]>
Commit | Line | Data |
---|---|---|
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" |
1da177e4 | 19 | #include "xfs_fs.h" |
a844f451 | 20 | #include "xfs_bit.h" |
1da177e4 | 21 | #include "xfs_log.h" |
a844f451 | 22 | #include "xfs_inum.h" |
1da177e4 LT |
23 | #include "xfs_trans.h" |
24 | #include "xfs_sb.h" | |
25 | #include "xfs_ag.h" | |
1da177e4 | 26 | #include "xfs_alloc.h" |
1da177e4 LT |
27 | #include "xfs_quota.h" |
28 | #include "xfs_mount.h" | |
1da177e4 | 29 | #include "xfs_bmap_btree.h" |
a844f451 | 30 | #include "xfs_alloc_btree.h" |
1da177e4 | 31 | #include "xfs_ialloc_btree.h" |
1da177e4 LT |
32 | #include "xfs_dinode.h" |
33 | #include "xfs_inode.h" | |
84803fb7 | 34 | #include "xfs_inode_item.h" |
a844f451 | 35 | #include "xfs_btree.h" |
1da177e4 | 36 | #include "xfs_bmap.h" |
1da177e4 LT |
37 | #include "xfs_rtalloc.h" |
38 | #include "xfs_error.h" | |
39 | #include "xfs_itable.h" | |
40 | #include "xfs_rw.h" | |
1da177e4 LT |
41 | #include "xfs_attr.h" |
42 | #include "xfs_buf_item.h" | |
43 | #include "xfs_trans_space.h" | |
44 | #include "xfs_utils.h" | |
45 | #include "xfs_iomap.h" | |
0b1b213f | 46 | #include "xfs_trace.h" |
1da177e4 | 47 | |
1da177e4 LT |
48 | |
49 | #define XFS_WRITEIO_ALIGN(mp,off) (((off) >> mp->m_writeio_log) \ | |
50 | << mp->m_writeio_log) | |
1da177e4 LT |
51 | #define XFS_WRITE_IMAPS XFS_BMAP_MAX_NMAP |
52 | ||
dd9f438e NS |
53 | STATIC int |
54 | xfs_iomap_eof_align_last_fsb( | |
55 | xfs_mount_t *mp, | |
541d7d3c | 56 | xfs_inode_t *ip, |
dd9f438e NS |
57 | xfs_extlen_t extsize, |
58 | xfs_fileoff_t *last_fsb) | |
59 | { | |
60 | xfs_fileoff_t new_last_fsb = 0; | |
bf322d98 | 61 | xfs_extlen_t align = 0; |
dd9f438e NS |
62 | int eof, error; |
63 | ||
bf322d98 CH |
64 | if (!XFS_IS_REALTIME_INODE(ip)) { |
65 | /* | |
66 | * Round up the allocation request to a stripe unit | |
67 | * (m_dalign) boundary if the file size is >= stripe unit | |
68 | * size, and we are allocating past the allocation eof. | |
69 | * | |
70 | * If mounted with the "-o swalloc" option the alignment is | |
71 | * increased from the strip unit size to the stripe width. | |
72 | */ | |
73 | if (mp->m_swidth && (mp->m_flags & XFS_MOUNT_SWALLOC)) | |
74 | align = mp->m_swidth; | |
75 | else if (mp->m_dalign) | |
76 | align = mp->m_dalign; | |
77 | ||
ce7ae151 | 78 | if (align && XFS_ISIZE(ip) >= XFS_FSB_TO_B(mp, align)) |
bf322d98 CH |
79 | new_last_fsb = roundup_64(*last_fsb, align); |
80 | } | |
dd9f438e NS |
81 | |
82 | /* | |
83 | * Always round up the allocation request to an extent boundary | |
84 | * (when file on a real-time subvolume or has di_extsize hint). | |
85 | */ | |
86 | if (extsize) { | |
87 | if (new_last_fsb) | |
88 | align = roundup_64(new_last_fsb, extsize); | |
89 | else | |
90 | align = extsize; | |
91 | new_last_fsb = roundup_64(*last_fsb, align); | |
92 | } | |
93 | ||
94 | if (new_last_fsb) { | |
541d7d3c | 95 | error = xfs_bmap_eof(ip, new_last_fsb, XFS_DATA_FORK, &eof); |
dd9f438e NS |
96 | if (error) |
97 | return error; | |
98 | if (eof) | |
99 | *last_fsb = new_last_fsb; | |
100 | } | |
101 | return 0; | |
102 | } | |
103 | ||
572d95f4 | 104 | STATIC int |
6d4a8ecb | 105 | xfs_alert_fsblock_zero( |
572d95f4 NS |
106 | xfs_inode_t *ip, |
107 | xfs_bmbt_irec_t *imap) | |
108 | { | |
6a19d939 | 109 | xfs_alert_tag(ip->i_mount, XFS_PTAG_FSBLOCK_ZERO, |
572d95f4 NS |
110 | "Access to block zero in inode %llu " |
111 | "start_block: %llx start_off: %llx " | |
112 | "blkcnt: %llx extent-state: %x\n", | |
113 | (unsigned long long)ip->i_ino, | |
114 | (unsigned long long)imap->br_startblock, | |
115 | (unsigned long long)imap->br_startoff, | |
116 | (unsigned long long)imap->br_blockcount, | |
117 | imap->br_state); | |
118 | return EFSCORRUPTED; | |
119 | } | |
120 | ||
a206c817 | 121 | int |
1da177e4 LT |
122 | xfs_iomap_write_direct( |
123 | xfs_inode_t *ip, | |
f403b7f4 | 124 | xfs_off_t offset, |
1da177e4 | 125 | size_t count, |
3070451e | 126 | xfs_bmbt_irec_t *imap, |
405f8042 | 127 | int nmaps) |
1da177e4 LT |
128 | { |
129 | xfs_mount_t *mp = ip->i_mount; | |
1da177e4 LT |
130 | xfs_fileoff_t offset_fsb; |
131 | xfs_fileoff_t last_fsb; | |
dd9f438e | 132 | xfs_filblks_t count_fsb, resaligned; |
1da177e4 | 133 | xfs_fsblock_t firstfsb; |
dd9f438e | 134 | xfs_extlen_t extsz, temp; |
0116d935 | 135 | int nimaps; |
1da177e4 | 136 | int bmapi_flag; |
06d10dd9 | 137 | int quota_flag; |
1da177e4 LT |
138 | int rt; |
139 | xfs_trans_t *tp; | |
1da177e4 | 140 | xfs_bmap_free_t free_list; |
dd9f438e | 141 | uint qblocks, resblks, resrtextents; |
1da177e4 | 142 | int committed; |
dd9f438e | 143 | int error; |
1da177e4 | 144 | |
507630b2 | 145 | error = xfs_qm_dqattach(ip, 0); |
1da177e4 LT |
146 | if (error) |
147 | return XFS_ERROR(error); | |
148 | ||
dd9f438e | 149 | rt = XFS_IS_REALTIME_INODE(ip); |
957d0ebe | 150 | extsz = xfs_get_extsz_hint(ip); |
1da177e4 | 151 | |
957d0ebe DC |
152 | offset_fsb = XFS_B_TO_FSBT(mp, offset); |
153 | last_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)(offset + count))); | |
ce7ae151 | 154 | if ((offset + count) > XFS_ISIZE(ip)) { |
9f6c92b9 | 155 | error = xfs_iomap_eof_align_last_fsb(mp, ip, extsz, &last_fsb); |
dd9f438e | 156 | if (error) |
507630b2 | 157 | return XFS_ERROR(error); |
1da177e4 | 158 | } else { |
405f8042 | 159 | if (nmaps && (imap->br_startblock == HOLESTARTBLOCK)) |
dd9f438e | 160 | last_fsb = MIN(last_fsb, (xfs_fileoff_t) |
3070451e CH |
161 | imap->br_blockcount + |
162 | imap->br_startoff); | |
1da177e4 | 163 | } |
dd9f438e NS |
164 | count_fsb = last_fsb - offset_fsb; |
165 | ASSERT(count_fsb > 0); | |
166 | ||
167 | resaligned = count_fsb; | |
168 | if (unlikely(extsz)) { | |
169 | if ((temp = do_mod(offset_fsb, extsz))) | |
170 | resaligned += temp; | |
171 | if ((temp = do_mod(resaligned, extsz))) | |
172 | resaligned += extsz - temp; | |
173 | } | |
174 | ||
175 | if (unlikely(rt)) { | |
176 | resrtextents = qblocks = resaligned; | |
177 | resrtextents /= mp->m_sb.sb_rextsize; | |
84e1e99f DC |
178 | resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0); |
179 | quota_flag = XFS_QMOPT_RES_RTBLKS; | |
180 | } else { | |
181 | resrtextents = 0; | |
dd9f438e | 182 | resblks = qblocks = XFS_DIOSTRAT_SPACE_RES(mp, resaligned); |
84e1e99f DC |
183 | quota_flag = XFS_QMOPT_RES_REGBLKS; |
184 | } | |
1da177e4 LT |
185 | |
186 | /* | |
06d10dd9 | 187 | * Allocate and setup the transaction |
1da177e4 | 188 | */ |
1da177e4 | 189 | tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT); |
1da177e4 | 190 | error = xfs_trans_reserve(tp, resblks, |
d52b44d0 | 191 | XFS_WRITE_LOG_RES(mp), resrtextents, |
1da177e4 LT |
192 | XFS_TRANS_PERM_LOG_RES, |
193 | XFS_WRITE_LOG_COUNT); | |
1da177e4 | 194 | /* |
06d10dd9 | 195 | * Check for running out of space, note: need lock to return |
1da177e4 | 196 | */ |
507630b2 | 197 | if (error) { |
1da177e4 | 198 | xfs_trans_cancel(tp, 0); |
507630b2 DC |
199 | return XFS_ERROR(error); |
200 | } | |
201 | ||
1da177e4 | 202 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
1da177e4 | 203 | |
7d095257 | 204 | error = xfs_trans_reserve_quota_nblks(tp, ip, qblocks, 0, quota_flag); |
dd9f438e | 205 | if (error) |
507630b2 | 206 | goto out_trans_cancel; |
1da177e4 | 207 | |
ddc3415a | 208 | xfs_trans_ijoin(tp, ip, 0); |
1da177e4 | 209 | |
c0dc7828 | 210 | bmapi_flag = 0; |
ce7ae151 | 211 | if (offset < XFS_ISIZE(ip) || extsz) |
1da177e4 LT |
212 | bmapi_flag |= XFS_BMAPI_PREALLOC; |
213 | ||
214 | /* | |
3070451e CH |
215 | * From this point onwards we overwrite the imap pointer that the |
216 | * caller gave to us. | |
1da177e4 | 217 | */ |
9d87c319 | 218 | xfs_bmap_init(&free_list, &firstfsb); |
06d10dd9 | 219 | nimaps = 1; |
c0dc7828 DC |
220 | error = xfs_bmapi_write(tp, ip, offset_fsb, count_fsb, bmapi_flag, |
221 | &firstfsb, 0, imap, &nimaps, &free_list); | |
06d10dd9 | 222 | if (error) |
507630b2 | 223 | goto out_bmap_cancel; |
1da177e4 LT |
224 | |
225 | /* | |
06d10dd9 | 226 | * Complete the transaction |
1da177e4 | 227 | */ |
f7c99b6f | 228 | error = xfs_bmap_finish(&tp, &free_list, &committed); |
06d10dd9 | 229 | if (error) |
507630b2 | 230 | goto out_bmap_cancel; |
1c72bf90 | 231 | error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); |
06d10dd9 | 232 | if (error) |
507630b2 | 233 | goto out_unlock; |
1da177e4 | 234 | |
06d10dd9 NS |
235 | /* |
236 | * Copy any maps to caller's array and return any error. | |
237 | */ | |
1da177e4 | 238 | if (nimaps == 0) { |
507630b2 DC |
239 | error = XFS_ERROR(ENOSPC); |
240 | goto out_unlock; | |
572d95f4 NS |
241 | } |
242 | ||
507630b2 | 243 | if (!(imap->br_startblock || XFS_IS_REALTIME_INODE(ip))) |
6d4a8ecb | 244 | error = xfs_alert_fsblock_zero(ip, imap); |
1da177e4 | 245 | |
507630b2 DC |
246 | out_unlock: |
247 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
248 | return error; | |
1da177e4 | 249 | |
507630b2 | 250 | out_bmap_cancel: |
1da177e4 | 251 | xfs_bmap_cancel(&free_list); |
7d095257 | 252 | xfs_trans_unreserve_quota_nblks(tp, ip, qblocks, 0, quota_flag); |
507630b2 | 253 | out_trans_cancel: |
1da177e4 | 254 | xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT); |
507630b2 | 255 | goto out_unlock; |
1da177e4 LT |
256 | } |
257 | ||
dd9f438e | 258 | /* |
8de2bf93 DC |
259 | * If the caller is doing a write at the end of the file, then extend the |
260 | * allocation out to the file system's write iosize. We clean up any extra | |
261 | * space left over when the file is closed in xfs_inactive(). | |
055388a3 DC |
262 | * |
263 | * If we find we already have delalloc preallocation beyond EOF, don't do more | |
264 | * preallocation as it it not needed. | |
dd9f438e NS |
265 | */ |
266 | STATIC int | |
267 | xfs_iomap_eof_want_preallocate( | |
268 | xfs_mount_t *mp, | |
541d7d3c | 269 | xfs_inode_t *ip, |
dd9f438e NS |
270 | xfs_off_t offset, |
271 | size_t count, | |
dd9f438e NS |
272 | xfs_bmbt_irec_t *imap, |
273 | int nimaps, | |
274 | int *prealloc) | |
275 | { | |
276 | xfs_fileoff_t start_fsb; | |
277 | xfs_filblks_t count_fsb; | |
278 | xfs_fsblock_t firstblock; | |
279 | int n, error, imaps; | |
055388a3 | 280 | int found_delalloc = 0; |
dd9f438e NS |
281 | |
282 | *prealloc = 0; | |
ce7ae151 | 283 | if (offset + count <= XFS_ISIZE(ip)) |
dd9f438e NS |
284 | return 0; |
285 | ||
286 | /* | |
287 | * If there are any real blocks past eof, then don't | |
288 | * do any speculative allocation. | |
289 | */ | |
290 | start_fsb = XFS_B_TO_FSBT(mp, ((xfs_ufsize_t)(offset + count - 1))); | |
291 | count_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp)); | |
292 | while (count_fsb > 0) { | |
293 | imaps = nimaps; | |
3ddb8fa9 | 294 | firstblock = NULLFSBLOCK; |
5c8ed202 DC |
295 | error = xfs_bmapi_read(ip, start_fsb, count_fsb, imap, &imaps, |
296 | 0); | |
dd9f438e NS |
297 | if (error) |
298 | return error; | |
299 | for (n = 0; n < imaps; n++) { | |
300 | if ((imap[n].br_startblock != HOLESTARTBLOCK) && | |
301 | (imap[n].br_startblock != DELAYSTARTBLOCK)) | |
302 | return 0; | |
303 | start_fsb += imap[n].br_blockcount; | |
304 | count_fsb -= imap[n].br_blockcount; | |
055388a3 DC |
305 | |
306 | if (imap[n].br_startblock == DELAYSTARTBLOCK) | |
307 | found_delalloc = 1; | |
dd9f438e NS |
308 | } |
309 | } | |
055388a3 DC |
310 | if (!found_delalloc) |
311 | *prealloc = 1; | |
dd9f438e NS |
312 | return 0; |
313 | } | |
314 | ||
055388a3 DC |
315 | /* |
316 | * If we don't have a user specified preallocation size, dynamically increase | |
317 | * the preallocation size as the size of the file grows. Cap the maximum size | |
318 | * at a single extent or less if the filesystem is near full. The closer the | |
319 | * filesystem is to full, the smaller the maximum prealocation. | |
320 | */ | |
321 | STATIC xfs_fsblock_t | |
322 | xfs_iomap_prealloc_size( | |
323 | struct xfs_mount *mp, | |
324 | struct xfs_inode *ip) | |
325 | { | |
326 | xfs_fsblock_t alloc_blocks = 0; | |
327 | ||
328 | if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) { | |
329 | int shift = 0; | |
330 | int64_t freesp; | |
331 | ||
b8fc8263 DC |
332 | /* |
333 | * rounddown_pow_of_two() returns an undefined result | |
334 | * if we pass in alloc_blocks = 0. Hence the "+ 1" to | |
335 | * ensure we always pass in a non-zero value. | |
336 | */ | |
ce7ae151 | 337 | alloc_blocks = XFS_B_TO_FSB(mp, XFS_ISIZE(ip)) + 1; |
055388a3 DC |
338 | alloc_blocks = XFS_FILEOFF_MIN(MAXEXTLEN, |
339 | rounddown_pow_of_two(alloc_blocks)); | |
340 | ||
341 | xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); | |
342 | freesp = mp->m_sb.sb_fdblocks; | |
343 | if (freesp < mp->m_low_space[XFS_LOWSP_5_PCNT]) { | |
344 | shift = 2; | |
345 | if (freesp < mp->m_low_space[XFS_LOWSP_4_PCNT]) | |
346 | shift++; | |
347 | if (freesp < mp->m_low_space[XFS_LOWSP_3_PCNT]) | |
348 | shift++; | |
349 | if (freesp < mp->m_low_space[XFS_LOWSP_2_PCNT]) | |
350 | shift++; | |
351 | if (freesp < mp->m_low_space[XFS_LOWSP_1_PCNT]) | |
352 | shift++; | |
353 | } | |
354 | if (shift) | |
355 | alloc_blocks >>= shift; | |
356 | } | |
357 | ||
358 | if (alloc_blocks < mp->m_writeio_blocks) | |
359 | alloc_blocks = mp->m_writeio_blocks; | |
360 | ||
361 | return alloc_blocks; | |
362 | } | |
363 | ||
a206c817 | 364 | int |
1da177e4 LT |
365 | xfs_iomap_write_delay( |
366 | xfs_inode_t *ip, | |
f403b7f4 | 367 | xfs_off_t offset, |
1da177e4 | 368 | size_t count, |
405f8042 | 369 | xfs_bmbt_irec_t *ret_imap) |
1da177e4 LT |
370 | { |
371 | xfs_mount_t *mp = ip->i_mount; | |
1da177e4 LT |
372 | xfs_fileoff_t offset_fsb; |
373 | xfs_fileoff_t last_fsb; | |
dd9f438e NS |
374 | xfs_off_t aligned_offset; |
375 | xfs_fileoff_t ioalign; | |
dd9f438e | 376 | xfs_extlen_t extsz; |
1da177e4 | 377 | int nimaps; |
1da177e4 | 378 | xfs_bmbt_irec_t imap[XFS_WRITE_IMAPS]; |
8de2bf93 | 379 | int prealloc, flushed = 0; |
dd9f438e | 380 | int error; |
1da177e4 | 381 | |
579aa9ca | 382 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
1da177e4 LT |
383 | |
384 | /* | |
385 | * Make sure that the dquots are there. This doesn't hold | |
386 | * the ilock across a disk read. | |
387 | */ | |
7d095257 | 388 | error = xfs_qm_dqattach_locked(ip, 0); |
1da177e4 LT |
389 | if (error) |
390 | return XFS_ERROR(error); | |
391 | ||
957d0ebe | 392 | extsz = xfs_get_extsz_hint(ip); |
dd9f438e NS |
393 | offset_fsb = XFS_B_TO_FSBT(mp, offset); |
394 | ||
055388a3 | 395 | |
9f6c92b9 | 396 | error = xfs_iomap_eof_want_preallocate(mp, ip, offset, count, |
405f8042 | 397 | imap, XFS_WRITE_IMAPS, &prealloc); |
dd9f438e NS |
398 | if (error) |
399 | return error; | |
1da177e4 | 400 | |
8de2bf93 | 401 | retry: |
dd9f438e | 402 | if (prealloc) { |
055388a3 DC |
403 | xfs_fsblock_t alloc_blocks = xfs_iomap_prealloc_size(mp, ip); |
404 | ||
1da177e4 LT |
405 | aligned_offset = XFS_WRITEIO_ALIGN(mp, (offset + count - 1)); |
406 | ioalign = XFS_B_TO_FSBT(mp, aligned_offset); | |
055388a3 | 407 | last_fsb = ioalign + alloc_blocks; |
dd9f438e NS |
408 | } else { |
409 | last_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)(offset + count))); | |
1da177e4 | 410 | } |
1da177e4 | 411 | |
dd9f438e | 412 | if (prealloc || extsz) { |
9f6c92b9 | 413 | error = xfs_iomap_eof_align_last_fsb(mp, ip, extsz, &last_fsb); |
dd9f438e | 414 | if (error) |
1da177e4 | 415 | return error; |
1da177e4 | 416 | } |
dd9f438e NS |
417 | |
418 | nimaps = XFS_WRITE_IMAPS; | |
4403280a CH |
419 | error = xfs_bmapi_delay(ip, offset_fsb, last_fsb - offset_fsb, |
420 | imap, &nimaps, XFS_BMAPI_ENTIRE); | |
055388a3 DC |
421 | switch (error) { |
422 | case 0: | |
423 | case ENOSPC: | |
424 | case EDQUOT: | |
425 | break; | |
426 | default: | |
1da177e4 | 427 | return XFS_ERROR(error); |
055388a3 | 428 | } |
dd9f438e | 429 | |
1da177e4 | 430 | /* |
055388a3 DC |
431 | * If bmapi returned us nothing, we got either ENOSPC or EDQUOT. For |
432 | * ENOSPC, * flush all other inodes with delalloc blocks to free up | |
433 | * some of the excess reserved metadata space. For both cases, retry | |
434 | * without EOF preallocation. | |
1da177e4 LT |
435 | */ |
436 | if (nimaps == 0) { | |
0b1b213f | 437 | trace_xfs_delalloc_enospc(ip, offset, count); |
8de2bf93 | 438 | if (flushed) |
055388a3 | 439 | return XFS_ERROR(error ? error : ENOSPC); |
1da177e4 | 440 | |
055388a3 DC |
441 | if (error == ENOSPC) { |
442 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
443 | xfs_flush_inodes(ip); | |
444 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
445 | } | |
8de2bf93 DC |
446 | |
447 | flushed = 1; | |
1da177e4 | 448 | error = 0; |
8de2bf93 | 449 | prealloc = 0; |
1da177e4 LT |
450 | goto retry; |
451 | } | |
452 | ||
86c4d623 | 453 | if (!(imap[0].br_startblock || XFS_IS_REALTIME_INODE(ip))) |
6d4a8ecb | 454 | return xfs_alert_fsblock_zero(ip, &imap[0]); |
dd9f438e NS |
455 | |
456 | *ret_imap = imap[0]; | |
1da177e4 LT |
457 | return 0; |
458 | } | |
459 | ||
460 | /* | |
461 | * Pass in a delayed allocate extent, convert it to real extents; | |
462 | * return to the caller the extent we create which maps on top of | |
463 | * the originating callers request. | |
464 | * | |
465 | * Called without a lock on the inode. | |
e4143a1c DC |
466 | * |
467 | * We no longer bother to look at the incoming map - all we have to | |
468 | * guarantee is that whatever we allocate fills the required range. | |
1da177e4 | 469 | */ |
a206c817 | 470 | int |
1da177e4 LT |
471 | xfs_iomap_write_allocate( |
472 | xfs_inode_t *ip, | |
f403b7f4 | 473 | xfs_off_t offset, |
24e17b5f | 474 | size_t count, |
405f8042 | 475 | xfs_bmbt_irec_t *imap) |
1da177e4 LT |
476 | { |
477 | xfs_mount_t *mp = ip->i_mount; | |
1da177e4 LT |
478 | xfs_fileoff_t offset_fsb, last_block; |
479 | xfs_fileoff_t end_fsb, map_start_fsb; | |
480 | xfs_fsblock_t first_block; | |
481 | xfs_bmap_free_t free_list; | |
482 | xfs_filblks_t count_fsb; | |
1da177e4 | 483 | xfs_trans_t *tp; |
e4143a1c | 484 | int nimaps, committed; |
1da177e4 LT |
485 | int error = 0; |
486 | int nres; | |
487 | ||
1da177e4 LT |
488 | /* |
489 | * Make sure that the dquots are there. | |
490 | */ | |
7d095257 CH |
491 | error = xfs_qm_dqattach(ip, 0); |
492 | if (error) | |
1da177e4 LT |
493 | return XFS_ERROR(error); |
494 | ||
24e17b5f | 495 | offset_fsb = XFS_B_TO_FSBT(mp, offset); |
3070451e CH |
496 | count_fsb = imap->br_blockcount; |
497 | map_start_fsb = imap->br_startoff; | |
1da177e4 LT |
498 | |
499 | XFS_STATS_ADD(xs_xstrat_bytes, XFS_FSB_TO_B(mp, count_fsb)); | |
500 | ||
501 | while (count_fsb != 0) { | |
502 | /* | |
503 | * Set up a transaction with which to allocate the | |
504 | * backing store for the file. Do allocations in a | |
505 | * loop until we get some space in the range we are | |
506 | * interested in. The other space that might be allocated | |
507 | * is in the delayed allocation extent on which we sit | |
508 | * but before our buffer starts. | |
509 | */ | |
510 | ||
511 | nimaps = 0; | |
512 | while (nimaps == 0) { | |
513 | tp = xfs_trans_alloc(mp, XFS_TRANS_STRAT_WRITE); | |
84e1e99f | 514 | tp->t_flags |= XFS_TRANS_RESERVE; |
1da177e4 LT |
515 | nres = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK); |
516 | error = xfs_trans_reserve(tp, nres, | |
517 | XFS_WRITE_LOG_RES(mp), | |
518 | 0, XFS_TRANS_PERM_LOG_RES, | |
519 | XFS_WRITE_LOG_COUNT); | |
1da177e4 LT |
520 | if (error) { |
521 | xfs_trans_cancel(tp, 0); | |
522 | return XFS_ERROR(error); | |
523 | } | |
524 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
ddc3415a | 525 | xfs_trans_ijoin(tp, ip, 0); |
1da177e4 | 526 | |
9d87c319 | 527 | xfs_bmap_init(&free_list, &first_block); |
1da177e4 | 528 | |
1da177e4 | 529 | /* |
e4143a1c DC |
530 | * it is possible that the extents have changed since |
531 | * we did the read call as we dropped the ilock for a | |
532 | * while. We have to be careful about truncates or hole | |
533 | * punchs here - we are not allowed to allocate | |
534 | * non-delalloc blocks here. | |
535 | * | |
536 | * The only protection against truncation is the pages | |
537 | * for the range we are being asked to convert are | |
538 | * locked and hence a truncate will block on them | |
539 | * first. | |
540 | * | |
541 | * As a result, if we go beyond the range we really | |
542 | * need and hit an delalloc extent boundary followed by | |
543 | * a hole while we have excess blocks in the map, we | |
544 | * will fill the hole incorrectly and overrun the | |
545 | * transaction reservation. | |
546 | * | |
547 | * Using a single map prevents this as we are forced to | |
548 | * check each map we look for overlap with the desired | |
549 | * range and abort as soon as we find it. Also, given | |
550 | * that we only return a single map, having one beyond | |
551 | * what we can return is probably a bit silly. | |
552 | * | |
553 | * We also need to check that we don't go beyond EOF; | |
554 | * this is a truncate optimisation as a truncate sets | |
555 | * the new file size before block on the pages we | |
556 | * currently have locked under writeback. Because they | |
557 | * are about to be tossed, we don't need to write them | |
558 | * back.... | |
1da177e4 | 559 | */ |
e4143a1c | 560 | nimaps = 1; |
ce7ae151 | 561 | end_fsb = XFS_B_TO_FSB(mp, XFS_ISIZE(ip)); |
7c9ef85c DC |
562 | error = xfs_bmap_last_offset(NULL, ip, &last_block, |
563 | XFS_DATA_FORK); | |
564 | if (error) | |
565 | goto trans_cancel; | |
566 | ||
1da177e4 LT |
567 | last_block = XFS_FILEOFF_MAX(last_block, end_fsb); |
568 | if ((map_start_fsb + count_fsb) > last_block) { | |
569 | count_fsb = last_block - map_start_fsb; | |
570 | if (count_fsb == 0) { | |
571 | error = EAGAIN; | |
572 | goto trans_cancel; | |
573 | } | |
574 | } | |
575 | ||
3070451e | 576 | /* |
3070451e CH |
577 | * From this point onwards we overwrite the imap |
578 | * pointer that the caller gave to us. | |
579 | */ | |
c0dc7828 DC |
580 | error = xfs_bmapi_write(tp, ip, map_start_fsb, |
581 | count_fsb, 0, &first_block, 1, | |
582 | imap, &nimaps, &free_list); | |
1da177e4 LT |
583 | if (error) |
584 | goto trans_cancel; | |
585 | ||
f7c99b6f | 586 | error = xfs_bmap_finish(&tp, &free_list, &committed); |
1da177e4 LT |
587 | if (error) |
588 | goto trans_cancel; | |
589 | ||
1c72bf90 | 590 | error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); |
1da177e4 LT |
591 | if (error) |
592 | goto error0; | |
593 | ||
594 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
595 | } | |
596 | ||
597 | /* | |
598 | * See if we were able to allocate an extent that | |
599 | * covers at least part of the callers request | |
600 | */ | |
3070451e | 601 | if (!(imap->br_startblock || XFS_IS_REALTIME_INODE(ip))) |
6d4a8ecb | 602 | return xfs_alert_fsblock_zero(ip, imap); |
86c4d623 | 603 | |
3070451e CH |
604 | if ((offset_fsb >= imap->br_startoff) && |
605 | (offset_fsb < (imap->br_startoff + | |
606 | imap->br_blockcount))) { | |
e4143a1c DC |
607 | XFS_STATS_INC(xs_xstrat_quick); |
608 | return 0; | |
1da177e4 LT |
609 | } |
610 | ||
e4143a1c DC |
611 | /* |
612 | * So far we have not mapped the requested part of the | |
1da177e4 LT |
613 | * file, just surrounding data, try again. |
614 | */ | |
3070451e CH |
615 | count_fsb -= imap->br_blockcount; |
616 | map_start_fsb = imap->br_startoff + imap->br_blockcount; | |
1da177e4 LT |
617 | } |
618 | ||
619 | trans_cancel: | |
620 | xfs_bmap_cancel(&free_list); | |
621 | xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT); | |
622 | error0: | |
623 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
624 | return XFS_ERROR(error); | |
625 | } | |
626 | ||
627 | int | |
628 | xfs_iomap_write_unwritten( | |
629 | xfs_inode_t *ip, | |
f403b7f4 | 630 | xfs_off_t offset, |
1da177e4 LT |
631 | size_t count) |
632 | { | |
633 | xfs_mount_t *mp = ip->i_mount; | |
1da177e4 LT |
634 | xfs_fileoff_t offset_fsb; |
635 | xfs_filblks_t count_fsb; | |
636 | xfs_filblks_t numblks_fsb; | |
dd9f438e NS |
637 | xfs_fsblock_t firstfsb; |
638 | int nimaps; | |
639 | xfs_trans_t *tp; | |
640 | xfs_bmbt_irec_t imap; | |
641 | xfs_bmap_free_t free_list; | |
84803fb7 | 642 | xfs_fsize_t i_size; |
dd9f438e | 643 | uint resblks; |
1da177e4 LT |
644 | int committed; |
645 | int error; | |
1da177e4 | 646 | |
0b1b213f | 647 | trace_xfs_unwritten_convert(ip, offset, count); |
1da177e4 LT |
648 | |
649 | offset_fsb = XFS_B_TO_FSBT(mp, offset); | |
650 | count_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + count); | |
651 | count_fsb = (xfs_filblks_t)(count_fsb - offset_fsb); | |
652 | ||
4ddd8bb1 LM |
653 | /* |
654 | * Reserve enough blocks in this transaction for two complete extent | |
655 | * btree splits. We may be converting the middle part of an unwritten | |
656 | * extent and in this case we will insert two new extents in the btree | |
657 | * each of which could cause a full split. | |
658 | * | |
659 | * This reservation amount will be used in the first call to | |
660 | * xfs_bmbt_split() to select an AG with enough space to satisfy the | |
661 | * rest of the operation. | |
662 | */ | |
dd9f438e | 663 | resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0) << 1; |
1da177e4 | 664 | |
dd9f438e | 665 | do { |
1da177e4 LT |
666 | /* |
667 | * set up a transaction to convert the range of extents | |
668 | * from unwritten to real. Do allocations in a loop until | |
669 | * we have covered the range passed in. | |
80641dc6 CH |
670 | * |
671 | * Note that we open code the transaction allocation here | |
672 | * to pass KM_NOFS--we can't risk to recursing back into | |
673 | * the filesystem here as we might be asked to write out | |
674 | * the same inode that we complete here and might deadlock | |
675 | * on the iolock. | |
1da177e4 | 676 | */ |
80641dc6 CH |
677 | xfs_wait_for_freeze(mp, SB_FREEZE_TRANS); |
678 | tp = _xfs_trans_alloc(mp, XFS_TRANS_STRAT_WRITE, KM_NOFS); | |
84e1e99f | 679 | tp->t_flags |= XFS_TRANS_RESERVE; |
dd9f438e | 680 | error = xfs_trans_reserve(tp, resblks, |
1da177e4 LT |
681 | XFS_WRITE_LOG_RES(mp), 0, |
682 | XFS_TRANS_PERM_LOG_RES, | |
683 | XFS_WRITE_LOG_COUNT); | |
684 | if (error) { | |
685 | xfs_trans_cancel(tp, 0); | |
572d95f4 | 686 | return XFS_ERROR(error); |
1da177e4 LT |
687 | } |
688 | ||
689 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
ddc3415a | 690 | xfs_trans_ijoin(tp, ip, 0); |
1da177e4 LT |
691 | |
692 | /* | |
693 | * Modify the unwritten extent state of the buffer. | |
694 | */ | |
9d87c319 | 695 | xfs_bmap_init(&free_list, &firstfsb); |
1da177e4 | 696 | nimaps = 1; |
c0dc7828 DC |
697 | error = xfs_bmapi_write(tp, ip, offset_fsb, count_fsb, |
698 | XFS_BMAPI_CONVERT, &firstfsb, | |
b4e9181e | 699 | 1, &imap, &nimaps, &free_list); |
1da177e4 LT |
700 | if (error) |
701 | goto error_on_bmapi_transaction; | |
702 | ||
84803fb7 CH |
703 | /* |
704 | * Log the updated inode size as we go. We have to be careful | |
705 | * to only log it up to the actual write offset if it is | |
706 | * halfway into a block. | |
707 | */ | |
708 | i_size = XFS_FSB_TO_B(mp, offset_fsb + count_fsb); | |
709 | if (i_size > offset + count) | |
710 | i_size = offset + count; | |
711 | ||
712 | i_size = xfs_new_eof(ip, i_size); | |
713 | if (i_size) { | |
714 | ip->i_d.di_size = i_size; | |
715 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
716 | } | |
717 | ||
718 | error = xfs_bmap_finish(&tp, &free_list, &committed); | |
1da177e4 LT |
719 | if (error) |
720 | goto error_on_bmapi_transaction; | |
721 | ||
1c72bf90 | 722 | error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); |
1da177e4 LT |
723 | xfs_iunlock(ip, XFS_ILOCK_EXCL); |
724 | if (error) | |
572d95f4 NS |
725 | return XFS_ERROR(error); |
726 | ||
86c4d623 | 727 | if (!(imap.br_startblock || XFS_IS_REALTIME_INODE(ip))) |
6d4a8ecb | 728 | return xfs_alert_fsblock_zero(ip, &imap); |
1da177e4 LT |
729 | |
730 | if ((numblks_fsb = imap.br_blockcount) == 0) { | |
731 | /* | |
732 | * The numblks_fsb value should always get | |
733 | * smaller, otherwise the loop is stuck. | |
734 | */ | |
735 | ASSERT(imap.br_blockcount); | |
736 | break; | |
737 | } | |
738 | offset_fsb += numblks_fsb; | |
739 | count_fsb -= numblks_fsb; | |
740 | } while (count_fsb > 0); | |
741 | ||
742 | return 0; | |
743 | ||
744 | error_on_bmapi_transaction: | |
745 | xfs_bmap_cancel(&free_list); | |
746 | xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT)); | |
747 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
1da177e4 LT |
748 | return XFS_ERROR(error); |
749 | } |