]>
Commit | Line | Data |
---|---|---|
2bd0ea18 | 1 | /* |
da23017d NS |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
2bd0ea18 | 4 | * |
da23017d 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 | |
2bd0ea18 NS |
7 | * published by the Free Software Foundation. |
8 | * | |
da23017d 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. | |
2bd0ea18 | 13 | * |
da23017d 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 | |
2bd0ea18 NS |
17 | */ |
18 | ||
9c799827 | 19 | #include "libxfs_priv.h" |
9542ae13 | 20 | #include "libxfs_io.h" |
b626fb59 DC |
21 | #include "init.h" |
22 | #include "xfs_fs.h" | |
23 | #include "xfs_shared.h" | |
24 | #include "xfs_format.h" | |
25 | #include "xfs_log_format.h" | |
26 | #include "xfs_trans_resv.h" | |
27 | #include "xfs_mount.h" | |
f944d3d0 | 28 | #include "xfs_defer.h" |
b626fb59 DC |
29 | #include "xfs_inode_buf.h" |
30 | #include "xfs_inode_fork.h" | |
31 | #include "xfs_inode.h" | |
32 | #include "xfs_trans.h" | |
33 | #include "xfs_bmap.h" | |
34 | #include "xfs_bmap_btree.h" | |
35 | #include "xfs_trans_space.h" | |
36 | #include "xfs_ialloc.h" | |
37 | #include "xfs_alloc.h" | |
9542ae13 | 38 | #include "xfs_bit.h" |
d15188a1 DW |
39 | #include "xfs_da_format.h" |
40 | #include "xfs_da_btree.h" | |
41 | #include "xfs_dir2_priv.h" | |
2bd0ea18 | 42 | |
88cd79be DC |
43 | /* |
44 | * Calculate the worst case log unit reservation for a given superblock | |
45 | * configuration. Copied and munged from the kernel code, and assumes a | |
46 | * worse case header usage (maximum log buffer sizes) | |
47 | */ | |
48 | int | |
49 | xfs_log_calc_unit_res( | |
50 | struct xfs_mount *mp, | |
51 | int unit_bytes) | |
52 | { | |
53 | int iclog_space; | |
54 | int iclog_header_size; | |
55 | int iclog_size; | |
56 | uint num_headers; | |
57 | ||
58 | if (xfs_sb_version_haslogv2(&mp->m_sb)) { | |
59 | iclog_size = XLOG_MAX_RECORD_BSIZE; | |
60 | iclog_header_size = BBTOB(iclog_size / XLOG_HEADER_CYCLE_SIZE); | |
61 | } else { | |
62 | iclog_size = XLOG_BIG_RECORD_BSIZE; | |
63 | iclog_header_size = BBSIZE; | |
64 | } | |
65 | ||
66 | /* | |
67 | * Permanent reservations have up to 'cnt'-1 active log operations | |
68 | * in the log. A unit in this case is the amount of space for one | |
69 | * of these log operations. Normal reservations have a cnt of 1 | |
70 | * and their unit amount is the total amount of space required. | |
71 | * | |
72 | * The following lines of code account for non-transaction data | |
73 | * which occupy space in the on-disk log. | |
74 | * | |
75 | * Normal form of a transaction is: | |
76 | * <oph><trans-hdr><start-oph><reg1-oph><reg1><reg2-oph>...<commit-oph> | |
77 | * and then there are LR hdrs, split-recs and roundoff at end of syncs. | |
78 | * | |
79 | * We need to account for all the leadup data and trailer data | |
80 | * around the transaction data. | |
81 | * And then we need to account for the worst case in terms of using | |
82 | * more space. | |
83 | * The worst case will happen if: | |
84 | * - the placement of the transaction happens to be such that the | |
85 | * roundoff is at its maximum | |
86 | * - the transaction data is synced before the commit record is synced | |
87 | * i.e. <transaction-data><roundoff> | <commit-rec><roundoff> | |
88 | * Therefore the commit record is in its own Log Record. | |
89 | * This can happen as the commit record is called with its | |
90 | * own region to xlog_write(). | |
91 | * This then means that in the worst case, roundoff can happen for | |
92 | * the commit-rec as well. | |
93 | * The commit-rec is smaller than padding in this scenario and so it is | |
94 | * not added separately. | |
95 | */ | |
96 | ||
97 | /* for trans header */ | |
98 | unit_bytes += sizeof(xlog_op_header_t); | |
99 | unit_bytes += sizeof(xfs_trans_header_t); | |
100 | ||
101 | /* for start-rec */ | |
102 | unit_bytes += sizeof(xlog_op_header_t); | |
103 | ||
104 | /* | |
105 | * for LR headers - the space for data in an iclog is the size minus | |
106 | * the space used for the headers. If we use the iclog size, then we | |
107 | * undercalculate the number of headers required. | |
108 | * | |
109 | * Furthermore - the addition of op headers for split-recs might | |
110 | * increase the space required enough to require more log and op | |
111 | * headers, so take that into account too. | |
112 | * | |
113 | * IMPORTANT: This reservation makes the assumption that if this | |
114 | * transaction is the first in an iclog and hence has the LR headers | |
115 | * accounted to it, then the remaining space in the iclog is | |
116 | * exclusively for this transaction. i.e. if the transaction is larger | |
117 | * than the iclog, it will be the only thing in that iclog. | |
118 | * Fundamentally, this means we must pass the entire log vector to | |
119 | * xlog_write to guarantee this. | |
120 | */ | |
121 | iclog_space = iclog_size - iclog_header_size; | |
122 | num_headers = howmany(unit_bytes, iclog_space); | |
123 | ||
124 | /* for split-recs - ophdrs added when data split over LRs */ | |
125 | unit_bytes += sizeof(xlog_op_header_t) * num_headers; | |
126 | ||
127 | /* add extra header reservations if we overrun */ | |
128 | while (!num_headers || | |
129 | howmany(unit_bytes, iclog_space) > num_headers) { | |
130 | unit_bytes += sizeof(xlog_op_header_t); | |
131 | num_headers++; | |
132 | } | |
133 | unit_bytes += iclog_header_size * num_headers; | |
134 | ||
135 | /* for commit-rec LR header - note: padding will subsume the ophdr */ | |
136 | unit_bytes += iclog_header_size; | |
137 | ||
138 | /* for roundoff padding for transaction data and one for commit record */ | |
139 | if (xfs_sb_version_haslogv2(&mp->m_sb) && mp->m_sb.sb_logsunit > 1) { | |
140 | /* log su roundoff */ | |
141 | unit_bytes += 2 * mp->m_sb.sb_logsunit; | |
142 | } else { | |
143 | /* BB roundoff */ | |
144 | unit_bytes += 2 * BBSIZE; | |
145 | } | |
146 | ||
147 | return unit_bytes; | |
148 | } | |
149 | ||
2bd0ea18 NS |
150 | /* |
151 | * Change the requested timestamp in the given inode. | |
5000d01d | 152 | * |
2bd0ea18 | 153 | * This was once shared with the kernel, but has diverged to the point |
ff1f79a7 | 154 | * where it's no longer worth the hassle of maintaining common code. |
2bd0ea18 NS |
155 | */ |
156 | void | |
56b2de80 DC |
157 | libxfs_trans_ichgtime( |
158 | struct xfs_trans *tp, | |
159 | struct xfs_inode *ip, | |
160 | int flags) | |
2bd0ea18 | 161 | { |
5000d01d | 162 | struct timespec tv; |
2bd0ea18 NS |
163 | struct timeval stv; |
164 | ||
165 | gettimeofday(&stv, (struct timezone *)0); | |
166 | tv.tv_sec = stv.tv_sec; | |
167 | tv.tv_nsec = stv.tv_usec * 1000; | |
1bc6cbe3 DC |
168 | if (flags & XFS_ICHGTIME_MOD) |
169 | VFS_I(ip)->i_mtime = tv; | |
170 | if (flags & XFS_ICHGTIME_CHG) | |
171 | VFS_I(ip)->i_ctime = tv; | |
41ce5f36 | 172 | if (flags & XFS_ICHGTIME_CREATE) { |
14f8b681 DW |
173 | ip->i_d.di_crtime.t_sec = (int32_t)tv.tv_sec; |
174 | ip->i_d.di_crtime.t_nsec = (int32_t)tv.tv_nsec; | |
41ce5f36 | 175 | } |
2bd0ea18 NS |
176 | } |
177 | ||
06b80354 DW |
178 | STATIC uint16_t |
179 | xfs_flags2diflags( | |
180 | struct xfs_inode *ip, | |
181 | unsigned int xflags) | |
182 | { | |
183 | /* can't set PREALLOC this way, just preserve it */ | |
184 | uint16_t di_flags = | |
185 | (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC); | |
186 | ||
187 | if (xflags & FS_XFLAG_IMMUTABLE) | |
188 | di_flags |= XFS_DIFLAG_IMMUTABLE; | |
189 | if (xflags & FS_XFLAG_APPEND) | |
190 | di_flags |= XFS_DIFLAG_APPEND; | |
191 | if (xflags & FS_XFLAG_SYNC) | |
192 | di_flags |= XFS_DIFLAG_SYNC; | |
193 | if (xflags & FS_XFLAG_NOATIME) | |
194 | di_flags |= XFS_DIFLAG_NOATIME; | |
195 | if (xflags & FS_XFLAG_NODUMP) | |
196 | di_flags |= XFS_DIFLAG_NODUMP; | |
197 | if (xflags & FS_XFLAG_NODEFRAG) | |
198 | di_flags |= XFS_DIFLAG_NODEFRAG; | |
199 | if (xflags & FS_XFLAG_FILESTREAM) | |
200 | di_flags |= XFS_DIFLAG_FILESTREAM; | |
201 | if (S_ISDIR(VFS_I(ip)->i_mode)) { | |
202 | if (xflags & FS_XFLAG_RTINHERIT) | |
203 | di_flags |= XFS_DIFLAG_RTINHERIT; | |
204 | if (xflags & FS_XFLAG_NOSYMLINKS) | |
205 | di_flags |= XFS_DIFLAG_NOSYMLINKS; | |
206 | if (xflags & FS_XFLAG_EXTSZINHERIT) | |
207 | di_flags |= XFS_DIFLAG_EXTSZINHERIT; | |
208 | if (xflags & FS_XFLAG_PROJINHERIT) | |
209 | di_flags |= XFS_DIFLAG_PROJINHERIT; | |
210 | } else if (S_ISREG(VFS_I(ip)->i_mode)) { | |
211 | if (xflags & FS_XFLAG_REALTIME) | |
212 | di_flags |= XFS_DIFLAG_REALTIME; | |
213 | if (xflags & FS_XFLAG_EXTSIZE) | |
214 | di_flags |= XFS_DIFLAG_EXTSIZE; | |
215 | } | |
216 | ||
217 | return di_flags; | |
218 | } | |
219 | ||
220 | STATIC uint64_t | |
221 | xfs_flags2diflags2( | |
222 | struct xfs_inode *ip, | |
223 | unsigned int xflags) | |
224 | { | |
225 | uint64_t di_flags2 = | |
226 | (ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK); | |
227 | ||
228 | if (xflags & FS_XFLAG_DAX) | |
229 | di_flags2 |= XFS_DIFLAG2_DAX; | |
230 | if (xflags & FS_XFLAG_COWEXTSIZE) | |
231 | di_flags2 |= XFS_DIFLAG2_COWEXTSIZE; | |
232 | ||
233 | return di_flags2; | |
234 | } | |
235 | ||
2bd0ea18 | 236 | /* |
ff1f79a7 | 237 | * Allocate an inode on disk and return a copy of its in-core version. |
2bd0ea18 NS |
238 | * Set mode, nlink, and rdev appropriately within the inode. |
239 | * The uid and gid for the inode are set according to the contents of | |
240 | * the given cred structure. | |
241 | * | |
242 | * This was once shared with the kernel, but has diverged to the point | |
ff1f79a7 | 243 | * where it's no longer worth the hassle of maintaining common code. |
2bd0ea18 | 244 | */ |
5e656dbb | 245 | int |
2bd0ea18 NS |
246 | libxfs_ialloc( |
247 | xfs_trans_t *tp, | |
248 | xfs_inode_t *pip, | |
249 | mode_t mode, | |
250 | nlink_t nlink, | |
63899e27 | 251 | xfs_dev_t rdev, |
9f064b7e NS |
252 | struct cred *cr, |
253 | struct fsxattr *fsx, | |
2bd0ea18 NS |
254 | int okalloc, |
255 | xfs_buf_t **ialloc_context, | |
2bd0ea18 NS |
256 | xfs_inode_t **ipp) |
257 | { | |
258 | xfs_ino_t ino; | |
259 | xfs_inode_t *ip; | |
260 | uint flags; | |
261 | int error; | |
262 | ||
263 | /* | |
264 | * Call the space management code to pick | |
265 | * the on-disk inode to be allocated. | |
266 | */ | |
267 | error = xfs_dialloc(tp, pip ? pip->i_ino : 0, mode, okalloc, | |
a2ceac1f | 268 | ialloc_context, &ino); |
2bd0ea18 NS |
269 | if (error != 0) |
270 | return error; | |
a2ceac1f | 271 | if (*ialloc_context || ino == NULLFSINO) { |
2bd0ea18 NS |
272 | *ipp = NULL; |
273 | return 0; | |
274 | } | |
275 | ASSERT(*ialloc_context == NULL); | |
276 | ||
46eca962 | 277 | error = xfs_trans_iget(tp->t_mountp, tp, ino, 0, 0, &ip); |
2bd0ea18 NS |
278 | if (error != 0) |
279 | return error; | |
280 | ASSERT(ip != NULL); | |
281 | ||
e37bf53c | 282 | VFS_I(ip)->i_mode = mode; |
bcbe04c1 | 283 | set_nlink(VFS_I(ip), nlink); |
2bd0ea18 NS |
284 | ip->i_d.di_uid = cr->cr_uid; |
285 | ip->i_d.di_gid = cr->cr_gid; | |
22bc10ed | 286 | xfs_set_projid(&ip->i_d, pip ? 0 : fsx->fsx_projid); |
41ce5f36 | 287 | xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG | XFS_ICHGTIME_MOD); |
2bd0ea18 NS |
288 | |
289 | /* | |
ff105f75 DC |
290 | * We only support filesystems that understand v2 format inodes. So if |
291 | * this is currently an old format inode, then change the inode version | |
292 | * number now. This way we only do the conversion here rather than here | |
293 | * and in the flush/logging code. | |
2bd0ea18 | 294 | */ |
ff105f75 | 295 | if (ip->i_d.di_version == 1) { |
56b2de80 | 296 | ip->i_d.di_version = 2; |
22bc10ed AM |
297 | /* |
298 | * old link count, projid_lo/hi field, pad field | |
299 | * already zeroed | |
300 | */ | |
5000d01d | 301 | } |
2bd0ea18 | 302 | |
e37bf53c | 303 | if (pip && (VFS_I(pip)->i_mode & S_ISGID)) { |
9f064b7e | 304 | ip->i_d.di_gid = pip->i_d.di_gid; |
e37bf53c DC |
305 | if ((VFS_I(pip)->i_mode & S_ISGID) && (mode & S_IFMT) == S_IFDIR) |
306 | VFS_I(ip)->i_mode |= S_ISGID; | |
9f064b7e NS |
307 | } |
308 | ||
2bd0ea18 NS |
309 | ip->i_d.di_size = 0; |
310 | ip->i_d.di_nextents = 0; | |
311 | ASSERT(ip->i_d.di_nblocks == 0); | |
9f064b7e | 312 | ip->i_d.di_extsize = pip ? 0 : fsx->fsx_extsize; |
2bd0ea18 NS |
313 | ip->i_d.di_dmevmask = 0; |
314 | ip->i_d.di_dmstate = 0; | |
06b80354 | 315 | ip->i_d.di_flags = pip ? 0 : xfs_flags2diflags(ip, fsx->fsx_xflags); |
41ce5f36 DC |
316 | |
317 | if (ip->i_d.di_version == 3) { | |
318 | ASSERT(ip->i_d.di_ino == ino); | |
299c0388 | 319 | ASSERT(uuid_equal(&ip->i_d.di_uuid, &mp->m_sb.sb_meta_uuid)); |
9abcc5cd | 320 | VFS_I(ip)->i_version = 1; |
06b80354 DW |
321 | ip->i_d.di_flags2 = pip ? 0 : xfs_flags2diflags2(ip, |
322 | fsx->fsx_xflags); | |
14f8b681 DW |
323 | ip->i_d.di_crtime.t_sec = (int32_t)VFS_I(ip)->i_mtime.tv_sec; |
324 | ip->i_d.di_crtime.t_nsec = (int32_t)VFS_I(ip)->i_mtime.tv_nsec; | |
06b80354 | 325 | ip->i_d.di_cowextsize = pip ? 0 : fsx->fsx_cowextsize; |
41ce5f36 DC |
326 | } |
327 | ||
2bd0ea18 | 328 | flags = XFS_ILOG_CORE; |
322f2a29 SL |
329 | switch (mode & S_IFMT) { |
330 | case S_IFIFO: | |
63899e27 NS |
331 | case S_IFSOCK: |
332 | /* doesn't make sense to set an rdev for these */ | |
333 | rdev = 0; | |
90ec25ed | 334 | /* FALLTHROUGH */ |
322f2a29 SL |
335 | case S_IFCHR: |
336 | case S_IFBLK: | |
2bd0ea18 | 337 | ip->i_d.di_format = XFS_DINODE_FMT_DEV; |
2bd0ea18 | 338 | flags |= XFS_ILOG_DEV; |
551174eb | 339 | VFS_I(ip)->i_rdev = rdev; |
2bd0ea18 | 340 | break; |
322f2a29 SL |
341 | case S_IFREG: |
342 | case S_IFDIR: | |
9f064b7e NS |
343 | if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) { |
344 | uint di_flags = 0; | |
345 | ||
346 | if ((mode & S_IFMT) == S_IFDIR) { | |
347 | if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT) | |
348 | di_flags |= XFS_DIFLAG_RTINHERIT; | |
349 | if (pip->i_d.di_flags & XFS_DIFLAG_EXTSZINHERIT) { | |
350 | di_flags |= XFS_DIFLAG_EXTSZINHERIT; | |
351 | ip->i_d.di_extsize = pip->i_d.di_extsize; | |
352 | } | |
353 | } else { | |
354 | if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT) { | |
355 | di_flags |= XFS_DIFLAG_REALTIME; | |
356 | } | |
357 | if (pip->i_d.di_flags & XFS_DIFLAG_EXTSZINHERIT) { | |
358 | di_flags |= XFS_DIFLAG_EXTSIZE; | |
359 | ip->i_d.di_extsize = pip->i_d.di_extsize; | |
360 | } | |
361 | } | |
362 | if (pip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) | |
363 | di_flags |= XFS_DIFLAG_PROJINHERIT; | |
364 | ip->i_d.di_flags |= di_flags; | |
365 | } | |
366 | /* FALLTHROUGH */ | |
322f2a29 | 367 | case S_IFLNK: |
2bd0ea18 NS |
368 | ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS; |
369 | ip->i_df.if_flags = XFS_IFEXTENTS; | |
370 | ip->i_df.if_bytes = ip->i_df.if_real_bytes = 0; | |
371 | ip->i_df.if_u1.if_extents = NULL; | |
372 | break; | |
373 | default: | |
374 | ASSERT(0); | |
375 | } | |
376 | /* Attribute fork settings for new inode. */ | |
377 | ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS; | |
378 | ip->i_d.di_anextents = 0; | |
379 | ||
ff105f75 DC |
380 | /* |
381 | * set up the inode ops structure that the libxfs code relies on | |
382 | */ | |
e37bf53c | 383 | if (XFS_ISDIR(ip)) |
ff105f75 DC |
384 | ip->d_ops = ip->i_mount->m_dir_inode_ops; |
385 | else | |
386 | ip->d_ops = ip->i_mount->m_nondir_inode_ops; | |
387 | ||
2bd0ea18 NS |
388 | /* |
389 | * Log the new values stuffed into the inode. | |
390 | */ | |
391 | xfs_trans_log_inode(tp, ip, flags); | |
392 | *ipp = ip; | |
393 | return 0; | |
394 | } | |
395 | ||
396 | void | |
5e656dbb BN |
397 | libxfs_iprint( |
398 | xfs_inode_t *ip) | |
2bd0ea18 | 399 | { |
138659f1 | 400 | struct xfs_icdinode *dip; |
5e656dbb | 401 | xfs_bmbt_rec_host_t *ep; |
db15fab1 NS |
402 | xfs_extnum_t i; |
403 | xfs_extnum_t nextents; | |
2bd0ea18 | 404 | |
5b64e00a | 405 | printf("Inode %lx\n", (unsigned long)ip); |
5b64e00a | 406 | printf(" i_ino %llx\n", (unsigned long long)ip->i_ino); |
2bd0ea18 NS |
407 | |
408 | if (ip->i_df.if_flags & XFS_IFEXTENTS) | |
409 | printf("EXTENTS "); | |
410 | printf("\n"); | |
411 | printf(" i_df.if_bytes %d\n", ip->i_df.if_bytes); | |
5b64e00a NS |
412 | printf(" i_df.if_u1.if_extents/if_data %lx\n", |
413 | (unsigned long)ip->i_df.if_u1.if_extents); | |
2bd0ea18 NS |
414 | if (ip->i_df.if_flags & XFS_IFEXTENTS) { |
415 | nextents = ip->i_df.if_bytes / (uint)sizeof(*ep); | |
f8149110 | 416 | for (ep = ip->i_df.if_u1.if_extents, i = 0; i < nextents; |
5e656dbb | 417 | i++, ep++) { |
5000d01d | 418 | xfs_bmbt_irec_t rec; |
2bd0ea18 NS |
419 | |
420 | xfs_bmbt_get_all(ep, &rec); | |
5b64e00a NS |
421 | printf("\t%d: startoff %llu, startblock 0x%llx," |
422 | " blockcount %llu, state %d\n", | |
423 | i, (unsigned long long)rec.br_startoff, | |
424 | (unsigned long long)rec.br_startblock, | |
425 | (unsigned long long)rec.br_blockcount, | |
2bd0ea18 NS |
426 | (int)rec.br_state); |
427 | } | |
428 | } | |
5b64e00a | 429 | printf(" i_df.if_broot %lx\n", (unsigned long)ip->i_df.if_broot); |
2bd0ea18 NS |
430 | printf(" i_df.if_broot_bytes %x\n", ip->i_df.if_broot_bytes); |
431 | ||
5e656dbb | 432 | dip = &ip->i_d; |
2bd0ea18 | 433 | printf("\nOn disk portion\n"); |
e37bf53c | 434 | printf(" di_mode %o\n", VFS_I(ip)->i_mode); |
2bd0ea18 NS |
435 | printf(" di_version %x\n", (uint)dip->di_version); |
436 | switch (ip->i_d.di_format) { | |
437 | case XFS_DINODE_FMT_LOCAL: | |
438 | printf(" Inline inode\n"); | |
439 | break; | |
440 | case XFS_DINODE_FMT_EXTENTS: | |
441 | printf(" Extents inode\n"); | |
442 | break; | |
443 | case XFS_DINODE_FMT_BTREE: | |
444 | printf(" B-tree inode\n"); | |
445 | break; | |
446 | default: | |
447 | printf(" Other inode\n"); | |
448 | break; | |
449 | } | |
bcbe04c1 | 450 | printf(" di_nlink %x\n", VFS_I(ip)->i_nlink); |
2bd0ea18 NS |
451 | printf(" di_uid %d\n", dip->di_uid); |
452 | printf(" di_gid %d\n", dip->di_gid); | |
453 | printf(" di_nextents %d\n", dip->di_nextents); | |
5b64e00a | 454 | printf(" di_size %llu\n", (unsigned long long)dip->di_size); |
6652c253 | 455 | printf(" di_gen %x\n", VFS_I(ip)->i_generation); |
2bd0ea18 NS |
456 | printf(" di_extsize %d\n", dip->di_extsize); |
457 | printf(" di_flags %x\n", dip->di_flags); | |
5b64e00a | 458 | printf(" di_nblocks %llu\n", (unsigned long long)dip->di_nblocks); |
2bd0ea18 NS |
459 | } |
460 | ||
461 | /* | |
462 | * Writes a modified inode's changes out to the inode's on disk home. | |
463 | * Originally based on xfs_iflush_int() from xfs_inode.c in the kernel. | |
464 | */ | |
465 | int | |
466 | libxfs_iflush_int(xfs_inode_t *ip, xfs_buf_t *bp) | |
467 | { | |
468 | xfs_inode_log_item_t *iip; | |
469 | xfs_dinode_t *dip; | |
470 | xfs_mount_t *mp; | |
471 | ||
472 | ASSERT(XFS_BUF_FSPRIVATE(bp, void *) != NULL); | |
473 | ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE || | |
474 | ip->i_d.di_nextents > ip->i_df.if_ext_max); | |
ff105f75 | 475 | ASSERT(ip->i_d.di_version > 1); |
2bd0ea18 NS |
476 | |
477 | iip = ip->i_itemp; | |
478 | mp = ip->i_mount; | |
479 | ||
480 | /* set *dip = inode's place in the buffer */ | |
92acb899 | 481 | dip = xfs_buf_offset(bp, ip->i_imap.im_boffset); |
2bd0ea18 | 482 | |
2bd0ea18 | 483 | ASSERT(ip->i_d.di_magic == XFS_DINODE_MAGIC); |
e37bf53c | 484 | if (XFS_ISREG(ip)) { |
2bd0ea18 NS |
485 | ASSERT( (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS) || |
486 | (ip->i_d.di_format == XFS_DINODE_FMT_BTREE) ); | |
ff0f39ea | 487 | } else if (XFS_ISDIR(ip)) { |
2bd0ea18 NS |
488 | ASSERT( (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS) || |
489 | (ip->i_d.di_format == XFS_DINODE_FMT_BTREE) || | |
490 | (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL) ); | |
491 | } | |
492 | ASSERT(ip->i_d.di_nextents+ip->i_d.di_anextents <= ip->i_d.di_nblocks); | |
493 | ASSERT(ip->i_d.di_forkoff <= mp->m_sb.sb_inodesize); | |
2bd0ea18 | 494 | |
41ce5f36 DC |
495 | /* bump the change count on v3 inodes */ |
496 | if (ip->i_d.di_version == 3) | |
9abcc5cd | 497 | VFS_I(ip)->i_version++; |
41ce5f36 | 498 | |
d15188a1 DW |
499 | /* Check the inline directory data. */ |
500 | if (S_ISDIR(VFS_I(ip)->i_mode) && | |
501 | ip->i_d.di_format == XFS_DINODE_FMT_LOCAL && | |
502 | xfs_dir2_sf_verify(ip)) | |
503 | return -EFSCORRUPTED; | |
504 | ||
2bd0ea18 NS |
505 | /* |
506 | * Copy the dirty parts of the inode into the on-disk | |
507 | * inode. We always copy out the core of the inode, | |
508 | * because if the inode is dirty at all the core must | |
509 | * be. | |
510 | */ | |
db17aebe | 511 | xfs_inode_to_disk(ip, dip, iip->ili_item.li_lsn); |
46eca962 | 512 | |
ff105f75 | 513 | xfs_iflush_fork(ip, dip, iip, XFS_DATA_FORK); |
f8149110 | 514 | if (XFS_IFORK_Q(ip)) |
ff105f75 | 515 | xfs_iflush_fork(ip, dip, iip, XFS_ATTR_FORK); |
2bd0ea18 | 516 | |
41ce5f36 DC |
517 | /* generate the checksum. */ |
518 | xfs_dinode_calc_crc(mp, dip); | |
519 | ||
2bd0ea18 NS |
520 | return 0; |
521 | } | |
522 | ||
2bd0ea18 | 523 | int |
5e656dbb | 524 | libxfs_mod_incore_sb( |
19ebedcf DC |
525 | struct xfs_mount *mp, |
526 | int field, | |
5e656dbb BN |
527 | int64_t delta, |
528 | int rsvd) | |
2bd0ea18 NS |
529 | { |
530 | long long lcounter; /* long counter for 64 bit fields */ | |
531 | ||
532 | switch (field) { | |
19ebedcf | 533 | case XFS_TRANS_SB_FDBLOCKS: |
2bd0ea18 NS |
534 | lcounter = (long long)mp->m_sb.sb_fdblocks; |
535 | lcounter += delta; | |
536 | if (lcounter < 0) | |
12b53197 | 537 | return -ENOSPC; |
2bd0ea18 | 538 | mp->m_sb.sb_fdblocks = lcounter; |
5e656dbb | 539 | return 0; |
2bd0ea18 NS |
540 | default: |
541 | ASSERT(0); | |
12b53197 | 542 | return -EINVAL; |
2bd0ea18 | 543 | } |
2bd0ea18 NS |
544 | } |
545 | ||
2bd0ea18 NS |
546 | /* |
547 | * This routine allocates disk space for the given file. | |
548 | * Originally derived from xfs_alloc_file_space(). | |
549 | */ | |
550 | int | |
551 | libxfs_alloc_file_space( | |
552 | xfs_inode_t *ip, | |
553 | xfs_off_t offset, | |
554 | xfs_off_t len, | |
555 | int alloc_type, | |
556 | int attr_flags) | |
557 | { | |
558 | xfs_mount_t *mp; | |
559 | xfs_off_t count; | |
560 | xfs_filblks_t datablocks; | |
561 | xfs_filblks_t allocated_fsb; | |
562 | xfs_filblks_t allocatesize_fsb; | |
563 | xfs_fsblock_t firstfsb; | |
6f530e9a | 564 | struct xfs_defer_ops free_list; |
5000d01d SL |
565 | xfs_bmbt_irec_t *imapp; |
566 | xfs_bmbt_irec_t imaps[1]; | |
2bd0ea18 NS |
567 | int reccount; |
568 | uint resblks; | |
569 | xfs_fileoff_t startoffset_fsb; | |
570 | xfs_trans_t *tp; | |
571 | int xfs_bmapi_flags; | |
2bd0ea18 NS |
572 | int error; |
573 | ||
574 | if (len <= 0) | |
12b53197 | 575 | return -EINVAL; |
2bd0ea18 NS |
576 | |
577 | count = len; | |
578 | error = 0; | |
579 | imapp = &imaps[0]; | |
580 | reccount = 1; | |
a2ceac1f | 581 | xfs_bmapi_flags = alloc_type ? XFS_BMAPI_PREALLOC : 0; |
2bd0ea18 NS |
582 | mp = ip->i_mount; |
583 | startoffset_fsb = XFS_B_TO_FSBT(mp, offset); | |
584 | allocatesize_fsb = XFS_B_TO_FSB(mp, count); | |
585 | ||
586 | /* allocate file space until done or until there is an error */ | |
587 | while (allocatesize_fsb && !error) { | |
588 | datablocks = allocatesize_fsb; | |
589 | ||
2bd0ea18 | 590 | resblks = (uint)XFS_DIOSTRAT_SPACE_RES(mp, datablocks); |
9074815c CH |
591 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, |
592 | 0, 0, &tp); | |
682fa9e0 ES |
593 | /* |
594 | * Check for running out of space | |
595 | */ | |
596 | if (error) { | |
12b53197 | 597 | ASSERT(error == -ENOSPC); |
2bd0ea18 | 598 | break; |
682fa9e0 | 599 | } |
2bd0ea18 | 600 | xfs_trans_ijoin(tp, ip, 0); |
2bd0ea18 | 601 | |
6f530e9a | 602 | xfs_defer_init(&free_list, &firstfsb); |
a2ceac1f | 603 | error = xfs_bmapi_write(tp, ip, startoffset_fsb, allocatesize_fsb, |
2bd0ea18 | 604 | xfs_bmapi_flags, &firstfsb, 0, imapp, |
56b2de80 DC |
605 | &reccount, &free_list); |
606 | ||
2bd0ea18 | 607 | if (error) |
682fa9e0 | 608 | goto error0; |
2bd0ea18 | 609 | |
5c33baee CH |
610 | /* |
611 | * Complete the transaction | |
612 | */ | |
613 | error = xfs_defer_finish(&tp, &free_list); | |
2bd0ea18 | 614 | if (error) |
682fa9e0 | 615 | goto error0; |
2bd0ea18 | 616 | |
de5a3f46 | 617 | error = xfs_trans_commit(tp); |
2bd0ea18 NS |
618 | if (error) |
619 | break; | |
620 | ||
621 | allocated_fsb = imapp->br_blockcount; | |
622 | if (reccount == 0) | |
12b53197 | 623 | return -ENOSPC; |
2bd0ea18 NS |
624 | |
625 | startoffset_fsb += allocated_fsb; | |
626 | allocatesize_fsb -= allocated_fsb; | |
627 | } | |
628 | return error; | |
682fa9e0 ES |
629 | |
630 | error0: /* Cancel bmap, cancel trans */ | |
6f530e9a | 631 | xfs_defer_cancel(&free_list); |
3d7434fe | 632 | xfs_trans_cancel(tp); |
682fa9e0 | 633 | return error; |
2bd0ea18 NS |
634 | } |
635 | ||
636 | unsigned int | |
637 | libxfs_log2_roundup(unsigned int i) | |
638 | { | |
639 | unsigned int rval; | |
640 | ||
641 | for (rval = 0; rval < NBBY * sizeof(i); rval++) { | |
642 | if ((1 << rval) >= i) | |
643 | break; | |
644 | } | |
645 | return rval; | |
646 | } | |
647 | ||
9f064b7e NS |
648 | /* |
649 | * Wrapper around call to libxfs_ialloc. Takes care of committing and | |
650 | * allocating a new transaction as needed. | |
651 | * | |
652 | * Originally there were two copies of this code - one in mkfs, the | |
653 | * other in repair - now there is just the one. | |
654 | */ | |
655 | int | |
656 | libxfs_inode_alloc( | |
657 | xfs_trans_t **tp, | |
658 | xfs_inode_t *pip, | |
659 | mode_t mode, | |
660 | nlink_t nlink, | |
661 | xfs_dev_t rdev, | |
662 | struct cred *cr, | |
663 | struct fsxattr *fsx, | |
664 | xfs_inode_t **ipp) | |
665 | { | |
9f064b7e NS |
666 | xfs_buf_t *ialloc_context; |
667 | xfs_inode_t *ip; | |
9f064b7e NS |
668 | int error; |
669 | ||
9f064b7e NS |
670 | ialloc_context = (xfs_buf_t *)0; |
671 | error = libxfs_ialloc(*tp, pip, mode, nlink, rdev, cr, fsx, | |
a2ceac1f DC |
672 | 1, &ialloc_context, &ip); |
673 | if (error) { | |
674 | *ipp = NULL; | |
9f064b7e | 675 | return error; |
a2ceac1f DC |
676 | } |
677 | if (!ialloc_context && !ip) { | |
678 | *ipp = NULL; | |
12b53197 | 679 | return -ENOSPC; |
a2ceac1f | 680 | } |
9f064b7e | 681 | |
a2ceac1f | 682 | if (ialloc_context) { |
48ea6cb9 | 683 | |
9f064b7e | 684 | xfs_trans_bhold(*tp, ialloc_context); |
d262295d | 685 | |
d67406c9 | 686 | error = xfs_trans_roll(tp); |
48ea6cb9 | 687 | if (error) { |
d262295d | 688 | fprintf(stderr, _("%s: cannot duplicate transaction: %s\n"), |
48ea6cb9 | 689 | progname, strerror(error)); |
9f064b7e NS |
690 | exit(1); |
691 | } | |
692 | xfs_trans_bjoin(*tp, ialloc_context); | |
693 | error = libxfs_ialloc(*tp, pip, mode, nlink, rdev, cr, | |
a2ceac1f | 694 | fsx, 1, &ialloc_context, &ip); |
9f064b7e | 695 | if (!ip) |
12b53197 | 696 | error = -ENOSPC; |
9f064b7e NS |
697 | if (error) |
698 | return error; | |
699 | } | |
9f064b7e NS |
700 | |
701 | *ipp = ip; | |
702 | return error; | |
703 | } | |
1552a820 NS |
704 | |
705 | /* | |
706 | * Userspace versions of common diagnostic routines (varargs fun). | |
707 | */ | |
708 | void | |
5e656dbb | 709 | libxfs_fs_repair_cmn_err(int level, xfs_mount_t *mp, char *fmt, ...) |
1552a820 NS |
710 | { |
711 | va_list ap; | |
712 | ||
713 | va_start(ap, fmt); | |
714 | vfprintf(stderr, fmt, ap); | |
715 | fprintf(stderr, " This is a bug.\n"); | |
89c4bb8e | 716 | fprintf(stderr, "%s version %s\n", progname, VERSION); |
130093ab DC |
717 | fprintf(stderr, |
718 | "Please capture the filesystem metadata with xfs_metadump and\n" | |
719 | "report it to linux-xfs@vger.kernel.org\n"); | |
1552a820 NS |
720 | va_end(ap); |
721 | } | |
722 | ||
723 | void | |
5e656dbb | 724 | libxfs_fs_cmn_err(int level, xfs_mount_t *mp, char *fmt, ...) |
1552a820 NS |
725 | { |
726 | va_list ap; | |
727 | ||
728 | va_start(ap, fmt); | |
729 | vfprintf(stderr, fmt, ap); | |
730 | fputs("\n", stderr); | |
731 | va_end(ap); | |
732 | } | |
733 | ||
734 | void | |
735 | cmn_err(int level, char *fmt, ...) | |
736 | { | |
737 | va_list ap; | |
738 | ||
739 | va_start(ap, fmt); | |
740 | vfprintf(stderr, fmt, ap); | |
741 | fputs("\n", stderr); | |
742 | va_end(ap); | |
743 | } | |
99c1ec96 DC |
744 | |
745 | /* | |
746 | * Warnings specifically for verifier errors. Differentiate CRC vs. invalid | |
747 | * values, and omit the stack trace unless the error level is tuned high. | |
748 | */ | |
749 | void | |
750 | xfs_verifier_error( | |
751 | struct xfs_buf *bp) | |
752 | { | |
a3fac935 | 753 | xfs_alert(NULL, "Metadata %s detected at %s block 0x%llx/0x%x", |
12b53197 | 754 | bp->b_error == -EFSBADCRC ? "CRC error" : "corruption", |
a3fac935 | 755 | bp->b_ops->name, bp->b_bn, BBTOB(bp->b_length)); |
99c1ec96 | 756 | } |
a65d8d29 | 757 | |
7d77349c BF |
758 | /* |
759 | * This is called from I/O verifiers on v5 superblock filesystems. In the | |
760 | * kernel, it validates the metadata LSN parameter against the current LSN of | |
761 | * the active log. We don't have an active log in userspace so this kind of | |
762 | * validation is not required. Therefore, this function always returns true in | |
763 | * userspace. | |
764 | * | |
765 | * xfs_repair piggybacks off this mechanism to help track the largest metadata | |
766 | * LSN in use on a filesystem. Keep a record of the largest LSN seen such that | |
767 | * repair can validate it against the state of the log. | |
768 | */ | |
769 | xfs_lsn_t libxfs_max_lsn = 0; | |
770 | pthread_mutex_t libxfs_max_lsn_lock = PTHREAD_MUTEX_INITIALIZER; | |
771 | ||
a65d8d29 BF |
772 | bool |
773 | xfs_log_check_lsn( | |
774 | struct xfs_mount *mp, | |
775 | xfs_lsn_t lsn) | |
776 | { | |
7d77349c BF |
777 | int cycle = CYCLE_LSN(lsn); |
778 | int block = BLOCK_LSN(lsn); | |
779 | int max_cycle; | |
780 | int max_block; | |
781 | ||
782 | if (lsn == NULLCOMMITLSN) | |
783 | return true; | |
784 | ||
785 | pthread_mutex_lock(&libxfs_max_lsn_lock); | |
786 | ||
787 | max_cycle = CYCLE_LSN(libxfs_max_lsn); | |
788 | max_block = BLOCK_LSN(libxfs_max_lsn); | |
789 | ||
790 | if ((cycle > max_cycle) || | |
791 | (cycle == max_cycle && block > max_block)) | |
792 | libxfs_max_lsn = lsn; | |
793 | ||
794 | pthread_mutex_unlock(&libxfs_max_lsn_lock); | |
795 | ||
a65d8d29 BF |
796 | return true; |
797 | } | |
9542ae13 DC |
798 | |
799 | static struct xfs_buftarg * | |
800 | xfs_find_bdev_for_inode( | |
801 | struct xfs_inode *ip) | |
802 | { | |
803 | struct xfs_mount *mp = ip->i_mount; | |
804 | ||
805 | if (XFS_IS_REALTIME_INODE(ip)) | |
806 | return mp->m_rtdev_targp; | |
807 | return mp->m_ddev_targp; | |
808 | } | |
809 | ||
810 | static xfs_daddr_t | |
811 | xfs_fsb_to_db(struct xfs_inode *ip, xfs_fsblock_t fsb) | |
812 | { | |
813 | if (XFS_IS_REALTIME_INODE(ip)) | |
814 | return XFS_FSB_TO_BB(ip->i_mount, fsb); | |
815 | return XFS_FSB_TO_DADDR(ip->i_mount, (fsb)); | |
816 | } | |
817 | ||
818 | int | |
819 | libxfs_zero_extent( | |
820 | struct xfs_inode *ip, | |
821 | xfs_fsblock_t start_fsb, | |
822 | xfs_off_t count_fsb) | |
823 | { | |
824 | xfs_daddr_t sector = xfs_fsb_to_db(ip, start_fsb); | |
825 | ssize_t size = XFS_FSB_TO_BB(ip->i_mount, count_fsb); | |
826 | ||
827 | return libxfs_device_zero(xfs_find_bdev_for_inode(ip), sector, size); | |
828 | } | |
0b90dda6 | 829 |