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
34STATIC int
35xfs_internal_inum(
36 xfs_mount_t *mp,
37 xfs_ino_t ino)
38{
39 return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino ||
40 (xfs_sb_version_hasquota(&mp->m_sb) &&
41 xfs_is_quota_inode(&mp->m_sb, ino)));
42}
43
44/*
45 * Return stat information for one inode.
46 * Return 0 if ok, else errno.
47 */
48int
49xfs_bulkstat_one_int(
50 struct xfs_mount *mp, /* mount point for filesystem */
51 xfs_ino_t ino, /* inode to get data for */
52 void __user *buffer, /* buffer to place output in */
53 int ubsize, /* size of buffer */
54 bulkstat_one_fmt_pf formatter, /* formatter, copy to user */
55 int *ubused, /* bytes used by me */
56 int *stat) /* BULKSTAT_RV_... */
57{
58 struct xfs_icdinode *dic; /* dinode core info pointer */
59 struct xfs_inode *ip; /* incore inode pointer */
60 struct inode *inode;
61 struct xfs_bstat *buf; /* return buffer */
62 int error = 0; /* error value */
63
64 *stat = BULKSTAT_RV_NOTHING;
65
66 if (!buffer || xfs_internal_inum(mp, ino))
67 return -EINVAL;
68
69 buf = kmem_zalloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL);
70 if (!buf)
71 return -ENOMEM;
72
73 error = xfs_iget(mp, NULL, ino,
74 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
75 XFS_ILOCK_SHARED, &ip);
76 if (error)
77 goto out_free;
78
79 ASSERT(ip != NULL);
80 ASSERT(ip->i_imap.im_blkno != 0);
81 inode = VFS_I(ip);
82
83 dic = &ip->i_d;
84
85 /* xfs_iget returns the following without needing
86 * further change.
87 */
88 buf->bs_projid_lo = dic->di_projid_lo;
89 buf->bs_projid_hi = dic->di_projid_hi;
90 buf->bs_ino = ino;
91 buf->bs_uid = dic->di_uid;
92 buf->bs_gid = dic->di_gid;
93 buf->bs_size = dic->di_size;
94
95 buf->bs_nlink = inode->i_nlink;
96 buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
97 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
98 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
99 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
100 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
101 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
102 buf->bs_gen = inode->i_generation;
103 buf->bs_mode = inode->i_mode;
104
105 buf->bs_xflags = xfs_ip2xflags(ip);
106 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
107 buf->bs_extents = dic->di_nextents;
108 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
109 buf->bs_dmevmask = dic->di_dmevmask;
110 buf->bs_dmstate = dic->di_dmstate;
111 buf->bs_aextents = dic->di_anextents;
112 buf->bs_forkoff = XFS_IFORK_BOFF(ip);
113
114 if (dic->di_version == 3) {
115 if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
116 buf->bs_cowextsize = dic->di_cowextsize <<
117 mp->m_sb.sb_blocklog;
118 }
119
120 switch (dic->di_format) {
121 case XFS_DINODE_FMT_DEV:
122 buf->bs_rdev = ip->i_df.if_u2.if_rdev;
123 buf->bs_blksize = BLKDEV_IOSIZE;
124 buf->bs_blocks = 0;
125 break;
126 case XFS_DINODE_FMT_LOCAL:
127 case XFS_DINODE_FMT_UUID:
128 buf->bs_rdev = 0;
129 buf->bs_blksize = mp->m_sb.sb_blocksize;
130 buf->bs_blocks = 0;
131 break;
132 case XFS_DINODE_FMT_EXTENTS:
133 case XFS_DINODE_FMT_BTREE:
134 buf->bs_rdev = 0;
135 buf->bs_blksize = mp->m_sb.sb_blocksize;
136 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
137 break;
138 }
139 xfs_iunlock(ip, XFS_ILOCK_SHARED);
140 IRELE(ip);
141
142 error = formatter(buffer, ubsize, ubused, buf);
143 if (!error)
144 *stat = BULKSTAT_RV_DIDONE;
145
146 out_free:
147 kmem_free(buf);
148 return error;
149}
150
151/* Return 0 on success or positive error */
152STATIC int
153xfs_bulkstat_one_fmt(
154 void __user *ubuffer,
155 int ubsize,
156 int *ubused,
157 const xfs_bstat_t *buffer)
158{
159 if (ubsize < sizeof(*buffer))
160 return -ENOMEM;
161 if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
162 return -EFAULT;
163 if (ubused)
164 *ubused = sizeof(*buffer);
165 return 0;
166}
167
168int
169xfs_bulkstat_one(
170 xfs_mount_t *mp, /* mount point for filesystem */
171 xfs_ino_t ino, /* inode number to get data for */
172 void __user *buffer, /* buffer to place output in */
173 int ubsize, /* size of buffer */
174 int *ubused, /* bytes used by me */
175 int *stat) /* BULKSTAT_RV_... */
176{
177 return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
178 xfs_bulkstat_one_fmt, ubused, stat);
179}
180
181/*
182 * Loop over all clusters in a chunk for a given incore inode allocation btree
183 * record. Do a readahead if there are any allocated inodes in that cluster.
184 */
185STATIC void
186xfs_bulkstat_ichunk_ra(
187 struct xfs_mount *mp,
188 xfs_agnumber_t agno,
189 struct xfs_inobt_rec_incore *irec)
190{
191 xfs_agblock_t agbno;
192 struct blk_plug plug;
193 int blks_per_cluster;
194 int inodes_per_cluster;
195 int i; /* inode chunk index */
196
197 agbno = XFS_AGINO_TO_AGBNO(mp, irec->ir_startino);
198 blks_per_cluster = xfs_icluster_size_fsb(mp);
199 inodes_per_cluster = blks_per_cluster << mp->m_sb.sb_inopblog;
200
201 blk_start_plug(&plug);
202 for (i = 0; i < XFS_INODES_PER_CHUNK;
203 i += inodes_per_cluster, agbno += blks_per_cluster) {
204 if (xfs_inobt_maskn(i, inodes_per_cluster) & ~irec->ir_free) {
205 xfs_btree_reada_bufs(mp, agno, agbno, blks_per_cluster,
206 &xfs_inode_buf_ops);
207 }
208 }
209 blk_finish_plug(&plug);
210}
211
212/*
213 * Lookup the inode chunk that the given inode lives in and then get the record
214 * if we found the chunk. If the inode was not the last in the chunk and there
215 * are some left allocated, update the data for the pointed-to record as well as
216 * return the count of grabbed inodes.
217 */
218STATIC int
219xfs_bulkstat_grab_ichunk(
220 struct xfs_btree_cur *cur, /* btree cursor */
221 xfs_agino_t agino, /* starting inode of chunk */
222 int *icount,/* return # of inodes grabbed */
223 struct xfs_inobt_rec_incore *irec) /* btree record */
224{
225 int idx; /* index into inode chunk */
226 int stat;
227 int error = 0;
228
229 /* Lookup the inode chunk that this inode lives in */
230 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &stat);
231 if (error)
232 return error;
233 if (!stat) {
234 *icount = 0;
235 return error;
236 }
237
238 /* Get the record, should always work */
239 error = xfs_inobt_get_rec(cur, irec, &stat);
240 if (error)
241 return error;
242 XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1);
243
244 /* Check if the record contains the inode in request */
245 if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) {
246 *icount = 0;
247 return 0;
248 }
249
250 idx = agino - irec->ir_startino + 1;
251 if (idx < XFS_INODES_PER_CHUNK &&
252 (xfs_inobt_maskn(idx, XFS_INODES_PER_CHUNK - idx) & ~irec->ir_free)) {
253 int i;
254
255 /* We got a right chunk with some left inodes allocated at it.
256 * Grab the chunk record. Mark all the uninteresting inodes
257 * free -- because they're before our start point.
258 */
259 for (i = 0; i < idx; i++) {
260 if (XFS_INOBT_MASK(i) & ~irec->ir_free)
261 irec->ir_freecount++;
262 }
263
264 irec->ir_free |= xfs_inobt_maskn(0, idx);
265 *icount = irec->ir_count - irec->ir_freecount;
266 }
267
268 return 0;
269}
270
271#define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size)
272
273struct xfs_bulkstat_agichunk {
274 char __user **ac_ubuffer;/* pointer into user's buffer */
275 int ac_ubleft; /* bytes left in user's buffer */
276 int ac_ubelem; /* spaces used in user's buffer */
277};
278
279/*
280 * Process inodes in chunk with a pointer to a formatter function
281 * that will iget the inode and fill in the appropriate structure.
282 */
283static int
284xfs_bulkstat_ag_ichunk(
285 struct xfs_mount *mp,
286 xfs_agnumber_t agno,
287 struct xfs_inobt_rec_incore *irbp,
288 bulkstat_one_pf formatter,
289 size_t statstruct_size,
290 struct xfs_bulkstat_agichunk *acp,
291 xfs_agino_t *last_agino)
292{
293 char __user **ubufp = acp->ac_ubuffer;
294 int chunkidx;
295 int error = 0;
296 xfs_agino_t agino = irbp->ir_startino;
297
298 for (chunkidx = 0; chunkidx < XFS_INODES_PER_CHUNK;
299 chunkidx++, agino++) {
300 int fmterror;
301 int ubused;
302
303 /* inode won't fit in buffer, we are done */
304 if (acp->ac_ubleft < statstruct_size)
305 break;
306
307 /* Skip if this inode is free */
308 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free)
309 continue;
310
311 /* Get the inode and fill in a single buffer */
312 ubused = statstruct_size;
313 error = formatter(mp, XFS_AGINO_TO_INO(mp, agno, agino),
314 *ubufp, acp->ac_ubleft, &ubused, &fmterror);
315
316 if (fmterror == BULKSTAT_RV_GIVEUP ||
317 (error && error != -ENOENT && error != -EINVAL)) {
318 acp->ac_ubleft = 0;
319 ASSERT(error);
320 break;
321 }
322
323 /* be careful not to leak error if at end of chunk */
324 if (fmterror == BULKSTAT_RV_NOTHING || error) {
325 error = 0;
326 continue;
327 }
328
329 *ubufp += ubused;
330 acp->ac_ubleft -= ubused;
331 acp->ac_ubelem++;
332 }
333
334 /*
335 * Post-update *last_agino. At this point, agino will always point one
336 * inode past the last inode we processed successfully. Hence we
337 * substract that inode when setting the *last_agino cursor so that we
338 * return the correct cookie to userspace. On the next bulkstat call,
339 * the inode under the lastino cookie will be skipped as we have already
340 * processed it here.
341 */
342 *last_agino = agino - 1;
343
344 return error;
345}
346
347/*
348 * Return stat information in bulk (by-inode) for the filesystem.
349 */
350int /* error status */
351xfs_bulkstat(
352 xfs_mount_t *mp, /* mount point for filesystem */
353 xfs_ino_t *lastinop, /* last inode returned */
354 int *ubcountp, /* size of buffer/count returned */
355 bulkstat_one_pf formatter, /* func that'd fill a single buf */
356 size_t statstruct_size, /* sizeof struct filling */
357 char __user *ubuffer, /* buffer with inode stats */
358 int *done) /* 1 if there are more stats to get */
359{
360 xfs_buf_t *agbp; /* agi header buffer */
361 xfs_agino_t agino; /* inode # in allocation group */
362 xfs_agnumber_t agno; /* allocation group number */
363 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */
364 size_t irbsize; /* size of irec buffer in bytes */
365 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */
366 int nirbuf; /* size of irbuf */
367 int ubcount; /* size of user's buffer */
368 struct xfs_bulkstat_agichunk ac;
369 int error = 0;
370
371 /*
372 * Get the last inode value, see if there's nothing to do.
373 */
374 agno = XFS_INO_TO_AGNO(mp, *lastinop);
375 agino = XFS_INO_TO_AGINO(mp, *lastinop);
376 if (agno >= mp->m_sb.sb_agcount ||
377 *lastinop != XFS_AGINO_TO_INO(mp, agno, agino)) {
378 *done = 1;
379 *ubcountp = 0;
380 return 0;
381 }
382
383 ubcount = *ubcountp; /* statstruct's */
384 ac.ac_ubuffer = &ubuffer;
385 ac.ac_ubleft = ubcount * statstruct_size; /* bytes */;
386 ac.ac_ubelem = 0;
387
388 *ubcountp = 0;
389 *done = 0;
390
391 irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4);
392 if (!irbuf)
393 return -ENOMEM;
394
395 nirbuf = irbsize / sizeof(*irbuf);
396
397 /*
398 * Loop over the allocation groups, starting from the last
399 * inode returned; 0 means start of the allocation group.
400 */
401 while (agno < mp->m_sb.sb_agcount) {
402 struct xfs_inobt_rec_incore *irbp = irbuf;
403 struct xfs_inobt_rec_incore *irbufend = irbuf + nirbuf;
404 bool end_of_ag = false;
405 int icount = 0;
406 int stat;
407
408 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
409 if (error)
410 break;
411 /*
412 * Allocate and initialize a btree cursor for ialloc btree.
413 */
414 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
415 XFS_BTNUM_INO);
416 if (agino > 0) {
417 /*
418 * In the middle of an allocation group, we need to get
419 * the remainder of the chunk we're in.
420 */
421 struct xfs_inobt_rec_incore r;
422
423 error = xfs_bulkstat_grab_ichunk(cur, agino, &icount, &r);
424 if (error)
425 goto del_cursor;
426 if (icount) {
427 irbp->ir_startino = r.ir_startino;
428 irbp->ir_holemask = r.ir_holemask;
429 irbp->ir_count = r.ir_count;
430 irbp->ir_freecount = r.ir_freecount;
431 irbp->ir_free = r.ir_free;
432 irbp++;
433 }
434 /* Increment to the next record */
435 error = xfs_btree_increment(cur, 0, &stat);
436 } else {
437 /* Start of ag. Lookup the first inode chunk */
438 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &stat);
439 }
440 if (error || stat == 0) {
441 end_of_ag = true;
442 goto del_cursor;
443 }
444
445 /*
446 * Loop through inode btree records in this ag,
447 * until we run out of inodes or space in the buffer.
448 */
449 while (irbp < irbufend && icount < ubcount) {
450 struct xfs_inobt_rec_incore r;
451
452 error = xfs_inobt_get_rec(cur, &r, &stat);
453 if (error || stat == 0) {
454 end_of_ag = true;
455 goto del_cursor;
456 }
457
458 /*
459 * If this chunk has any allocated inodes, save it.
460 * Also start read-ahead now for this chunk.
461 */
462 if (r.ir_freecount < r.ir_count) {
463 xfs_bulkstat_ichunk_ra(mp, agno, &r);
464 irbp->ir_startino = r.ir_startino;
465 irbp->ir_holemask = r.ir_holemask;
466 irbp->ir_count = r.ir_count;
467 irbp->ir_freecount = r.ir_freecount;
468 irbp->ir_free = r.ir_free;
469 irbp++;
470 icount += r.ir_count - r.ir_freecount;
471 }
472 error = xfs_btree_increment(cur, 0, &stat);
473 if (error || stat == 0) {
474 end_of_ag = true;
475 goto del_cursor;
476 }
477 cond_resched();
478 }
479
480 /*
481 * Drop the btree buffers and the agi buffer as we can't hold any
482 * of the locks these represent when calling iget. If there is a
483 * pending error, then we are done.
484 */
485del_cursor:
486 xfs_btree_del_cursor(cur, error ?
487 XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
488 xfs_buf_relse(agbp);
489 if (error)
490 break;
491 /*
492 * Now format all the good inodes into the user's buffer. The
493 * call to xfs_bulkstat_ag_ichunk() sets up the agino pointer
494 * for the next loop iteration.
495 */
496 irbufend = irbp;
497 for (irbp = irbuf;
498 irbp < irbufend && ac.ac_ubleft >= statstruct_size;
499 irbp++) {
500 error = xfs_bulkstat_ag_ichunk(mp, agno, irbp,
501 formatter, statstruct_size, &ac,
502 &agino);
503 if (error)
504 break;
505
506 cond_resched();
507 }
508
509 /*
510 * If we've run out of space or had a formatting error, we
511 * are now done
512 */
513 if (ac.ac_ubleft < statstruct_size || error)
514 break;
515
516 if (end_of_ag) {
517 agno++;
518 agino = 0;
519 }
520 }
521 /*
522 * Done, we're either out of filesystem or space to put the data.
523 */
524 kmem_free(irbuf);
525 *ubcountp = ac.ac_ubelem;
526
527 /*
528 * We found some inodes, so clear the error status and return them.
529 * The lastino pointer will point directly at the inode that triggered
530 * any error that occurred, so on the next call the error will be
531 * triggered again and propagated to userspace as there will be no
532 * formatted inodes in the buffer.
533 */
534 if (ac.ac_ubelem)
535 error = 0;
536
537 /*
538 * If we ran out of filesystem, lastino will point off the end of
539 * the filesystem so the next call will return immediately.
540 */
541 *lastinop = XFS_AGINO_TO_INO(mp, agno, agino);
542 if (agno >= mp->m_sb.sb_agcount)
543 *done = 1;
544
545 return error;
546}
547
548int
549xfs_inumbers_fmt(
550 void __user *ubuffer, /* buffer to write to */
551 const struct xfs_inogrp *buffer, /* buffer to read from */
552 long count, /* # of elements to read */
553 long *written) /* # of bytes written */
554{
555 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
556 return -EFAULT;
557 *written = count * sizeof(*buffer);
558 return 0;
559}
560
561/*
562 * Return inode number table for the filesystem.
563 */
564int /* error status */
565xfs_inumbers(
566 struct xfs_mount *mp,/* mount point for filesystem */
567 xfs_ino_t *lastino,/* last inode returned */
568 int *count,/* size of buffer/count returned */
569 void __user *ubuffer,/* buffer with inode descriptions */
570 inumbers_fmt_pf formatter)
571{
572 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, *lastino);
573 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, *lastino);
574 struct xfs_btree_cur *cur = NULL;
575 struct xfs_buf *agbp = NULL;
576 struct xfs_inogrp *buffer;
577 int bcount;
578 int left = *count;
579 int bufidx = 0;
580 int error = 0;
581
582 *count = 0;
583 if (agno >= mp->m_sb.sb_agcount ||
584 *lastino != XFS_AGINO_TO_INO(mp, agno, agino))
585 return error;
586
587 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
588 buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
589 do {
590 struct xfs_inobt_rec_incore r;
591 int stat;
592
593 if (!agbp) {
594 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
595 if (error)
596 break;
597
598 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno,
599 XFS_BTNUM_INO);
600 error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
601 &stat);
602 if (error)
603 break;
604 if (!stat)
605 goto next_ag;
606 }
607
608 error = xfs_inobt_get_rec(cur, &r, &stat);
609 if (error)
610 break;
611 if (!stat)
612 goto next_ag;
613
614 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
615 buffer[bufidx].xi_startino =
616 XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
617 buffer[bufidx].xi_alloccount = r.ir_count - r.ir_freecount;
618 buffer[bufidx].xi_allocmask = ~r.ir_free;
619 if (++bufidx == bcount) {
620 long written;
621
622 error = formatter(ubuffer, buffer, bufidx, &written);
623 if (error)
624 break;
625 ubuffer += written;
626 *count += bufidx;
627 bufidx = 0;
628 }
629 if (!--left)
630 break;
631
632 error = xfs_btree_increment(cur, 0, &stat);
633 if (error)
634 break;
635 if (stat)
636 continue;
637
638next_ag:
639 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
640 cur = NULL;
641 xfs_buf_relse(agbp);
642 agbp = NULL;
643 agino = 0;
644 agno++;
645 } while (agno < mp->m_sb.sb_agcount);
646
647 if (!error) {
648 if (bufidx) {
649 long written;
650
651 error = formatter(ubuffer, buffer, bufidx, &written);
652 if (!error)
653 *count += bufidx;
654 }
655 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
656 }
657
658 kmem_free(buffer);
659 if (cur)
660 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
661 XFS_BTREE_NOERROR));
662 if (agbp)
663 xfs_buf_relse(agbp);
664
665 return error;
666}
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 user_namespace *mnt_userns,
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 ASSERT(ip != NULL);
84 ASSERT(ip->i_imap.im_blkno != 0);
85 inode = VFS_I(ip);
86 vfsuid = i_uid_into_vfsuid(mnt_userns, inode);
87 vfsgid = i_gid_into_vfsgid(mnt_userns, inode);
88
89 /* xfs_iget returns the following without needing
90 * further change.
91 */
92 buf->bs_projectid = ip->i_projid;
93 buf->bs_ino = ino;
94 buf->bs_uid = from_kuid(sb_userns, vfsuid_into_kuid(vfsuid));
95 buf->bs_gid = from_kgid(sb_userns, vfsgid_into_kgid(vfsgid));
96 buf->bs_size = ip->i_disk_size;
97
98 buf->bs_nlink = inode->i_nlink;
99 buf->bs_atime = inode->i_atime.tv_sec;
100 buf->bs_atime_nsec = inode->i_atime.tv_nsec;
101 buf->bs_mtime = inode->i_mtime.tv_sec;
102 buf->bs_mtime_nsec = inode->i_mtime.tv_nsec;
103 buf->bs_ctime = inode->i_ctime.tv_sec;
104 buf->bs_ctime_nsec = inode->i_ctime.tv_nsec;
105 buf->bs_gen = inode->i_generation;
106 buf->bs_mode = inode->i_mode;
107
108 buf->bs_xflags = xfs_ip2xflags(ip);
109 buf->bs_extsize_blks = ip->i_extsize;
110
111 nextents = xfs_ifork_nextents(&ip->i_df);
112 if (!(bc->breq->flags & XFS_IBULK_NREXT64))
113 buf->bs_extents = min(nextents, XFS_MAX_EXTCNT_DATA_FORK_SMALL);
114 else
115 buf->bs_extents64 = nextents;
116
117 xfs_bulkstat_health(ip, buf);
118 buf->bs_aextents = xfs_ifork_nextents(&ip->i_af);
119 buf->bs_forkoff = xfs_inode_fork_boff(ip);
120 buf->bs_version = XFS_BULKSTAT_VERSION_V5;
121
122 if (xfs_has_v3inodes(mp)) {
123 buf->bs_btime = ip->i_crtime.tv_sec;
124 buf->bs_btime_nsec = ip->i_crtime.tv_nsec;
125 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
126 buf->bs_cowextsize_blks = ip->i_cowextsize;
127 }
128
129 switch (ip->i_df.if_format) {
130 case XFS_DINODE_FMT_DEV:
131 buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
132 buf->bs_blksize = BLKDEV_IOSIZE;
133 buf->bs_blocks = 0;
134 break;
135 case XFS_DINODE_FMT_LOCAL:
136 buf->bs_rdev = 0;
137 buf->bs_blksize = mp->m_sb.sb_blocksize;
138 buf->bs_blocks = 0;
139 break;
140 case XFS_DINODE_FMT_EXTENTS:
141 case XFS_DINODE_FMT_BTREE:
142 buf->bs_rdev = 0;
143 buf->bs_blksize = mp->m_sb.sb_blocksize;
144 buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks;
145 break;
146 }
147 xfs_iunlock(ip, XFS_ILOCK_SHARED);
148 xfs_irele(ip);
149
150 error = bc->formatter(bc->breq, buf);
151 if (error == -ECANCELED)
152 goto out_advance;
153 if (error)
154 goto out;
155
156out_advance:
157 /*
158 * Advance the cursor to the inode that comes after the one we just
159 * looked at. We want the caller to move along if the bulkstat
160 * information was copied successfully; if we tried to grab the inode
161 * but it's no longer allocated; or if it's internal metadata.
162 */
163 bc->breq->startino = ino + 1;
164out:
165 return error;
166}
167
168/* Bulkstat a single inode. */
169int
170xfs_bulkstat_one(
171 struct xfs_ibulk *breq,
172 bulkstat_one_fmt_pf formatter)
173{
174 struct xfs_bstat_chunk bc = {
175 .formatter = formatter,
176 .breq = breq,
177 };
178 struct xfs_trans *tp;
179 int error;
180
181 if (breq->mnt_userns != &init_user_ns) {
182 xfs_warn_ratelimited(breq->mp,
183 "bulkstat not supported inside of idmapped mounts.");
184 return -EINVAL;
185 }
186
187 ASSERT(breq->icount == 1);
188
189 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
190 KM_MAYFAIL);
191 if (!bc.buf)
192 return -ENOMEM;
193
194 /*
195 * Grab an empty transaction so that we can use its recursive buffer
196 * locking abilities to detect cycles in the inobt without deadlocking.
197 */
198 error = xfs_trans_alloc_empty(breq->mp, &tp);
199 if (error)
200 goto out;
201
202 error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, tp,
203 breq->startino, &bc);
204 xfs_trans_cancel(tp);
205out:
206 kmem_free(bc.buf);
207
208 /*
209 * If we reported one inode to userspace then we abort because we hit
210 * the end of the buffer. Don't leak that back to userspace.
211 */
212 if (error == -ECANCELED)
213 error = 0;
214
215 return error;
216}
217
218static int
219xfs_bulkstat_iwalk(
220 struct xfs_mount *mp,
221 struct xfs_trans *tp,
222 xfs_ino_t ino,
223 void *data)
224{
225 struct xfs_bstat_chunk *bc = data;
226 int error;
227
228 error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data);
229 /* bulkstat just skips over missing inodes */
230 if (error == -ENOENT || error == -EINVAL)
231 return 0;
232 return error;
233}
234
235/*
236 * Check the incoming lastino parameter.
237 *
238 * We allow any inode value that could map to physical space inside the
239 * filesystem because if there are no inodes there, bulkstat moves on to the
240 * next chunk. In other words, the magic agino value of zero takes us to the
241 * first chunk in the AG, and an agino value past the end of the AG takes us to
242 * the first chunk in the next AG.
243 *
244 * Therefore we can end early if the requested inode is beyond the end of the
245 * filesystem or doesn't map properly.
246 */
247static inline bool
248xfs_bulkstat_already_done(
249 struct xfs_mount *mp,
250 xfs_ino_t startino)
251{
252 xfs_agnumber_t agno = XFS_INO_TO_AGNO(mp, startino);
253 xfs_agino_t agino = XFS_INO_TO_AGINO(mp, startino);
254
255 return agno >= mp->m_sb.sb_agcount ||
256 startino != XFS_AGINO_TO_INO(mp, agno, agino);
257}
258
259/* Return stat information in bulk (by-inode) for the filesystem. */
260int
261xfs_bulkstat(
262 struct xfs_ibulk *breq,
263 bulkstat_one_fmt_pf formatter)
264{
265 struct xfs_bstat_chunk bc = {
266 .formatter = formatter,
267 .breq = breq,
268 };
269 struct xfs_trans *tp;
270 unsigned int iwalk_flags = 0;
271 int error;
272
273 if (breq->mnt_userns != &init_user_ns) {
274 xfs_warn_ratelimited(breq->mp,
275 "bulkstat not supported inside of idmapped mounts.");
276 return -EINVAL;
277 }
278 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
279 return 0;
280
281 bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
282 KM_MAYFAIL);
283 if (!bc.buf)
284 return -ENOMEM;
285
286 /*
287 * Grab an empty transaction so that we can use its recursive buffer
288 * locking abilities to detect cycles in the inobt without deadlocking.
289 */
290 error = xfs_trans_alloc_empty(breq->mp, &tp);
291 if (error)
292 goto out;
293
294 if (breq->flags & XFS_IBULK_SAME_AG)
295 iwalk_flags |= XFS_IWALK_SAME_AG;
296
297 error = xfs_iwalk(breq->mp, tp, breq->startino, iwalk_flags,
298 xfs_bulkstat_iwalk, breq->icount, &bc);
299 xfs_trans_cancel(tp);
300out:
301 kmem_free(bc.buf);
302
303 /*
304 * We found some inodes, so clear the error status and return them.
305 * The lastino pointer will point directly at the inode that triggered
306 * any error that occurred, so on the next call the error will be
307 * triggered again and propagated to userspace as there will be no
308 * formatted inodes in the buffer.
309 */
310 if (breq->ocount > 0)
311 error = 0;
312
313 return error;
314}
315
316/* Convert bulkstat (v5) to bstat (v1). */
317void
318xfs_bulkstat_to_bstat(
319 struct xfs_mount *mp,
320 struct xfs_bstat *bs1,
321 const struct xfs_bulkstat *bstat)
322{
323 /* memset is needed here because of padding holes in the structure. */
324 memset(bs1, 0, sizeof(struct xfs_bstat));
325 bs1->bs_ino = bstat->bs_ino;
326 bs1->bs_mode = bstat->bs_mode;
327 bs1->bs_nlink = bstat->bs_nlink;
328 bs1->bs_uid = bstat->bs_uid;
329 bs1->bs_gid = bstat->bs_gid;
330 bs1->bs_rdev = bstat->bs_rdev;
331 bs1->bs_blksize = bstat->bs_blksize;
332 bs1->bs_size = bstat->bs_size;
333 bs1->bs_atime.tv_sec = bstat->bs_atime;
334 bs1->bs_mtime.tv_sec = bstat->bs_mtime;
335 bs1->bs_ctime.tv_sec = bstat->bs_ctime;
336 bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
337 bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
338 bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
339 bs1->bs_blocks = bstat->bs_blocks;
340 bs1->bs_xflags = bstat->bs_xflags;
341 bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
342 bs1->bs_extents = bstat->bs_extents;
343 bs1->bs_gen = bstat->bs_gen;
344 bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
345 bs1->bs_forkoff = bstat->bs_forkoff;
346 bs1->bs_projid_hi = bstat->bs_projectid >> 16;
347 bs1->bs_sick = bstat->bs_sick;
348 bs1->bs_checked = bstat->bs_checked;
349 bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
350 bs1->bs_dmevmask = 0;
351 bs1->bs_dmstate = 0;
352 bs1->bs_aextents = bstat->bs_aextents;
353}
354
355struct xfs_inumbers_chunk {
356 inumbers_fmt_pf formatter;
357 struct xfs_ibulk *breq;
358};
359
360/*
361 * INUMBERS
362 * ========
363 * This is how we export inode btree records to userspace, so that XFS tools
364 * can figure out where inodes are allocated.
365 */
366
367/*
368 * Format the inode group structure and report it somewhere.
369 *
370 * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
371 * through the filesystem so we move it forward unless there was a runtime
372 * error. If the formatter tells us the buffer is now full we also move the
373 * cursor forward and abort the walk.
374 */
375STATIC int
376xfs_inumbers_walk(
377 struct xfs_mount *mp,
378 struct xfs_trans *tp,
379 xfs_agnumber_t agno,
380 const struct xfs_inobt_rec_incore *irec,
381 void *data)
382{
383 struct xfs_inumbers inogrp = {
384 .xi_startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
385 .xi_alloccount = irec->ir_count - irec->ir_freecount,
386 .xi_allocmask = ~irec->ir_free,
387 .xi_version = XFS_INUMBERS_VERSION_V5,
388 };
389 struct xfs_inumbers_chunk *ic = data;
390 int error;
391
392 error = ic->formatter(ic->breq, &inogrp);
393 if (error && error != -ECANCELED)
394 return error;
395
396 ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
397 XFS_INODES_PER_CHUNK;
398 return error;
399}
400
401/*
402 * Return inode number table for the filesystem.
403 */
404int
405xfs_inumbers(
406 struct xfs_ibulk *breq,
407 inumbers_fmt_pf formatter)
408{
409 struct xfs_inumbers_chunk ic = {
410 .formatter = formatter,
411 .breq = breq,
412 };
413 struct xfs_trans *tp;
414 int error = 0;
415
416 if (xfs_bulkstat_already_done(breq->mp, breq->startino))
417 return 0;
418
419 /*
420 * Grab an empty transaction so that we can use its recursive buffer
421 * locking abilities to detect cycles in the inobt without deadlocking.
422 */
423 error = xfs_trans_alloc_empty(breq->mp, &tp);
424 if (error)
425 goto out;
426
427 error = xfs_inobt_walk(breq->mp, tp, breq->startino, breq->flags,
428 xfs_inumbers_walk, breq->icount, &ic);
429 xfs_trans_cancel(tp);
430out:
431
432 /*
433 * We found some inode groups, so clear the error status and return
434 * them. The lastino pointer will point directly at the inode that
435 * triggered any error that occurred, so on the next call the error
436 * will be triggered again and propagated to userspace as there will be
437 * no formatted inode groups in the buffer.
438 */
439 if (breq->ocount > 0)
440 error = 0;
441
442 return error;
443}
444
445/* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
446void
447xfs_inumbers_to_inogrp(
448 struct xfs_inogrp *ig1,
449 const struct xfs_inumbers *ig)
450{
451 /* memset is needed here because of padding holes in the structure. */
452 memset(ig1, 0, sizeof(struct xfs_inogrp));
453 ig1->xi_startino = ig->xi_startino;
454 ig1->xi_alloccount = ig->xi_alloccount;
455 ig1->xi_allocmask = ig->xi_allocmask;
456}