xref: /linux/fs/xfs/xfs_itable.c (revision d39d0ed196aa1685bb24771e92f78633c66ac9cb)
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_types.h"
21 #include "xfs_bit.h"
22 #include "xfs_log.h"
23 #include "xfs_inum.h"
24 #include "xfs_trans.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_mount.h"
28 #include "xfs_bmap_btree.h"
29 #include "xfs_alloc_btree.h"
30 #include "xfs_ialloc_btree.h"
31 #include "xfs_dinode.h"
32 #include "xfs_inode.h"
33 #include "xfs_ialloc.h"
34 #include "xfs_itable.h"
35 #include "xfs_error.h"
36 #include "xfs_btree.h"
37 #include "xfs_trace.h"
38 
39 STATIC int
40 xfs_internal_inum(
41 	xfs_mount_t	*mp,
42 	xfs_ino_t	ino)
43 {
44 	return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino ||
45 		(xfs_sb_version_hasquota(&mp->m_sb) &&
46 		 (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino)));
47 }
48 
49 /*
50  * Return stat information for one inode.
51  * Return 0 if ok, else errno.
52  */
53 int
54 xfs_bulkstat_one_int(
55 	struct xfs_mount	*mp,		/* mount point for filesystem */
56 	xfs_ino_t		ino,		/* inode to get data for */
57 	void __user		*buffer,	/* buffer to place output in */
58 	int			ubsize,		/* size of buffer */
59 	bulkstat_one_fmt_pf	formatter,	/* formatter, copy to user */
60 	int			*ubused,	/* bytes used by me */
61 	int			*stat)		/* BULKSTAT_RV_... */
62 {
63 	struct xfs_icdinode	*dic;		/* dinode core info pointer */
64 	struct xfs_inode	*ip;		/* incore inode pointer */
65 	struct inode		*inode;
66 	struct xfs_bstat	*buf;		/* return buffer */
67 	int			error = 0;	/* error value */
68 
69 	*stat = BULKSTAT_RV_NOTHING;
70 
71 	if (!buffer || xfs_internal_inum(mp, ino))
72 		return XFS_ERROR(EINVAL);
73 
74 	buf = kmem_alloc(sizeof(*buf), KM_SLEEP | KM_MAYFAIL);
75 	if (!buf)
76 		return XFS_ERROR(ENOMEM);
77 
78 	error = xfs_iget(mp, NULL, ino,
79 			 XFS_IGET_UNTRUSTED, XFS_ILOCK_SHARED, &ip);
80 	if (error) {
81 		*stat = BULKSTAT_RV_NOTHING;
82 		goto out_free;
83 	}
84 
85 	ASSERT(ip != NULL);
86 	ASSERT(ip->i_imap.im_blkno != 0);
87 
88 	dic = &ip->i_d;
89 	inode = VFS_I(ip);
90 
91 	/* xfs_iget returns the following without needing
92 	 * further change.
93 	 */
94 	buf->bs_nlink = dic->di_nlink;
95 	buf->bs_projid = dic->di_projid;
96 	buf->bs_ino = ino;
97 	buf->bs_mode = dic->di_mode;
98 	buf->bs_uid = dic->di_uid;
99 	buf->bs_gid = dic->di_gid;
100 	buf->bs_size = dic->di_size;
101 
102 	/*
103 	 * We need to read the timestamps from the Linux inode because
104 	 * the VFS keeps writing directly into the inode structure instead
105 	 * of telling us about the updates.
106 	 */
107 	buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
108 	buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
109 	buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
110 	buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
111 	buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
112 	buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
113 
114 	buf->bs_xflags = xfs_ip2xflags(ip);
115 	buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
116 	buf->bs_extents = dic->di_nextents;
117 	buf->bs_gen = dic->di_gen;
118 	memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
119 	buf->bs_dmevmask = dic->di_dmevmask;
120 	buf->bs_dmstate = dic->di_dmstate;
121 	buf->bs_aextents = dic->di_anextents;
122 	buf->bs_forkoff = XFS_IFORK_BOFF(ip);
123 
124 	switch (dic->di_format) {
125 	case XFS_DINODE_FMT_DEV:
126 		buf->bs_rdev = ip->i_df.if_u2.if_rdev;
127 		buf->bs_blksize = BLKDEV_IOSIZE;
128 		buf->bs_blocks = 0;
129 		break;
130 	case XFS_DINODE_FMT_LOCAL:
131 	case XFS_DINODE_FMT_UUID:
132 		buf->bs_rdev = 0;
133 		buf->bs_blksize = mp->m_sb.sb_blocksize;
134 		buf->bs_blocks = 0;
135 		break;
136 	case XFS_DINODE_FMT_EXTENTS:
137 	case XFS_DINODE_FMT_BTREE:
138 		buf->bs_rdev = 0;
139 		buf->bs_blksize = mp->m_sb.sb_blocksize;
140 		buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
141 		break;
142 	}
143 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
144 	IRELE(ip);
145 
146 	error = formatter(buffer, ubsize, ubused, buf);
147 
148 	if (!error)
149 		*stat = BULKSTAT_RV_DIDONE;
150 
151  out_free:
152 	kmem_free(buf);
153 	return error;
154 }
155 
156 /* Return 0 on success or positive error */
157 STATIC int
158 xfs_bulkstat_one_fmt(
159 	void			__user *ubuffer,
160 	int			ubsize,
161 	int			*ubused,
162 	const xfs_bstat_t	*buffer)
163 {
164 	if (ubsize < sizeof(*buffer))
165 		return XFS_ERROR(ENOMEM);
166 	if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
167 		return XFS_ERROR(EFAULT);
168 	if (ubused)
169 		*ubused = sizeof(*buffer);
170 	return 0;
171 }
172 
173 int
174 xfs_bulkstat_one(
175 	xfs_mount_t	*mp,		/* mount point for filesystem */
176 	xfs_ino_t	ino,		/* inode number to get data for */
177 	void		__user *buffer,	/* buffer to place output in */
178 	int		ubsize,		/* size of buffer */
179 	int		*ubused,	/* bytes used by me */
180 	int		*stat)		/* BULKSTAT_RV_... */
181 {
182 	return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
183 				    xfs_bulkstat_one_fmt, ubused, stat);
184 }
185 
186 #define XFS_BULKSTAT_UBLEFT(ubleft)	((ubleft) >= statstruct_size)
187 
188 /*
189  * Return stat information in bulk (by-inode) for the filesystem.
190  */
191 int					/* error status */
192 xfs_bulkstat(
193 	xfs_mount_t		*mp,	/* mount point for filesystem */
194 	xfs_ino_t		*lastinop, /* last inode returned */
195 	int			*ubcountp, /* size of buffer/count returned */
196 	bulkstat_one_pf		formatter, /* func that'd fill a single buf */
197 	size_t			statstruct_size, /* sizeof struct filling */
198 	char			__user *ubuffer, /* buffer with inode stats */
199 	int			*done)	/* 1 if there are more stats to get */
200 {
201 	xfs_agblock_t		agbno=0;/* allocation group block number */
202 	xfs_buf_t		*agbp;	/* agi header buffer */
203 	xfs_agi_t		*agi;	/* agi header data */
204 	xfs_agino_t		agino;	/* inode # in allocation group */
205 	xfs_agnumber_t		agno;	/* allocation group number */
206 	xfs_daddr_t		bno;	/* inode cluster start daddr */
207 	int			chunkidx; /* current index into inode chunk */
208 	int			clustidx; /* current index into inode cluster */
209 	xfs_btree_cur_t		*cur;	/* btree cursor for ialloc btree */
210 	int			end_of_ag; /* set if we've seen the ag end */
211 	int			error;	/* error code */
212 	int                     fmterror;/* bulkstat formatter result */
213 	int			i;	/* loop index */
214 	int			icount;	/* count of inodes good in irbuf */
215 	size_t			irbsize; /* size of irec buffer in bytes */
216 	xfs_ino_t		ino;	/* inode number (filesystem) */
217 	xfs_inobt_rec_incore_t	*irbp;	/* current irec buffer pointer */
218 	xfs_inobt_rec_incore_t	*irbuf;	/* start of irec buffer */
219 	xfs_inobt_rec_incore_t	*irbufend; /* end of good irec buffer entries */
220 	xfs_ino_t		lastino; /* last inode number returned */
221 	int			nbcluster; /* # of blocks in a cluster */
222 	int			nicluster; /* # of inodes in a cluster */
223 	int			nimask;	/* mask for inode clusters */
224 	int			nirbuf;	/* size of irbuf */
225 	int			rval;	/* return value error code */
226 	int			tmp;	/* result value from btree calls */
227 	int			ubcount; /* size of user's buffer */
228 	int			ubleft;	/* bytes left in user's buffer */
229 	char			__user *ubufp;	/* pointer into user's buffer */
230 	int			ubelem;	/* spaces used in user's buffer */
231 	int			ubused;	/* bytes used by formatter */
232 	xfs_buf_t		*bp;	/* ptr to on-disk inode cluster buf */
233 
234 	/*
235 	 * Get the last inode value, see if there's nothing to do.
236 	 */
237 	ino = (xfs_ino_t)*lastinop;
238 	lastino = ino;
239 	agno = XFS_INO_TO_AGNO(mp, ino);
240 	agino = XFS_INO_TO_AGINO(mp, ino);
241 	if (agno >= mp->m_sb.sb_agcount ||
242 	    ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
243 		*done = 1;
244 		*ubcountp = 0;
245 		return 0;
246 	}
247 	if (!ubcountp || *ubcountp <= 0) {
248 		return EINVAL;
249 	}
250 	ubcount = *ubcountp; /* statstruct's */
251 	ubleft = ubcount * statstruct_size; /* bytes */
252 	*ubcountp = ubelem = 0;
253 	*done = 0;
254 	fmterror = 0;
255 	ubufp = ubuffer;
256 	nicluster = mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp) ?
257 		mp->m_sb.sb_inopblock :
258 		(XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog);
259 	nimask = ~(nicluster - 1);
260 	nbcluster = nicluster >> mp->m_sb.sb_inopblog;
261 	irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4);
262 	if (!irbuf)
263 		return ENOMEM;
264 
265 	nirbuf = irbsize / sizeof(*irbuf);
266 
267 	/*
268 	 * Loop over the allocation groups, starting from the last
269 	 * inode returned; 0 means start of the allocation group.
270 	 */
271 	rval = 0;
272 	while (XFS_BULKSTAT_UBLEFT(ubleft) && agno < mp->m_sb.sb_agcount) {
273 		cond_resched();
274 		bp = NULL;
275 		error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
276 		if (error) {
277 			/*
278 			 * Skip this allocation group and go to the next one.
279 			 */
280 			agno++;
281 			agino = 0;
282 			continue;
283 		}
284 		agi = XFS_BUF_TO_AGI(agbp);
285 		/*
286 		 * Allocate and initialize a btree cursor for ialloc btree.
287 		 */
288 		cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
289 		irbp = irbuf;
290 		irbufend = irbuf + nirbuf;
291 		end_of_ag = 0;
292 		/*
293 		 * If we're returning in the middle of an allocation group,
294 		 * we need to get the remainder of the chunk we're in.
295 		 */
296 		if (agino > 0) {
297 			xfs_inobt_rec_incore_t r;
298 
299 			/*
300 			 * Lookup the inode chunk that this inode lives in.
301 			 */
302 			error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE,
303 						 &tmp);
304 			if (!error &&	/* no I/O error */
305 			    tmp &&	/* lookup succeeded */
306 					/* got the record, should always work */
307 			    !(error = xfs_inobt_get_rec(cur, &r, &i)) &&
308 			    i == 1 &&
309 					/* this is the right chunk */
310 			    agino < r.ir_startino + XFS_INODES_PER_CHUNK &&
311 					/* lastino was not last in chunk */
312 			    (chunkidx = agino - r.ir_startino + 1) <
313 				    XFS_INODES_PER_CHUNK &&
314 					/* there are some left allocated */
315 			    xfs_inobt_maskn(chunkidx,
316 				    XFS_INODES_PER_CHUNK - chunkidx) &
317 				    ~r.ir_free) {
318 				/*
319 				 * Grab the chunk record.  Mark all the
320 				 * uninteresting inodes (because they're
321 				 * before our start point) free.
322 				 */
323 				for (i = 0; i < chunkidx; i++) {
324 					if (XFS_INOBT_MASK(i) & ~r.ir_free)
325 						r.ir_freecount++;
326 				}
327 				r.ir_free |= xfs_inobt_maskn(0, chunkidx);
328 				irbp->ir_startino = r.ir_startino;
329 				irbp->ir_freecount = r.ir_freecount;
330 				irbp->ir_free = r.ir_free;
331 				irbp++;
332 				agino = r.ir_startino + XFS_INODES_PER_CHUNK;
333 				icount = XFS_INODES_PER_CHUNK - r.ir_freecount;
334 			} else {
335 				/*
336 				 * If any of those tests failed, bump the
337 				 * inode number (just in case).
338 				 */
339 				agino++;
340 				icount = 0;
341 			}
342 			/*
343 			 * In any case, increment to the next record.
344 			 */
345 			if (!error)
346 				error = xfs_btree_increment(cur, 0, &tmp);
347 		} else {
348 			/*
349 			 * Start of ag.  Lookup the first inode chunk.
350 			 */
351 			error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &tmp);
352 			icount = 0;
353 		}
354 		/*
355 		 * Loop through inode btree records in this ag,
356 		 * until we run out of inodes or space in the buffer.
357 		 */
358 		while (irbp < irbufend && icount < ubcount) {
359 			xfs_inobt_rec_incore_t r;
360 
361 			/*
362 			 * Loop as long as we're unable to read the
363 			 * inode btree.
364 			 */
365 			while (error) {
366 				agino += XFS_INODES_PER_CHUNK;
367 				if (XFS_AGINO_TO_AGBNO(mp, agino) >=
368 						be32_to_cpu(agi->agi_length))
369 					break;
370 				error = xfs_inobt_lookup(cur, agino,
371 							 XFS_LOOKUP_GE, &tmp);
372 				cond_resched();
373 			}
374 			/*
375 			 * If ran off the end of the ag either with an error,
376 			 * or the normal way, set end and stop collecting.
377 			 */
378 			if (error) {
379 				end_of_ag = 1;
380 				break;
381 			}
382 
383 			error = xfs_inobt_get_rec(cur, &r, &i);
384 			if (error || i == 0) {
385 				end_of_ag = 1;
386 				break;
387 			}
388 
389 			/*
390 			 * If this chunk has any allocated inodes, save it.
391 			 * Also start read-ahead now for this chunk.
392 			 */
393 			if (r.ir_freecount < XFS_INODES_PER_CHUNK) {
394 				/*
395 				 * Loop over all clusters in the next chunk.
396 				 * Do a readahead if there are any allocated
397 				 * inodes in that cluster.
398 				 */
399 				agbno = XFS_AGINO_TO_AGBNO(mp, r.ir_startino);
400 				for (chunkidx = 0;
401 				     chunkidx < XFS_INODES_PER_CHUNK;
402 				     chunkidx += nicluster,
403 				     agbno += nbcluster) {
404 					if (xfs_inobt_maskn(chunkidx, nicluster)
405 							& ~r.ir_free)
406 						xfs_btree_reada_bufs(mp, agno,
407 							agbno, nbcluster);
408 				}
409 				irbp->ir_startino = r.ir_startino;
410 				irbp->ir_freecount = r.ir_freecount;
411 				irbp->ir_free = r.ir_free;
412 				irbp++;
413 				icount += XFS_INODES_PER_CHUNK - r.ir_freecount;
414 			}
415 			/*
416 			 * Set agino to after this chunk and bump the cursor.
417 			 */
418 			agino = r.ir_startino + XFS_INODES_PER_CHUNK;
419 			error = xfs_btree_increment(cur, 0, &tmp);
420 			cond_resched();
421 		}
422 		/*
423 		 * Drop the btree buffers and the agi buffer.
424 		 * We can't hold any of the locks these represent
425 		 * when calling iget.
426 		 */
427 		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
428 		xfs_buf_relse(agbp);
429 		/*
430 		 * Now format all the good inodes into the user's buffer.
431 		 */
432 		irbufend = irbp;
433 		for (irbp = irbuf;
434 		     irbp < irbufend && XFS_BULKSTAT_UBLEFT(ubleft); irbp++) {
435 			/*
436 			 * Now process this chunk of inodes.
437 			 */
438 			for (agino = irbp->ir_startino, chunkidx = clustidx = 0;
439 			     XFS_BULKSTAT_UBLEFT(ubleft) &&
440 				irbp->ir_freecount < XFS_INODES_PER_CHUNK;
441 			     chunkidx++, clustidx++, agino++) {
442 				ASSERT(chunkidx < XFS_INODES_PER_CHUNK);
443 				/*
444 				 * Recompute agbno if this is the
445 				 * first inode of the cluster.
446 				 *
447 				 * Careful with clustidx.   There can be
448 				 * multiple clusters per chunk, a single
449 				 * cluster per chunk or a cluster that has
450 				 * inodes represented from several different
451 				 * chunks (if blocksize is large).
452 				 *
453 				 * Because of this, the starting clustidx is
454 				 * initialized to zero in this loop but must
455 				 * later be reset after reading in the cluster
456 				 * buffer.
457 				 */
458 				if ((chunkidx & (nicluster - 1)) == 0) {
459 					agbno = XFS_AGINO_TO_AGBNO(mp,
460 							irbp->ir_startino) +
461 						((chunkidx & nimask) >>
462 						 mp->m_sb.sb_inopblog);
463 				}
464 				ino = XFS_AGINO_TO_INO(mp, agno, agino);
465 				bno = XFS_AGB_TO_DADDR(mp, agno, agbno);
466 				/*
467 				 * Skip if this inode is free.
468 				 */
469 				if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) {
470 					lastino = ino;
471 					continue;
472 				}
473 				/*
474 				 * Count used inodes as free so we can tell
475 				 * when the chunk is used up.
476 				 */
477 				irbp->ir_freecount++;
478 
479 				/*
480 				 * Get the inode and fill in a single buffer.
481 				 */
482 				ubused = statstruct_size;
483 				error = formatter(mp, ino, ubufp, ubleft,
484 						  &ubused, &fmterror);
485 				if (fmterror == BULKSTAT_RV_NOTHING) {
486 					if (error && error != ENOENT &&
487 						error != EINVAL) {
488 						ubleft = 0;
489 						rval = error;
490 						break;
491 					}
492 					lastino = ino;
493 					continue;
494 				}
495 				if (fmterror == BULKSTAT_RV_GIVEUP) {
496 					ubleft = 0;
497 					ASSERT(error);
498 					rval = error;
499 					break;
500 				}
501 				if (ubufp)
502 					ubufp += ubused;
503 				ubleft -= ubused;
504 				ubelem++;
505 				lastino = ino;
506 			}
507 
508 			cond_resched();
509 		}
510 
511 		if (bp)
512 			xfs_buf_relse(bp);
513 
514 		/*
515 		 * Set up for the next loop iteration.
516 		 */
517 		if (XFS_BULKSTAT_UBLEFT(ubleft)) {
518 			if (end_of_ag) {
519 				agno++;
520 				agino = 0;
521 			} else
522 				agino = XFS_INO_TO_AGINO(mp, lastino);
523 		} else
524 			break;
525 	}
526 	/*
527 	 * Done, we're either out of filesystem or space to put the data.
528 	 */
529 	kmem_free_large(irbuf);
530 	*ubcountp = ubelem;
531 	/*
532 	 * Found some inodes, return them now and return the error next time.
533 	 */
534 	if (ubelem)
535 		rval = 0;
536 	if (agno >= mp->m_sb.sb_agcount) {
537 		/*
538 		 * If we ran out of filesystem, mark lastino as off
539 		 * the end of the filesystem, so the next call
540 		 * will return immediately.
541 		 */
542 		*lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0);
543 		*done = 1;
544 	} else
545 		*lastinop = (xfs_ino_t)lastino;
546 
547 	return rval;
548 }
549 
550 /*
551  * Return stat information in bulk (by-inode) for the filesystem.
552  * Special case for non-sequential one inode bulkstat.
553  */
554 int					/* error status */
555 xfs_bulkstat_single(
556 	xfs_mount_t		*mp,	/* mount point for filesystem */
557 	xfs_ino_t		*lastinop, /* inode to return */
558 	char			__user *buffer, /* buffer with inode stats */
559 	int			*done)	/* 1 if there are more stats to get */
560 {
561 	int			count;	/* count value for bulkstat call */
562 	int			error;	/* return value */
563 	xfs_ino_t		ino;	/* filesystem inode number */
564 	int			res;	/* result from bs1 */
565 
566 	/*
567 	 * note that requesting valid inode numbers which are not allocated
568 	 * to inodes will most likely cause xfs_itobp to generate warning
569 	 * messages about bad magic numbers. This is ok. The fact that
570 	 * the inode isn't actually an inode is handled by the
571 	 * error check below. Done this way to make the usual case faster
572 	 * at the expense of the error case.
573 	 */
574 
575 	ino = (xfs_ino_t)*lastinop;
576 	error = xfs_bulkstat_one(mp, ino, buffer, sizeof(xfs_bstat_t), 0, &res);
577 	if (error) {
578 		/*
579 		 * Special case way failed, do it the "long" way
580 		 * to see if that works.
581 		 */
582 		(*lastinop)--;
583 		count = 1;
584 		if (xfs_bulkstat(mp, lastinop, &count, xfs_bulkstat_one,
585 				sizeof(xfs_bstat_t), buffer, done))
586 			return error;
587 		if (count == 0 || (xfs_ino_t)*lastinop != ino)
588 			return error == EFSCORRUPTED ?
589 				XFS_ERROR(EINVAL) : error;
590 		else
591 			return 0;
592 	}
593 	*done = 0;
594 	return 0;
595 }
596 
597 int
598 xfs_inumbers_fmt(
599 	void			__user *ubuffer, /* buffer to write to */
600 	const xfs_inogrp_t	*buffer,	/* buffer to read from */
601 	long			count,		/* # of elements to read */
602 	long			*written)	/* # of bytes written */
603 {
604 	if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
605 		return -EFAULT;
606 	*written = count * sizeof(*buffer);
607 	return 0;
608 }
609 
610 /*
611  * Return inode number table for the filesystem.
612  */
613 int					/* error status */
614 xfs_inumbers(
615 	xfs_mount_t	*mp,		/* mount point for filesystem */
616 	xfs_ino_t	*lastino,	/* last inode returned */
617 	int		*count,		/* size of buffer/count returned */
618 	void		__user *ubuffer,/* buffer with inode descriptions */
619 	inumbers_fmt_pf	formatter)
620 {
621 	xfs_buf_t	*agbp;
622 	xfs_agino_t	agino;
623 	xfs_agnumber_t	agno;
624 	int		bcount;
625 	xfs_inogrp_t	*buffer;
626 	int		bufidx;
627 	xfs_btree_cur_t	*cur;
628 	int		error;
629 	xfs_inobt_rec_incore_t r;
630 	int		i;
631 	xfs_ino_t	ino;
632 	int		left;
633 	int		tmp;
634 
635 	ino = (xfs_ino_t)*lastino;
636 	agno = XFS_INO_TO_AGNO(mp, ino);
637 	agino = XFS_INO_TO_AGINO(mp, ino);
638 	left = *count;
639 	*count = 0;
640 	bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
641 	buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
642 	error = bufidx = 0;
643 	cur = NULL;
644 	agbp = NULL;
645 	while (left > 0 && agno < mp->m_sb.sb_agcount) {
646 		if (agbp == NULL) {
647 			error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
648 			if (error) {
649 				/*
650 				 * If we can't read the AGI of this ag,
651 				 * then just skip to the next one.
652 				 */
653 				ASSERT(cur == NULL);
654 				agbp = NULL;
655 				agno++;
656 				agino = 0;
657 				continue;
658 			}
659 			cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
660 			error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
661 						 &tmp);
662 			if (error) {
663 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
664 				cur = NULL;
665 				xfs_buf_relse(agbp);
666 				agbp = NULL;
667 				/*
668 				 * Move up the last inode in the current
669 				 * chunk.  The lookup_ge will always get
670 				 * us the first inode in the next chunk.
671 				 */
672 				agino += XFS_INODES_PER_CHUNK - 1;
673 				continue;
674 			}
675 		}
676 		error = xfs_inobt_get_rec(cur, &r, &i);
677 		if (error || i == 0) {
678 			xfs_buf_relse(agbp);
679 			agbp = NULL;
680 			xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
681 			cur = NULL;
682 			agno++;
683 			agino = 0;
684 			continue;
685 		}
686 		agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
687 		buffer[bufidx].xi_startino =
688 			XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
689 		buffer[bufidx].xi_alloccount =
690 			XFS_INODES_PER_CHUNK - r.ir_freecount;
691 		buffer[bufidx].xi_allocmask = ~r.ir_free;
692 		bufidx++;
693 		left--;
694 		if (bufidx == bcount) {
695 			long written;
696 			if (formatter(ubuffer, buffer, bufidx, &written)) {
697 				error = XFS_ERROR(EFAULT);
698 				break;
699 			}
700 			ubuffer += written;
701 			*count += bufidx;
702 			bufidx = 0;
703 		}
704 		if (left) {
705 			error = xfs_btree_increment(cur, 0, &tmp);
706 			if (error) {
707 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
708 				cur = NULL;
709 				xfs_buf_relse(agbp);
710 				agbp = NULL;
711 				/*
712 				 * The agino value has already been bumped.
713 				 * Just try to skip up to it.
714 				 */
715 				agino += XFS_INODES_PER_CHUNK;
716 				continue;
717 			}
718 		}
719 	}
720 	if (!error) {
721 		if (bufidx) {
722 			long written;
723 			if (formatter(ubuffer, buffer, bufidx, &written))
724 				error = XFS_ERROR(EFAULT);
725 			else
726 				*count += bufidx;
727 		}
728 		*lastino = XFS_AGINO_TO_INO(mp, agno, agino);
729 	}
730 	kmem_free(buffer);
731 	if (cur)
732 		xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
733 					   XFS_BTREE_NOERROR));
734 	if (agbp)
735 		xfs_buf_relse(agbp);
736 	return error;
737 }
738