blob: cf6754a3c5b3e07a1e660c11ead329de21ef314c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Nathan Scott7b718762005-11-02 14:58:39 +11002 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
Nathan Scott7b718762005-11-02 14:58:39 +11005 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * published by the Free Software Foundation.
8 *
Nathan Scott7b718762005-11-02 14:58:39 +11009 * 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 *
Nathan Scott7b718762005-11-02 14:58:39 +110014 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include "xfs.h"
Nathan Scotta844f452005-11-02 14:38:42 +110019#include "xfs_fs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include "xfs_types.h"
Nathan Scotta844f452005-11-02 14:38:42 +110021#include "xfs_bit.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include "xfs_log.h"
Nathan Scotta844f452005-11-02 14:38:42 +110023#include "xfs_inum.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include "xfs_trans.h"
25#include "xfs_sb.h"
Nathan Scotta844f452005-11-02 14:38:42 +110026#include "xfs_ag.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include "xfs_dir2.h"
28#include "xfs_dmapi.h"
29#include "xfs_mount.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include "xfs_bmap_btree.h"
Nathan Scotta844f452005-11-02 14:38:42 +110031#include "xfs_alloc_btree.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include "xfs_ialloc_btree.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include "xfs_dir2_sf.h"
Nathan Scotta844f452005-11-02 14:38:42 +110034#include "xfs_attr_sf.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include "xfs_dinode.h"
36#include "xfs_inode.h"
37#include "xfs_ialloc.h"
38#include "xfs_itable.h"
39#include "xfs_error.h"
Nathan Scotta844f452005-11-02 14:38:42 +110040#include "xfs_btree.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Vlad Apostolov6f1f2162006-09-28 11:06:15 +100042int
43xfs_internal_inum(
44 xfs_mount_t *mp,
45 xfs_ino_t ino)
46{
47 return (ino == mp->m_sb.sb_rbmino || ino == mp->m_sb.sb_rsumino ||
Eric Sandeen62118702008-03-06 13:44:28 +110048 (xfs_sb_version_hasquota(&mp->m_sb) &&
Vlad Apostolov6f1f2162006-09-28 11:06:15 +100049 (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino)));
50}
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052STATIC int
53xfs_bulkstat_one_iget(
54 xfs_mount_t *mp, /* mount point for filesystem */
55 xfs_ino_t ino, /* inode number to get data for */
56 xfs_daddr_t bno, /* starting bno of inode cluster */
57 xfs_bstat_t *buf, /* return buffer */
58 int *stat) /* BULKSTAT_RV_... */
59{
Christoph Hellwig347d1c02007-08-28 13:57:51 +100060 xfs_icdinode_t *dic; /* dinode core info pointer */
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 xfs_inode_t *ip; /* incore inode pointer */
62 int error;
63
Nathan Scott745b1f472006-09-28 11:02:23 +100064 error = xfs_iget(mp, NULL, ino,
65 XFS_IGET_BULKSTAT, XFS_ILOCK_SHARED, &ip, bno);
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 if (error) {
67 *stat = BULKSTAT_RV_NOTHING;
68 return error;
69 }
70
71 ASSERT(ip != NULL);
72 ASSERT(ip->i_blkno != (xfs_daddr_t)0);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74 dic = &ip->i_d;
75
76 /* xfs_iget returns the following without needing
77 * further change.
78 */
79 buf->bs_nlink = dic->di_nlink;
80 buf->bs_projid = dic->di_projid;
81 buf->bs_ino = ino;
82 buf->bs_mode = dic->di_mode;
83 buf->bs_uid = dic->di_uid;
84 buf->bs_gid = dic->di_gid;
85 buf->bs_size = dic->di_size;
Christoph Hellwigdf80c932008-08-13 16:22:09 +100086 vn_atime_to_bstime(VFS_I(ip), &buf->bs_atime);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 buf->bs_mtime.tv_sec = dic->di_mtime.t_sec;
88 buf->bs_mtime.tv_nsec = dic->di_mtime.t_nsec;
89 buf->bs_ctime.tv_sec = dic->di_ctime.t_sec;
90 buf->bs_ctime.tv_nsec = dic->di_ctime.t_nsec;
91 buf->bs_xflags = xfs_ip2xflags(ip);
92 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
93 buf->bs_extents = dic->di_nextents;
94 buf->bs_gen = dic->di_gen;
95 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
96 buf->bs_dmevmask = dic->di_dmevmask;
97 buf->bs_dmstate = dic->di_dmstate;
98 buf->bs_aextents = dic->di_anextents;
99
100 switch (dic->di_format) {
101 case XFS_DINODE_FMT_DEV:
102 buf->bs_rdev = ip->i_df.if_u2.if_rdev;
103 buf->bs_blksize = BLKDEV_IOSIZE;
104 buf->bs_blocks = 0;
105 break;
106 case XFS_DINODE_FMT_LOCAL:
107 case XFS_DINODE_FMT_UUID:
108 buf->bs_rdev = 0;
109 buf->bs_blksize = mp->m_sb.sb_blocksize;
110 buf->bs_blocks = 0;
111 break;
112 case XFS_DINODE_FMT_EXTENTS:
113 case XFS_DINODE_FMT_BTREE:
114 buf->bs_rdev = 0;
115 buf->bs_blksize = mp->m_sb.sb_blocksize;
116 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
117 break;
118 }
119
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 xfs_iput(ip, XFS_ILOCK_SHARED);
121 return error;
122}
123
David Chinner7b073392008-04-10 12:24:04 +1000124STATIC void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125xfs_bulkstat_one_dinode(
126 xfs_mount_t *mp, /* mount point for filesystem */
127 xfs_ino_t ino, /* inode number to get data for */
128 xfs_dinode_t *dip, /* dinode inode pointer */
129 xfs_bstat_t *buf) /* return buffer */
130{
131 xfs_dinode_core_t *dic; /* dinode core info pointer */
132
133 dic = &dip->di_core;
134
135 /*
136 * The inode format changed when we moved the link count and
137 * made it 32 bits long. If this is an old format inode,
138 * convert it in memory to look like a new one. If it gets
139 * flushed to disk we will convert back before flushing or
140 * logging it. We zero out the new projid field and the old link
141 * count field. We'll handle clearing the pad field (the remains
142 * of the old uuid field) when we actually convert the inode to
143 * the new format. We don't change the version number so that we
144 * can distinguish this from a real new format inode.
145 */
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000146 if (dic->di_version == XFS_DINODE_VERSION_1) {
147 buf->bs_nlink = be16_to_cpu(dic->di_onlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 buf->bs_projid = 0;
149 } else {
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000150 buf->bs_nlink = be32_to_cpu(dic->di_nlink);
151 buf->bs_projid = be16_to_cpu(dic->di_projid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 }
153
154 buf->bs_ino = ino;
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000155 buf->bs_mode = be16_to_cpu(dic->di_mode);
156 buf->bs_uid = be32_to_cpu(dic->di_uid);
157 buf->bs_gid = be32_to_cpu(dic->di_gid);
158 buf->bs_size = be64_to_cpu(dic->di_size);
159 buf->bs_atime.tv_sec = be32_to_cpu(dic->di_atime.t_sec);
160 buf->bs_atime.tv_nsec = be32_to_cpu(dic->di_atime.t_nsec);
161 buf->bs_mtime.tv_sec = be32_to_cpu(dic->di_mtime.t_sec);
162 buf->bs_mtime.tv_nsec = be32_to_cpu(dic->di_mtime.t_nsec);
163 buf->bs_ctime.tv_sec = be32_to_cpu(dic->di_ctime.t_sec);
164 buf->bs_ctime.tv_nsec = be32_to_cpu(dic->di_ctime.t_nsec);
Christoph Hellwig45ba5982007-12-07 14:07:20 +1100165 buf->bs_xflags = xfs_dic2xflags(dip);
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000166 buf->bs_extsize = be32_to_cpu(dic->di_extsize) << mp->m_sb.sb_blocklog;
167 buf->bs_extents = be32_to_cpu(dic->di_nextents);
168 buf->bs_gen = be32_to_cpu(dic->di_gen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000170 buf->bs_dmevmask = be32_to_cpu(dic->di_dmevmask);
171 buf->bs_dmstate = be16_to_cpu(dic->di_dmstate);
172 buf->bs_aextents = be16_to_cpu(dic->di_anextents);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000174 switch (dic->di_format) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 case XFS_DINODE_FMT_DEV:
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000176 buf->bs_rdev = be32_to_cpu(dip->di_u.di_dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 buf->bs_blksize = BLKDEV_IOSIZE;
178 buf->bs_blocks = 0;
179 break;
180 case XFS_DINODE_FMT_LOCAL:
181 case XFS_DINODE_FMT_UUID:
182 buf->bs_rdev = 0;
183 buf->bs_blksize = mp->m_sb.sb_blocksize;
184 buf->bs_blocks = 0;
185 break;
186 case XFS_DINODE_FMT_EXTENTS:
187 case XFS_DINODE_FMT_BTREE:
188 buf->bs_rdev = 0;
189 buf->bs_blksize = mp->m_sb.sb_blocksize;
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000190 buf->bs_blocks = be64_to_cpu(dic->di_nblocks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 break;
192 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
Michal Marekfaa63e92007-07-11 11:10:19 +1000195STATIC int
196xfs_bulkstat_one_fmt(
197 void __user *ubuffer,
198 const xfs_bstat_t *buffer)
199{
200 if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
201 return -EFAULT;
202 return sizeof(*buffer);
203}
204
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205/*
206 * Return stat information for one inode.
207 * Return 0 if ok, else errno.
208 */
209int /* error status */
210xfs_bulkstat_one(
211 xfs_mount_t *mp, /* mount point for filesystem */
212 xfs_ino_t ino, /* inode number to get data for */
213 void __user *buffer, /* buffer to place output in */
214 int ubsize, /* size of buffer */
215 void *private_data, /* my private data */
216 xfs_daddr_t bno, /* starting bno of inode cluster */
217 int *ubused, /* bytes used by me */
218 void *dibuff, /* on-disk inode buffer */
219 int *stat) /* BULKSTAT_RV_... */
220{
221 xfs_bstat_t *buf; /* return buffer */
222 int error = 0; /* error value */
223 xfs_dinode_t *dip; /* dinode inode pointer */
Michal Marekfaa63e92007-07-11 11:10:19 +1000224 bulkstat_one_fmt_pf formatter = private_data ? : xfs_bulkstat_one_fmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
226 dip = (xfs_dinode_t *)dibuff;
Vlad Apostolov6f1f2162006-09-28 11:06:15 +1000227 *stat = BULKSTAT_RV_NOTHING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228
Vlad Apostolov6f1f2162006-09-28 11:06:15 +1000229 if (!buffer || xfs_internal_inum(mp, ino))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return XFS_ERROR(EINVAL);
Vlad Apostolov6f1f2162006-09-28 11:06:15 +1000231 if (ubsize < sizeof(*buf))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 return XFS_ERROR(ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
234 buf = kmem_alloc(sizeof(*buf), KM_SLEEP);
235
236 if (dip == NULL) {
237 /* We're not being passed a pointer to a dinode. This happens
238 * if BULKSTAT_FG_IGET is selected. Do the iget.
239 */
240 error = xfs_bulkstat_one_iget(mp, ino, bno, buf, stat);
241 if (error)
242 goto out_free;
243 } else {
244 xfs_bulkstat_one_dinode(mp, ino, dip, buf);
245 }
246
Michal Marekfaa63e92007-07-11 11:10:19 +1000247 error = formatter(buffer, buf);
248 if (error < 0) {
Vlad Apostolov6f1f2162006-09-28 11:06:15 +1000249 error = EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 goto out_free;
251 }
252
253 *stat = BULKSTAT_RV_DIDONE;
254 if (ubused)
Michal Marekfaa63e92007-07-11 11:10:19 +1000255 *ubused = error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
257 out_free:
Denys Vlasenkof0e2d932008-05-19 16:31:57 +1000258 kmem_free(buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 return error;
260}
261
262/*
Nathan Scott8b56f082006-09-28 11:01:46 +1000263 * Test to see whether we can use the ondisk inode directly, based
264 * on the given bulkstat flags, filling in dipp accordingly.
265 * Returns zero if the inode is dodgey.
266 */
267STATIC int
268xfs_bulkstat_use_dinode(
269 xfs_mount_t *mp,
270 int flags,
271 xfs_buf_t *bp,
272 int clustidx,
273 xfs_dinode_t **dipp)
274{
275 xfs_dinode_t *dip;
276 unsigned int aformat;
277
278 *dipp = NULL;
279 if (!bp || (flags & BULKSTAT_FG_IGET))
280 return 1;
281 dip = (xfs_dinode_t *)
282 xfs_buf_offset(bp, clustidx << mp->m_sb.sb_inodelog);
Vlad Apostolov859d7182007-10-11 17:44:18 +1000283 /*
Vlad Apostolovc319b582007-11-23 16:27:51 +1100284 * Check the buffer containing the on-disk inode for di_mode == 0.
Vlad Apostolov859d7182007-10-11 17:44:18 +1000285 * This is to prevent xfs_bulkstat from picking up just reclaimed
286 * inodes that have their in-core state initialized but not flushed
287 * to disk yet. This is a temporary hack that would require a proper
288 * fix in the future.
289 */
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000290 if (be16_to_cpu(dip->di_core.di_magic) != XFS_DINODE_MAGIC ||
Vlad Apostolov859d7182007-10-11 17:44:18 +1000291 !XFS_DINODE_GOOD_VERSION(dip->di_core.di_version) ||
Vlad Apostolovc319b582007-11-23 16:27:51 +1100292 !dip->di_core.di_mode)
Nathan Scott8b56f082006-09-28 11:01:46 +1000293 return 0;
294 if (flags & BULKSTAT_FG_QUICK) {
295 *dipp = dip;
296 return 1;
297 }
298 /* BULKSTAT_FG_INLINE: if attr fork is local, or not there, use it */
Christoph Hellwig347d1c02007-08-28 13:57:51 +1000299 aformat = dip->di_core.di_aformat;
Christoph Hellwig45ba5982007-12-07 14:07:20 +1100300 if ((XFS_DFORK_Q(dip) == 0) ||
Nathan Scott8b56f082006-09-28 11:01:46 +1000301 (aformat == XFS_DINODE_FMT_LOCAL) ||
302 (aformat == XFS_DINODE_FMT_EXTENTS && !dip->di_core.di_anextents)) {
303 *dipp = dip;
304 return 1;
305 }
306 return 1;
307}
308
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100309#define XFS_BULKSTAT_UBLEFT(ubleft) ((ubleft) >= statstruct_size)
310
Nathan Scott8b56f082006-09-28 11:01:46 +1000311/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 * Return stat information in bulk (by-inode) for the filesystem.
313 */
314int /* error status */
315xfs_bulkstat(
316 xfs_mount_t *mp, /* mount point for filesystem */
317 xfs_ino_t *lastinop, /* last inode returned */
318 int *ubcountp, /* size of buffer/count returned */
319 bulkstat_one_pf formatter, /* func that'd fill a single buf */
320 void *private_data,/* private data for formatter */
321 size_t statstruct_size, /* sizeof struct filling */
322 char __user *ubuffer, /* buffer with inode stats */
323 int flags, /* defined in xfs_itable.h */
Nathan Scottc41564b2006-03-29 08:55:14 +1000324 int *done) /* 1 if there are more stats to get */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325{
326 xfs_agblock_t agbno=0;/* allocation group block number */
327 xfs_buf_t *agbp; /* agi header buffer */
328 xfs_agi_t *agi; /* agi header data */
329 xfs_agino_t agino; /* inode # in allocation group */
330 xfs_agnumber_t agno; /* allocation group number */
331 xfs_daddr_t bno; /* inode cluster start daddr */
332 int chunkidx; /* current index into inode chunk */
333 int clustidx; /* current index into inode cluster */
334 xfs_btree_cur_t *cur; /* btree cursor for ialloc btree */
335 int end_of_ag; /* set if we've seen the ag end */
336 int error; /* error code */
337 int fmterror;/* bulkstat formatter result */
338 __int32_t gcnt; /* current btree rec's count */
339 xfs_inofree_t gfree; /* current btree rec's free mask */
340 xfs_agino_t gino; /* current btree rec's start inode */
341 int i; /* loop index */
342 int icount; /* count of inodes good in irbuf */
Nathan Scott215101c2006-09-28 11:04:43 +1000343 size_t irbsize; /* size of irec buffer in bytes */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 xfs_ino_t ino; /* inode number (filesystem) */
Nathan Scott262750932006-09-28 11:02:03 +1000345 xfs_inobt_rec_incore_t *irbp; /* current irec buffer pointer */
346 xfs_inobt_rec_incore_t *irbuf; /* start of irec buffer */
347 xfs_inobt_rec_incore_t *irbufend; /* end of good irec buffer entries */
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100348 xfs_ino_t lastino; /* last inode number returned */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 int nbcluster; /* # of blocks in a cluster */
350 int nicluster; /* # of inodes in a cluster */
351 int nimask; /* mask for inode clusters */
352 int nirbuf; /* size of irbuf */
353 int rval; /* return value error code */
354 int tmp; /* result value from btree calls */
355 int ubcount; /* size of user's buffer */
356 int ubleft; /* bytes left in user's buffer */
357 char __user *ubufp; /* pointer into user's buffer */
358 int ubelem; /* spaces used in user's buffer */
359 int ubused; /* bytes used by formatter */
360 xfs_buf_t *bp; /* ptr to on-disk inode cluster buf */
361 xfs_dinode_t *dip; /* ptr into bp for specific inode */
362 xfs_inode_t *ip; /* ptr to in-core inode struct */
363
364 /*
365 * Get the last inode value, see if there's nothing to do.
366 */
367 ino = (xfs_ino_t)*lastinop;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100368 lastino = ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 dip = NULL;
370 agno = XFS_INO_TO_AGNO(mp, ino);
371 agino = XFS_INO_TO_AGINO(mp, ino);
372 if (agno >= mp->m_sb.sb_agcount ||
373 ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
374 *done = 1;
375 *ubcountp = 0;
376 return 0;
377 }
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100378 if (!ubcountp || *ubcountp <= 0) {
379 return EINVAL;
380 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 ubcount = *ubcountp; /* statstruct's */
382 ubleft = ubcount * statstruct_size; /* bytes */
383 *ubcountp = ubelem = 0;
384 *done = 0;
385 fmterror = 0;
386 ubufp = ubuffer;
387 nicluster = mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp) ?
388 mp->m_sb.sb_inopblock :
389 (XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog);
390 nimask = ~(nicluster - 1);
391 nbcluster = nicluster >> mp->m_sb.sb_inopblog;
Tim Shimmine6a4b372007-11-23 16:30:42 +1100392 irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4,
Nathan Scott77e46352006-09-28 11:03:27 +1000393 KM_SLEEP | KM_MAYFAIL | KM_LARGE);
Nathan Scottbb3c7d22006-09-28 11:02:09 +1000394 nirbuf = irbsize / sizeof(*irbuf);
395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 /*
397 * Loop over the allocation groups, starting from the last
398 * inode returned; 0 means start of the allocation group.
399 */
400 rval = 0;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100401 while (XFS_BULKSTAT_UBLEFT(ubleft) && agno < mp->m_sb.sb_agcount) {
402 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 bp = NULL;
404 down_read(&mp->m_peraglock);
405 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
406 up_read(&mp->m_peraglock);
407 if (error) {
408 /*
409 * Skip this allocation group and go to the next one.
410 */
411 agno++;
412 agino = 0;
413 continue;
414 }
415 agi = XFS_BUF_TO_AGI(agbp);
416 /*
417 * Allocate and initialize a btree cursor for ialloc btree.
418 */
419 cur = xfs_btree_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_INO,
Nathan Scott262750932006-09-28 11:02:03 +1000420 (xfs_inode_t *)0, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 irbp = irbuf;
422 irbufend = irbuf + nirbuf;
423 end_of_ag = 0;
424 /*
425 * If we're returning in the middle of an allocation group,
426 * we need to get the remainder of the chunk we're in.
427 */
428 if (agino > 0) {
429 /*
430 * Lookup the inode chunk that this inode lives in.
431 */
432 error = xfs_inobt_lookup_le(cur, agino, 0, 0, &tmp);
433 if (!error && /* no I/O error */
434 tmp && /* lookup succeeded */
435 /* got the record, should always work */
436 !(error = xfs_inobt_get_rec(cur, &gino, &gcnt,
437 &gfree, &i)) &&
438 i == 1 &&
439 /* this is the right chunk */
440 agino < gino + XFS_INODES_PER_CHUNK &&
441 /* lastino was not last in chunk */
442 (chunkidx = agino - gino + 1) <
443 XFS_INODES_PER_CHUNK &&
444 /* there are some left allocated */
445 XFS_INOBT_MASKN(chunkidx,
446 XFS_INODES_PER_CHUNK - chunkidx) & ~gfree) {
447 /*
448 * Grab the chunk record. Mark all the
449 * uninteresting inodes (because they're
450 * before our start point) free.
451 */
452 for (i = 0; i < chunkidx; i++) {
453 if (XFS_INOBT_MASK(i) & ~gfree)
454 gcnt++;
455 }
456 gfree |= XFS_INOBT_MASKN(0, chunkidx);
Nathan Scott262750932006-09-28 11:02:03 +1000457 irbp->ir_startino = gino;
458 irbp->ir_freecount = gcnt;
459 irbp->ir_free = gfree;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 irbp++;
461 agino = gino + XFS_INODES_PER_CHUNK;
462 icount = XFS_INODES_PER_CHUNK - gcnt;
463 } else {
464 /*
465 * If any of those tests failed, bump the
466 * inode number (just in case).
467 */
468 agino++;
469 icount = 0;
470 }
471 /*
472 * In any case, increment to the next record.
473 */
474 if (!error)
475 error = xfs_inobt_increment(cur, 0, &tmp);
476 } else {
477 /*
478 * Start of ag. Lookup the first inode chunk.
479 */
480 error = xfs_inobt_lookup_ge(cur, 0, 0, 0, &tmp);
481 icount = 0;
482 }
483 /*
484 * Loop through inode btree records in this ag,
485 * until we run out of inodes or space in the buffer.
486 */
487 while (irbp < irbufend && icount < ubcount) {
488 /*
489 * Loop as long as we're unable to read the
490 * inode btree.
491 */
492 while (error) {
493 agino += XFS_INODES_PER_CHUNK;
494 if (XFS_AGINO_TO_AGBNO(mp, agino) >=
Christoph Hellwig16259e72005-11-02 15:11:25 +1100495 be32_to_cpu(agi->agi_length))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 break;
497 error = xfs_inobt_lookup_ge(cur, agino, 0, 0,
498 &tmp);
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100499 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 }
501 /*
502 * If ran off the end of the ag either with an error,
503 * or the normal way, set end and stop collecting.
504 */
505 if (error ||
506 (error = xfs_inobt_get_rec(cur, &gino, &gcnt,
507 &gfree, &i)) ||
508 i == 0) {
509 end_of_ag = 1;
510 break;
511 }
512 /*
513 * If this chunk has any allocated inodes, save it.
Nathan Scott262750932006-09-28 11:02:03 +1000514 * Also start read-ahead now for this chunk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 */
516 if (gcnt < XFS_INODES_PER_CHUNK) {
Nathan Scott262750932006-09-28 11:02:03 +1000517 /*
518 * Loop over all clusters in the next chunk.
519 * Do a readahead if there are any allocated
520 * inodes in that cluster.
521 */
522 for (agbno = XFS_AGINO_TO_AGBNO(mp, gino),
523 chunkidx = 0;
524 chunkidx < XFS_INODES_PER_CHUNK;
525 chunkidx += nicluster,
526 agbno += nbcluster) {
527 if (XFS_INOBT_MASKN(chunkidx,
528 nicluster) & ~gfree)
529 xfs_btree_reada_bufs(mp, agno,
530 agbno, nbcluster);
531 }
532 irbp->ir_startino = gino;
533 irbp->ir_freecount = gcnt;
534 irbp->ir_free = gfree;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 irbp++;
536 icount += XFS_INODES_PER_CHUNK - gcnt;
537 }
538 /*
539 * Set agino to after this chunk and bump the cursor.
540 */
541 agino = gino + XFS_INODES_PER_CHUNK;
542 error = xfs_inobt_increment(cur, 0, &tmp);
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100543 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 }
545 /*
546 * Drop the btree buffers and the agi buffer.
547 * We can't hold any of the locks these represent
548 * when calling iget.
549 */
550 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
551 xfs_buf_relse(agbp);
552 /*
553 * Now format all the good inodes into the user's buffer.
554 */
555 irbufend = irbp;
556 for (irbp = irbuf;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100557 irbp < irbufend && XFS_BULKSTAT_UBLEFT(ubleft); irbp++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 * Now process this chunk of inodes.
560 */
Nathan Scott262750932006-09-28 11:02:03 +1000561 for (agino = irbp->ir_startino, chunkidx = clustidx = 0;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100562 XFS_BULKSTAT_UBLEFT(ubleft) &&
Nathan Scott262750932006-09-28 11:02:03 +1000563 irbp->ir_freecount < XFS_INODES_PER_CHUNK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 chunkidx++, clustidx++, agino++) {
565 ASSERT(chunkidx < XFS_INODES_PER_CHUNK);
566 /*
567 * Recompute agbno if this is the
568 * first inode of the cluster.
569 *
570 * Careful with clustidx. There can be
571 * multple clusters per chunk, a single
572 * cluster per chunk or a cluster that has
573 * inodes represented from several different
574 * chunks (if blocksize is large).
575 *
576 * Because of this, the starting clustidx is
577 * initialized to zero in this loop but must
578 * later be reset after reading in the cluster
579 * buffer.
580 */
581 if ((chunkidx & (nicluster - 1)) == 0) {
582 agbno = XFS_AGINO_TO_AGBNO(mp,
Nathan Scott262750932006-09-28 11:02:03 +1000583 irbp->ir_startino) +
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 ((chunkidx & nimask) >>
585 mp->m_sb.sb_inopblog);
586
Nathan Scott8b56f082006-09-28 11:01:46 +1000587 if (flags & (BULKSTAT_FG_QUICK |
588 BULKSTAT_FG_INLINE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 ino = XFS_AGINO_TO_INO(mp, agno,
590 agino);
591 bno = XFS_AGB_TO_DADDR(mp, agno,
592 agbno);
593
594 /*
595 * Get the inode cluster buffer
596 */
597 ASSERT(xfs_inode_zone != NULL);
598 ip = kmem_zone_zalloc(xfs_inode_zone,
599 KM_SLEEP);
600 ip->i_ino = ino;
601 ip->i_mount = mp;
David Chinnerf273ab82006-09-28 11:06:03 +1000602 spin_lock_init(&ip->i_flags_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 if (bp)
604 xfs_buf_relse(bp);
605 error = xfs_itobp(mp, NULL, ip,
Nathan Scottb12dd342006-03-17 17:26:04 +1100606 &dip, &bp, bno,
David Chinnera3f74ff2008-03-06 13:43:42 +1100607 XFS_IMAP_BULKSTAT,
608 XFS_BUF_LOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 if (!error)
610 clustidx = ip->i_boffset / mp->m_sb.sb_inodesize;
611 kmem_zone_free(xfs_inode_zone, ip);
612 if (XFS_TEST_ERROR(error != 0,
613 mp, XFS_ERRTAG_BULKSTAT_READ_CHUNK,
614 XFS_RANDOM_BULKSTAT_READ_CHUNK)) {
615 bp = NULL;
Nathan Scottb12dd342006-03-17 17:26:04 +1100616 ubleft = 0;
617 rval = error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 break;
619 }
620 }
621 }
Lachlan McIlroyc2cba572007-10-12 11:12:20 +1000622 ino = XFS_AGINO_TO_INO(mp, agno, agino);
623 bno = XFS_AGB_TO_DADDR(mp, agno, agbno);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 /*
625 * Skip if this inode is free.
626 */
Lachlan McIlroyc2cba572007-10-12 11:12:20 +1000627 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) {
628 lastino = ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 continue;
Lachlan McIlroyc2cba572007-10-12 11:12:20 +1000630 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 /*
632 * Count used inodes as free so we can tell
633 * when the chunk is used up.
634 */
Nathan Scott262750932006-09-28 11:02:03 +1000635 irbp->ir_freecount++;
Nathan Scott8b56f082006-09-28 11:01:46 +1000636 if (!xfs_bulkstat_use_dinode(mp, flags, bp,
Lachlan McIlroyc2cba572007-10-12 11:12:20 +1000637 clustidx, &dip)) {
638 lastino = ino;
Nathan Scott8b56f082006-09-28 11:01:46 +1000639 continue;
Lachlan McIlroyc2cba572007-10-12 11:12:20 +1000640 }
Nathan Scott8b56f082006-09-28 11:01:46 +1000641 /*
642 * If we need to do an iget, cannot hold bp.
643 * Drop it, until starting the next cluster.
644 */
645 if ((flags & BULKSTAT_FG_INLINE) && !dip) {
646 if (bp)
647 xfs_buf_relse(bp);
648 bp = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 }
650
651 /*
652 * Get the inode and fill in a single buffer.
653 * BULKSTAT_FG_QUICK uses dip to fill it in.
654 * BULKSTAT_FG_IGET uses igets.
Nathan Scott8b56f082006-09-28 11:01:46 +1000655 * BULKSTAT_FG_INLINE uses dip if we have an
656 * inline attr fork, else igets.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 * See: xfs_bulkstat_one & xfs_dm_bulkstat_one.
658 * This is also used to count inodes/blks, etc
659 * in xfs_qm_quotacheck.
660 */
661 ubused = statstruct_size;
662 error = formatter(mp, ino, ubufp,
663 ubleft, private_data,
664 bno, &ubused, dip, &fmterror);
665 if (fmterror == BULKSTAT_RV_NOTHING) {
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100666 if (error && error != ENOENT &&
667 error != EINVAL) {
Vlad Apostolove132f542006-09-28 11:04:31 +1000668 ubleft = 0;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100669 rval = error;
670 break;
671 }
672 lastino = ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 continue;
674 }
675 if (fmterror == BULKSTAT_RV_GIVEUP) {
676 ubleft = 0;
677 ASSERT(error);
678 rval = error;
679 break;
680 }
681 if (ubufp)
682 ubufp += ubused;
683 ubleft -= ubused;
684 ubelem++;
685 lastino = ino;
686 }
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100687
688 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 }
690
691 if (bp)
692 xfs_buf_relse(bp);
693
694 /*
695 * Set up for the next loop iteration.
696 */
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100697 if (XFS_BULKSTAT_UBLEFT(ubleft)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 if (end_of_ag) {
699 agno++;
700 agino = 0;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100701 } else
702 agino = XFS_INO_TO_AGINO(mp, lastino);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 } else
704 break;
705 }
706 /*
707 * Done, we're either out of filesystem or space to put the data.
708 */
Denys Vlasenkof0e2d932008-05-19 16:31:57 +1000709 kmem_free(irbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 *ubcountp = ubelem;
Lachlan McIlroycd57e592007-11-23 16:30:32 +1100711 /*
712 * Found some inodes, return them now and return the error next time.
713 */
714 if (ubelem)
715 rval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 if (agno >= mp->m_sb.sb_agcount) {
717 /*
718 * If we ran out of filesystem, mark lastino as off
719 * the end of the filesystem, so the next call
720 * will return immediately.
721 */
722 *lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0);
723 *done = 1;
724 } else
725 *lastinop = (xfs_ino_t)lastino;
726
727 return rval;
728}
729
730/*
731 * Return stat information in bulk (by-inode) for the filesystem.
732 * Special case for non-sequential one inode bulkstat.
733 */
734int /* error status */
735xfs_bulkstat_single(
736 xfs_mount_t *mp, /* mount point for filesystem */
737 xfs_ino_t *lastinop, /* inode to return */
738 char __user *buffer, /* buffer with inode stats */
Nathan Scottc41564b2006-03-29 08:55:14 +1000739 int *done) /* 1 if there are more stats to get */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740{
741 int count; /* count value for bulkstat call */
742 int error; /* return value */
743 xfs_ino_t ino; /* filesystem inode number */
744 int res; /* result from bs1 */
745
746 /*
747 * note that requesting valid inode numbers which are not allocated
748 * to inodes will most likely cause xfs_itobp to generate warning
749 * messages about bad magic numbers. This is ok. The fact that
750 * the inode isn't actually an inode is handled by the
751 * error check below. Done this way to make the usual case faster
752 * at the expense of the error case.
753 */
754
755 ino = (xfs_ino_t)*lastinop;
756 error = xfs_bulkstat_one(mp, ino, buffer, sizeof(xfs_bstat_t),
757 NULL, 0, NULL, NULL, &res);
758 if (error) {
759 /*
760 * Special case way failed, do it the "long" way
761 * to see if that works.
762 */
763 (*lastinop)--;
764 count = 1;
765 if (xfs_bulkstat(mp, lastinop, &count, xfs_bulkstat_one,
766 NULL, sizeof(xfs_bstat_t), buffer,
767 BULKSTAT_FG_IGET, done))
768 return error;
769 if (count == 0 || (xfs_ino_t)*lastinop != ino)
770 return error == EFSCORRUPTED ?
771 XFS_ERROR(EINVAL) : error;
772 else
773 return 0;
774 }
775 *done = 0;
776 return 0;
777}
778
Michal Marekfaa63e92007-07-11 11:10:19 +1000779int
780xfs_inumbers_fmt(
781 void __user *ubuffer, /* buffer to write to */
782 const xfs_inogrp_t *buffer, /* buffer to read from */
783 long count, /* # of elements to read */
784 long *written) /* # of bytes written */
785{
786 if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
787 return -EFAULT;
788 *written = count * sizeof(*buffer);
789 return 0;
790}
791
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792/*
793 * Return inode number table for the filesystem.
794 */
795int /* error status */
796xfs_inumbers(
797 xfs_mount_t *mp, /* mount point for filesystem */
798 xfs_ino_t *lastino, /* last inode returned */
799 int *count, /* size of buffer/count returned */
Michal Marekfaa63e92007-07-11 11:10:19 +1000800 void __user *ubuffer,/* buffer with inode descriptions */
801 inumbers_fmt_pf formatter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802{
803 xfs_buf_t *agbp;
804 xfs_agino_t agino;
805 xfs_agnumber_t agno;
806 int bcount;
807 xfs_inogrp_t *buffer;
808 int bufidx;
809 xfs_btree_cur_t *cur;
810 int error;
811 __int32_t gcnt;
812 xfs_inofree_t gfree;
813 xfs_agino_t gino;
814 int i;
815 xfs_ino_t ino;
816 int left;
817 int tmp;
818
819 ino = (xfs_ino_t)*lastino;
820 agno = XFS_INO_TO_AGNO(mp, ino);
821 agino = XFS_INO_TO_AGINO(mp, ino);
822 left = *count;
823 *count = 0;
Tim Shimmine6a4b372007-11-23 16:30:42 +1100824 bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
826 error = bufidx = 0;
827 cur = NULL;
828 agbp = NULL;
829 while (left > 0 && agno < mp->m_sb.sb_agcount) {
830 if (agbp == NULL) {
831 down_read(&mp->m_peraglock);
832 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
833 up_read(&mp->m_peraglock);
834 if (error) {
835 /*
836 * If we can't read the AGI of this ag,
837 * then just skip to the next one.
838 */
839 ASSERT(cur == NULL);
840 agbp = NULL;
841 agno++;
842 agino = 0;
843 continue;
844 }
845 cur = xfs_btree_init_cursor(mp, NULL, agbp, agno,
846 XFS_BTNUM_INO, (xfs_inode_t *)0, 0);
847 error = xfs_inobt_lookup_ge(cur, agino, 0, 0, &tmp);
848 if (error) {
849 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
850 cur = NULL;
851 xfs_buf_relse(agbp);
852 agbp = NULL;
853 /*
Michael Opdenacker59c51592007-05-09 08:57:56 +0200854 * Move up the last inode in the current
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 * chunk. The lookup_ge will always get
856 * us the first inode in the next chunk.
857 */
858 agino += XFS_INODES_PER_CHUNK - 1;
859 continue;
860 }
861 }
862 if ((error = xfs_inobt_get_rec(cur, &gino, &gcnt, &gfree,
863 &i)) ||
864 i == 0) {
865 xfs_buf_relse(agbp);
866 agbp = NULL;
867 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
868 cur = NULL;
869 agno++;
870 agino = 0;
871 continue;
872 }
873 agino = gino + XFS_INODES_PER_CHUNK - 1;
874 buffer[bufidx].xi_startino = XFS_AGINO_TO_INO(mp, agno, gino);
875 buffer[bufidx].xi_alloccount = XFS_INODES_PER_CHUNK - gcnt;
876 buffer[bufidx].xi_allocmask = ~gfree;
877 bufidx++;
878 left--;
879 if (bufidx == bcount) {
Michal Marekfaa63e92007-07-11 11:10:19 +1000880 long written;
881 if (formatter(ubuffer, buffer, bufidx, &written)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 error = XFS_ERROR(EFAULT);
883 break;
884 }
Michal Marekfaa63e92007-07-11 11:10:19 +1000885 ubuffer += written;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 *count += bufidx;
887 bufidx = 0;
888 }
889 if (left) {
890 error = xfs_inobt_increment(cur, 0, &tmp);
891 if (error) {
892 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
893 cur = NULL;
894 xfs_buf_relse(agbp);
895 agbp = NULL;
896 /*
897 * The agino value has already been bumped.
898 * Just try to skip up to it.
899 */
900 agino += XFS_INODES_PER_CHUNK;
901 continue;
902 }
903 }
904 }
905 if (!error) {
906 if (bufidx) {
Michal Marekfaa63e92007-07-11 11:10:19 +1000907 long written;
908 if (formatter(ubuffer, buffer, bufidx, &written))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 error = XFS_ERROR(EFAULT);
910 else
911 *count += bufidx;
912 }
913 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
914 }
Denys Vlasenkof0e2d932008-05-19 16:31:57 +1000915 kmem_free(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 if (cur)
917 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
918 XFS_BTREE_NOERROR));
919 if (agbp)
920 xfs_buf_relse(agbp);
921 return error;
922}