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_mount.h"
25#include "xfs_inode.h"
26#include "xfs_btree.h"
27#include "xfs_ialloc.h"
28#include "xfs_ialloc_btree.h"
29#include "xfs_itable.h"
30#include "xfs_error.h"
31#include "xfs_trace.h"
32#include "xfs_icache.h"
33
34/*
35 * Return stat information for one inode.
36 * Return 0 if ok, else errno.
37 */
38int
39xfs_bulkstat_one_int(
40 struct xfs_mount *mp, /* mount point for filesystem */
41 xfs_ino_t ino, /* inode to get data for */
42 void __user *buffer, /* buffer to place output in */
43 int ubsize, /* size of buffer */
44 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */
45 int *ubused, /* bytes used by me */
46 int *stat) /* BULKSTAT_RV_... */
47{
48 struct xfs_icdinode *dic; /* dinode core info pointer */
49 struct xfs_inode *ip; /* incore inode pointer */
50 struct inode *inode;
51 struct xfs_bstat *buf; /* return buffer */
52 int error = 0; /* error value */
53
54 *stat = BULKSTAT_RV_NOTHING;
55
56 if (!buffer || xfs_internal_inum(mp, ino))
57 return -EINVAL;
58
59 buf = kmem_zalloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL);
60 if (!buf)
61 return -ENOMEM;
62
63 error = xfs_iget(mp, NULL, ino,
64 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
65 XFS_ILOCK_SHARED, &ip);
66 if (error)
67 goto out_free;
68
69 ASSERT(ip != NULL);
70 ASSERT(ip->i_imap.im_blkno != 0);
71 inode = VFS_I(ip);
72
73 dic = &ip->i_d;
74
75 /* xfs_iget returns the following without needing
76 * further change.
77 */
78 buf->bs_projid_lo = dic->di_projid_lo;
79 buf->bs_projid_hi = dic->di_projid_hi;
80 buf->bs_ino = ino;
81 buf->bs_uid = dic->di_uid;
82 buf->bs_gid = dic->di_gid;
83 buf->bs_size = dic->di_size;
84
85 buf->bs_nlink = inode->i_nlink;
86 buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
87 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
88 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
89 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
90 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
91 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
92 buf->bs_gen = inode->i_generation;
93 buf->bs_mode = inode->i_mode;
94
95 buf->bs_xflags = xfs_ip2xflags(ip);
96 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
97 buf->bs_extents = dic->di_nextents;
98 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
99 buf->bs_dmevmask = dic->di_dmevmask;
100 buf->bs_dmstate = dic->di_dmstate;
101 buf->bs_aextents = dic->di_anextents;
102 buf->bs_forkoff = XFS_IFORK_BOFF(ip);
103
104 if (dic->di_version == 3) {
105 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
106 buf->bs_cowextsize = dic->di_cowextsize <<
107 mp->m_sb.sb_blocklog;
108 }
109
110 switch (dic->di_format) {
111 case XFS_DINODE_FMT_DEV:
112 buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
113 buf->bs_blksize = BLKDEV_IOSIZE;
114 buf->bs_blocks = 0;
115 break;
116 case XFS_DINODE_FMT_LOCAL:
117 buf->bs_rdev = 0;
118 buf->bs_blksize = mp->m_sb.sb_blocksize;
119 buf->bs_blocks = 0;
120 break;
121 case XFS_DINODE_FMT_EXTENTS:
122 case XFS_DINODE_FMT_BTREE:
123 buf->bs_rdev = 0;
124 buf->bs_blksize = mp->m_sb.sb_blocksize;
125 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
126 break;
127 }
128 xfs_iunlock(ip, XFS_ILOCK_SHARED);
129 IRELE(ip);
130
131 error = formatter(buffer, ubsize, ubused, buf);
132 if (!error)
133 *stat = BULKSTAT_RV_DIDONE;
134
135 out_free:
136 kmem_free(buf);
137 return error;
138}
139
140/* Return 0 on success or positive error */
141STATIC int
142xfs_bulkstat_one_fmt(
143 void __user *ubuffer,
144 int ubsize,
145 int *ubused,
146 const xfs_bstat_t *buffer)
147{
148 if (ubsize < sizeof(*buffer))
149 return -ENOMEM;
150 if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
151 return -EFAULT;
152 if (ubused)
153 *ubused = sizeof(*buffer);
154 return 0;
155}
156
157int
158xfs_bulkstat_one(
159 xfs_mount_t *mp, /* mount point for filesystem */
160 xfs_ino_t ino, /* inode number to get data for */
161 void __user *buffer, /* buffer to place output in */
162 int ubsize, /* size of buffer */
163 int *ubused, /* bytes used by me */
164 int *stat) /* BULKSTAT_RV_... */
165{
166 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
167 xfs_bulkstat_one_fmt, ubused, stat);
168}
169
170/*
171 * Loop over all clusters in a chunk for a given incore inode allocation btree
172 * record. Do a readahead if there are any allocated inodes in that cluster.
173 */
174STATIC void
175xfs_bulkstat_ichunk_ra(
176 struct xfs_mount *mp,
177 xfs_agnumber_t agno,
178 struct xfs_inobt_rec_incore *irec)
179{
180 xfs_agblock_t agbno;
181 struct blk_plug plug;
182 int blks_per_cluster;
183 int inodes_per_cluster;
184 int i; /* inode chunk index */
185
186 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino);
187 blks_per_cluster = xfs_icluster_size_fsb(mp);
188 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
189
190 blk_start_plug(&plug);
191 for (i = 0; i < XFS_INODES_PER_CHUNK;
192 i += inodes_per_cluster, agbno += blks_per_cluster) {
193 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) {
194 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster,
195 &xfs_inode_buf_ops);
196 }
197 }
198 blk_finish_plug(&plug);
199}
200
201/*
202 * Lookup the inode chunk that the given inode lives in and then get the record
203 * if we found the chunk. If the inode was not the last in the chunk and there
204 * are some left allocated, update the data for the pointed-to record as well as
205 * return the count of grabbed inodes.
206 */
207STATIC int
208xfs_bulkstat_grab_ichunk(
209 struct xfs_btree_cur *cur, /* btree cursor */
210 xfs_agino_t agino, /* starting inode of chunk */
211 int *icount,/* return # of inodes grabbed */
212 struct xfs_inobt_rec_incore *irec) /* btree record */
213{
214 int idx; /* index into inode chunk */
215 int stat;
216 int error = 0;
217
218 /* Lookup the inode chunk that this inode lives in */
219 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat);
220 if (error)
221 return error;
222 if (!stat) {
223 *icount = 0;
224 return error;
225 }
226
227 /* Get the record, should always work */
228 error = xfs_inobt_get_rec(cur, irec, &stat);
229 if (error)
230 return error;
231 XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1);
232
233 /* Check if the record contains the inode in request */
234 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) {
235 *icount = 0;
236 return 0;
237 }
238
239 idx = agino - irec->ir_startino + 1;
240 if (idx < XFS_INODES_PER_CHUNK &&
241 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) {
242 int i;
243
244 /* We got a right chunk with some left inodes allocated at it.
245 * Grab the chunk record. Mark all the uninteresting inodes
246 * free -- because they're before our start point.
247 */
248 for (i = 0; i < idx; i++) {
249 if (XFS_INOBT_MASK(i) & ~irec->ir_free)
250 irec->ir_freecount++;
251 }
252
253 irec->ir_free |= xfs_inobt_maskn(0, idx);
254 *icount = irec->ir_count - irec->ir_freecount;
255 }
256
257 return 0;
258}
259
260#define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size)
261
262struct xfs_bulkstat_agichunk {
263 char __user **ac_ubuffer;/* pointer into user's buffer */
264 int ac_ubleft; /* bytes left in user's buffer */
265 int ac_ubelem; /* spaces used in user's buffer */
266};
267
268/*
269 * Process inodes in chunk with a pointer to a formatter function
270 * that will iget the inode and fill in the appropriate structure.
271 */
272static int
273xfs_bulkstat_ag_ichunk(
274 struct xfs_mount *mp,
275 xfs_agnumber_t agno,
276 struct xfs_inobt_rec_incore *irbp,
277 bulkstat_one_pf formatter,
278 size_t statstruct_size,
279 struct xfs_bulkstat_agichunk *acp,
280 xfs_agino_t *last_agino)
281{
282 char __user **ubufp = acp->ac_ubuffer;
283 int chunkidx;
284 int error = 0;
285 xfs_agino_t agino = irbp->ir_startino;
286
287 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK;
288 chunkidx++, agino++) {
289 int fmterror;
290 int ubused;
291
292 /* inode won't fit in buffer, we are done */
293 if (acp->ac_ubleft < statstruct_size)
294 break;
295
296 /* Skip if this inode is free */
297 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free)
298 continue;
299
300 /* Get the inode and fill in a single buffer */
301 ubused = statstruct_size;
302 error = formatter(mp, XFS_AGINO_TO_INO(mp, agno, agino),
303 *ubufp, acp->ac_ubleft, &ubused, &fmterror);
304
305 if (fmterror == BULKSTAT_RV_GIVEUP ||
306 (error && error != -ENOENT && error != -EINVAL)) {
307 acp->ac_ubleft = 0;
308 ASSERT(error);
309 break;
310 }
311
312 /* be careful not to leak error if at end of chunk */
313 if (fmterror == BULKSTAT_RV_NOTHING || error) {
314 error = 0;
315 continue;
316 }
317
318 *ubufp += ubused;
319 acp->ac_ubleft -= ubused;
320 acp->ac_ubelem++;
321 }
322
323 /*
324 * Post-update *last_agino. At this point, agino will always point one
325 * inode past the last inode we processed successfully. Hence we
326 * substract that inode when setting the *last_agino cursor so that we
327 * return the correct cookie to userspace. On the next bulkstat call,
328 * the inode under the lastino cookie will be skipped as we have already
329 * processed it here.
330 */
331 *last_agino = agino - 1;
332
333 return error;
334}
335
336/*
337 * Return stat information in bulk (by-inode) for the filesystem.
338 */
339int /* error status */
340xfs_bulkstat(
341 xfs_mount_t *mp, /* mount point for filesystem */
342 xfs_ino_t *lastinop, /* last inode returned */
343 int *ubcountp, /* size of buffer/count returned */
344 bulkstat_one_pf formatter, /* func that'd fill a single buf */
345 size_t statstruct_size, /* sizeof struct filling */
346 char __user *ubuffer, /* buffer with inode stats */
347 int *done) /* 1 if there are more stats to get */
348{
349 xfs_buf_t *agbp; /* agi header buffer */
350 xfs_agino_t agino; /* inode # in allocation group */
351 xfs_agnumber_t agno; /* allocation group number */
352 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */
353 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */
354 int nirbuf; /* size of irbuf */
355 int ubcount; /* size of user's buffer */
356 struct xfs_bulkstat_agichunk ac;
357 int error = 0;
358
359 /*
360 * Get the last inode value, see if there's nothing to do.
361 */
362 agno = XFS_INO_TO_AGNO(mp, *lastinop);
363 agino = XFS_INO_TO_AGINO(mp, *lastinop);
364 if (agno >= mp->m_sb.sb_agcount ||
365 *lastinop != XFS_AGINO_TO_INO(mp, agno, agino)) {
366 *done = 1;
367 *ubcountp = 0;
368 return 0;
369 }
370
371 ubcount = *ubcountp; /* statstruct's */
372 ac.ac_ubuffer = &ubuffer;
373 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */;
374 ac.ac_ubelem = 0;
375
376 *ubcountp = 0;
377 *done = 0;
378
379 irbuf = kmem_zalloc_large(PAGE_SIZE * 4, KM_SLEEP);
380 if (!irbuf)
381 return -ENOMEM;
382 nirbuf = (PAGE_SIZE * 4) / sizeof(*irbuf);
383
384 /*
385 * Loop over the allocation groups, starting from the last
386 * inode returned; 0 means start of the allocation group.
387 */
388 while (agno < mp->m_sb.sb_agcount) {
389 struct xfs_inobt_rec_incore *irbp = irbuf;
390 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf;
391 bool end_of_ag = false;
392 int icount = 0;
393 int stat;
394
395 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
396 if (error)
397 break;
398 /*
399 * Allocate and initialize a btree cursor for ialloc btree.
400 */
401 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
402 XFS_BTNUM_INO);
403 if (agino > 0) {
404 /*
405 * In the middle of an allocation group, we need to get
406 * the remainder of the chunk we're in.
407 */
408 struct xfs_inobt_rec_incore r;
409
410 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r);
411 if (error)
412 goto del_cursor;
413 if (icount) {
414 irbp->ir_startino = r.ir_startino;
415 irbp->ir_holemask = r.ir_holemask;
416 irbp->ir_count = r.ir_count;
417 irbp->ir_freecount = r.ir_freecount;
418 irbp->ir_free = r.ir_free;
419 irbp++;
420 }
421 /* Increment to the next record */
422 error = xfs_btree_increment(cur, 0, &stat);
423 } else {
424 /* Start of ag. Lookup the first inode chunk */
425 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat);
426 }
427 if (error || stat == 0) {
428 end_of_ag = true;
429 goto del_cursor;
430 }
431
432 /*
433 * Loop through inode btree records in this ag,
434 * until we run out of inodes or space in the buffer.
435 */
436 while (irbp < irbufend && icount < ubcount) {
437 struct xfs_inobt_rec_incore r;
438
439 error = xfs_inobt_get_rec(cur, &r, &stat);
440 if (error || stat == 0) {
441 end_of_ag = true;
442 goto del_cursor;
443 }
444
445 /*
446 * If this chunk has any allocated inodes, save it.
447 * Also start read-ahead now for this chunk.
448 */
449 if (r.ir_freecount < r.ir_count) {
450 xfs_bulkstat_ichunk_ra(mp, agno, &r);
451 irbp->ir_startino = r.ir_startino;
452 irbp->ir_holemask = r.ir_holemask;
453 irbp->ir_count = r.ir_count;
454 irbp->ir_freecount = r.ir_freecount;
455 irbp->ir_free = r.ir_free;
456 irbp++;
457 icount += r.ir_count - r.ir_freecount;
458 }
459 error = xfs_btree_increment(cur, 0, &stat);
460 if (error || stat == 0) {
461 end_of_ag = true;
462 goto del_cursor;
463 }
464 cond_resched();
465 }
466
467 /*
468 * Drop the btree buffers and the agi buffer as we can't hold any
469 * of the locks these represent when calling iget. If there is a
470 * pending error, then we are done.
471 */
472del_cursor:
473 xfs_btree_del_cursor(cur, error ?
474 XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
475 xfs_buf_relse(agbp);
476 if (error)
477 break;
478 /*
479 * Now format all the good inodes into the user's buffer. The
480 * call to xfs_bulkstat_ag_ichunk() sets up the agino pointer
481 * for the next loop iteration.
482 */
483 irbufend = irbp;
484 for (irbp = irbuf;
485 irbp < irbufend && ac.ac_ubleft >= statstruct_size;
486 irbp++) {
487 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp,
488 formatter, statstruct_size, &ac,
489 &agino);
490 if (error)
491 break;
492
493 cond_resched();
494 }
495
496 /*
497 * If we've run out of space or had a formatting error, we
498 * are now done
499 */
500 if (ac.ac_ubleft < statstruct_size || error)
501 break;
502
503 if (end_of_ag) {
504 agno++;
505 agino = 0;
506 }
507 }
508 /*
509 * Done, we're either out of filesystem or space to put the data.
510 */
511 kmem_free(irbuf);
512 *ubcountp = ac.ac_ubelem;
513
514 /*
515 * We found some inodes, so clear the error status and return them.
516 * The lastino pointer will point directly at the inode that triggered
517 * any error that occurred, so on the next call the error will be
518 * triggered again and propagated to userspace as there will be no
519 * formatted inodes in the buffer.
520 */
521 if (ac.ac_ubelem)
522 error = 0;
523
524 /*
525 * If we ran out of filesystem, lastino will point off the end of
526 * the filesystem so the next call will return immediately.
527 */
528 *lastinop = XFS_AGINO_TO_INO(mp, agno, agino);
529 if (agno >= mp->m_sb.sb_agcount)
530 *done = 1;
531
532 return error;
533}
534
535int
536xfs_inumbers_fmt(
537 void __user *ubuffer, /* buffer to write to */
538 const struct xfs_inogrp *buffer, /* buffer to read from */
539 long count, /* # of elements to read */
540 long *written) /* # of bytes written */
541{
542 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
543 return -EFAULT;
544 *written = count * sizeof(*buffer);
545 return 0;
546}
547
548/*
549 * Return inode number table for the filesystem.
550 */
551int /* error status */
552xfs_inumbers(
553 struct xfs_mount *mp,/* mount point for filesystem */
554 xfs_ino_t *lastino,/* last inode returned */
555 int *count,/* size of buffer/count returned */
556 void __user *ubuffer,/* buffer with inode descriptions */
557 inumbers_fmt_pf formatter)
558{
559 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino);
560 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino);
561 struct xfs_btree_cur *cur = NULL;
562 struct xfs_buf *agbp = NULL;
563 struct xfs_inogrp *buffer;
564 int bcount;
565 int left = *count;
566 int bufidx = 0;
567 int error = 0;
568
569 *count = 0;
570 if (agno >= mp->m_sb.sb_agcount ||
571 *lastino != XFS_AGINO_TO_INO(mp, agno, agino))
572 return error;
573
574 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
575 buffer = kmem_zalloc(bcount * sizeof(*buffer), KM_SLEEP);
576 do {
577 struct xfs_inobt_rec_incore r;
578 int stat;
579
580 if (!agbp) {
581 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
582 if (error)
583 break;
584
585 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
586 XFS_BTNUM_INO);
587 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
588 &stat);
589 if (error)
590 break;
591 if (!stat)
592 goto next_ag;
593 }
594
595 error = xfs_inobt_get_rec(cur, &r, &stat);
596 if (error)
597 break;
598 if (!stat)
599 goto next_ag;
600
601 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
602 buffer[bufidx].xi_startino =
603 XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
604 buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount;
605 buffer[bufidx].xi_allocmask = ~r.ir_free;
606 if (++bufidx == bcount) {
607 long written;
608
609 error = formatter(ubuffer, buffer, bufidx, &written);
610 if (error)
611 break;
612 ubuffer += written;
613 *count += bufidx;
614 bufidx = 0;
615 }
616 if (!--left)
617 break;
618
619 error = xfs_btree_increment(cur, 0, &stat);
620 if (error)
621 break;
622 if (stat)
623 continue;
624
625next_ag:
626 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
627 cur = NULL;
628 xfs_buf_relse(agbp);
629 agbp = NULL;
630 agino = 0;
631 agno++;
632 } while (agno < mp->m_sb.sb_agcount);
633
634 if (!error) {
635 if (bufidx) {
636 long written;
637
638 error = formatter(ubuffer, buffer, bufidx, &written);
639 if (!error)
640 *count += bufidx;
641 }
642 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
643 }
644
645 kmem_free(buffer);
646 if (cur)
647 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
648 XFS_BTREE_NOERROR));
649 if (agbp)
650 xfs_buf_relse(agbp);
651
652 return error;
653}
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
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 */
55STATIC int
56xfs_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 struct user_namespace *sb_userns = mp->m_super->s_user_ns;
64 struct xfs_inode *ip; /* incore inode pointer */
65 struct inode *inode;
66 struct xfs_bulkstat *buf = bc->buf;
67 xfs_extnum_t nextents;
68 int error = -EINVAL;
69 vfsuid_t vfsuid;
70 vfsgid_t vfsgid;
71
72 if (xfs_internal_inum(mp, ino))
73 goto out_advance;
74
75 error = xfs_iget(mp, tp, ino,
76 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
77 XFS_ILOCK_SHARED, &ip);
78 if (error == -ENOENT || error == -EINVAL)
79 goto out_advance;
80 if (error)
81 goto out;
82
83 /* Reload the incore unlinked list to avoid failure in inodegc. */
84 if (xfs_inode_unlinked_incomplete(ip)) {
85 error = xfs_inode_reload_unlinked_bucket(tp, ip);
86 if (error) {
87 xfs_iunlock(ip, XFS_ILOCK_SHARED);
88 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
89 xfs_irele(ip);
90 return error;
91 }
92 }
93
94 ASSERT(ip != NULL);
95 ASSERT(ip->i_imap.im_blkno != 0);
96 inode = VFS_I(ip);
97 vfsuid = i_uid_into_vfsuid(idmap, inode);
98 vfsgid = i_gid_into_vfsgid(idmap, inode);
99
100 /* xfs_iget returns the following without needing
101 * further change.
102 */
103 buf->bs_projectid = ip->i_projid;
104 buf->bs_ino = ino;
105 buf->bs_uid = from_kuid(sb_userns, vfsuid_into_kuid(vfsuid));
106 buf->bs_gid = from_kgid(sb_userns, vfsgid_into_kgid(vfsgid));
107 buf->bs_size = ip->i_disk_size;
108
109 buf->bs_nlink = inode->i_nlink;
110 buf->bs_atime = inode_get_atime_sec(inode);
111 buf->bs_atime_nsec = inode_get_atime_nsec(inode);
112 buf->bs_mtime = inode_get_mtime_sec(inode);
113 buf->bs_mtime_nsec = inode_get_mtime_nsec(inode);
114 buf->bs_ctime = inode_get_ctime_sec(inode);
115 buf->bs_ctime_nsec = inode_get_ctime_nsec(inode);
116 buf->bs_gen = inode->i_generation;
117 buf->bs_mode = inode->i_mode;
118
119 buf->bs_xflags = xfs_ip2xflags(ip);
120 buf->bs_extsize_blks = ip->i_extsize;
121
122 nextents = xfs_ifork_nextents(&ip->i_df);
123 if (!(bc->breq->flags & XFS_IBULK_NREXT64))
124 buf->bs_extents = min(nextents, XFS_MAX_EXTCNT_DATA_FORK_SMALL);
125 else
126 buf->bs_extents64 = nextents;
127
128 xfs_bulkstat_health(ip, buf);
129 buf->bs_aextents = xfs_ifork_nextents(&ip->i_af);
130 buf->bs_forkoff = xfs_inode_fork_boff(ip);
131 buf->bs_version = XFS_BULKSTAT_VERSION_V5;
132
133 if (xfs_has_v3inodes(mp)) {
134 buf->bs_btime = ip->i_crtime.tv_sec;
135 buf->bs_btime_nsec = ip->i_crtime.tv_nsec;
136 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
137 buf->bs_cowextsize_blks = ip->i_cowextsize;
138 }
139
140 switch (ip->i_df.if_format) {
141 case XFS_DINODE_FMT_DEV:
142 buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
143 buf->bs_blksize = BLKDEV_IOSIZE;
144 buf->bs_blocks = 0;
145 break;
146 case XFS_DINODE_FMT_LOCAL:
147 buf->bs_rdev = 0;
148 buf->bs_blksize = mp->m_sb.sb_blocksize;
149 buf->bs_blocks = 0;
150 break;
151 case XFS_DINODE_FMT_EXTENTS:
152 case XFS_DINODE_FMT_BTREE:
153 buf->bs_rdev = 0;
154 buf->bs_blksize = mp->m_sb.sb_blocksize;
155 buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks;
156 break;
157 }
158 xfs_iunlock(ip, XFS_ILOCK_SHARED);
159 xfs_irele(ip);
160
161 error = bc->formatter(bc->breq, buf);
162 if (error == -ECANCELED)
163 goto out_advance;
164 if (error)
165 goto out;
166
167out_advance:
168 /*
169 * Advance the cursor to the inode that comes after the one we just
170 * looked at. We want the caller to move along if the bulkstat
171 * information was copied successfully; if we tried to grab the inode
172 * but it's no longer allocated; or if it's internal metadata.
173 */
174 bc->breq->startino = ino + 1;
175out:
176 return error;
177}
178
179/* Bulkstat a single inode. */
180int
181xfs_bulkstat_one(
182 struct xfs_ibulk *breq,
183 bulkstat_one_fmt_pf formatter)
184{
185 struct xfs_bstat_chunk bc = {
186 .formatter = formatter,
187 .breq = breq,
188 };
189 struct xfs_trans *tp;
190 int error;
191
192 if (breq->idmap != &nop_mnt_idmap) {
193 xfs_warn_ratelimited(breq->mp,
194 "bulkstat not supported inside of idmapped mounts.");
195 return -EINVAL;
196 }
197
198 ASSERT(breq->icount == 1);
199
200 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
201 KM_MAYFAIL);
202 if (!bc.buf)
203 return -ENOMEM;
204
205 /*
206 * Grab an empty transaction so that we can use its recursive buffer
207 * locking abilities to detect cycles in the inobt without deadlocking.
208 */
209 error = xfs_trans_alloc_empty(breq->mp, &tp);
210 if (error)
211 goto out;
212
213 error = xfs_bulkstat_one_int(breq->mp, breq->idmap, tp,
214 breq->startino, &bc);
215 xfs_trans_cancel(tp);
216out:
217 kmem_free(bc.buf);
218
219 /*
220 * If we reported one inode to userspace then we abort because we hit
221 * the end of the buffer. Don't leak that back to userspace.
222 */
223 if (error == -ECANCELED)
224 error = 0;
225
226 return error;
227}
228
229static int
230xfs_bulkstat_iwalk(
231 struct xfs_mount *mp,
232 struct xfs_trans *tp,
233 xfs_ino_t ino,
234 void *data)
235{
236 struct xfs_bstat_chunk *bc = data;
237 int error;
238
239 error = xfs_bulkstat_one_int(mp, bc->breq->idmap, tp, ino, data);
240 /* bulkstat just skips over missing inodes */
241 if (error == -ENOENT || error == -EINVAL)
242 return 0;
243 return error;
244}
245
246/*
247 * Check the incoming lastino parameter.
248 *
249 * We allow any inode value that could map to physical space inside the
250 * filesystem because if there are no inodes there, bulkstat moves on to the
251 * next chunk. In other words, the magic agino value of zero takes us to the
252 * first chunk in the AG, and an agino value past the end of the AG takes us to
253 * the first chunk in the next AG.
254 *
255 * Therefore we can end early if the requested inode is beyond the end of the
256 * filesystem or doesn't map properly.
257 */
258static inline bool
259xfs_bulkstat_already_done(
260 struct xfs_mount *mp,
261 xfs_ino_t startino)
262{
263 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino);
264 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino);
265
266 return agno >= mp->m_sb.sb_agcount ||
267 startino != XFS_AGINO_TO_INO(mp, agno, agino);
268}
269
270/* Return stat information in bulk (by-inode) for the filesystem. */
271int
272xfs_bulkstat(
273 struct xfs_ibulk *breq,
274 bulkstat_one_fmt_pf formatter)
275{
276 struct xfs_bstat_chunk bc = {
277 .formatter = formatter,
278 .breq = breq,
279 };
280 struct xfs_trans *tp;
281 unsigned int iwalk_flags = 0;
282 int error;
283
284 if (breq->idmap != &nop_mnt_idmap) {
285 xfs_warn_ratelimited(breq->mp,
286 "bulkstat not supported inside of idmapped mounts.");
287 return -EINVAL;
288 }
289 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
290 return 0;
291
292 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
293 KM_MAYFAIL);
294 if (!bc.buf)
295 return -ENOMEM;
296
297 /*
298 * Grab an empty transaction so that we can use its recursive buffer
299 * locking abilities to detect cycles in the inobt without deadlocking.
300 */
301 error = xfs_trans_alloc_empty(breq->mp, &tp);
302 if (error)
303 goto out;
304
305 if (breq->flags & XFS_IBULK_SAME_AG)
306 iwalk_flags |= XFS_IWALK_SAME_AG;
307
308 error = xfs_iwalk(breq->mp, tp, breq->startino, iwalk_flags,
309 xfs_bulkstat_iwalk, breq->icount, &bc);
310 xfs_trans_cancel(tp);
311out:
312 kmem_free(bc.buf);
313
314 /*
315 * We found some inodes, so clear the error status and return them.
316 * The lastino pointer will point directly at the inode that triggered
317 * any error that occurred, so on the next call the error will be
318 * triggered again and propagated to userspace as there will be no
319 * formatted inodes in the buffer.
320 */
321 if (breq->ocount > 0)
322 error = 0;
323
324 return error;
325}
326
327/* Convert bulkstat (v5) to bstat (v1). */
328void
329xfs_bulkstat_to_bstat(
330 struct xfs_mount *mp,
331 struct xfs_bstat *bs1,
332 const struct xfs_bulkstat *bstat)
333{
334 /* memset is needed here because of padding holes in the structure. */
335 memset(bs1, 0, sizeof(struct xfs_bstat));
336 bs1->bs_ino = bstat->bs_ino;
337 bs1->bs_mode = bstat->bs_mode;
338 bs1->bs_nlink = bstat->bs_nlink;
339 bs1->bs_uid = bstat->bs_uid;
340 bs1->bs_gid = bstat->bs_gid;
341 bs1->bs_rdev = bstat->bs_rdev;
342 bs1->bs_blksize = bstat->bs_blksize;
343 bs1->bs_size = bstat->bs_size;
344 bs1->bs_atime.tv_sec = bstat->bs_atime;
345 bs1->bs_mtime.tv_sec = bstat->bs_mtime;
346 bs1->bs_ctime.tv_sec = bstat->bs_ctime;
347 bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
348 bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
349 bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
350 bs1->bs_blocks = bstat->bs_blocks;
351 bs1->bs_xflags = bstat->bs_xflags;
352 bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
353 bs1->bs_extents = bstat->bs_extents;
354 bs1->bs_gen = bstat->bs_gen;
355 bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
356 bs1->bs_forkoff = bstat->bs_forkoff;
357 bs1->bs_projid_hi = bstat->bs_projectid >> 16;
358 bs1->bs_sick = bstat->bs_sick;
359 bs1->bs_checked = bstat->bs_checked;
360 bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
361 bs1->bs_dmevmask = 0;
362 bs1->bs_dmstate = 0;
363 bs1->bs_aextents = bstat->bs_aextents;
364}
365
366struct xfs_inumbers_chunk {
367 inumbers_fmt_pf formatter;
368 struct xfs_ibulk *breq;
369};
370
371/*
372 * INUMBERS
373 * ========
374 * This is how we export inode btree records to userspace, so that XFS tools
375 * can figure out where inodes are allocated.
376 */
377
378/*
379 * Format the inode group structure and report it somewhere.
380 *
381 * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
382 * through the filesystem so we move it forward unless there was a runtime
383 * error. If the formatter tells us the buffer is now full we also move the
384 * cursor forward and abort the walk.
385 */
386STATIC int
387xfs_inumbers_walk(
388 struct xfs_mount *mp,
389 struct xfs_trans *tp,
390 xfs_agnumber_t agno,
391 const struct xfs_inobt_rec_incore *irec,
392 void *data)
393{
394 struct xfs_inumbers inogrp = {
395 .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
396 .xi_alloccount = irec->ir_count - irec->ir_freecount,
397 .xi_allocmask = ~irec->ir_free,
398 .xi_version = XFS_INUMBERS_VERSION_V5,
399 };
400 struct xfs_inumbers_chunk *ic = data;
401 int error;
402
403 error = ic->formatter(ic->breq, &inogrp);
404 if (error && error != -ECANCELED)
405 return error;
406
407 ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
408 XFS_INODES_PER_CHUNK;
409 return error;
410}
411
412/*
413 * Return inode number table for the filesystem.
414 */
415int
416xfs_inumbers(
417 struct xfs_ibulk *breq,
418 inumbers_fmt_pf formatter)
419{
420 struct xfs_inumbers_chunk ic = {
421 .formatter = formatter,
422 .breq = breq,
423 };
424 struct xfs_trans *tp;
425 int error = 0;
426
427 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
428 return 0;
429
430 /*
431 * Grab an empty transaction so that we can use its recursive buffer
432 * locking abilities to detect cycles in the inobt without deadlocking.
433 */
434 error = xfs_trans_alloc_empty(breq->mp, &tp);
435 if (error)
436 goto out;
437
438 error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags,
439 xfs_inumbers_walk, breq->icount, &ic);
440 xfs_trans_cancel(tp);
441out:
442
443 /*
444 * We found some inode groups, so clear the error status and return
445 * them. The lastino pointer will point directly at the inode that
446 * triggered any error that occurred, so on the next call the error
447 * will be triggered again and propagated to userspace as there will be
448 * no formatted inode groups in the buffer.
449 */
450 if (breq->ocount > 0)
451 error = 0;
452
453 return error;
454}
455
456/* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
457void
458xfs_inumbers_to_inogrp(
459 struct xfs_inogrp *ig1,
460 const struct xfs_inumbers *ig)
461{
462 /* memset is needed here because of padding holes in the structure. */
463 memset(ig1, 0, sizeof(struct xfs_inogrp));
464 ig1->xi_startino = ig->xi_startino;
465 ig1->xi_alloccount = ig->xi_alloccount;
466 ig1->xi_allocmask = ig->xi_allocmask;
467}