fs/xfs/xfs_rw.c

   1 /*
   2  * Copyright (c) 2000-2006 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_inode_item.h"
  38 #include "xfs_itable.h"
  39 #include "xfs_btree.h"
  40 #include "xfs_alloc.h"
  41 #include "xfs_ialloc.h"
  42 #include "xfs_attr.h"
  43 #include "xfs_bmap.h"
  44 #include "xfs_error.h"
  45 #include "xfs_buf_item.h"
  46 #include "xfs_rw.h"
  47
  48 /*
  49  * This is a subroutine for xfs_write() and other writers (xfs_ioctl)
  50  * which clears the setuid and setgid bits when a file is written.
  51  */
  52 int
  53 xfs_write_clear_setuid(
  54         xfs_inode_t     *ip)
  55 {
  56         xfs_mount_t     *mp;
  57         xfs_trans_t     *tp;
  58         int             error;
  59
  60         mp = ip->i_mount;
  61         tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
  62         if ((error = xfs_trans_reserve(tp, 0,
  63                                       XFS_WRITEID_LOG_RES(mp),
  64                                       0, 0, 0))) {
  65                 xfs_trans_cancel(tp, 0);
  66                 return error;
  67         }
  68         xfs_ilock(ip, XFS_ILOCK_EXCL);
  69         xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
  70         xfs_trans_ihold(tp, ip);
  71         ip->i_d.di_mode &= ~S_ISUID;
  72
  73         /*
  74          * Note that we don't have to worry about mandatory
  75          * file locking being disabled here because we only
  76          * clear the S_ISGID bit if the Group execute bit is
  77          * on, but if it was on then mandatory locking wouldn't
  78          * have been enabled.
  79          */
  80         if (ip->i_d.di_mode & S_IXGRP) {
  81                 ip->i_d.di_mode &= ~S_ISGID;
  82         }
  83         xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
  84         xfs_trans_set_sync(tp);
  85         error = xfs_trans_commit(tp, 0);
  86         xfs_iunlock(ip, XFS_ILOCK_EXCL);
  87         return 0;
  88 }
  89
  90 /*
  91  * Force a shutdown of the filesystem instantly while keeping
  92  * the filesystem consistent. We don't do an unmount here; just shutdown
  93  * the shop, make sure that absolutely nothing persistent happens to
  94  * this filesystem after this point.
  95  */
  96 void
  97 xfs_do_force_shutdown(
  98         xfs_mount_t     *mp,
  99         int             flags,
 100         char            *fname,
 101         int             lnnum)
 102 {
 103         int             logerror;
 104
 105         logerror = flags & SHUTDOWN_LOG_IO_ERROR;
 106
 107         if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
 108                 cmn_err(CE_NOTE, "xfs_force_shutdown(%s,0x%x) called from "
 109                                  "line %d of file %s.  Return address = 0x%p",
 110                         mp->m_fsname, flags, lnnum, fname, __return_address);
 111         }
 112         /*
 113          * No need to duplicate efforts.
 114          */
 115         if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
 116                 return;
 117
 118         /*
 119          * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
 120          * queue up anybody new on the log reservations, and wakes up
 121          * everybody who's sleeping on log reservations to tell them
 122          * the bad news.
 123          */
 124         if (xfs_log_force_umount(mp, logerror))
 125                 return;
 126
 127         if (flags & SHUTDOWN_CORRUPT_INCORE) {
 128                 xfs_cmn_err(XFS_PTAG_SHUTDOWN_CORRUPT, CE_ALERT, mp,
 129     "Corruption of in-memory data detected.  Shutting down filesystem: %s",
 130                         mp->m_fsname);
 131                 if (XFS_ERRLEVEL_HIGH <= xfs_error_level) {
 132                         xfs_stack_trace();
 133                 }
 134         } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
 135                 if (logerror) {
 136                         xfs_cmn_err(XFS_PTAG_SHUTDOWN_LOGERROR, CE_ALERT, mp,
 137                 "Log I/O Error Detected.  Shutting down filesystem: %s",
 138                                 mp->m_fsname);
 139                 } else if (flags & SHUTDOWN_DEVICE_REQ) {
 140                         xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
 141                 "All device paths lost.  Shutting down filesystem: %s",
 142                                 mp->m_fsname);
 143                 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
 144                         xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
 145                 "I/O Error Detected.  Shutting down filesystem: %s",
 146                                 mp->m_fsname);
 147                 }
 148         }
 149         if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
 150                 cmn_err(CE_ALERT, "Please umount the filesystem, "
 151                                   "and rectify the problem(s)");
 152         }
 153 }
 154
 155
 156 /*
 157  * Called when we want to stop a buffer from getting written or read.
 158  * We attach the EIO error, muck with its flags, and call biodone
 159  * so that the proper iodone callbacks get called.
 160  */
 161 int
 162 xfs_bioerror(
 163         xfs_buf_t *bp)
 164 {
 165
 166 #ifdef XFSERRORDEBUG
 167         ASSERT(XFS_BUF_ISREAD(bp) || bp->b_iodone);
 168 #endif
 169
 170         /*
 171          * No need to wait until the buffer is unpinned.
 172          * We aren't flushing it.
 173          */
 174         xfs_buftrace("XFS IOERROR", bp);
 175         XFS_BUF_ERROR(bp, EIO);
 176         /*
 177          * We're calling biodone, so delete B_DONE flag. Either way
 178          * we have to call the iodone callback, and calling biodone
 179          * probably is the best way since it takes care of
 180          * GRIO as well.
 181          */
 182         XFS_BUF_UNREAD(bp);
 183         XFS_BUF_UNDELAYWRITE(bp);
 184         XFS_BUF_UNDONE(bp);
 185         XFS_BUF_STALE(bp);
 186
 187         XFS_BUF_CLR_BDSTRAT_FUNC(bp);
 188         xfs_biodone(bp);
 189
 190         return (EIO);
 191 }
 192
 193 /*
 194  * Same as xfs_bioerror, except that we are releasing the buffer
 195  * here ourselves, and avoiding the biodone call.
 196  * This is meant for userdata errors; metadata bufs come with
 197  * iodone functions attached, so that we can track down errors.
 198  */
 199 int
 200 xfs_bioerror_relse(
 201         xfs_buf_t *bp)
 202 {
 203         int64_t fl;
 204
 205         ASSERT(XFS_BUF_IODONE_FUNC(bp) != xfs_buf_iodone_callbacks);
 206         ASSERT(XFS_BUF_IODONE_FUNC(bp) != xlog_iodone);
 207
 208         xfs_buftrace("XFS IOERRELSE", bp);
 209         fl = XFS_BUF_BFLAGS(bp);
 210         /*
 211          * No need to wait until the buffer is unpinned.
 212          * We aren't flushing it.
 213          *
 214          * chunkhold expects B_DONE to be set, whether
 215          * we actually finish the I/O or not. We don't want to
 216          * change that interface.
 217          */
 218         XFS_BUF_UNREAD(bp);
 219         XFS_BUF_UNDELAYWRITE(bp);
 220         XFS_BUF_DONE(bp);
 221         XFS_BUF_STALE(bp);
 222         XFS_BUF_CLR_IODONE_FUNC(bp);
 223         XFS_BUF_CLR_BDSTRAT_FUNC(bp);
 224         if (!(fl & XFS_B_ASYNC)) {
 225                 /*
 226                  * Mark b_error and B_ERROR _both_.
 227                  * Lot's of chunkcache code assumes that.
 228                  * There's no reason to mark error for
 229                  * ASYNC buffers.
 230                  */
 231                 XFS_BUF_ERROR(bp, EIO);
 232                 XFS_BUF_FINISH_IOWAIT(bp);
 233         } else {
 234                 xfs_buf_relse(bp);
 235         }
 236         return (EIO);
 237 }
 238
 239 /*
 240  * Prints out an ALERT message about I/O error.
 241  */
 242 void
 243 xfs_ioerror_alert(
 244         char                    *func,
 245         struct xfs_mount        *mp,
 246         xfs_buf_t               *bp,
 247         xfs_daddr_t             blkno)
 248 {
 249         cmn_err(CE_ALERT,
 250  "I/O error in filesystem (\"%s\") meta-data dev %s block 0x%llx"
 251  "       (\"%s\") error %d buf count %zd",
 252                 (!mp || !mp->m_fsname) ? "(fs name not set)" : mp->m_fsname,
 253                 XFS_BUFTARG_NAME(XFS_BUF_TARGET(bp)),
 254                 (__uint64_t)blkno, func,
 255                 XFS_BUF_GETERROR(bp), XFS_BUF_COUNT(bp));
 256 }
 257
 258 /*
 259  * This isn't an absolute requirement, but it is
 260  * just a good idea to call xfs_read_buf instead of
 261  * directly doing a read_buf call. For one, we shouldn't
 262  * be doing this disk read if we are in SHUTDOWN state anyway,
 263  * so this stops that from happening. Secondly, this does all
 264  * the error checking stuff and the brelse if appropriate for
 265  * the caller, so the code can be a little leaner.
 266  */
 267
 268 int
 269 xfs_read_buf(
 270         struct xfs_mount *mp,
 271         xfs_buftarg_t    *target,
 272         xfs_daddr_t      blkno,
 273         int              len,
 274         uint             flags,
 275         xfs_buf_t        **bpp)
 276 {
 277         xfs_buf_t        *bp;
 278         int              error;
 279
 280         if (flags)
 281                 bp = xfs_buf_read_flags(target, blkno, len, flags);
 282         else
 283                 bp = xfs_buf_read(target, blkno, len, flags);
 284         if (!bp)
 285                 return XFS_ERROR(EIO);
 286         error = XFS_BUF_GETERROR(bp);
 287         if (bp && !error && !XFS_FORCED_SHUTDOWN(mp)) {
 288                 *bpp = bp;
 289         } else {
 290                 *bpp = NULL;
 291                 if (error) {
 292                         xfs_ioerror_alert("xfs_read_buf", mp, bp, XFS_BUF_ADDR(bp));
 293                 } else {
 294                         error = XFS_ERROR(EIO);
 295                 }
 296                 if (bp) {
 297                         XFS_BUF_UNDONE(bp);
 298                         XFS_BUF_UNDELAYWRITE(bp);
 299                         XFS_BUF_STALE(bp);
 300                         /*
 301                          * brelse clears B_ERROR and b_error
 302                          */
 303                         xfs_buf_relse(bp);
 304                 }
 305         }
 306         return (error);
 307 }
 308
 309 /*
 310  * Wrapper around bwrite() so that we can trap
 311  * write errors, and act accordingly.
 312  */
 313 int
 314 xfs_bwrite(
 315         struct xfs_mount *mp,
 316         struct xfs_buf   *bp)
 317 {
 318         int     error;
 319
 320         /*
 321          * XXXsup how does this work for quotas.
 322          */
 323         XFS_BUF_SET_BDSTRAT_FUNC(bp, xfs_bdstrat_cb);
 324         bp->b_mount = mp;
 325         XFS_BUF_WRITE(bp);
 326
 327         if ((error = XFS_bwrite(bp))) {
 328                 ASSERT(mp);
 329                 /*
 330                  * Cannot put a buftrace here since if the buffer is not
 331                  * B_HOLD then we will brelse() the buffer before returning
 332                  * from bwrite and we could be tracing a buffer that has
 333                  * been reused.
 334                  */
 335                 xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
 336         }
 337         return (error);
 338 }