /* error */
1da177e4
LT
6060xfs_bmap_isaeof(
6061 xfs_inode_t *ip, /* incore inode pointer */
6062 xfs_fileoff_t off, /* file offset in fsblocks */
6063 int whichfork, /* data or attribute fork */
6064 char *aeof) /* return value */
6065{
6066 int error; /* error return value */
6067 xfs_ifork_t *ifp; /* inode fork pointer */
a6f64d4a 6068 xfs_bmbt_rec_host_t *lastrec; /* extent record pointer */
4eea22f0
MK
6069 xfs_extnum_t nextents; /* number of file extents */
6070 xfs_bmbt_irec_t s; /* expanded extent record */
1da177e4
LT
6071
6072 ASSERT(whichfork == XFS_DATA_FORK);
6073 ifp = XFS_IFORK_PTR(ip, whichfork);
6074 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
6075 (error = xfs_iread_extents(NULL, ip, whichfork)))
6076 return error;
6077 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
6078 if (nextents == 0) {
6079 *aeof = 1;
6080 return 0;
6081 }
6082 /*
6083 * Go to the last extent
6084 */
4eea22f0 6085 lastrec = xfs_iext_get_ext(ifp, nextents - 1);
1da177e4
LT
6086 xfs_bmbt_get_all(lastrec, &s);
6087 /*
6088 * Check we are allocating in the last extent (for delayed allocations)
6089 * or past the last extent for non-delayed allocations.
6090 */
6091 *aeof = (off >= s.br_startoff &&
6092 off < s.br_startoff + s.br_blockcount &&
9d87c319 6093 isnullstartblock(s.br_startblock)) ||
1da177e4
LT
6094 off >= s.br_startoff + s.br_blockcount;
6095 return 0;
6096}
6097
6098/*
6099 * Check if the endoff is outside the last extent. If so the caller will grow
6100 * the allocation to a stripe unit boundary.
6101 */
6102int /* error */
6103xfs_bmap_eof(
6104 xfs_inode_t *ip, /* incore inode pointer */
6105 xfs_fileoff_t endoff, /* file offset in fsblocks */
6106 int whichfork, /* data or attribute fork */
6107 int *eof) /* result value */
6108{
6109 xfs_fsblock_t blockcount; /* extent block count */
6110 int error; /* error return value */
6111 xfs_ifork_t *ifp; /* inode fork pointer */
a6f64d4a 6112 xfs_bmbt_rec_host_t *lastrec; /* extent record pointer */
4eea22f0 6113 xfs_extnum_t nextents; /* number of file extents */
1da177e4
LT
6114 xfs_fileoff_t startoff; /* extent starting file offset */
6115
6116 ASSERT(whichfork == XFS_DATA_FORK);
6117 ifp = XFS_IFORK_PTR(ip, whichfork);
6118 if (!(ifp->if_flags & XFS_IFEXTENTS) &&
6119 (error = xfs_iread_extents(NULL, ip, whichfork)))
6120 return error;
6121 nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
6122 if (nextents == 0) {
6123 *eof = 1;
6124 return 0;
6125 }
6126 /*
6127 * Go to the last extent
6128 */
4eea22f0 6129 lastrec = xfs_iext_get_ext(ifp, nextents - 1);
1da177e4
LT
6130 startoff = xfs_bmbt_get_startoff(lastrec);
6131 blockcount = xfs_bmbt_get_blockcount(lastrec);
6132 *eof = endoff >= startoff + blockcount;
6133 return 0;
6134}
6135
6136#ifdef DEBUG
1da177e4
LT
6137STATIC
6138xfs_buf_t *
6139xfs_bmap_get_bp(
6140 xfs_btree_cur_t *cur,
6141 xfs_fsblock_t bno)
6142{
6143 int i;
6144 xfs_buf_t *bp;
6145
6146 if (!cur)
6147 return(NULL);
6148
6149 bp = NULL;
6150 for(i = 0; i < XFS_BTREE_MAXLEVELS; i++) {
6151 bp = cur->bc_bufs[i];
6152 if (!bp) break;
6153 if (XFS_BUF_ADDR(bp) == bno)
6154 break; /* Found it */
6155 }
6156 if (i == XFS_BTREE_MAXLEVELS)
6157 bp = NULL;
6158
6159 if (!bp) { /* Chase down all the log items to see if the bp is there */
6160 xfs_log_item_chunk_t *licp;
6161 xfs_trans_t *tp;
6162
6163 tp = cur->bc_tp;
6164 licp = &tp->t_items;
6165 while (!bp && licp != NULL) {
39dab9d7 6166 if (xfs_lic_are_all_free(licp)) {
1da177e4
LT
6167 licp = licp->lic_next;
6168 continue;
6169 }
6170 for (i = 0; i < licp->lic_unused; i++) {
6171 xfs_log_item_desc_t *lidp;
6172 xfs_log_item_t *lip;
6173 xfs_buf_log_item_t *bip;
6174 xfs_buf_t *lbp;
6175
39dab9d7 6176 if (xfs_lic_isfree(licp, i)) {
1da177e4
LT
6177 continue;
6178 }
6179
39dab9d7 6180 lidp = xfs_lic_slot(licp, i);
1da177e4
LT
6181 lip = lidp->lid_item;
6182 if (lip->li_type != XFS_LI_BUF)
6183 continue;
6184
6185 bip = (xfs_buf_log_item_t *)lip;
6186 lbp = bip->bli_buf;
6187
6188 if (XFS_BUF_ADDR(lbp) == bno) {
6189 bp = lbp;
6190 break; /* Found it */
6191 }
6192 }
6193 licp = licp->lic_next;
6194 }
6195 }
6196 return(bp);
6197}
6198
3180e66d 6199STATIC void
1da177e4 6200xfs_check_block(
7cc95a82 6201 struct xfs_btree_block *block,
1da177e4
LT
6202 xfs_mount_t *mp,
6203 int root,
6204 short sz)
6205{
6206 int i, j, dmxr;
576039cf 6207 __be64 *pp, *thispa; /* pointer to block address */
1da177e4
LT
6208 xfs_bmbt_key_t *prevp, *keyp;
6209
16259e7d 6210 ASSERT(be16_to_cpu(block->bb_level) > 0);
1da177e4
LT
6211
6212 prevp = NULL;
7cc95a82 6213 for( i = 1; i <= xfs_btree_get_numrecs(block); i++) {
1da177e4 6214 dmxr = mp->m_bmap_dmxr[0];
136341b4 6215 keyp = XFS_BMBT_KEY_ADDR(mp, block, i);
1da177e4
LT
6216
6217 if (prevp) {
4a26e66e
CH
6218 ASSERT(be64_to_cpu(prevp->br_startoff) <
6219 be64_to_cpu(keyp->br_startoff));
1da177e4
LT
6220 }
6221 prevp = keyp;
6222
6223 /*
6224 * Compare the block numbers to see if there are dups.
6225 */
136341b4 6226 if (root)
60197e8d 6227 pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, i, sz);
136341b4
CH
6228 else
6229 pp = XFS_BMBT_PTR_ADDR(mp, block, i, dmxr);
6230
16259e7d 6231 for (j = i+1; j <= be16_to_cpu(block->bb_numrecs); j++) {
136341b4 6232 if (root)
60197e8d 6233 thispa = XFS_BMAP_BROOT_PTR_ADDR(mp, block, j, sz);
136341b4
CH
6234 else
6235 thispa = XFS_BMBT_PTR_ADDR(mp, block, j, dmxr);
576039cf 6236 if (*thispa == *pp) {
1da177e4 6237 cmn_err(CE_WARN, "%s: thispa(%d) == pp(%d) %Ld",
34a622b2 6238 __func__, j, i,
576039cf 6239 (unsigned long long)be64_to_cpu(*thispa));
1da177e4 6240 panic("%s: ptrs are equal in node\n",
34a622b2 6241 __func__);
1da177e4
LT
6242 }
6243 }
6244 }
6245}
6246
6247/*
6248 * Check that the extents for the inode ip are in the right order in all
6249 * btree leaves.
6250 */
6251
6252STATIC void
6253xfs_bmap_check_leaf_extents(
6254 xfs_btree_cur_t *cur, /* btree cursor or null */
6255 xfs_inode_t *ip, /* incore inode pointer */
6256 int whichfork) /* data or attr fork */
6257{
7cc95a82 6258 struct xfs_btree_block *block; /* current btree block */
1da177e4
LT
6259 xfs_fsblock_t bno; /* block # of "block" */
6260 xfs_buf_t *bp; /* buffer for "block" */
6261 int error; /* error return value */
4eea22f0 6262 xfs_extnum_t i=0, j; /* index into the extents list */
1da177e4
LT
6263 xfs_ifork_t *ifp; /* fork structure */
6264 int level; /* btree level, for checking */
6265 xfs_mount_t *mp; /* file system mount structure */
576039cf 6266 __be64 *pp; /* pointer to block address */
4eea22f0 6267 xfs_bmbt_rec_t *ep; /* pointer to current extent */
2abdb8c8 6268 xfs_bmbt_rec_t last = {0, 0}; /* last extent in prev block */
4eea22f0 6269 xfs_bmbt_rec_t *nextp; /* pointer to next extent */
1da177e4
LT
6270 int bp_release = 0;
6271
6272 if (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE) {
6273 return;
6274 }
6275
6276 bno = NULLFSBLOCK;
6277 mp = ip->i_mount;
6278 ifp = XFS_IFORK_PTR(ip, whichfork);
6279 block = ifp->if_broot;
6280 /*
6281 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
6282 */
16259e7d
CH
6283 level = be16_to_cpu(block->bb_level);
6284 ASSERT(level > 0);
1da177e4 6285 xfs_check_block(block, mp, 1, ifp->if_broot_bytes);
60197e8d 6286 pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
576039cf
CH
6287 bno = be64_to_cpu(*pp);
6288
6289 ASSERT(bno != NULLDFSBNO);
6290 ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
6291 ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
6292
1da177e4
LT
6293 /*
6294 * Go down the tree until leaf level is reached, following the first
6295 * pointer (leftmost) at each level.
6296 */
6297 while (level-- > 0) {
6298 /* See if buf is in cur first */
6299 bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
6300 if (bp) {
6301 bp_release = 0;
6302 } else {
6303 bp_release = 1;
6304 }
6305 if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
6306 XFS_BMAP_BTREE_REF)))
6307 goto error_norelse;
7cc95a82 6308 block = XFS_BUF_TO_BLOCK(bp);
1da177e4 6309 XFS_WANT_CORRUPTED_GOTO(
4e8938fe 6310 xfs_bmap_sanity_check(mp, bp, level),
1da177e4
LT
6311 error0);
6312 if (level == 0)
6313 break;
6314
6315 /*
6316 * Check this block for basic sanity (increasing keys and
6317 * no duplicate blocks).
6318 */
6319
6320 xfs_check_block(block, mp, 0, 0);
136341b4 6321 pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
576039cf
CH
6322 bno = be64_to_cpu(*pp);
6323 XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0);
1da177e4
LT
6324 if (bp_release) {
6325 bp_release = 0;
6326 xfs_trans_brelse(NULL, bp);
6327 }
6328 }
6329
6330 /*
6331 * Here with bp and block set to the leftmost leaf node in the tree.
6332 */
6333 i = 0;
6334
6335 /*
6336 * Loop over all leaf nodes checking that all extents are in the right order.
6337 */
1da177e4 6338 for (;;) {
1da177e4
LT
6339 xfs_fsblock_t nextbno;
6340 xfs_extnum_t num_recs;
6341
6342
7cc95a82 6343 num_recs = xfs_btree_get_numrecs(block);
1da177e4
LT
6344
6345 /*
6346 * Read-ahead the next leaf block, if any.
6347 */
6348
7cc95a82 6349 nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
1da177e4
LT
6350
6351 /*
6352 * Check all the extents to make sure they are OK.
6353 * If we had a previous block, the last entry should
6354 * conform with the first entry in this one.
6355 */
6356
136341b4 6357 ep = XFS_BMBT_REC_ADDR(mp, block, 1);
2abdb8c8 6358 if (i) {
4a26e66e
CH
6359 ASSERT(xfs_bmbt_disk_get_startoff(&last) +
6360 xfs_bmbt_disk_get_blockcount(&last) <=
6361 xfs_bmbt_disk_get_startoff(ep));
2abdb8c8 6362 }
4eea22f0 6363 for (j = 1; j < num_recs; j++) {
136341b4 6364 nextp = XFS_BMBT_REC_ADDR(mp, block, j + 1);
4a26e66e
CH
6365 ASSERT(xfs_bmbt_disk_get_startoff(ep) +
6366 xfs_bmbt_disk_get_blockcount(ep) <=
6367 xfs_bmbt_disk_get_startoff(nextp));
4eea22f0 6368 ep = nextp;
1da177e4 6369 }
1da177e4 6370
2abdb8c8 6371 last = *ep;
1da177e4
LT
6372 i += num_recs;
6373 if (bp_release) {
6374 bp_release = 0;
6375 xfs_trans_brelse(NULL, bp);
6376 }
6377 bno = nextbno;
6378 /*
6379 * If we've reached the end, stop.
6380 */
6381 if (bno == NULLFSBLOCK)
6382 break;
6383
6384 bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
6385 if (bp) {
6386 bp_release = 0;
6387 } else {
6388 bp_release = 1;
6389 }
6390 if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
6391 XFS_BMAP_BTREE_REF)))
6392 goto error_norelse;
7cc95a82 6393 block = XFS_BUF_TO_BLOCK(bp);
1da177e4
LT
6394 }
6395 if (bp_release) {
6396 bp_release = 0;
6397 xfs_trans_brelse(NULL, bp);
6398 }
6399 return;
6400
6401error0:
34a622b2 6402 cmn_err(CE_WARN, "%s: at error0", __func__);
1da177e4
LT
6403 if (bp_release)
6404 xfs_trans_brelse(NULL, bp);
6405error_norelse:
6406 cmn_err(CE_WARN, "%s: BAD after btree leaves for %d extents",
34a622b2
HH
6407 __func__, i);
6408 panic("%s: CORRUPTED BTREE OR SOMETHING", __func__);
1da177e4
LT
6409 return;
6410}
6411#endif
6412
6413/*
6414 * Count fsblocks of the given fork.
6415 */
6416int /* error */
6417xfs_bmap_count_blocks(
6418 xfs_trans_t *tp, /* transaction pointer */
6419 xfs_inode_t *ip, /* incore inode */
6420 int whichfork, /* data or attr fork */
6421 int *count) /* out: count of blocks */
6422{
7cc95a82 6423 struct xfs_btree_block *block; /* current btree block */
1da177e4
LT
6424 xfs_fsblock_t bno; /* block # of "block" */
6425 xfs_ifork_t *ifp; /* fork structure */
6426 int level; /* btree level, for checking */
6427 xfs_mount_t *mp; /* file system mount structure */
576039cf 6428 __be64 *pp; /* pointer to block address */
1da177e4
LT
6429
6430 bno = NULLFSBLOCK;
6431 mp = ip->i_mount;
6432 ifp = XFS_IFORK_PTR(ip, whichfork);
6433 if ( XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS ) {
c94312de 6434 xfs_bmap_count_leaves(ifp, 0,
1da177e4 6435 ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t),
c94312de 6436 count);
1da177e4
LT
6437 return 0;
6438 }
6439
6440 /*
6441 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
6442 */
6443 block = ifp->if_broot;
16259e7d
CH
6444 level = be16_to_cpu(block->bb_level);
6445 ASSERT(level > 0);
60197e8d 6446 pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
576039cf
CH
6447 bno = be64_to_cpu(*pp);
6448 ASSERT(bno != NULLDFSBNO);
6449 ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
6450 ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
1da177e4 6451
4eea22f0 6452 if (unlikely(xfs_bmap_count_tree(mp, tp, ifp, bno, level, count) < 0)) {
1da177e4
LT
6453 XFS_ERROR_REPORT("xfs_bmap_count_blocks(2)", XFS_ERRLEVEL_LOW,
6454 mp);
6455 return XFS_ERROR(EFSCORRUPTED);
6456 }
6457
6458 return 0;
6459}
6460
6461/*
6462 * Recursively walks each level of a btree
6463 * to count total fsblocks is use.
6464 */
a8272ce0 6465STATIC int /* error */
1da177e4
LT
6466xfs_bmap_count_tree(
6467 xfs_mount_t *mp, /* file system mount point */
6468 xfs_trans_t *tp, /* transaction pointer */
4eea22f0 6469 xfs_ifork_t *ifp, /* inode fork pointer */
1da177e4
LT
6470 xfs_fsblock_t blockno, /* file system block number */
6471 int levelin, /* level in btree */
6472 int *count) /* Count of blocks */
6473{
6474 int error;
6475 xfs_buf_t *bp, *nbp;
6476 int level = levelin;
576039cf 6477 __be64 *pp;
1da177e4
LT
6478 xfs_fsblock_t bno = blockno;
6479 xfs_fsblock_t nextbno;
7cc95a82 6480 struct xfs_btree_block *block, *nextblock;
1da177e4 6481 int numrecs;
1da177e4
LT
6482
6483 if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF)))
6484 return error;
6485 *count += 1;
7cc95a82 6486 block = XFS_BUF_TO_BLOCK(bp);
1da177e4
LT
6487
6488 if (--level) {
9da096fd 6489 /* Not at node above leaves, count this level of nodes */
7cc95a82 6490 nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
1da177e4
LT
6491 while (nextbno != NULLFSBLOCK) {
6492 if ((error = xfs_btree_read_bufl(mp, tp, nextbno,
6493 0, &nbp, XFS_BMAP_BTREE_REF)))
6494 return error;
6495 *count += 1;
7cc95a82
CH
6496 nextblock = XFS_BUF_TO_BLOCK(nbp);
6497 nextbno = be64_to_cpu(nextblock->bb_u.l.bb_rightsib);
1da177e4
LT
6498 xfs_trans_brelse(tp, nbp);
6499 }
6500
6501 /* Dive to the next level */
136341b4 6502 pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
576039cf 6503 bno = be64_to_cpu(*pp);
1da177e4 6504 if (unlikely((error =
4eea22f0 6505 xfs_bmap_count_tree(mp, tp, ifp, bno, level, count)) < 0)) {
1da177e4
LT
6506 xfs_trans_brelse(tp, bp);
6507 XFS_ERROR_REPORT("xfs_bmap_count_tree(1)",
6508 XFS_ERRLEVEL_LOW, mp);
6509 return XFS_ERROR(EFSCORRUPTED);
6510 }
6511 xfs_trans_brelse(tp, bp);
6512 } else {
6513 /* count all level 1 nodes and their leaves */
6514 for (;;) {
7cc95a82 6515 nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
16259e7d 6516 numrecs = be16_to_cpu(block->bb_numrecs);
136341b4 6517 xfs_bmap_disk_count_leaves(mp, block, numrecs, count);
1da177e4
LT
6518 xfs_trans_brelse(tp, bp);
6519 if (nextbno == NULLFSBLOCK)
6520 break;
6521 bno = nextbno;
6522 if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
6523 XFS_BMAP_BTREE_REF)))
6524 return error;
6525 *count += 1;
7cc95a82 6526 block = XFS_BUF_TO_BLOCK(bp);
1da177e4
LT
6527 }
6528 }
6529 return 0;
6530}
6531
6532/*
4eea22f0 6533 * Count leaf blocks given a range of extent records.
1da177e4 6534 */
c94312de 6535STATIC void
1da177e4 6536xfs_bmap_count_leaves(
4eea22f0
MK
6537 xfs_ifork_t *ifp,
6538 xfs_extnum_t idx,
1da177e4
LT
6539 int numrecs,
6540 int *count)
6541{
6542 int b;
6543
4eea22f0 6544 for (b = 0; b < numrecs; b++) {
a6f64d4a 6545 xfs_bmbt_rec_host_t *frp = xfs_iext_get_ext(ifp, idx + b);
91e11088 6546 *count += xfs_bmbt_get_blockcount(frp);
4eea22f0 6547 }
91e11088
YL
6548}
6549
6550/*
4eea22f0
MK
6551 * Count leaf blocks given a range of extent records originally
6552 * in btree format.
91e11088 6553 */
c94312de 6554STATIC void
91e11088 6555xfs_bmap_disk_count_leaves(
136341b4 6556 struct xfs_mount *mp,
7cc95a82 6557 struct xfs_btree_block *block,
91e11088
YL
6558 int numrecs,
6559 int *count)
6560{
6561 int b;
4eea22f0 6562 xfs_bmbt_rec_t *frp;
91e11088 6563
4eea22f0 6564 for (b = 1; b <= numrecs; b++) {
136341b4 6565 frp = XFS_BMBT_REC_ADDR(mp, block, b);
1da177e4 6566 *count += xfs_bmbt_disk_get_blockcount(frp);
4eea22f0 6567 }
1da177e4 6568}