]>
Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
68988114 DC |
2 | /* |
3 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. | |
c24b5dfa | 4 | * Copyright (c) 2012 Red Hat, Inc. |
68988114 | 5 | * All Rights Reserved. |
68988114 DC |
6 | */ |
7 | #include "xfs.h" | |
8 | #include "xfs_fs.h" | |
70a9883c | 9 | #include "xfs_shared.h" |
239880ef DC |
10 | #include "xfs_format.h" |
11 | #include "xfs_log_format.h" | |
12 | #include "xfs_trans_resv.h" | |
68988114 | 13 | #include "xfs_bit.h" |
68988114 | 14 | #include "xfs_mount.h" |
3ab78df2 | 15 | #include "xfs_defer.h" |
68988114 DC |
16 | #include "xfs_inode.h" |
17 | #include "xfs_btree.h" | |
239880ef | 18 | #include "xfs_trans.h" |
68988114 DC |
19 | #include "xfs_alloc.h" |
20 | #include "xfs_bmap.h" | |
21 | #include "xfs_bmap_util.h" | |
a4fbe6ab | 22 | #include "xfs_bmap_btree.h" |
68988114 DC |
23 | #include "xfs_rtalloc.h" |
24 | #include "xfs_error.h" | |
25 | #include "xfs_quota.h" | |
26 | #include "xfs_trans_space.h" | |
27 | #include "xfs_trace.h" | |
c24b5dfa | 28 | #include "xfs_icache.h" |
f86f4037 DW |
29 | #include "xfs_iomap.h" |
30 | #include "xfs_reflink.h" | |
68988114 DC |
31 | |
32 | /* Kernel only BMAP related definitions and functions */ | |
33 | ||
34 | /* | |
35 | * Convert the given file system block to a disk block. We have to treat it | |
36 | * differently based on whether the file is a real time file or not, because the | |
37 | * bmap code does. | |
38 | */ | |
39 | xfs_daddr_t | |
40 | xfs_fsb_to_db(struct xfs_inode *ip, xfs_fsblock_t fsb) | |
41 | { | |
ecfc28a4 CH |
42 | if (XFS_IS_REALTIME_INODE(ip)) |
43 | return XFS_FSB_TO_BB(ip->i_mount, fsb); | |
44 | return XFS_FSB_TO_DADDR(ip->i_mount, fsb); | |
68988114 DC |
45 | } |
46 | ||
3fbbbea3 DC |
47 | /* |
48 | * Routine to zero an extent on disk allocated to the specific inode. | |
49 | * | |
50 | * The VFS functions take a linearised filesystem block offset, so we have to | |
51 | * convert the sparse xfs fsb to the right format first. | |
52 | * VFS types are real funky, too. | |
53 | */ | |
54 | int | |
55 | xfs_zero_extent( | |
56 | struct xfs_inode *ip, | |
57 | xfs_fsblock_t start_fsb, | |
58 | xfs_off_t count_fsb) | |
59 | { | |
60 | struct xfs_mount *mp = ip->i_mount; | |
61 | xfs_daddr_t sector = xfs_fsb_to_db(ip, start_fsb); | |
62 | sector_t block = XFS_BB_TO_FSBT(mp, sector); | |
3fbbbea3 | 63 | |
3dc29161 MW |
64 | return blkdev_issue_zeroout(xfs_find_bdev_for_inode(VFS_I(ip)), |
65 | block << (mp->m_super->s_blocksize_bits - 9), | |
66 | count_fsb << (mp->m_super->s_blocksize_bits - 9), | |
ee472d83 | 67 | GFP_NOFS, 0); |
3fbbbea3 DC |
68 | } |
69 | ||
bb9c2e54 | 70 | #ifdef CONFIG_XFS_RT |
68988114 DC |
71 | int |
72 | xfs_bmap_rtalloc( | |
73 | struct xfs_bmalloca *ap) /* bmap alloc argument struct */ | |
74 | { | |
68988114 DC |
75 | int error; /* error return value */ |
76 | xfs_mount_t *mp; /* mount point structure */ | |
77 | xfs_extlen_t prod = 0; /* product factor for allocators */ | |
0703a8e1 | 78 | xfs_extlen_t mod = 0; /* product factor for allocators */ |
68988114 DC |
79 | xfs_extlen_t ralen = 0; /* realtime allocation length */ |
80 | xfs_extlen_t align; /* minimum allocation alignment */ | |
81 | xfs_rtblock_t rtb; | |
82 | ||
83 | mp = ap->ip->i_mount; | |
84 | align = xfs_get_extsz_hint(ap->ip); | |
85 | prod = align / mp->m_sb.sb_rextsize; | |
86 | error = xfs_bmap_extsize_align(mp, &ap->got, &ap->prev, | |
87 | align, 1, ap->eof, 0, | |
88 | ap->conv, &ap->offset, &ap->length); | |
89 | if (error) | |
90 | return error; | |
91 | ASSERT(ap->length); | |
92 | ASSERT(ap->length % mp->m_sb.sb_rextsize == 0); | |
93 | ||
94 | /* | |
95 | * If the offset & length are not perfectly aligned | |
96 | * then kill prod, it will just get us in trouble. | |
97 | */ | |
0703a8e1 DC |
98 | div_u64_rem(ap->offset, align, &mod); |
99 | if (mod || ap->length % align) | |
68988114 DC |
100 | prod = 1; |
101 | /* | |
102 | * Set ralen to be the actual requested length in rtextents. | |
103 | */ | |
104 | ralen = ap->length / mp->m_sb.sb_rextsize; | |
105 | /* | |
106 | * If the old value was close enough to MAXEXTLEN that | |
107 | * we rounded up to it, cut it back so it's valid again. | |
108 | * Note that if it's a really large request (bigger than | |
109 | * MAXEXTLEN), we don't hear about that number, and can't | |
110 | * adjust the starting point to match it. | |
111 | */ | |
112 | if (ralen * mp->m_sb.sb_rextsize >= MAXEXTLEN) | |
113 | ralen = MAXEXTLEN / mp->m_sb.sb_rextsize; | |
114 | ||
115 | /* | |
4b680afb | 116 | * Lock out modifications to both the RT bitmap and summary inodes |
68988114 | 117 | */ |
f4a0660d | 118 | xfs_ilock(mp->m_rbmip, XFS_ILOCK_EXCL|XFS_ILOCK_RTBITMAP); |
68988114 | 119 | xfs_trans_ijoin(ap->tp, mp->m_rbmip, XFS_ILOCK_EXCL); |
f4a0660d | 120 | xfs_ilock(mp->m_rsumip, XFS_ILOCK_EXCL|XFS_ILOCK_RTSUM); |
4b680afb | 121 | xfs_trans_ijoin(ap->tp, mp->m_rsumip, XFS_ILOCK_EXCL); |
68988114 DC |
122 | |
123 | /* | |
124 | * If it's an allocation to an empty file at offset 0, | |
125 | * pick an extent that will space things out in the rt area. | |
126 | */ | |
127 | if (ap->eof && ap->offset == 0) { | |
128 | xfs_rtblock_t uninitialized_var(rtx); /* realtime extent no */ | |
129 | ||
130 | error = xfs_rtpick_extent(mp, ap->tp, ralen, &rtx); | |
131 | if (error) | |
132 | return error; | |
133 | ap->blkno = rtx * mp->m_sb.sb_rextsize; | |
134 | } else { | |
135 | ap->blkno = 0; | |
136 | } | |
137 | ||
138 | xfs_bmap_adjacent(ap); | |
139 | ||
140 | /* | |
141 | * Realtime allocation, done through xfs_rtallocate_extent. | |
142 | */ | |
68988114 DC |
143 | do_div(ap->blkno, mp->m_sb.sb_rextsize); |
144 | rtb = ap->blkno; | |
145 | ap->length = ralen; | |
089ec2f8 CH |
146 | error = xfs_rtallocate_extent(ap->tp, ap->blkno, 1, ap->length, |
147 | &ralen, ap->wasdel, prod, &rtb); | |
148 | if (error) | |
68988114 | 149 | return error; |
089ec2f8 | 150 | |
68988114 DC |
151 | ap->blkno = rtb; |
152 | if (ap->blkno != NULLFSBLOCK) { | |
153 | ap->blkno *= mp->m_sb.sb_rextsize; | |
154 | ralen *= mp->m_sb.sb_rextsize; | |
155 | ap->length = ralen; | |
156 | ap->ip->i_d.di_nblocks += ralen; | |
157 | xfs_trans_log_inode(ap->tp, ap->ip, XFS_ILOG_CORE); | |
158 | if (ap->wasdel) | |
159 | ap->ip->i_delayed_blks -= ralen; | |
160 | /* | |
161 | * Adjust the disk quota also. This was reserved | |
162 | * earlier. | |
163 | */ | |
164 | xfs_trans_mod_dquot_byino(ap->tp, ap->ip, | |
165 | ap->wasdel ? XFS_TRANS_DQ_DELRTBCOUNT : | |
166 | XFS_TRANS_DQ_RTBCOUNT, (long) ralen); | |
3fbbbea3 DC |
167 | |
168 | /* Zero the extent if we were asked to do so */ | |
292378ed | 169 | if (ap->datatype & XFS_ALLOC_USERDATA_ZERO) { |
3fbbbea3 DC |
170 | error = xfs_zero_extent(ap->ip, ap->blkno, ap->length); |
171 | if (error) | |
172 | return error; | |
173 | } | |
68988114 DC |
174 | } else { |
175 | ap->length = 0; | |
176 | } | |
177 | return 0; | |
178 | } | |
bb9c2e54 | 179 | #endif /* CONFIG_XFS_RT */ |
68988114 | 180 | |
68988114 DC |
181 | /* |
182 | * Check if the endoff is outside the last extent. If so the caller will grow | |
183 | * the allocation to a stripe unit boundary. All offsets are considered outside | |
184 | * the end of file for an empty fork, so 1 is returned in *eof in that case. | |
185 | */ | |
186 | int | |
187 | xfs_bmap_eof( | |
188 | struct xfs_inode *ip, | |
189 | xfs_fileoff_t endoff, | |
190 | int whichfork, | |
191 | int *eof) | |
192 | { | |
193 | struct xfs_bmbt_irec rec; | |
194 | int error; | |
195 | ||
196 | error = xfs_bmap_last_extent(NULL, ip, whichfork, &rec, eof); | |
197 | if (error || *eof) | |
198 | return error; | |
199 | ||
200 | *eof = endoff >= rec.br_startoff + rec.br_blockcount; | |
201 | return 0; | |
202 | } | |
203 | ||
204 | /* | |
205 | * Extent tree block counting routines. | |
206 | */ | |
207 | ||
208 | /* | |
d29cb3e4 DW |
209 | * Count leaf blocks given a range of extent records. Delayed allocation |
210 | * extents are not counted towards the totals. | |
68988114 | 211 | */ |
e17a5c6f | 212 | xfs_extnum_t |
68988114 | 213 | xfs_bmap_count_leaves( |
d29cb3e4 | 214 | struct xfs_ifork *ifp, |
e7f5d5ca | 215 | xfs_filblks_t *count) |
68988114 | 216 | { |
b2b1712a | 217 | struct xfs_iext_cursor icur; |
e17a5c6f | 218 | struct xfs_bmbt_irec got; |
b2b1712a | 219 | xfs_extnum_t numrecs = 0; |
68988114 | 220 | |
b2b1712a | 221 | for_each_xfs_iext(ifp, &icur, &got) { |
e17a5c6f CH |
222 | if (!isnullstartblock(got.br_startblock)) { |
223 | *count += got.br_blockcount; | |
224 | numrecs++; | |
d29cb3e4 | 225 | } |
68988114 | 226 | } |
b2b1712a | 227 | |
e17a5c6f | 228 | return numrecs; |
68988114 DC |
229 | } |
230 | ||
231 | /* | |
232 | * Count leaf blocks given a range of extent records originally | |
233 | * in btree format. | |
234 | */ | |
235 | STATIC void | |
236 | xfs_bmap_disk_count_leaves( | |
237 | struct xfs_mount *mp, | |
238 | struct xfs_btree_block *block, | |
239 | int numrecs, | |
e7f5d5ca | 240 | xfs_filblks_t *count) |
68988114 DC |
241 | { |
242 | int b; | |
243 | xfs_bmbt_rec_t *frp; | |
244 | ||
245 | for (b = 1; b <= numrecs; b++) { | |
246 | frp = XFS_BMBT_REC_ADDR(mp, block, b); | |
247 | *count += xfs_bmbt_disk_get_blockcount(frp); | |
248 | } | |
249 | } | |
250 | ||
251 | /* | |
252 | * Recursively walks each level of a btree | |
8be11e92 | 253 | * to count total fsblocks in use. |
68988114 | 254 | */ |
e7f5d5ca | 255 | STATIC int |
68988114 | 256 | xfs_bmap_count_tree( |
e7f5d5ca DW |
257 | struct xfs_mount *mp, |
258 | struct xfs_trans *tp, | |
259 | struct xfs_ifork *ifp, | |
260 | xfs_fsblock_t blockno, | |
261 | int levelin, | |
262 | xfs_extnum_t *nextents, | |
263 | xfs_filblks_t *count) | |
68988114 DC |
264 | { |
265 | int error; | |
e7f5d5ca | 266 | struct xfs_buf *bp, *nbp; |
68988114 DC |
267 | int level = levelin; |
268 | __be64 *pp; | |
269 | xfs_fsblock_t bno = blockno; | |
270 | xfs_fsblock_t nextbno; | |
271 | struct xfs_btree_block *block, *nextblock; | |
272 | int numrecs; | |
273 | ||
f5b999c0 | 274 | error = xfs_btree_read_bufl(mp, tp, bno, &bp, XFS_BMAP_BTREE_REF, |
68988114 DC |
275 | &xfs_bmbt_buf_ops); |
276 | if (error) | |
277 | return error; | |
278 | *count += 1; | |
279 | block = XFS_BUF_TO_BLOCK(bp); | |
280 | ||
281 | if (--level) { | |
282 | /* Not at node above leaves, count this level of nodes */ | |
283 | nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib); | |
284 | while (nextbno != NULLFSBLOCK) { | |
f5b999c0 | 285 | error = xfs_btree_read_bufl(mp, tp, nextbno, &nbp, |
68988114 DC |
286 | XFS_BMAP_BTREE_REF, |
287 | &xfs_bmbt_buf_ops); | |
288 | if (error) | |
289 | return error; | |
290 | *count += 1; | |
291 | nextblock = XFS_BUF_TO_BLOCK(nbp); | |
292 | nextbno = be64_to_cpu(nextblock->bb_u.l.bb_rightsib); | |
293 | xfs_trans_brelse(tp, nbp); | |
294 | } | |
295 | ||
296 | /* Dive to the next level */ | |
297 | pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); | |
298 | bno = be64_to_cpu(*pp); | |
e7f5d5ca DW |
299 | error = xfs_bmap_count_tree(mp, tp, ifp, bno, level, nextents, |
300 | count); | |
301 | if (error) { | |
68988114 DC |
302 | xfs_trans_brelse(tp, bp); |
303 | XFS_ERROR_REPORT("xfs_bmap_count_tree(1)", | |
304 | XFS_ERRLEVEL_LOW, mp); | |
2451337d | 305 | return -EFSCORRUPTED; |
68988114 DC |
306 | } |
307 | xfs_trans_brelse(tp, bp); | |
308 | } else { | |
309 | /* count all level 1 nodes and their leaves */ | |
310 | for (;;) { | |
311 | nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib); | |
312 | numrecs = be16_to_cpu(block->bb_numrecs); | |
e7f5d5ca | 313 | (*nextents) += numrecs; |
68988114 DC |
314 | xfs_bmap_disk_count_leaves(mp, block, numrecs, count); |
315 | xfs_trans_brelse(tp, bp); | |
316 | if (nextbno == NULLFSBLOCK) | |
317 | break; | |
318 | bno = nextbno; | |
f5b999c0 | 319 | error = xfs_btree_read_bufl(mp, tp, bno, &bp, |
68988114 DC |
320 | XFS_BMAP_BTREE_REF, |
321 | &xfs_bmbt_buf_ops); | |
322 | if (error) | |
323 | return error; | |
324 | *count += 1; | |
325 | block = XFS_BUF_TO_BLOCK(bp); | |
326 | } | |
327 | } | |
328 | return 0; | |
329 | } | |
330 | ||
331 | /* | |
d29cb3e4 DW |
332 | * Count fsblocks of the given fork. Delayed allocation extents are |
333 | * not counted towards the totals. | |
68988114 | 334 | */ |
e7f5d5ca | 335 | int |
68988114 | 336 | xfs_bmap_count_blocks( |
e7f5d5ca DW |
337 | struct xfs_trans *tp, |
338 | struct xfs_inode *ip, | |
339 | int whichfork, | |
340 | xfs_extnum_t *nextents, | |
341 | xfs_filblks_t *count) | |
68988114 | 342 | { |
e7f5d5ca DW |
343 | struct xfs_mount *mp; /* file system mount structure */ |
344 | __be64 *pp; /* pointer to block address */ | |
68988114 | 345 | struct xfs_btree_block *block; /* current btree block */ |
e7f5d5ca | 346 | struct xfs_ifork *ifp; /* fork structure */ |
68988114 | 347 | xfs_fsblock_t bno; /* block # of "block" */ |
68988114 | 348 | int level; /* btree level, for checking */ |
e7f5d5ca | 349 | int error; |
68988114 DC |
350 | |
351 | bno = NULLFSBLOCK; | |
352 | mp = ip->i_mount; | |
e7f5d5ca DW |
353 | *nextents = 0; |
354 | *count = 0; | |
68988114 | 355 | ifp = XFS_IFORK_PTR(ip, whichfork); |
e7f5d5ca | 356 | if (!ifp) |
68988114 | 357 | return 0; |
68988114 | 358 | |
e7f5d5ca DW |
359 | switch (XFS_IFORK_FORMAT(ip, whichfork)) { |
360 | case XFS_DINODE_FMT_EXTENTS: | |
e17a5c6f | 361 | *nextents = xfs_bmap_count_leaves(ifp, count); |
e7f5d5ca DW |
362 | return 0; |
363 | case XFS_DINODE_FMT_BTREE: | |
364 | if (!(ifp->if_flags & XFS_IFEXTENTS)) { | |
365 | error = xfs_iread_extents(tp, ip, whichfork); | |
366 | if (error) | |
367 | return error; | |
368 | } | |
369 | ||
370 | /* | |
371 | * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out. | |
372 | */ | |
373 | block = ifp->if_broot; | |
374 | level = be16_to_cpu(block->bb_level); | |
375 | ASSERT(level > 0); | |
376 | pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes); | |
377 | bno = be64_to_cpu(*pp); | |
378 | ASSERT(bno != NULLFSBLOCK); | |
379 | ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount); | |
380 | ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks); | |
381 | ||
382 | error = xfs_bmap_count_tree(mp, tp, ifp, bno, level, | |
383 | nextents, count); | |
384 | if (error) { | |
385 | XFS_ERROR_REPORT("xfs_bmap_count_blocks(2)", | |
386 | XFS_ERRLEVEL_LOW, mp); | |
387 | return -EFSCORRUPTED; | |
388 | } | |
389 | return 0; | |
68988114 DC |
390 | } |
391 | ||
392 | return 0; | |
393 | } | |
394 | ||
abbf9e8a CH |
395 | static int |
396 | xfs_getbmap_report_one( | |
397 | struct xfs_inode *ip, | |
398 | struct getbmapx *bmv, | |
232b5194 | 399 | struct kgetbmap *out, |
abbf9e8a CH |
400 | int64_t bmv_end, |
401 | struct xfs_bmbt_irec *got) | |
f86f4037 | 402 | { |
232b5194 | 403 | struct kgetbmap *p = out + bmv->bmv_entries; |
d392bc81 | 404 | bool shared = false; |
abbf9e8a | 405 | int error; |
f86f4037 | 406 | |
d392bc81 | 407 | error = xfs_reflink_trim_around_shared(ip, got, &shared); |
f86f4037 DW |
408 | if (error) |
409 | return error; | |
410 | ||
abbf9e8a CH |
411 | if (isnullstartblock(got->br_startblock) || |
412 | got->br_startblock == DELAYSTARTBLOCK) { | |
f86f4037 | 413 | /* |
abbf9e8a CH |
414 | * Delalloc extents that start beyond EOF can occur due to |
415 | * speculative EOF allocation when the delalloc extent is larger | |
416 | * than the largest freespace extent at conversion time. These | |
417 | * extents cannot be converted by data writeback, so can exist | |
418 | * here even if we are not supposed to be finding delalloc | |
419 | * extents. | |
f86f4037 | 420 | */ |
abbf9e8a CH |
421 | if (got->br_startoff < XFS_B_TO_FSB(ip->i_mount, XFS_ISIZE(ip))) |
422 | ASSERT((bmv->bmv_iflags & BMV_IF_DELALLOC) != 0); | |
423 | ||
424 | p->bmv_oflags |= BMV_OF_DELALLOC; | |
425 | p->bmv_block = -2; | |
f86f4037 | 426 | } else { |
abbf9e8a | 427 | p->bmv_block = xfs_fsb_to_db(ip, got->br_startblock); |
f86f4037 DW |
428 | } |
429 | ||
abbf9e8a CH |
430 | if (got->br_state == XFS_EXT_UNWRITTEN && |
431 | (bmv->bmv_iflags & BMV_IF_PREALLOC)) | |
432 | p->bmv_oflags |= BMV_OF_PREALLOC; | |
433 | ||
434 | if (shared) | |
435 | p->bmv_oflags |= BMV_OF_SHARED; | |
436 | ||
437 | p->bmv_offset = XFS_FSB_TO_BB(ip->i_mount, got->br_startoff); | |
438 | p->bmv_length = XFS_FSB_TO_BB(ip->i_mount, got->br_blockcount); | |
439 | ||
440 | bmv->bmv_offset = p->bmv_offset + p->bmv_length; | |
441 | bmv->bmv_length = max(0LL, bmv_end - bmv->bmv_offset); | |
442 | bmv->bmv_entries++; | |
f86f4037 DW |
443 | return 0; |
444 | } | |
445 | ||
abbf9e8a CH |
446 | static void |
447 | xfs_getbmap_report_hole( | |
448 | struct xfs_inode *ip, | |
449 | struct getbmapx *bmv, | |
232b5194 | 450 | struct kgetbmap *out, |
abbf9e8a CH |
451 | int64_t bmv_end, |
452 | xfs_fileoff_t bno, | |
453 | xfs_fileoff_t end) | |
454 | { | |
232b5194 | 455 | struct kgetbmap *p = out + bmv->bmv_entries; |
abbf9e8a CH |
456 | |
457 | if (bmv->bmv_iflags & BMV_IF_NO_HOLES) | |
458 | return; | |
459 | ||
460 | p->bmv_block = -1; | |
461 | p->bmv_offset = XFS_FSB_TO_BB(ip->i_mount, bno); | |
462 | p->bmv_length = XFS_FSB_TO_BB(ip->i_mount, end - bno); | |
463 | ||
464 | bmv->bmv_offset = p->bmv_offset + p->bmv_length; | |
465 | bmv->bmv_length = max(0LL, bmv_end - bmv->bmv_offset); | |
466 | bmv->bmv_entries++; | |
467 | } | |
468 | ||
469 | static inline bool | |
470 | xfs_getbmap_full( | |
471 | struct getbmapx *bmv) | |
472 | { | |
473 | return bmv->bmv_length == 0 || bmv->bmv_entries >= bmv->bmv_count - 1; | |
474 | } | |
475 | ||
476 | static bool | |
477 | xfs_getbmap_next_rec( | |
478 | struct xfs_bmbt_irec *rec, | |
479 | xfs_fileoff_t total_end) | |
480 | { | |
481 | xfs_fileoff_t end = rec->br_startoff + rec->br_blockcount; | |
482 | ||
483 | if (end == total_end) | |
484 | return false; | |
485 | ||
486 | rec->br_startoff += rec->br_blockcount; | |
487 | if (!isnullstartblock(rec->br_startblock) && | |
488 | rec->br_startblock != DELAYSTARTBLOCK) | |
489 | rec->br_startblock += rec->br_blockcount; | |
490 | rec->br_blockcount = total_end - end; | |
491 | return true; | |
492 | } | |
493 | ||
68988114 DC |
494 | /* |
495 | * Get inode's extents as described in bmv, and format for output. | |
496 | * Calls formatter to fill the user's buffer until all extents | |
497 | * are mapped, until the passed-in bmv->bmv_count slots have | |
498 | * been filled, or until the formatter short-circuits the loop, | |
499 | * if it is tracking filled-in extents on its own. | |
500 | */ | |
501 | int /* error code */ | |
502 | xfs_getbmap( | |
232b5194 | 503 | struct xfs_inode *ip, |
68988114 | 504 | struct getbmapx *bmv, /* user bmap structure */ |
232b5194 | 505 | struct kgetbmap *out) |
68988114 | 506 | { |
abbf9e8a CH |
507 | struct xfs_mount *mp = ip->i_mount; |
508 | int iflags = bmv->bmv_iflags; | |
232b5194 | 509 | int whichfork, lock, error = 0; |
abbf9e8a CH |
510 | int64_t bmv_end, max_len; |
511 | xfs_fileoff_t bno, first_bno; | |
512 | struct xfs_ifork *ifp; | |
abbf9e8a CH |
513 | struct xfs_bmbt_irec got, rec; |
514 | xfs_filblks_t len; | |
b2b1712a | 515 | struct xfs_iext_cursor icur; |
68988114 | 516 | |
232b5194 CH |
517 | if (bmv->bmv_iflags & ~BMV_IF_VALID) |
518 | return -EINVAL; | |
f86f4037 DW |
519 | #ifndef DEBUG |
520 | /* Only allow CoW fork queries if we're debugging. */ | |
521 | if (iflags & BMV_IF_COWFORK) | |
522 | return -EINVAL; | |
523 | #endif | |
524 | if ((iflags & BMV_IF_ATTRFORK) && (iflags & BMV_IF_COWFORK)) | |
525 | return -EINVAL; | |
526 | ||
abbf9e8a CH |
527 | if (bmv->bmv_length < -1) |
528 | return -EINVAL; | |
abbf9e8a CH |
529 | bmv->bmv_entries = 0; |
530 | if (bmv->bmv_length == 0) | |
531 | return 0; | |
532 | ||
f86f4037 DW |
533 | if (iflags & BMV_IF_ATTRFORK) |
534 | whichfork = XFS_ATTR_FORK; | |
535 | else if (iflags & BMV_IF_COWFORK) | |
536 | whichfork = XFS_COW_FORK; | |
537 | else | |
538 | whichfork = XFS_DATA_FORK; | |
abbf9e8a | 539 | ifp = XFS_IFORK_PTR(ip, whichfork); |
f86f4037 | 540 | |
abbf9e8a | 541 | xfs_ilock(ip, XFS_IOLOCK_SHARED); |
f86f4037 DW |
542 | switch (whichfork) { |
543 | case XFS_ATTR_FORK: | |
abbf9e8a CH |
544 | if (!XFS_IFORK_Q(ip)) |
545 | goto out_unlock_iolock; | |
68988114 | 546 | |
abbf9e8a CH |
547 | max_len = 1LL << 32; |
548 | lock = xfs_ilock_attr_map_shared(ip); | |
f86f4037 DW |
549 | break; |
550 | case XFS_COW_FORK: | |
abbf9e8a CH |
551 | /* No CoW fork? Just return */ |
552 | if (!ifp) | |
553 | goto out_unlock_iolock; | |
68988114 | 554 | |
abbf9e8a CH |
555 | if (xfs_get_cowextsz_hint(ip)) |
556 | max_len = mp->m_super->s_maxbytes; | |
557 | else | |
558 | max_len = XFS_ISIZE(ip); | |
68988114 | 559 | |
abbf9e8a CH |
560 | lock = XFS_ILOCK_SHARED; |
561 | xfs_ilock(ip, lock); | |
562 | break; | |
f86f4037 | 563 | case XFS_DATA_FORK: |
efa70be1 CH |
564 | if (!(iflags & BMV_IF_DELALLOC) && |
565 | (ip->i_delayed_blks || XFS_ISIZE(ip) > ip->i_d.di_size)) { | |
2451337d | 566 | error = filemap_write_and_wait(VFS_I(ip)->i_mapping); |
68988114 DC |
567 | if (error) |
568 | goto out_unlock_iolock; | |
efa70be1 CH |
569 | |
570 | /* | |
571 | * Even after flushing the inode, there can still be | |
572 | * delalloc blocks on the inode beyond EOF due to | |
573 | * speculative preallocation. These are not removed | |
574 | * until the release function is called or the inode | |
575 | * is inactivated. Hence we cannot assert here that | |
576 | * ip->i_delayed_blks == 0. | |
577 | */ | |
68988114 | 578 | } |
68988114 | 579 | |
abbf9e8a CH |
580 | if (xfs_get_extsz_hint(ip) || |
581 | (ip->i_d.di_flags & | |
582 | (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))) | |
583 | max_len = mp->m_super->s_maxbytes; | |
584 | else | |
585 | max_len = XFS_ISIZE(ip); | |
586 | ||
efa70be1 | 587 | lock = xfs_ilock_data_map_shared(ip); |
f86f4037 | 588 | break; |
efa70be1 | 589 | } |
68988114 | 590 | |
abbf9e8a CH |
591 | switch (XFS_IFORK_FORMAT(ip, whichfork)) { |
592 | case XFS_DINODE_FMT_EXTENTS: | |
593 | case XFS_DINODE_FMT_BTREE: | |
594 | break; | |
595 | case XFS_DINODE_FMT_LOCAL: | |
596 | /* Local format inode forks report no extents. */ | |
68988114 | 597 | goto out_unlock_ilock; |
abbf9e8a CH |
598 | default: |
599 | error = -EINVAL; | |
600 | goto out_unlock_ilock; | |
601 | } | |
68988114 | 602 | |
abbf9e8a CH |
603 | if (bmv->bmv_length == -1) { |
604 | max_len = XFS_FSB_TO_BB(mp, XFS_B_TO_FSB(mp, max_len)); | |
605 | bmv->bmv_length = max(0LL, max_len - bmv->bmv_offset); | |
68988114 DC |
606 | } |
607 | ||
abbf9e8a | 608 | bmv_end = bmv->bmv_offset + bmv->bmv_length; |
68988114 | 609 | |
abbf9e8a CH |
610 | first_bno = bno = XFS_BB_TO_FSBT(mp, bmv->bmv_offset); |
611 | len = XFS_BB_TO_FSB(mp, bmv->bmv_length); | |
68988114 | 612 | |
abbf9e8a CH |
613 | if (!(ifp->if_flags & XFS_IFEXTENTS)) { |
614 | error = xfs_iread_extents(NULL, ip, whichfork); | |
615 | if (error) | |
616 | goto out_unlock_ilock; | |
617 | } | |
f86f4037 | 618 | |
b2b1712a | 619 | if (!xfs_iext_lookup_extent(ip, ifp, bno, &icur, &got)) { |
abbf9e8a CH |
620 | /* |
621 | * Report a whole-file hole if the delalloc flag is set to | |
622 | * stay compatible with the old implementation. | |
623 | */ | |
624 | if (iflags & BMV_IF_DELALLOC) | |
625 | xfs_getbmap_report_hole(ip, bmv, out, bmv_end, bno, | |
626 | XFS_B_TO_FSB(mp, XFS_ISIZE(ip))); | |
627 | goto out_unlock_ilock; | |
628 | } | |
68988114 | 629 | |
abbf9e8a CH |
630 | while (!xfs_getbmap_full(bmv)) { |
631 | xfs_trim_extent(&got, first_bno, len); | |
68988114 | 632 | |
abbf9e8a CH |
633 | /* |
634 | * Report an entry for a hole if this extent doesn't directly | |
635 | * follow the previous one. | |
636 | */ | |
637 | if (got.br_startoff > bno) { | |
638 | xfs_getbmap_report_hole(ip, bmv, out, bmv_end, bno, | |
639 | got.br_startoff); | |
640 | if (xfs_getbmap_full(bmv)) | |
641 | break; | |
642 | } | |
68988114 | 643 | |
abbf9e8a CH |
644 | /* |
645 | * In order to report shared extents accurately, we report each | |
646 | * distinct shared / unshared part of a single bmbt record with | |
647 | * an individual getbmapx record. | |
648 | */ | |
649 | bno = got.br_startoff + got.br_blockcount; | |
650 | rec = got; | |
651 | do { | |
652 | error = xfs_getbmap_report_one(ip, bmv, out, bmv_end, | |
653 | &rec); | |
654 | if (error || xfs_getbmap_full(bmv)) | |
655 | goto out_unlock_ilock; | |
656 | } while (xfs_getbmap_next_rec(&rec, bno)); | |
657 | ||
b2b1712a | 658 | if (!xfs_iext_next_extent(ifp, &icur, &got)) { |
abbf9e8a CH |
659 | xfs_fileoff_t end = XFS_B_TO_FSB(mp, XFS_ISIZE(ip)); |
660 | ||
661 | out[bmv->bmv_entries - 1].bmv_oflags |= BMV_OF_LAST; | |
662 | ||
663 | if (whichfork != XFS_ATTR_FORK && bno < end && | |
664 | !xfs_getbmap_full(bmv)) { | |
665 | xfs_getbmap_report_hole(ip, bmv, out, bmv_end, | |
666 | bno, end); | |
c364b6d0 | 667 | } |
abbf9e8a | 668 | break; |
68988114 | 669 | } |
68988114 | 670 | |
abbf9e8a CH |
671 | if (bno >= first_bno + len) |
672 | break; | |
673 | } | |
674 | ||
675 | out_unlock_ilock: | |
01f4f327 | 676 | xfs_iunlock(ip, lock); |
abbf9e8a | 677 | out_unlock_iolock: |
68988114 | 678 | xfs_iunlock(ip, XFS_IOLOCK_SHARED); |
68988114 DC |
679 | return error; |
680 | } | |
681 | ||
682 | /* | |
e2ac8363 CH |
683 | * Dead simple method of punching delalyed allocation blocks from a range in |
684 | * the inode. This will always punch out both the start and end blocks, even | |
685 | * if the ranges only partially overlap them, so it is up to the caller to | |
686 | * ensure that partial blocks are not passed in. | |
68988114 DC |
687 | */ |
688 | int | |
689 | xfs_bmap_punch_delalloc_range( | |
690 | struct xfs_inode *ip, | |
691 | xfs_fileoff_t start_fsb, | |
692 | xfs_fileoff_t length) | |
693 | { | |
e2ac8363 CH |
694 | struct xfs_ifork *ifp = &ip->i_df; |
695 | xfs_fileoff_t end_fsb = start_fsb + length; | |
696 | struct xfs_bmbt_irec got, del; | |
697 | struct xfs_iext_cursor icur; | |
68988114 DC |
698 | int error = 0; |
699 | ||
0065b541 | 700 | ASSERT(ifp->if_flags & XFS_IFEXTENTS); |
68988114 | 701 | |
0065b541 | 702 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
e2ac8363 | 703 | if (!xfs_iext_lookup_extent_before(ip, ifp, &end_fsb, &icur, &got)) |
d4380177 | 704 | goto out_unlock; |
68988114 | 705 | |
e2ac8363 CH |
706 | while (got.br_startoff + got.br_blockcount > start_fsb) { |
707 | del = got; | |
708 | xfs_trim_extent(&del, start_fsb, length); | |
68988114 DC |
709 | |
710 | /* | |
e2ac8363 CH |
711 | * A delete can push the cursor forward. Step back to the |
712 | * previous extent on non-delalloc or extents outside the | |
713 | * target range. | |
68988114 | 714 | */ |
e2ac8363 CH |
715 | if (!del.br_blockcount || |
716 | !isnullstartblock(del.br_startblock)) { | |
717 | if (!xfs_iext_prev_extent(ifp, &icur, &got)) | |
718 | break; | |
719 | continue; | |
720 | } | |
68988114 | 721 | |
e2ac8363 CH |
722 | error = xfs_bmap_del_extent_delay(ip, XFS_DATA_FORK, &icur, |
723 | &got, &del); | |
724 | if (error || !xfs_iext_get_extent(ifp, &icur, &got)) | |
725 | break; | |
726 | } | |
68988114 | 727 | |
d4380177 CH |
728 | out_unlock: |
729 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
68988114 DC |
730 | return error; |
731 | } | |
c24b5dfa DC |
732 | |
733 | /* | |
734 | * Test whether it is appropriate to check an inode for and free post EOF | |
735 | * blocks. The 'force' parameter determines whether we should also consider | |
736 | * regular files that are marked preallocated or append-only. | |
737 | */ | |
738 | bool | |
739 | xfs_can_free_eofblocks(struct xfs_inode *ip, bool force) | |
740 | { | |
741 | /* prealloc/delalloc exists only on regular files */ | |
c19b3b05 | 742 | if (!S_ISREG(VFS_I(ip)->i_mode)) |
c24b5dfa DC |
743 | return false; |
744 | ||
745 | /* | |
746 | * Zero sized files with no cached pages and delalloc blocks will not | |
747 | * have speculative prealloc/delalloc blocks to remove. | |
748 | */ | |
749 | if (VFS_I(ip)->i_size == 0 && | |
2667c6f9 | 750 | VFS_I(ip)->i_mapping->nrpages == 0 && |
c24b5dfa DC |
751 | ip->i_delayed_blks == 0) |
752 | return false; | |
753 | ||
754 | /* If we haven't read in the extent list, then don't do it now. */ | |
755 | if (!(ip->i_df.if_flags & XFS_IFEXTENTS)) | |
756 | return false; | |
757 | ||
758 | /* | |
759 | * Do not free real preallocated or append-only files unless the file | |
760 | * has delalloc blocks and we are forced to remove them. | |
761 | */ | |
762 | if (ip->i_d.di_flags & (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) | |
763 | if (!force || ip->i_delayed_blks == 0) | |
764 | return false; | |
765 | ||
766 | return true; | |
767 | } | |
768 | ||
769 | /* | |
3b4683c2 BF |
770 | * This is called to free any blocks beyond eof. The caller must hold |
771 | * IOLOCK_EXCL unless we are in the inode reclaim path and have the only | |
772 | * reference to the inode. | |
c24b5dfa DC |
773 | */ |
774 | int | |
775 | xfs_free_eofblocks( | |
a36b9261 | 776 | struct xfs_inode *ip) |
c24b5dfa | 777 | { |
a36b9261 BF |
778 | struct xfs_trans *tp; |
779 | int error; | |
780 | xfs_fileoff_t end_fsb; | |
781 | xfs_fileoff_t last_fsb; | |
782 | xfs_filblks_t map_len; | |
783 | int nimaps; | |
784 | struct xfs_bmbt_irec imap; | |
785 | struct xfs_mount *mp = ip->i_mount; | |
786 | ||
c24b5dfa DC |
787 | /* |
788 | * Figure out if there are any blocks beyond the end | |
789 | * of the file. If not, then there is nothing to do. | |
790 | */ | |
791 | end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_ISIZE(ip)); | |
792 | last_fsb = XFS_B_TO_FSB(mp, mp->m_super->s_maxbytes); | |
793 | if (last_fsb <= end_fsb) | |
794 | return 0; | |
795 | map_len = last_fsb - end_fsb; | |
796 | ||
797 | nimaps = 1; | |
798 | xfs_ilock(ip, XFS_ILOCK_SHARED); | |
799 | error = xfs_bmapi_read(ip, end_fsb, map_len, &imap, &nimaps, 0); | |
800 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
801 | ||
a36b9261 BF |
802 | /* |
803 | * If there are blocks after the end of file, truncate the file to its | |
804 | * current size to free them up. | |
805 | */ | |
c24b5dfa DC |
806 | if (!error && (nimaps != 0) && |
807 | (imap.br_startblock != HOLESTARTBLOCK || | |
808 | ip->i_delayed_blks)) { | |
809 | /* | |
810 | * Attach the dquots to the inode up front. | |
811 | */ | |
c14cfcca | 812 | error = xfs_qm_dqattach(ip); |
c24b5dfa DC |
813 | if (error) |
814 | return error; | |
815 | ||
e4229d6b BF |
816 | /* wait on dio to ensure i_size has settled */ |
817 | inode_dio_wait(VFS_I(ip)); | |
818 | ||
253f4911 CH |
819 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, |
820 | &tp); | |
c24b5dfa DC |
821 | if (error) { |
822 | ASSERT(XFS_FORCED_SHUTDOWN(mp)); | |
c24b5dfa DC |
823 | return error; |
824 | } | |
825 | ||
826 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
827 | xfs_trans_ijoin(tp, ip, 0); | |
828 | ||
829 | /* | |
830 | * Do not update the on-disk file size. If we update the | |
831 | * on-disk file size and then the system crashes before the | |
832 | * contents of the file are flushed to disk then the files | |
833 | * may be full of holes (ie NULL files bug). | |
834 | */ | |
4e529339 BF |
835 | error = xfs_itruncate_extents_flags(&tp, ip, XFS_DATA_FORK, |
836 | XFS_ISIZE(ip), XFS_BMAPI_NODISCARD); | |
c24b5dfa DC |
837 | if (error) { |
838 | /* | |
839 | * If we get an error at this point we simply don't | |
840 | * bother truncating the file. | |
841 | */ | |
4906e215 | 842 | xfs_trans_cancel(tp); |
c24b5dfa | 843 | } else { |
70393313 | 844 | error = xfs_trans_commit(tp); |
c24b5dfa DC |
845 | if (!error) |
846 | xfs_inode_clear_eofblocks_tag(ip); | |
847 | } | |
848 | ||
849 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
c24b5dfa DC |
850 | } |
851 | return error; | |
852 | } | |
853 | ||
83aee9e4 | 854 | int |
c24b5dfa | 855 | xfs_alloc_file_space( |
83aee9e4 | 856 | struct xfs_inode *ip, |
c24b5dfa DC |
857 | xfs_off_t offset, |
858 | xfs_off_t len, | |
5f8aca8b | 859 | int alloc_type) |
c24b5dfa DC |
860 | { |
861 | xfs_mount_t *mp = ip->i_mount; | |
862 | xfs_off_t count; | |
863 | xfs_filblks_t allocated_fsb; | |
864 | xfs_filblks_t allocatesize_fsb; | |
865 | xfs_extlen_t extsz, temp; | |
866 | xfs_fileoff_t startoffset_fsb; | |
e093c4be | 867 | xfs_fileoff_t endoffset_fsb; |
c24b5dfa DC |
868 | int nimaps; |
869 | int quota_flag; | |
870 | int rt; | |
871 | xfs_trans_t *tp; | |
872 | xfs_bmbt_irec_t imaps[1], *imapp; | |
c24b5dfa | 873 | uint qblocks, resblks, resrtextents; |
c24b5dfa DC |
874 | int error; |
875 | ||
876 | trace_xfs_alloc_file_space(ip); | |
877 | ||
878 | if (XFS_FORCED_SHUTDOWN(mp)) | |
2451337d | 879 | return -EIO; |
c24b5dfa | 880 | |
c14cfcca | 881 | error = xfs_qm_dqattach(ip); |
c24b5dfa DC |
882 | if (error) |
883 | return error; | |
884 | ||
885 | if (len <= 0) | |
2451337d | 886 | return -EINVAL; |
c24b5dfa DC |
887 | |
888 | rt = XFS_IS_REALTIME_INODE(ip); | |
889 | extsz = xfs_get_extsz_hint(ip); | |
890 | ||
891 | count = len; | |
892 | imapp = &imaps[0]; | |
893 | nimaps = 1; | |
894 | startoffset_fsb = XFS_B_TO_FSBT(mp, offset); | |
e093c4be MR |
895 | endoffset_fsb = XFS_B_TO_FSB(mp, offset + count); |
896 | allocatesize_fsb = endoffset_fsb - startoffset_fsb; | |
c24b5dfa DC |
897 | |
898 | /* | |
899 | * Allocate file space until done or until there is an error | |
900 | */ | |
901 | while (allocatesize_fsb && !error) { | |
902 | xfs_fileoff_t s, e; | |
903 | ||
904 | /* | |
905 | * Determine space reservations for data/realtime. | |
906 | */ | |
907 | if (unlikely(extsz)) { | |
908 | s = startoffset_fsb; | |
909 | do_div(s, extsz); | |
910 | s *= extsz; | |
911 | e = startoffset_fsb + allocatesize_fsb; | |
0703a8e1 DC |
912 | div_u64_rem(startoffset_fsb, extsz, &temp); |
913 | if (temp) | |
c24b5dfa | 914 | e += temp; |
0703a8e1 DC |
915 | div_u64_rem(e, extsz, &temp); |
916 | if (temp) | |
c24b5dfa DC |
917 | e += extsz - temp; |
918 | } else { | |
919 | s = 0; | |
920 | e = allocatesize_fsb; | |
921 | } | |
922 | ||
923 | /* | |
924 | * The transaction reservation is limited to a 32-bit block | |
925 | * count, hence we need to limit the number of blocks we are | |
926 | * trying to reserve to avoid an overflow. We can't allocate | |
927 | * more than @nimaps extents, and an extent is limited on disk | |
928 | * to MAXEXTLEN (21 bits), so use that to enforce the limit. | |
929 | */ | |
930 | resblks = min_t(xfs_fileoff_t, (e - s), (MAXEXTLEN * nimaps)); | |
931 | if (unlikely(rt)) { | |
932 | resrtextents = qblocks = resblks; | |
933 | resrtextents /= mp->m_sb.sb_rextsize; | |
934 | resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0); | |
935 | quota_flag = XFS_QMOPT_RES_RTBLKS; | |
936 | } else { | |
937 | resrtextents = 0; | |
938 | resblks = qblocks = XFS_DIOSTRAT_SPACE_RES(mp, resblks); | |
939 | quota_flag = XFS_QMOPT_RES_REGBLKS; | |
940 | } | |
941 | ||
942 | /* | |
943 | * Allocate and setup the transaction. | |
944 | */ | |
253f4911 CH |
945 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, |
946 | resrtextents, 0, &tp); | |
947 | ||
c24b5dfa DC |
948 | /* |
949 | * Check for running out of space | |
950 | */ | |
951 | if (error) { | |
952 | /* | |
953 | * Free the transaction structure. | |
954 | */ | |
2451337d | 955 | ASSERT(error == -ENOSPC || XFS_FORCED_SHUTDOWN(mp)); |
c24b5dfa DC |
956 | break; |
957 | } | |
958 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
959 | error = xfs_trans_reserve_quota_nblks(tp, ip, qblocks, | |
960 | 0, quota_flag); | |
961 | if (error) | |
962 | goto error1; | |
963 | ||
964 | xfs_trans_ijoin(tp, ip, 0); | |
965 | ||
c24b5dfa | 966 | error = xfs_bmapi_write(tp, ip, startoffset_fsb, |
a7beabea BF |
967 | allocatesize_fsb, alloc_type, resblks, |
968 | imapp, &nimaps); | |
f6106efa | 969 | if (error) |
c24b5dfa | 970 | goto error0; |
c24b5dfa DC |
971 | |
972 | /* | |
973 | * Complete the transaction | |
974 | */ | |
70393313 | 975 | error = xfs_trans_commit(tp); |
c24b5dfa | 976 | xfs_iunlock(ip, XFS_ILOCK_EXCL); |
f6106efa | 977 | if (error) |
c24b5dfa | 978 | break; |
c24b5dfa DC |
979 | |
980 | allocated_fsb = imapp->br_blockcount; | |
981 | ||
982 | if (nimaps == 0) { | |
2451337d | 983 | error = -ENOSPC; |
c24b5dfa DC |
984 | break; |
985 | } | |
986 | ||
987 | startoffset_fsb += allocated_fsb; | |
988 | allocatesize_fsb -= allocated_fsb; | |
989 | } | |
990 | ||
991 | return error; | |
992 | ||
c8eac49e | 993 | error0: /* unlock inode, unreserve quota blocks, cancel trans */ |
c24b5dfa DC |
994 | xfs_trans_unreserve_quota_nblks(tp, ip, (long)qblocks, 0, quota_flag); |
995 | ||
996 | error1: /* Just cancel transaction */ | |
4906e215 | 997 | xfs_trans_cancel(tp); |
c24b5dfa DC |
998 | xfs_iunlock(ip, XFS_ILOCK_EXCL); |
999 | return error; | |
1000 | } | |
1001 | ||
bdb0d04f CH |
1002 | static int |
1003 | xfs_unmap_extent( | |
1004 | struct xfs_inode *ip, | |
1005 | xfs_fileoff_t startoffset_fsb, | |
1006 | xfs_filblks_t len_fsb, | |
1007 | int *done) | |
c24b5dfa | 1008 | { |
bdb0d04f CH |
1009 | struct xfs_mount *mp = ip->i_mount; |
1010 | struct xfs_trans *tp; | |
bdb0d04f CH |
1011 | uint resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0); |
1012 | int error; | |
c24b5dfa | 1013 | |
bdb0d04f CH |
1014 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, 0, 0, &tp); |
1015 | if (error) { | |
1016 | ASSERT(error == -ENOSPC || XFS_FORCED_SHUTDOWN(mp)); | |
1017 | return error; | |
1018 | } | |
c24b5dfa | 1019 | |
bdb0d04f CH |
1020 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
1021 | error = xfs_trans_reserve_quota(tp, mp, ip->i_udquot, ip->i_gdquot, | |
1022 | ip->i_pdquot, resblks, 0, XFS_QMOPT_RES_REGBLKS); | |
1023 | if (error) | |
1024 | goto out_trans_cancel; | |
c24b5dfa | 1025 | |
bdb0d04f | 1026 | xfs_trans_ijoin(tp, ip, 0); |
4f317369 | 1027 | |
2af52842 | 1028 | error = xfs_bunmapi(tp, ip, startoffset_fsb, len_fsb, 0, 2, done); |
bdb0d04f | 1029 | if (error) |
c8eac49e | 1030 | goto out_trans_cancel; |
4f317369 | 1031 | |
bdb0d04f CH |
1032 | error = xfs_trans_commit(tp); |
1033 | out_unlock: | |
1034 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
1035 | return error; | |
4f69f578 | 1036 | |
bdb0d04f CH |
1037 | out_trans_cancel: |
1038 | xfs_trans_cancel(tp); | |
1039 | goto out_unlock; | |
1040 | } | |
4f69f578 | 1041 | |
2c307174 | 1042 | int |
bdb0d04f CH |
1043 | xfs_flush_unmap_range( |
1044 | struct xfs_inode *ip, | |
1045 | xfs_off_t offset, | |
1046 | xfs_off_t len) | |
1047 | { | |
1048 | struct xfs_mount *mp = ip->i_mount; | |
1049 | struct inode *inode = VFS_I(ip); | |
1050 | xfs_off_t rounding, start, end; | |
1051 | int error; | |
1052 | ||
1053 | /* wait for the completion of any pending DIOs */ | |
1054 | inode_dio_wait(inode); | |
1055 | ||
1056 | rounding = max_t(xfs_off_t, 1 << mp->m_sb.sb_blocklog, PAGE_SIZE); | |
1057 | start = round_down(offset, rounding); | |
1058 | end = round_up(offset + len, rounding) - 1; | |
1059 | ||
1060 | error = filemap_write_and_wait_range(inode->i_mapping, start, end); | |
1061 | if (error) | |
1062 | return error; | |
1063 | truncate_pagecache_range(inode, start, end); | |
1064 | return 0; | |
c24b5dfa DC |
1065 | } |
1066 | ||
83aee9e4 | 1067 | int |
c24b5dfa | 1068 | xfs_free_file_space( |
83aee9e4 | 1069 | struct xfs_inode *ip, |
c24b5dfa | 1070 | xfs_off_t offset, |
5f8aca8b | 1071 | xfs_off_t len) |
c24b5dfa | 1072 | { |
bdb0d04f | 1073 | struct xfs_mount *mp = ip->i_mount; |
c24b5dfa | 1074 | xfs_fileoff_t startoffset_fsb; |
bdb0d04f | 1075 | xfs_fileoff_t endoffset_fsb; |
3c2bdc91 | 1076 | int done = 0, error; |
c24b5dfa DC |
1077 | |
1078 | trace_xfs_free_file_space(ip); | |
1079 | ||
c14cfcca | 1080 | error = xfs_qm_dqattach(ip); |
c24b5dfa DC |
1081 | if (error) |
1082 | return error; | |
1083 | ||
c24b5dfa | 1084 | if (len <= 0) /* if nothing being freed */ |
bdb0d04f | 1085 | return 0; |
c24b5dfa | 1086 | |
bdb0d04f | 1087 | error = xfs_flush_unmap_range(ip, offset, len); |
c24b5dfa | 1088 | if (error) |
bdb0d04f CH |
1089 | return error; |
1090 | ||
1091 | startoffset_fsb = XFS_B_TO_FSB(mp, offset); | |
1092 | endoffset_fsb = XFS_B_TO_FSBT(mp, offset + len); | |
c24b5dfa DC |
1093 | |
1094 | /* | |
daa79bae | 1095 | * Need to zero the stuff we're not freeing, on disk. |
c24b5dfa | 1096 | */ |
3c2bdc91 CH |
1097 | if (endoffset_fsb > startoffset_fsb) { |
1098 | while (!done) { | |
1099 | error = xfs_unmap_extent(ip, startoffset_fsb, | |
1100 | endoffset_fsb - startoffset_fsb, &done); | |
1101 | if (error) | |
1102 | return error; | |
c24b5dfa | 1103 | } |
c24b5dfa DC |
1104 | } |
1105 | ||
3c2bdc91 CH |
1106 | /* |
1107 | * Now that we've unmap all full blocks we'll have to zero out any | |
f5c54717 CH |
1108 | * partial block at the beginning and/or end. iomap_zero_range is smart |
1109 | * enough to skip any holes, including those we just created, but we | |
1110 | * must take care not to zero beyond EOF and enlarge i_size. | |
3c2bdc91 | 1111 | */ |
3dd09d5a CO |
1112 | if (offset >= XFS_ISIZE(ip)) |
1113 | return 0; | |
3dd09d5a CO |
1114 | if (offset + len > XFS_ISIZE(ip)) |
1115 | len = XFS_ISIZE(ip) - offset; | |
f150b423 CH |
1116 | error = iomap_zero_range(VFS_I(ip), offset, len, NULL, |
1117 | &xfs_buffered_write_iomap_ops); | |
e53c4b59 DW |
1118 | if (error) |
1119 | return error; | |
1120 | ||
1121 | /* | |
1122 | * If we zeroed right up to EOF and EOF straddles a page boundary we | |
1123 | * must make sure that the post-EOF area is also zeroed because the | |
1124 | * page could be mmap'd and iomap_zero_range doesn't do that for us. | |
1125 | * Writeback of the eof page will do this, albeit clumsily. | |
1126 | */ | |
a579121f | 1127 | if (offset + len >= XFS_ISIZE(ip) && offset_in_page(offset + len) > 0) { |
e53c4b59 | 1128 | error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, |
a579121f | 1129 | round_down(offset + len, PAGE_SIZE), LLONG_MAX); |
e53c4b59 DW |
1130 | } |
1131 | ||
1132 | return error; | |
c24b5dfa DC |
1133 | } |
1134 | ||
5d11fb4b BF |
1135 | /* |
1136 | * Preallocate and zero a range of a file. This mechanism has the allocation | |
1137 | * semantics of fallocate and in addition converts data in the range to zeroes. | |
1138 | */ | |
865e9446 | 1139 | int |
c24b5dfa DC |
1140 | xfs_zero_file_space( |
1141 | struct xfs_inode *ip, | |
1142 | xfs_off_t offset, | |
5f8aca8b | 1143 | xfs_off_t len) |
c24b5dfa DC |
1144 | { |
1145 | struct xfs_mount *mp = ip->i_mount; | |
5d11fb4b | 1146 | uint blksize; |
c24b5dfa DC |
1147 | int error; |
1148 | ||
897b73b6 DC |
1149 | trace_xfs_zero_file_space(ip); |
1150 | ||
5d11fb4b | 1151 | blksize = 1 << mp->m_sb.sb_blocklog; |
c24b5dfa DC |
1152 | |
1153 | /* | |
5d11fb4b BF |
1154 | * Punch a hole and prealloc the range. We use hole punch rather than |
1155 | * unwritten extent conversion for two reasons: | |
1156 | * | |
1157 | * 1.) Hole punch handles partial block zeroing for us. | |
1158 | * | |
1159 | * 2.) If prealloc returns ENOSPC, the file range is still zero-valued | |
1160 | * by virtue of the hole punch. | |
c24b5dfa | 1161 | */ |
5d11fb4b | 1162 | error = xfs_free_file_space(ip, offset, len); |
66ae56a5 CH |
1163 | if (error || xfs_is_always_cow_inode(ip)) |
1164 | return error; | |
c24b5dfa | 1165 | |
66ae56a5 | 1166 | return xfs_alloc_file_space(ip, round_down(offset, blksize), |
5d11fb4b BF |
1167 | round_up(offset + len, blksize) - |
1168 | round_down(offset, blksize), | |
1169 | XFS_BMAPI_PREALLOC); | |
c24b5dfa DC |
1170 | } |
1171 | ||
72c1a739 | 1172 | static int |
4ed36c6b CH |
1173 | xfs_prepare_shift( |
1174 | struct xfs_inode *ip, | |
1175 | loff_t offset) | |
e1d8fb88 | 1176 | { |
e1d8fb88 | 1177 | int error; |
e1d8fb88 | 1178 | |
f71721d0 BF |
1179 | /* |
1180 | * Trim eofblocks to avoid shifting uninitialized post-eof preallocation | |
1181 | * into the accessible region of the file. | |
1182 | */ | |
41b9d726 | 1183 | if (xfs_can_free_eofblocks(ip, true)) { |
a36b9261 | 1184 | error = xfs_free_eofblocks(ip); |
41b9d726 BF |
1185 | if (error) |
1186 | return error; | |
1187 | } | |
1669a8ca | 1188 | |
f71721d0 BF |
1189 | /* |
1190 | * Writeback and invalidate cache for the remainder of the file as we're | |
a904b1ca | 1191 | * about to shift down every extent from offset to EOF. |
f71721d0 | 1192 | */ |
7f9f71be | 1193 | error = xfs_flush_unmap_range(ip, offset, XFS_ISIZE(ip)); |
1749d1ea BF |
1194 | if (error) |
1195 | return error; | |
e1d8fb88 | 1196 | |
a904b1ca | 1197 | /* |
3af423b0 DW |
1198 | * Clean out anything hanging around in the cow fork now that |
1199 | * we've flushed all the dirty data out to disk to avoid having | |
1200 | * CoW extents at the wrong offsets. | |
1201 | */ | |
51d62690 | 1202 | if (xfs_inode_has_cow_data(ip)) { |
3af423b0 DW |
1203 | error = xfs_reflink_cancel_cow_range(ip, offset, NULLFILEOFF, |
1204 | true); | |
1205 | if (error) | |
1206 | return error; | |
1207 | } | |
1208 | ||
4ed36c6b CH |
1209 | return 0; |
1210 | } | |
1211 | ||
1212 | /* | |
1213 | * xfs_collapse_file_space() | |
1214 | * This routine frees disk space and shift extent for the given file. | |
1215 | * The first thing we do is to free data blocks in the specified range | |
1216 | * by calling xfs_free_file_space(). It would also sync dirty data | |
1217 | * and invalidate page cache over the region on which collapse range | |
1218 | * is working. And Shift extent records to the left to cover a hole. | |
1219 | * RETURNS: | |
1220 | * 0 on success | |
1221 | * errno on error | |
1222 | * | |
1223 | */ | |
1224 | int | |
1225 | xfs_collapse_file_space( | |
1226 | struct xfs_inode *ip, | |
1227 | xfs_off_t offset, | |
1228 | xfs_off_t len) | |
1229 | { | |
4ed36c6b CH |
1230 | struct xfs_mount *mp = ip->i_mount; |
1231 | struct xfs_trans *tp; | |
1232 | int error; | |
4ed36c6b CH |
1233 | xfs_fileoff_t next_fsb = XFS_B_TO_FSB(mp, offset + len); |
1234 | xfs_fileoff_t shift_fsb = XFS_B_TO_FSB(mp, len); | |
1235 | uint resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0); | |
ecfea3f0 | 1236 | bool done = false; |
4ed36c6b CH |
1237 | |
1238 | ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); | |
9ad1a23a CH |
1239 | ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL)); |
1240 | ||
4ed36c6b CH |
1241 | trace_xfs_collapse_file_space(ip); |
1242 | ||
1243 | error = xfs_free_file_space(ip, offset, len); | |
1244 | if (error) | |
1245 | return error; | |
1246 | ||
1247 | error = xfs_prepare_shift(ip, offset); | |
1248 | if (error) | |
1249 | return error; | |
a904b1ca | 1250 | |
e1d8fb88 | 1251 | while (!error && !done) { |
48af96ab BF |
1252 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, 0, 0, |
1253 | &tp); | |
253f4911 | 1254 | if (error) |
e1d8fb88 | 1255 | break; |
e1d8fb88 NJ |
1256 | |
1257 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
1258 | error = xfs_trans_reserve_quota(tp, mp, ip->i_udquot, | |
48af96ab | 1259 | ip->i_gdquot, ip->i_pdquot, resblks, 0, |
e1d8fb88 NJ |
1260 | XFS_QMOPT_RES_REGBLKS); |
1261 | if (error) | |
d4a97a04 | 1262 | goto out_trans_cancel; |
a904b1ca | 1263 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); |
e1d8fb88 | 1264 | |
ecfea3f0 | 1265 | error = xfs_bmap_collapse_extents(tp, ip, &next_fsb, shift_fsb, |
333f950c | 1266 | &done); |
e1d8fb88 | 1267 | if (error) |
c8eac49e | 1268 | goto out_trans_cancel; |
e1d8fb88 | 1269 | |
70393313 | 1270 | error = xfs_trans_commit(tp); |
e1d8fb88 NJ |
1271 | } |
1272 | ||
1273 | return error; | |
1274 | ||
d4a97a04 | 1275 | out_trans_cancel: |
4906e215 | 1276 | xfs_trans_cancel(tp); |
e1d8fb88 NJ |
1277 | return error; |
1278 | } | |
1279 | ||
a904b1ca NJ |
1280 | /* |
1281 | * xfs_insert_file_space() | |
1282 | * This routine create hole space by shifting extents for the given file. | |
1283 | * The first thing we do is to sync dirty data and invalidate page cache | |
1284 | * over the region on which insert range is working. And split an extent | |
1285 | * to two extents at given offset by calling xfs_bmap_split_extent. | |
1286 | * And shift all extent records which are laying between [offset, | |
1287 | * last allocated extent] to the right to reserve hole range. | |
1288 | * RETURNS: | |
1289 | * 0 on success | |
1290 | * errno on error | |
1291 | */ | |
1292 | int | |
1293 | xfs_insert_file_space( | |
1294 | struct xfs_inode *ip, | |
1295 | loff_t offset, | |
1296 | loff_t len) | |
1297 | { | |
4ed36c6b CH |
1298 | struct xfs_mount *mp = ip->i_mount; |
1299 | struct xfs_trans *tp; | |
1300 | int error; | |
4ed36c6b CH |
1301 | xfs_fileoff_t stop_fsb = XFS_B_TO_FSB(mp, offset); |
1302 | xfs_fileoff_t next_fsb = NULLFSBLOCK; | |
1303 | xfs_fileoff_t shift_fsb = XFS_B_TO_FSB(mp, len); | |
ecfea3f0 | 1304 | bool done = false; |
4ed36c6b | 1305 | |
a904b1ca | 1306 | ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); |
9ad1a23a CH |
1307 | ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL)); |
1308 | ||
a904b1ca NJ |
1309 | trace_xfs_insert_file_space(ip); |
1310 | ||
f62cb48e DW |
1311 | error = xfs_bmap_can_insert_extents(ip, stop_fsb, shift_fsb); |
1312 | if (error) | |
1313 | return error; | |
1314 | ||
4ed36c6b CH |
1315 | error = xfs_prepare_shift(ip, offset); |
1316 | if (error) | |
1317 | return error; | |
1318 | ||
1319 | /* | |
1320 | * The extent shifting code works on extent granularity. So, if stop_fsb | |
1321 | * is not the starting block of extent, we need to split the extent at | |
1322 | * stop_fsb. | |
1323 | */ | |
1324 | error = xfs_bmap_split_extent(ip, stop_fsb); | |
1325 | if (error) | |
1326 | return error; | |
1327 | ||
1328 | while (!error && !done) { | |
1329 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, 0, 0, 0, | |
1330 | &tp); | |
1331 | if (error) | |
1332 | break; | |
1333 | ||
1334 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
1335 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
ecfea3f0 | 1336 | error = xfs_bmap_insert_extents(tp, ip, &next_fsb, shift_fsb, |
333f950c | 1337 | &done, stop_fsb); |
4ed36c6b | 1338 | if (error) |
c8eac49e | 1339 | goto out_trans_cancel; |
4ed36c6b | 1340 | |
4ed36c6b CH |
1341 | error = xfs_trans_commit(tp); |
1342 | } | |
1343 | ||
1344 | return error; | |
1345 | ||
c8eac49e | 1346 | out_trans_cancel: |
4ed36c6b CH |
1347 | xfs_trans_cancel(tp); |
1348 | return error; | |
a904b1ca NJ |
1349 | } |
1350 | ||
a133d952 DC |
1351 | /* |
1352 | * We need to check that the format of the data fork in the temporary inode is | |
1353 | * valid for the target inode before doing the swap. This is not a problem with | |
1354 | * attr1 because of the fixed fork offset, but attr2 has a dynamically sized | |
1355 | * data fork depending on the space the attribute fork is taking so we can get | |
1356 | * invalid formats on the target inode. | |
1357 | * | |
1358 | * E.g. target has space for 7 extents in extent format, temp inode only has | |
1359 | * space for 6. If we defragment down to 7 extents, then the tmp format is a | |
1360 | * btree, but when swapped it needs to be in extent format. Hence we can't just | |
1361 | * blindly swap data forks on attr2 filesystems. | |
1362 | * | |
1363 | * Note that we check the swap in both directions so that we don't end up with | |
1364 | * a corrupt temporary inode, either. | |
1365 | * | |
1366 | * Note that fixing the way xfs_fsr sets up the attribute fork in the source | |
1367 | * inode will prevent this situation from occurring, so all we do here is | |
1368 | * reject and log the attempt. basically we are putting the responsibility on | |
1369 | * userspace to get this right. | |
1370 | */ | |
1371 | static int | |
1372 | xfs_swap_extents_check_format( | |
e06259aa DW |
1373 | struct xfs_inode *ip, /* target inode */ |
1374 | struct xfs_inode *tip) /* tmp inode */ | |
a133d952 DC |
1375 | { |
1376 | ||
1377 | /* Should never get a local format */ | |
1378 | if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || | |
1379 | tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) | |
2451337d | 1380 | return -EINVAL; |
a133d952 DC |
1381 | |
1382 | /* | |
1383 | * if the target inode has less extents that then temporary inode then | |
1384 | * why did userspace call us? | |
1385 | */ | |
1386 | if (ip->i_d.di_nextents < tip->i_d.di_nextents) | |
2451337d | 1387 | return -EINVAL; |
a133d952 | 1388 | |
1f08af52 DW |
1389 | /* |
1390 | * If we have to use the (expensive) rmap swap method, we can | |
1391 | * handle any number of extents and any format. | |
1392 | */ | |
1393 | if (xfs_sb_version_hasrmapbt(&ip->i_mount->m_sb)) | |
1394 | return 0; | |
1395 | ||
a133d952 DC |
1396 | /* |
1397 | * if the target inode is in extent form and the temp inode is in btree | |
1398 | * form then we will end up with the target inode in the wrong format | |
1399 | * as we already know there are less extents in the temp inode. | |
1400 | */ | |
1401 | if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS && | |
1402 | tip->i_d.di_format == XFS_DINODE_FMT_BTREE) | |
2451337d | 1403 | return -EINVAL; |
a133d952 DC |
1404 | |
1405 | /* Check temp in extent form to max in target */ | |
1406 | if (tip->i_d.di_format == XFS_DINODE_FMT_EXTENTS && | |
1407 | XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) > | |
1408 | XFS_IFORK_MAXEXT(ip, XFS_DATA_FORK)) | |
2451337d | 1409 | return -EINVAL; |
a133d952 DC |
1410 | |
1411 | /* Check target in extent form to max in temp */ | |
1412 | if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS && | |
1413 | XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) > | |
1414 | XFS_IFORK_MAXEXT(tip, XFS_DATA_FORK)) | |
2451337d | 1415 | return -EINVAL; |
a133d952 DC |
1416 | |
1417 | /* | |
1418 | * If we are in a btree format, check that the temp root block will fit | |
1419 | * in the target and that it has enough extents to be in btree format | |
1420 | * in the target. | |
1421 | * | |
1422 | * Note that we have to be careful to allow btree->extent conversions | |
1423 | * (a common defrag case) which will occur when the temp inode is in | |
1424 | * extent format... | |
1425 | */ | |
1426 | if (tip->i_d.di_format == XFS_DINODE_FMT_BTREE) { | |
0cbe48cc | 1427 | if (XFS_IFORK_Q(ip) && |
a133d952 | 1428 | XFS_BMAP_BMDR_SPACE(tip->i_df.if_broot) > XFS_IFORK_BOFF(ip)) |
2451337d | 1429 | return -EINVAL; |
a133d952 DC |
1430 | if (XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) <= |
1431 | XFS_IFORK_MAXEXT(ip, XFS_DATA_FORK)) | |
2451337d | 1432 | return -EINVAL; |
a133d952 DC |
1433 | } |
1434 | ||
1435 | /* Reciprocal target->temp btree format checks */ | |
1436 | if (ip->i_d.di_format == XFS_DINODE_FMT_BTREE) { | |
0cbe48cc | 1437 | if (XFS_IFORK_Q(tip) && |
a133d952 | 1438 | XFS_BMAP_BMDR_SPACE(ip->i_df.if_broot) > XFS_IFORK_BOFF(tip)) |
2451337d | 1439 | return -EINVAL; |
a133d952 DC |
1440 | if (XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) <= |
1441 | XFS_IFORK_MAXEXT(tip, XFS_DATA_FORK)) | |
2451337d | 1442 | return -EINVAL; |
a133d952 DC |
1443 | } |
1444 | ||
1445 | return 0; | |
1446 | } | |
1447 | ||
7abbb8f9 | 1448 | static int |
4ef897a2 DC |
1449 | xfs_swap_extent_flush( |
1450 | struct xfs_inode *ip) | |
1451 | { | |
1452 | int error; | |
1453 | ||
1454 | error = filemap_write_and_wait(VFS_I(ip)->i_mapping); | |
1455 | if (error) | |
1456 | return error; | |
1457 | truncate_pagecache_range(VFS_I(ip), 0, -1); | |
1458 | ||
1459 | /* Verify O_DIRECT for ftmp */ | |
1460 | if (VFS_I(ip)->i_mapping->nrpages) | |
1461 | return -EINVAL; | |
4ef897a2 DC |
1462 | return 0; |
1463 | } | |
1464 | ||
1f08af52 DW |
1465 | /* |
1466 | * Move extents from one file to another, when rmap is enabled. | |
1467 | */ | |
1468 | STATIC int | |
1469 | xfs_swap_extent_rmap( | |
1470 | struct xfs_trans **tpp, | |
1471 | struct xfs_inode *ip, | |
1472 | struct xfs_inode *tip) | |
1473 | { | |
7a7943c7 | 1474 | struct xfs_trans *tp = *tpp; |
1f08af52 DW |
1475 | struct xfs_bmbt_irec irec; |
1476 | struct xfs_bmbt_irec uirec; | |
1477 | struct xfs_bmbt_irec tirec; | |
1478 | xfs_fileoff_t offset_fsb; | |
1479 | xfs_fileoff_t end_fsb; | |
1480 | xfs_filblks_t count_fsb; | |
1f08af52 DW |
1481 | int error; |
1482 | xfs_filblks_t ilen; | |
1483 | xfs_filblks_t rlen; | |
1484 | int nimaps; | |
c8ce540d | 1485 | uint64_t tip_flags2; |
1f08af52 DW |
1486 | |
1487 | /* | |
1488 | * If the source file has shared blocks, we must flag the donor | |
1489 | * file as having shared blocks so that we get the shared-block | |
1490 | * rmap functions when we go to fix up the rmaps. The flags | |
1491 | * will be switch for reals later. | |
1492 | */ | |
1493 | tip_flags2 = tip->i_d.di_flags2; | |
1494 | if (ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK) | |
1495 | tip->i_d.di_flags2 |= XFS_DIFLAG2_REFLINK; | |
1496 | ||
1497 | offset_fsb = 0; | |
1498 | end_fsb = XFS_B_TO_FSB(ip->i_mount, i_size_read(VFS_I(ip))); | |
1499 | count_fsb = (xfs_filblks_t)(end_fsb - offset_fsb); | |
1500 | ||
1501 | while (count_fsb) { | |
1502 | /* Read extent from the donor file */ | |
1503 | nimaps = 1; | |
1504 | error = xfs_bmapi_read(tip, offset_fsb, count_fsb, &tirec, | |
1505 | &nimaps, 0); | |
1506 | if (error) | |
1507 | goto out; | |
1508 | ASSERT(nimaps == 1); | |
1509 | ASSERT(tirec.br_startblock != DELAYSTARTBLOCK); | |
1510 | ||
1511 | trace_xfs_swap_extent_rmap_remap(tip, &tirec); | |
1512 | ilen = tirec.br_blockcount; | |
1513 | ||
1514 | /* Unmap the old blocks in the source file. */ | |
1515 | while (tirec.br_blockcount) { | |
c8eac49e | 1516 | ASSERT(tp->t_firstblock == NULLFSBLOCK); |
1f08af52 DW |
1517 | trace_xfs_swap_extent_rmap_remap_piece(tip, &tirec); |
1518 | ||
1519 | /* Read extent from the source file */ | |
1520 | nimaps = 1; | |
1521 | error = xfs_bmapi_read(ip, tirec.br_startoff, | |
1522 | tirec.br_blockcount, &irec, | |
1523 | &nimaps, 0); | |
1524 | if (error) | |
d5a2e289 | 1525 | goto out; |
1f08af52 DW |
1526 | ASSERT(nimaps == 1); |
1527 | ASSERT(tirec.br_startoff == irec.br_startoff); | |
1528 | trace_xfs_swap_extent_rmap_remap_piece(ip, &irec); | |
1529 | ||
1530 | /* Trim the extent. */ | |
1531 | uirec = tirec; | |
1532 | uirec.br_blockcount = rlen = min_t(xfs_filblks_t, | |
1533 | tirec.br_blockcount, | |
1534 | irec.br_blockcount); | |
1535 | trace_xfs_swap_extent_rmap_remap_piece(tip, &uirec); | |
1536 | ||
1537 | /* Remove the mapping from the donor file. */ | |
3e08f42a | 1538 | xfs_bmap_unmap_extent(tp, tip, &uirec); |
1f08af52 DW |
1539 | |
1540 | /* Remove the mapping from the source file. */ | |
3e08f42a | 1541 | xfs_bmap_unmap_extent(tp, ip, &irec); |
1f08af52 DW |
1542 | |
1543 | /* Map the donor file's blocks into the source file. */ | |
3e08f42a | 1544 | xfs_bmap_map_extent(tp, ip, &uirec); |
1f08af52 DW |
1545 | |
1546 | /* Map the source file's blocks into the donor file. */ | |
3e08f42a | 1547 | xfs_bmap_map_extent(tp, tip, &irec); |
1f08af52 | 1548 | |
9e28a242 | 1549 | error = xfs_defer_finish(tpp); |
7a7943c7 | 1550 | tp = *tpp; |
1f08af52 | 1551 | if (error) |
9b1f4e98 | 1552 | goto out; |
1f08af52 DW |
1553 | |
1554 | tirec.br_startoff += rlen; | |
1555 | if (tirec.br_startblock != HOLESTARTBLOCK && | |
1556 | tirec.br_startblock != DELAYSTARTBLOCK) | |
1557 | tirec.br_startblock += rlen; | |
1558 | tirec.br_blockcount -= rlen; | |
1559 | } | |
1560 | ||
1561 | /* Roll on... */ | |
1562 | count_fsb -= ilen; | |
1563 | offset_fsb += ilen; | |
1564 | } | |
1565 | ||
1566 | tip->i_d.di_flags2 = tip_flags2; | |
1567 | return 0; | |
1568 | ||
1f08af52 DW |
1569 | out: |
1570 | trace_xfs_swap_extent_rmap_error(ip, error, _RET_IP_); | |
1571 | tip->i_d.di_flags2 = tip_flags2; | |
1572 | return error; | |
1573 | } | |
1574 | ||
39aff5fd DW |
1575 | /* Swap the extents of two files by swapping data forks. */ |
1576 | STATIC int | |
1577 | xfs_swap_extent_forks( | |
1578 | struct xfs_trans *tp, | |
1579 | struct xfs_inode *ip, | |
1580 | struct xfs_inode *tip, | |
1581 | int *src_log_flags, | |
1582 | int *target_log_flags) | |
a133d952 | 1583 | { |
e7f5d5ca DW |
1584 | xfs_filblks_t aforkblks = 0; |
1585 | xfs_filblks_t taforkblks = 0; | |
1586 | xfs_extnum_t junk; | |
c8ce540d | 1587 | uint64_t tmp; |
39aff5fd | 1588 | int error; |
a133d952 | 1589 | |
a133d952 DC |
1590 | /* |
1591 | * Count the number of extended attribute blocks | |
1592 | */ | |
1593 | if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && | |
1594 | (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
e7f5d5ca | 1595 | error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &junk, |
39aff5fd | 1596 | &aforkblks); |
a133d952 | 1597 | if (error) |
39aff5fd | 1598 | return error; |
a133d952 DC |
1599 | } |
1600 | if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && | |
1601 | (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
e7f5d5ca | 1602 | error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, &junk, |
39aff5fd | 1603 | &taforkblks); |
a133d952 | 1604 | if (error) |
39aff5fd | 1605 | return error; |
a133d952 DC |
1606 | } |
1607 | ||
21b5c978 | 1608 | /* |
6fb10d6d BF |
1609 | * Btree format (v3) inodes have the inode number stamped in the bmbt |
1610 | * block headers. We can't start changing the bmbt blocks until the | |
1611 | * inode owner change is logged so recovery does the right thing in the | |
1612 | * event of a crash. Set the owner change log flags now and leave the | |
1613 | * bmbt scan as the last step. | |
21b5c978 | 1614 | */ |
21b5c978 | 1615 | if (ip->i_d.di_version == 3 && |
6fb10d6d | 1616 | ip->i_d.di_format == XFS_DINODE_FMT_BTREE) |
39aff5fd | 1617 | (*target_log_flags) |= XFS_ILOG_DOWNER; |
21b5c978 | 1618 | if (tip->i_d.di_version == 3 && |
6fb10d6d | 1619 | tip->i_d.di_format == XFS_DINODE_FMT_BTREE) |
39aff5fd | 1620 | (*src_log_flags) |= XFS_ILOG_DOWNER; |
21b5c978 | 1621 | |
a133d952 DC |
1622 | /* |
1623 | * Swap the data forks of the inodes | |
1624 | */ | |
897992b7 | 1625 | swap(ip->i_df, tip->i_df); |
a133d952 DC |
1626 | |
1627 | /* | |
1628 | * Fix the on-disk inode values | |
1629 | */ | |
c8ce540d | 1630 | tmp = (uint64_t)ip->i_d.di_nblocks; |
a133d952 DC |
1631 | ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; |
1632 | tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; | |
1633 | ||
897992b7 GS |
1634 | swap(ip->i_d.di_nextents, tip->i_d.di_nextents); |
1635 | swap(ip->i_d.di_format, tip->i_d.di_format); | |
a133d952 DC |
1636 | |
1637 | /* | |
1638 | * The extents in the source inode could still contain speculative | |
1639 | * preallocation beyond EOF (e.g. the file is open but not modified | |
1640 | * while defrag is in progress). In that case, we need to copy over the | |
1641 | * number of delalloc blocks the data fork in the source inode is | |
1642 | * tracking beyond EOF so that when the fork is truncated away when the | |
1643 | * temporary inode is unlinked we don't underrun the i_delayed_blks | |
1644 | * counter on that inode. | |
1645 | */ | |
1646 | ASSERT(tip->i_delayed_blks == 0); | |
1647 | tip->i_delayed_blks = ip->i_delayed_blks; | |
1648 | ip->i_delayed_blks = 0; | |
1649 | ||
a133d952 DC |
1650 | switch (ip->i_d.di_format) { |
1651 | case XFS_DINODE_FMT_EXTENTS: | |
39aff5fd | 1652 | (*src_log_flags) |= XFS_ILOG_DEXT; |
a133d952 DC |
1653 | break; |
1654 | case XFS_DINODE_FMT_BTREE: | |
21b5c978 | 1655 | ASSERT(ip->i_d.di_version < 3 || |
39aff5fd DW |
1656 | (*src_log_flags & XFS_ILOG_DOWNER)); |
1657 | (*src_log_flags) |= XFS_ILOG_DBROOT; | |
a133d952 DC |
1658 | break; |
1659 | } | |
1660 | ||
a133d952 DC |
1661 | switch (tip->i_d.di_format) { |
1662 | case XFS_DINODE_FMT_EXTENTS: | |
39aff5fd | 1663 | (*target_log_flags) |= XFS_ILOG_DEXT; |
a133d952 DC |
1664 | break; |
1665 | case XFS_DINODE_FMT_BTREE: | |
39aff5fd | 1666 | (*target_log_flags) |= XFS_ILOG_DBROOT; |
21b5c978 | 1667 | ASSERT(tip->i_d.di_version < 3 || |
39aff5fd | 1668 | (*target_log_flags & XFS_ILOG_DOWNER)); |
a133d952 DC |
1669 | break; |
1670 | } | |
1671 | ||
39aff5fd DW |
1672 | return 0; |
1673 | } | |
1674 | ||
2dd3d709 BF |
1675 | /* |
1676 | * Fix up the owners of the bmbt blocks to refer to the current inode. The | |
1677 | * change owner scan attempts to order all modified buffers in the current | |
1678 | * transaction. In the event of ordered buffer failure, the offending buffer is | |
1679 | * physically logged as a fallback and the scan returns -EAGAIN. We must roll | |
1680 | * the transaction in this case to replenish the fallback log reservation and | |
1681 | * restart the scan. This process repeats until the scan completes. | |
1682 | */ | |
1683 | static int | |
1684 | xfs_swap_change_owner( | |
1685 | struct xfs_trans **tpp, | |
1686 | struct xfs_inode *ip, | |
1687 | struct xfs_inode *tmpip) | |
1688 | { | |
1689 | int error; | |
1690 | struct xfs_trans *tp = *tpp; | |
1691 | ||
1692 | do { | |
1693 | error = xfs_bmbt_change_owner(tp, ip, XFS_DATA_FORK, ip->i_ino, | |
1694 | NULL); | |
1695 | /* success or fatal error */ | |
1696 | if (error != -EAGAIN) | |
1697 | break; | |
1698 | ||
1699 | error = xfs_trans_roll(tpp); | |
1700 | if (error) | |
1701 | break; | |
1702 | tp = *tpp; | |
1703 | ||
1704 | /* | |
1705 | * Redirty both inodes so they can relog and keep the log tail | |
1706 | * moving forward. | |
1707 | */ | |
1708 | xfs_trans_ijoin(tp, ip, 0); | |
1709 | xfs_trans_ijoin(tp, tmpip, 0); | |
1710 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
1711 | xfs_trans_log_inode(tp, tmpip, XFS_ILOG_CORE); | |
1712 | } while (true); | |
1713 | ||
1714 | return error; | |
1715 | } | |
1716 | ||
39aff5fd DW |
1717 | int |
1718 | xfs_swap_extents( | |
1719 | struct xfs_inode *ip, /* target inode */ | |
1720 | struct xfs_inode *tip, /* tmp inode */ | |
1721 | struct xfs_swapext *sxp) | |
1722 | { | |
1723 | struct xfs_mount *mp = ip->i_mount; | |
1724 | struct xfs_trans *tp; | |
1725 | struct xfs_bstat *sbp = &sxp->sx_stat; | |
1726 | int src_log_flags, target_log_flags; | |
1727 | int error = 0; | |
1728 | int lock_flags; | |
c8ce540d | 1729 | uint64_t f; |
2dd3d709 | 1730 | int resblks = 0; |
39aff5fd DW |
1731 | |
1732 | /* | |
1733 | * Lock the inodes against other IO, page faults and truncate to | |
1734 | * begin with. Then we can ensure the inodes are flushed and have no | |
1735 | * page cache safely. Once we have done this we can take the ilocks and | |
1736 | * do the rest of the checks. | |
1737 | */ | |
65523218 CH |
1738 | lock_two_nondirectories(VFS_I(ip), VFS_I(tip)); |
1739 | lock_flags = XFS_MMAPLOCK_EXCL; | |
7c2d238a | 1740 | xfs_lock_two_inodes(ip, XFS_MMAPLOCK_EXCL, tip, XFS_MMAPLOCK_EXCL); |
39aff5fd DW |
1741 | |
1742 | /* Verify that both files have the same format */ | |
1743 | if ((VFS_I(ip)->i_mode & S_IFMT) != (VFS_I(tip)->i_mode & S_IFMT)) { | |
1744 | error = -EINVAL; | |
1745 | goto out_unlock; | |
1746 | } | |
1747 | ||
1748 | /* Verify both files are either real-time or non-realtime */ | |
1749 | if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) { | |
1750 | error = -EINVAL; | |
1751 | goto out_unlock; | |
1752 | } | |
1753 | ||
1754 | error = xfs_swap_extent_flush(ip); | |
1755 | if (error) | |
1756 | goto out_unlock; | |
1757 | error = xfs_swap_extent_flush(tip); | |
1758 | if (error) | |
1759 | goto out_unlock; | |
1760 | ||
96987eea CH |
1761 | if (xfs_inode_has_cow_data(tip)) { |
1762 | error = xfs_reflink_cancel_cow_range(tip, 0, NULLFILEOFF, true); | |
1763 | if (error) | |
1764 | return error; | |
1765 | } | |
1766 | ||
1f08af52 DW |
1767 | /* |
1768 | * Extent "swapping" with rmap requires a permanent reservation and | |
1769 | * a block reservation because it's really just a remap operation | |
1770 | * performed with log redo items! | |
1771 | */ | |
1772 | if (xfs_sb_version_hasrmapbt(&mp->m_sb)) { | |
b3fed434 BF |
1773 | int w = XFS_DATA_FORK; |
1774 | uint32_t ipnext = XFS_IFORK_NEXTENTS(ip, w); | |
1775 | uint32_t tipnext = XFS_IFORK_NEXTENTS(tip, w); | |
1776 | ||
1777 | /* | |
1778 | * Conceptually this shouldn't affect the shape of either bmbt, | |
1779 | * but since we atomically move extents one by one, we reserve | |
1780 | * enough space to rebuild both trees. | |
1781 | */ | |
1782 | resblks = XFS_SWAP_RMAP_SPACE_RES(mp, ipnext, w); | |
1783 | resblks += XFS_SWAP_RMAP_SPACE_RES(mp, tipnext, w); | |
1784 | ||
1f08af52 | 1785 | /* |
b3fed434 BF |
1786 | * Handle the corner case where either inode might straddle the |
1787 | * btree format boundary. If so, the inode could bounce between | |
1788 | * btree <-> extent format on unmap -> remap cycles, freeing and | |
1789 | * allocating a bmapbt block each time. | |
1f08af52 | 1790 | */ |
b3fed434 BF |
1791 | if (ipnext == (XFS_IFORK_MAXEXT(ip, w) + 1)) |
1792 | resblks += XFS_IFORK_MAXEXT(ip, w); | |
1793 | if (tipnext == (XFS_IFORK_MAXEXT(tip, w) + 1)) | |
1794 | resblks += XFS_IFORK_MAXEXT(tip, w); | |
2dd3d709 BF |
1795 | } |
1796 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, 0, 0, &tp); | |
39aff5fd DW |
1797 | if (error) |
1798 | goto out_unlock; | |
1799 | ||
1800 | /* | |
1801 | * Lock and join the inodes to the tansaction so that transaction commit | |
1802 | * or cancel will unlock the inodes from this point onwards. | |
1803 | */ | |
7c2d238a | 1804 | xfs_lock_two_inodes(ip, XFS_ILOCK_EXCL, tip, XFS_ILOCK_EXCL); |
39aff5fd DW |
1805 | lock_flags |= XFS_ILOCK_EXCL; |
1806 | xfs_trans_ijoin(tp, ip, 0); | |
1807 | xfs_trans_ijoin(tp, tip, 0); | |
1808 | ||
1809 | ||
1810 | /* Verify all data are being swapped */ | |
1811 | if (sxp->sx_offset != 0 || | |
1812 | sxp->sx_length != ip->i_d.di_size || | |
1813 | sxp->sx_length != tip->i_d.di_size) { | |
1814 | error = -EFAULT; | |
1815 | goto out_trans_cancel; | |
1816 | } | |
1817 | ||
1818 | trace_xfs_swap_extent_before(ip, 0); | |
1819 | trace_xfs_swap_extent_before(tip, 1); | |
1820 | ||
1821 | /* check inode formats now that data is flushed */ | |
1822 | error = xfs_swap_extents_check_format(ip, tip); | |
1823 | if (error) { | |
1824 | xfs_notice(mp, | |
1825 | "%s: inode 0x%llx format is incompatible for exchanging.", | |
1826 | __func__, ip->i_ino); | |
1827 | goto out_trans_cancel; | |
1828 | } | |
1829 | ||
1830 | /* | |
1831 | * Compare the current change & modify times with that | |
1832 | * passed in. If they differ, we abort this swap. | |
1833 | * This is the mechanism used to ensure the calling | |
1834 | * process that the file was not changed out from | |
1835 | * under it. | |
1836 | */ | |
1837 | if ((sbp->bs_ctime.tv_sec != VFS_I(ip)->i_ctime.tv_sec) || | |
1838 | (sbp->bs_ctime.tv_nsec != VFS_I(ip)->i_ctime.tv_nsec) || | |
1839 | (sbp->bs_mtime.tv_sec != VFS_I(ip)->i_mtime.tv_sec) || | |
1840 | (sbp->bs_mtime.tv_nsec != VFS_I(ip)->i_mtime.tv_nsec)) { | |
1841 | error = -EBUSY; | |
1842 | goto out_trans_cancel; | |
1843 | } | |
1844 | ||
1845 | /* | |
1846 | * Note the trickiness in setting the log flags - we set the owner log | |
1847 | * flag on the opposite inode (i.e. the inode we are setting the new | |
1848 | * owner to be) because once we swap the forks and log that, log | |
1849 | * recovery is going to see the fork as owned by the swapped inode, | |
1850 | * not the pre-swapped inodes. | |
1851 | */ | |
1852 | src_log_flags = XFS_ILOG_CORE; | |
1853 | target_log_flags = XFS_ILOG_CORE; | |
1854 | ||
1f08af52 DW |
1855 | if (xfs_sb_version_hasrmapbt(&mp->m_sb)) |
1856 | error = xfs_swap_extent_rmap(&tp, ip, tip); | |
1857 | else | |
1858 | error = xfs_swap_extent_forks(tp, ip, tip, &src_log_flags, | |
1859 | &target_log_flags); | |
39aff5fd DW |
1860 | if (error) |
1861 | goto out_trans_cancel; | |
1862 | ||
f0bc4d13 DW |
1863 | /* Do we have to swap reflink flags? */ |
1864 | if ((ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK) ^ | |
1865 | (tip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK)) { | |
1866 | f = ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK; | |
1867 | ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK; | |
1868 | ip->i_d.di_flags2 |= tip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK; | |
1869 | tip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK; | |
1870 | tip->i_d.di_flags2 |= f & XFS_DIFLAG2_REFLINK; | |
52bfcdd7 DW |
1871 | } |
1872 | ||
1873 | /* Swap the cow forks. */ | |
1874 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { | |
52bfcdd7 DW |
1875 | ASSERT(ip->i_cformat == XFS_DINODE_FMT_EXTENTS); |
1876 | ASSERT(tip->i_cformat == XFS_DINODE_FMT_EXTENTS); | |
1877 | ||
897992b7 GS |
1878 | swap(ip->i_cnextents, tip->i_cnextents); |
1879 | swap(ip->i_cowfp, tip->i_cowfp); | |
52bfcdd7 | 1880 | |
5bcffe30 | 1881 | if (ip->i_cowfp && ip->i_cowfp->if_bytes) |
52bfcdd7 DW |
1882 | xfs_inode_set_cowblocks_tag(ip); |
1883 | else | |
1884 | xfs_inode_clear_cowblocks_tag(ip); | |
5bcffe30 | 1885 | if (tip->i_cowfp && tip->i_cowfp->if_bytes) |
52bfcdd7 DW |
1886 | xfs_inode_set_cowblocks_tag(tip); |
1887 | else | |
1888 | xfs_inode_clear_cowblocks_tag(tip); | |
f0bc4d13 DW |
1889 | } |
1890 | ||
a133d952 DC |
1891 | xfs_trans_log_inode(tp, ip, src_log_flags); |
1892 | xfs_trans_log_inode(tp, tip, target_log_flags); | |
1893 | ||
6fb10d6d BF |
1894 | /* |
1895 | * The extent forks have been swapped, but crc=1,rmapbt=0 filesystems | |
1896 | * have inode number owner values in the bmbt blocks that still refer to | |
1897 | * the old inode. Scan each bmbt to fix up the owner values with the | |
1898 | * inode number of the current inode. | |
1899 | */ | |
1900 | if (src_log_flags & XFS_ILOG_DOWNER) { | |
2dd3d709 | 1901 | error = xfs_swap_change_owner(&tp, ip, tip); |
6fb10d6d BF |
1902 | if (error) |
1903 | goto out_trans_cancel; | |
1904 | } | |
1905 | if (target_log_flags & XFS_ILOG_DOWNER) { | |
2dd3d709 | 1906 | error = xfs_swap_change_owner(&tp, tip, ip); |
6fb10d6d BF |
1907 | if (error) |
1908 | goto out_trans_cancel; | |
1909 | } | |
1910 | ||
a133d952 DC |
1911 | /* |
1912 | * If this is a synchronous mount, make sure that the | |
1913 | * transaction goes to disk before returning to the user. | |
1914 | */ | |
1915 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
1916 | xfs_trans_set_sync(tp); | |
1917 | ||
70393313 | 1918 | error = xfs_trans_commit(tp); |
a133d952 DC |
1919 | |
1920 | trace_xfs_swap_extent_after(ip, 0); | |
1921 | trace_xfs_swap_extent_after(tip, 1); | |
a133d952 | 1922 | |
65523218 | 1923 | out_unlock: |
81217683 DC |
1924 | xfs_iunlock(ip, lock_flags); |
1925 | xfs_iunlock(tip, lock_flags); | |
65523218 | 1926 | unlock_two_nondirectories(VFS_I(ip), VFS_I(tip)); |
39aff5fd | 1927 | return error; |
a133d952 DC |
1928 | |
1929 | out_trans_cancel: | |
4906e215 | 1930 | xfs_trans_cancel(tp); |
65523218 | 1931 | goto out_unlock; |
a133d952 | 1932 | } |