fs/xfs/xfs_itable.c

   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_dir2.h"
  28 #include "xfs_dmapi.h"
  29 #include "xfs_mount.h"
  30 #include "xfs_bmap_btree.h"
  31 #include "xfs_alloc_btree.h"
  32 #include "xfs_ialloc_btree.h"
  33 #include "xfs_dir2_sf.h"
  34 #include "xfs_attr_sf.h"
  35 #include "xfs_dinode.h"
  36 #include "xfs_inode.h"
  37 #include "xfs_ialloc.h"
  38 #include "xfs_itable.h"
  39 #include "xfs_error.h"
  40 #include "xfs_btree.h"
  41
  42 STATIC int
  43 xfs_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 ||
  48                 (xfs_sb_version_hasquota(&mp->m_sb) &&
  49                  (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino)));
  50 }
  51
  52 STATIC int
  53 xfs_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 {
  60         xfs_icdinode_t  *dic;   /* dinode core info pointer */
  61         xfs_inode_t     *ip;            /* incore inode pointer */
  62         struct inode    *inode;
  63         int             error;
  64
  65         error = xfs_iget(mp, NULL, ino,
  66                          XFS_IGET_BULKSTAT, XFS_ILOCK_SHARED, &ip, bno);
  67         if (error) {
  68                 *stat = BULKSTAT_RV_NOTHING;
  69                 return error;
  70         }
  71
  72         ASSERT(ip != NULL);
  73         ASSERT(ip->i_imap.im_blkno != 0);
  74
  75         dic = &ip->i_d;
  76         inode = VFS_I(ip);
  77
  78         /* xfs_iget returns the following without needing
  79          * further change.
  80          */
  81         buf->bs_nlink = dic->di_nlink;
  82         buf->bs_projid = dic->di_projid;
  83         buf->bs_ino = ino;
  84         buf->bs_mode = dic->di_mode;
  85         buf->bs_uid = dic->di_uid;
  86         buf->bs_gid = dic->di_gid;
  87         buf->bs_size = dic->di_size;
  88
  89         /*
  90          * We need to read the timestamps from the Linux inode because
  91          * the VFS keeps writing directly into the inode structure instead
  92          * of telling us about the updates.
  93          */
  94         buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
  95         buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
  96         buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
  97         buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
  98         buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
  99         buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
 100
 101         buf->bs_xflags = xfs_ip2xflags(ip);
 102         buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
 103         buf->bs_extents = dic->di_nextents;
 104         buf->bs_gen = dic->di_gen;
 105         memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
 106         buf->bs_dmevmask = dic->di_dmevmask;
 107         buf->bs_dmstate = dic->di_dmstate;
 108         buf->bs_aextents = dic->di_anextents;
 109         buf->bs_forkoff = XFS_IFORK_BOFF(ip);
 110
 111         switch (dic->di_format) {
 112         case XFS_DINODE_FMT_DEV:
 113                 buf->bs_rdev = ip->i_df.if_u2.if_rdev;
 114                 buf->bs_blksize = BLKDEV_IOSIZE;
 115                 buf->bs_blocks = 0;
 116                 break;
 117         case XFS_DINODE_FMT_LOCAL:
 118         case XFS_DINODE_FMT_UUID:
 119                 buf->bs_rdev = 0;
 120                 buf->bs_blksize = mp->m_sb.sb_blocksize;
 121                 buf->bs_blocks = 0;
 122                 break;
 123         case XFS_DINODE_FMT_EXTENTS:
 124         case XFS_DINODE_FMT_BTREE:
 125                 buf->bs_rdev = 0;
 126                 buf->bs_blksize = mp->m_sb.sb_blocksize;
 127                 buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
 128                 break;
 129         }
 130
 131         xfs_iput(ip, XFS_ILOCK_SHARED);
 132         return error;
 133 }
 134
 135 STATIC void
 136 xfs_bulkstat_one_dinode(
 137         xfs_mount_t     *mp,            /* mount point for filesystem */
 138         xfs_ino_t       ino,            /* inode number to get data for */
 139         xfs_dinode_t    *dic,           /* dinode inode pointer */
 140         xfs_bstat_t     *buf)           /* return buffer */
 141 {
 142         /*
 143          * The inode format changed when we moved the link count and
 144          * made it 32 bits long.  If this is an old format inode,
 145          * convert it in memory to look like a new one.  If it gets
 146          * flushed to disk we will convert back before flushing or
 147          * logging it.  We zero out the new projid field and the old link
 148          * count field.  We'll handle clearing the pad field (the remains
 149          * of the old uuid field) when we actually convert the inode to
 150          * the new format. We don't change the version number so that we
 151          * can distinguish this from a real new format inode.
 152          */
 153         if (dic->di_version == 1) {
 154                 buf->bs_nlink = be16_to_cpu(dic->di_onlink);
 155                 buf->bs_projid = 0;
 156         } else {
 157                 buf->bs_nlink = be32_to_cpu(dic->di_nlink);
 158                 buf->bs_projid = be16_to_cpu(dic->di_projid);
 159         }
 160
 161         buf->bs_ino = ino;
 162         buf->bs_mode = be16_to_cpu(dic->di_mode);
 163         buf->bs_uid = be32_to_cpu(dic->di_uid);
 164         buf->bs_gid = be32_to_cpu(dic->di_gid);
 165         buf->bs_size = be64_to_cpu(dic->di_size);
 166         buf->bs_atime.tv_sec = be32_to_cpu(dic->di_atime.t_sec);
 167         buf->bs_atime.tv_nsec = be32_to_cpu(dic->di_atime.t_nsec);
 168         buf->bs_mtime.tv_sec = be32_to_cpu(dic->di_mtime.t_sec);
 169         buf->bs_mtime.tv_nsec = be32_to_cpu(dic->di_mtime.t_nsec);
 170         buf->bs_ctime.tv_sec = be32_to_cpu(dic->di_ctime.t_sec);
 171         buf->bs_ctime.tv_nsec = be32_to_cpu(dic->di_ctime.t_nsec);
 172         buf->bs_xflags = xfs_dic2xflags(dic);
 173         buf->bs_extsize = be32_to_cpu(dic->di_extsize) << mp->m_sb.sb_blocklog;
 174         buf->bs_extents = be32_to_cpu(dic->di_nextents);
 175         buf->bs_gen = be32_to_cpu(dic->di_gen);
 176         memset(buf->bs_pad, 0, sizeof(buf->bs_pad));
 177         buf->bs_dmevmask = be32_to_cpu(dic->di_dmevmask);
 178         buf->bs_dmstate = be16_to_cpu(dic->di_dmstate);
 179         buf->bs_aextents = be16_to_cpu(dic->di_anextents);
 180         buf->bs_forkoff = XFS_DFORK_BOFF(dic);
 181
 182         switch (dic->di_format) {
 183         case XFS_DINODE_FMT_DEV:
 184                 buf->bs_rdev = xfs_dinode_get_rdev(dic);
 185                 buf->bs_blksize = BLKDEV_IOSIZE;
 186                 buf->bs_blocks = 0;
 187                 break;
 188         case XFS_DINODE_FMT_LOCAL:
 189         case XFS_DINODE_FMT_UUID:
 190                 buf->bs_rdev = 0;
 191                 buf->bs_blksize = mp->m_sb.sb_blocksize;
 192                 buf->bs_blocks = 0;
 193                 break;
 194         case XFS_DINODE_FMT_EXTENTS:
 195         case XFS_DINODE_FMT_BTREE:
 196                 buf->bs_rdev = 0;
 197                 buf->bs_blksize = mp->m_sb.sb_blocksize;
 198                 buf->bs_blocks = be64_to_cpu(dic->di_nblocks);
 199                 break;
 200         }
 201 }
 202
 203 /* Return 0 on success or positive error */
 204 STATIC int
 205 xfs_bulkstat_one_fmt(
 206         void                    __user *ubuffer,
 207         int                     ubsize,
 208         int                     *ubused,
 209         const xfs_bstat_t       *buffer)
 210 {
 211         if (ubsize < sizeof(*buffer))
 212                 return XFS_ERROR(ENOMEM);
 213         if (copy_to_user(ubuffer, buffer, sizeof(*buffer)))
 214                 return XFS_ERROR(EFAULT);
 215         if (ubused)
 216                 *ubused = sizeof(*buffer);
 217         return 0;
 218 }
 219
 220 /*
 221  * Return stat information for one inode.
 222  * Return 0 if ok, else errno.
 223  */
 224 int                                     /* error status */
 225 xfs_bulkstat_one_int(
 226         xfs_mount_t     *mp,            /* mount point for filesystem */
 227         xfs_ino_t       ino,            /* inode number to get data for */
 228         void            __user *buffer, /* buffer to place output in */
 229         int             ubsize,         /* size of buffer */
 230         bulkstat_one_fmt_pf formatter,  /* formatter, copy to user */
 231         xfs_daddr_t     bno,            /* starting bno of inode cluster */
 232         int             *ubused,        /* bytes used by me */
 233         void            *dibuff,        /* on-disk inode buffer */
 234         int             *stat)          /* BULKSTAT_RV_... */
 235 {
 236         xfs_bstat_t     *buf;           /* return buffer */
 237         int             error = 0;      /* error value */
 238         xfs_dinode_t    *dip;           /* dinode inode pointer */
 239
 240         dip = (xfs_dinode_t *)dibuff;
 241         *stat = BULKSTAT_RV_NOTHING;
 242
 243         if (!buffer || xfs_internal_inum(mp, ino))
 244                 return XFS_ERROR(EINVAL);
 245
 246         buf = kmem_alloc(sizeof(*buf), KM_SLEEP);
 247
 248         if (dip == NULL) {
 249                 /* We're not being passed a pointer to a dinode.  This happens
 250                  * if BULKSTAT_FG_IGET is selected.  Do the iget.
 251                  */
 252                 error = xfs_bulkstat_one_iget(mp, ino, bno, buf, stat);
 253                 if (error)
 254                         goto out_free;
 255         } else {
 256                 xfs_bulkstat_one_dinode(mp, ino, dip, buf);
 257         }
 258
 259         error = formatter(buffer, ubsize, ubused, buf);
 260         if (error)
 261                 goto out_free;
 262
 263         *stat = BULKSTAT_RV_DIDONE;
 264
 265  out_free:
 266         kmem_free(buf);
 267         return error;
 268 }
 269
 270 int
 271 xfs_bulkstat_one(
 272         xfs_mount_t     *mp,            /* mount point for filesystem */
 273         xfs_ino_t       ino,            /* inode number to get data for */
 274         void            __user *buffer, /* buffer to place output in */
 275         int             ubsize,         /* size of buffer */
 276         void            *private_data,  /* my private data */
 277         xfs_daddr_t     bno,            /* starting bno of inode cluster */
 278         int             *ubused,        /* bytes used by me */
 279         void            *dibuff,        /* on-disk inode buffer */
 280         int             *stat)          /* BULKSTAT_RV_... */
 281 {
 282         return xfs_bulkstat_one_int(mp, ino, buffer, ubsize,
 283                                     xfs_bulkstat_one_fmt, bno,
 284                                     ubused, dibuff, stat);
 285 }
 286
 287 /*
 288  * Test to see whether we can use the ondisk inode directly, based
 289  * on the given bulkstat flags, filling in dipp accordingly.
 290  * Returns zero if the inode is dodgey.
 291  */
 292 STATIC int
 293 xfs_bulkstat_use_dinode(
 294         xfs_mount_t     *mp,
 295         int             flags,
 296         xfs_buf_t       *bp,
 297         int             clustidx,
 298         xfs_dinode_t    **dipp)
 299 {
 300         xfs_dinode_t    *dip;
 301         unsigned int    aformat;
 302
 303         *dipp = NULL;
 304         if (!bp || (flags & BULKSTAT_FG_IGET))
 305                 return 1;
 306         dip = (xfs_dinode_t *)
 307                         xfs_buf_offset(bp, clustidx << mp->m_sb.sb_inodelog);
 308         /*
 309          * Check the buffer containing the on-disk inode for di_mode == 0.
 310          * This is to prevent xfs_bulkstat from picking up just reclaimed
 311          * inodes that have their in-core state initialized but not flushed
 312          * to disk yet. This is a temporary hack that would require a proper
 313          * fix in the future.
 314          */
 315         if (be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC ||
 316             !XFS_DINODE_GOOD_VERSION(dip->di_version) ||
 317             !dip->di_mode)
 318                 return 0;
 319         if (flags & BULKSTAT_FG_QUICK) {
 320                 *dipp = dip;
 321                 return 1;
 322         }
 323         /* BULKSTAT_FG_INLINE: if attr fork is local, or not there, use it */
 324         aformat = dip->di_aformat;
 325         if ((XFS_DFORK_Q(dip) == 0) ||
 326             (aformat == XFS_DINODE_FMT_LOCAL) ||
 327             (aformat == XFS_DINODE_FMT_EXTENTS && !dip->di_anextents)) {
 328                 *dipp = dip;
 329                 return 1;
 330         }
 331         return 1;
 332 }
 333
 334 #define XFS_BULKSTAT_UBLEFT(ubleft)     ((ubleft) >= statstruct_size)
 335
 336 /*
 337  * Return stat information in bulk (by-inode) for the filesystem.
 338  */
 339 int                                     /* error status */
 340 xfs_bulkstat(
 341         xfs_mount_t             *mp,    /* mount point for filesystem */
 342         xfs_ino_t               *lastinop, /* last inode returned */
 343         int                     *ubcountp, /* size of buffer/count returned */
 344         bulkstat_one_pf         formatter, /* func that'd fill a single buf */
 345         void                    *private_data,/* private data for formatter */
 346         size_t                  statstruct_size, /* sizeof struct filling */
 347         char                    __user *ubuffer, /* buffer with inode stats */
 348         int                     flags,  /* defined in xfs_itable.h */
 349         int                     *done)  /* 1 if there are more stats to get */
 350 {
 351         xfs_agblock_t           agbno=0;/* allocation group block number */
 352         xfs_buf_t               *agbp;  /* agi header buffer */
 353         xfs_agi_t               *agi;   /* agi header data */
 354         xfs_agino_t             agino;  /* inode # in allocation group */
 355         xfs_agnumber_t          agno;   /* allocation group number */
 356         xfs_daddr_t             bno;    /* inode cluster start daddr */
 357         int                     chunkidx; /* current index into inode chunk */
 358         int                     clustidx; /* current index into inode cluster */
 359         xfs_btree_cur_t         *cur;   /* btree cursor for ialloc btree */
 360         int                     end_of_ag; /* set if we've seen the ag end */
 361         int                     error;  /* error code */
 362         int                     fmterror;/* bulkstat formatter result */
 363         int                     i;      /* loop index */
 364         int                     icount; /* count of inodes good in irbuf */
 365         size_t                  irbsize; /* size of irec buffer in bytes */
 366         xfs_ino_t               ino;    /* inode number (filesystem) */
 367         xfs_inobt_rec_incore_t  *irbp;  /* current irec buffer pointer */
 368         xfs_inobt_rec_incore_t  *irbuf; /* start of irec buffer */
 369         xfs_inobt_rec_incore_t  *irbufend; /* end of good irec buffer entries */
 370         xfs_ino_t               lastino; /* last inode number returned */
 371         int                     nbcluster; /* # of blocks in a cluster */
 372         int                     nicluster; /* # of inodes in a cluster */
 373         int                     nimask; /* mask for inode clusters */
 374         int                     nirbuf; /* size of irbuf */
 375         int                     rval;   /* return value error code */
 376         int                     tmp;    /* result value from btree calls */
 377         int                     ubcount; /* size of user's buffer */
 378         int                     ubleft; /* bytes left in user's buffer */
 379         char                    __user *ubufp;  /* pointer into user's buffer */
 380         int                     ubelem; /* spaces used in user's buffer */
 381         int                     ubused; /* bytes used by formatter */
 382         xfs_buf_t               *bp;    /* ptr to on-disk inode cluster buf */
 383         xfs_dinode_t            *dip;   /* ptr into bp for specific inode */
 384
 385         /*
 386          * Get the last inode value, see if there's nothing to do.
 387          */
 388         ino = (xfs_ino_t)*lastinop;
 389         lastino = ino;
 390         dip = NULL;
 391         agno = XFS_INO_TO_AGNO(mp, ino);
 392         agino = XFS_INO_TO_AGINO(mp, ino);
 393         if (agno >= mp->m_sb.sb_agcount ||
 394             ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
 395                 *done = 1;
 396                 *ubcountp = 0;
 397                 return 0;
 398         }
 399         if (!ubcountp || *ubcountp <= 0) {
 400                 return EINVAL;
 401         }
 402         ubcount = *ubcountp; /* statstruct's */
 403         ubleft = ubcount * statstruct_size; /* bytes */
 404         *ubcountp = ubelem = 0;
 405         *done = 0;
 406         fmterror = 0;
 407         ubufp = ubuffer;
 408         nicluster = mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp) ?
 409                 mp->m_sb.sb_inopblock :
 410                 (XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog);
 411         nimask = ~(nicluster - 1);
 412         nbcluster = nicluster >> mp->m_sb.sb_inopblog;
 413         irbuf = kmem_zalloc_greedy(&irbsize, PAGE_SIZE, PAGE_SIZE * 4);
 414         if (!irbuf)
 415                 return ENOMEM;
 416
 417         nirbuf = irbsize / sizeof(*irbuf);
 418
 419         /*
 420          * Loop over the allocation groups, starting from the last
 421          * inode returned; 0 means start of the allocation group.
 422          */
 423         rval = 0;
 424         while (XFS_BULKSTAT_UBLEFT(ubleft) && agno < mp->m_sb.sb_agcount) {
 425                 cond_resched();
 426                 bp = NULL;
 427                 error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
 428                 if (error) {
 429                         /*
 430                          * Skip this allocation group and go to the next one.
 431                          */
 432                         agno++;
 433                         agino = 0;
 434                         continue;
 435                 }
 436                 agi = XFS_BUF_TO_AGI(agbp);
 437                 /*
 438                  * Allocate and initialize a btree cursor for ialloc btree.
 439                  */
 440                 cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
 441                 irbp = irbuf;
 442                 irbufend = irbuf + nirbuf;
 443                 end_of_ag = 0;
 444                 /*
 445                  * If we're returning in the middle of an allocation group,
 446                  * we need to get the remainder of the chunk we're in.
 447                  */
 448                 if (agino > 0) {
 449                         xfs_inobt_rec_incore_t r;
 450
 451                         /*
 452                          * Lookup the inode chunk that this inode lives in.
 453                          */
 454                         error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE,
 455                                                  &tmp);
 456                         if (!error &&   /* no I/O error */
 457                             tmp &&      /* lookup succeeded */
 458                                         /* got the record, should always work */
 459                             !(error = xfs_inobt_get_rec(cur, &r, &i)) &&
 460                             i == 1 &&
 461                                         /* this is the right chunk */
 462                             agino < r.ir_startino + XFS_INODES_PER_CHUNK &&
 463                                         /* lastino was not last in chunk */
 464                             (chunkidx = agino - r.ir_startino + 1) <
 465                                     XFS_INODES_PER_CHUNK &&
 466                                         /* there are some left allocated */
 467                             xfs_inobt_maskn(chunkidx,
 468                                     XFS_INODES_PER_CHUNK - chunkidx) &
 469                                     ~r.ir_free) {
 470                                 /*
 471                                  * Grab the chunk record.  Mark all the
 472                                  * uninteresting inodes (because they're
 473                                  * before our start point) free.
 474                                  */
 475                                 for (i = 0; i < chunkidx; i++) {
 476                                         if (XFS_INOBT_MASK(i) & ~r.ir_free)
 477                                                 r.ir_freecount++;
 478                                 }
 479                                 r.ir_free |= xfs_inobt_maskn(0, chunkidx);
 480                                 irbp->ir_startino = r.ir_startino;
 481                                 irbp->ir_freecount = r.ir_freecount;
 482                                 irbp->ir_free = r.ir_free;
 483                                 irbp++;
 484                                 agino = r.ir_startino + XFS_INODES_PER_CHUNK;
 485                                 icount = XFS_INODES_PER_CHUNK - r.ir_freecount;
 486                         } else {
 487                                 /*
 488                                  * If any of those tests failed, bump the
 489                                  * inode number (just in case).
 490                                  */
 491                                 agino++;
 492                                 icount = 0;
 493                         }
 494                         /*
 495                          * In any case, increment to the next record.
 496                          */
 497                         if (!error)
 498                                 error = xfs_btree_increment(cur, 0, &tmp);
 499                 } else {
 500                         /*
 501                          * Start of ag.  Lookup the first inode chunk.
 502                          */
 503                         error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &tmp);
 504                         icount = 0;
 505                 }
 506                 /*
 507                  * Loop through inode btree records in this ag,
 508                  * until we run out of inodes or space in the buffer.
 509                  */
 510                 while (irbp < irbufend && icount < ubcount) {
 511                         xfs_inobt_rec_incore_t r;
 512
 513                         /*
 514                          * Loop as long as we're unable to read the
 515                          * inode btree.
 516                          */
 517                         while (error) {
 518                                 agino += XFS_INODES_PER_CHUNK;
 519                                 if (XFS_AGINO_TO_AGBNO(mp, agino) >=
 520                                                 be32_to_cpu(agi->agi_length))
 521                                         break;
 522                                 error = xfs_inobt_lookup(cur, agino,
 523                                                          XFS_LOOKUP_GE, &tmp);
 524                                 cond_resched();
 525                         }
 526                         /*
 527                          * If ran off the end of the ag either with an error,
 528                          * or the normal way, set end and stop collecting.
 529                          */
 530                         if (error) {
 531                                 end_of_ag = 1;
 532                                 break;
 533                         }
 534
 535                         error = xfs_inobt_get_rec(cur, &r, &i);
 536                         if (error || i == 0) {
 537                                 end_of_ag = 1;
 538                                 break;
 539                         }
 540
 541                         /*
 542                          * If this chunk has any allocated inodes, save it.
 543                          * Also start read-ahead now for this chunk.
 544                          */
 545                         if (r.ir_freecount < XFS_INODES_PER_CHUNK) {
 546                                 /*
 547                                  * Loop over all clusters in the next chunk.
 548                                  * Do a readahead if there are any allocated
 549                                  * inodes in that cluster.
 550                                  */
 551                                 agbno = XFS_AGINO_TO_AGBNO(mp, r.ir_startino);
 552                                 for (chunkidx = 0;
 553                                      chunkidx < XFS_INODES_PER_CHUNK;
 554                                      chunkidx += nicluster,
 555                                      agbno += nbcluster) {
 556                                         if (xfs_inobt_maskn(chunkidx, nicluster)
 557                                                         & ~r.ir_free)
 558                                                 xfs_btree_reada_bufs(mp, agno,
 559                                                         agbno, nbcluster);
 560                                 }
 561                                 irbp->ir_startino = r.ir_startino;
 562                                 irbp->ir_freecount = r.ir_freecount;
 563                                 irbp->ir_free = r.ir_free;
 564                                 irbp++;
 565                                 icount += XFS_INODES_PER_CHUNK - r.ir_freecount;
 566                         }
 567                         /*
 568                          * Set agino to after this chunk and bump the cursor.
 569                          */
 570                         agino = r.ir_startino + XFS_INODES_PER_CHUNK;
 571                         error = xfs_btree_increment(cur, 0, &tmp);
 572                         cond_resched();
 573                 }
 574                 /*
 575                  * Drop the btree buffers and the agi buffer.
 576                  * We can't hold any of the locks these represent
 577                  * when calling iget.
 578                  */
 579                 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
 580                 xfs_buf_relse(agbp);
 581                 /*
 582                  * Now format all the good inodes into the user's buffer.
 583                  */
 584                 irbufend = irbp;
 585                 for (irbp = irbuf;
 586                      irbp < irbufend && XFS_BULKSTAT_UBLEFT(ubleft); irbp++) {
 587                         /*
 588                          * Now process this chunk of inodes.
 589                          */
 590                         for (agino = irbp->ir_startino, chunkidx = clustidx = 0;
 591                              XFS_BULKSTAT_UBLEFT(ubleft) &&
 592                                 irbp->ir_freecount < XFS_INODES_PER_CHUNK;
 593                              chunkidx++, clustidx++, agino++) {
 594                                 ASSERT(chunkidx < XFS_INODES_PER_CHUNK);
 595                                 /*
 596                                  * Recompute agbno if this is the
 597                                  * first inode of the cluster.
 598                                  *
 599                                  * Careful with clustidx.   There can be
 600                                  * multiple clusters per chunk, a single
 601                                  * cluster per chunk or a cluster that has
 602                                  * inodes represented from several different
 603                                  * chunks (if blocksize is large).
 604                                  *
 605                                  * Because of this, the starting clustidx is
 606                                  * initialized to zero in this loop but must
 607                                  * later be reset after reading in the cluster
 608                                  * buffer.
 609                                  */
 610                                 if ((chunkidx & (nicluster - 1)) == 0) {
 611                                         agbno = XFS_AGINO_TO_AGBNO(mp,
 612                                                         irbp->ir_startino) +
 613                                                 ((chunkidx & nimask) >>
 614                                                  mp->m_sb.sb_inopblog);
 615
 616                                         if (flags & (BULKSTAT_FG_QUICK |
 617                                                      BULKSTAT_FG_INLINE)) {
 618                                                 int offset;
 619
 620                                                 ino = XFS_AGINO_TO_INO(mp, agno,
 621                                                                        agino);
 622                                                 bno = XFS_AGB_TO_DADDR(mp, agno,
 623                                                                        agbno);
 624
 625                                                 /*
 626                                                  * Get the inode cluster buffer
 627                                                  */
 628                                                 if (bp)
 629                                                         xfs_buf_relse(bp);
 630
 631                                                 error = xfs_inotobp(mp, NULL, ino, &dip,
 632                                                                     &bp, &offset,
 633                                                                     XFS_IGET_BULKSTAT);
 634
 635                                                 if (!error)
 636                                                         clustidx = offset / mp->m_sb.sb_inodesize;
 637                                                 if (XFS_TEST_ERROR(error != 0,
 638                                                                    mp, XFS_ERRTAG_BULKSTAT_READ_CHUNK,
 639                                                                    XFS_RANDOM_BULKSTAT_READ_CHUNK)) {
 640                                                         bp = NULL;
 641                                                         ubleft = 0;
 642                                                         rval = error;
 643                                                         break;
 644                                                 }
 645                                         }
 646                                 }
 647                                 ino = XFS_AGINO_TO_INO(mp, agno, agino);
 648                                 bno = XFS_AGB_TO_DADDR(mp, agno, agbno);
 649                                 /*
 650                                  * Skip if this inode is free.
 651                                  */
 652                                 if (XFS_INOBT_MASK(chunkidx) & irbp->ir_free) {
 653                                         lastino = ino;
 654                                         continue;
 655                                 }
 656                                 /*
 657                                  * Count used inodes as free so we can tell
 658                                  * when the chunk is used up.
 659                                  */
 660                                 irbp->ir_freecount++;
 661                                 if (!xfs_bulkstat_use_dinode(mp, flags, bp,
 662                                                              clustidx, &dip)) {
 663                                         lastino = ino;
 664                                         continue;
 665                                 }
 666                                 /*
 667                                  * If we need to do an iget, cannot hold bp.
 668                                  * Drop it, until starting the next cluster.
 669                                  */
 670                                 if ((flags & BULKSTAT_FG_INLINE) && !dip) {
 671                                         if (bp)
 672                                                 xfs_buf_relse(bp);
 673                                         bp = NULL;
 674                                 }
 675
 676                                 /*
 677                                  * Get the inode and fill in a single buffer.
 678                                  * BULKSTAT_FG_QUICK uses dip to fill it in.
 679                                  * BULKSTAT_FG_IGET uses igets.
 680                                  * BULKSTAT_FG_INLINE uses dip if we have an
 681                                  * inline attr fork, else igets.
 682                                  * See: xfs_bulkstat_one & xfs_dm_bulkstat_one.
 683                                  * This is also used to count inodes/blks, etc
 684                                  * in xfs_qm_quotacheck.
 685                                  */
 686                                 ubused = statstruct_size;
 687                                 error = formatter(mp, ino, ubufp,
 688                                                 ubleft, private_data,
 689                                                 bno, &ubused, dip, &fmterror);
 690                                 if (fmterror == BULKSTAT_RV_NOTHING) {
 691                                         if (error && error != ENOENT &&
 692                                                 error != EINVAL) {
 693                                                 ubleft = 0;
 694                                                 rval = error;
 695                                                 break;
 696                                         }
 697                                         lastino = ino;
 698                                         continue;
 699                                 }
 700                                 if (fmterror == BULKSTAT_RV_GIVEUP) {
 701                                         ubleft = 0;
 702                                         ASSERT(error);
 703                                         rval = error;
 704                                         break;
 705                                 }
 706                                 if (ubufp)
 707                                         ubufp += ubused;
 708                                 ubleft -= ubused;
 709                                 ubelem++;
 710                                 lastino = ino;
 711                         }
 712
 713                         cond_resched();
 714                 }
 715
 716                 if (bp)
 717                         xfs_buf_relse(bp);
 718
 719                 /*
 720                  * Set up for the next loop iteration.
 721                  */
 722                 if (XFS_BULKSTAT_UBLEFT(ubleft)) {
 723                         if (end_of_ag) {
 724                                 agno++;
 725                                 agino = 0;
 726                         } else
 727                                 agino = XFS_INO_TO_AGINO(mp, lastino);
 728                 } else
 729                         break;
 730         }
 731         /*
 732          * Done, we're either out of filesystem or space to put the data.
 733          */
 734         kmem_free_large(irbuf);
 735         *ubcountp = ubelem;
 736         /*
 737          * Found some inodes, return them now and return the error next time.
 738          */
 739         if (ubelem)
 740                 rval = 0;
 741         if (agno >= mp->m_sb.sb_agcount) {
 742                 /*
 743                  * If we ran out of filesystem, mark lastino as off
 744                  * the end of the filesystem, so the next call
 745                  * will return immediately.
 746                  */
 747                 *lastinop = (xfs_ino_t)XFS_AGINO_TO_INO(mp, agno, 0);
 748                 *done = 1;
 749         } else
 750                 *lastinop = (xfs_ino_t)lastino;
 751
 752         return rval;
 753 }
 754
 755 /*
 756  * Return stat information in bulk (by-inode) for the filesystem.
 757  * Special case for non-sequential one inode bulkstat.
 758  */
 759 int                                     /* error status */
 760 xfs_bulkstat_single(
 761         xfs_mount_t             *mp,    /* mount point for filesystem */
 762         xfs_ino_t               *lastinop, /* inode to return */
 763         char                    __user *buffer, /* buffer with inode stats */
 764         int                     *done)  /* 1 if there are more stats to get */
 765 {
 766         int                     count;  /* count value for bulkstat call */
 767         int                     error;  /* return value */
 768         xfs_ino_t               ino;    /* filesystem inode number */
 769         int                     res;    /* result from bs1 */
 770
 771         /*
 772          * note that requesting valid inode numbers which are not allocated
 773          * to inodes will most likely cause xfs_itobp to generate warning
 774          * messages about bad magic numbers. This is ok. The fact that
 775          * the inode isn't actually an inode is handled by the
 776          * error check below. Done this way to make the usual case faster
 777          * at the expense of the error case.
 778          */
 779
 780         ino = (xfs_ino_t)*lastinop;
 781         error = xfs_bulkstat_one(mp, ino, buffer, sizeof(xfs_bstat_t),
 782                                  NULL, 0, NULL, NULL, &res);
 783         if (error) {
 784                 /*
 785                  * Special case way failed, do it the "long" way
 786                  * to see if that works.
 787                  */
 788                 (*lastinop)--;
 789                 count = 1;
 790                 if (xfs_bulkstat(mp, lastinop, &count, xfs_bulkstat_one,
 791                                 NULL, sizeof(xfs_bstat_t), buffer,
 792                                 BULKSTAT_FG_IGET, done))
 793                         return error;
 794                 if (count == 0 || (xfs_ino_t)*lastinop != ino)
 795                         return error == EFSCORRUPTED ?
 796                                 XFS_ERROR(EINVAL) : error;
 797                 else
 798                         return 0;
 799         }
 800         *done = 0;
 801         return 0;
 802 }
 803
 804 int
 805 xfs_inumbers_fmt(
 806         void                    __user *ubuffer, /* buffer to write to */
 807         const xfs_inogrp_t      *buffer,        /* buffer to read from */
 808         long                    count,          /* # of elements to read */
 809         long                    *written)       /* # of bytes written */
 810 {
 811         if (copy_to_user(ubuffer, buffer, count * sizeof(*buffer)))
 812                 return -EFAULT;
 813         *written = count * sizeof(*buffer);
 814         return 0;
 815 }
 816
 817 /*
 818  * Return inode number table for the filesystem.
 819  */
 820 int                                     /* error status */
 821 xfs_inumbers(
 822         xfs_mount_t     *mp,            /* mount point for filesystem */
 823         xfs_ino_t       *lastino,       /* last inode returned */
 824         int             *count,         /* size of buffer/count returned */
 825         void            __user *ubuffer,/* buffer with inode descriptions */
 826         inumbers_fmt_pf formatter)
 827 {
 828         xfs_buf_t       *agbp;
 829         xfs_agino_t     agino;
 830         xfs_agnumber_t  agno;
 831         int             bcount;
 832         xfs_inogrp_t    *buffer;
 833         int             bufidx;
 834         xfs_btree_cur_t *cur;
 835         int             error;
 836         xfs_inobt_rec_incore_t r;
 837         int             i;
 838         xfs_ino_t       ino;
 839         int             left;
 840         int             tmp;
 841
 842         ino = (xfs_ino_t)*lastino;
 843         agno = XFS_INO_TO_AGNO(mp, ino);
 844         agino = XFS_INO_TO_AGINO(mp, ino);
 845         left = *count;
 846         *count = 0;
 847         bcount = MIN(left, (int)(PAGE_SIZE / sizeof(*buffer)));
 848         buffer = kmem_alloc(bcount * sizeof(*buffer), KM_SLEEP);
 849         error = bufidx = 0;
 850         cur = NULL;
 851         agbp = NULL;
 852         while (left > 0 && agno < mp->m_sb.sb_agcount) {
 853                 if (agbp == NULL) {
 854                         error = xfs_ialloc_read_agi(mp, NULL, agno, &agbp);
 855                         if (error) {
 856                                 /*
 857                                  * If we can't read the AGI of this ag,
 858                                  * then just skip to the next one.
 859                                  */
 860                                 ASSERT(cur == NULL);
 861                                 agbp = NULL;
 862                                 agno++;
 863                                 agino = 0;
 864                                 continue;
 865                         }
 866                         cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
 867                         error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_GE,
 868                                                  &tmp);
 869                         if (error) {
 870                                 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 871                                 cur = NULL;
 872                                 xfs_buf_relse(agbp);
 873                                 agbp = NULL;
 874                                 /*
 875                                  * Move up the last inode in the current
 876                                  * chunk.  The lookup_ge will always get
 877                                  * us the first inode in the next chunk.
 878                                  */
 879                                 agino += XFS_INODES_PER_CHUNK - 1;
 880                                 continue;
 881                         }
 882                 }
 883                 error = xfs_inobt_get_rec(cur, &r, &i);
 884                 if (error || i == 0) {
 885                         xfs_buf_relse(agbp);
 886                         agbp = NULL;
 887                         xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
 888                         cur = NULL;
 889                         agno++;
 890                         agino = 0;
 891                         continue;
 892                 }
 893                 agino = r.ir_startino + XFS_INODES_PER_CHUNK - 1;
 894                 buffer[bufidx].xi_startino =
 895                         XFS_AGINO_TO_INO(mp, agno, r.ir_startino);
 896                 buffer[bufidx].xi_alloccount =
 897                         XFS_INODES_PER_CHUNK - r.ir_freecount;
 898                 buffer[bufidx].xi_allocmask = ~r.ir_free;
 899                 bufidx++;
 900                 left--;
 901                 if (bufidx == bcount) {
 902                         long written;
 903                         if (formatter(ubuffer, buffer, bufidx, &written)) {
 904                                 error = XFS_ERROR(EFAULT);
 905                                 break;
 906                         }
 907                         ubuffer += written;
 908                         *count += bufidx;
 909                         bufidx = 0;
 910                 }
 911                 if (left) {
 912                         error = xfs_btree_increment(cur, 0, &tmp);
 913                         if (error) {
 914                                 xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 915                                 cur = NULL;
 916                                 xfs_buf_relse(agbp);
 917                                 agbp = NULL;
 918                                 /*
 919                                  * The agino value has already been bumped.
 920                                  * Just try to skip up to it.
 921                                  */
 922                                 agino += XFS_INODES_PER_CHUNK;
 923                                 continue;
 924                         }
 925                 }
 926         }
 927         if (!error) {
 928                 if (bufidx) {
 929                         long written;
 930                         if (formatter(ubuffer, buffer, bufidx, &written))
 931                                 error = XFS_ERROR(EFAULT);
 932                         else
 933                                 *count += bufidx;
 934                 }
 935                 *lastino = XFS_AGINO_TO_INO(mp, agno, agino);
 936         }
 937         kmem_free(buffer);
 938         if (cur)
 939                 xfs_btree_del_cursor(cur, (error ? XFS_BTREE_ERROR :
 940                                            XFS_BTREE_NOERROR));
 941         if (agbp)
 942                 xfs_buf_relse(agbp);
 943         return error;
 944 }