Loading...
1/*
2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
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
7 * published by the Free Software Foundation.
8 *
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.
13 *
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
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_shared.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_inum.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_mount.h"
28#include "xfs_inode.h"
29#include "xfs_btree.h"
30#include "xfs_ialloc.h"
31#include "xfs_ialloc_btree.h"
32#include "xfs_itable.h"
33#include "xfs_error.h"
34#include "xfs_trace.h"
35#include "xfs_icache.h"
36#include "xfs_dinode.h"
37
38STATIC int
39xfs_internal_inum(
40 xfs_mount_t *mp,
41 xfs_ino_t ino)
42{
43 return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino ||
44 (xfs_sb_version_hasquota(&mp->m_sb) &&
45 xfs_is_quota_inode(&mp->m_sb, ino)));
46}
47
48/*
49 * Return stat information for one inode.
50 * Return 0 if ok, else errno.
51 */
52int
53xfs_bulkstat_one_int(
54 struct xfs_mount *mp, /* mount point for filesystem */
55 xfs_ino_t ino, /* inode to get data for */
56 void __user *buffer, /* buffer to place output in */
57 int ubsize, /* size of buffer */
58 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */
59 int *ubused, /* bytes used by me */
60 int *stat) /* BULKSTAT_RV_... */
61{
62 struct xfs_icdinode *dic; /* dinode core info pointer */
63 struct xfs_inode *ip; /* incore inode pointer */
64 struct xfs_bstat *buf; /* return buffer */
65 int error = 0; /* error value */
66
67 *stat = BULKSTAT_RV_NOTHING;
68
69 if (!buffer || xfs_internal_inum(mp, ino))
70 return XFS_ERROR(EINVAL);
71
72 buf = kmem_alloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL);
73 if (!buf)
74 return XFS_ERROR(ENOMEM);
75
76 error = xfs_iget(mp, NULL, ino,
77 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
78 XFS_ILOCK_SHARED, &ip);
79 if (error) {
80 *stat = BULKSTAT_RV_NOTHING;
81 goto out_free;
82 }
83
84 ASSERT(ip != NULL);
85 ASSERT(ip->i_imap.im_blkno != 0);
86
87 dic = &ip->i_d;
88
89 /* xfs_iget returns the following without needing
90 * further change.
91 */
92 buf->bs_nlink = dic->di_nlink;
93 buf->bs_projid_lo = dic->di_projid_lo;
94 buf->bs_projid_hi = dic->di_projid_hi;
95 buf->bs_ino = ino;
96 buf->bs_mode = dic->di_mode;
97 buf->bs_uid = dic->di_uid;
98 buf->bs_gid = dic->di_gid;
99 buf->bs_size = dic->di_size;
100 buf->bs_atime.tv_sec = dic->di_atime.t_sec;
101 buf->bs_atime.tv_nsec = dic->di_atime.t_nsec;
102 buf->bs_mtime.tv_sec = dic->di_mtime.t_sec;
103 buf->bs_mtime.tv_nsec = dic->di_mtime.t_nsec;
104 buf->bs_ctime.tv_sec = dic->di_ctime.t_sec;
105 buf->bs_ctime.tv_nsec = dic->di_ctime.t_nsec;
106 buf->bs_xflags = xfs_ip2xflags(ip);
107 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
108 buf->bs_extents = dic->di_nextents;
109 buf->bs_gen = dic->di_gen;
110 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
111 buf->bs_dmevmask = dic->di_dmevmask;
112 buf->bs_dmstate = dic->di_dmstate;
113 buf->bs_aextents = dic->di_anextents;
114 buf->bs_forkoff = XFS_IFORK_BOFF(ip);
115
116 switch (dic->di_format) {
117 case XFS_DINODE_FMT_DEV:
118 buf->bs_rdev = ip->i_df.if_u2.if_rdev;
119 buf->bs_blksize = BLKDEV_IOSIZE;
120 buf->bs_blocks = 0;
121 break;
122 case XFS_DINODE_FMT_LOCAL:
123 case XFS_DINODE_FMT_UUID:
124 buf->bs_rdev = 0;
125 buf->bs_blksize = mp->m_sb.sb_blocksize;
126 buf->bs_blocks = 0;
127 break;
128 case XFS_DINODE_FMT_EXTENTS:
129 case XFS_DINODE_FMT_BTREE:
130 buf->bs_rdev = 0;
131 buf->bs_blksize = mp->m_sb.sb_blocksize;
132 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
133 break;
134 }
135 xfs_iunlock(ip, XFS_ILOCK_SHARED);
136 IRELE(ip);
137
138 error = formatter(buffer, ubsize, ubused, buf);
139
140 if (!error)
141 *stat = BULKSTAT_RV_DIDONE;
142
143 out_free:
144 kmem_free(buf);
145 return error;
146}
147
148/* Return 0 on success or positive error */
149STATIC int
150xfs_bulkstat_one_fmt(
151 void __user *ubuffer,
152 int ubsize,
153 int *ubused,
154 const xfs_bstat_t *buffer)
155{
156 if (ubsize < sizeof(*buffer))
157 return XFS_ERROR(ENOMEM);
158 if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
159 return XFS_ERROR(EFAULT);
160 if (ubused)
161 *ubused = sizeof(*buffer);
162 return 0;
163}
164
165int
166xfs_bulkstat_one(
167 xfs_mount_t *mp, /* mount point for filesystem */
168 xfs_ino_t ino, /* inode number to get data for */
169 void __user *buffer, /* buffer to place output in */
170 int ubsize, /* size of buffer */
171 int *ubused, /* bytes used by me */
172 int *stat) /* BULKSTAT_RV_... */
173{
174 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
175 xfs_bulkstat_one_fmt, ubused, stat);
176}
177
178#define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size)
179
180/*
181 * Return stat information in bulk (by-inode) for the filesystem.
182 */
183int /* error status */
184xfs_bulkstat(
185 xfs_mount_t *mp, /* mount point for filesystem */
186 xfs_ino_t *lastinop, /* last inode returned */
187 int *ubcountp, /* size of buffer/count returned */
188 bulkstat_one_pf formatter, /* func that'd fill a single buf */
189 size_t statstruct_size, /* sizeof struct filling */
190 char __user *ubuffer, /* buffer with inode stats */
191 int *done) /* 1 if there are more stats to get */
192{
193 xfs_agblock_t agbno=0;/* allocation group block number */
194 xfs_buf_t *agbp; /* agi header buffer */
195 xfs_agi_t *agi; /* agi header data */
196 xfs_agino_t agino; /* inode # in allocation group */
197 xfs_agnumber_t agno; /* allocation group number */
198 int chunkidx; /* current index into inode chunk */
199 int clustidx; /* current index into inode cluster */
200 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */
201 int end_of_ag; /* set if we've seen the ag end */
202 int error; /* error code */
203 int fmterror;/* bulkstat formatter result */
204 int i; /* loop index */
205 int icount; /* count of inodes good in irbuf */
206 size_t irbsize; /* size of irec buffer in bytes */
207 xfs_ino_t ino; /* inode number (filesystem) */
208 xfs_inobt_rec_incore_t *irbp; /* current irec buffer pointer */
209 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */
210 xfs_inobt_rec_incore_t *irbufend; /* end of good irec buffer entries */
211 xfs_ino_t lastino; /* last inode number returned */
212 int blks_per_cluster; /* # of blocks per cluster */
213 int inodes_per_cluster;/* # of inodes per cluster */
214 int nirbuf; /* size of irbuf */
215 int rval; /* return value error code */
216 int tmp; /* result value from btree calls */
217 int ubcount; /* size of user's buffer */
218 int ubleft; /* bytes left in user's buffer */
219 char __user *ubufp; /* pointer into user's buffer */
220 int ubelem; /* spaces used in user's buffer */
221 int ubused; /* bytes used by formatter */
222
223 /*
224 * Get the last inode value, see if there's nothing to do.
225 */
226 ino = (xfs_ino_t)*lastinop;
227 lastino = ino;
228 agno = XFS_INO_TO_AGNO(mp, ino);
229 agino = XFS_INO_TO_AGINO(mp, ino);
230 if (agno >= mp->m_sb.sb_agcount ||
231 ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
232 *done = 1;
233 *ubcountp = 0;
234 return 0;
235 }
236 if (!ubcountp || *ubcountp <= 0) {
237 return EINVAL;
238 }
239 ubcount = *ubcountp; /* statstruct's */
240 ubleft = ubcount * statstruct_size; /* bytes */
241 *ubcountp = ubelem = 0;
242 *done = 0;
243 fmterror = 0;
244 ubufp = ubuffer;
245 blks_per_cluster = xfs_icluster_size_fsb(mp);
246 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
247 irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4);
248 if (!irbuf)
249 return ENOMEM;
250
251 nirbuf = irbsize / sizeof(*irbuf);
252
253 /*
254 * Loop over the allocation groups, starting from the last
255 * inode returned; 0 means start of the allocation group.
256 */
257 rval = 0;
258 while (XFS_BULKSTAT_UBLEFT(ubleft) && agno < mp->m_sb.sb_agcount) {
259 cond_resched();
260 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
261 if (error) {
262 /*
263 * Skip this allocation group and go to the next one.
264 */
265 agno++;
266 agino = 0;
267 continue;
268 }
269 agi = XFS_BUF_TO_AGI(agbp);
270 /*
271 * Allocate and initialize a btree cursor for ialloc btree.
272 */
273 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
274 irbp = irbuf;
275 irbufend = irbuf + nirbuf;
276 end_of_ag = 0;
277 /*
278 * If we're returning in the middle of an allocation group,
279 * we need to get the remainder of the chunk we're in.
280 */
281 if (agino > 0) {
282 xfs_inobt_rec_incore_t r;
283
284 /*
285 * Lookup the inode chunk that this inode lives in.
286 */
287 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE,
288 &tmp);
289 if (!error && /* no I/O error */
290 tmp && /* lookup succeeded */
291 /* got the record, should always work */
292 !(error = xfs_inobt_get_rec(cur, &r, &i)) &&
293 i == 1 &&
294 /* this is the right chunk */
295 agino < r.ir_startino + XFS_INODES_PER_CHUNK &&
296 /* lastino was not last in chunk */
297 (chunkidx = agino - r.ir_startino + 1) <
298 XFS_INODES_PER_CHUNK &&
299 /* there are some left allocated */
300 xfs_inobt_maskn(chunkidx,
301 XFS_INODES_PER_CHUNK - chunkidx) &
302 ~r.ir_free) {
303 /*
304 * Grab the chunk record. Mark all the
305 * uninteresting inodes (because they're
306 * before our start point) free.
307 */
308 for (i = 0; i < chunkidx; i++) {
309 if (XFS_INOBT_MASK(i) & ~r.ir_free)
310 r.ir_freecount++;
311 }
312 r.ir_free |= xfs_inobt_maskn(0, chunkidx);
313 irbp->ir_startino = r.ir_startino;
314 irbp->ir_freecount = r.ir_freecount;
315 irbp->ir_free = r.ir_free;
316 irbp++;
317 agino = r.ir_startino + XFS_INODES_PER_CHUNK;
318 icount = XFS_INODES_PER_CHUNK - r.ir_freecount;
319 } else {
320 /*
321 * If any of those tests failed, bump the
322 * inode number (just in case).
323 */
324 agino++;
325 icount = 0;
326 }
327 /*
328 * In any case, increment to the next record.
329 */
330 if (!error)
331 error = xfs_btree_increment(cur, 0, &tmp);
332 } else {
333 /*
334 * Start of ag. Lookup the first inode chunk.
335 */
336 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &tmp);
337 icount = 0;
338 }
339 /*
340 * Loop through inode btree records in this ag,
341 * until we run out of inodes or space in the buffer.
342 */
343 while (irbp < irbufend && icount < ubcount) {
344 xfs_inobt_rec_incore_t r;
345
346 /*
347 * Loop as long as we're unable to read the
348 * inode btree.
349 */
350 while (error) {
351 agino += XFS_INODES_PER_CHUNK;
352 if (XFS_AGINO_TO_AGBNO(mp, agino) >=
353 be32_to_cpu(agi->agi_length))
354 break;
355 error = xfs_inobt_lookup(cur, agino,
356 XFS_LOOKUP_GE, &tmp);
357 cond_resched();
358 }
359 /*
360 * If ran off the end of the ag either with an error,
361 * or the normal way, set end and stop collecting.
362 */
363 if (error) {
364 end_of_ag = 1;
365 break;
366 }
367
368 error = xfs_inobt_get_rec(cur, &r, &i);
369 if (error || i == 0) {
370 end_of_ag = 1;
371 break;
372 }
373
374 /*
375 * If this chunk has any allocated inodes, save it.
376 * Also start read-ahead now for this chunk.
377 */
378 if (r.ir_freecount < XFS_INODES_PER_CHUNK) {
379 struct blk_plug plug;
380 /*
381 * Loop over all clusters in the next chunk.
382 * Do a readahead if there are any allocated
383 * inodes in that cluster.
384 */
385 blk_start_plug(&plug);
386 agbno = XFS_AGINO_TO_AGBNO(mp, r.ir_startino);
387 for (chunkidx = 0;
388 chunkidx < XFS_INODES_PER_CHUNK;
389 chunkidx += inodes_per_cluster,
390 agbno += blks_per_cluster) {
391 if (xfs_inobt_maskn(chunkidx,
392 inodes_per_cluster) & ~r.ir_free)
393 xfs_btree_reada_bufs(mp, agno,
394 agbno, blks_per_cluster,
395 &xfs_inode_buf_ops);
396 }
397 blk_finish_plug(&plug);
398 irbp->ir_startino = r.ir_startino;
399 irbp->ir_freecount = r.ir_freecount;
400 irbp->ir_free = r.ir_free;
401 irbp++;
402 icount += XFS_INODES_PER_CHUNK - r.ir_freecount;
403 }
404 /*
405 * Set agino to after this chunk and bump the cursor.
406 */
407 agino = r.ir_startino + XFS_INODES_PER_CHUNK;
408 error = xfs_btree_increment(cur, 0, &tmp);
409 cond_resched();
410 }
411 /*
412 * Drop the btree buffers and the agi buffer.
413 * We can't hold any of the locks these represent
414 * when calling iget.
415 */
416 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
417 xfs_buf_relse(agbp);
418 /*
419 * Now format all the good inodes into the user's buffer.
420 */
421 irbufend = irbp;
422 for (irbp = irbuf;
423 irbp < irbufend && XFS_BULKSTAT_UBLEFT(ubleft); irbp++) {
424 /*
425 * Now process this chunk of inodes.
426 */
427 for (agino = irbp->ir_startino, chunkidx = clustidx = 0;
428 XFS_BULKSTAT_UBLEFT(ubleft) &&
429 irbp->ir_freecount < XFS_INODES_PER_CHUNK;
430 chunkidx++, clustidx++, agino++) {
431 ASSERT(chunkidx < XFS_INODES_PER_CHUNK);
432
433 ino = XFS_AGINO_TO_INO(mp, agno, agino);
434 /*
435 * Skip if this inode is free.
436 */
437 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) {
438 lastino = ino;
439 continue;
440 }
441 /*
442 * Count used inodes as free so we can tell
443 * when the chunk is used up.
444 */
445 irbp->ir_freecount++;
446
447 /*
448 * Get the inode and fill in a single buffer.
449 */
450 ubused = statstruct_size;
451 error = formatter(mp, ino, ubufp, ubleft,
452 &ubused, &fmterror);
453 if (fmterror == BULKSTAT_RV_NOTHING) {
454 if (error && error != ENOENT &&
455 error != EINVAL) {
456 ubleft = 0;
457 rval = error;
458 break;
459 }
460 lastino = ino;
461 continue;
462 }
463 if (fmterror == BULKSTAT_RV_GIVEUP) {
464 ubleft = 0;
465 ASSERT(error);
466 rval = error;
467 break;
468 }
469 if (ubufp)
470 ubufp += ubused;
471 ubleft -= ubused;
472 ubelem++;
473 lastino = ino;
474 }
475
476 cond_resched();
477 }
478 /*
479 * Set up for the next loop iteration.
480 */
481 if (XFS_BULKSTAT_UBLEFT(ubleft)) {
482 if (end_of_ag) {
483 agno++;
484 agino = 0;
485 } else
486 agino = XFS_INO_TO_AGINO(mp, lastino);
487 } else
488 break;
489 }
490 /*
491 * Done, we're either out of filesystem or space to put the data.
492 */
493 kmem_free(irbuf);
494 *ubcountp = ubelem;
495 /*
496 * Found some inodes, return them now and return the error next time.
497 */
498 if (ubelem)
499 rval = 0;
500 if (agno >= mp->m_sb.sb_agcount) {
501 /*
502 * If we ran out of filesystem, mark lastino as off
503 * the end of the filesystem, so the next call
504 * will return immediately.
505 */
506 *lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0);
507 *done = 1;
508 } else
509 *lastinop = (xfs_ino_t)lastino;
510
511 return rval;
512}
513
514/*
515 * Return stat information in bulk (by-inode) for the filesystem.
516 * Special case for non-sequential one inode bulkstat.
517 */
518int /* error status */
519xfs_bulkstat_single(
520 xfs_mount_t *mp, /* mount point for filesystem */
521 xfs_ino_t *lastinop, /* inode to return */
522 char __user *buffer, /* buffer with inode stats */
523 int *done) /* 1 if there are more stats to get */
524{
525 int count; /* count value for bulkstat call */
526 int error; /* return value */
527 xfs_ino_t ino; /* filesystem inode number */
528 int res; /* result from bs1 */
529
530 /*
531 * note that requesting valid inode numbers which are not allocated
532 * to inodes will most likely cause xfs_imap_to_bp to generate warning
533 * messages about bad magic numbers. This is ok. The fact that
534 * the inode isn't actually an inode is handled by the
535 * error check below. Done this way to make the usual case faster
536 * at the expense of the error case.
537 */
538
539 ino = *lastinop;
540 error = xfs_bulkstat_one(mp, ino, buffer, sizeof(xfs_bstat_t),
541 NULL, &res);
542 if (error) {
543 /*
544 * Special case way failed, do it the "long" way
545 * to see if that works.
546 */
547 (*lastinop)--;
548 count = 1;
549 if (xfs_bulkstat(mp, lastinop, &count, xfs_bulkstat_one,
550 sizeof(xfs_bstat_t), buffer, done))
551 return error;
552 if (count == 0 || (xfs_ino_t)*lastinop != ino)
553 return error == EFSCORRUPTED ?
554 XFS_ERROR(EINVAL) : error;
555 else
556 return 0;
557 }
558 *done = 0;
559 return 0;
560}
561
562int
563xfs_inumbers_fmt(
564 void __user *ubuffer, /* buffer to write to */
565 const xfs_inogrp_t *buffer, /* buffer to read from */
566 long count, /* # of elements to read */
567 long *written) /* # of bytes written */
568{
569 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
570 return -EFAULT;
571 *written = count * sizeof(*buffer);
572 return 0;
573}
574
575/*
576 * Return inode number table for the filesystem.
577 */
578int /* error status */
579xfs_inumbers(
580 xfs_mount_t *mp, /* mount point for filesystem */
581 xfs_ino_t *lastino, /* last inode returned */
582 int *count, /* size of buffer/count returned */
583 void __user *ubuffer,/* buffer with inode descriptions */
584 inumbers_fmt_pf formatter)
585{
586 xfs_buf_t *agbp;
587 xfs_agino_t agino;
588 xfs_agnumber_t agno;
589 int bcount;
590 xfs_inogrp_t *buffer;
591 int bufidx;
592 xfs_btree_cur_t *cur;
593 int error;
594 xfs_inobt_rec_incore_t r;
595 int i;
596 xfs_ino_t ino;
597 int left;
598 int tmp;
599
600 ino = (xfs_ino_t)*lastino;
601 agno = XFS_INO_TO_AGNO(mp, ino);
602 agino = XFS_INO_TO_AGINO(mp, ino);
603 left = *count;
604 *count = 0;
605 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
606 buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
607 error = bufidx = 0;
608 cur = NULL;
609 agbp = NULL;
610 while (left > 0 && agno < mp->m_sb.sb_agcount) {
611 if (agbp == NULL) {
612 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
613 if (error) {
614 /*
615 * If we can't read the AGI of this ag,
616 * then just skip to the next one.
617 */
618 ASSERT(cur == NULL);
619 agbp = NULL;
620 agno++;
621 agino = 0;
622 continue;
623 }
624 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
625 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
626 &tmp);
627 if (error) {
628 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
629 cur = NULL;
630 xfs_buf_relse(agbp);
631 agbp = NULL;
632 /*
633 * Move up the last inode in the current
634 * chunk. The lookup_ge will always get
635 * us the first inode in the next chunk.
636 */
637 agino += XFS_INODES_PER_CHUNK - 1;
638 continue;
639 }
640 }
641 error = xfs_inobt_get_rec(cur, &r, &i);
642 if (error || i == 0) {
643 xfs_buf_relse(agbp);
644 agbp = NULL;
645 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
646 cur = NULL;
647 agno++;
648 agino = 0;
649 continue;
650 }
651 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
652 buffer[bufidx].xi_startino =
653 XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
654 buffer[bufidx].xi_alloccount =
655 XFS_INODES_PER_CHUNK - r.ir_freecount;
656 buffer[bufidx].xi_allocmask = ~r.ir_free;
657 bufidx++;
658 left--;
659 if (bufidx == bcount) {
660 long written;
661 if (formatter(ubuffer, buffer, bufidx, &written)) {
662 error = XFS_ERROR(EFAULT);
663 break;
664 }
665 ubuffer += written;
666 *count += bufidx;
667 bufidx = 0;
668 }
669 if (left) {
670 error = xfs_btree_increment(cur, 0, &tmp);
671 if (error) {
672 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
673 cur = NULL;
674 xfs_buf_relse(agbp);
675 agbp = NULL;
676 /*
677 * The agino value has already been bumped.
678 * Just try to skip up to it.
679 */
680 agino += XFS_INODES_PER_CHUNK;
681 continue;
682 }
683 }
684 }
685 if (!error) {
686 if (bufidx) {
687 long written;
688 if (formatter(ubuffer, buffer, bufidx, &written))
689 error = XFS_ERROR(EFAULT);
690 else
691 *count += bufidx;
692 }
693 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
694 }
695 kmem_free(buffer);
696 if (cur)
697 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
698 XFS_BTREE_NOERROR));
699 if (agbp)
700 xfs_buf_relse(agbp);
701 return error;
702}
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
33struct xfs_bstat_chunk {
34 bulkstat_one_fmt_pf formatter;
35 struct xfs_ibulk *breq;
36 struct xfs_bulkstat *buf;
37};
38
39static inline bool
40want_metadir_file(
41 struct xfs_inode *ip,
42 struct xfs_ibulk *breq)
43{
44 return xfs_is_metadir_inode(ip) && (breq->flags & XFS_IBULK_METADIR);
45}
46
47/*
48 * Fill out the bulkstat info for a single inode and report it somewhere.
49 *
50 * bc->breq->lastino is effectively the inode cursor as we walk through the
51 * filesystem. Therefore, we update it any time we need to move the cursor
52 * forward, regardless of whether or not we're sending any bstat information
53 * back to userspace. If the inode is internal metadata or, has been freed
54 * out from under us, we just simply keep going.
55 *
56 * However, if any other type of error happens we want to stop right where we
57 * are so that userspace will call back with exact number of the bad inode and
58 * we can send back an error code.
59 *
60 * Note that if the formatter tells us there's no space left in the buffer we
61 * move the cursor forward and abort the walk.
62 */
63STATIC int
64xfs_bulkstat_one_int(
65 struct xfs_mount *mp,
66 struct mnt_idmap *idmap,
67 struct xfs_trans *tp,
68 xfs_ino_t ino,
69 struct xfs_bstat_chunk *bc)
70{
71 struct user_namespace *sb_userns = mp->m_super->s_user_ns;
72 struct xfs_inode *ip; /* incore inode pointer */
73 struct inode *inode;
74 struct xfs_bulkstat *buf = bc->buf;
75 xfs_extnum_t nextents;
76 int error = -EINVAL;
77 vfsuid_t vfsuid;
78 vfsgid_t vfsgid;
79
80 error = xfs_iget(mp, tp, ino,
81 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
82 XFS_ILOCK_SHARED, &ip);
83 if (error == -ENOENT || error == -EINVAL)
84 goto out_advance;
85 if (error)
86 goto out;
87
88 /* Reload the incore unlinked list to avoid failure in inodegc. */
89 if (xfs_inode_unlinked_incomplete(ip)) {
90 error = xfs_inode_reload_unlinked_bucket(tp, ip);
91 if (error) {
92 xfs_iunlock(ip, XFS_ILOCK_SHARED);
93 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
94 xfs_irele(ip);
95 return error;
96 }
97 }
98
99 ASSERT(ip != NULL);
100 ASSERT(ip->i_imap.im_blkno != 0);
101 inode = VFS_I(ip);
102 vfsuid = i_uid_into_vfsuid(idmap, inode);
103 vfsgid = i_gid_into_vfsgid(idmap, inode);
104
105 /*
106 * If caller wants files from the metadata directories, push out the
107 * bare minimum information for enabling scrub.
108 */
109 if (want_metadir_file(ip, bc->breq)) {
110 memset(buf, 0, sizeof(*buf));
111 buf->bs_ino = ino;
112 buf->bs_gen = inode->i_generation;
113 buf->bs_mode = inode->i_mode & S_IFMT;
114 xfs_bulkstat_health(ip, buf);
115 buf->bs_version = XFS_BULKSTAT_VERSION_V5;
116 xfs_iunlock(ip, XFS_ILOCK_SHARED);
117 xfs_irele(ip);
118
119 error = bc->formatter(bc->breq, buf);
120 if (!error || error == -ECANCELED)
121 goto out_advance;
122 goto out;
123 }
124
125 /* If this is a private inode, don't leak its details to userspace. */
126 if (IS_PRIVATE(inode) || xfs_is_sb_inum(mp, ino)) {
127 xfs_iunlock(ip, XFS_ILOCK_SHARED);
128 xfs_irele(ip);
129 error = -EINVAL;
130 goto out_advance;
131 }
132
133 /* xfs_iget returns the following without needing
134 * further change.
135 */
136 buf->bs_projectid = ip->i_projid;
137 buf->bs_ino = ino;
138 buf->bs_uid = from_kuid(sb_userns, vfsuid_into_kuid(vfsuid));
139 buf->bs_gid = from_kgid(sb_userns, vfsgid_into_kgid(vfsgid));
140 buf->bs_size = ip->i_disk_size;
141
142 buf->bs_nlink = inode->i_nlink;
143 buf->bs_atime = inode_get_atime_sec(inode);
144 buf->bs_atime_nsec = inode_get_atime_nsec(inode);
145 buf->bs_mtime = inode_get_mtime_sec(inode);
146 buf->bs_mtime_nsec = inode_get_mtime_nsec(inode);
147 buf->bs_ctime = inode_get_ctime_sec(inode);
148 buf->bs_ctime_nsec = inode_get_ctime_nsec(inode);
149 buf->bs_gen = inode->i_generation;
150 buf->bs_mode = inode->i_mode;
151
152 buf->bs_xflags = xfs_ip2xflags(ip);
153 buf->bs_extsize_blks = ip->i_extsize;
154
155 nextents = xfs_ifork_nextents(&ip->i_df);
156 if (!(bc->breq->flags & XFS_IBULK_NREXT64))
157 buf->bs_extents = min(nextents, XFS_MAX_EXTCNT_DATA_FORK_SMALL);
158 else
159 buf->bs_extents64 = nextents;
160
161 xfs_bulkstat_health(ip, buf);
162 buf->bs_aextents = xfs_ifork_nextents(&ip->i_af);
163 buf->bs_forkoff = xfs_inode_fork_boff(ip);
164 buf->bs_version = XFS_BULKSTAT_VERSION_V5;
165
166 if (xfs_has_v3inodes(mp)) {
167 buf->bs_btime = ip->i_crtime.tv_sec;
168 buf->bs_btime_nsec = ip->i_crtime.tv_nsec;
169 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
170 buf->bs_cowextsize_blks = ip->i_cowextsize;
171 }
172
173 switch (ip->i_df.if_format) {
174 case XFS_DINODE_FMT_DEV:
175 buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
176 buf->bs_blksize = BLKDEV_IOSIZE;
177 buf->bs_blocks = 0;
178 break;
179 case XFS_DINODE_FMT_LOCAL:
180 buf->bs_rdev = 0;
181 buf->bs_blksize = mp->m_sb.sb_blocksize;
182 buf->bs_blocks = 0;
183 break;
184 case XFS_DINODE_FMT_EXTENTS:
185 case XFS_DINODE_FMT_BTREE:
186 buf->bs_rdev = 0;
187 buf->bs_blksize = mp->m_sb.sb_blocksize;
188 buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks;
189 break;
190 }
191 xfs_iunlock(ip, XFS_ILOCK_SHARED);
192 xfs_irele(ip);
193
194 error = bc->formatter(bc->breq, buf);
195 if (error == -ECANCELED)
196 goto out_advance;
197 if (error)
198 goto out;
199
200out_advance:
201 /*
202 * Advance the cursor to the inode that comes after the one we just
203 * looked at. We want the caller to move along if the bulkstat
204 * information was copied successfully; if we tried to grab the inode
205 * but it's no longer allocated; or if it's internal metadata.
206 */
207 bc->breq->startino = ino + 1;
208out:
209 return error;
210}
211
212/* Bulkstat a single inode. */
213int
214xfs_bulkstat_one(
215 struct xfs_ibulk *breq,
216 bulkstat_one_fmt_pf formatter)
217{
218 struct xfs_bstat_chunk bc = {
219 .formatter = formatter,
220 .breq = breq,
221 };
222 struct xfs_trans *tp;
223 int error;
224
225 if (breq->idmap != &nop_mnt_idmap) {
226 xfs_warn_ratelimited(breq->mp,
227 "bulkstat not supported inside of idmapped mounts.");
228 return -EINVAL;
229 }
230
231 ASSERT(breq->icount == 1);
232
233 bc.buf = kzalloc(sizeof(struct xfs_bulkstat),
234 GFP_KERNEL | __GFP_RETRY_MAYFAIL);
235 if (!bc.buf)
236 return -ENOMEM;
237
238 /*
239 * Grab an empty transaction so that we can use its recursive buffer
240 * locking abilities to detect cycles in the inobt without deadlocking.
241 */
242 error = xfs_trans_alloc_empty(breq->mp, &tp);
243 if (error)
244 goto out;
245
246 error = xfs_bulkstat_one_int(breq->mp, breq->idmap, tp,
247 breq->startino, &bc);
248 xfs_trans_cancel(tp);
249out:
250 kfree(bc.buf);
251
252 /*
253 * If we reported one inode to userspace then we abort because we hit
254 * the end of the buffer. Don't leak that back to userspace.
255 */
256 if (error == -ECANCELED)
257 error = 0;
258
259 return error;
260}
261
262static int
263xfs_bulkstat_iwalk(
264 struct xfs_mount *mp,
265 struct xfs_trans *tp,
266 xfs_ino_t ino,
267 void *data)
268{
269 struct xfs_bstat_chunk *bc = data;
270 int error;
271
272 error = xfs_bulkstat_one_int(mp, bc->breq->idmap, tp, ino, data);
273 /* bulkstat just skips over missing inodes */
274 if (error == -ENOENT || error == -EINVAL)
275 return 0;
276 return error;
277}
278
279/*
280 * Check the incoming lastino parameter.
281 *
282 * We allow any inode value that could map to physical space inside the
283 * filesystem because if there are no inodes there, bulkstat moves on to the
284 * next chunk. In other words, the magic agino value of zero takes us to the
285 * first chunk in the AG, and an agino value past the end of the AG takes us to
286 * the first chunk in the next AG.
287 *
288 * Therefore we can end early if the requested inode is beyond the end of the
289 * filesystem or doesn't map properly.
290 */
291static inline bool
292xfs_bulkstat_already_done(
293 struct xfs_mount *mp,
294 xfs_ino_t startino)
295{
296 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino);
297 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino);
298
299 return agno >= mp->m_sb.sb_agcount ||
300 startino != XFS_AGINO_TO_INO(mp, agno, agino);
301}
302
303/* Return stat information in bulk (by-inode) for the filesystem. */
304int
305xfs_bulkstat(
306 struct xfs_ibulk *breq,
307 bulkstat_one_fmt_pf formatter)
308{
309 struct xfs_bstat_chunk bc = {
310 .formatter = formatter,
311 .breq = breq,
312 };
313 struct xfs_trans *tp;
314 unsigned int iwalk_flags = 0;
315 int error;
316
317 if (breq->idmap != &nop_mnt_idmap) {
318 xfs_warn_ratelimited(breq->mp,
319 "bulkstat not supported inside of idmapped mounts.");
320 return -EINVAL;
321 }
322 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
323 return 0;
324
325 bc.buf = kzalloc(sizeof(struct xfs_bulkstat),
326 GFP_KERNEL | __GFP_RETRY_MAYFAIL);
327 if (!bc.buf)
328 return -ENOMEM;
329
330 /*
331 * Grab an empty transaction so that we can use its recursive buffer
332 * locking abilities to detect cycles in the inobt without deadlocking.
333 */
334 error = xfs_trans_alloc_empty(breq->mp, &tp);
335 if (error)
336 goto out;
337
338 if (breq->flags & XFS_IBULK_SAME_AG)
339 iwalk_flags |= XFS_IWALK_SAME_AG;
340
341 error = xfs_iwalk(breq->mp, tp, breq->startino, iwalk_flags,
342 xfs_bulkstat_iwalk, breq->icount, &bc);
343 xfs_trans_cancel(tp);
344out:
345 kfree(bc.buf);
346
347 /*
348 * We found some inodes, so clear the error status and return them.
349 * The lastino pointer will point directly at the inode that triggered
350 * any error that occurred, so on the next call the error will be
351 * triggered again and propagated to userspace as there will be no
352 * formatted inodes in the buffer.
353 */
354 if (breq->ocount > 0)
355 error = 0;
356
357 return error;
358}
359
360/* Convert bulkstat (v5) to bstat (v1). */
361void
362xfs_bulkstat_to_bstat(
363 struct xfs_mount *mp,
364 struct xfs_bstat *bs1,
365 const struct xfs_bulkstat *bstat)
366{
367 /* memset is needed here because of padding holes in the structure. */
368 memset(bs1, 0, sizeof(struct xfs_bstat));
369 bs1->bs_ino = bstat->bs_ino;
370 bs1->bs_mode = bstat->bs_mode;
371 bs1->bs_nlink = bstat->bs_nlink;
372 bs1->bs_uid = bstat->bs_uid;
373 bs1->bs_gid = bstat->bs_gid;
374 bs1->bs_rdev = bstat->bs_rdev;
375 bs1->bs_blksize = bstat->bs_blksize;
376 bs1->bs_size = bstat->bs_size;
377 bs1->bs_atime.tv_sec = bstat->bs_atime;
378 bs1->bs_mtime.tv_sec = bstat->bs_mtime;
379 bs1->bs_ctime.tv_sec = bstat->bs_ctime;
380 bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
381 bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
382 bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
383 bs1->bs_blocks = bstat->bs_blocks;
384 bs1->bs_xflags = bstat->bs_xflags;
385 bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
386 bs1->bs_extents = bstat->bs_extents;
387 bs1->bs_gen = bstat->bs_gen;
388 bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
389 bs1->bs_forkoff = bstat->bs_forkoff;
390 bs1->bs_projid_hi = bstat->bs_projectid >> 16;
391 bs1->bs_sick = bstat->bs_sick;
392 bs1->bs_checked = bstat->bs_checked;
393 bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
394 bs1->bs_dmevmask = 0;
395 bs1->bs_dmstate = 0;
396 bs1->bs_aextents = bstat->bs_aextents;
397}
398
399struct xfs_inumbers_chunk {
400 inumbers_fmt_pf formatter;
401 struct xfs_ibulk *breq;
402};
403
404/*
405 * INUMBERS
406 * ========
407 * This is how we export inode btree records to userspace, so that XFS tools
408 * can figure out where inodes are allocated.
409 */
410
411/*
412 * Format the inode group structure and report it somewhere.
413 *
414 * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
415 * through the filesystem so we move it forward unless there was a runtime
416 * error. If the formatter tells us the buffer is now full we also move the
417 * cursor forward and abort the walk.
418 */
419STATIC int
420xfs_inumbers_walk(
421 struct xfs_mount *mp,
422 struct xfs_trans *tp,
423 xfs_agnumber_t agno,
424 const struct xfs_inobt_rec_incore *irec,
425 void *data)
426{
427 struct xfs_inumbers inogrp = {
428 .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
429 .xi_alloccount = irec->ir_count - irec->ir_freecount,
430 .xi_allocmask = ~irec->ir_free,
431 .xi_version = XFS_INUMBERS_VERSION_V5,
432 };
433 struct xfs_inumbers_chunk *ic = data;
434 int error;
435
436 error = ic->formatter(ic->breq, &inogrp);
437 if (error && error != -ECANCELED)
438 return error;
439
440 ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
441 XFS_INODES_PER_CHUNK;
442 return error;
443}
444
445/*
446 * Return inode number table for the filesystem.
447 */
448int
449xfs_inumbers(
450 struct xfs_ibulk *breq,
451 inumbers_fmt_pf formatter)
452{
453 struct xfs_inumbers_chunk ic = {
454 .formatter = formatter,
455 .breq = breq,
456 };
457 struct xfs_trans *tp;
458 int error = 0;
459
460 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
461 return 0;
462
463 /*
464 * Grab an empty transaction so that we can use its recursive buffer
465 * locking abilities to detect cycles in the inobt without deadlocking.
466 */
467 error = xfs_trans_alloc_empty(breq->mp, &tp);
468 if (error)
469 goto out;
470
471 error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags,
472 xfs_inumbers_walk, breq->icount, &ic);
473 xfs_trans_cancel(tp);
474out:
475
476 /*
477 * We found some inode groups, so clear the error status and return
478 * them. The lastino pointer will point directly at the inode that
479 * triggered any error that occurred, so on the next call the error
480 * will be triggered again and propagated to userspace as there will be
481 * no formatted inode groups in the buffer.
482 */
483 if (breq->ocount > 0)
484 error = 0;
485
486 return error;
487}
488
489/* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
490void
491xfs_inumbers_to_inogrp(
492 struct xfs_inogrp *ig1,
493 const struct xfs_inumbers *ig)
494{
495 /* memset is needed here because of padding holes in the structure. */
496 memset(ig1, 0, sizeof(struct xfs_inogrp));
497 ig1->xi_startino = ig->xi_startino;
498 ig1->xi_alloccount = ig->xi_alloccount;
499 ig1->xi_allocmask = ig->xi_allocmask;
500}