xref: /openbmc/linux/fs/xfs/xfs_itable.c (revision 565d76cb)
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_lo = dic->di_projid_lo;
96 	buf->bs_projid_hi = dic->di_projid_hi;
97 	buf->bs_ino = ino;
98 	buf->bs_mode = dic->di_mode;
99 	buf->bs_uid = dic->di_uid;
100 	buf->bs_gid = dic->di_gid;
101 	buf->bs_size = dic->di_size;
102 
103 	/*
104 	 * We need to read the timestamps from the Linux inode because
105 	 * the VFS keeps writing directly into the inode structure instead
106 	 * of telling us about the updates.
107 	 */
108 	buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
109 	buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
110 	buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
111 	buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
112 	buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
113 	buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
114 
115 	buf->bs_xflags = xfs_ip2xflags(ip);
116 	buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
117 	buf->bs_extents = dic->di_nextents;
118 	buf->bs_gen = dic->di_gen;
119 	memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
120 	buf->bs_dmevmask = dic->di_dmevmask;
121 	buf->bs_dmstate = dic->di_dmstate;
122 	buf->bs_aextents = dic->di_anextents;
123 	buf->bs_forkoff = XFS_IFORK_BOFF(ip);
124 
125 	switch (dic->di_format) {
126 	case XFS_DINODE_FMT_DEV:
127 		buf->bs_rdev = ip->i_df.if_u2.if_rdev;
128 		buf->bs_blksize = BLKDEV_IOSIZE;
129 		buf->bs_blocks = 0;
130 		break;
131 	case XFS_DINODE_FMT_LOCAL:
132 	case XFS_DINODE_FMT_UUID:
133 		buf->bs_rdev = 0;
134 		buf->bs_blksize = mp->m_sb.sb_blocksize;
135 		buf->bs_blocks = 0;
136 		break;
137 	case XFS_DINODE_FMT_EXTENTS:
138 	case XFS_DINODE_FMT_BTREE:
139 		buf->bs_rdev = 0;
140 		buf->bs_blksize = mp->m_sb.sb_blocksize;
141 		buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
142 		break;
143 	}
144 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
145 	IRELE(ip);
146 
147 	error = formatter(buffer, ubsize, ubused, buf);
148 
149 	if (!error)
150 		*stat = BULKSTAT_RV_DIDONE;
151 
152  out_free:
153 	kmem_free(buf);
154 	return error;
155 }
156 
157 /* Return 0 on success or positive error */
158 STATIC int
159 xfs_bulkstat_one_fmt(
160 	void			__user *ubuffer,
161 	int			ubsize,
162 	int			*ubused,
163 	const xfs_bstat_t	*buffer)
164 {
165 	if (ubsize < sizeof(*buffer))
166 		return XFS_ERROR(ENOMEM);
167 	if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
168 		return XFS_ERROR(EFAULT);
169 	if (ubused)
170 		*ubused = sizeof(*buffer);
171 	return 0;
172 }
173 
174 int
175 xfs_bulkstat_one(
176 	xfs_mount_t	*mp,		/* mount point for filesystem */
177 	xfs_ino_t	ino,		/* inode number to get data for */
178 	void		__user *buffer,	/* buffer to place output in */
179 	int		ubsize,		/* size of buffer */
180 	int		*ubused,	/* bytes used by me */
181 	int		*stat)		/* BULKSTAT_RV_... */
182 {
183 	return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
184 				    xfs_bulkstat_one_fmt, ubused, stat);
185 }
186 
187 #define XFS_BULKSTAT_UBLEFT(ubleft)	((ubleft) >= statstruct_size)
188 
189 /*
190  * Return stat information in bulk (by-inode) for the filesystem.
191  */
192 int					/* error status */
193 xfs_bulkstat(
194 	xfs_mount_t		*mp,	/* mount point for filesystem */
195 	xfs_ino_t		*lastinop, /* last inode returned */
196 	int			*ubcountp, /* size of buffer/count returned */
197 	bulkstat_one_pf		formatter, /* func that'd fill a single buf */
198 	size_t			statstruct_size, /* sizeof struct filling */
199 	char			__user *ubuffer, /* buffer with inode stats */
200 	int			*done)	/* 1 if there are more stats to get */
201 {
202 	xfs_agblock_t		agbno=0;/* allocation group block number */
203 	xfs_buf_t		*agbp;	/* agi header buffer */
204 	xfs_agi_t		*agi;	/* agi header data */
205 	xfs_agino_t		agino;	/* inode # in allocation group */
206 	xfs_agnumber_t		agno;	/* allocation group number */
207 	xfs_daddr_t		bno;	/* inode cluster start daddr */
208 	int			chunkidx; /* current index into inode chunk */
209 	int			clustidx; /* current index into inode cluster */
210 	xfs_btree_cur_t		*cur;	/* btree cursor for ialloc btree */
211 	int			end_of_ag; /* set if we've seen the ag end */
212 	int			error;	/* error code */
213 	int                     fmterror;/* bulkstat formatter result */
214 	int			i;	/* loop index */
215 	int			icount;	/* count of inodes good in irbuf */
216 	size_t			irbsize; /* size of irec buffer in bytes */
217 	xfs_ino_t		ino;	/* inode number (filesystem) */
218 	xfs_inobt_rec_incore_t	*irbp;	/* current irec buffer pointer */
219 	xfs_inobt_rec_incore_t	*irbuf;	/* start of irec buffer */
220 	xfs_inobt_rec_incore_t	*irbufend; /* end of good irec buffer entries */
221 	xfs_ino_t		lastino; /* last inode number returned */
222 	int			nbcluster; /* # of blocks in a cluster */
223 	int			nicluster; /* # of inodes in a cluster */
224 	int			nimask;	/* mask for inode clusters */
225 	int			nirbuf;	/* size of irbuf */
226 	int			rval;	/* return value error code */
227 	int			tmp;	/* result value from btree calls */
228 	int			ubcount; /* size of user's buffer */
229 	int			ubleft;	/* bytes left in user's buffer */
230 	char			__user *ubufp;	/* pointer into user's buffer */
231 	int			ubelem;	/* spaces used in user's buffer */
232 	int			ubused;	/* bytes used by formatter */
233 	xfs_buf_t		*bp;	/* ptr to on-disk inode cluster buf */
234 
235 	/*
236 	 * Get the last inode value, see if there's nothing to do.
237 	 */
238 	ino = (xfs_ino_t)*lastinop;
239 	lastino = ino;
240 	agno = XFS_INO_TO_AGNO(mp, ino);
241 	agino = XFS_INO_TO_AGINO(mp, ino);
242 	if (agno >= mp->m_sb.sb_agcount ||
243 	    ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
244 		*done = 1;
245 		*ubcountp = 0;
246 		return 0;
247 	}
248 	if (!ubcountp || *ubcountp <= 0) {
249 		return EINVAL;
250 	}
251 	ubcount = *ubcountp; /* statstruct's */
252 	ubleft = ubcount * statstruct_size; /* bytes */
253 	*ubcountp = ubelem = 0;
254 	*done = 0;
255 	fmterror = 0;
256 	ubufp = ubuffer;
257 	nicluster = mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp) ?
258 		mp->m_sb.sb_inopblock :
259 		(XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog);
260 	nimask = ~(nicluster - 1);
261 	nbcluster = nicluster >> mp->m_sb.sb_inopblog;
262 	irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4);
263 	if (!irbuf)
264 		return ENOMEM;
265 
266 	nirbuf = irbsize / sizeof(*irbuf);
267 
268 	/*
269 	 * Loop over the allocation groups, starting from the last
270 	 * inode returned; 0 means start of the allocation group.
271 	 */
272 	rval = 0;
273 	while (XFS_BULKSTAT_UBLEFT(ubleft) && agno < mp->m_sb.sb_agcount) {
274 		cond_resched();
275 		bp = NULL;
276 		error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
277 		if (error) {
278 			/*
279 			 * Skip this allocation group and go to the next one.
280 			 */
281 			agno++;
282 			agino = 0;
283 			continue;
284 		}
285 		agi = XFS_BUF_TO_AGI(agbp);
286 		/*
287 		 * Allocate and initialize a btree cursor for ialloc btree.
288 		 */
289 		cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
290 		irbp = irbuf;
291 		irbufend = irbuf + nirbuf;
292 		end_of_ag = 0;
293 		/*
294 		 * If we're returning in the middle of an allocation group,
295 		 * we need to get the remainder of the chunk we're in.
296 		 */
297 		if (agino > 0) {
298 			xfs_inobt_rec_incore_t r;
299 
300 			/*
301 			 * Lookup the inode chunk that this inode lives in.
302 			 */
303 			error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE,
304 						 &tmp);
305 			if (!error &&	/* no I/O error */
306 			    tmp &&	/* lookup succeeded */
307 					/* got the record, should always work */
308 			    !(error = xfs_inobt_get_rec(cur, &r, &i)) &&
309 			    i == 1 &&
310 					/* this is the right chunk */
311 			    agino < r.ir_startino + XFS_INODES_PER_CHUNK &&
312 					/* lastino was not last in chunk */
313 			    (chunkidx = agino - r.ir_startino + 1) <
314 				    XFS_INODES_PER_CHUNK &&
315 					/* there are some left allocated */
316 			    xfs_inobt_maskn(chunkidx,
317 				    XFS_INODES_PER_CHUNK - chunkidx) &
318 				    ~r.ir_free) {
319 				/*
320 				 * Grab the chunk record.  Mark all the
321 				 * uninteresting inodes (because they're
322 				 * before our start point) free.
323 				 */
324 				for (i = 0; i < chunkidx; i++) {
325 					if (XFS_INOBT_MASK(i) & ~r.ir_free)
326 						r.ir_freecount++;
327 				}
328 				r.ir_free |= xfs_inobt_maskn(0, chunkidx);
329 				irbp->ir_startino = r.ir_startino;
330 				irbp->ir_freecount = r.ir_freecount;
331 				irbp->ir_free = r.ir_free;
332 				irbp++;
333 				agino = r.ir_startino + XFS_INODES_PER_CHUNK;
334 				icount = XFS_INODES_PER_CHUNK - r.ir_freecount;
335 			} else {
336 				/*
337 				 * If any of those tests failed, bump the
338 				 * inode number (just in case).
339 				 */
340 				agino++;
341 				icount = 0;
342 			}
343 			/*
344 			 * In any case, increment to the next record.
345 			 */
346 			if (!error)
347 				error = xfs_btree_increment(cur, 0, &tmp);
348 		} else {
349 			/*
350 			 * Start of ag.  Lookup the first inode chunk.
351 			 */
352 			error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &tmp);
353 			icount = 0;
354 		}
355 		/*
356 		 * Loop through inode btree records in this ag,
357 		 * until we run out of inodes or space in the buffer.
358 		 */
359 		while (irbp < irbufend && icount < ubcount) {
360 			xfs_inobt_rec_incore_t r;
361 
362 			/*
363 			 * Loop as long as we're unable to read the
364 			 * inode btree.
365 			 */
366 			while (error) {
367 				agino += XFS_INODES_PER_CHUNK;
368 				if (XFS_AGINO_TO_AGBNO(mp, agino) >=
369 						be32_to_cpu(agi->agi_length))
370 					break;
371 				error = xfs_inobt_lookup(cur, agino,
372 							 XFS_LOOKUP_GE, &tmp);
373 				cond_resched();
374 			}
375 			/*
376 			 * If ran off the end of the ag either with an error,
377 			 * or the normal way, set end and stop collecting.
378 			 */
379 			if (error) {
380 				end_of_ag = 1;
381 				break;
382 			}
383 
384 			error = xfs_inobt_get_rec(cur, &r, &i);
385 			if (error || i == 0) {
386 				end_of_ag = 1;
387 				break;
388 			}
389 
390 			/*
391 			 * If this chunk has any allocated inodes, save it.
392 			 * Also start read-ahead now for this chunk.
393 			 */
394 			if (r.ir_freecount < XFS_INODES_PER_CHUNK) {
395 				/*
396 				 * Loop over all clusters in the next chunk.
397 				 * Do a readahead if there are any allocated
398 				 * inodes in that cluster.
399 				 */
400 				agbno = XFS_AGINO_TO_AGBNO(mp, r.ir_startino);
401 				for (chunkidx = 0;
402 				     chunkidx < XFS_INODES_PER_CHUNK;
403 				     chunkidx += nicluster,
404 				     agbno += nbcluster) {
405 					if (xfs_inobt_maskn(chunkidx, nicluster)
406 							& ~r.ir_free)
407 						xfs_btree_reada_bufs(mp, agno,
408 							agbno, nbcluster);
409 				}
410 				irbp->ir_startino = r.ir_startino;
411 				irbp->ir_freecount = r.ir_freecount;
412 				irbp->ir_free = r.ir_free;
413 				irbp++;
414 				icount += XFS_INODES_PER_CHUNK - r.ir_freecount;
415 			}
416 			/*
417 			 * Set agino to after this chunk and bump the cursor.
418 			 */
419 			agino = r.ir_startino + XFS_INODES_PER_CHUNK;
420 			error = xfs_btree_increment(cur, 0, &tmp);
421 			cond_resched();
422 		}
423 		/*
424 		 * Drop the btree buffers and the agi buffer.
425 		 * We can't hold any of the locks these represent
426 		 * when calling iget.
427 		 */
428 		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
429 		xfs_buf_relse(agbp);
430 		/*
431 		 * Now format all the good inodes into the user's buffer.
432 		 */
433 		irbufend = irbp;
434 		for (irbp = irbuf;
435 		     irbp < irbufend && XFS_BULKSTAT_UBLEFT(ubleft); irbp++) {
436 			/*
437 			 * Now process this chunk of inodes.
438 			 */
439 			for (agino = irbp->ir_startino, chunkidx = clustidx = 0;
440 			     XFS_BULKSTAT_UBLEFT(ubleft) &&
441 				irbp->ir_freecount < XFS_INODES_PER_CHUNK;
442 			     chunkidx++, clustidx++, agino++) {
443 				ASSERT(chunkidx < XFS_INODES_PER_CHUNK);
444 				/*
445 				 * Recompute agbno if this is the
446 				 * first inode of the cluster.
447 				 *
448 				 * Careful with clustidx.   There can be
449 				 * multiple clusters per chunk, a single
450 				 * cluster per chunk or a cluster that has
451 				 * inodes represented from several different
452 				 * chunks (if blocksize is large).
453 				 *
454 				 * Because of this, the starting clustidx is
455 				 * initialized to zero in this loop but must
456 				 * later be reset after reading in the cluster
457 				 * buffer.
458 				 */
459 				if ((chunkidx & (nicluster - 1)) == 0) {
460 					agbno = XFS_AGINO_TO_AGBNO(mp,
461 							irbp->ir_startino) +
462 						((chunkidx & nimask) >>
463 						 mp->m_sb.sb_inopblog);
464 				}
465 				ino = XFS_AGINO_TO_INO(mp, agno, agino);
466 				bno = XFS_AGB_TO_DADDR(mp, agno, agbno);
467 				/*
468 				 * Skip if this inode is free.
469 				 */
470 				if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) {
471 					lastino = ino;
472 					continue;
473 				}
474 				/*
475 				 * Count used inodes as free so we can tell
476 				 * when the chunk is used up.
477 				 */
478 				irbp->ir_freecount++;
479 
480 				/*
481 				 * Get the inode and fill in a single buffer.
482 				 */
483 				ubused = statstruct_size;
484 				error = formatter(mp, ino, ubufp, ubleft,
485 						  &ubused, &fmterror);
486 				if (fmterror == BULKSTAT_RV_NOTHING) {
487 					if (error && error != ENOENT &&
488 						error != EINVAL) {
489 						ubleft = 0;
490 						rval = error;
491 						break;
492 					}
493 					lastino = ino;
494 					continue;
495 				}
496 				if (fmterror == BULKSTAT_RV_GIVEUP) {
497 					ubleft = 0;
498 					ASSERT(error);
499 					rval = error;
500 					break;
501 				}
502 				if (ubufp)
503 					ubufp += ubused;
504 				ubleft -= ubused;
505 				ubelem++;
506 				lastino = ino;
507 			}
508 
509 			cond_resched();
510 		}
511 
512 		if (bp)
513 			xfs_buf_relse(bp);
514 
515 		/*
516 		 * Set up for the next loop iteration.
517 		 */
518 		if (XFS_BULKSTAT_UBLEFT(ubleft)) {
519 			if (end_of_ag) {
520 				agno++;
521 				agino = 0;
522 			} else
523 				agino = XFS_INO_TO_AGINO(mp, lastino);
524 		} else
525 			break;
526 	}
527 	/*
528 	 * Done, we're either out of filesystem or space to put the data.
529 	 */
530 	kmem_free_large(irbuf);
531 	*ubcountp = ubelem;
532 	/*
533 	 * Found some inodes, return them now and return the error next time.
534 	 */
535 	if (ubelem)
536 		rval = 0;
537 	if (agno >= mp->m_sb.sb_agcount) {
538 		/*
539 		 * If we ran out of filesystem, mark lastino as off
540 		 * the end of the filesystem, so the next call
541 		 * will return immediately.
542 		 */
543 		*lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0);
544 		*done = 1;
545 	} else
546 		*lastinop = (xfs_ino_t)lastino;
547 
548 	return rval;
549 }
550 
551 /*
552  * Return stat information in bulk (by-inode) for the filesystem.
553  * Special case for non-sequential one inode bulkstat.
554  */
555 int					/* error status */
556 xfs_bulkstat_single(
557 	xfs_mount_t		*mp,	/* mount point for filesystem */
558 	xfs_ino_t		*lastinop, /* inode to return */
559 	char			__user *buffer, /* buffer with inode stats */
560 	int			*done)	/* 1 if there are more stats to get */
561 {
562 	int			count;	/* count value for bulkstat call */
563 	int			error;	/* return value */
564 	xfs_ino_t		ino;	/* filesystem inode number */
565 	int			res;	/* result from bs1 */
566 
567 	/*
568 	 * note that requesting valid inode numbers which are not allocated
569 	 * to inodes will most likely cause xfs_itobp to generate warning
570 	 * messages about bad magic numbers. This is ok. The fact that
571 	 * the inode isn't actually an inode is handled by the
572 	 * error check below. Done this way to make the usual case faster
573 	 * at the expense of the error case.
574 	 */
575 
576 	ino = (xfs_ino_t)*lastinop;
577 	error = xfs_bulkstat_one(mp, ino, buffer, sizeof(xfs_bstat_t), 0, &res);
578 	if (error) {
579 		/*
580 		 * Special case way failed, do it the "long" way
581 		 * to see if that works.
582 		 */
583 		(*lastinop)--;
584 		count = 1;
585 		if (xfs_bulkstat(mp, lastinop, &count, xfs_bulkstat_one,
586 				sizeof(xfs_bstat_t), buffer, done))
587 			return error;
588 		if (count == 0 || (xfs_ino_t)*lastinop != ino)
589 			return error == EFSCORRUPTED ?
590 				XFS_ERROR(EINVAL) : error;
591 		else
592 			return 0;
593 	}
594 	*done = 0;
595 	return 0;
596 }
597 
598 int
599 xfs_inumbers_fmt(
600 	void			__user *ubuffer, /* buffer to write to */
601 	const xfs_inogrp_t	*buffer,	/* buffer to read from */
602 	long			count,		/* # of elements to read */
603 	long			*written)	/* # of bytes written */
604 {
605 	if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
606 		return -EFAULT;
607 	*written = count * sizeof(*buffer);
608 	return 0;
609 }
610 
611 /*
612  * Return inode number table for the filesystem.
613  */
614 int					/* error status */
615 xfs_inumbers(
616 	xfs_mount_t	*mp,		/* mount point for filesystem */
617 	xfs_ino_t	*lastino,	/* last inode returned */
618 	int		*count,		/* size of buffer/count returned */
619 	void		__user *ubuffer,/* buffer with inode descriptions */
620 	inumbers_fmt_pf	formatter)
621 {
622 	xfs_buf_t	*agbp;
623 	xfs_agino_t	agino;
624 	xfs_agnumber_t	agno;
625 	int		bcount;
626 	xfs_inogrp_t	*buffer;
627 	int		bufidx;
628 	xfs_btree_cur_t	*cur;
629 	int		error;
630 	xfs_inobt_rec_incore_t r;
631 	int		i;
632 	xfs_ino_t	ino;
633 	int		left;
634 	int		tmp;
635 
636 	ino = (xfs_ino_t)*lastino;
637 	agno = XFS_INO_TO_AGNO(mp, ino);
638 	agino = XFS_INO_TO_AGINO(mp, ino);
639 	left = *count;
640 	*count = 0;
641 	bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
642 	buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
643 	error = bufidx = 0;
644 	cur = NULL;
645 	agbp = NULL;
646 	while (left > 0 && agno < mp->m_sb.sb_agcount) {
647 		if (agbp == NULL) {
648 			error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
649 			if (error) {
650 				/*
651 				 * If we can't read the AGI of this ag,
652 				 * then just skip to the next one.
653 				 */
654 				ASSERT(cur == NULL);
655 				agbp = NULL;
656 				agno++;
657 				agino = 0;
658 				continue;
659 			}
660 			cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
661 			error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
662 						 &tmp);
663 			if (error) {
664 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
665 				cur = NULL;
666 				xfs_buf_relse(agbp);
667 				agbp = NULL;
668 				/*
669 				 * Move up the last inode in the current
670 				 * chunk.  The lookup_ge will always get
671 				 * us the first inode in the next chunk.
672 				 */
673 				agino += XFS_INODES_PER_CHUNK - 1;
674 				continue;
675 			}
676 		}
677 		error = xfs_inobt_get_rec(cur, &r, &i);
678 		if (error || i == 0) {
679 			xfs_buf_relse(agbp);
680 			agbp = NULL;
681 			xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
682 			cur = NULL;
683 			agno++;
684 			agino = 0;
685 			continue;
686 		}
687 		agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
688 		buffer[bufidx].xi_startino =
689 			XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
690 		buffer[bufidx].xi_alloccount =
691 			XFS_INODES_PER_CHUNK - r.ir_freecount;
692 		buffer[bufidx].xi_allocmask = ~r.ir_free;
693 		bufidx++;
694 		left--;
695 		if (bufidx == bcount) {
696 			long written;
697 			if (formatter(ubuffer, buffer, bufidx, &written)) {
698 				error = XFS_ERROR(EFAULT);
699 				break;
700 			}
701 			ubuffer += written;
702 			*count += bufidx;
703 			bufidx = 0;
704 		}
705 		if (left) {
706 			error = xfs_btree_increment(cur, 0, &tmp);
707 			if (error) {
708 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
709 				cur = NULL;
710 				xfs_buf_relse(agbp);
711 				agbp = NULL;
712 				/*
713 				 * The agino value has already been bumped.
714 				 * Just try to skip up to it.
715 				 */
716 				agino += XFS_INODES_PER_CHUNK;
717 				continue;
718 			}
719 		}
720 	}
721 	if (!error) {
722 		if (bufidx) {
723 			long written;
724 			if (formatter(ubuffer, buffer, bufidx, &written))
725 				error = XFS_ERROR(EFAULT);
726 			else
727 				*count += bufidx;
728 		}
729 		*lastino = XFS_AGINO_TO_INO(mp, agno, agino);
730 	}
731 	kmem_free(buffer);
732 	if (cur)
733 		xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
734 					   XFS_BTREE_NOERROR));
735 	if (agbp)
736 		xfs_buf_relse(agbp);
737 	return error;
738 }
739