Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
7b718769 NS |
3 | * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
1da177e4 | 5 | */ |
1da177e4 | 6 | #include "xfs.h" |
a844f451 | 7 | #include "xfs_fs.h" |
70a9883c | 8 | #include "xfs_shared.h" |
a4fbe6ab | 9 | #include "xfs_format.h" |
239880ef DC |
10 | #include "xfs_log_format.h" |
11 | #include "xfs_trans_resv.h" | |
1da177e4 | 12 | #include "xfs_mount.h" |
1da177e4 | 13 | #include "xfs_inode.h" |
a4fbe6ab | 14 | #include "xfs_btree.h" |
1da177e4 | 15 | #include "xfs_ialloc.h" |
a4fbe6ab | 16 | #include "xfs_ialloc_btree.h" |
2810bd68 | 17 | #include "xfs_iwalk.h" |
1da177e4 LT |
18 | #include "xfs_itable.h" |
19 | #include "xfs_error.h" | |
33479e05 | 20 | #include "xfs_icache.h" |
89d139d5 | 21 | #include "xfs_health.h" |
a6343e4d | 22 | #include "xfs_trans.h" |
1da177e4 | 23 | |
7dce11db | 24 | /* |
2810bd68 DW |
25 | * Bulk Stat |
26 | * ========= | |
27 | * | |
7035f972 | 28 | * Use the inode walking functions to fill out struct xfs_bulkstat for every |
2810bd68 DW |
29 | * allocated inode, then pass the stat information to some externally provided |
30 | * iteration function. | |
7dce11db | 31 | */ |
2810bd68 DW |
32 | |
33 | struct xfs_bstat_chunk { | |
34 | bulkstat_one_fmt_pf formatter; | |
35 | struct xfs_ibulk *breq; | |
7035f972 | 36 | struct xfs_bulkstat *buf; |
2810bd68 DW |
37 | }; |
38 | ||
39 | /* | |
40 | * Fill out the bulkstat info for a single inode and report it somewhere. | |
41 | * | |
42 | * bc->breq->lastino is effectively the inode cursor as we walk through the | |
43 | * filesystem. Therefore, we update it any time we need to move the cursor | |
44 | * forward, regardless of whether or not we're sending any bstat information | |
45 | * back to userspace. If the inode is internal metadata or, has been freed | |
46 | * out from under us, we just simply keep going. | |
47 | * | |
48 | * However, if any other type of error happens we want to stop right where we | |
49 | * are so that userspace will call back with exact number of the bad inode and | |
50 | * we can send back an error code. | |
51 | * | |
52 | * Note that if the formatter tells us there's no space left in the buffer we | |
53 | * move the cursor forward and abort the walk. | |
54 | */ | |
55 | STATIC int | |
7dce11db | 56 | xfs_bulkstat_one_int( |
2810bd68 | 57 | struct xfs_mount *mp, |
f736d93d | 58 | struct user_namespace *mnt_userns, |
2810bd68 DW |
59 | struct xfs_trans *tp, |
60 | xfs_ino_t ino, | |
61 | struct xfs_bstat_chunk *bc) | |
1da177e4 | 62 | { |
f736d93d | 63 | struct user_namespace *sb_userns = mp->m_super->s_user_ns; |
7dce11db | 64 | struct xfs_inode *ip; /* incore inode pointer */ |
3987848c | 65 | struct inode *inode; |
7035f972 | 66 | struct xfs_bulkstat *buf = bc->buf; |
2810bd68 | 67 | int error = -EINVAL; |
7dce11db | 68 | |
2810bd68 DW |
69 | if (xfs_internal_inum(mp, ino)) |
70 | goto out_advance; | |
7dce11db | 71 | |
2810bd68 | 72 | error = xfs_iget(mp, tp, ino, |
5132ba8f DC |
73 | (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED), |
74 | XFS_ILOCK_SHARED, &ip); | |
2810bd68 DW |
75 | if (error == -ENOENT || error == -EINVAL) |
76 | goto out_advance; | |
8fe65776 | 77 | if (error) |
2810bd68 | 78 | goto out; |
1da177e4 LT |
79 | |
80 | ASSERT(ip != NULL); | |
92bfc6e7 | 81 | ASSERT(ip->i_imap.im_blkno != 0); |
3987848c | 82 | inode = VFS_I(ip); |
1da177e4 | 83 | |
1da177e4 LT |
84 | /* xfs_iget returns the following without needing |
85 | * further change. | |
86 | */ | |
ceaf603c | 87 | buf->bs_projectid = ip->i_projid; |
1da177e4 | 88 | buf->bs_ino = ino; |
f736d93d CH |
89 | buf->bs_uid = from_kuid(sb_userns, i_uid_into_mnt(mnt_userns, inode)); |
90 | buf->bs_gid = from_kgid(sb_userns, i_gid_into_mnt(mnt_userns, inode)); | |
13d2c10b | 91 | buf->bs_size = ip->i_disk_size; |
3987848c | 92 | |
54d7b5c1 | 93 | buf->bs_nlink = inode->i_nlink; |
7035f972 DW |
94 | buf->bs_atime = inode->i_atime.tv_sec; |
95 | buf->bs_atime_nsec = inode->i_atime.tv_nsec; | |
96 | buf->bs_mtime = inode->i_mtime.tv_sec; | |
97 | buf->bs_mtime_nsec = inode->i_mtime.tv_nsec; | |
98 | buf->bs_ctime = inode->i_ctime.tv_sec; | |
99 | buf->bs_ctime_nsec = inode->i_ctime.tv_nsec; | |
9e9a2674 | 100 | buf->bs_gen = inode->i_generation; |
c19b3b05 | 101 | buf->bs_mode = inode->i_mode; |
3987848c | 102 | |
1da177e4 | 103 | buf->bs_xflags = xfs_ip2xflags(ip); |
031474c2 | 104 | buf->bs_extsize_blks = ip->i_extsize; |
daf83964 | 105 | buf->bs_extents = xfs_ifork_nextents(&ip->i_df); |
89d139d5 | 106 | xfs_bulkstat_health(ip, buf); |
daf83964 | 107 | buf->bs_aextents = xfs_ifork_nextents(ip->i_afp); |
07000ee6 | 108 | buf->bs_forkoff = XFS_IFORK_BOFF(ip); |
7035f972 | 109 | buf->bs_version = XFS_BULKSTAT_VERSION_V5; |
1da177e4 | 110 | |
38c26bfd | 111 | if (xfs_has_v3inodes(mp)) { |
e98d5e88 CH |
112 | buf->bs_btime = ip->i_crtime.tv_sec; |
113 | buf->bs_btime_nsec = ip->i_crtime.tv_nsec; | |
3e09ab8f | 114 | if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE) |
b33ce57d | 115 | buf->bs_cowextsize_blks = ip->i_cowextsize; |
f7ca3522 DW |
116 | } |
117 | ||
f7e67b20 | 118 | switch (ip->i_df.if_format) { |
1da177e4 | 119 | case XFS_DINODE_FMT_DEV: |
66f36464 | 120 | buf->bs_rdev = sysv_encode_dev(inode->i_rdev); |
1da177e4 LT |
121 | buf->bs_blksize = BLKDEV_IOSIZE; |
122 | buf->bs_blocks = 0; | |
123 | break; | |
124 | case XFS_DINODE_FMT_LOCAL: | |
1da177e4 LT |
125 | buf->bs_rdev = 0; |
126 | buf->bs_blksize = mp->m_sb.sb_blocksize; | |
127 | buf->bs_blocks = 0; | |
128 | break; | |
129 | case XFS_DINODE_FMT_EXTENTS: | |
130 | case XFS_DINODE_FMT_BTREE: | |
131 | buf->bs_rdev = 0; | |
132 | buf->bs_blksize = mp->m_sb.sb_blocksize; | |
6e73a545 | 133 | buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks; |
1da177e4 LT |
134 | break; |
135 | } | |
f2d67614 | 136 | xfs_iunlock(ip, XFS_ILOCK_SHARED); |
44a8736b | 137 | xfs_irele(ip); |
1da177e4 | 138 | |
2810bd68 | 139 | error = bc->formatter(bc->breq, buf); |
e7ee96df | 140 | if (error == -ECANCELED) |
2810bd68 DW |
141 | goto out_advance; |
142 | if (error) | |
143 | goto out; | |
1da177e4 | 144 | |
2810bd68 DW |
145 | out_advance: |
146 | /* | |
147 | * Advance the cursor to the inode that comes after the one we just | |
148 | * looked at. We want the caller to move along if the bulkstat | |
149 | * information was copied successfully; if we tried to grab the inode | |
150 | * but it's no longer allocated; or if it's internal metadata. | |
151 | */ | |
152 | bc->breq->startino = ino + 1; | |
153 | out: | |
7dce11db | 154 | return error; |
1da177e4 LT |
155 | } |
156 | ||
2810bd68 | 157 | /* Bulkstat a single inode. */ |
2ee4fa5c | 158 | int |
159 | xfs_bulkstat_one( | |
2810bd68 DW |
160 | struct xfs_ibulk *breq, |
161 | bulkstat_one_fmt_pf formatter) | |
2ee4fa5c | 162 | { |
2810bd68 DW |
163 | struct xfs_bstat_chunk bc = { |
164 | .formatter = formatter, | |
165 | .breq = breq, | |
166 | }; | |
a6343e4d | 167 | struct xfs_trans *tp; |
2810bd68 DW |
168 | int error; |
169 | ||
8723d5ba CH |
170 | if (breq->mnt_userns != &init_user_ns) { |
171 | xfs_warn_ratelimited(breq->mp, | |
172 | "bulkstat not supported inside of idmapped mounts."); | |
173 | return -EINVAL; | |
174 | } | |
175 | ||
2810bd68 DW |
176 | ASSERT(breq->icount == 1); |
177 | ||
7035f972 | 178 | bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), |
707e0dda | 179 | KM_MAYFAIL); |
2810bd68 DW |
180 | if (!bc.buf) |
181 | return -ENOMEM; | |
182 | ||
a6343e4d DW |
183 | /* |
184 | * Grab an empty transaction so that we can use its recursive buffer | |
185 | * locking abilities to detect cycles in the inobt without deadlocking. | |
186 | */ | |
187 | error = xfs_trans_alloc_empty(breq->mp, &tp); | |
188 | if (error) | |
189 | goto out; | |
2810bd68 | 190 | |
a6343e4d DW |
191 | error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, tp, |
192 | breq->startino, &bc); | |
193 | xfs_trans_cancel(tp); | |
194 | out: | |
2810bd68 DW |
195 | kmem_free(bc.buf); |
196 | ||
197 | /* | |
198 | * If we reported one inode to userspace then we abort because we hit | |
199 | * the end of the buffer. Don't leak that back to userspace. | |
200 | */ | |
e7ee96df | 201 | if (error == -ECANCELED) |
2810bd68 DW |
202 | error = 0; |
203 | ||
204 | return error; | |
8b56f083 NS |
205 | } |
206 | ||
bf4a5af2 | 207 | static int |
2810bd68 DW |
208 | xfs_bulkstat_iwalk( |
209 | struct xfs_mount *mp, | |
210 | struct xfs_trans *tp, | |
211 | xfs_ino_t ino, | |
212 | void *data) | |
1e773c49 | 213 | { |
f736d93d | 214 | struct xfs_bstat_chunk *bc = data; |
2810bd68 | 215 | int error; |
00275899 | 216 | |
f736d93d | 217 | error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data); |
2810bd68 DW |
218 | /* bulkstat just skips over missing inodes */ |
219 | if (error == -ENOENT || error == -EINVAL) | |
220 | return 0; | |
1e773c49 JL |
221 | return error; |
222 | } | |
223 | ||
1da177e4 | 224 | /* |
2810bd68 DW |
225 | * Check the incoming lastino parameter. |
226 | * | |
227 | * We allow any inode value that could map to physical space inside the | |
228 | * filesystem because if there are no inodes there, bulkstat moves on to the | |
229 | * next chunk. In other words, the magic agino value of zero takes us to the | |
230 | * first chunk in the AG, and an agino value past the end of the AG takes us to | |
231 | * the first chunk in the next AG. | |
232 | * | |
233 | * Therefore we can end early if the requested inode is beyond the end of the | |
234 | * filesystem or doesn't map properly. | |
1da177e4 | 235 | */ |
2810bd68 DW |
236 | static inline bool |
237 | xfs_bulkstat_already_done( | |
238 | struct xfs_mount *mp, | |
239 | xfs_ino_t startino) | |
1da177e4 | 240 | { |
2810bd68 DW |
241 | xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino); |
242 | xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino); | |
1da177e4 | 243 | |
2810bd68 DW |
244 | return agno >= mp->m_sb.sb_agcount || |
245 | startino != XFS_AGINO_TO_INO(mp, agno, agino); | |
246 | } | |
296dfd7f | 247 | |
2810bd68 DW |
248 | /* Return stat information in bulk (by-inode) for the filesystem. */ |
249 | int | |
250 | xfs_bulkstat( | |
251 | struct xfs_ibulk *breq, | |
252 | bulkstat_one_fmt_pf formatter) | |
253 | { | |
254 | struct xfs_bstat_chunk bc = { | |
255 | .formatter = formatter, | |
256 | .breq = breq, | |
257 | }; | |
a6343e4d | 258 | struct xfs_trans *tp; |
2810bd68 | 259 | int error; |
bf4a5af2 | 260 | |
f736d93d CH |
261 | if (breq->mnt_userns != &init_user_ns) { |
262 | xfs_warn_ratelimited(breq->mp, | |
263 | "bulkstat not supported inside of idmapped mounts."); | |
264 | return -EINVAL; | |
265 | } | |
2810bd68 DW |
266 | if (xfs_bulkstat_already_done(breq->mp, breq->startino)) |
267 | return 0; | |
bf4a5af2 | 268 | |
7035f972 | 269 | bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat), |
707e0dda | 270 | KM_MAYFAIL); |
2810bd68 | 271 | if (!bc.buf) |
2451337d | 272 | return -ENOMEM; |
bf4a5af2 | 273 | |
a6343e4d DW |
274 | /* |
275 | * Grab an empty transaction so that we can use its recursive buffer | |
276 | * locking abilities to detect cycles in the inobt without deadlocking. | |
277 | */ | |
278 | error = xfs_trans_alloc_empty(breq->mp, &tp); | |
279 | if (error) | |
280 | goto out; | |
6e57c542 | 281 | |
a6343e4d DW |
282 | error = xfs_iwalk(breq->mp, tp, breq->startino, breq->flags, |
283 | xfs_bulkstat_iwalk, breq->icount, &bc); | |
284 | xfs_trans_cancel(tp); | |
285 | out: | |
2810bd68 | 286 | kmem_free(bc.buf); |
febe3cbe | 287 | |
cd57e594 | 288 | /* |
febe3cbe DC |
289 | * We found some inodes, so clear the error status and return them. |
290 | * The lastino pointer will point directly at the inode that triggered | |
291 | * any error that occurred, so on the next call the error will be | |
292 | * triggered again and propagated to userspace as there will be no | |
293 | * formatted inodes in the buffer. | |
cd57e594 | 294 | */ |
2810bd68 | 295 | if (breq->ocount > 0) |
febe3cbe DC |
296 | error = 0; |
297 | ||
febe3cbe | 298 | return error; |
1da177e4 LT |
299 | } |
300 | ||
7035f972 DW |
301 | /* Convert bulkstat (v5) to bstat (v1). */ |
302 | void | |
303 | xfs_bulkstat_to_bstat( | |
304 | struct xfs_mount *mp, | |
305 | struct xfs_bstat *bs1, | |
306 | const struct xfs_bulkstat *bstat) | |
307 | { | |
2e616d9f | 308 | /* memset is needed here because of padding holes in the structure. */ |
7035f972 DW |
309 | memset(bs1, 0, sizeof(struct xfs_bstat)); |
310 | bs1->bs_ino = bstat->bs_ino; | |
311 | bs1->bs_mode = bstat->bs_mode; | |
312 | bs1->bs_nlink = bstat->bs_nlink; | |
313 | bs1->bs_uid = bstat->bs_uid; | |
314 | bs1->bs_gid = bstat->bs_gid; | |
315 | bs1->bs_rdev = bstat->bs_rdev; | |
316 | bs1->bs_blksize = bstat->bs_blksize; | |
317 | bs1->bs_size = bstat->bs_size; | |
318 | bs1->bs_atime.tv_sec = bstat->bs_atime; | |
319 | bs1->bs_mtime.tv_sec = bstat->bs_mtime; | |
320 | bs1->bs_ctime.tv_sec = bstat->bs_ctime; | |
321 | bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec; | |
322 | bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec; | |
323 | bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec; | |
324 | bs1->bs_blocks = bstat->bs_blocks; | |
325 | bs1->bs_xflags = bstat->bs_xflags; | |
326 | bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks); | |
327 | bs1->bs_extents = bstat->bs_extents; | |
328 | bs1->bs_gen = bstat->bs_gen; | |
329 | bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF; | |
330 | bs1->bs_forkoff = bstat->bs_forkoff; | |
331 | bs1->bs_projid_hi = bstat->bs_projectid >> 16; | |
332 | bs1->bs_sick = bstat->bs_sick; | |
333 | bs1->bs_checked = bstat->bs_checked; | |
334 | bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks); | |
335 | bs1->bs_dmevmask = 0; | |
336 | bs1->bs_dmstate = 0; | |
337 | bs1->bs_aextents = bstat->bs_aextents; | |
338 | } | |
339 | ||
677717fb DW |
340 | struct xfs_inumbers_chunk { |
341 | inumbers_fmt_pf formatter; | |
342 | struct xfs_ibulk *breq; | |
343 | }; | |
344 | ||
345 | /* | |
346 | * INUMBERS | |
347 | * ======== | |
348 | * This is how we export inode btree records to userspace, so that XFS tools | |
349 | * can figure out where inodes are allocated. | |
350 | */ | |
351 | ||
352 | /* | |
353 | * Format the inode group structure and report it somewhere. | |
354 | * | |
355 | * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk | |
356 | * through the filesystem so we move it forward unless there was a runtime | |
357 | * error. If the formatter tells us the buffer is now full we also move the | |
358 | * cursor forward and abort the walk. | |
359 | */ | |
360 | STATIC int | |
361 | xfs_inumbers_walk( | |
362 | struct xfs_mount *mp, | |
363 | struct xfs_trans *tp, | |
364 | xfs_agnumber_t agno, | |
365 | const struct xfs_inobt_rec_incore *irec, | |
366 | void *data) | |
faa63e95 | 367 | { |
5f19c7fc | 368 | struct xfs_inumbers inogrp = { |
677717fb DW |
369 | .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino), |
370 | .xi_alloccount = irec->ir_count - irec->ir_freecount, | |
371 | .xi_allocmask = ~irec->ir_free, | |
5f19c7fc | 372 | .xi_version = XFS_INUMBERS_VERSION_V5, |
677717fb DW |
373 | }; |
374 | struct xfs_inumbers_chunk *ic = data; | |
677717fb DW |
375 | int error; |
376 | ||
377 | error = ic->formatter(ic->breq, &inogrp); | |
e7ee96df | 378 | if (error && error != -ECANCELED) |
677717fb DW |
379 | return error; |
380 | ||
0df5c39b DW |
381 | ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) + |
382 | XFS_INODES_PER_CHUNK; | |
677717fb | 383 | return error; |
faa63e95 MM |
384 | } |
385 | ||
1da177e4 LT |
386 | /* |
387 | * Return inode number table for the filesystem. | |
388 | */ | |
677717fb | 389 | int |
1da177e4 | 390 | xfs_inumbers( |
677717fb | 391 | struct xfs_ibulk *breq, |
549fa006 | 392 | inumbers_fmt_pf formatter) |
1da177e4 | 393 | { |
677717fb DW |
394 | struct xfs_inumbers_chunk ic = { |
395 | .formatter = formatter, | |
396 | .breq = breq, | |
397 | }; | |
a6343e4d | 398 | struct xfs_trans *tp; |
549fa006 JL |
399 | int error = 0; |
400 | ||
677717fb DW |
401 | if (xfs_bulkstat_already_done(breq->mp, breq->startino)) |
402 | return 0; | |
549fa006 | 403 | |
a6343e4d DW |
404 | /* |
405 | * Grab an empty transaction so that we can use its recursive buffer | |
406 | * locking abilities to detect cycles in the inobt without deadlocking. | |
407 | */ | |
408 | error = xfs_trans_alloc_empty(breq->mp, &tp); | |
409 | if (error) | |
410 | goto out; | |
411 | ||
412 | error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags, | |
677717fb | 413 | xfs_inumbers_walk, breq->icount, &ic); |
a6343e4d DW |
414 | xfs_trans_cancel(tp); |
415 | out: | |
c7cb51dc | 416 | |
677717fb DW |
417 | /* |
418 | * We found some inode groups, so clear the error status and return | |
419 | * them. The lastino pointer will point directly at the inode that | |
420 | * triggered any error that occurred, so on the next call the error | |
421 | * will be triggered again and propagated to userspace as there will be | |
422 | * no formatted inode groups in the buffer. | |
423 | */ | |
424 | if (breq->ocount > 0) | |
425 | error = 0; | |
c7cb51dc | 426 | |
1da177e4 LT |
427 | return error; |
428 | } | |
5f19c7fc DW |
429 | |
430 | /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */ | |
431 | void | |
432 | xfs_inumbers_to_inogrp( | |
433 | struct xfs_inogrp *ig1, | |
434 | const struct xfs_inumbers *ig) | |
435 | { | |
2e616d9f DW |
436 | /* memset is needed here because of padding holes in the structure. */ |
437 | memset(ig1, 0, sizeof(struct xfs_inogrp)); | |
5f19c7fc DW |
438 | ig1->xi_startino = ig->xi_startino; |
439 | ig1->xi_alloccount = ig->xi_alloccount; | |
440 | ig1->xi_allocmask = ig->xi_allocmask; | |
441 | } |