Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0
2 : /*
3 : * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
4 : * All Rights Reserved.
5 : */
6 : #include "xfs.h"
7 : #include "xfs_fs.h"
8 : #include "xfs_shared.h"
9 : #include "xfs_format.h"
10 : #include "xfs_log_format.h"
11 : #include "xfs_trans_resv.h"
12 : #include "xfs_mount.h"
13 : #include "xfs_inode.h"
14 : #include "xfs_btree.h"
15 : #include "xfs_ialloc.h"
16 : #include "xfs_ialloc_btree.h"
17 : #include "xfs_iwalk.h"
18 : #include "xfs_itable.h"
19 : #include "xfs_error.h"
20 : #include "xfs_icache.h"
21 : #include "xfs_health.h"
22 : #include "xfs_trans.h"
23 :
24 : /*
25 : * Bulk Stat
26 : * =========
27 : *
28 : * Use the inode walking functions to fill out struct xfs_bulkstat for every
29 : * allocated inode, then pass the stat information to some externally provided
30 : * iteration function.
31 : */
32 :
33 : struct xfs_bstat_chunk {
34 : bulkstat_one_fmt_pf formatter;
35 : struct xfs_ibulk *breq;
36 : struct xfs_bulkstat *buf;
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
56 72152196266 : xfs_bulkstat_one_int(
57 : struct xfs_mount *mp,
58 : struct mnt_idmap *idmap,
59 : struct xfs_trans *tp,
60 : xfs_ino_t ino,
61 : struct xfs_bstat_chunk *bc)
62 : {
63 72152196266 : struct user_namespace *sb_userns = mp->m_super->s_user_ns;
64 72152196266 : struct xfs_inode *ip; /* incore inode pointer */
65 72152196266 : struct inode *inode;
66 72152196266 : struct xfs_bulkstat *buf = bc->buf;
67 72152196266 : xfs_extnum_t nextents;
68 72152196266 : int error = -EINVAL;
69 72152196266 : vfsuid_t vfsuid;
70 72152196266 : vfsgid_t vfsgid;
71 :
72 72152196266 : if (xfs_internal_inum(mp, ino))
73 7223448 : goto out_advance;
74 :
75 72133422910 : error = xfs_iget(mp, tp, ino,
76 : (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
77 : XFS_ILOCK_SHARED, &ip);
78 73063276634 : if (error == -ENOENT || error == -EINVAL)
79 2392466 : goto out_advance;
80 73060884168 : if (error)
81 3022 : goto out;
82 :
83 73060881146 : ASSERT(ip != NULL);
84 73060881146 : ASSERT(ip->i_imap.im_blkno != 0);
85 73060881146 : inode = VFS_I(ip);
86 73060881146 : vfsuid = i_uid_into_vfsuid(idmap, inode);
87 72762031761 : vfsgid = i_gid_into_vfsgid(idmap, inode);
88 :
89 : /* If this is a private inode, don't leak its details to userspace. */
90 72650574302 : if (IS_PRIVATE(inode)) {
91 91953 : xfs_iunlock(ip, XFS_ILOCK_SHARED);
92 91924 : xfs_irele(ip);
93 91948 : error = -EINVAL;
94 91948 : goto out_advance;
95 : }
96 :
97 : /* xfs_iget returns the following without needing
98 : * further change.
99 : */
100 72650482349 : buf->bs_projectid = ip->i_projid;
101 72650482349 : buf->bs_ino = ino;
102 72650482349 : buf->bs_uid = from_kuid(sb_userns, vfsuid_into_kuid(vfsuid));
103 72709226862 : buf->bs_gid = from_kgid(sb_userns, vfsgid_into_kgid(vfsgid));
104 72776251397 : buf->bs_size = ip->i_disk_size;
105 :
106 72776251397 : buf->bs_nlink = inode->i_nlink;
107 72776251397 : buf->bs_atime = inode->i_atime.tv_sec;
108 72776251397 : buf->bs_atime_nsec = inode->i_atime.tv_nsec;
109 72776251397 : buf->bs_mtime = inode->i_mtime.tv_sec;
110 72776251397 : buf->bs_mtime_nsec = inode->i_mtime.tv_nsec;
111 72776251397 : buf->bs_ctime = inode->i_ctime.tv_sec;
112 72776251397 : buf->bs_ctime_nsec = inode->i_ctime.tv_nsec;
113 72776251397 : buf->bs_gen = inode->i_generation;
114 72776251397 : buf->bs_mode = inode->i_mode;
115 :
116 72776251397 : buf->bs_xflags = xfs_ip2xflags(ip);
117 72378490227 : buf->bs_extsize_blks = ip->i_extsize;
118 :
119 72378490227 : nextents = xfs_ifork_nextents(&ip->i_df);
120 72378490227 : if (!(bc->breq->flags & XFS_IBULK_NREXT64))
121 72040333025 : buf->bs_extents = min(nextents, XFS_MAX_EXTCNT_DATA_FORK_SMALL);
122 : else
123 338157202 : buf->bs_extents64 = nextents;
124 :
125 72378490227 : xfs_bulkstat_health(ip, buf);
126 73364525327 : buf->bs_aextents = xfs_ifork_nextents(&ip->i_af);
127 73364525327 : buf->bs_forkoff = xfs_inode_fork_boff(ip);
128 73364525327 : buf->bs_version = XFS_BULKSTAT_VERSION_V5;
129 :
130 73364525327 : if (xfs_has_v3inodes(mp)) {
131 72259098787 : buf->bs_btime = ip->i_crtime.tv_sec;
132 72259098787 : buf->bs_btime_nsec = ip->i_crtime.tv_nsec;
133 72259098787 : if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
134 17716 : buf->bs_cowextsize_blks = ip->i_cowextsize;
135 : }
136 :
137 73364525327 : switch (ip->i_df.if_format) {
138 27265607787 : case XFS_DINODE_FMT_DEV:
139 27265607787 : buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
140 27265607787 : buf->bs_blksize = BLKDEV_IOSIZE;
141 27265607787 : buf->bs_blocks = 0;
142 27265607787 : break;
143 14160749828 : case XFS_DINODE_FMT_LOCAL:
144 14160749828 : buf->bs_rdev = 0;
145 14160749828 : buf->bs_blksize = mp->m_sb.sb_blocksize;
146 14160749828 : buf->bs_blocks = 0;
147 14160749828 : break;
148 31938167712 : case XFS_DINODE_FMT_EXTENTS:
149 : case XFS_DINODE_FMT_BTREE:
150 31938167712 : buf->bs_rdev = 0;
151 31938167712 : buf->bs_blksize = mp->m_sb.sb_blocksize;
152 31938167712 : buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks;
153 31938167712 : break;
154 : }
155 73364525327 : xfs_iunlock(ip, XFS_ILOCK_SHARED);
156 71679568588 : xfs_irele(ip);
157 :
158 72959087013 : error = bc->formatter(bc->breq, buf);
159 72115152039 : if (error == -ECANCELED)
160 484366933 : goto out_advance;
161 71630785106 : if (error)
162 0 : goto out;
163 :
164 71630785106 : out_advance:
165 : /*
166 : * Advance the cursor to the inode that comes after the one we just
167 : * looked at. We want the caller to move along if the bulkstat
168 : * information was copied successfully; if we tried to grab the inode
169 : * but it's no longer allocated; or if it's internal metadata.
170 : */
171 72124859901 : bc->breq->startino = ino + 1;
172 72124862923 : out:
173 72124862923 : return error;
174 : }
175 :
176 : /* Bulkstat a single inode. */
177 : int
178 1808270 : xfs_bulkstat_one(
179 : struct xfs_ibulk *breq,
180 : bulkstat_one_fmt_pf formatter)
181 : {
182 1808270 : struct xfs_bstat_chunk bc = {
183 : .formatter = formatter,
184 : .breq = breq,
185 : };
186 1808270 : struct xfs_trans *tp;
187 1808270 : int error;
188 :
189 1808270 : if (breq->idmap != &nop_mnt_idmap) {
190 0 : xfs_warn_ratelimited(breq->mp,
191 : "bulkstat not supported inside of idmapped mounts.");
192 0 : return -EINVAL;
193 : }
194 :
195 1808270 : ASSERT(breq->icount == 1);
196 :
197 1808270 : bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
198 : KM_MAYFAIL);
199 1808307 : if (!bc.buf)
200 : return -ENOMEM;
201 :
202 : /*
203 : * Grab an empty transaction so that we can use its recursive buffer
204 : * locking abilities to detect cycles in the inobt without deadlocking.
205 : */
206 1808307 : error = xfs_trans_alloc_empty(breq->mp, &tp);
207 1808477 : if (error)
208 0 : goto out;
209 :
210 1808477 : error = xfs_bulkstat_one_int(breq->mp, breq->idmap, tp,
211 : breq->startino, &bc);
212 1808461 : xfs_trans_cancel(tp);
213 1808577 : out:
214 1808577 : kmem_free(bc.buf);
215 :
216 : /*
217 : * If we reported one inode to userspace then we abort because we hit
218 : * the end of the buffer. Don't leak that back to userspace.
219 : */
220 1808553 : if (error == -ECANCELED)
221 890581 : error = 0;
222 :
223 : return error;
224 : }
225 :
226 : static int
227 71908731841 : xfs_bulkstat_iwalk(
228 : struct xfs_mount *mp,
229 : struct xfs_trans *tp,
230 : xfs_ino_t ino,
231 : void *data)
232 : {
233 71908731841 : struct xfs_bstat_chunk *bc = data;
234 71908731841 : int error;
235 :
236 71908731841 : error = xfs_bulkstat_one_int(mp, bc->breq->idmap, tp, ino, data);
237 : /* bulkstat just skips over missing inodes */
238 72235751192 : if (error == -ENOENT || error == -EINVAL)
239 8789893 : return 0;
240 : return error;
241 : }
242 :
243 : /*
244 : * Check the incoming lastino parameter.
245 : *
246 : * We allow any inode value that could map to physical space inside the
247 : * filesystem because if there are no inodes there, bulkstat moves on to the
248 : * next chunk. In other words, the magic agino value of zero takes us to the
249 : * first chunk in the AG, and an agino value past the end of the AG takes us to
250 : * the first chunk in the next AG.
251 : *
252 : * Therefore we can end early if the requested inode is beyond the end of the
253 : * filesystem or doesn't map properly.
254 : */
255 : static inline bool
256 492620133 : xfs_bulkstat_already_done(
257 : struct xfs_mount *mp,
258 : xfs_ino_t startino)
259 : {
260 492620133 : xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino);
261 492620133 : xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino);
262 :
263 492620133 : return agno >= mp->m_sb.sb_agcount ||
264 492600223 : startino != XFS_AGINO_TO_INO(mp, agno, agino);
265 : }
266 :
267 : /* Return stat information in bulk (by-inode) for the filesystem. */
268 : int
269 486958043 : xfs_bulkstat(
270 : struct xfs_ibulk *breq,
271 : bulkstat_one_fmt_pf formatter)
272 : {
273 486958043 : struct xfs_bstat_chunk bc = {
274 : .formatter = formatter,
275 : .breq = breq,
276 : };
277 486958043 : struct xfs_trans *tp;
278 486958043 : unsigned int iwalk_flags = 0;
279 486958043 : int error;
280 :
281 486958043 : if (breq->idmap != &nop_mnt_idmap) {
282 0 : xfs_warn_ratelimited(breq->mp,
283 : "bulkstat not supported inside of idmapped mounts.");
284 0 : return -EINVAL;
285 : }
286 486958043 : if (xfs_bulkstat_already_done(breq->mp, breq->startino))
287 : return 0;
288 :
289 486940741 : bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
290 : KM_MAYFAIL);
291 486946199 : if (!bc.buf)
292 : return -ENOMEM;
293 :
294 : /*
295 : * Grab an empty transaction so that we can use its recursive buffer
296 : * locking abilities to detect cycles in the inobt without deadlocking.
297 : */
298 486946199 : error = xfs_trans_alloc_empty(breq->mp, &tp);
299 486971883 : if (error)
300 0 : goto out;
301 :
302 486971883 : if (breq->flags & XFS_IBULK_SAME_AG)
303 87736 : iwalk_flags |= XFS_IWALK_SAME_AG;
304 :
305 486971883 : error = xfs_iwalk(breq->mp, tp, breq->startino, iwalk_flags,
306 : xfs_bulkstat_iwalk, breq->icount, &bc);
307 487016426 : xfs_trans_cancel(tp);
308 487027407 : out:
309 487027407 : kmem_free(bc.buf);
310 :
311 : /*
312 : * We found some inodes, so clear the error status and return them.
313 : * The lastino pointer will point directly at the inode that triggered
314 : * any error that occurred, so on the next call the error will be
315 : * triggered again and propagated to userspace as there will be no
316 : * formatted inodes in the buffer.
317 : */
318 487032977 : if (breq->ocount > 0)
319 485277181 : error = 0;
320 :
321 : return error;
322 : }
323 :
324 : /* Convert bulkstat (v5) to bstat (v1). */
325 : void
326 72173953102 : xfs_bulkstat_to_bstat(
327 : struct xfs_mount *mp,
328 : struct xfs_bstat *bs1,
329 : const struct xfs_bulkstat *bstat)
330 : {
331 : /* memset is needed here because of padding holes in the structure. */
332 72173953102 : memset(bs1, 0, sizeof(struct xfs_bstat));
333 72173953102 : bs1->bs_ino = bstat->bs_ino;
334 72173953102 : bs1->bs_mode = bstat->bs_mode;
335 72173953102 : bs1->bs_nlink = bstat->bs_nlink;
336 72173953102 : bs1->bs_uid = bstat->bs_uid;
337 72173953102 : bs1->bs_gid = bstat->bs_gid;
338 72173953102 : bs1->bs_rdev = bstat->bs_rdev;
339 72173953102 : bs1->bs_blksize = bstat->bs_blksize;
340 72173953102 : bs1->bs_size = bstat->bs_size;
341 72173953102 : bs1->bs_atime.tv_sec = bstat->bs_atime;
342 72173953102 : bs1->bs_mtime.tv_sec = bstat->bs_mtime;
343 72173953102 : bs1->bs_ctime.tv_sec = bstat->bs_ctime;
344 72173953102 : bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
345 72173953102 : bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
346 72173953102 : bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
347 72173953102 : bs1->bs_blocks = bstat->bs_blocks;
348 72173953102 : bs1->bs_xflags = bstat->bs_xflags;
349 72173953102 : bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
350 72173953102 : bs1->bs_extents = bstat->bs_extents;
351 72173953102 : bs1->bs_gen = bstat->bs_gen;
352 72173953102 : bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
353 72173953102 : bs1->bs_forkoff = bstat->bs_forkoff;
354 72173953102 : bs1->bs_projid_hi = bstat->bs_projectid >> 16;
355 72173953102 : bs1->bs_sick = bstat->bs_sick;
356 72173953102 : bs1->bs_checked = bstat->bs_checked;
357 72173953102 : bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
358 72173953102 : bs1->bs_dmevmask = 0;
359 72173953102 : bs1->bs_dmstate = 0;
360 72173953102 : bs1->bs_aextents = bstat->bs_aextents;
361 72173953102 : }
362 :
363 : struct xfs_inumbers_chunk {
364 : inumbers_fmt_pf formatter;
365 : struct xfs_ibulk *breq;
366 : };
367 :
368 : /*
369 : * INUMBERS
370 : * ========
371 : * This is how we export inode btree records to userspace, so that XFS tools
372 : * can figure out where inodes are allocated.
373 : */
374 :
375 : /*
376 : * Format the inode group structure and report it somewhere.
377 : *
378 : * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
379 : * through the filesystem so we move it forward unless there was a runtime
380 : * error. If the formatter tells us the buffer is now full we also move the
381 : * cursor forward and abort the walk.
382 : */
383 : STATIC int
384 8101199 : xfs_inumbers_walk(
385 : struct xfs_mount *mp,
386 : struct xfs_trans *tp,
387 : xfs_agnumber_t agno,
388 : const struct xfs_inobt_rec_incore *irec,
389 : void *data)
390 : {
391 16202398 : struct xfs_inumbers inogrp = {
392 8101199 : .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
393 8101199 : .xi_alloccount = irec->ir_count - irec->ir_freecount,
394 8101199 : .xi_allocmask = ~irec->ir_free,
395 : .xi_version = XFS_INUMBERS_VERSION_V5,
396 : };
397 8101199 : struct xfs_inumbers_chunk *ic = data;
398 8101199 : int error;
399 :
400 8101199 : error = ic->formatter(ic->breq, &inogrp);
401 8127993 : if (error && error != -ECANCELED)
402 : return error;
403 :
404 8127993 : ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
405 : XFS_INODES_PER_CHUNK;
406 8127993 : return error;
407 : }
408 :
409 : /*
410 : * Return inode number table for the filesystem.
411 : */
412 : int
413 5669925 : xfs_inumbers(
414 : struct xfs_ibulk *breq,
415 : inumbers_fmt_pf formatter)
416 : {
417 5669925 : struct xfs_inumbers_chunk ic = {
418 : .formatter = formatter,
419 : .breq = breq,
420 : };
421 5669925 : struct xfs_trans *tp;
422 5669925 : int error = 0;
423 :
424 5669925 : if (xfs_bulkstat_already_done(breq->mp, breq->startino))
425 : return 0;
426 :
427 : /*
428 : * Grab an empty transaction so that we can use its recursive buffer
429 : * locking abilities to detect cycles in the inobt without deadlocking.
430 : */
431 5663208 : error = xfs_trans_alloc_empty(breq->mp, &tp);
432 5667586 : if (error)
433 0 : goto out;
434 :
435 5667586 : error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags,
436 : xfs_inumbers_walk, breq->icount, &ic);
437 5669755 : xfs_trans_cancel(tp);
438 5661544 : out:
439 :
440 : /*
441 : * We found some inode groups, so clear the error status and return
442 : * them. The lastino pointer will point directly at the inode that
443 : * triggered any error that occurred, so on the next call the error
444 : * will be triggered again and propagated to userspace as there will be
445 : * no formatted inode groups in the buffer.
446 : */
447 5661544 : if (breq->ocount > 0)
448 5226362 : error = 0;
449 :
450 : return error;
451 : }
452 :
453 : /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
454 : void
455 71975 : xfs_inumbers_to_inogrp(
456 : struct xfs_inogrp *ig1,
457 : const struct xfs_inumbers *ig)
458 : {
459 : /* memset is needed here because of padding holes in the structure. */
460 71975 : memset(ig1, 0, sizeof(struct xfs_inogrp));
461 71975 : ig1->xi_startino = ig->xi_startino;
462 71975 : ig1->xi_alloccount = ig->xi_alloccount;
463 71975 : ig1->xi_allocmask = ig->xi_allocmask;
464 71975 : }
|